Discovering new biology with drug-resistance alleles

Abstract

Small molecule drugs form the backbone of modern medicine’s therapeutic arsenal. Often less appreciated is the role that small molecules have had in advancing basic biology. In this Review, we highlight how resistance mutations have unlocked the potential of small molecule chemical probes to discover new biology. We describe key instances in which resistance mutations and related genetic variants yielded foundational biological insight and categorize these examples on the basis of their role in the discovery of novel molecular mechanisms, protein allostery, physiology and cell signaling. Next, we suggest ways in which emerging technologies can be leveraged to systematically introduce and characterize resistance mutations to catalyze basic biology research and drug discovery. By recognizing how resistance mutations have propelled biological discovery, we can better harness new technologies and maximize the potential of small molecules to advance our understanding of biology and improve human health.

From artemisinin in wormwood plants to statins in a vial, small molecules throughout history have served as transformative therapeutic agents, altering disease courses in nearly every area of medicine. Beyond their direct impact in human health, small molecules have also led to countless biological advances. For instance, the serendipitous discovery of penicillin ushered in the field of bacterial cell wall biology, while deciphering the mechanism of action of the antirheumatic agent colchicine led to the isolation of tubulin and jumpstarted the field of microtubule biology^1,2. More recently, the characterization of small molecules identified by phenotypic screening led to the identification of ferroptosis, a new form of cell death³. These small molecules and numerous others were critical to the advancement of basic biology.

Although it is clear that small molecules have led to fundamental biological insights, their ability to reveal molecular mechanisms hinges on an understanding of which cellular target(s) they might engage to elicit their observed effects. Many small molecule-initiated discoveries arose from finding these very cellular targets through a process called target identification. As small molecules can bind to many different macromolecules within the cell, deconvoluting and identifying those responsible and/or sufficient for the biological effects observed can be challenging.

While many approaches exist to identify the protein target(s) of a small molecule, identification of a mutation in the presumed protein target that suppresses the effects of the small molecule is the gold-standard method for confirming on-target engagement⁴. We will refer to such a genetic variant as a resistance mutation throughout this Review. In the development and use of targeted therapies, resistance mutations in a drug’s target are often encountered in the clinic, leading to continued disease progression. While a major challenge from a clinical perspective, these drug-resistance mutations ultimately validate that perturbing the intended target mediates the therapeutic effect of the small molecule.

Beyond confirming small molecule on-target engagement for drug discovery, resistance mutations have had pivotal roles in advancing the mechanistic understanding of fundamental biological processes. In this Review, we highlight several examples where the identification and use of resistance mutations led to advances in basic biology. In our discussion, we will consider only coding resistance mutations that occur in the target and maintain protein expression (for example, missense mutations, in-frame insertions or in-frame deletions). Although of great importance, resistance mutations that occur outside the target protein or lead to complete loss of protein expression (that is, ‘knockout’) are outside the scope of this Review^5,6. We have grouped these selected examples into four themes that reflect the ways in which resistance mutations and related genetic variants have advanced biology across diverse fields (Fig. 1a). Informed by these precedents, we conclude with perspectives on how emerging mutagenesis technologies may be used to systematically identify resistance mutations to address questions in biology and medicine.

Fig. 1 |. Resistance mutations identify small molecule targets and reveal new biology.

a, Schematic illustrating the themes outlined in this piece. PDB: 5HXB, 5MO4. b, Top, chemical structure of rapamycin. Bottom, structural view of the ternary complex containing rapamycin (yellow), Homo sapiens FKBP (gray) and the H. sapiens FRB domain of mTOR (green). The FKBP residues homologous to those identified in Heitman et al. are shown in blue¹³. PDB: 1FAP. c, Left, chemical structure of indisulam. Right, structural view of the ternary complex containing an analog of indisulam (E7820, yellow), RBM39 (gray) and DCAF15 (green). The RBM39 residues identified in Han et al. are shown in blue²¹. PDB: 6Q0R.

Uncovering molecular mechanisms by target identification

Small molecule natural products have not only been exploited for the treatment of disease for millennia but also have had a pivotal role in the advancement of chemistry and biology. Scientists have used resistance mutations to discover how these natural products function, primarily through target identification. In doing so, they have discovered many previously uncharacterized proteins, complexes and molecular mechanisms that are of broad biological relevance.

In the mid-twentieth century, studies involving resistance mutations had critical roles in characterizing the key players involved in the central dogma. In these studies, scientists looking to characterize the mechanism of action of influential natural product therapeutic agents serendipitously discovered critical components and aspects of the transcription and translation machinery. Identification of resistance mutations to the rifamycins, an antibiotic class critical for the treatment of tuberculosis and leprosy, led to mapping of the first RNA polymerase gene in any organism^7,8. In the area of translation, resistance mutations to the antibiotic streptomycin were used to characterize fundamental properties of the ribosome^9–11. These studies demonstrated for the first time that translational fidelity is dictated in part by the speed of translation by the ribosome, contrary to early hypotheses that translational accuracy resulted solely from the free energy provided by codon–anticodon pairing¹². In these seminal studies, resistance mutations transformed natural product therapeutic agents into chemical probes that yielded broad, fundamental biological discoveries.

