Features of molecular recognition of intrinsically disordered proteins via coupled folding and binding

Abstract

The sequence–structure–function paradigm of proteins has been revolutionized by the discovery of intrinsically disordered proteins (IDPs) or intrinsically disordered regions (IDRs). In contrast to traditional ordered proteins, IDPs/IDRs are unstructured under physiological conditions. The absence of well‐defined three‐dimensional structures in the free state of IDPs/IDRs is fundamental to their function. Folding upon binding is an important mode of molecular recognition for IDPs/IDRs. While great efforts have been devoted to investigating the complex structures and binding kinetics and affinities, our knowledge on the binding mechanisms of IDPs/IDRs remains very limited. Here, we review recent advances on the binding mechanisms of IDPs/IDRs. The structures and kinetic parameters of IDPs/IDRs can vary greatly, and the binding mechanisms can be highly dependent on the structural properties of IDPs/IDRs. IDPs/IDRs can employ various combinations of conformational selection and induced fit in a binding process, which can be templated by the target and/or encoded by the IDP/IDR. Further studies should provide deeper insights into the molecular recognition of IDPs/IDRs and enable the rational design of IDP/IDR binding mechanisms in the future.

Short abstract

PDB Code(s): 2NL9, 2FO1, 2LQI, 2N3K, 2AZE, 3KXY

Abbreviations

3D

three‐dimensional

AS

alternative splicing

IDPs

intrinsically disordered proteins

IDRs

intrinsically disordered regions

LFER

linear free‐energy relationships

LLPS

liquid–liquid phase separation

MoRFs

molecular recognition features

PTM

posttranslational modification

SLiMs

short linear motifs

1. INTRODUCTION

Proteins are important biological molecules. The three‐dimensional (3D) structure, which is determined by the primary amino‐acid sequence, is critical for a protein to carry out its functions. Traditionally, proteins are classified as being either ordered (folded) or disordered (unfolded) by analyzing their conformational states. Ordered proteins have well‐defined 3D structures and exhibit small‐scale structural fluctuations under physiological conditions. On the contrary, intrinsically disordered proteins could sample an ensemble of conformations which may be compact (molten globule‐like) or extended (coil‐like or pre‐molten globule‐like).1, 2, 3, 4, 5 Furthermore, proteins can be entirely disordered polypeptides (IDPs) or a combination of disordered regions (IDRs) and ordered domains.6, 7, 8

Based on bioinformatics predictions, IDPs/IDRs are abundant in all species.9, 10, 11 By analyzing the proteomes of 3,484 species and correlating the fraction of disordered residues with proteome size, it is shown that eukaryotes have more disordered residues than prokaryotes.12 A recent comprehensive analysis of over 6 million proteins characterized intrinsic disorder at proteomic and protein levels indicates that IDPs/IDRs are more abundant in eukaryotes and certain functions are exclusively implemented by IDPs/IDRs.13 The correlation between the organism complexity and the amount of intrinsic disorder are consistent with the extensive involvement of IDPs/IDRs in regulatory and signaling functions and the increased disorder content in eukaryotic proteomes might be used by nature to deal with the increased cellular complexity.12, 14, 15

IDPs/IDRs are involved in various biological functions. In a comprehensive bioinformatics study carried out by Xie et al.,16, 17 a positive correlation between the functional annotation of the SwissProt database and the predicted intrinsic disorder has been found. Generally, IDPs/IDRs are enriched in proteins involved in signaling and regulatory functions, including transcription regulation, cell cycle, mRNA processing, scaffolding, and apoptosis.6, 14, 15, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 Consequently, dysregulation of IDPs/IDRs are associated with a variety of human diseases.32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44 Most recently, many IDPs/IDRs are found to be able to undergo liquid–liquid phase separation (LLPS), which is related to the assembling of membraneless organelles in vivo.45, 46, 47, 48, 49, 50, 51, 52, 53 So far, studies on IDPs/IDRs have greatly extended our understanding on the sequence–structure–function relationship of proteins.29, 54, 55, 56 Recently, the protein structure–function continuum concept was proposed by Uversky to illustrate the numerous biological functions of p53 through multiple proteoforms by various mechanisms and may be extended to many multi‐function IDPs.57

In this review, we will summarize recent advances of our understanding on the molecular recognition of IDPs/IDRs. We will focus on specific interactions between IDPs/IDRs and their targets, which usually result in folding of the IDPs/IDRs upon target binding. We will discuss the mechanistic features of molecular recognition inferred from kinetics, thermodynamics, and structure investigations.

