Multicomponent reaction–derived covalent inhibitor space

Electrophiles are used in covalent drug discovery and synthesized from electrophilic building blocks diversified by MCRs.

Abstract

The area of covalent inhibitors is gaining momentum due to recently introduced clinical drugs, but libraries of these compounds are scarce. Multicomponent reaction (MCR) chemistry is well known for its easy access to a very large and diverse chemical space. Here, we show that MCRs are highly suitable to generate libraries of electrophiles based on different scaffolds and three-dimensional shapes and highly compatible with multiple functional groups. According to the building block principle of MCR, acrylamide, acrylic acid ester, sulfurylfluoride, chloroacetic acid amide, nitrile, and α,β-unsaturated sulfonamide warheads can be easily incorporated into many different scaffolds. We show examples of each electrophile on 10 different scaffolds on a preparative scale as well as in a high-throughput synthesis mode on a nanoscale to produce libraries of potential covalent binders in a resource- and time-saving manner. Our operational procedure is simple, mild, and step economical to facilitate future covalent library synthesis.

INTRODUCTION

Covalent inhibitors have a rich tradition as drugs exemplified in the classic and lifesaving β-lactam antibiotics (1). More than 25% of approved enzyme targeting drugs work through a covalent mechanism (2). Recently, covalent inhibitors have experienced a renaissance, for example, with the clinical introduction of the Bruton’s tyrosine kinase inhibitor zanubrutinib or the experimental G12C RAS inhibitor AMG510 (Fig. 1, A and B) (3, 4).

Fig. 1. Covalent inhibitors in medicinal chemistry and natural products.

(A and B) Three compounds covalently binding to their protein targets. AMG510, a G12C RAS inhibitor [Protein Data Bank (PDB) ID: 6OIM], cyclostreptin binding to His²²⁹ of tubulin (PDB ID: 6QTN), and Bruton’s tyrosine kinase inhibitor zanubrutinib (PDB ID: 6J6M). (C) Common electrophilic warheads and nucleophilic amino acid targets. (D) Synthetic strategies to covalent inhibitors.

In addition, nature extensively uses the principle of covalent modification exemplified by the tubulin binder cyclostreptin (Fig. 1, A and B) (5). Covalent drugs potentially offer several advantages over noncovalent drugs, including increased potency and therefore lower dosing, selectivity, duration of action, and resistance to mutations (6). However, covalent drugs have been controversial because of their potential off-target binding leading to unforeseeable, e.g., idiosyncratic or off-target toxicity (7). Selectivity issues and side effects could be potentially mitigated by maximizing selectivity of binding to the target protein, maintaining a low dose, and avoiding reactive metabolites (8). Outside drug discovery, covalent inhibitors also play an important role as tool compounds in chemical biology to identify ligands in proteome-wide screens (9). For example, a chemical proteomic approach has recently led to the discovery of a selective probe for the difficult to target mitochondrial pyruvate carrier complex (10). Covalent targeting of kinases based on endogenous or mutated cysteines is also a proven clinical anticancer strategy (11). The targets of covalent inhibitors in proteins are the nucleophilic amino acid side chains—mostly cysteine and serine but also threonine, histidine, tyrosine, lysine, arginine, or tryptophan (Fig. 1C). Moreover, aspartate and glutamate have been targeted by covalent modifications (12). Typical electrophiles used in covalent inhibitors include acrylamide (13), acrylic acid ester, sulfurylfluoride (14), boronic acid (15), chloroacetic acid amide (16), nitrile (17), and α,β-unsaturated sulfonamide warheads (Fig. 1C) (18). Another cysteine-specific technology is disulfide trapping, which is based on the principle of reversible thiol-disulfide exchange between the protein target and the disulfide containing screening compounds (19). Among the different electrophiles, a certain degree of nucleophile selectivity can be reached (20). For instance, boronic acid is an oxophile and has a preference for hydroxyl amino acids, while acrylamides are thiophiles and react often with cysteines in a Michael addition. To discover hits and initial leads, screening of diverse libraries based on multiple electrophiles imprinted on multiple scaffolds and decorated with many additional functional groups is desirable. Despite the growing importance of covalent targeting, diverse screening libraries decorated with a range of chemical functionalities are scarce. Past covalent library screening efforts were mostly performed on a small scale (21–23). Libraries of covalent inhibitors are often synthesized from (commercial) building blocks and additions of an electrophile by late-stage functionalization, e.g., acrylamide (24). Moreover, the few commercially available covalent screening libraries are of limited size and diversity.

