Allosteric regulation and inhibition of protein kinases

Abstract

The human genome encodes more than 500 different protein kinases: signaling enzymes with tightly regulated activity. Enzymatic activity within the conserved kinase domain is influenced by numerous regulatory inputs including the binding of regulatory domains, substrates, and the effect of posttranslational modifications such as autophosphorylation. Integration of these diverse inputs occurs via allosteric sites that relate signals via networks of amino acid residues to the active site and ensures controlled phosphorylation of kinase substrates. Here, we review mechanisms of allosteric regulation of protein kinases and recent advances in the field.

Introduction

Eukaryotic protein kinases regulate a wide array of cellular processes through the catalytic transfer of a phosphate group from adenosine triphosphate (ATP) onto Ser/Thr or Tyr residues of a substrate protein ¹. The timing, localization, and duration of kinase activity must be tightly regulated and responsive to signaling inputs. Protein kinases are allosterically regulated enzymes: a perturbation at one site of the protein (allosteric site) is transmitted and functionally linked to the active site (orthosteric site) through networks of coupled amino acids, or “spines.” The orthosteric ATP-binding site is highly conserved while allosteric sites are less conserved among kinases, reflecting variety in mechanisms of enzymatic regulation ². Allosteric site perturbation can occur in several ways, including regulatory domain interactions, protein dimerization, protein-protein interactions, post-translational modifications, and ligand binding. Amino acid spines relay such perturbations to the orthosteric site where they produce subtle changes in enzymatic action, affinity for kinase substrates, or even larger domain movements that dynamically alter the population of kinases in the active or inactive states. The diversity of allosteric regulatory mechanisms and the diverse outputs facilitate the specific regulation of protein kinases.

Allosteric signal integration in kinases

The protein kinase catalytic domain is composed of approximately 250 amino acid residues that form a bilobal structure, containing a small amino-terminal N-lobe and larger carboxy-terminal C- lobe. A deep cleft situated between the two lobes forms the conserved ATP-binding active site. A glycine-rich phosphate binding loop and the activation loop form two sides of the active site and mediate ATP and substrate peptide binding. Regulatory elements in the kinase domain participate in conserved intramolecular networks that span both the N-lobe and C-lobe, which include the regulatory spine (R-spine) and catalytic spine (C-spine) ^{3, 4}. The conserved R-spine adopts an assembled conformation in virtually all active kinase structures and can disassemble in different ways, leading to kinase inactivation. Additionally, a dynamically-coupled spine (DC spine) has been observed in some kinases to mediate negative cooperativity between ATP and substrate peptide binding sites ⁵. These spines demonstrate how larger networks of interacting residues relay regulatory inputs (e.g., binding of regulatory domains and allosteric modulators, phosphorylation events, etc.) to the catalytic machinery.

Allosteric regulation via interdomain communication

Intramolecular interactions often autoregulate protein kinase activity. Roughly half of all human non-receptor tyrosine kinases share a common global domain architecture: an N-terminal SH3 (Src-homology) domain, followed by an SH2 domain and the kinase domain (also referred to as the SH1 domain). Evolutionary variants of this so called “Src module” contain additional binding domains, incorporate regulatory post-translational modification sites, and altered conformational preferences ⁶. The autoinhibitory mechanism of the SH3/SH2 domain interaction with the kinase domain is well understood and has been described in detail for Src family kinases. In brief, binding of the SH2 domain to a phosphorylated C-terminal tyrosine clamps the SH2 domain down onto the kinase domain, while the SH3 domain binds a poly-proline sequence in the linker region between the SH2 and kinase domains. The autoinhibitory conformation adopted by the SH3/SH2 domains locks the kinase domain in an inactive conformation.

Variations on the theme of autoinhibition mediated by the SH3/SH2 domains have been observed in several other kinases. Abl and Tec families lack the C-terminal phosphorylation site that engages the SH2 domain during autoinhibition, and instead an N-terminal myristoyl group binds to the C-lobe of the kinase domain which leads to SH2 domain binding ⁷. In the Tec family kinase, BTK, a negatively charged sidechain (D656) provides the SH2 binding site ⁸. The SH3 domain in ACK1 kinase binds to the N-terminal lobe of the kinase domain and inhibits its activity ⁹. Similarly, the Brk SH3 domain autoinhibits kinase activity and proline-rich peptides enhanced the kinase activity ¹⁰. Even the serine/threonine kinase, MLK3, has evolved a similar autoinhibitory mechanism ¹¹.

