Chemical Proteomics-Guided Discovery of Covalent Ligands for Cancer Proteins

Abstract

Advances in genome sequencing and editing technologies have enriched our understanding of the biochemical pathways that drive tumorigenesis. Translating this knowledge into new medicines for cancer treatment, however, remains challenging, and many oncogenic proteins have proven recalcitrant to conventional approaches for chemical probe and drug discovery. Here, we discuss how innovations in chemical proteomics and covalent chemistry are being integrated to identify and advance first-in-class small molecules that target cancer-relevant proteins. Mechanistic studies have revealed that covalent compounds perturb protein functions in cancer cells in diverse ways that include the remodeling of protein-protein and protein-RNA complexes, as well as through alterations in post-translational modification. We speculate on the attributes of chemical proteomics and covalent chemistry that have enabled targeting of previously inaccessible cancer-relevant pathways and consider technical challenges that remain to be addressed in order to fully realize the druggability of the cancer proteome.

1. Introduction

Proteins that drive cancer cell growth have been identified in multiple ways. The mapping of frequent, or hotspot, somatic mutations and fusion events in cancer has proven to be a particularly rich source of oncogenic proteins for targeted therapy (e.g., kinases like ABL, ALK, EGFR, and FGFR) (Chakravarty & Solit 2021). Lineage factors that, even in unmutated form, are required for the growth of specific cancer types are another category of validated cancer targets, particularly for expendable tissues, such as breast (e.g., estrogen receptor), prostate (e.g., androgen receptor), or certain immune cell types (e.g., BTK) (Garraway & Sellers 2006) A third type of cancer target exhibits synthetic lethality with, for instance, specific mutational states in cancer (e.g., PARP1 in BRCA-mutant cancers) (O’Neil et al 2017). More recently, genome-wide functional screens using CRISPR/Cas9 gene editing have generated a Cancer Dependency Map (Tsherniak et al 2017) that includes a striking number of targets required for the growth of specific subset of genetically- or lineage-defined cancer cells.

In combination, the aforementioned types of cancer-relevant proteins provide an incredibly rich foundation for the pursuit of next-generated targeted therapies. Nonetheless, chemical probes are lacking for most of these proteins. For instance, among 628 cancer priority targets from CRISPR–Cas9 screens in 324 human cancer cell lines (Behan et al 2019), only 40 are targeted by FDA-approved drugs, and approximately 50% of the remaining priority proteins originate from classes such as transcription factors and adaptor proteins that have proven historically recalcitrant to targeting with small molecules (Bushweller 2019, Dang et al 2017). Addressing this challenge has become a source for both chemical and technological innovations.

Many complementary approaches have been introduced to discover small-molecule binders for cancer-relevant proteins, including fragment-based screening (Erlanson et al 2016), DNA-encoded libraries (Goodnow et al 2017), and chemical proteomics (Spradlin et al 2021). This review will focus on chemical proteomic strategies for ligand discovery in cancer with an emphasis on integrating these methods with covalent chemistry (Gehringer & Laufer 2019, Johnson et al 2010). We will briefly discuss the origins of the activity-based protein profiling (ABPP) technology as a chemical proteomic method to assess dynamic changes in enzyme activities in native biological systems, and how the concepts of ABPP have since been extended to furnish global maps of covalent small molecule-protein interactions in cancer cells. We will describe how such maps can be used to discover allosteric and first-in-class ligands for cancer-relevant proteins from diverse structural and functional classes, as well as to understand and optimize the proteome-wide specificity of advanced covalent probes and drug candidates targeting oncogenic proteins. Finally, we will discuss future opportunities and challenges for cancer chemical probe and drug discovery.

2. Origins and principles of ABPP.

Original platforms for ABPP utilized active site-directed chemical probes to functionally characterize large numbers of mechanistically related enzymes in native biological systems (Cravatt et al 2008, Sanman & Bogyo 2014). Such “class-specific” ABPP probes have been developed for diverse enzymes families, including serine hydrolases (Liu et al 1999), cysteine proteases (Greenbaum et al 2000), histone deacetylases (Salisbury & Cravatt 2007), kinases (Patricelli et al 2007), and phosphatases (Lo et al 2002) (Figure 1), and used to identify deregulated enzymes in disease states (Jessani et al 2004, Nomura et al 2010), as well as to identify first-in-class enzyme inhibitors (Niphakis & Cravatt 2014). Of note, many of the novel enzyme inhibitors advanced with ABPP are covalent, including several targeting cancer-relevant enzymes (Gruner et al 2016, Remsberg et al 2021), and others that have entered clinical development for indications like neurological disease (Cisar et al 2018, Johnson et al 2011). In these programs, ABPP provided not only a versatile target engagement assay, but also global maps of inhibitor selectivity across the target enzyme class and the broader proteome. In doing so, ABPP acquired some of the first compelling evidence to support the high selectivity that optimized covalent inhibitors can display in vivo. Nonetheless, covalent inhibitors of enzymes often target aberrantly nucleophilic catalytic residues in well-defined active sites poised to bind small molecules. Whether covalent chemistry could be extended for probe and drug development across the broader “non-enzymatic” proteome remained an open question.

Figure 1. Activity-based protein profiling (ABPP) of enzymatic and non-enzymatic proteins in cancer cells.