2. MOLECULAR RECOGNITION FEATURES

The flexible structures of IDPs/IDRs make them suitable for cellular regulatory and dynamic signaling processes.14 Several functional modes have been summarized for IDPs/IDRs, including entropic chains, effectors, scavengers, assemblers, display sites, and chaperones.7, 30, 58 A common module for molecular recognition within IDPs/IDRs is often known as molecular recognition features (MoRFs) or short linear motifs.59, 60, 61, 62, 63 The sequence features of MoRFs are distinct from the rest portion of IDPs/IDRs, enabling development of predictors to identify MoRFs.64 For example, ANCHOR predicts disordered binding regions based on the pairwise energy estimation from IUPred.65, 66, 67, 68 Usually, upon binding to their partners, MoRFs undergo disorder‐to‐order transitions. This process is termed coupled folding‐binding.69, 70 The structures of MoRFs adopted upon binding can be divided into three types: α‐helix, β‐strand, and irregular secondary structure. IDPs/IDRs can utilize multiple MoRFs simultaneously when interacting with their binding partners (Figure 1).59

Figure 1.

Examples of intrinsically disordered protein (IDP) complex with various combinations of molecular recognition features (MoRFs). IDPs are shown in rainbow color and the targets are shown in gray. PDB IDs are: Bim/MCL‐1 (2NL9), CSL/notch (2FO1), FOXO3a/KIX (2LQI), MLV IN/Brd4 (2N3K), Rb/E2F1‐DP1 (2AZE), and ExsC/ExsE (3KXY)

Studying the recognition mechanisms of IDPs/IDRs with their partners is not a trivial task. In recent years, however, large progress has been made from a close collaboration between experimental and computational studies. A spectrum of techniques have been applied to study IDPs/IDRs, providing valuable information on their structures, dynamic properties and binding mechanisms.2, 3, 71, 72, 73, 74, 75 In parallel, molecular modeling and computer simulations provide atomic pictures of conformation ensembles and binding processes as well as reveal important underlying physical principles.76, 77, 78, 79, 80, 81, 82

3. RATE CONSTANTS

Coupled folding with binding has been suggested to enhance the binding rates of IDPs/IDRs.83 Theoretical analysis and computer simulations predicted that the “fly‐casting” effect accelerates the binding rate by twofolds to threefolds.83, 84, 85 Consistent with this prediction, rate constants of IDP/IDR‐protein interactions from the literature show differences from those of ordered proteins in a general trend.85 The binding kinetics of IDPs/IDRs is affected not only by the overall structure flexibility, but also by the local conformation preference of MoRFs. Stabilizing the preformed conformation of MoRFs has been found to accelerate the association rate constants (k _on), due to an increase of the probability of converting collision complexes to bound state.86, 87, 88, 89, 90 On the other hand, increasing the degree of disorder has been found to significantly increase the dissociation rate constant (k _off), suggesting that the dominant effect of disorder on molecular recognition may be to accelerate dissociation rather than association.87, 91 While computer simulations provide detailed correlation between conformation disorder and binding/unbinding rate constants, it is difficult to test the actual role of disorder in binding kinetics experimentally as it is hard to quantify the extent of disorder and the influence on binding kinetics could be resulted from changes in the interactions between the target and IDP/IDR or changes in the binding mechanism.