Apart from autoinhibition, the SH2 domains in Src, Abl and BTK can also activate the kinase domain by binding to the N-lobe of the kinase domain. Mutations or ligands that disrupt these activating interactions between the SH2 domain and the N-lobe can deactivate the kinase ^12–14.

Allosteric modulation of kinase activity via the Pseudokinase domain

The Janus kinase (JAK) family (JAK1/2/3 and TYK2) are non-receptor tyrosine kinases (NRTKs) that contain an N-terminal FERM domain, an SH2-like domain, a catalytically-dead pseudokinase domain (JAK homology 2, JH2), and a C-terminal tyrosine kinase domain (JAK homology 1, JH1). While mechanisms of autoregulation of the primary catalytic JH1 domain can vary between members of this kinase family, the JH2 domain of JAK2 has been shown to negatively regulate JH1 activity via autophosphorylation on S523 and Y570 ¹⁵. Additionally, the JH2 domain of JAK2 and TYK2 can play an allosteric role by stabilizing an inactive conformation of the JH1 domain ^{16, 17}. A crystal structure of TYK2 JH2-JH1 and computational work on JAK2 revealed an extensive interaction interface between the JH2 domain N-lobe and the JH1 domain ¹⁸. This allosteric regulatory role of the JH2 domain explains how observed mutations in the JH2-JH1 interface destabilize autoinhibitory interactions ¹⁹.

Allosteric modulation of kinase activity via the ubiquitin-associated domain

LKB1, a tumor suppressor Ser/Thr protein kinase, phosphorylates and activates a group of protein kinases closely related to AMPK (AMP-activated protein kinase)²⁰. In ten of the 14 AMPK-related (AMPKR) kinases in this family that LBK1 activates, a ubiquitin-associated (UBA) domain is located C-terminal to the kinase domain and consists of a unique three-helix bundle not found in other human kinases ²⁰. LKB1-mediated phosphorylation of the UBA domain and UBA-kinase domain interactions may be essential for kinase activity through autoregulatory interactions with the N-lobe. Structural studies of MARK1, an AMPKR family member, revealed that the UBA domain of MARK1 binds to helix αC in the N-lobe, thereby locking the kinase domain in the inactive conformation ²¹. AMPK itself contains a C-terminal autoinhibitory domain (AID) that shares structural homology with the UBA domain. The structure of unphosphorylated S. pombe AMPK revealed extensive interactions between AMPK AID, helices αC and αE, and the hinge region, which restrains movement of the αC helix, favoring the inactive kinase conformation ²².

Allosteric regulation triggered by protein-protein interactions

The formation of dimers and higher-order oligomers is a feature of receptor tyrosine kinases, a class of transmembrane signaling kinases activated by the binding of extracellular ligands (reviewed in ²³). Receptor homodimers can exist prior to ligand binding (e.g., KIT or FGFRs ²⁴), or they can assemble as a result of bivalent ligand binding (e.g.., EGFR). Two monomers of EGFR are activated through binding epidermal growth factor (EGF), which changes the relative conformations of their extracellular ligand-binding domains. This binding event rearranges the intracellular kinase domains of each monomer as a head-to tail asymmetric dimer, where the C-lobe of one kinase domain stabilizes the N-lobe of the second kinase domain in an active conformation ^{25, 26}. EGFR serves as a model for studying these two major modes of allosteric control: transduction of extracellular signals to intracellular domains and kinase-dimer dynamics that promote activation of one of the kinase domains.

Heterodimerization with a non-kinase protein is exemplified by cyclin-dependent kinases (CDKs), a class of Ser/Thr kinases that regulate eukaryotic cell-cycle progression. Monomeric Cdk2 adopts an autoinhibitory conformation where helix-αC is positioned outwards (αC -out) and the activation loop is positioned into the active site (A-loop in) ²⁷. The autoinhibitory conformation breaks the R-spine and thereby prevents the formation of a salt bridge between the catalytic lysine and a glutamate in helix-αC critical to ATP coordination. Phosphorylation of Thr160 in the activation loop of Cdk2 by a CDK-activating kinase complex is followed by the binding of a cognate cyclin subunit ²⁸. Cyclin binding to Cdk2 induces conformational transitions to the αC-in, A-loop-out conformation, facilitating R-spine restoration and Glu-Lys salt bridge formation to allow for ATP-coordination and catalytic activity ²⁹. Additionally, Cdk2 phosphorylation at T160 increases Cdk2-cyclin binding, Cdk2-cyclin allosteric coupling, and subsequent sampling of the Cdk2 A-loop out, αC-in active conformation ³⁰.