One paradigmatic example of how resistance mutations pioneered new biology is the discovery of the target of rapamycin (TOR) proteins. In 1991, yeast geneticists investigated the mechanism of action of the immunosuppressant rapamycin (Fig. 1b). Taking advantage of rapamycin’s antifungal activity, rapamycin resistance mutations were identified by selecting for yeast strains that were able to grow in the presence of rapamycin¹³. Reassuringly, the geneticists identified mutations in FPR1, the yeast gene encoding FKBP, the previously identified binding partner of rapamycin (Fig. 1b)¹⁴. Unexpectedly, they also discovered mutations in two unidentified genes, which they named target of rapamycin (TOR1 and TOR2). They noted that these resistance mutations in TOR1 and TOR2 exhibited nonallelic noncomplementation with the resistance mutations in FPR1, a genetic phenomenon that suggests that the proteins might participate in the same protein complex. On the basis of this result, the authors hypothesized that the mechanism of rapamycin likely involves the formation of a protein complex between FKBP and the novel protein TOR¹³. Following this seminal discovery, the mammalian homolog of TOR (mTOR) was identified and recognized as a vital nutrient sensor in mammalian cells^15–17. The original hypothesis that rapamycin is a mediator of chemically induced protein heterodimerization was later confirmed, making rapamycin, along with the related molecules cyclosporin A and FK506, the first identified ‘molecular glue’ compounds inducing de novo binding of two proteins^18,19.

Resistance mutations discovered with the goal of deciphering small molecule mechanism of action have led to the identification of ‘molecular glue degraders’. In one recent study, scientists identified resistance mutations to the anticancer sulfonamides using an approach similar to the previously described method DrugTargetSeqR^20–22. These methods use cancer cell lines with defects in mismatch repair, enabling accelerated formation of resistance mutations that are then identified by transcriptome or exome sequencing. Using this strategy, several resistance mutations to the anticancer sulfonamide indisulam were identified in the RNA splicing factor RBM39 (Fig. 1c). The authors then found that RBM39 was degraded in a proteasome-dependent manner upon indisulam treatment owing to chemically induced recruitment of RBM39 to the ubiquitin E3 ligase substrate receptor DCAF15. With these findings and those of subsequent studies, the anticancer sulfonamides were nominated as one of only a handful of molecular glue degrader scaffolds that have helped give rise to the burgeoning field of targeted protein degradation.

Discovering allostery through mutagenesis

Mutational analyses have served as important tools for determining the structure and function of many protein targets. In this manner, resistance mutations have been used strategically to nominate specific genetic variants that alter protein function for in-depth investigation. These mutations have been identified by strategies ranging from structure-guided site-directed mutagenesis to unbiased mutational scanning and, once identified, have been used with great success to uncover important biology in a wide range of fields.

Historically, resistance mutations were used for the initial characterization of proteins that were challenging to interrogate through traditional structural methods. Studies that identified mutant sodium channels resistant to tetrodotoxin (TTX) (Fig. 2a), a potent small molecule neurotoxin that inhibits voltage-gated sodium channels, were particularly influential in identifying key functional residues within the sodium channel pore. Specifically, a series of studies investigating resistance mutations that block TTX action led to correct identification of the sodium channel selectivity filter—the residues responsible for the selectivity of the sodium channel for sodium over divalent cations such as calcium—20 years before publication of the first high-resolution sodium channel structure (Fig. 2a)^23–25. In another example, resistance mutations to a natural product toxin, α-amanitin, were used to elucidate the function of different structural elements of RNA polymerase II in mammalian cells²⁶. One such study using α-amanitin resistance mutations was the first to demonstrate the essentiality of the RNA polymerase II C-terminal domain, which is now established to be a key regulatory center for transcription²⁷. Given the challenging nature of solving membrane protein and protein complex structures, the use of resistance mutations was critical for the advancement of these fields.

Fig. 2 |. Resistance mutations unveil protein structural information and mechanisms of protein allosteric regulation.

a, Top, chemical structure of TTX. Bottom, structural view of TTX (yellow) engaging the Periplaneta americana Na_vPaS channel (gray). The Na_vPaS residues homologous to those identified as resistance mutations to TTX in Noda et al. and Terlau et al. are shown in blue^23,24. These residues are responsible for the selectivity of the channel for sodium over other cations. PDB: 6A95. b, Left, chemical structure of imatinib. Right, a structural view of ABL kinase is shown with resistance mutations identified by Azam et al. highlighted (red, active residues; blue, distal site residues)²⁸. PDB: 5MO4. c, Left, chemical structure of palbociclib. Right, a structural view of CDK6 kinase is shown with resistance mutations identified by Persky et al. highlighted (red, active site residues; blue, distal site residue)³⁵. PDB: 2EUF.