Here, we introduce the use of multicomponent reaction (MCR) chemistry for the production of unmatched diverse libraries of covalent inhibitors. In combination with newly reported building blocks, our approach circumvents problems ascribed to traditional electrophile library synthesis such as slow sequential synthesis and limited library diversity. We report the synthesis of 10 MCR-compatible and 10 commercially available building blocks that were introduced into 10 different scaffolds to produce 102 compounds with a diverse set of electrophiles. The diversity of warheads included 10 different classes of electrophiles. To satisfy different compound quantity demands, we demonstrated the synthesis on different scales covering four orders of magnitude. Two synthesis protocols were reported in an automated fashion on a nanoscale or on a millimole scale for repeated use of the screening libraries.

RESULTS AND DISCUSSION

Chemistry and design considerations

To design a reactive library, we have to consider the nature of the electrophile, the structure of the noncovalent diversity elements, and the linker between them. The nature and thus intrinsic reactivity of the electrophile moiety can vary a lot (20, 25, 26). In our design, the electrophile is often attached to an aliphatic linker, which overall should normalize the intrinsic reactivity of the warhead. The distance between the electrophilic warhead and the diversity element is important because an effective compound must make productive interactions with the protein by displaying the electrophile at the correct distance and orientation to react with the nucleophile residue. The flexibility of the linker element also needs considerations not only to adapt to the shape of the receptor pocket but also to potentially reduce conformational space while improving drug-like properties, e.g., oral bioavailability by increasing membrane permeation. Thus, linker lengths and geometries are important design elements for electrophile libraries. The composition of the linker can also have important effects on the chemical reactivity of the electrophile (27). In addition, the nature of the electrophile is of great importance to address specific nucleophilic amino acid side chains and to fine-tune reactivity, e.g., oxophile versus thiophile. Last, the shape and three-dimensional (3D) pharmacophore distribution of ligands encoded in the scaffold is of high importance for noncovalent receptor interaction to provide a negative imprint of the binding pocket. Very few libraries containing a diverse selection of electrophiles, linker elements, scaffolds, and conformational space are available for purchase from commercial vendors, perhaps because of the long-standing bias toward noncovalent binders (28, 29). To create libraries of electrophiles, we used a convergent one-pot MCR approach using an array of available or easily accessible electrophile building blocks. The electrophile moiety was introduced into different kinds of building blocks with an orthogonal functional group required for a variety of different MCRs (Fig. 2). To keep the overall average molecular weight of the target compounds low, we designed rather small electrophile building blocks based on small aliphatic chains or six-membered rings to reflect flexibility and stiffness. However, it should be noted that our choice of linkers is arbitrary and many other building blocks are possible and would not be limited by the chemistry used. Additional considerations in our design were to enrich the library with related analogs to allow preliminary structure-activity relationships to be deduced directly from a primary screen.

Fig. 2. Design of electrophile libraries.

(A) The use of multiple MCRs allows for great scaffold diversity. (B) The electrophile building blocks consist of three parts: the electrophile functional group, the linker, and the MCR-compatible functional group. The colored forms graphically represent different types of building blocks.

To provide a broad range of scaffolds of different shape and 3D pharmacophore distribution, we used 10 different MCRs (Table 1). While hundreds of MCR scaffolds were previously described, we selected a subset to represent a broad range of chemotypes that have been previously used in the discovery of bioactive matter, including semirigid bis-amides (a) (30), heterocyclic basic α-amino tetrazoles (b) (31), heterocyclic planar imidazoles (c) (32), hydantoins (d) (30, 33), flexible hydroxyacylcarboxy amides (e) (34), bicyclic planar heterocycles (f) (35, 36), heterocyclic nonbasic α-hydroxy tetrazole (g) (37, 38), elongated basic bis-amides (h) (39), cyclic bis-amides with cis conformation (i) (40), and heteroaromatic conformationally constrained thiophene carboxamides (j) (41).