Allosteric inhibitors

Allosteric inhibitors that target protein kinases are of considerable interest for both their potential as therapeutics in human disease and utility as research probes for interrogating mechanisms of kinase regulation. The structural diversity underpinning allosteric regulatory mechanisms also offers opportunities for selective kinase inhibition and overcoming resistance to orthosteric inhibitors. Some orthosteric inhibitor resistance mutations may allosterically promote unfavorable kinase conformations, such as the active state for type II inhibitors that preferentially bind the inactive conformation ³¹.

Allosteric inhibitors can be subdivided into two main subclasses: inhibitors that bind adjacent to the ATP-site but do not overlap with it (type III) and inhibitors that bind an allosteric site distal to the ATP-site (type-IV) ³². Although type-III and type-IV inhibitors present potential advantages, they remain far less common than other classes of inhibitors and only a small number have been approved by the FDA (Table 1). The lack of allosteric inhibitors is largely due to the challenges associated with 1) identifying allosteric regulatory sites and 2) determining their mechanisms of action. Fortunately, recent advances in structural techniques, including X-ray crystallography, cryo-EM, and NMR have driven the validation of allosteric sites and advanced the structural views on allosteric modulation ³³. In addition, advances in the computational prediction of allosteric sites have been complemented by experimental methods as well ³⁴. Given the diversity of allosteric regulatory mechanisms, allosteric modulators can alter kinase dimerization, interdomain interactions, and even influence binding of orthosteric ligands.

Table 1. Selected Allosteric Kinase Inhibitors.

Clinical trials information was obtained by using search terms associated with the compound names at clinicaltrials.gov, and FDA approval status was identified by using associated search terms at www.accessdata.fda.gov/scripts/cder/daf/index.cfm. References for PDBs are as follows: 5D41 ³⁷, 6DUK ³⁸, 5MO4 ⁴¹, 3K5V ⁴⁰, 3PXF ⁴⁵, 6NZP ⁶⁰, 7JUX ⁶¹, 5KCV ⁵², 4RQK ⁶², 4U6R ⁶³, 7QIE ⁶⁴.

Kinase	Compound	Strategy	Pocket	Inhibitor MOA	PDB	Selected Clinical Trials	FDA Approval / Indications
EGFR		Stabilizes the inactive kinase domain conformation (both)	Binds pocket adjacent to N-lobe αC helix (both)	Prevents the kinase from adopting the active conformation, inhibition relieved by asymmetric EGFR dimer formation (both)	5D41 (EAI045) 6DUK (JBJ-04-125-02)	None (both)	None (both)
BCR-Abl		Stabilizes the inactive conformational state (both)	Myristate binding site (both)	Causes bending of aI-helix and allows SH3/SH2 domains to bind, restrict KD movement, and adopt the autoinhibited conformational state (both)	5MO4 (Asciminib) 3K5V (GNF-2 )	NCT02081378 (asciminib) NCT03106779 (asciminib)	Philadelphia chromosome-positive chronic myeloid leukemia (Ph+ CML) in chronic phase (CP) with the T315I BCR-Abl mutation, or Ph+ CML in CP previously treated with two or more TKI (asciminib)
Src		Interacts with allosteric regulatory spine	Novel G-loop site	Type V non-competitive inhibitor	None	None	None
Cdk2		Prevents formation of active complexes	Multiple sites	Displaces helix- αC and prevents activation by cyclin A	3PXF	None	None
TYK2		Leverage JH2-JH1 allosteric crosstalk to inactivate TYK2	Alanine pocket of TYK2 JH2	Binds and stabilizes JH2 domain to prevent TYK2 signaling	6NZP	NCT03934216 NCT04536961	Adults with moderate-to-severe plaque psoriasis who are candidates for systemic therapy or phototherapy
MEK1/2		Bridges MEK-KSR interaction interface	ATP-adjacent binding pocket and interface formed by KSR-binding	Facilitates regulatory interaction with KSR	7JUX	NCT03232892 NCT03428126 NCT01941927	Patients with unresectable or metastatic melanoma with BRAF V600E or V600K mutations as detected by an FDA-approved test
Akt		Achieves selectivity: binding requires pleckstrin homology domain (both)	Pocket between the kinase and pleckstrin homology domains (both)	Stabilizes the autoinhibited conformation (miransertib)	5KCV (miransertib)	NCT01473095 (miransertib) NCT01186705 (MK-2206) NCT01169649 (MK-2206) NCT01283035 (MK-2206)	None (both)
PDK1		Target conserved PDK1-interacting fragment (PIF) site	Binds in the N-lobe PIF pocket	Enhances the activity of ATP-competitive orthosteric inhibitors	4RQK	None	None
IRE1α		Stabilizes inactive-like conformation	Binds in an allosteric pocket formed following αC-helix shift	Promotes αC-helix shift away from the active state, and a DFG-out conformation in the activation loop	4U6R	None	None
PI5P4Kγ		Stabilizes an inactive conformation	Occupies a putative lipid-binding allosteric pocket	Stabilizes the activation loop in a conformation that occludes the ATP-binding site	7QIE	None	None