Electrostatic interactions could play important roles in the coupled folding‐binding process of IDPs/IDRs as many MoRFs contain charged residues.92 For many studied IDPs/IDRs, the k _on values are reduced for about ten folds when the salt concentration increases from low (~50 mM) to high (~500 mM), that is, ∂log(k_on)/∂log(C_salt) ≈  − 1 (Table 1), indicating the presence of favorable electrostatic interactions. Molecular dynamics simulations found that long‐range electrostatic interactions accelerate the binding rate in a range consistent with experimental results.97, 104, 105, 106, 107, 108 More importantly, simulations revealed that electrostatic forces enhance the binding kinetics not only by increasing the encounter rate but also by enhancing the efficiency of IDPs/IDRs evolving to the bound states upon encounter.105, 108 Advances in single molecule techniques allow the detection of transient species during a coupled folding‐binding process. Interestingly, the transition path times of ACTR/NCBD interaction is much longer than the transition path times of protein folding, indicating the presence of stable intermediate state along the binding process.109 Furthermore, the lifetime of transient complexes for ACTR/NCBD is also longer than that for barnase/barstar,110 consistent with previous simulation predictions.85

Table 1.

Effect of salt concentration on the association rate constant of IDPs

IDP	Number of charges		Target	∂log(kon)/∂log(Csalt)	Reference
p5313‐61	–11	+1	NCBD	−2.10	93
HPV E7	−4	+0	Rb	−1.56	94
MLL	−6	+1	KIX	−1.52	95
STAT2	−10	+4	TAZ1	−1.32	96
ACTR	−11	+4	NCBD	−1.26	93
p27	−16	+14	Cdk2/cyclin A	−1.21	97
SRC1	−11	+5	NCBD	−1.15	93
E3IDP	−12	+20	Im3	−1.06	98
TIF2	−9	+2	NCBD	−0.86	93
PUMA	−10	+6	MCL‐1	−0.68	99
WASP	−21	+21	Cdc42	−0.62	100
p5313‐63	−12	+1	Mdm2	−0.56	101
p7311‐25	−3	+0	Mdm2	−0.38	101
c‐Myb	−5	+5	KIX	−0.25	102
p5315‐29	−3	+1	Mdm2	−0.21	101
p6352‐65	−3	+0	Mdm2	0.13	101
eIF4G	−2	+4	eIF4E	0.75	103

Abbreviation: IDP, intrinsically disordered protein.

It is noted that as the conformations of IDPs/IDRs are highly dynamic, electrostatic interactions between IDPs/IDRs and their targets during the encounter process may be different from those of the corresponding ordered proteins, which may be reflected from a recent study on the interactions between the colicin E3 rRNase domain (E3) and the immunity protein Im3 (Figure 2a).98 Although E3 is a folded domain, disorder prediction suggests that it contains high disorder propensities (Figure 2b,c). Actually, a single alanine mutation at Tyr507 within the hydrophobic core of E3 causes the protein to become an IDP (E3^IDP). Kinetics studies show that k _on of E3^WT with Im3 is decreased by three orders of magnitude when the salt concentration is increased. However, under the same range of salt concentration, k _on of E3^IDP with Im3 is only decreased by less than 40 folds (Figure 2d).98 However, as the mechanism of E3^IDP binding to Im3 is unclear, it is unknown whether E3^IDP folds before binding or folds during binding. In this context, the salt dependence of the association rate constant for E3^IDP/Im3 interaction remains elusive.

Figure 2.

Binding of E3 to Im3. (a) Crystal structure of E3/Im3 complex. E3 is shown in rainbow color and Im3 in gray. (b) Disorder propensity prediction of E3 using three different predictors: IUPred2 (black), MFDp2 (red), and PONDR VL‐XT (blue). (c) Location of E3 on the charge‐hydrophobicity plot. The black line indicates the boundary between the intrinsically disordered proteins (IDPs) region and the ordered proteins. (d) Effect of salt concentration on k _on for E3^WT and E3^IDP 98

4. CONFORMATIONAL SELECTION, INDUCED FIT AND BEYOND

Recently, Dunker and Oldfield111 suggested that the interaction between an IDP/IDR and its partner should not be described as induced fit where the protein is folded but can adjust its structure to fit the substrate. However, since the discovery of IDPs/IDRs, the sequence–structure paradigm has been revolutionized. In this context, it should be reasonable to expand the concepts of induced fit to analyze the binding processes of IDPs/IDRs. Thus, in an induced fit process, unfolded conformations of an IDP/IDR are able to weakly interact with the target to form encounter complexes which induce the unfolded conformations fold into the bound conformations. On the contrary, in a conformational selection process, an IDP/IDR samples unfolded conformation as well as pre‐folded conformations and only the prefolded conformations are binding competent.