Table 1. The MCRs used to create libraries with electrophilic moieties.

The different scaffolds are boxed in different colors that will be used throughout the manuscript.

For the creation of our electrophile libraries, we required specific bifunctional building blocks. The synthesis of each building block was performed on a gram scale (Fig. 3). Primary amine-containing building blocks 6 and 7 are short and rather flexible aliphatic diamine-derived mono acylates with two and three rotatable bonds (nrb), respectively. Acrylate 10 features an acyl hydrazine moiety. Piperidines 14 and 24 are cyclic motifs featuring a vinyl sulfonamide and an acrylamide warhead along with a secondary amine and ketone as an MCR-compatible functional group, respectively. β-Cyanoethyl isocyanide 18 featuring a soft nitrile electrophile was synthesized (42). A flexible (nrb = 2) ethylenediamine-derived mono isocyanide mono acrylamide 22 was accessed by a short three-step sequence. Last, acrylamide carboxylic acids 27 and 28 were produced with two different short-length linkers. Moreover, additional building blocks were used, such as unsubstituted and differentially substituted acrylic acids 29 to 32 (Fig. 7), chloroacetic acid 33, and chloroacetyl chloride 1 and 2-butynoic acid 35 (Fig. 8). The but-2-ynamide electrophile has been used recently in the discovery of the covalent Bruton’s tyrosine kinase inhibitor branebrutinib as a superior electrophile over acrylamides and vinyl sulfonamides (43).

Fig. 3. Synthesis of specific bifunctional electrophile building blocks.

The electrophile and the MCR functional groups are marked in yellow and blue, respectively. rt, room temperature.

Fig. 7. Millimole-scale synthesis of electrophiles based on acrylic acid building blocks 29 to 32.

The color of the boxes refers to the scaffolds shown in Table 1.

Fig. 8. Millimole-scale synthesis of electrophiles based on singleton building blocks 10, 14, 18, 27, 28, and 33 to 37 (*yield over two steps).

Boxed color code is according to scaffold type of Table 1.

To show the feasibility of creating complex electrophiles as potential covalent binders, we performed model reactions for each scaffold using a selection of electrophile building blocks (Figs. 4 to 8). Ethylenediamine-derived building block 6 was introduced in five different Ugi four-component reactions (U-4CRs) (Table 1, a) with an average yield of 50% (Fig. 4). Moreover, we synthesized four highly substituted α-amino tetrazoles with an average yield of 63%. Substituted and unsubstituted aromatic, aliphatic, and alicyclic reagents worked well to satisfying. We also incorporated 6 into a flat heterocyclic imidazole scaffold, which can be accessed in two steps via an Ugi reaction followed by cyclization (6c1 and 6c2; Table 1, c) as well as into three imino-hydantoin examples (6d1 to 6d3; Table 1, d). The functional group compatibility of screening compounds is of high importance as it increases the chance to capture interactions with the receptors.

Fig. 4. Millimole-scale synthesis of electrophiles based on acrylamide amine building blocks 6 and 7.

(*yield over two steps). The color of the boxes refers to the scaffolds shown in Table 1.

Building block 22 is an ethylenediamine-derived acrylamide isocyanide and was synthesized in a three-step sequence on a gram scale (Fig. 3). The isocyanide was introduced in two variations of the Ugi reaction [U-4CR and Ugi tetrazole four-component reaction (UT-4CR)] as well as two variations of the Passerini reaction [Passerini three-component reaction (P-3CR) and Passerini tetrazole three component reaction (PT-3CR)]. In total, 24 derivatives were produced in an average of 75% yield (Fig. 5). Bis-allylamine (22b13) and bromo phenyl (22b7, 22e7, and 22e8) compounds could be further reacted, for example, by an ring-closing metathesis or Pd-catalyzed cross-coupling reactions (e.g., Suzuki-Miyaura reaction).

