Fuzziness in enzymatic catalysis

Abstract

Intrinsically disordered proteins/regions (IDPs/IDRs) frequently engage in dynamic charge:charge interactions, commonly referred to as ‘fuzzy’ interactions. These fuzzy interactions play critical roles in enzymatic regulation and substrate recruitment, especially for protein kinases and protein phosphatases. Here, we review recent advances that demonstrate how inter- and intramolecular fuzzy interactions among kinases and phosphatases and their cognate regulators and substrates allow for enzyme assembly, activation and substrate recruitment. We also highlight a unique mechanism of protein inhibition, where a protein phosphatase is inhibited by dynamic fuzzy interactions with its active site metals.

Introduction

Interactions between biomolecules, especially proteins, are essential for cellular communication. Historically, these interactions were thought to be exclusive to structured proteins, allowing the formation of complexes. However, over last 25 years, the essential roles of intrinsically disordered proteins/regions (IDPs/IDRs) in macromolecular assembly, specificity determination, and functional regulation have been recognized [1].

IDPs/IDRs are typically enriched with polar and charged amino acids and therefore largely lacking the hydrophobic and bulky amino acids characteristic of folded proteins. Their conformational plasticity enables them to form sequence-driven, multivalent protein:protein interactions, resulting in functionalities not accessible to folded proteins [2,3]. These interactions often involve dynamic charge:charge contacts that can only be described in the context of multi-state ensembles. Collectively, they are referred to as “fuzzy” interactions [4]. IDPs engaged in fuzzy interactions with their cognate, typically folded binding partners remain locally flexible, exhibiting varying degrees of disorder (Fig. 1a). Critically, these interactions can be as strong as rigid body interactions, since the entropic differences between the bound and unbound states are much more favorable for the overall Gibbs free energy change, allowing for effective energy compensation [5]. Moreover, the overall interaction affinity can be enhanced by avidity due to the synergistic effects of multiple fuzzy interactions within a single complex [6].

Figure 1: Fuzzy interactions regulate enzymatic functions.

a) Classification of enzyme-regulating fuzzy interactions by IDRs. Schematic models of IDRs engaged in intermolecular, intramolecular or context–dependent bifunctional interactions. IDR ensembles exhibit a heterogenous array of disorder to achieve regulation of enzymatic functions. The examples include static (multiple distinct conformations) and dynamic (fluctuating conformational ensembles) disorder, where the latter can be classified into clamp (disordered region connecting ordered regions), flanking (ordered region flanked by disordered regions) and random (retained disorder) models. b-c) The catalytic activity of kinases and phosphatases is regulated by fuzzy interactions. Examples of such interactions discovered using solution NMR in combination with other biophysical approaches are discussed in the context of specific phosphoprotein phosphatases (PPPs), protein tyrosine phosphatases (PTPs) and kinases.

One of the problems in characterizing fuzzy interactions is the lack of atomic-resolution techniques that can provide insights into the conformational sampling of IDPs/IDRs and the complexes they form with other proteins (Fig. 1b). The canonical techniques used to study protein:protein interactions at atomic resolution, X-ray crystallography and cryo-EM, are unable to detect fuzzy interactions as their dynamic nature does not allow for diffraction to occur (crystallography) or enough identical ensembles to be confidently identified (cryo-EM). Moreover, AlphaFold and other protein structure prediction programs predict only single conformations and thus do not accurately describe ensemble interactions, nor are there enough ensemble structures available to allow for machine learning to be performed. However, computational advances, including the development of sequence-based platforms utilizing coarse-grained force fields to predict intermolecular IDR interactions, are ongoing [7]. Nevertheless, nuclear magnetic resonance (NMR) spectroscopy is still the sole technique for obtaining atomic-resolution molecular descriptions of fuzzy interactions. NMR studies are often complemented by additional molecular techniques, such as single molecule fluorescence as well as isothermal titration calorimetry (ITC). Here we summarize recent reports describing fuzzy interactions to uncover their essential roles in mediating function, with a focus on the catalytic regulation of kinases and phosphatases (Fig. 1c).