While site-directed mutagenesis can be used to explicitly test and identify resistance mutations involving preconceived candidate residues, identifying resistance mutations in an unbiased fashion can uncover unanticipated protein regions and structural motifs within a target that are relevant to small molecule mechanism of action and target function. In a key example, resistance mutations distal to the binding site of the tyrosine kinase inhibitor imatinib were identified by a random high-throughput mutagenesis approach²⁸. These resistance mutations had an important role in defining the structural mechanism of autoinhibition of ABL kinase (Fig. 2b)²⁹. To accomplish this, the investigators cloned the ABL kinase sequence using a repair-deficient bacterial strain, which resulted in a diverse library of ABL variants. Upon overexpressing these ABL variants in cells and treating with imatinib, ABL resistance mutations were identified within the imatinib-binding site that recapitulated known clinical resistance mechanisms. Simultaneously, several ABL mutations conferring partial resistance were identified outside the imatinib-binding site (Fig. 2b, blue spheres). At the time of this study, only the kinase domain of ABL had been structurally resolved. In the absence of a full ABL structure, the authors mapped these resistance mutations to their locations on a homologous kinase, SRC. In doing so, the authors found that the resistance mutations outside the small molecule-binding site clustered in structural regions known to be important for SRC autoinhibition. This led the authors to propose a model of ABL allostery in which these domains, highlighted by resistance mutations, are critical for ABL autoinhibition. The model was validated by structural and biochemical studies published simultaneously^30,31. Discovery of the ABL autoinhibition mechanism by these studies led to the development of allosteric ABL inhibitors that are now in late-stage clinical development for the treatment of BCR-ABL-positive leukemias^32,33.

Akin to this systematic approach, in a more recent study, investigators used a saturation mutagenesis strategy known as mutagenesis by integrated tiles and sequencing (MITE-seq) to uncover resistance mutations to kinase inhibitors in an unbiased manner^34,35. In MITE-seq, each amino acid in the protein-coding sequence is individually varied to each of the 19 other canonical amino acids, allowing for an exhaustive sampling of all possible amino acid substitutions. In addition to identifying mutations within the well-annotated binding site of ATP-competitive kinase inhibitors, the investigators found resistance mutations in an unannotated region that rendered the kinases hyperactive (Fig. 2c). With these results, they hypothesized that this allosteric site, which they termed the Keymaster site, might represent a novel and general autoinhibitory kinase mechanism. Together, these examples highlight how resistance mutations outside the small molecule-binding site can be used to find novel allosteric sites in protein targets that may represent vulnerabilities that can be exploited therapeutically.

Characterizing physiology with genetic variation

Many small molecules can have distinct effects across different tissues, individuals or species. Among individuals and species, these differences can often arise from genetic variation, and such genetic variants can be conceptualized as resistance mutations when present in organisms resistant to the effects of a small molecule. These resistance mutations, applied across different tissues, organisms and species, have deepened our understanding of organismal physiology and pharmacology.

Small molecules can have pleiotropic effects on physiology by binding multiple targets across various tissues, where even engaging the same target in different tissue types might elicit different cell-type-specific responses. One can exploit resistance mutations in the target gene specifically in the tissue of interest to disentangle this pleiotropy. This concept has been applied to investigate the mechanism of action of benzodiazepines, a class of anxiolytic agents that allosterically pan-activates γ-aminobutyric acid type A (GABA_A) receptors in the brain (Fig. 3a). There are several GABA_A receptor subtypes present across distinct functional and anatomical regions of the brain, each eliciting specific effects in response to activation by classical benzodiazepines such as diazepam and alprazolam. However, before the late 1990s, how activation of each of these receptor subtypes and thus brain regions by diazepam contributed to the overall behavioral response to systemic diazepam administration remained unknown. By generating mouse models with diazepam-resistance mutations within each of the GABA_A receptor subtypes, investigators were able to modulate the effect of diazepam in specific brain regions and hence deconvolute the different receptor subtypes and brain regions that were responsible for each aspect of a mouse’s response to benzodiazepines (Fig. 3a). In a series of studies, investigators determined that the α₁ GABA_A receptor subtype present in the thalamus and cortex mediated sedative, amnesic and anticonvulsant effects, whereas the α₂ GABA_A receptor subtype present in the limbic system mediated anxiolytic effects (Fig. 3b)^36–38. This finding prompted a search for selective α₂ GABA_A receptor subtype agonists, which yielded compounds that entered clinical development as nonsedative anxiolytics³⁹. In this context, resistance mutations were critical in elucidating the functions of different regions and receptor subtypes in the brain, ultimately spurring the development of a novel class of anxiolytic therapeutic candidates.