Fig. 5. Millimole-scale synthesis of electrophiles based on acrylamide isocyanide building block 25.

The color of the boxes refers to the scaffolds shown in Table 1.

The oxopiperidine N-acrylate building block 24 was introduced in four different scaffolds including U-4CR, P-3CR, PT-3CR, and UT-4CR (Table 1), in 16 different examples in an average yield of 55% (Fig. 6). Compound 24g1, for example, can be synthesized from simple building blocks in one step involving a newly described Passerini tetrazole reaction in 32% isolated yields (37).

Fig. 6. Millimole-scale synthesis of electrophiles based on acrylamide ketone building block 24.

The color of the boxes refers to the scaffolds shown in Table 1.

Next, we evaluated the reactivity of acrylic acid and several derivatives thereof (Fig. 7). Substituted acrylates have been used to fine-tune the reactivity toward nucleophiles (44). Four different acrylic acid derivatives have been used here in the U-4CR, the P-3CR, and the split U-4CR to yield 20 compounds at an average yield of 69%.

Last, we scouted other building blocks to increase the structural diversity of the electrophile library by performing singleton or small number syntheses (Fig. 8). In the UT-4CR, not only primary and secondary amines but also acryl hydrazones reacted smoothly. Thus, we synthesized acryl hydrazone building block 10 (Fig. 3) and reacted it in the UT-4CR to yield acrylhydrazone tetrazole 10b1. Vinyl sulfonamide building block 14 was combined similarly in the UT-4CR, yielding the reactive vinyl sulfonamide electrophiles (14b1 and 14b2). Arylsulfonylfluorides were described as privileged warheads in chemical biology with the right balance of biocompatibility and protein reactivity, modifying not only reactive serines but also context-specific threonine, lysine, tyrosine, cysteine, and histidine residues (45). The cyano group is not only a potent electrophile but the β-cyanoethyl group was recently also described as a protecting group to yield N-unsubstituted tetrazoles (42, 46). β-Cyanoethyl isocyanide 18 was introduced in different heterocyclic rings including imidazole (18c1), imidazopyridine (18f1), and imidazothiazole (18f2). Glycine-N-acrylate 27 and 4-amino butanoic acid N-acrylate 28 are examples of amino acid building blocks. Chloroacetic acid 33 works well in the Ugi and Passerini reactions and provides straightforward access to libraries of diversified chloroacetates, e.g., 33a1 and 33e1 and 33e2. Arylsulfonylfluoride building block 34 was reacted as a carboxylic acid in an U-4CR yielding 34a1 in 58% isolated yield or to the spirocycle 34i1 in 76% yield. Worthwhile to mention is the water-solubilizing tetrahydropyrane ring in 34i1. Butinyl carboxylic acid 35 yielded Ugi product 35a1, Passerini product 35e1, and spirocycle 35i1 in moderate to good yield. α-Cyanocyclopropyl carboxylic acid 36 yielded spirocycle 36i1. The α-cyano cyclopropyl moiety can often be found in reversible cysteine protease inhibitors (47). Acrylic acid chloride 1, and chloroacetyl chloride 37, were used for the late-stage functionalization of 2-amino thiophenes produced by a Gewald three-component reaction (GW-3CR; 1j1 to 1j4 and 37j1 to 37j4) (41).