Ser/Thr Phosphoprotein Phosphatases (PPP)

Protein Phosphatase 1 is inhibited via a fuzzy metal interaction

The vast majority of Protein Phosphatase 1 (PP1) interacting proteins—regulators, substrates and inhibitors—are IDPs that bind PP1 via a variety of short linear motifs (SLiMs; short, conserved sequences in IDPs [typically 3–10 amino acids] that mediate protein-protein interactions; Fig. 2a) [8]. These interactions typically result in the rigidification of the SLiM in the PP1-bound state, resulting in a tight interaction. PP1 associates with an estimated >200 IDP regulators to form distinct, specific PP1 holoenzymes that define subcellular localization, substrate recruitment (if the regulator itself is not a substrate) and site-specific dephosphorylation [9,10].

Figure 2: Ser/Thr Phosphoprotein phosphatase regulation by fuzzy interactions.

a) PP1 recruits its regulatory proteins via different SLiM interactions to create functional holoenzymes. b) PP1:I3 holoenzyme (PDBID: 8U5G), I3 (purple) binds at the PP1 RVxF and SILK motif binding pockets (left). I3 also leverages electrostatic interactions to bind to the PP1 acidic substrate binding groove and uses the fuzzy CCC motif to interact with the PP1 active site metals for tight binding (right). c) Schematic representation of the PP2A:B56 holoenzyme, highlighting the LxxIxE SLiM and acidic groove binding sites on B56. d) Crystal structure of B56 (electrostatic surface; red, negative charge; blue, positive charge) in complex with pBUBR1 (teal; PDBID: 5SWF), bound at the canonical LxxIxE motif binding pocket and a conserved B56 acidic binding groove (fuzzy interaction). e) Schematic representation of the PP2A:B55-pARRP19 inhibitor/substrate complex. f) Cryo-EM structures of PP2A:B55 bound to pARPP19 (orange, PDBID: 8TTB) and FAM122A (magenta, PDBID: 8SO0). ARPP19 α2 engages dynamically to B55 allowing FAM122A to simultaneously interact with B55. g) Schematic representation of CN:NHE1 SLiM and fuzzy interactions. h) Crystal structure of the CN:NHE1 complex (PDBID: 6NUC). NHE1 (blue) binds to the PxIxIT and LxVP binding pockets on CN. The 27-residue linker connecting both SLiMs (highlighted as dashed blue lines), engages the CN surface dynamically via electrostatic interactions at acidic patches, restricting active site accessibility.

PP1 is also inhibited by specific protein inhibitors, namely Inhibitor-1 (I1), Inhibitor-2 (I2) and Inhibitor-3 (I3) [11–13]. These inhibitors are IDPs that inhibit PP1 by occupying the PP1 active site as well as by blocking the substrate recruitment sites (Fig. 2a). Although the binding and inhibition mechanism of I2 is well-established [12], only recently has the interaction of I3 been molecularly described [11]. I3 binds PP1 via an RVxF motif, which is shared by nearly all PP1 regulators and inhibitors, and a SILK motif (Fig. 2b). However, the I3-mediated inhibition of PP1 is distinct from I2. First, multiple basic I3 residues bind the PP1 acidic substrate binding groove, an interaction that provides a blueprint for how substrates bind PP1 for dephosphorylation. Second, this interaction positions the I3 CCC (cys-cys-cys) motif over the PP1 active site (Fig. 2b). NMR combined with inhibition assays revealed that the I3 CCC motif binds and inhibits PP1 in an unexpected fuzzy manner via transient engagement with the PP1 active site metals. Furthermore, mutagenesis studies showed that inhibition was dependent solely on the availability of one cysteine, independent of its position. This highlights how fuzzy interactions guide the dynamic positioning of the inhibitory tri-cys element over PP1 active site, allowing for potent inhibition.