Conformational selection and induced fit have been widely applied to explain the coupled folding‐binding process of IDPs/IDRs.75, 112, 113 Which mechanism dominates during the binding process depends on several factors, including the structure preference and conformational dynamics of the IDPs/IDRs, the association rate, and the concentration as well.114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125 It has been established that IDPs/IDRs sample a variety of conformations rapidly.126, 127, 128, 129 At one extreme, if the conformation ensemble of an IDP/IDR in the unbound state is completely different from that in the bound state, it is expected that the binding process proceeds via the induced fit mechanism. Except this extreme condition, (partially) bound‐like conformations could be sampled by the free IDPs/IDRs. It is plausible that these preformed bound‐like conformations can also initiate the binding process. Under such circumstance, the observed binding mechanism is determined by a competition between the flux of conformational selection and that of induced fit.119, 120, 121, 122 The flux from unbound state to bound state is determined by the folding/unfolding rate constants, association/dissociation rate constants as well as protein concentrations.119 The flux description predicts that conformational selection is favored when the folding kinetics of free IDPs/IDRs is fast, affinity for inactive conformations is low, and protein concentration is low.121 In another study, similar conclusions are reached for the effect of protein concentration via molecular dynamics simulations; however, the effect of conformation transition kinetics is opposite.120 Sampling of the bound‐like conformations in the unbound state is necessary but not sufficient for a conformational selection process. As conformational transitions occur in the unbound state as well as in the loosely bound state, increasing conformation transition kinetics will push the mechanism toward induced fit.120, 122

Several experimental strategies have been proposed to distinguish conformational selection from induced fit. Weikl and Deuster114 proposed a framework by perturbing the conformational equilibrium between the inactive conformations and active conformations via introducing mutation far from ligand binding site. In the case of conformational selection, such mutations will mainly change the association rate, whereas in the case of induced fit, the dissociation rate will be mainly affected. In the interactions between the BH3 motif of PUMA and the structured protein MCL‐1, the helical structure of PUMA was modulated by mutating solvent‐exposed residues to proline or glycine.87, 130 The mutations resulted in a modest effect on k _on but a significant effect on k _off, suggesting that the PUMA/MCL‐1 interaction is an induced fit process. Stabilizing the helical conformation by trifluoroethanol may be applied to perturb the conformational equilibrium as well. Increasing the trifluoroethanol concentration increased the helix content of c‐Myb and decreased the dissociation rate of c‐Myb/KIX complex, suggesting that folding of c‐Myb is induced after KIX binding.131 The ACTR/NCBD interaction was investigated by selectively perturbing the amount of secondary structure in free ACTR via mutation and k _on and k _off were affected to similar extent,86, 89 suggesting that conformational selection is involved in the ACTR/NCBD binding process. While mutational analysis provides clues to speculate the binding mechanism, a correlation between helix propensity and rate constants is not a proof for conformational selection.116, 132 Other proposed strategies rely on measuring the observed rate constant under various ligand or/and target concentrations and investigating the dependence of observed rate constant on concentration.115, 116, 117, 118 For example, a comparison of the observed rate constant for various ACTR/NCBD concentrations and N_TAIL/XD concentrations suggest that their binding processes are induced fit.116, 133, 134

From studies on IDP/IDR‐protein interactions, it is likely that the binding processes are induced fit combined with various degree of conformational selection.75, 135 As discussed above, the relative flux through these two pathways is determined by the protein concentrations, association rate and conformation transition kinetics. It is plausible that the partially preformed bound‐like conformations play a role in the initial binding step, forming weak encounter complexes which further evolve into the bound conformation.90 Recently, through NMR investigation, Schneider et al.136 found that the free‐state conformational equilibrium of N_TAIL is funneled by interactions with XD, leading to preformed bound‐like conformations in the encounter complex. Thus, the free‐state conformational transition of an IDP/IDR and its interactions with the target are coupled in the “conformational funneling” description of the folding‐binding process.132, 136 Structure information on the encounter complexes, the intermediates, and transition states will be of great value for comprehensive understanding of the entire binding process.