Nanoscale synthesis

Automated, accelerated, nanoscale synthesis of compound libraries for the purpose of reaction evaluation and screening for biologically active compounds recently became an important alternative to manual macroscopic syntheses (40, 48–54). Nanoscale synthesis not only allows for fast and automated production of large compound collections but also is highly sustainable as much fewer valuable reagents, solvent, and consumables are used (55). We have recently introduced acoustic droplet ejection (ADE) as a suitable tool for the automated nanoscale synthesis of libraries of small molecules (40, 48, 54). Here, we describe the Ugi-4CR of carboxylic acids, primary amines, formaldehyde, and N-(2-isocyanoethyl)acrylamide 22 as an example for the nanoscale synthesis of electrophiles for potential covalent biological space scouting. We chose formaldehyde as a constant component because the methylene group renders the compounds more flexible as opposed to the Ugi scaffold incorporating substituted aldehydes and ketones. The nanosynthesis was performed as recently described by us (Supplementary Materials) (40, 48, 54). In brief, we used stock solutions of the appropriate building block carboxylic acids, formaldehyde, amines, and N-(2-isocyanoethyl)acrylamide 22 in ethylene glycol or 2-methoxy ethanol depending on solubility as 0.5 M stock solutions in 384-well source plates. The building blocks were automatically transferred into 384-well destination plates using an Echo 555 instrument. From each building block, 750 nl was sequentially transferred. The transfer time to charge one 384-well plate was ~150 min. The plates were then covered with a sealing foil and shaken for 24 hours. Then, the plates were unsealed and each well was diluted with 100 μl of ethylene glycol. The reactions were analyzed by direct injection into a mass spectrometer as recently described (40, 48, 54). The heights of the molecular ion peak or that of a derivative served to create a crude reaction classifier. The analytical outcome of the reaction is shown in Fig. 9. Of the 192 reactions, 53% gave the expected compound as a major product (depicted in green), while 20% of the reactions give no product at all (depicted in blue). In total, four 384-well plates were created, potentially yielding 1536 electrophiles. Noteworthy is the building block diversity used to assemble this array of electrophiles and some exemplary products are shown in Fig. 9. We used building blocks with two differentially reactive amines, an aniline and a benzylamine (A-14), o-amino phenol (A-5), o-amino-biphenyl (A-3), bulky trityl amine (A-19) and small methyl (A-18) or isopropylamine (A-21), or symmetrical butane 1,4-diamine (A-30) which can react mono or bis, highly substituted pyridyl-3-amine (A-23) that can undergo further nucleophilic aromatic substitution reactions, anthranilic acid (A-28), and pentane-5-olamine (A-15). Different benzoic acid building blocks were introduced, including phenol (B-1, B-50), boronic acid (B-64), o-biphenyl (B-20), and α,β-unsaturated carboxylic acids (B-23, B-33, B-41, B-44, B-45, B-49, and B-54). Especially worthwhile to mention is maleic acid derivative (B-49) which can undergo further addition reactions (56). A great diversity of heterocyclic building blocks was introduced, e.g., thiophene (B-15), benzofuran (B-12), indole (B-4, B-18), imidazole (B-43), furan (B-54), and cyanoacetic acid (B-3). Exemplary reaction products are also shown in Fig. 9, for example, the constrained proline derivative (B11) or isoquinoline derivative (B12). Other products include a diversity of unprotected functional groups such as aliphatic hydroxy (A2, A19, and B19), phenol (C15), amino group (A7 and B11), Boc-protected amine, and free carboxylic acid group (A9). B6, B7, B20, B21, C15, and D20 are examples of bis-α,β-unsaturated carbonyl scaffolds. Last, we resynthesized some of the library compounds out for the plates to confirm scalability and provide full analytical data and yields. The three products are F23 (57%), H1 (44%), and H9 (65%) and feature biphenyl, o-, and p-phenol as interesting structural moieties. In summary, the ADE-enabled nanoscale synthesis of electrophiles allows for the rapid and automated assembly of a diverse functional group and shape rich chemical space.

Fig. 9. Automated nanoscale synthesis of acrylamides using ADE.

(A) ADE-enabled nanoscale automated electrophile synthesis based on the U-4CR of isocyanide building block 22. (B) Heat plot of 192 compounds based on mass spectrometry analysis: green for major product formation, yellow for medium product formation, blue for no product formation, and white for Echo reagent transfer failure. (C) Exemplary amine and carboxylic acid building blocks used. (D) Exemplary product structures of the first two columns A and B. (E) Structures of resynthesized compounds on a millimole scale (red-boxed in heat plot).