Original formats for ABPP employed active site-directed chemical probes to investigate specific enzyme families (shown for acylphosphate probes targeting catalytic lyines in kinases (Patricelli et al 2007) and fluorophosphonate probes targeting catalytic serines in serine hydrolases (Liu et al 1999)). More recent platforms for ABPP use chemical probes that broadly react with a specific type of amino acid in proteins (shown for iodoacetamide probes targeting cysteines), allowing for analysis of functionality (Weerapana et al 2010) and small-molecule interactions (Wang et al 2014) of many sites on proteins from diverse classes.

3. Extending the principles of ABPP for proteome-wide investigations of covalent chemistry.

Additional evidence for the greater potential of covalent chemistry emerged from medicinal chemistry efforts focused on cancer-relevant kinases, such as Bruton’s tyrosine kinase (BTK) and epidermal growth factor receptor (EGFR). The structure-guided incorporation of electrophilic groups into advanced reversible inhibitors targeting these kinases resulted in covalent modification of a shared ATP-binding pocket cysteine (Honigberg et al 2010, Li et al 2008, Pan et al 2007, Zhou et al 2009). ABPP studies further supported that covalent BTK and EGFR inhibitors displayed good overall selectivity across the broader proteome both in cancer cells and in vivo (Lanning et al 2014, Pan et al 2007). While still formally representing active site-directed covalent inhibitors (Figure 1), the resulting compounds, including ibrutinib (Honigberg et al 2010, Pan et al 2007), acalbrutinib (Byrd et al 2016), afatinib (Li et al 2008, Solca et al 2012), osmertinib (Finlay et al 2014), and others (Ferguson & Gray 2018), showcased the opportunity to create transformative cancer therapies that acted by covalently modifying non-catalytic cysteines in functional protein pockets.

Wondering whether covalent chemistry could be extended beyond enzyme active sites, we set out to create advanced ABPP platforms for the global investigation of electrophilic small molecule-cysteine interactions across the proteome (Figure 1). In this approach (referred to hereafter as cysteine-directed ABPP), electrophilic compounds are added to a biological system (e.g., cancer cell, tumor) and their proteome-wide interactions determined using a broad-spectrum iodoacetamide (IA) probe that reacts with many thousands of cysteines (Backus et al 2016, Wang et al 2014, Weerapana et al 2010) (Figure 2). Initial cysteine-directed ABPP experiments investigating a small library of electrophilic fragments uncovered many hundreds of ligandable cysteines on structurally and functionally diverse proteins in human cancer cells (Figure 1), including alpha-chloroacetamides that engage the inactive precursor form of caspase-8 (Backus et al 2016, Xu et al 2020), pointing to the potential for covalent chemistry to target specific proteoforms of proteins. More generally, the vast majority of ligandable cysteines were found on proteins lacking previous evidence of binding to small molecules, highlighting the potential for covalent chemistry to greatly expand the druggable proteome. In the following sections, we will summarize select applications of cysteine-directed ABPP for the discovery and/or advancement of covalent ligands for cancer-relevant proteins.

Figure 2. Cysteine-directed ABPP for the discovery of covalent ligands.

In cysteine-directed ABPP experiments, a biological system of interest (e.g., cancer cell) is pre-treated with electrophilic compounds bearing a cysteine-directed reactive group (CRG), followed by exposure to a broad cysteine-reactive probe (e.g., iodoacetamide (IA) coupled to a reporter tag like desthiobiotin), and affinity enrichment and identification of IA-reactive cysteines by mass spectrometry (MS)-based proteomics. Covalently liganded cysteines show decreased IA probe reactivity in samples treated with an electrophilic compound, which is shown being readout in a multiplexed (tandem mass tagging (TMT))-MS proteomics format.

3.1. Cysteine-directed ABPP of genetically defined cancer states

Targeting co-dependency proteins in genetically defined cancers presents an opportunity to selectively target a specific oncogenic state, as opposed to inhibiting proteins that are essential for the viability of many cell types. Bar-Peled et al. employed cysteine-directed ABPP to identify ligandable cysteines in co-dependency proteins in KEAP1 (Kelch-like ECH-associated protein 1)-mutant non-small cell lung cancer (NSCLC) lines (Bar-Peled et al 2017). KEAP1 is an E3 ligase responsible for degrading the transcription factor NRF2 (or NFE2L2, nuclear factor-erythroid factor 2-related factor 2), and cancer cells harboring deleterious mutations in KEAP1, which are found in >20% of NSCLCs, show enhanced NRF2 activity and dependence on NRF2 for growth (Berger et al 2016, Cancer Genome Atlas Research 2014, Tsherniak et al 2017). The authors discovered that an orphan nuclear receptor NR0B1 (nuclear receptor subfamily 0 group B member 1) showing restricted expression in KEAP1-mutant NSCLC cells harbored a ligandable cysteine (C274) at its predicted protein-protein interface (Sablin et al 2008) (Figure 3a). By screening a focused collection of electrophilic small molecules, they identified a compound, BPK-29, that covalently engaged NR0B1_C274 and disrupted an NR0B1 protein complex that regulates the transcriptional output and growth of KEAP1-mutant NSCLC cells (Figure 3b). This study thus showed how cysteine-directed ABPP can identify a novel, druggable co-dependency protein in a genetically defined cancer along with chemical probes that can perturb this protein’s biomolecular interactions and function.