PP2A:B55 and PP2A:B56 – fuzzy interactions direct regulator recruitment

The recruitment of substrates by protein phosphatase 2A (PP2A) is not fully understood, limiting our understanding of PP2A-regulated signaling. PP2A is a trimeric holoenzyme comprised of a catalytic subunit, PP2Ac, a scaffolding subunit, PP2Aa, and one of several regulatory subunits, commonly referred to as B subunits (B55, B56, PR72, PR93). The first PP2A consensus binding motif identified was LxxIxE, which binds PP2A:B56 holoenzymes [14]. However, most validated LxxIxE motifs bind PP2A:B56 with micromolar affinities, highlighting the possibility that additional motifs and/or mechanisms exist to modulate PP2A:B56 binding (Fig. 2c). For example, LxxIxE phosphorylation (LpSPIxE) enhances binding [15]. Additionally, a subset of PP2A:B56 interactors strengthen their interaction with B56 using a dynamic fuzzy interaction, in which basic residues N-terminal to the LxxIxE motif engage with a conserved, negatively charged B56 groove that is immediately adjacent to the LxxIxE binding pocket (Fig. 2d) [16].

Recently, multiple structures of the PP2A:B55 holoenzyme bound to protein inhibitors or regulators/substrates were reported [17–19]. Different from PP2A:B56, PP2A:B55 does not bind linear SLiM sequences to achieve substrate/regulator recruitment but instead relies on α-helical interaction motifs (Fig. 2e). In addition, NMR data showed that the interactions of at least a subset of regulators/inhibitors also leverage fuzzy interactions to direct PP2A:B55 activity. For example, regulators (e.g. FAM122A or p107) that bind to a common PP2A:B55 binding site; known as the B55 platform, can bind in the presence of the protein inhibitor ARPP19 (Fig. 2f) [17]. Although ARPP19 releases the B55 platform in the presence of these regulators, it stays bound to B55 via a dynamic charge:charge interaction mediated by ARPP19 α-helix 2 (although this interaction is only weakly observed in the cryo-EM structure, the interaction of α-helix 2 with B55 both in the absence and presence of B55 platform binding regulators is readily detected by NMR spectroscopy). These examples show how fuzzy interactions regulate inhibition, recruitment, and regulator binding in PP2A holoenzymes.

Calcineurin leverages dynamic interactions for substrate specificity

Calcineurin (CN, Protein Phosphatase 2B [PP2B], Protein Phosphatase 3 [PP3]) recognizes regulators, substrates and inhibitors using two SLiMs (PxIxIT/LxVP; Fig. 2g) [20,21]. CN itself is a two-protein complex, with the A subunit, CNA, containing the structured catalytic domain and a long-extended IDR domain that includes an autoinhibitory motif and recruitment sequences for the regulatory B subunit, CNB (which binds Ca²⁺) and calmodulin. CN is activated by calcium (CNB binds Ca²⁺ to order the LxVP recruitment pocket and Ca²⁺-loaded calmodulin binds CNA, leading to the release of the autoinhibitory domain and freeing the catalytic domain to act on CN substrates).

Regulators bind CNA via a PxIxIT SLiM and/or the CNA/B interface via an LxVP SLiM. While the CN PxIxIT binding pocket is always accessible, access to LxVP binding pocket requires Ca²⁺ activation. These motifs were first identified and characterized in the canonical CN substrates, the nuclear factor transcription factors (NFATs), and have since been identified in RCAN1, NHE1, AKAP79, among many others [22,23]. However, recent studies have shown that in addition to SLiM-based recruitment, CN substrate specificity is also achieved through additional mechanisms, including both fuzzy interactions and active site recognition motifs that are unique to CN. For example, NMR data showed that the 27-residue linker connecting the PxIxIT and LxVP SLiMs in the Na⁺/H⁺-exchanger 1 (NHE1) engages with CN in a fuzzy manner (Fig. 2h), making dynamic electrostatic interactions with an expansive conserved acidic patch on CN [22]. Consequently, this fuzzy interaction restricts substrate access to the CN active site, which, in turn, ensures the specificity of CN dephosphorylation towards a single phosphosite of NHE1, T779 (i), whose binding to CN is enhanced via a CN-specific active site recognition motif. Specifically, a binding pocket adjacent to the CN active site, which is unique to CN amongst the PPP phosphatases, coordinates the NHE1 i+3 proline. This enabled the discovery of the first active site recognition motif that is specific to CN, S/TxxP, revealing an additional mechanism by which CN achieves substrate specificity. Together, these data demonstrate how fuzzy linker interactions facilitate specificity of substrate dephosphorylation by CN.