5. THE TRANSITION STATES

It is important to analyze the transition state to understand how a coupled folding‐binding process crosses the free energy barrier. This can be achieved by ϕ‐value analysis and linear free‐energy relationships (LFERs) analysis. In a coupled‐folding binding process, the ϕ‐values are calculated from the free energy change for the transition state (ΔΔG ^‡) and at equilibrium (ΔΔG _Eq):

$$\varphi =\frac{\mathrm{\Delta \Delta }{G}^{‡}}{\mathrm{\Delta \Delta }{G}_{\mathrm{Eq}}},$$ (1)

$$\mathrm{\Delta \Delta }{G}^{‡}=\mathit{RT}\ln \left(\frac{k_{\mathrm{on}}^{\text{wild}‐\text{type}}}{k_{\mathrm{on}}^{\text{mutant}}}\right),$$ (2)

$$\mathrm{\Delta \Delta }{G}_{\mathrm{Eq}}=\mathit{RT}\ln \left(\frac{K_{\mathrm{d}}^{\text{mutant}}}{K_{\mathrm{d}}^{\text{wild}‐\text{type}}}\right).$$ (3)

By comparing the influence of a point mutation on k _on and K _d, the ϕ‐value of a residue provides information on the proportion of native contacts (either intermolecular or intramolecular) it makes at the transition state. In general, residues with ϕ ≈ 0 and ϕ ≈ 1 indicate their structures in the transition state resemble those in the unbound state and the bound state, respectively.

Although with a lower resolution, the location of transition state can also be inferred from the LFER analysis. Small structure alteration on the unbound molecules will result in changes in the complex stability and association kinetics by:

$$\log \frac{k_{\mathrm{on}}^{\mathrm{MT}}}{k_{\mathrm{on}}^{\mathrm{WT}}}=\mathrm{\alpha }\log \frac{K_{\mathrm{d}}^{\mathrm{WT}}}{K_{\mathrm{d}}^{\mathrm{MT}}}.$$ (4)

The parameter α (0 ≤ α ≤ 1) measures the location of the transition state along the binding path. Binding processes with α ≈ 0 and α ≈ 1 mean that the transition state is unbound‐like and bound‐like, respectively. The LFER analysis is helpful to identify the location of transition state when mutations result in small changes in stability, prohibiting reliable calculations of ϕ‐values.

ϕ‐value analysis and LFER analysis have been applied to many IDP/IDR complexes (Table 2). In most studied cases, low fractional values of ϕ and α are commonly observed, indicating that IDPs/IDRs remain largely unstructured in the transition states. This is manifested in the conformation ensemble of transition states obtained via molecular dynamics simulations using ϕ‐values as restraints.138, 147 It is noted that ϕ‐values are not evenly distributed along the sequences. Residues with high ϕ‐value may serve as the anchor sites to stabilize the encounter complexes, allowing the encounter complexes to cross the free energy barrier and evolve to native bound states. This picture resembles the dock‐and‐coalesce mechanism proposed by Zhou et al.150 Furthermore, low values of ϕ and α highlight the importance of non‐specific interactions (including electrostatic and hydrophobic interactions) in the initial stage of binding, probably stabilizing the encounter complexes.

Table 2.

