5.6: Binding - Conformational Selections and Intrinsically Disordered Proteins

Search Fundamentals of Biochemistry

Learning Goals

(Learning goals written by Claude, Sonnet 4.6, Anthropic)

Conformational Selection vs. Induced Fit

• Distinguish conformational selection (ligand binds preferentially to one member of a pre-existing ensemble of protein conformations without inducing subsequent structural change) from induced fit (ligand binds loosely to a single conformation and subsequently stabilizes a rearranged, tighter complex), and explain how each model generates distinct kinetic signatures — non-linear vs. linear ligand-concentration dependence of the slow and fast kinetic phases — that allow experimental differentiation by stopped-flow fluorescence or NMR techniques.
• Interpret key experimental evidence supporting conformational selection — including (1) one antibody binding two structurally distinct ligands through two pre-existing conformations, (2) ubiquitin's NMR-determined solution ensemble matching the 46 distinct conformations seen in crystal structures of its ligand complexes, and (3) AFM energy landscape data showing that mastoparan stabilizes the folded form of calmodulin without affecting the folding rate (selecting a pre-existing conformation) while Ca²⁺ increases the folding rate (stabilizing the transition state) — and explain what each result implies about the nature of the binding mechanism.
• Explain the "fly casting model" as a variant of induced fit relevant to partially or fully disordered proteins: initial rapid hydrophobic collapse onto the ligand (fast, linearly ligand-dependent phase) is followed by slow conformational rearrangement into the native-like bound complex (slow, ligand-independent phase at high [L]), and contrast this with classic induced fit and with conformational selection in terms of the order of binding and folding events.

Molecular Recognition by Intrinsically Disordered Proteins

• Define Molecular Recognition Features (MoRFs) as short disordered segments (< ~70 amino acids) within IDPs that adopt defined secondary structures — α-MoRFs (helices), β-MoRFs (strands), ι-MoRFs (irregular), or complex-MoRFs — upon encountering a binding partner, and explain how this coupled folding-and-binding mechanism allows IDPs enriched in polar residues to form binding interfaces with hydrophobic character comparable to or larger than ordered protein interfaces.
• Identify the physicochemical properties that distinguish MoRF binding interfaces from typical ordered protein surfaces — enrichment in hydrophobic and methionine residues at the interface despite the IDP's overall polar composition, few prolines in α-MoRFs, and large solvent-exposed surface area before binding — and explain why these properties make MoRFs computationally predictable and why Met is particularly well-suited to hydrophobic binding interfaces.
• Describe the three classes of IDP:target interaction (folded complex formation, fuzzy complex formation with retained disorder, and IDR:IDR disordered complex), explain how computational tools like FINCHES-online use coarse-grained force fields to predict these interactions from sequence, and describe how RFdiffusion-designed binders that read the IDP backbone rather than inducing secondary structure formation represent a novel strategy for targeting IDPs with nanomolar affinity — with potential applications as therapeutic agents and probes of IDP function in biomolecular condensates.

Conformational Selection

In our study of hemoglobin structure in the MWC model, we showed that hemoglobin exists in solution in two forms, the taut and relaxed forms, which are pre-existing and interconvertible even in the absence of dioxygen. Oxygen was presumed to bind preferentially to the relaxed form. In the KNF model, we observed that ligand binding can induce conformational changes in adjacent subunits, thereby promoting ligand cooperativity. These two models generally distill down to combinations of two simpler models. The first might be called conformational selection, in which the ligand binds tightly to a preexisting conformation in a "lock and key" manner without inducing subsequent macromolecular conformational change. Alternatively, the ligand might bind loosely and alter the macromolecular conformation to produce tighter binding, an example of the induced fit model. For the binding of dioxygen to hemoglobin, thermodynamic cycles could be drawn showing either the binding of the ligand and subsequent conformational changes in protein structure, or conformational changes in protein structure followed by binding. Is there additional evidence to support the conformational selection model of ligand binding to a protein that can exist in two conformations without a ligand? The answer is yes.