Figure 3. Cysteine-directed ABPP identifies NR0B1 as a co-dependency cancer target in KEAP1-mutant NSCLCs.

(a) Cysteine-directed ABPP of KEAP1-WT and KEAP1-mutant NSCLC cell lines led to the discovery of a ligandable cysteine (C274) in NR0B1 – protein with restricted expression in KEAP1-mutant NSCLC (Bar-Peled et al 2017). (b) The electrophilic compound BPK-29 disrupts the NR0B1 and SNW1 interaction by engaging NR0B1_C274, resulting in the suppression of growth of KEAP1-mutant NSCLCs. Yellow star, electrophilic group.

3.2. Integrating cysteine-directed ABPP with phenotypic screening

The phenotypic screening of small-molecule libraries offers a powerful strategy to identify compounds that inhibit the pro-tumorigenic properties of cancer cells (Moffat et al 2014). Understanding the direct targets and mechanisms of action of bioactive compounds identified in phenotypic screens, however, remains technically challenging. Covalent chemistry combined with ABPP offers an attractive way to address this problem, as the permanent adduct formed by electrophilic compounds with protein targets can be readily characterized by chemical proteomics. Here we will review examples of cancer-relevant phenotypic screens conducted with electrophilic compound libraries, where ABPP has been used to gain insights into the mechanisms of action for hit compounds.

3.2.1. Covalent compounds that suppress androgen receptor expression in prostate cancer by targeting the RNA-binding protein NONO.

RNA-binding proteins (RBPs) play fundamental roles in cellular physiology and disease, but remain underexplored as targets of chemical probes (Julio & Backus 2021). RBPs are considered a challenging protein class for chemical probe discovery due to a general lack of conventional ligand-binding pockets and functional assays (D’Agostino et al 2019, Julio & Backus 2021, Nasti et al 2017). Kathman et al. aimed to identify electrophilic compounds that decrease the transcripts encoding androgen receptor (AR) and its major splice variants in prostate cancer cells (Kathman et al 2023). AR is a major driver of prostate cancer, and AR antagonists are frontline therapies for this disease (Shen & Balk 2009). However, prostate cancers often develop resistance mechanisms to AR antagonists that include amplifying AR expression and generating point mutants or splice variants of AR that are insensitive to antagonists (Schmidt et al 2021). In searching for new mechanisms to regulate AR expression in prostate cancer, the authors phenotypically screened a focused library of electrophilic compounds and identified a series of alpha-chloroacetamides that rapidly suppress (within 4–6 h) the transcripts encoding full-length AR and the major AR-V7 splice variant.

Using cysteine-directed ABPP of a set of active and structurally related inactive control compounds (including a key pair of active and inactive enantiomers (R)-SKBG-1 and (S)-SKBG-1), the authors identified a cysteine residue (C145) in the DBHS family RNA-binding protein NONO that exclusively reacted with active compounds in prostate cancer cells. This cysteine was not found in the related DBHS proteins PSPC1 and SPFQ, indicating that the covalent ligands should show paralog-restricted activity. Surprisingly, genetic disruption of NONO did not phenocopy, but instead blocked the activity of the compounds on AR expression, suggesting that covalent ligands targeting NONO_C145 did not act as simple antagonists. Consistent with this conclusion, the authors found that the NONO ligands stabilized the binding of NONO to mRNAs, which appears to in turn hinder the processing and maturation of these transcripts (Figure 4). These ligand-stabilized NONO-mRNA complexes additionally may prevent the compensatory action of paralogs PSPC1 and SFPQ, which are upregulated in NONO ligand-treated cells or in cells with genetic loss of NONO. The authors finally demonstrated that covalent NONO ligands modulate the broader expression of gene networks in cancer cells that are enriched in RNA homeostasis and signal transduction, and suppress the growth of cancer cells in a manner that is dependent on NONO and rescued by expression of a C145S-NONO mutant. These findings designate NONO as a druggable RPB capable of being co-opted by covalent small molecules to suppress pro-tumorigenic transcriptional networks in cancer cells.

Figure 4. Integrated phenotypic screening and cysteine-directed ABPP identifies electrophilic compounds that remodel cancer transcriptomes by targeting NONO.

The electrophilic compound (R)-SKBG-1 stereoselectively reacts with C145 (red ball) of NONO, leading to stalled mRNA processing and degradation in cancer cells and preventing compensatory effects mediated by paralogous proteins PSPC1 and SFPQ (Kathman et al 2023).

3.2.2. Covalent compounds that suppress colon cancer cell growth by targeting RTN4.

Bateman et al. screened a focused cysteine-reactive compound library and found a compound DKM 3–30 that inhibits the survival of the colorectal cancer cell line SW620 (Bateman et al 2017). Using cysteine-directed ABPP, they identified C1101 on the poorly characterized ER protein, reticulon 4 (RTN4), as a target of DKM 3–30. An alkynylated DKM 3–30 was synthesized and used to confirm the engagement of endogenous RTN4 by DKM 3–30. Knocking down RTN4 by RNA-interference impaired the proliferation of SW620 cells in culture and tumor growth in a SW620-xenografted mouse model, indicating an important role for RTN4 in SW620 cell survival and proliferation. Further mechanistic studies showed that DKM 3–30 disrupted endoplasmic reticulum (ER) and nuclear envelope (NE) morphology, pointing to a role for RTN4 in the maintenance of these cellular structures.