Fig. 3 |. Resistance mutations elucidate physiology and provide insight into species differences in small molecule mechanism of action.

a, Left, chemical structure of alprazolam. Right, structural view of alprazolam (yellow) engaging the H. sapiens GABA_A receptor (gray). The resistance mutation used in Rudolph et al. is shown in blue³⁶. PDB: 6HUO. b, Brain regions expressing α₁ and α₂ GABA_A receptor subtypes responsible for the physiological effects of benzodiazepine action. The location of the α₁ subtype receptors in the thalamus and cortex are highlighted in blue, whereas the location of the α₂ subtype receptors in the limbic system are highlighted in red. c, Left, chemical structure of abacavir. Right, structural view of abacavir binding the protein encoded by the HLA-B*57:01 allele. Sites of natural genetic variation in an individual’s HLA locus that block abacavir binding are shown in blue⁴¹. PDB: 3VRJ. d, Left, chemical structure of thalidomide and its structural analog, pomalidomide. Right, structural view of pomalidomide (yellow) forming a ternary complex with H. sapiens CrBN (gray) and H. sapiens SALL4 zinc-finger 2 (purple). The locations of Mus musculus Sall4 residues that differ from those found in H. sapiens SALL4 are highlighted in light blue. The location of the M. musculus CRBN (mCRBN) Ile391 residue is highlighted in dark blue. Bottom, the sequences of M. musculus Sall4 and H. sapiens SALL4. Variation is highlighted in light blue. PDB: 3WX2, 6UML.

Genetic variation can account for differential side effect profiles of small molecule therapeutic agents among different individuals. Such genetic variants were critical for uncovering the cause of toxicity in some individuals from abacavir, a reverse transcriptase inhibitor used to treat HIV/AIDS (Fig. 3c). Abacavir hypersensitivity syndrome (AHS) is a drug reaction that predominantly afflicts individuals with a specific allele in their human leukocyte antigen (HLA) locus⁴⁰. Individuals without this AHS-sensitizing HLA allele are resistant to the side effect profile and may be thought of as having resistance ‘mutations’ in their HLA locus. The HLA protein encoded by this locus is responsible for presentation of self-peptides, which guides the immune system to distinguish self from foreign antigens. In a study comparing AHS-sensitizing and AHS-resistant HLA alleles, it was found that abacavir binds HLA protein encoded by the AHS-sensitizing allele and causes it to display an altered peptide repertoire (Fig. 3c)⁴¹. This leads to alteration of the immune system’s recognition of self and to widespread inappropriate immune activation that manifests as AHS. HLA proteins encoded by AHS-resistant alleles do not bind abacavir, and thus individuals with these alleles are spared from AHS. Before this study, it was unknown that small molecules could bind to HLA receptors and change their function to alter immune response. In this study, through the characterization of sensitizing and resistant HLA alleles, small molecule binding to HLA protein was proposed as a general mechanism underlying HLA-associated drug toxicities, advancing the understanding of immune-related drug reactions.

Subtle natural genetic variation in the target of a small molecule across different species can have a large impact on a small molecule’s effect in different organisms. This was infamously demonstrated in the tragic medical disaster of thalidomide (Fig. 3d). In the late 1950s, thalidomide, a molecule marketed as an antiemetic for the treatment of morning sickness, was shown to cause numerous teratogenic effects in newborn infants and was removed from the market⁴². Despite thalidomide’s devastating side effects in pregnant women, structural analogs were later found to be an effective treatment for a variety of hematologic malignancies and are now approved by the US Food and Drug Administration (FDA) for the treatment of leprosy, multiple myeloma and myelodysplastic syndrome. Although humans experience teratogenicity from this molecule, several model organisms do not. Characterization of genetic variations in organisms resistant to thalidomide’s effects led to both an understanding of this medical disaster and fundamental discoveries regarding small molecule mechanism of action.