Antibodies are immune system protein molecules that can bind "foreign" molecules and target them for biological neutralization. Many crystal structures of antibodies have been determined in the presence or absence of a "foreign" ligand molecule. In these cases, the conformation of the bound antibody differs from that of the free antibody. An induced fit model for ligand binding or a lock-and-key model for ligand binding to one of two pre-existing antibody conformations could account for this observation. These different mechanisms could be differentiated experimentally using stop-flow kinetic techniques, since both exhibit slow and fast phases that are affected differently by ligand concentration. Theoretically, in the induced fit model, only one ligand type could bind to the antibody, which would undergo a conformational rearrangement to produce tighter binding. However, a different structural ligand might bind to the two main antibody conformations in the preexisting conformational models. James et al. have recently shown, using stop-flow kinetics (to investigate binding) and x-ray crystallography (to investigate final structures), that one antibody molecule can, by existing in two different preexisting conformations, bind two different ligands (antigens). One antibody conformation binds small aromatic molecules with low affinity (including the immunizing molecule, or hapten, 2,4-dinitrophenol). Then it rearranges to produce a high-affinity binding complex in which the DNP is bound in a narrow cavity (reducing the bound ligand's effective off rate (k_off). A second antibody conformation binds a ligand over a broad, flat binding site of the antibody molecule.

Lange et al. (2008), using an NMR technique, residual dipolar coupling, which allows sampling of structures on the microsecond time scale, have shown that the solution structure of ubiquitin (which we modeled in our first lab) in the absence of ligand exists as an ensemble of conformational states. More importantly, these different conformational states are identical to those found in the 46 crystal structures of ligands complexed to various protein ligands, strongly supporting the concept of conformational selection. In all likelihood, a combination of both induced fit and conformational selection probably occurs within a 3D energy landscape in which an initial binding encounter by either a lock-and-key fit to the "optimal fit" conformer or to a higher-energy conformer in which the bound state relaxes to a lower energy through the induction of shape changes in the binding protein.

Figure \(\PageIndex{1}\) shows a cartoon illustrating the differences between conformational selection and induced fit binding (after Boehr and Wright, Science 320, 1429 (2008)).

Figure \(\PageIndex{1}\): Conformational Selection vs Induced Fit Binding (after Boehr and Wright, Science 320, 1429 (2008))

Rea et al. proposed an interesting experimental model to distinguish between conformational selection and induced ligand binding. They studied rabbit ileal bile acid binding protein (I-BABP). The wild-type protein has a helix-turn-helix motif at its N-terminus. They produced a mutant (Δa-I-BABP) that replaced this motif with a Gly-Gly-Ser-Gly linker, leading to protein unfolding. Next, they conducted binding and folding studies on the addition of taurochenodeoxycholate (TCDC) using stopped-flow fluorescence to measure binding behavior. They wished to distinguish between two distinct mechanisms – folding before binding (or conformational selection) and binding before folding (or the induced-fit model). The data support a two-phase model. One phase did not depend on the ligand, whereas the other did (suggesting that binding was followed by a conformational change).

Conformational Selection

Equation\(\PageIndex{1}\) below describes the equilibria involved in the conformation selection model. The forward rate constants are shown as k_n, while the reverse ones are k_-n.

\begin{equation}
P \underset{k-1}{\stackrel{k_{1}}{\leftrightarrow}} P^{*}+L \underset{k_{-2}}{\stackrel{k_{2}}{\leftrightarrow}} P^{*} L
\end{equation}

P* in the conformational selection model represents a high-affinity, pre-existing protein conformation.

Induced Fit

Equation\(\PageIndex{2}\) below describes the equilibria involved in the induced fit model.

\begin{equation}
P+L \underset{k-1}{\stackrel{k_{1}}{\leftrightarrow}} P L \underset{k_{-2}}{\stackrel{k_{2}}{\leftrightarrow}} P^{*} L
\end{equation}

P* in the induced fit models results when a high ligand shifts the equilibrium to the right.