Protein Tyrosine Phosphatases (PTP)

PTPN2 and PTPN22 are dynamically autoinhibited/regulated by C-terminal IDRs

Protein tyrosine phosphatases (PTPs) use a cysteine-based hydrolysis mechanism for catalysis, differing from metal-based mechanism used by PPPs [24]. The PTP catalytic domain is often part of a large multidomain protein, with extended IDRs linking the domains. These IDRs, which can be modified by diverse posttranslational modifications, typically function as protein recruitment platforms and/or mediate PTP allostery. More than 20 years ago, it was shown that the C-terminal IDR inhibits the activity of the T-Cell Protein Tyrosine Phosphatase (TCPTP, PTPN2), a nonreceptor type PTP that is ubiquitously expressed in human cells [25]. However, the molecular basis for this inhibition was only recently discovered. NMR showed that TCPTP’s IDR C-terminus (~100 aa) wraps around its catalytic domain engaging in fuzzy interactions both at the active site and other areas of the protein surface (Fig. 3a) [26,27]. These interactions are autoinhibitory, functioning like a ‘windshield-wiper’ whose dynamic motion, rather than a single bound conformation, restricts substrate access to the active site (Fig. 3b). In distinct cellular compartments, the intrinsically disordered cytosolic tails of membrane receptors compete with and displace the TCPTP autoinhibitory tail, allowing for: (1) the full activation of TCPTP and (2) its localization to the enzymatic point of action. A similarly important regulatory function was also shown for the IDR of protein tyrosine phosphatase nonreceptor type 22 (PTPN22, lymphoid-specific tyrosine phosphatase [Lyp]; important for T cell receptor signaling and thus a drug target for cancer immunotherapy) [28]. PTPN22 has an N-terminal catalytic domain, an IDR interdomain and four poly-Proline repeats. Recent work has shown that phosphorylation of the interdomain (likely by glycogen synthase kinase-3 [GSK3]) at S325 enhances its fuzzy interaction with the catalytic domain, leading to reduction in PTPN22 activity (Fig. 3c) [29]. These examples demonstrate how fuzzy interactions can achieve rapid, reversible autoinhibition and thus local control of enzyme activity.

Figure 3: Tyr phosphatase regulation by fuzzy interactions.

a) Schematic representation of autoinhibition of PTPN2 (TCPTP) by a fuzzy interaction of its C-terminal tail. The inhibition is released when substrates or regulators compete with the interaction site on PTPN2. b) Model of PTPN2 autoinhibition showing PTPN2 residues ~300–340 in an ensemble of conformations blocking the active site (yellow) [26]. c) Schematic representation of autoinhibition of PTPN22 by phosphorylation of Ser325 in its C-terminal region.

Ser/Thr Kinases

In contrast to phosphatases, kinases are well known for using active site recognition motifs (consensus motifs) for substrate recognition. Since these motifs are commonly preserved among kinase families, they can be leveraged to identify substrates. Nevertheless, kinases require additional autoinhibitory elements, pseudo-substrate/phospho-switch motifs and sequence-specific secondary docking sites recognizing kinase specific SLiMs to achieve higher levels of substrate specificity needed for the signaling fidelity in cells [30]. As recently discovered, many of these specificity-defining interactions leverage dynamic charge:charge (fuzzy) interactions.

Mitogen-activated protein kinases (MAPKs) leverage fuzzy interactions via SLiM motifs

MAPKs are terminal effectors of a sequential signaling triad of kinases, that regulate cell survival, apoptosis, gene expression, and differentiation [31,32]. p38, extracellular signal-regulated kinases (ERK) 1/2, ERK5 and c-JUN N-terminal kinase (JNK) form MAPK subfamilies. They share a common catalytic domain containing a T-x-Y motif, which upon dual-phosphorylation, activates the kinase [33]. MAPKs are proline-directed (phosphorylate S/T-P sequences); however, additional interactions ensure substrate specificity and upstream MAPK kinase (MKK) recognition to preserve signaling fidelity [34,35].