Thalidomide acts as a molecular glue degrader that induces de novo protein–protein interactions between the ubiquitin ligase substrate receptor cereblon (CRBN) and several zinc-finger transcription factors, causing their proteolytic degradation^43,44. It was well known that mice and rats were resistant to both the teratogenic and antineoplastic properties of thalidomide⁴⁵. The molecular basis for this resistance was recently uncovered by characterizing the genetic differences between mouse and human CRBN and SALL4, a key developmental transcription factor degraded by thalidomide^46–48. It was found that a single residue in mouse Crbn, Ile391, when mutated to valine—the corresponding residue found in human CRBN—could render mice sensitive to the antineoplastic effects of thalidomide treatment⁴⁹. Mutating five additional residues in mouse Sall4 to the corresponding residues in human SALL4 was found to induce Sall4 degradation (Fig. 3d)⁴⁷. It was previously shown that SALL4 loss-of-function mutations lead to teratogenic syndromes with effects similar to those from thalidomide-induced teratogenicity, suggesting that SALL4 degradation may be responsible for these effects⁵⁰. In this sense, the mouse Crbn Ile391 residue and five mouse Sall4 residues may be thought of as natural thalidomide resistance mutations that block thalidomide’s teratogenic activity in mice. Interestingly, mouse Crbn still binds to thalidomide, despite the fact that mice are resistant to its antineoplastic effects. However, once bound, thalidomide is not able to induce de novo protein–protein interactions with many of the proteins corresponding to human substrates owing to the increased steric bulk of the mouse isoleucine residue in comparison to the human valine residue (Fig. 3d)⁴⁶. This suggests that further characterizing resistance mutations to molecular glue compounds may yield highly sought-after design principles for this important and rapidly growing small molecule modality. Furthermore, the lack of thalidomide-mediated teratogenicity in rodents may have been a major contributor to the unanticipated wave of tragic thalidomide-induced birth defects, emphasizing the importance of considering species-specific effects and natural genetic variation in preclinical drug development⁵¹.

Modulating both sides of the protein–ligand interface

As chemistry and genetics have advanced, the ability to rapidly create small molecule and protein variants has greatly increased. Advancement of these fields has allowed for simultaneous perturbation of both sides of the small molecule–protein interface, resulting in allele-selective compounds that are selective for proteins with engineered sensitizing mutations. In this sense, proteins without a sensitizing mutation contain a resistance ‘mutation’ blocking the action of the compounds. The side-by-side use of resistance mutations and allele-selective compounds has advanced knowledge of small molecule mechanism of action and allowed the dissection of complex cell signaling pathways with exquisite control.

Such approaches were essential in uncovering the importance of dimerization in signaling by the oncogenic kinase BRAF. BRAF is a component of the growth-promoting MAPK signaling pathway that is commonly mutated and activated across many cancer subtypes (Fig. 4a). BRAF is frequently hyperactivated in melanoma by a point mutation that results in substitution of Val600 with a glutamic acid residue (BRAF(V600E)) (Fig. 4a). Inhibitors targeting the BRAF(V600E) mutant were developed and showed remarkable clinical efficacy, resulting in the FDA approval of the BRAF inhibitor vemurafenib for metastatic melanoma in 2011 (refs.^52,53). The clinical efficacy of these BRAF inhibitors was initially surprising given that inhibitors of downstream kinases in the MAPK pathway (that is, MEK and ERK) had comparatively modest clinical response rates and exhibited dose-limiting toxicities⁵⁴. The mechanism underlying this surprising discrepancy was uncovered by a series of studies detailing differences in the effects of vemurafenib between cells harboring wild-type and mutant BRAF.

Fig. 4 |. Allele-selective kinase inhibitors and resistance mutations uncover the importance of BRAF dimerization in the MAPK signaling pathway.

a, Left, chemical structure of vemurafenib. Middle, schematic illustrating the MAPK signaling pathway in cells with wild-type BRAF. In these cells, BRAF is inactive in its monomeric form. Upon activation of Ras by upstream growth factor receptors, BRAF is phosphorylated, which leads to its dimerization. BRAF dimers represent the active form of the kinase and mediate downstream MAPK pathway activation through phosphorylation of the MEK and ERK kinases. Right, schematic illustrating the MAPK signaling pathway in cells with BRAF (V600E). BRAF (V600E) is active as a monomer. Thus, constitutive MAPK pathway activation through phosphorylation of MEK and ERK occurs in the absence of active Ras. b, Top, chemical structure of JAB34. The phenylbromide highlighted in green acts as a bump, complementary to the hole mutation engineered into SRC kinase. The acrylamide highlighted in blue is the cysteine-reactive portion of the molecule. Bottom, wild-type (WT) and mutant SRC kinases and their sensitivity to inhibition by JAB34. Half-maximal inhibitory concentration (IC₅₀) values are from Blair et al.⁶⁰. c, Schematic illustrating the use of JAB34 to uncover the mechanism of BRAF inhibitor-mediated dimer transactivation. d, Schematic showing the mechanism underlying resistance of the BRAF (V600E) p61 isoform to vemurafenib inhibition.