3.2.3. Covalent compounds that promote targeted protein degradation through engaging E3 ligases.

Targeted protein degradation (TPD) is a powerful chemical biology strategy that employs small-molecule probes to induce proximity between target proteins and E3 ligases, leading to target protein ubiquitination and proteasomal degradation (Bekes et al 2022, Bondeson & Crews 2017, Cromm & Crews 2017). TPD can be mediated by two types of small molecules: 1) monofunctional compounds termed “molecular glues” that form tripartite complexes with specific E3 ligases and neo-substrate proteins (Schreiber 2021) and include drugs such as the IMiDs (Kronke et al 2015, Kronke et al 2014, Lu et al 2014, Matyskiela et al 2016); and 2) heterobifunctional compounds, known as PROTACs (proteolysis targeting chimeras), that couple E3 ligase ligands to substrate ligands via a variable linker structure (Lu et al 2015, Sakamoto et al 2001, Winter et al 2015, Zorba et al 2018). The majority of TPD studies thus far have employed established ligands for two E3 ligases, CRBN and VHL (Kannt & Dikic 2021, Schapira et al 2019). The discovery of other ligandable E3 ligases could expand the substrate scope of TPD (Bondeson et al 2018, Huang et al 2018), as well as the potential for drug resistance arising from genetic mutation of CRBN or VHL (Zhang et al 2019a).

Phenotypic screening of candidate electrophilic PROTACs combined with cysteine-directed ABPP has facilitated the identification of diverse covalent ligands that support TPD, including electrophilic natural products such as nimbolide (Spradlin et al 2019) and fragment electrophiles (Henning et al 2022b, Ward et al 2019). Nimbolide was found to target C8 on the E3 ligase RNF114, and subsequent studies have identified a fully synthetic molecule EN219 that also engages this cysteine and was converted into PROTACs that degrade the oncogenic fusion protein BCR-ABL (Luo et al 2021). Another productive approach for the discovery of E3 ligases for TPD has been to phenotypically screen focused collection of bifunctional compounds consisting of a target-binding ligand coupled to broad-spectrum electrophilic fragments, which has led to the discovery of DCAF16 and DCAF11 as E3 ligases that can support TPD through reacting with covalent PROTACs (Zhang et al 2019b, Zhang et al 2021) (Figure 5). One of the interesting features of these systems is that only modest fractional engagement of cysteines on the E3 ligases (10–40%) appears sufficient to promote robust substrate protein degradation, highlighting the catalytic nature of TPD (Bondeson et al 2015) and the potential for covalent PROTACs and molecular glues to support TPD while minimally perturbing the endogenous substrates of the co-opted E3 ligases. Other E3 ligases that have been found to support TPD through covalent PROTAC or molecular glue mechanisms include RNF4, KEAP1, FEM1B, and RNF126 (Henning et al 2022b, Pei et al 2023, Tong et al 2020, Toriki et al 2023, Ward et al 2019).

Figure 5. Integrated phenotypic screening and cysteine-directed ABPP to identify covalent PROTAC-E3 ligase systems that support targeted protein degradation (TPD).

Heterobifunctional compounds containing a specific target-binding ligand coupled to a broad-spectrum electrophilic fragment were synthesized and evaluated for the promotion of target degradation. Affinity purification-mass spectrometry (AP-MS) and ABPP were used to identify the E3 ligases (DCAF11, DCAF16) mediating TPD (Zhang et al 2019b, Zhang et al 2021).

3.3. Integrating cysteine-directed ABPP with structural biology

The Janus tyrosine kinase (JAK) family of non-receptor tyrosine kinases has four isoforms (JAK1, JAK2, JAK3, and TYK2) and plays a crucial role in signal transduction downstream of various cytokine receptors (Schwartz et al 2017). JAK inhibitors are important therapies for immunological disorders and hematological cancers, but their full utility has been hindered by undesirable side effects thought to arise from poor isoform selectivity. This is a common challenge for inhibitors that target the ATP-binding pocket of kinases (Spinelli et al 2021). Using cysteine-directed ABPP, Kavanagh et al. recently discovered an isoform-restricted, ligandable cysteine (C817) on JAK1 (Kavanagh et al 2022). Interestingly, this cysteine is located in the pseudokinase domain of JAK1 proximal to a pocket that is structurally related to the autoinhibitory myristoylation pocket in the kinase ABL that is targeted by the allosteric inhibitor asciminib used to treat BCR-ABL-driven chronic myeloid leukemia (CML) (Schoepfer et al 2018). These data suggested that the ligandable pocket in JAK1 might also be functional, and the authors proceeded to optimize covalent compounds targeting JAK1_C817 by cysteine-directed ABPP to identify a potent and selective chemical probe VVD-118313 (Figure 6) that blocked JAK1 signaling in cancer and immune cells without affecting other JAK isoform pathways (including TYK2, for which VVD-118313 appears to act as a largely silent ligand) (Kavanagh et al 2022). While the mechanism for allosteric inhibition of JAK1 by covalent ligands targeting C817 is not yet fully understood, these compounds block JAK1 transphosphorylation in cancer cells (Kavanagh et al 2022), and recent structural studies suggest that covalent ligands engaging C817 may destabilize the pose of JAK1 required to support transactivation (Caveney et al 2023).