One way to differentiate these models is to examine the dependence of the different kinetic phases on ligand concentration. In the conformation selection model, the slow step is the formation of the high-affinity protein form, P*. The first slow step depends nonlinearly on L, while the second, fast step depends linearly. The data did not fit this model well.

\begin{equation}
\begin{aligned}
&k_{\text {slow }}=k_{-2}+\frac{k_{2}}{1+\frac{L}{\left(\frac{k_{-1}}{k_{1}}\right)}} \\
&k_{\text {fast }}=k_{-1}+k_{1} L
\end{aligned}
\end{equation}

In the induced fit model, the ligand binds to a low-affinity and perhaps unfolded form of the protein, which subsequently collapses to the bound form in a slow step.

\begin{equation}
\begin{aligned}
&k_{\text {slow }}=k_{-2}+\frac{k_{2} L}{\left(L+\frac{k_{-1}}{k_{1}}\right)} \\
&k_{\text {fast }}=k_{-1}+k_{1}[L]
\end{aligned}
\end{equation}

Both ligand-dependent and independent phases are evident in the equation for the slow step for the induced fit mechanism. At high ligand concentration (when L >> k_-1/k1), the slow step in the induced fit would be independent of ligand (k_slow = k_-2 + k₂). The authors state that the data are consistent with a variant of induced fit called the "fly casting model." In this model, the protein first encounters a ligand and forms a hydrophobic collapse intermediate (PL) in a fast step characterized by a linear dependence on ligand concentration. Then, the intermediate slowly interconverts into a wild-type-like complex through conformational rearrangement. Wild-type protein binds the ligand 1000x as quickly, suggesting entropic barriers to binding of the ligand to the unfolded state and rearrangement of the protein thereafter.

Junker et al. used atomic force microscopy (AFM) to observe the effects of ligand binding on the folding/unfolding fluctuations of a single calmodulin (CaM) molecule. This calcium-binding protein binds amphiphilic helical peptides, leading to a large conformational change in the protein. To do this, they sandwiched a single CaM molecule between filamins, which serve as attachment points for the AFM tip, and the surface. A slow pulling force was applied to the molecule, and the length gain was measured as the protein unfolded. The rapid fluctuations between folded and unfolded states were quantified and used to derive a complete energy landscape for CaM folding. They conducted these experiments in the presence of two ligands, Ca²⁺ and mastoparan (Mas), a wasp venom peptide. They found that Mas does not affect the folding rate of CaM, although it does stabilize the already folded form. This suggests that Mas does not bind to the transition state or the unfolded protein but rather selects a particular conformation from an ensemble of possible choices. Ca²⁺, however, increases the folding rate, which suggests that it stabilizes both the transition state and the folded state. AFM offers considerable precision in drawing energy landscapes of protein folding and unfolding, and it has several applications yet to be explored.

Binding to Intrinsically Disordered Proteins and MORFs

As described above, the binding of a protein to a ligand (including another protein) could occur by a lock-and-key mechanism, possibly through a conformational selection process or through an induced fit when a conformational rearrangement follows an initial binding event to form a more tightly bound complex. But how does binding to a completely intrinsically disordered protein (which has been documented) occur? These cases are excluded from those envisioned by simple induced-fit mechanisms. Binding to IDPs might occur through specific Molecular Recognition Features (MoRFs).

MoRFs are typically contiguous but disordered protein sections that first encounter a binding partner (a protein, for example). Mohan et al. conducted a structural study of MoRFs in the Protein Data Bank by selecting short regions (less than 70 amino acids) from mostly disordered proteins bound to larger proteins (>100 amino acids). They chose a sequence length of 70 amino acids or fewer, since these sequences are most likely to display conformational flexibility before binding to a target. 2512 proteins fit their criteria. For comparison, they created a similar database of ordered monomeric proteins. The analysis showed that after they encounter a binding surface on another protein, the MoRF would adopt or "morph" into several types of new conformations, including alpha-helices (a-MoRFs), beta-strands (b-MoRFs), irregular strands (i-MoRFs), and combined secondary structure (complex-MoRFs), as shown in the figure below.