Although initial BRAF inhibitors such as vemurafenib potently inhibit BRAF activity in melanoma cells with the BRAF^V600E mutation, they paradoxically activate BRAF activity in cells with wild-type BRAF present in all other tissues throughout the body (Fig. 4a)⁵⁵. The mechanism underlying this counterintuitive activation was established using allele-selective kinase inhibitors^56,57. Allele-selective inhibitors are those designed to selectively inhibit a target of interest with engineered sensitizing mutations. This can be accomplished through the use of a steric protrusion (‘bump’) on the inhibitor that is complementary to a mutation in the inhibitor-binding site (‘hole’) or through the inclusion of an electrophilic functional group on the inhibitor that will covalently bind to a nucleophilic amino acid engineered into the inhibitor-binding site (Fig. 4b)^58–60. Wild-type proteins without the sensitizing mutation are thus resistant to the action of these inhibitors, allowing for selective inhibition of a protein of interest within the complex cellular environment without the development of novel chemical scaffolds. Although approaches using bump–hole technology are highly related to the field of allele-selective inhibitors and resistance mutations, they have been reviewed extensively elsewhere⁶¹.

JAB34 is one such allele-selective kinase inhibitor initially developed to target kinase mutants with two substitutions, one introducing a cysteine and one incorporating a hole into the kinase active site (Fig. 4b). In a seminal study, JAB34 was applied to uncover the mechanism underlying the paradoxical activation by BRAF inhibitors⁵⁶. Previously, it had been shown that, while wild-type BRAF is catalytically active only as a dimer, which is formed upon activation by Ras, BRAF(V600E) is active as a monomer, independently of Ras activation (Fig. 4a)⁶². Using allele-selective kinase inhibitors to inhibit only one of the two protomers of the BRAF dimer, which harbored a sensitizing mutation, investigators demonstrated that canonical BRAF inhibitors such as vemurafenib activate BRAF signaling through activation of one protomer of the BRAF dimer upon drug binding to the other protomer. It was determined that canonical BRAF inhibitors could only activate BRAF signaling when the BRAF protein resistant to inhibition by the allele-selective kinase inhibitor was catalytically active. When a kinase-dead allele was engineered for the non-inhibited protomer, inhibitor-dependent BRAF activation disappeared, demonstrating the presence of drug-dependent transactivation (Fig. 4c). It was later shown that, in cells with wild-type BRAF, drug binding promoted BRAF dimer formation by increasing the interaction of BRAF with GTP-bound Ras⁶³. Furthermore, within the context of these BRAF dimers, drug binding to one protomer of the BRAF dimer caused a structural shift within the second protomer of the dimer that both inhibited drug binding and locked the second protomer into an active conformation^63,64.

The relevance of BRAF dimerization in the mechanism of action of vemurafenib was confirmed by characterization of a vemurafenib resistance mutation⁶⁵. In this study, a shortened isoform of BRAF lacking exons 4–8, known as the p61 isoform, was characterized in several resistant clonal cell lines isolated after 2 months of vemurafenib exposure. It was found that the p61 BRAF isoform constitutively formed dimers, which were refractory to inhibition by vemurafenib (Fig. 4d), supporting the notion that vemurafenib inhibits monomeric BRAF signaling only in the context of BRAF(V600E)-encoding cells. Cells that encode wild-type BRAF signal through dimeric BRAF and thus are resistant to inhibition by vemurafenib. In line with this idea, tumors that activate the MAPK pathway through dimeric BRAF downstream of Ras activation are unresponsive to first-generation BRAF inhibitors such as vemurafenib⁶⁶. The resistance of BRAF dimers to vemurafenib inhibition underlies the lack of toxicity and thus clinical success of these initial BRAF inhibitors in BRAF^V600E melanoma.

Recently, the BRAF p61 isoform has been used in screens to identify novel BRAF inhibitors that selectively inhibit dimeric BRAF over monomeric BRAF^67,68. This selectivity profile is opposite to that of first-generation BRAF inhibitors such as vemurafenib that inhibit monomeric BRAF over dimeric BRAF. In a preliminary case report, it was found that combined use of a monomeric BRAF inhibitor and a next-generation BRAF inhibitor identified through this screening approach was complementary and synergistic, maintaining therapeutic efficacy while minimizing toxicity⁶⁸. In this way, the study of the BRAF p61 isoform both confirmed the mechanism of action of first-generation BRAF inhibitors such as vemurafenib and led to the identification of next-generation therapeutic candidates targeting this pathway.