Figure 6. Identification of electrophilic compounds targeting an isoform-restricted allosteric cysteine in JAK1.

Cysteine-directed ABPP identified a ligandable cysteine in the pseudokinase domain of JAK1 (C817) and TYK2 (C838), which is unique to these isoforms and absent in JAK2 and JAK3. A targeted cysteine-directed ABPP approach facilitated optimization of a potent and selective covalent probe (VVD-118313_ for JAK1_C817 (Kavanagh et al 2022).

Structure-guided and computational approaches combined with ABPP have also led to the discovery of the first cell-active, covalent inhibitors of the cancer-related peptidyl-prolyl isomerase Pin1 (Dubiella et al 2021, Pinch et al 2020). Pin1 is overexpressed in various cancers and modulates Ras signaling (Zhou & Lu 2016). In a series of recent studies, highly potent and selective PIN1 inhibitors were discovered by complementary approaches of structure-guided optimization of first-generation peptide inhibitors (Zhang et al 2007) and electrophilic fragment screening coupled with covalent docking (Dubiella et al 2021). Both compounds form a covalent interaction with Pin1_C113 and were shown to be highly selective by cysteine-directed ABPP (Dubiella et al 2021, Pinch et al 2020). These chemical proteomic experiments involved not only the use of an IA probe, but also a dethiobiotin-conjugated analog of the PIN covalent inhibitors in a method termed Covalent Inhibitor Target Site Identification (CITe-Id) (Browne et al 2019). Covalent PIN1 inhibitors have demonstrated a range of intriguing anti-cancer activities, including anti-proliferative and tumor-suppressive activity in neuroblastoma and PDAC models (Dubiella et al 2021, Pinch et al 2020), and, in the latter cancer models, show synergy with anti-PD-1 immunotherapy (Koikawa et al 2021).

In another example of integrating structural biology with cysteine-directed ABPP, Tao et al. recently reported an azetidine acrylamide ligand that stereoselectively and site-specifically engages C1113 on DCAF1 (Tao et al 2022). This cysteine is located in the WD40 domain of DCAF1 in proximity to the interface for binding to viral proteins that co-opt DCAF1 to promote degradation of host immune proteins such as SAMHD1 (Schwefel et al 2014). Covalent ligands targeting DCAF1_C1113 were then elaborated into PROTACs that stereoselectively and site-specifically (i.e., through C1113) promote the proteasomal degradation of proteins in cancer cells. Other research groups have also identified reversible ligands binding the WD40 domain of DCAF1 (Li et al 2023, Martin Schröder 2023) and have developed these compounds into PROTACs (Martin Schröder 2023).

3.4. “Function-first” genomic and proteomic strategies for prioritizing covalent compound-cysteine interactions in cancer-relevant proteins.

Complementing phenotypic screening and structure-guided approaches, multiple proteomic and genomic assays have been introduced to illuminate covalent liganding events that confer functional effects on cancer-relevant proteins by perturbing, for instance, protein-protein, protein-DNA, and protein-RNA interactions.

3.4.1. Function-first assays to discover covalent ligands that perturb protein-protein interactions

Proteins often function as components of larger complexes (Tsitsiridis et al 2023), and small-molecules that inhibit or stabilize protein-protein interactions (PPIs) can act as valuable chemical probes and therapeutic agents (Arkin et al 2014, Jin et al 2014, Schreiber 2021). Lazear et al. described a chemical proteomic platform that evaluates the functional effects of cysteine-directed electrophilic compounds on the complexation state of proteins in native biological systems and then integrates these data with cysteine-directed ABPP to understand mechanism of action (Lazear et al 2023) (Figure 7). Key to this approach was the investigation of focused sets of stereochemically defined electrophilic compounds such that the effects on protein complexes in cancer cells could be assigned as stereoselective shifts in protein migration measured by size-exclusion chromatography (SEC). This platform led to the discovery of a azetidine acrylamides and butynamides that target C22 of the adaptor protein PSME1 to disrupt the proteasome activator complex PA28 and cause global changes in MHC-I antigenic peptide presentation. Additionally, tryptoline acrylamides were found to stabilize a transient state of the spliceosome by stereoselectively engaging C1111 of the splicing factor SF3B1, leading to altered spliceosome composition, perturbations in RNA splicing, and impairments in cancer cell growth. The integration of proteomic data acquired by ABPP and SEC thus provided a function-first approach to identify covalent small-molecule modulators of key protein complexes involved in cancer-relevant processes such as antigen presentation and mRNA splicing.

Figure 7. A size-exclusion chromatography (SEC) approach to discover electrophilic compounds that perturb protein complexes in cancer cells.