Allosteric inhibitors targeted at EGFR

On-target resistance to orthosteric EGFR inhibitors is common in non-small cell lung cancers in which 10–30% of patients harbor EGFR mutations ³⁵. For example, the T790M mutation increases the affinity of EGFR for ATP, outcompeting the reversible orthosteric inhibitors erlotinib and gefitinib. Resistance has also been shown to arise against EGFR mutant-selective inhibitors. For example, EGFR-C797S, the location of the required cysteine for irreversible inhibitor binding, promotes resistance against the orthosteric inhibitor osimertinib ³⁶. These resistance mechanisms in the orthosteric site reveal the need for alternate strategies of EGFR inhibition. EAI045 is a selective EGFR allosteric inhibitor that circumvents the on-target C797S mutation ³⁷. EAI045 binds EGFR at a site adjacent to the ATP-binding pocket made available following the movement of αC helix into an inactive conformation. However, due to the asymmetric dimer in active EGFR, EAI045 was found to be ineffective as a single agent; the allosteric promotion of the active conformation of the receiver kinase domain mediated by the C-lobe of the activator kinase domain prevented formation of the EAI045 binding pocket through stabilization of the αC -helix in the active conformation. However, when EAI045 was combined with cetuximab, an anti-EGFR antibody that disrupts dimer formation, its efficacy was markedly increased ³⁷. Further work led to the development of JBJ-04–125-02, an EGFR-L858R/T790M selective allosteric inhibitor that binds to a pocket similarly formed by the displacement of the αC -helix adopted in the inactive EGFR conformation ³⁸. JBJ-04-125-02 also potently inhibited growth of a Ba/F3 cell model of L858R/T790M/C797S EGFR. Due to its binding mode, EGFR dimerization also reduced JBJ-04-125-02 efficacy, although co-treatment with the orthosteric inhibitor osimertinib was found to be more effective than either compound alone. EGFR allosteric inhibitors represent a promising approach for combination therapy to circumvent common on-target mechanisms of resistance to orthosteric inhibitors.

Allosteric inhibitors targeted at BCR-Abl

Allosteric ligands have been found to help overcome resistance to ATP-competitive inhibitors (e.g., imatinib, nilotinib, dasatinib, bosutinib etc.) of BCR-Abl, a fusion protein that drives the majority of malignancies in chronic myelogenous leukemia ³⁹. Early allosteric type-IV BCR-Abl inhibitors GNF-2 and GNF-5 bind to the c-Abl myristate binding site located at the C-terminus of the Abl kinase domain, which leads to autoinhibition of the kinase and reduced peptide substrate binding ⁴⁰. A more recently developed inhibitor, Asciminib (ABL001), potently binds to the myristate binding site of the Abl kinase domain with a Kd of 0.5–0.8 nM and exhibits a similar binding mode to GNF-2 ^{41, 42}. Notably, Asciminib monotreatment can inhibit all orthosteric site mutants of BCR-Abl with nanomolar potency, including the T315 mutants that markedly decrease the affinity of orthosteric inhibitors. These characteristics allowed asciminib to achieve the distinction of the first type-IV allosteric BCR-Abl inhibitor to enter clinical trials and achieve FDA-approval⁴². Although resistance profiles for asciminib have been observed, these mutations are not found to overlap with resistance mutations to orthosteric inhibitors such as nilotinib⁴¹. Furthermore, combination therapy with allosteric and orthosteric inhibitors appears to reduce the probability of resistance emerging ⁴³.