Figure: Types of Molecular Recognition Features in Intrinsically Disordered Proteins

Figure \(\PageIndex{2}\): Types of Molecular Recognition Features in Intrinsically Disordered Proteins. (A) α-MoRF, Proteinase Inhibitor IA3, bound to Proteinase A (PDB entry 1DP5). (B) A β-MoRF, viral protein pVIc, bound to Human Adenovirus 2 Proteinase (PDB entry 1AVP). (C) An ι-MoRF, Amphiphysin, bound to α-adaptin C (PDB entry 1KY7). (D) A complex-MoRF, β-amyloid precursor protein (βAPP), bound to the PTB domain of the neuron specific protein X11 (PDB entry 1X11). Partner interfaces (gray surface) are also indicated. Vacic, V. et al. Journal of Proteome Research 6, 2351 (2007). Permission from Copyright Clearance Center's Rightslink /American Chemical Society

Figure \(\PageIndex{8}\) shows interactive iCn3D models of the types of molecular recognition features in intrinsically disordered proteins

(A) α-MoRF, Proteinase Inhibitor IA3, bound to Proteinase A (1DP5) (Copyright; author via source). Click the image for a popup or use this external link: https://structure.ncbi.nlm.nih.gov/i...tk45aWAeMrTct7	(B) A β-MoRF, viral protein pVIc, bound to Human Adenovirus 2 Proteinase (1AVP) (Copyright; author via source). Click the image for a popup or use this external link:https://structure.ncbi.nlm.nih.gov/i...EFmanRy2T7zjX6
(C) An ι-MoRF, Amphiphysin, bound to α-adaptin C (1KY7) (Copyright; author via source). Click the image for a popup or use this external link: https://structure.ncbi.nlm.nih.gov/i...JEBC4VHUA9C718	(D) A complex-MoRF, β-amyloid precursor protein (βAPP), bound to the PTB domain of the neuron specific protein X11 (1X11) (Copyright; author via source). Click the image for a popup or use this external link: https://structure.ncbi.nlm.nih.gov/i...XtQ7retidnksT9

Vacic et al. have further characterized the binding interfaces between MoRFs and their binding partners using structural data from the PDB. Interfaces were studied by determining differences in accessible surface area between MoRFs and their binding partners, and between the protein in unbound states. These were compared to ordered protein complexes, including homodimers and antibody-protein antigen interactions not characterized by disordered interactions. Their findings are summarized below.

• MoRF interfaces have more hydrophobic groups and fewer polar groups than the monomer surface. This is true even as the overall amino acid composition of intrinsically disordered proteins is enriched in polar amino acids, leading them to adopt various unfixed solution conformations.
• a-MoRFs have few prolines, which is expected as prolines are helix breakers.
• Methionine is enriched in both MoRFs and in their binding partner interface. Methionine is unbranched, flexible, and contains sulfur, which is large and polarizable, making it an ideal side chain for participating in London dispersion forces in a hydrophobic environment.
• Even though MoRFs have few residues, their binding interfaces are similar in size to or larger than those of other protein binding interfaces, and this also applies to IDPs as a whole. MoRF interfaces also exhibit a larger solvent-exposed surface area, similar to that of IDPs. This is consistent with the notion that MoRFs are disordered before binding and that a defined structure is impossible with little buried surface area.
• As MoRFs have a significant nonpolar character within an IDP highly enriched in polar amino acids, MoRFs should be highly predictable by search algorithms.