(a) A schematic outlining the screening process for electrophilic compounds that perturb protein complexes using an SEC platform. Integration of compound-induced shifts in protein migration by SEC with cysteine-directed ABPP is then used to understand mechanism of action of compounds (Lazear et al 2023). (b) Top, an azetidine butynamide MY-45B stereoselectively engages PSME1_C22, resulting in impaired MHC-I antigenic peptide presentation. Bottom, a tryptoline acrylamide WX-02–23 stereoselectively engages SF3B1_C1111, leading to modulation of spliceosome structure and function, altered protein abundance, and inhibition of cancer cell proliferation.

3.4.2. Function-first assays to discover covalent ligands that perturb protein-DNA interactions

DNA-binding proteins, such as transcription factors, epigenetic proteins, and proteins involved in maintenance of genome integrity, play important roles in cancer. Ruprecht et al. introduced a platform for elucidating the proteomic binding landscape of 92 DNA sequences, which enabled the profiling of DNA-binding proteins in disease-relevant contexts and the impact of cysteine-reactive electrophilic compounds on these protein-DNA interactions. Integrating the data with cysteine-directed ABPP identifeid specific covalent liganding events that perturb DNA-protein interactions, including an electrophilic compound that engages C252 on MSH3 and disrupts the binding of the DNA-repair complex MSH2-MSH3 to DNA. One can envision the future extension of this platform to study how electrophilic compound-cysteine interactions perturb the functions of DNA-binding proteins in cancer cells by, for instance, ChIP-Seq or ATAC-Seq analysis.

MYC is a transcription factor that promotes cancer cell growth and survival (Dang 2012). As one of the most commonly amplified oncogenes, MYC has been a source of major interest to academic and pharmaceutical researchers for decades (McKeown & Bradner 2014). Multiple strategies have been identified to inhibit MYC regulatory pathways, including inhibitors of the bromodomain-containing protein BRD4 (Filippakopoulos et al 2010, Garnier et al 2014) and cyclin-dependent kinases CDK7/9 (Hashiguchi et al 2019, Minzel et al 2018), as well as stabilizers of the Max homodimer (Struntz et al 2019). However, directly targeting MYC with small molecules has proven challenging. Recently, Boike et al. screened a focused electrophilic compound library to identify compounds that can prevent MYC-MAX dimerization in vitro and inhibit MYC transcriptional activity in cells. (Boike et al 2021). Cysteine-directed ABPP revealed that a hit compound acrylamide EN4 reacted with C171 on MYC, which is located in an intrinsically disordered region, suggesting that such flexible regions might adopt transiently structured states that can engage and be trapped by electrophilic small molecules.

3.4.3. Function-first assays to discover essential cysteines in cancer-dependency proteins

ABPP experiments performed to date have identified covalent ligands (fragments and/or more elaborated compounds) for many hundreds of cysteines in the human proteome (Boatner et al 2023). However, the functional impact of most of these electrophilic small molecule-cysteine interactions remains unknown. In a recent study (Haoxin Li 2022), Li et al. have introduced an integrated base editing and ABPP strategy to globally assess the functionality and ligandability of cysteine residues in the context of cancer dependency proteins. The authors specifically altered the codons encoding cysteines and neighboring residues using programmable base editors (Gaudelli et al 2017, Komor et al 2016), which led to evidence of essentiality for > 1,300 cysteines on diverse classes of cancer dependency proteins. Combining these data with cysteine-directed ABPP led to a list of ~110 essential and ligandable cysteines, including sites targeted by covalent drugs (e.g., C797 on EGFR; C528 on XPO1) and additional sites on proteins with Common Essential or Strongly Selective designations in the Cancer Dependency Map (Tsherniak et al 2017). In related work, Benns et al. have used a CRISPR-based oligo recombineering approach to assess the functionality of cysteines in Toxoplasma gondii (Benns et al 2022). The emergence of gene editing platforms that can be combined with ABPP to offer residue-level functional and ligandability maps should offer valuable guidance for the future development of covalent chemical probes and drugs.

3.5. Assessing the proteome-wide selectivity of covalent cancer drugs by cysteine-directed ABPP

ABPP has not only proven useful for the discovery of “first-in-class” covalent ligands for cancer-relevant proteins, but also for evaluating the global selectivity of advanced covalent probes and drug candidates. Original applications of ABPP assessing covalent kinase inhibitors, such as ibrutinib, afatinib, and osmeritinb, have helped to assign off-target cysteines/proteins for these drugs across the human proteome (Lanning et al 2014, Niessen et al 2017). In general, these compounds have been found to show good proteome-wide selectivity, with handfuls of notable off-targets, but one could argue that this specificity benefited from recognition elements that have strong reversible binding affinity to the ATP-binding pocket of kinases. More recent covalent inhibitors targeting oncogenic G12C-KRAS originated from electrophilic fragments (Figure 8), and it was therefore of great interest to understand the types of global specificity that could be achieved with this approach to covalent drug development.

Figure 8. Covalent inhibitors targeting G12C-KRAS G12C and the utilization of cysteine-directed ABPP to assess their proteome-wide selectivity.

Cysteine-directed ABPP data: ARS-1620 (Janes et al 2018), sotorasib (Canon et al 2019), adagrasib (Fell et al 2020).