ϕ‐Value analysis of coupled folding‐binding process

IDP	Target	ϕ‐Value distribution	Reference	IDP	Target	ϕ‐Value distribution	Reference
HIF‐1α	TAZ1		137	c‐Myb	KIX		88, 138, 139
STAT2	TAZ1		96	pKID	KIX		140
PUMA	A1		141	MLL	KIX		142
PUMA	MCL‐1		130, 141	E6 peptide	PDZ2		143
BID	A1		141	S peptide	S protein		144
BID	MCL‐1		141	NTAIL	X domain		145
α‐Spectrin	β‐Spectrin		146	ACTR	NCBD		86, 147, 148
C‐terminal tail of nNOS PDZ	Syntrophin PDZ		149

Note: The IDPs are colored in light green and the targets are colored in gray. Residues with low (ϕ ≤ 0.25), medium (0.25 < ϕ < 0.6), and high (ϕ ≥ 0.6) ϕ‐values are highlighted in blue, magenta, and red, respectively.

6. IDPS/IDRS ENCODED BINDING VERSUS TARGET TEMPLATED FOLDING

Since IDPs/IDRs are mainly unfolded in their unbound states, their folded structures observed in the complex state should be induced or stabilized by the binding partners. It remains unclear how a coupled folding‐binding process is encoded. The sequence–structure relationship of proteins tells that the 3D structure of a protein is primarily encoded by its sequence. Extending this paradigm to the coupled folding‐binding of IDPs/IDRs, it is expected that the folded structure of an IDP/IDR in its bound state and its binding mechanism are determined by its sequence and/or the target's sequence (thus the target's structure).

An IDP/IDR binds to diverse targets and folds into similar structures resembling its free conformation ensemble should support that the folded structure of an IDP/IDR is encoded by its sequence. An example is the N‐terminal transactivation domain of p53, which forms similar α‐helical structures upon binding to Mdm2 (PDB 1YCR), MdmX (PDB 2MWY), TAZ1 (PDB 5HOU), TAZ2 (PDB 2MZD), tfb1 PH domain (PDB 2GS0), RPA70 (PDB 2B3G), HMGB1 (PDB 2LY4), and NCBD (PDB 2L14). Besides the folded structure, the binding mechanism can also be encoded by the IDPs/IDRs. Clarke et al141 compared the transition states between the disordered BH3‐only proteins PUMA and BID and the folded BCL‐2–like proteins A1 and MCL‐1 using ϕ‐value analysis. They found that the ϕ‐value profiles for PUMA and BID are conserved when binding to different partners, suggesting that the binding processes of PUMA and BID are encoded by the IDPs/IDRs but not templated by the partners (Figure 3a). Recently, Wu and Zhou altered the binding pathways of WASP with Cdc42, either suppressed the original dominant pathway or promoted a new dominant pathway through manipulating the charged residues on WASP,151 further emphasized the role of IDPs/IDRs in encoding the binding process.

Figure 3.

Illustrations of intrinsically disordered protein (IDP) encoded binding and target templated binding. (a) ϕ‐values for the disordered BH3‐only protein PUMA binding with BCL‐2–like proteins A1 and MCL‐1.141 (b) correlation of α‐value from linear free‐energy relationships (LFER) analysis for c‐Myb with the hydrophobic solvent accessible surface area in the binding site on KIX138

On the other hand, there are also evidences showing that the folding/binding process is templated by the target. Although they bind to the same pocket of S100B, the C‐terminal segment of p53 and TRTK‐12 show low sequence similarity and form different bound conformations. However, computer simulations show strong similarities in the binding intermediate states of the two peptides, suggesting that S100B templates the binding process.152 Toto et al. perturbed the hydrophobic network of KIX using site‐directed mutagenesis and investigated the folding mechanism of c‐Myb with wild‐type and mutated KIX.138 By performing a LFER analysis, they found that the α‐value decreases from 0.89 for wild‐type KIX to around 0.5 for I26V and L43A mutants and to 0.19 for I72V mutant. The decrease in α‐value appears to be correlated to the decrease in the hydrophobic solvent accessible surface area in the binding site on KIX [Figure 3b]. It is surprised to find that the structure of KIX dictates the folding mechanism of c‐Myb as c‐Myb has a strong propensity for α‐helix formation in its N‐terminus and the coupled folding‐binding process of c‐Myb to KIX has been suggested to involve elements of conformational selection.88 The templated folding mechanism has been suggested to enable IDPs/IDRs to be specifically recognized by multiple targets.138 A similar strategy was applied to investigate the interactions between N_TAIL and XD and revealed that the binding process of N_TAIL is very malleable and is affected by the structure of XD.153 On the contrary, mutagenesis and ϕ‐value analysis revealed that the transition state of ACTR/NCBD complex is highly heterogeneous and is robust with respect to most mutations for ACTR or NCBD.147