Recently, a new program, FINCHES-online, lets you predict interactions between disordered regions using sequence as input. It uses a new "coarse-grained" force field (such as those used in molecular dynamics simulations) to predict interactions between the IDR of a protein and target proteins. As surmised from the above discussions, three types of interactions of an IDR binder with a target protein are possible:

1. IDR binder + Folded Domain Target ↔ Folded IDR:Folded Domain Target Complex (i.e., folding on binding)
2. IDR binder + Folded Domain Target ↔ IDR:Folded Domain "Fuzzy Complex" (i.e., the IDR remains unfolded on binding)
3. IDR binder + IDR target ↔ IDR:IDR (i.e., both binding and target remain disordered.

The program is currently best suited for category 3.

Recent Updates:  9/16/25

Designing Binders for IDPs

Machine learning and AI approaches, including RFdiffusion, have been used to design a library of protein-binding pockets for IDPs, particularly IDRs.  The binders were designed to interact with peptide backbones and then refined to maximize interactions with side chains.  The binders often display nanomolar binding affinity (K_D around 1 nM).  The IDRs are induced upon binding to adopt conformations that fit the binding pocket.  One example is the IDP and neuropeptide dynorphin A.  Its main physiological target is the kappa-opioid receptor (K_D around 200 nM), but it can also bind other targets as it is an IDP. It is bound in an extended form to the synthesized binder with an even lower Kd (around 1 nM).  Figure \(\PageIndex{9}\) is an interactive iCn3D model showing the interactions of the IDP dynorphin A with the synthetic binder (9CCE).  The binder, an alpha-helical bundle, is shown in magenta, and the dynorphin A peptide backbone is shown in cyan.  The noncovalent side chain interactions between the two are shown in dotted lines between the side chains (in sticks).

Figure \(\PageIndex{9}\): Interactions of the IDP dynorphin A with the synthetic binder (9CCE).  (Copyright; author via source). Click the image for a popup or use this external link: https://www.ncbi.nlm.nih.gov/Structu...1be4e15ad2f3ec

The binding pocket is an elongated groove to which the IDP binds in an extended conformation.  On binding, the IDP is not induced to form secondary structure. Rather, it is the IDP backbone that is "read" by the binding pocket.  This binding mode is ideal for IDRs that lack a specific secondary structure.  Note in the iCn3D model that many of the hydrogen bonds between dynorphin A and the binder are between peptide bonds in the IDR to side chains in the binder.  The hydrogen bonds are often "bidentate," involving hydrogen bonds between the amide H and the carbonyl O of the peptide.  Additional specificity arises from interactions between the side chain of the IDR and the binder.

Binders can be designed and synthesized that fit many different types of sequences, with varying lengths and polarities.  There are many more disordered states for the IDR than highly ordered ones (such as all alpha helical). The binder forces the IDR peptide into an extended state, so there is an induced-fit aspect to binding.  The IDR can bind in an optimal extended conformation to provide a complementary match to the binding groove in the binder.  Hence, a binder could be made that specifically recognizes a given IDR.  Designed binders might eventually be used as drugs, similar to antibodies, and as probes for the normal functions of IDFs.  For example, biomolecular condensates are often enriched in proteins with IDRs.

Summary

(Summary written by Claude, Sonnet 4.6, Anthropic)

This chapter addresses the mechanistic basis of protein-ligand binding beyond simple equilibrium thermodynamics, focusing on how conformational heterogeneity in proteins and disorder in ligands shape the binding process. Two paradigms — conformational selection and induced fit — are developed, experimentally tested, and extended to the special case of intrinsically disordered proteins.