Oncogenic KRAS is a small GTPase that carries gain-of-function mutations in about 30% of all human cancers, making it a long-standing target of great promise for drug development (Simanshu et al 2017). Historically, RAS proteins have been considered “undruggable” due to an apparent lack of obvious allosteric regulatory sites and a picomolar binding affinity for GTP/GDP (Cox et al 2014). However, the G12C-KRAS mutation, present in approximately 10–13% of lung adenocarcinoma cases, provided a unique opportunity for covalent inhibitor development (Skoulidis et al 2021). Ostrem et al. discovered electrophilic small molecules that target a new pocket beneath the Switch II loop region specifically in the GDP-bound state of G12C-KRAS (Figure 8), thereby blocking Son of Sevenless (SOS)-mediated GDP/GTP exchange (Ostrem et al 2013). From this exciting starting point, Patricelli, Janes et al. reported more advanced covalent probes ARS-853 and ARS-1620 that potently inhibited G12C-KRAS in cancer cells and in vivo (Figure 8) (Janes et al 2018, Patricelli et al 2016), and, through cysteine-directed ABPP, the authors further demonstrated remarkable proteome-wide specificity for these compounds, with ARS-1620 only engaging two additional cysteines across more than 8,500 quantified cysteines in cancer cells. These milestone discoveries, together with other efforts in the pharmaceutical industry, have led to the development of multiple FDA-approved G12C-KRAS inhibitors (sotorasib (also known as AMG510) (Nakajima et al 2021) and adagrasib (Fell et al 2020)), for treating non-small cell lung cancer (NSCLC) (Figure 8). Each of these drugs has been evaluated by cysteine-directed ABPP as part of its development and found to show excellent proteome-wide selectivity for G12C-KRAS (Canon et al 2019, Fell et al 2020, Nakajima et al 2021). These advances showcase the high level of specificity that can be achieved with covalent chemistry, even when targeting a protein previously considered undruggable, and demonstrate how ABPP can assist in the assessment and optimization of the proteome-wide selectivity of covalent cancer drugs.

5. Summary and outlook

In this Review, we have attempted to highlight the growing impact of covalent chemistry and chemical proteomics on chemical probe and drug development in cancer. While covalent chemistry can offer several advantages over more classical reversibly binding compounds (Boike et al 2022, Singh et al 2011), we have focused here on what we view as the most exciting opportunity – to expand the druggability of cancer proteomes. The realization of this objective also underscores the value of the integrated application of innovations in covalent chemistry and chemical proteomics, where novel electrophilic compound designs can be screened by ABPP against 20,000+ cysteines on proteins from diverse structural and mechanistic classes (Backus et al 2016, Bar-Peled et al 2017, Blewett et al 2016, Crowley et al 2021, Vinogradova et al 2020, Weerapana et al 2010). These screens can be performed in living cancer cells, enabling the discovery of cryptic druggable pockets that may have been overlooked by more traditional studies with purified proteins. One compelling example is the recent discovery of the allosteric, isotype-restricted JAK1 inhibitor VVD-118313 (Kavanagh et al 2022). Initial chemical proteomic data would suggest that this site on JAK1 is quite ‘druggable’ in that it has been discovered to react with a range of electrophilic fragments and more elaborated covalent compounds and, in some cases, with excellent potency (low-nM engagement in cells) and proteome-wide selectivity (Kavanagh et al 2022). Given the extensive efforts within the pharmaceutical industry to develop JAK inhibitors (Schwartz et al 2017) it merits asking – why wasn’t this allosteric site discovered previously? We would posit that such an allosteric, druggable site may be paradoxically much easier to discover in cells than with purified proteins, which are often studied as simplified domains (e.g., the kinase domain in the case of JAKs) that may obscure allosteric modes of regulation. We accordingly wonder if terms like druggable and undruggable are misguided, and we might be better served considering terms such as “assayable” and “unassayable” when judging protein targets. To this end, chemical proteomics offers a unique way to assay any endogenous protein for small molecule interactions directly in native biological systems.

Beyond serving as a source of first-in-class chemical probes for cancer-relevant proteins, covalent chemistry has had an incredible impact on modern cancer drug development. Covalent inhibitors targeting a diverse range of targets, including kinases (e.g., BTK, EGFR, FGFR), nuclear import proteins (XPO1), and GTPases (G12C-KRAS), have been approved for treating cancers in recent years. The various attributes of these drugs have been extensively covered in other reviews (Boike et al 2022, Singh et al 2011), but we wanted to call attention to one attractive feature – the targeting of paralog-restricted cysteines. While this is most obvious for G12C-KRAS, which possesses a ‘neo-cysteine’ introduced through somatic mutation in cancers, original covalent kinase inhibitors were constructed on scaffolds that reversibly bound many additional kinases lacking the covalently drugged cysteine shared by BTK and EGFR. In the course of our cysteine-directed ABPP studies, we have been struck by the number of paralog-restricted ligandable cysteines in the human proteome. We noted JAK1_C817 above, but the additional discovery of functional covalent ligands targeting paralog-restricted cysteines in other classes of proteins such as RNA-binding proteins (NONO_C145 (Kathman et al 2023), TOE1_C80 (Haoxin Li 2022)) and adaptor proteins (PSME1_C22 (Lazear et al 2023)) points to the potential for covalent chemistry to offer powerful small-molecule probes for studying biological processes that might otherwise be vulnerable to paralog compensation when genetically manipulated (Kathman et al 2023).