Resistance mutations and allele-selective compounds have also proven influential outside the field of kinase biology. In a recent study from our laboratory, we used a CRISPR–Cas9-based mutagenesis strategy termed CRISPR-suppressor scanning to identify resistance mutations in the histone demethylase LSD1 to the the LSD1 inhibitor, GSK-LSD1, in acute myeloid leukemia (AML)⁶⁹. GSK-LSD1 and related inhibitors were designed to be activity-based covalent inhibitors of the histone demethylase activity of LSD1. Through CRISPR-suppressor scanning, we identified resistance mutations to GSK-LSD1 in the LSD1 active site that also inactivated LSD1’s histone demethylase activity. The viability of cells with these enzyme-inactive mutants suggested that LSD1’s histone demethylase activity is not essential in the context of AML and that inhibition of this activity is not the relevant mechanism of action for LSD1 inhibitors in AML. Through characterization of these resistant cell lines, we demonstrated that the relevant mechanism of action of LSD1 inhibitors in AML is disruption of a protein–protein interaction between LSD1 and the transcription factor GFI1/GFI1B (Fig. 5a). Through an in-depth characterization of small molecule resistance, we and others revised the mechanism of action of LSD1 inhibitors and identified the relevant function of LSD1 in leukemia^69,70.

Fig. 5 |. Resistance mutations uncover the mechanism of action of LSD1 inhibitors in AML.

a, Schematic summarizing the prior and revised models of LSD1 inhibition in AML. PDB: 1AOI. b, Top, schematic illustrating the mechanism of the drug-complementary GFI1B allele. Bottom, structural views of the LSD1 catalytic site labeled by GSK-LSD1 adduct bound to wild-type Snail/Gfi-1 (SNAG) peptide (left) or a SNAG(F5A) peptide (right). PDB: 2UXX, 2Y48.

In our study on LSD1, we sought to investigate the necessity of the LSD1–GFI1B interaction in AML. Inspired by bump–hole approaches, as described above, we engineered a compensatory hole in a GFI1B mutant that caused GSK-LSD1 to induce complexation of LSD1 and GFI1B, converting GSK-LSD1 from a protein–protein interaction inhibitor into a molecular glue (Fig. 5b). This mutant GFI1B rescued growth of AML cells, thereby functioning as a resistance ‘mutation’ that established GFI1B as a relevant target responsible for GSK-LSD1’s antiproliferative activity. We propose that such ‘drug-complementary alleles’ could be important tools for interrogating protein–protein interactions with precise chemical control. Furthermore, we believe that characterizing resistance mutations for protein–protein interaction-disrupting small molecules with respect to both protein partners involved could aid in determining the structure and function of these intricate interactions.

Conclusions and future perspectives

Although resistance mutations are most commonly used to identify and confirm the target of a small molecule candidate, they can also be exploited as powerful reagents to enable biological advances. Often these discoveries were unanticipated and involved identification of the small molecule target itself, unveiling an unexpected mechanism of action by which a small molecule elicited its effects or by which resistance occurred. It is clear that small molecules can have a variety of unanticipated effects on their targets beyond modulation of the target’s canonical activity, ranging from protein activation to the creation of de novo protein–protein interactions^71–74. Using the tools of chemistry and genetics to exploit these diverse mechanisms, new biological phenomena can be discovered.

Historically, resistance mutations were serendipitously discovered and subsequently characterized. However, in recent years, deliberate identification of drug-resistance alleles has been enabled by the advent of numerous targeted and highly parallel mutagenesis strategies. Advances in molecular cloning and DNA synthesis have prompted the development of techniques such as MITE-seq that allow for an exhaustive sampling of all possible point mutations through massively parallel variant overexpression (Fig. 6a)^34,75. As discussed in the preceding sections, each of these strategies has been used with success to identify resistance mutations in a variety of contexts^35,76–81. While such techniques offer the potential to survey an unprecedented number of genetic variants in a single experiment, they rely on overexpression to introduce the variant of interest, which may convolute interpretation of results owing to dosage effects. Technologies using CRISPR–Cas9 genome editing have also recently been applied to accelerate the discovery of resistance mutations in a target of interest in its biologically relevant setting (Fig. 6b). These strategies introduce variants of interest through genome editing of the endogenous locus, allowing for genetic variants to be evaluated at the target’s native expression level. Mutagenesis strategies that repair Cas9 double-stranded breaks with error-prone nonhomologous end joining (NHEJ) introduce a wide diversity of insertion and deletion mutations into the genetic loci targeted by guide RNAs, while those that use homology-directed repair (HDR) can introduce specific mutations in a massively parallel fashion. Both of these approaches have been developed and successfully applied as high-throughput mutagenesis strategies^69,82–87. Base editing strategies that directly chemically convert DNA bases through nucleobase deamination have been used to uncover missense resistance mutations in a variety of contexts^88–93. Two recent examples used cytosine base editors to profile clinical missense variants present in the ClinVar database for small molecule resistance^94,95. New innovative technologies such as prime editing will only assist in systematically introducing protein-coding mutations in endogenous systems^96,97. While base editing strategies allow for precise introduction of point mutations through a simple, scalable workflow, the mutational scope is limited by the nucleotide transitions possible with current base editor technology. Strategies such as CRISPR-suppressor scanning that rely on NHEJ to introduce genetic variation vastly increase mutational scope to include insertion and deletion mutations but do not allow for user control of the variant library and present challenges when predicting and identifying introduced mutations. Approaches that rely on HDR to introduce genetic variation as well as prime editing allow for precise user control of the variant library but present challenges when scaling to large sequences as each variant to be evaluated must be explicitly included in the library.