Conformational selection posits that proteins exist in solution not as a single structure but as an ensemble of interconverting conformations, and that ligand binding occurs preferentially to a pre-existing high-affinity conformer rather than inducing a structural rearrangement. This view is supported by the MWC model of hemoglobin, which posits that the T and R states preexist in the absence of oxygen. More direct evidence comes from three experimental systems. First, a single antibody molecule binds two structurally distinct antigens through two pre-existing conformations — one with a narrow cavity that captures small aromatic molecules with high affinity after initial low-affinity encounter and conformational rearrangement, and one with a flat surface suited for a different ligand — demonstrating that conformational diversity enables functional versatility. Second, NMR residual dipolar coupling measurements of ubiquitin in the absence of any ligand reveal an ensemble of conformational states that are essentially identical to those seen in 46 different crystal structures of ubiquitin bound to diverse protein ligands — the most direct experimental confirmation that bound conformers pre-exist in solution before the ligand arrives. Third, single-molecule AFM force spectroscopy on calmodulin distinguishes between the two mechanisms: mastoparan (a wasp venom peptide) stabilizes only the folded state without altering the folding rate, indicating it selects from the existing ensemble (conformational selection); Ca²⁺ increases the folding rate, indicating it stabilizes the transition state and promotes folding, a feature of induced fit.

Induced fit describes binding to a single lower-affinity conformation followed by a conformational rearrangement that tightens the complex. In reality, most binding events probably combine elements of both mechanisms within a three-dimensional energy landscape: an initial encounter (by either lock-and-key to the optimal conformer or to a higher-energy conformer) followed by relaxation to the lowest-energy bound state. For partially unfolded proteins, the fly casting model represents a variant of induced fit in which the disordered protein rapidly collapses onto the ligand through initial hydrophobic contacts (a fast, linearly ligand-concentration-dependent phase), then slowly reorganizes into a well-defined bound conformation (a slow, ligand-independent phase at high [L]). This was demonstrated for rabbit ileal bile acid binding protein (I-BABP), in which a mutant lacking the N-terminal helix-turn-helix adopts an unfolded conformation that still binds taurochenodeoxycholate — though ~1000-fold more slowly than wild-type — consistent with entropic barriers to binding and subsequent folding in the disordered state.

Binding to intrinsically disordered proteins requires a conceptual extension of both models. IDPs, which lack stable tertiary structure due to their high polar and charged amino acid content, nonetheless bind specific partners with defined specificity and often nanomolar affinity through Molecular Recognition Features (MoRFs) — short disordered segments (~<70 amino acids) that adopt secondary structure upon encountering a binding partner. Statistical analysis of the PDB reveals four types: α-MoRFs (forming helices upon binding), β-MoRFs (forming strands), ι-MoRFs (forming irregular structures), and complex-MoRFs (forming combined secondary structures). Despite the polar composition of IDPs overall, MoRF binding interfaces are enriched in hydrophobic residues — consistent with the thermodynamic principle that the hydrophobic effect drives high-affinity protein-protein interactions — and in methionine, whose unbranched, flexible, large-sulfur side chain is ideally suited to London dispersion interactions in hydrophobic environments. MoRF interfaces are also unusually large, with solvent-exposed surface areas consistent with disorder before binding. Together, these properties make MoRFs computationally tractable for prediction, and a new tool (FINCHES-online) uses coarse-grained force fields to predict IDR:IDR and IDR:folded domain interactions from sequence alone.

A recent and striking advance is the de novo design of synthetic binders for IDPs using RFdiffusion-based machine learning. Rather than inducing a secondary structure in the IDP, the designed binders present an elongated groove that reads the peptide backbone directly via hydrogen bonds to amide NH and carbonyl O groups, with additional specificity conferred by side chain interactions. This "backbone reading" mode is particularly powerful for IDRs that lack a canonical secondary structure preference. The IDP neuropeptide dynorphin A, normally binding the kappa-opioid receptor with KD ~200 nM, is bound in an extended conformation by a designed α-helical bundle binder with KD ~1 nM — a 200-fold affinity improvement. This demonstration that IDPs can be targeted with high specificity by designed binders opens a new frontier in drug development and chemical biology, with potential applications as therapeutic agents and as probes of IDP function in biomolecular condensates, transcription factor complexes, and signaling networks.