We should also comment on the limitations of chemical proteomics-guided discovery of covalent ligands. Early applications of cysteine-directed ABPP that mostly surveyed electrophilic fragments were confronted with challenges of converting low-potency liganding events of unclear structure-activity relationships (SARs) into advanced chemical probes suitable for cell biological studies. This concern has recently been addressed by deploying more structurally sophisticated covalent compound libraries constructed, for instance, on the principles of diversity-oriented synthesis (Gerry & Schreiber 2020) that emphasize densely functionalized sp3-rich cores with stereochemically defined centers such that each small molecule-cysteine interactions mapped in cancer cells can be assessed for stereoselectivity, an SAR property that may be considered predictive of actionable, druggable pockets (Scott et al 2022). Stereoselective covalent ligands have now been discovered for numerous cysteines on diverse protein classes (Evert Njomen 2023, Tao et al 2022, Vinogradova et al 2020), which highlights another challenge – assigning functional consequences to these small molecule-protein interactions. Original covalent ligands that engaged catalytic or non-catalyic residues in enzyme active sites could be inferred to act as inhibitors of these proteins; but, many covalent ligands discovered by chemical proteomics engage cysteines found at cryptic and functionally ill-defined pockets on proteins. We have attempted to highlight in this review how complementary approaches, including phenotypic screening and the large-scale analysis of general biochemical and cellular properties such as protein-protein/DNA interactions and cancer cell proliferation, are providing a way to assign functionality to covalent ligand-cysteine interactions. Nonetheless, further innovation is needed on this front, especially considering the diverse ways that covalent ligands have been found to alter protein functions, including not only blocking (Boike et al 2022), but also stabilizing protein-protein (Lucero et al 2023) and protein-RNA (Kathman et al 2023) interactions, as well as modulating (Kavanagh et al 2022) and specifically targeting (Fu et al 2023, Shi & Carroll 2020) the post-translational modification state of proteins in cancer cells. Additionally, as has been shown for the E3 ligases, covalent ligands can be co-opted to induce neo-functional outcomes, such as targeted protein degradation (Lu et al 2022). This concept has recently been extended to deubiquitinases, where covalent ligands targeting C23 in OTUB1 were converted into heterobifunctional compounds that stabilize post-translationally downregulated tumor suppressors, offering a new strategy for targeting highly proliferated tumors (Henning et al 2022a).

We conclude by emphasizing the importance of future innovations in the design of covalent compound libraries for continuing to expand the druggable cancer proteome. A cursory comparison of recent cysteine-directed ABPP data indicates very limited overlap in the cysteines that are stereoselectively engaged by focused electrophilic compound sets constructed from different scaffolds (e.g., azetidine (Lazear et al 2023) vs tryptoline (Lazear et al 2023, Vinogradova et al 2020) acrylamides), suggesting that continued efforts to diversify recognition chemistry should lead to the discovery of new covalent ligand-cysteine interactions. Another productive source of diversity is the reactive chemistry itself, where the acrylamide can be exchanged for an increasingly large number of groups that have tempered electrophilicity suitable for selective covalent modification of cysteines in the proteome (e.g., butynamides (Kavanagh et al 2022), fluorochloroacetamides (Sato et al 2020)). Finally, while cysteine, being the most intrinsically nucleophilic amino acid in proteins, has been the focus for most covalent chemistry efforts to date, its frequency of occurrence in protein is low (0.7%) (Brooks & Fresco 2002), and many druggable pockets may be cysteine-poor or -deficient. Ongoing efforts to expand the scope of covalent chemistry include targeting amino acids such as lysine (Abbasov et al 2021, Cuesta & Taunton 2019, Hacker et al 2017, Ward et al 2017, Yang et al 2022, Zhao et al 2017), serine (Fadeyi et al 2017), tyrosine (Hahm et al 2020), histidine (Jia et al 2019), aspartate/glutamate (Bach et al 2020), and methionine (Lin et al 2017). Most of these covalent chemistries can also be combined with residue-directed ABPP platforms, thus providing a broad emerging foundation for the discovery of new covalent chemical probes and drugs against cancer-relevant proteins.

Abbreviation

ABPP

activity-based protein profiling

AP-MS

affinity purification-mass spectrometry

AR

androgen receptor

BTK

Bruton’s tyrosine kinase

CITe-Id

covalent inhibitor target site identification

CML

chronic myeloid leukemia

CRG

cysteine-directed reactive group

EGFR

epidermal growth factor receptor

ER

endoplasmic reticulum

IA

iodoacetamide

JAK

Janus tyrosine kinase

KEAP1

Kelch-like ECH-associated protein 1

MS

mass spectrometry

NE

nuclear envelope

NFE2L2

nuclear factor-erythroid factor 2-related factor 2

NR0B1

nuclear receptor subfamily 0 group B member 1

NSCLC

non-small cell lung cancer

PPI

protein-protein interaction

PROTAC

proteolysis targeting chimeras

SAR

structure-activity relationship

SEC

size-exclusion chromatography

SOS

son of sevenless

TMT

tandem mass tagging

TPD

targeted protein degradation