Allosteric inhibitors targeted at Cdk2

Cdk2 is an important therapeutic target because it controls cell cycle checkpoints and can also facilitate resistance to Cdk4/6 inhibitors ⁴⁴. The development of selective Cdk2 orthosteric inhibitors has been severely challenged by the structural similarity observed among Cdks. The most well-characterized allosteric binding pocket in Cdk2 binds to the commercial dye 8-anilino-1-napthalene sulfonic acid (ANS). A crystal structure of Cdk2 in complex with ANS revealed two ANS molecules binding adjacent to the ATP-binding site and perturbing the conformation of helix-αC such that it disrupts the binding interface necessary for cyclin A binding ⁴⁵. Allosteric mechanisms may also be concentration-dependent; for example, at higher ANS concentrations it is possible for a third ANS molecule to bind directly to the ATP binding site ⁴⁵. ANS binds to Cdk2 with a K_d of 37 μM and is readily ejected in the presence of cyclin A ⁴⁵. Structural analysis supports this observation and suggests that when Cdk2 is bound to cyclin A, the ANS binding pocket is relatively inaccessible ⁴⁵. However, a synergistic mechanism with orthosteric inhibitors may also exist, as exemplified by certain inhibitors that can enhance ANS binding affinity 5- to 10-fold ⁴⁶. These observations render the Cdk2 ANS allosteric pocket a potential site for lead compound identification and future pharmacological intervention ⁴⁷.

Allosteric inhibitors targeted at Src

Allosteric sites can be challenging to predict and identify, as they are typically associated with high-energy conformations that require ligand-induced stabilization of the binding pocket. Recent work using long unbiased molecular dynamics simulations revealed that orthosteric Src inhibitors may transiently bind to lower affinity binding sites before binding to the high-affinity orthosteric site ⁴⁸. These distal sites are occluded in crystal structures of Src, and ligands appear to induce the transient opening of these pockets. Two of the observed sites, the PIF and Myr binding pockets, had been previously characterized as allosteric sites. However, the third site in the C-lobe, the G-loop site, was novel. Virtual screening of compounds that bind to the G-loop binding pocket resulted in the identification of compound 1C, a weak non-substrate competitive inhibitor of Src kinase ⁴⁸. This work illustrated one strategy to identify cryptic allosteric sites through interrogation of alternate lower-affinity sites of other compounds.

Allosteric inhibitors targeted at Akt

The Ser/Thr Akt kinases are a major therapeutic target in several cancers due to their promotion of cell survival through the apoptotic pathway ⁴⁹. Notably, some orthosteric inhibitors have been observed to induce hyperphosphorylation of Thr308 and Ser473 in the absence of pathway feedback effects, suggesting that allosteric inhibitors may be a more attractive method to inhibit Akt ⁵⁰. Several allosteric Akt inhibitors have been developed, including Akt1/Akt2 isoform selective compounds, the relatively selective compound 21a (ARQ 092/miransertib) which has been studied in a limited set of clinical trials, and MK-2206 which is currently being studied in phase II clinical trials (Table 1) ^51–53. Notably, since Akt mediates crosstalk between numerous cellular pathways, the upregulation of other signaling effectors can lead to Akt inhibitor resistance necessitating the implementation of combination therapies in clinical practice.