Covalent libraries are of great use in biology and drug discovery. The potential of electrophile (-fragment) screening as a practical and efficient tool for covalent-ligand discovery is well described (25). Few generalized approaches toward covalent libraries have been described while the need for biological screening of electrophiles is increasing (24, 25, 57, 58). The strategic formation of covalent bonds between small molecules and proteins has many other underappreciated applications as enabling platforms for drug discovery (59). For example, targeting of noncatalytic cysteine residues in an allele-specific manner with small molecules is drawing attention from drug discovery scientists and chemical biologists (27). Acquired cysteines in cancer can be targeted by covalent inhibitors and several clinical trials are under way (60), whereas reactive-cysteine profiling for drug discovery has been described in a proteome-wide scale (26). However, the construction of diverse libraries of electrophiles is challenging, and general access is demanding and underexplored. Therefore, we report here a holistic covalent library synthesis approach. The greatest advantage of our approach is the simple one-pot synthesis of electrophiles instead of sequential multistep synthesis, and as an example, we exercised the approach with 10 different scaffolds. The MCRs allow for the exploration of a very large chemical space based on a great number of archetypical commercially available building blocks such as primary amines, carboxylic acids, and aldehydes. For this, we synthesized nine different bifunctional building blocks on a gram scale or used commercially available building blocks incorporating an electrophile, a linker and an orthogonal MCR-compatible functional group (e.g., isocyanide, carbocyclic acid, and amine). These different electrophile building blocks were introduced in 10 different scaffolds. The generality and usefulness of our approach are reflected in the compatibility with many different functional groups, shapes, and electronic features. The generality of our approach is also reflected in the breath of scale from nanomole to millimole, the great number of synthesized compounds, and the diverse chemistries used. The syntheses were performed on a millimole scale, but also in an automated fashion on a nanoscale in 384- and 1536-well plate formats using ADE technology. Automated ADE-enabled chemistry allows the synthesis of tens of thousands of electrophiles. This methodology is amenable to a variety of MCRs as well as multiple classes of electrophile building blocks, allowing single-step access to a diverse array of products. Another benefit of the nanosynthesis is that any compound hitting a target during screening can be instantaneously resynthesized on a larger millimole scale for validation. One of the challenges of covalent library screening is that intrinsic reactivity of the warheads can vary greatly (16, 20, 25, 26). However, a key advantage of our work is that, in many of our scaffold designs, the aliphatic linker between the warhead and main scaffold will normalize the intrinsic reactivity. Ongoing applications of our electrophile synthesis platform in our laboratory will be reported in due course.

MATERIALS AND METHODS

All reagents and solvents were purchased from Sigma-Aldrich, Abcr GmbH, Acros, Fluorochem, and AK Scientific and were used without further purification. All isocyanides were prepared in-house (see the Supplementary Materials). All microwave irradiation reactions were carried out in a Biotage Initiator Microwave Synthesizer. All sonication is performed in an ultrasonic cleaner (220/240 V, 25 A, and frequency of 50/60 Hz). Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker AVANCE 500 spectrometer ¹H NMR (500 MHz) and ¹³C NMR (126 MHz). Chemical shifts for ¹H NMR were reported as δ values, and coupling constants were in hertz (Hz). The following abbreviations were used for spin multiplicity: s = singlet, br s = broad singlet, d = doublet, br d = broad doublet, t = triplet, br t = broad triplet, q = quartet, dd = double of doublets, ddd = double of doublet of doublets, and m = multiplet. Chemical shifts for ¹³C NMR reported in parts per million relative to the solvent peak. Thin-layer chromatography was performed using precoated silica gel 60 F₂₅₄ plates (Merck, Darmstadt), and the spots were visualized with ultraviolet light at 254 nm. Flash chromatography was performed on a Reveleris X2 Flash Chromatography, using Grace Reveleris Silica flash cartridges (12 g). Mass spectra were measured on a Waters Investigator Supercritical Fluid Chromatograph with a 3100 Mass Detector (electrospray ionization) using a solvent system of methanol and CO₂ on a Viridis silica gel column (4.6 × 250 mm, 5-μm particle size) or Viridis 2-ethyl pyridine column (4.6 × 250 mm, 5-μm particle size). High-resolution mass spectra were recorded using a LTQ-Orbitrap-Velos Pro (Thermo Fisher Scientific) in ESI-positive mode at a resolution of 60000 at m/z 400.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/6/eabd9307/DC1

View/request a protocol for this paper from Bio-protocol.