Specificity within the MAPK interactome is achieved via two SLiM docking pockets on the catalytic domain: the D-recruitment site (DRS; or kinase interaction motif; KIM) and the F-recruitment site (FRS; Fig. 4a), which recruit the D/KIM-motif and F-motif SLiMs, respectively. D-motifs consist of multiple basic residues connected to a hydrophobic ϕ-x-ϕ motif via a variable linker, while the F-motif consensus sequence is F-x-F-P [36]. Although canonical D-motif interactions have been characterized, recent studies show these views are incomplete. For example, NMR, ITC and X-ray crystallography identified differences in the conformational preferences of the three MKK7 (MAP Kinase Kinase 7) D-motifs, and multi-state conformational exchange of MKK7 D-motifs with JNK1 (Fig. 4b) [37]. In addition, recent efforts identified MAPK selectivity markers that correlate residue composition in the variable linkers of D-motifs to MAPK preferences (e.g., Lys residues for p38α; Pro residues for ERK2) [38]. The fuzzy interactions of Lys-rich linkers with p38α, highlight how fuzzy interactions confer further selectivity to the D-motifs in the MAPK interactome.

Figure 4: Kinase regulation mediated by fuzzy interactions.

a) Schematic representation of MAPK interactions with substrates/regulators via DRS and FRS-specific SLiM motifs. b) Crystal structure of JNK1 in complex with the MKK7 D2 peptide at the DRS (PDB: 4UX9). Chains A, C and D are superimposed to compare bound poses of the peptide. c) Schematic representation of CK1 regulation via its disordered, auto-phosphorylated C-terminal tail. d) Crystal structure of CK1δ bound to the p63 PAD peptide (tiple phosphorylated, PDB:6RU8). The fuzzy electrostatic interactions of the auto-phosphorylated C-tail at this region affect processive phosphorylation kinetics of p63 PAD in site-specific manner. e) Schematic representation of the functionally modular architecture of the C-terminal tail of DCLK1 and the imparted multifaceted autoregulation. The specific functions carried out by various regions are numbered in the order as described in the main text. f) Representative timepoints from MD simulations (source data ref [47]) highlighting functionally distinct regions and interspersed fuzzy regions of the C-tail occupying the DCLK1.2 substrate binding pocket. g) Schematic representation of PKC regulation via its disordered C-terminal V5 region. The intramolecular interactions of PKC V5 with its C2 domain is implicated in priming the kinase for Ca²⁺-mediated activation; and the intermolecular bivalent interactions of V5 with Pin1 implicated in downregulation are depicted. h) NMR structure of PKCα C2 domain in complex with pHM peptide of V5 (PDB: 5W4S), showing three lowest-energy conformers of the bound peptide. i) NMR structure of PKCβII pV5 in complex with Pin1 (PDB: 8SG2), showing three lowest-energy conformers of the bound pV5.

Finally, NMR analysis of ERK2 interactions with the transcription factor Ets-1 identified a bipartite recognition mode mediated by two suboptimal motifs engaging both the DRS and FRS of ERK2 [39]. The N-terminus of Ets-1 mimics the hydrophobic ϕ-x-ϕ motif while lacking the contribution from basic residues for D-sites engagement; this results in fuzzy, low-affinity binding. The C-terminal region of the Ets-1 engages with the FRS via weak rigid body interactions. Dual anchoring enables an ensemble of Ets-1 phospho-acceptor states access to the kinase active site and facilitates phosphorylation in a proximity-dependent manner.

Casein kinase 1 (CK1) is regulated by fuzzy interactions of its auto-phosphorylated C-terminal

The CK1 family of kinases regulate cell division, DNA repair, apoptosis and circadian rhythms [40]. CK1 is constitutively active and prefers ‘primed’ substrates (already phosphorylated at the canonical pS/T-x-x-S/T motifs). Nevertheless, some CK1 substrates can be phosphorylated in the absence of priming, suggesting recruitment mechanisms vary for different substrates. Recent data suggests that CK1 substrate specificity and phosphorylation processivity is directed by autophosphorylation of its IDR C-terminus (Fig. 4c) [41,42]. For instance, CK1 phosphorylation of the p53 family transcription regulator TAp63α shows biphasic kinetics [43] in which the first two sites are rapidly phosphorylated, while the functionally decisive third site is phosphorylated about ~20-fold more slowly (Fig. 4d). NMR data showed that the auto-phosphorylated IDR C-terminus selectively inhibits the phosphorylation at the third TAp63α site via competing fuzzy interactions with the substrate pocket [44]. Furthermore, CK1δ splice variants (CK1δ1/2) essentially have identical kinase domains but differ in their variably spliced extreme C-terminus (XCT; last 16 residues). Consequently, their effect on circadian rhythms is distinct. NMR and H/D exchange MS showed that this difference is due to differential autophosphorylation of the XCT, leading to distinct autoinhibitory interactions with their cognate kinase domains [41]. Together, these data highlight the role of the C-terminal tail in CK1 regulation by phosphorylation-dependent fuzzy interactions.