Fig. 6 |. Mutagenesis strategies enhance resistance mutation identification and enable novel methods to investigate the small molecule–protein interface.

a, Schematic illustrating the workflow of mutagenesis strategies that use variant overexpression approaches. Refs.^28,76–78 describe the use of error-prone cloning methods; refs.^{34,35,75,79–81} describe the use of massively parallel synthesis methods. CDS, coding sequence. b, Schematic illustrating the workflow of mutagenesis strategies that use CRISPR-based technologies. Refs.^69,82–84 describe the use of CRISPR-NHEJ approaches; refs.^85–87,97 describe the use of CRISPR-HDR approaches and prime editing; and refs.^94,95 describe the use of CRISPR base editing. c, Schematic illustrating the potential application of a resistance mutation profile to functionally classify small molecule analogs in a high-throughput manner. d, Schematic detailing the use of RADD to inform small molecule design. SAR, structure–activity relationship.

As new genome editing and mutagenesis technologies greatly increase the speed, throughput and ease of resistance mutation discovery, comprehensive identification of all target-driven resistance mechanisms may become routine. Development of computational methods to analyze large datasets of resistance mutations will be critical for realizing the full potential of these advances. This compendium of resistance mutations could be used to classify small molecule mechanism of action and/or target function, potentially providing an important fingerprint for the discovery of new molecules exhibiting similar mechanisms and even new mechanisms yet to be uncovered. We hypothesize that resistance mutation profiles identified by high-throughput mutagenesis could ‘fingerprint’ different small molecule analogs and point to important mechanistic differences that may relate not only to therapeutic efficacy but also to new biology (Fig. 6c). Supporting this idea, recent studies have found that inhibitors with different mechanisms of action have distinct landscapes of resistance mutations across a given protein target^33,79. We believe that this type of protein-wide structure–activity relationship could be complementary to and leveraged alongside classical small molecule structure–activity relationship profiling, allowing combinatorial exploration of protein–small molecule interactions from both sides of the molecular interface. Along these lines, resistance analysis during design (RADD) is a recently described approach that uses resistance mutations to guide the development of selective chemical probes and predict compound-binding modes in the absence of structural information (Fig. 6d)^98,99. Additionally, in our studies on LSD1, we found that different small molecule analogs with subtle structural perturbations can substantially enrich for different resistance mutations⁶⁹. High-throughput mutagenesis strategies could potentially inform the drug discovery process and catalyze biological discovery by bridging protein structure–function information with medicinal chemistry lead optimization efforts.

Moreover, the acceleration in resistance mutation discovery enabled by deep mutational scanning has the potential to advance biology in many areas. By exhaustively characterizing the resistance mutations of natural products and molecules identified by phenotypic screening, we may uncover additional hidden pathways and cellular processes that both yield new biological insights and provide new therapeutic opportunities. The use of resistance mutations to identify small molecule mechanism of action may be an important tool for the discovery of molecules with novel mechanisms such as chemically induced proximity as well as new mechanisms yet to be uncovered. As more molecules with these mechanisms are discovered, it may even be possible to identify design rules governing these emerging mechanistic classes. By characterizing unexpected resistance mutations, especially those outside the small molecule-binding site, novel mechanisms of protein function and allostery may also be discovered. Pairing of resistance mutations found by saturation mutagenesis with structural data from X-ray crystallography and cryo-electron microscopy may suggest the relevant protein structures and/or dynamics adopted in a particular cellular context. Additionally, resistance mutations found in unstructured or disordered regions, especially when paired with methods such as cross-linking mass spectrometry or hydrogen deuterium exchange mass spectrometry, may hint at the dynamics of these structurally elusive regions in a cellular context. Resistance mutations to drugs commonly used in the clinic applied across a diverse set of tissues in relevant model systems could aid in the detailed characterization of physiology and pathology. These studies may point toward common mechanisms underlying the diversity in therapeutic efficacy and side effect profiles across different individuals and species. Finally, resistance mutations used creatively through approaches such as allele-selective kinase inhibitors and drug-complementary alleles may aid in characterizing intricate biological interactions, particularly those occurring within large protein complexes, that have proven difficult to define using traditional techniques.

Expanded mutagenesis technologies have caused a paradigm shift in molecular biology and genetics. By understanding the important role resistance mutations have had in advancing biology, chemical biologists can creatively apply these mutagenesis technologies to identify novel resistance mutations and accelerate the discovery of new biology and medicines.