As IDPs/IDRs possess various sequence and structure preferences, the above discussions indicate that the coupled folding‐binding process of an IDP/IDR could be templated by the partner as well as encoded by its sequence. Further mechanistic studies are required to reveal the microscopic details on how a target templates or how an IDP/IDR encodes the binding/folding process.

7. EFFECT OF MACROMOLECULAR CROWDING

The intracellular environment is very crowded since up to 40% of the volume of a cell is occupied by biological macromolecules.154 Macromolecular crowding can affect protein–target binding and protein folding.155 In particular, the malleability of IDPs/IDRs makes them susceptible to the influence of macromolecular crowders.156, 157 Conformational compaction of IDPs/IDRs by macromolecular crowders has been observed, where the effect depends not only on the crowder size and concentration, but also on the properties of IDPs/IDRs.158, 159, 160, 161, 162, 163 MAP2c, p21^Cip1, and FlgM show global compaction and local structuring in crowded conditions.164, 165 The distal helix of calcineurin and transiently helical regions of ACTR are also stabilized when crowded by synthetic polymers.166, 167 However, conformational compaction induced by crowders is not necessary to promote secondary structure formation for IDPs/ IDRs. For example, α‐Casein, the C‐terminal activation domain of c‐Fos, and the kinase‐inhibition domain of p27^Kip1 shows little structural changes under crowded conditions.164, 168

Besides modulating the conformational properties of IDPs/IDRs, macromolecular crowding also affects their diffusion properties. In general the translational and rotational diffusions of IDPs/IDRs are reduced.169, 170 Interestingly, the effect of crowding on the diffusion of IDPs/IDRs is less than that on folded proteins.169, 170 Consequently, larger IDPs/IDRs may diffuse faster than smaller folded proteins in cells.169 Recently, study on FlgM under crowded condition reveals the presence of extended conformations which snake through interstitial crevices and bind multiple crowders simultaneously.171 It is probable that such extended conformations may facilitate recognition of IDPs/IDRs under crowded conditions.

It is also important to directly study the effect of macromolecular crowding on the molecular recognition process. Binding of calmodulin with CaMKI peptide was investigated under crowded conditions.172 It was found that the on‐ and off‐rates are reduced by about two folds in a compensatory fashion, thus the binding affinity is almost not changed. The reduction of association rate constant suggests that binding of CaMKI peptide with calmodulin is under diffusion control and crowding slows down the diffusion process.155, 172 For reaction control binding process, it is expected that the association rate constant will be increased.155 In another study, computer simulation on the coupled folding‐binding of pKID with KIX showed that the folding‐binding mechanism observed in bulk solution remains unchanged under highly crowded conditions.173 It seems that molecular crowding has small effect on the binding mechanism of IDPs/IDRs.

8. DYNAMIC CONTACTS AND FUZZY INTERACTIONS

While the main recognition elements are folded upon binding for many IDPs/IDRs, they may still exhibit conformational dynamics in the complex state.174, 175 For example, the TAD of STAT2 only undergoes a partial disorder‐to‐order transition upon binding with TAZ1 and retains subnanosecond motions.176 Conformational dynamics in the bound state enables the IDPs/IDRs to form polymorphic contacts with the partners.177 Such heterogeneity in the bound form is referred to as fuzziness.175, 178 Fuzziness and dynamic binding are universal in the molecular recognition of IDPs/IDRs and are beneficial for their function.8, 174, 179 The presence of fuzzy interacting regions adjacent to the main binding elements can regulate binding affinity, specificity, and selectivity.180, 181, 182, 183 Fuzzy regions also facilitate IDPs/IDRs‐mediated allosteric communication.184 Furthermore, transient binding interactions can promote formation of non‐native interactions stabilizing the encounter complexes, thus enhance the binding kinetics.185 Dynamic interactions can also modulate the folding‐binding mechanism of IDPs/IDRs.186 As discussed above, folding of c‐Myb is templated by KIX, where transient non‐native hydrophobic interactions between c‐Myb and KIX populate when the hydrophobic surface in the binding site of KIX is enlarged.138