Allosteric inhibitors targeted at the JAK Family

JAK inhibitors have emerged as major therapeutic targets for treating autoimmune diseases, including rheumatoid arthritis and ulcerative colitis ^{54, 55}. The structural conservation of the orthosteric site across JAK1, JAK2, JAK3, and TYK2 has complicated the discovery of isoform-specific inhibitors that target the active site of the catalytic JH1 domain ⁵⁶. For example, although tofacitinib has moderate selectivity for both JAK1 and JAK3, it is still active against JAK2 and TYK2 ⁵⁷. However, isoform-selective inhibition is possible, as exemplified by fedratinib which is fairly selective for JAK2 ⁵⁷. Although JAK inhibitors have been shown to be effective in treating autoimmune diseases, toxicity is observed at therapeutic doses of some JAK inhibitors which hampers clinical efficacy, leaving an unmet need for improved treatments ⁵⁸. Selective TYK2 inhibition has emerged as a potentially promising strategy for treating a number of autoimmune diseases, including both Crohn’s disease and ulcerative colitis ⁵⁵. Allosteric inhibitors that target the JH2 pseudo-kinase domain are highly selective and inhibit TYK2 by binding and stabilizing the JH2 domain, whereby the JH2 domain restricts the JH1 domain in an inactive conformation ^16–18. Hit-to-lead efforts against TYK2 produced a series of inhibitors that showed in vivo efficacy against CD40 agonist-induced colitis in a murine model of inflammatory bowel disease ⁵⁹. Inhibitor optimization resulted in the discovery of BMS-986165, deucravacitinib, which is currently in two phase II clinical trials for ulcerative colitis and Crohn’s disease and is FDA-approved for plaque psoriasis (Table 1) ⁵⁵. The X-ray co-crystal structure of TYK2 bound to deucravacitinib showed that the inhibitor binds to the hinge region, extended hinge region, and an “alanine pocket” of the JH2 domain that confers high TYK2 JH2 selectivity ^{59, 60}. Although the full mechanism of JH2-mediated allosteric regulation over the TYK2 JH1 domain has yet to be determined, increased TYK2 selectivity has the potential to improve side effects for patients over pan-JAK inhibitors.

Perspectives

Protein kinases are allosterically regulated through interdomain contacts, formation of kinase dimers, and interactions with other partner proteins. Some inhibitors can induce structural changes that promote dimerization and, in some cases, activate kinase dimerization partners ⁶⁵. The interplay between small molecule inhibition and functional higher-order arrangements of kinase dimers and other partner proteins is of great interest for understanding the biology of kinases and their unique allosteric regulatory mechanisms.

Unlike orthosteric inhibitors that can preferentially bind the ATP-binding pocket in the active or inactive kinase conformation, allosteric inhibitors are subject to- and can influence a wider range of conformational dynamics through modulation of alternate binding sites. Synergistic kinase inhibition is possible with both orthosteric inhibitors and antibodies that facilitate the conformational demands of inhibitor binding and isolate kinase domains from other allosteric regulators ^{37, 46}. Allosteric ligands can also be integrated into proteolysis-targeting chimera (PROTAC) molecules that promote ubiquitination and degradation of target proteins, providing an additional mechanism of synergy between allosteric and orthosteric inhibitors ⁶⁶.

Discovery of allosteric small molecule inhibitors remains a challenge, but techniques including x-ray crystallographic screening of compounds or fragments, molecular dynamics simulations, thiol-containing inhibitor disulfide tethering, and DNA-encoded libraries (DEL) present opportunities for screening large compound libraries ^67–71. The addition of an orthosteric inhibitor into DEL reactions mixes has been suggested as a strategy to preferentially identify non-orthosteric binders in this assay format ⁷². Adding an orthosteric inhibitor to x-ray crystallographic or DEL screens could also identify fragments that may offer synergistic effects, since they would preferentially bind to the same orthosteric inhibitor-bound kinase conformation. Further work in this area will build on our understanding of allosteric regulation and allosteric inhibition strategies, including combinatorial approaches with orthosteric inhibitors or antibodies, in order to mitigate inhibitor resistance and provide alternate therapeutic approaches.

Figure 1. The Regulatory, Catalytic, and Dynamically Coupled Spines in Src Kinase.

Regulatory spine (R-spine) (cyan), catalytic spine (C-spine) (green), dynamically coupled spine (DC spine) (blue and cyan). R-spine residues are also shared with the DC spine. PDB 3G5D.

Figure 2. Binding Modes of Selected Allosteric Kinase Inhibitors.

The allosteric inhibitor (blue) that binds each kinase domain (grey) are indicated. Only kinase domain residues and the inhibitors are shown for simplicity, and all other non-protein atoms, including ions, waters, and other ligands, are omitted. Chain A was selected for visualization from crystal structures with multiple copies of the kinase in the asymmetric unit.