Doublecortin-like kinase 1 (DCLK1) are autoregulated by its modular C-terminus

DCLK family kinases contain an N-terminal microtubule-binding doublecortin-like (DCX) domain, a Ca²⁺/calmodulin-dependent protein kinase (CAMK)-like domain and a C-terminal autoinhibitory IDR [45]. Recent data identified autoregulatory roles of the DCLK1 IDR C-terminus via different sub-regions (Fig. 4e) [46,47]. Region 1 constitutes a fuzzy IDR segment. Region 2 contains a pseudo-substrate mimic that binds and blocks the substrate binding groove via competing fuzzy interactions (Fig. 4f). Region 3 blocks ATP binding by completing the C-spine and provides a γ-phosphate mimic in the form of an auto-phosphorylatable thr residue. Region 4 leverages the C-terminus to create a pocket formed by the transient formation of complementary β-strand adjacent to ATP-binding pocket. Fuzzy region 5 blocks substrate access in a manner akin to a pseudo-substrate. Lastly, region 6 is a CAMK-tether that makes persistent contact with the CAMK-specific insert. Together, these multiple regions control substrate access, kinase activity, and ATP binding. This “swiss army knife” design of the DCLK1 C-terminal IDR highlights how fuzzy interactions leverage interspersed functionally-specific modules to achieve efficient kinase regulation.

Protein Kinase C (PKC) is regulated by context-dependent interactions of C-terminal IDR

PKC is a family of lipid-activated kinases with distinct domain organizations and second messenger sensitivities at the regulatory N-terminus [48]. In the cytosol, the pseudo-substrate containing N-terminus autoinhibits kinase activation (Fig. 4g). During activation, PKC is anchored to the membrane via its C1 (DAG-binding) and/or C2 (Ca²⁺/PtdSer/PtdIns(4,5)P₂-binding) domains, allowing the release of the pseudo-substrate inhibition. The C-terminal V5 domain is intrinsically disordered and contains two functionally relevant phosphosites: the turn motif (TM) and the hydrophobic motif (HF). NMR characterization of the V5 domain revealed a context-dependent functionality for this domain. For PKCα, the NMR-derived structural ensemble of the C2α:Ca²⁺:pHM ternary complex (Fig. 4g,h) shows that pHM engages intramolecularly with the conserved PtdIns(4,5)P₂-binding region of C2 domain via electrostatic and aromatic interactions [49]. This interaction supports the autoinhibitory assembly, while increasing C2 Ca²⁺sensitivity. Moreover, recent data identified a novel, bivalent intermolecular interaction of the PKCα and PKCβII V5 with the peptidyl-prolyl isomerase Pin1 (Fig. 4g,i) [50]. In this complex, the native linker connecting pHM and pTM motifs retains flexibility, suggesting an extended array of Pin1 conformations can be supported. Thus, the fuzziness afforded to Pin1 may affect PKC downregulation.

Concluding remarks and perspective

Formation of macromolecular assemblies and their regulation is essential for high-fidelity signaling. IDPs/IDRs provide an adaptable, specificity-defining, and yet energetically efficient platform for protein:protein interactions due to their ability to engage in multivalent fuzzy interactions. Efficient molecular recognition is particularly important for catalysis, evidenced by the abundance of fuzzy interactions involved in the regulation of kinases and phosphatases. It is important to note that given at least a third of human proteome is made up of disordered sequences, the characterization of novel mechanisms leveraging fuzzy interactions in protein regulation is far from complete.

Data availability

No data were used for the research described in this article.