Multivalent dynamic interactions are the main driving forces of LLPS.187, 188 Though many IDPs/IDRs involved in LLPS apply multisite electrostatic and aromatic interactions, dynamic coupled folding‐binding interactions mediated by specific recognition elements can also drive LLPS. For example, interactions between SH3 domain and proline‐rich motif are involved in the LLPS of the nephrin–NCK–N‐WASP system and the RIM–RIM‐BP system.189, 190 PDZ domain‐mediated binding is required for phase separation of PSD scaffold proteins.191, 192 Multivalent arginine‐rich linear motifs interact with the NPM1 pentamer, leading to LLPS.193 Within the protein‐rich droplets, non‐native transient interactions are expected to become more populated than in dilute solution. Nevertheless, the specific binding between the recognition motif and the target domain should remain unchanged.

9. POSTTRANSLATIONAL MODIFICATIONS

IDPs/IDRs are enriched in posttranslational modification (PTM) sites, such as phosphorylation, acetylation, and methylation.16, 194, 195 PTMs can regulate molecular recognition of IDPs/IDRs in various ways.18 For example, PTMs can alter the free energy landscapes of IDPs/IDRs, leading to changes in the conformation ensembles. Bah et al. showed that phosphorylation of 4E‐BP2 at T37 and T46 induces folding of 4E‐BP2 into a four β‐strand structure, sequestering the eIF4E‐binding motif and blocking its accessibility to eIF4E.196 However, the structural changes induced by PTMs could also be subtle, as observed in the N‐terminal transactivation domain of p53.197 PTMs located at the binding interface can directly regulate the interactions between IDPs/IDRs and the targets, for example, phosphorylation will introduce electrostatic interactions between the phosphate moiety and the binding partner.198 Interestingly, PTMs alter not only the equilibrium conformation ensemble but also conformational exchange among different conformations.199 Consequently, the binding mechanisms can also be modulated by PTMs.

10. ALTERNATIVE SPLICING

Alternative splicing (AS) generates various protein forms from a single gene. Previous studies have revealed that AS sites are often located within IDRs which are enriched in molecular recognition motifs.200, 201 As molecular recognition processes of IDPs/IDRs are mainly mediated by short recognition elements, removal of recognition elements by AS will eliminate existed molecular interactions or enable new interactions when competitive interactions are removed. Some proteins contain auto‐inhibition segments that mask the binding sites or compete with other molecules for binding.202 Removal of the auto‐inhibition segments by AS will switch the proteins into active states or increase the binding affinities for other molecules. Consequently, removing disordered segments containing different functional or signaling elements allows for rewiring the cellular signaling pathways.19, 26, 203

11. CONCLUSIONS AND FUTURE PERSPECTIVES

IDPs/IDRs are abundant in all species and involved in vital biological processes. Coupled folding upon binding is an important mode of molecular recognition for IDPs/IDRs. IDPs/IDRs can employ various combinations of conformational selection and induced fit mechanisms and the binding process can be templated by the target and encoded by the IDP/IDR as well. The coupled folding‐binding process can also be heterogeneous or fuzzy. While great efforts have been devoted to investigating the complex structures and binding kinetics and affinities, our knowledge on the binding mechanism of IDPs/IDRs remains very limited. Application of advanced kinetic techniques and NMR will provide deeper understanding on the features/mechanisms of molecular recognition of IDPs/IDRs in the future, which may enable rational design of IDP/IDR binding mechanisms.

Yang J, Gao M, Xiong J, Su Z, Huang Y. Features of molecular recognition of intrinsically disordered proteins via coupled folding and binding. Protein Science. 2019;28:1952–1965. 10.1002/pro.3718