Fast Association and Slow Transitions in the Interaction between Two Intrinsically Disordered Protein Domains*

Background: Intrinsically disordered proteins are common regulators of protein-protein interactions, but little is known about their mechanisms of interaction. Results: Two intrinsically disordered protein domains, from ACTR and CREB-binding protein, interact through rapid association and slow conformational changes. Conclusion: Electrostatics governs the fast association, but the overall reaction is multistep. Significance: The slow conformational search may be common among intrinsically disordered proteins with many binding partners. Proteins that contain long disordered regions are prevalent in the proteome and frequently associated with diseases. However, the mechanisms by which such intrinsically disordered proteins (IDPs) recognize their targets are not well understood. Here, we report the first experimental investigation of the interaction kinetics of the nuclear co-activator binding domain of CREB-binding protein and the activation domain from the p160 transcriptional co-activator for thyroid hormone and retinoid receptors. Both protein domains are intrinsically disordered in the free state and synergistically fold upon binding each other. Using the stopped-flow technique, we found that the binding reaction is fast, with an association rate constant of 3 × 107 m−1 s−1 at 277 K. Mutation of a conserved buried intermolecular salt bridge showed that electrostatics govern the rapid association. Furthermore, upon mutation of the salt bridge or at high salt concentration, an additional kinetic phase was detected (∼20 and ∼40 s−1, respectively, at 277 K), suggesting that the salt bridge may steer formation of the productive bimolecular complex in an intramolecular step. Finally, we directly measured slow kinetics for the IDP domains (∼1 s−1 at 277 K) related to conformational transitions upon binding. Together, the experiments demonstrate that the interaction involves several steps and accumulation of intermediate states. Our data are consistent with an induced fit mechanism, in agreement with previous simulations. We propose that the slow transitions may be a consequence of the multipartner interactions of IDPs.

and flattened energy landscapes, resulting in the absence of a well defined three-dimensional structure at physiological conditions when unbound, but they often undergo a coupled folding and binding event when interacting with their ligands. IDPs have important roles in various critical cellular regulatory processes, for instance, in signaling, transcription, cell cycle control, and translation (3,4). The abundance in the proteome, together with their functional importance, and frequent association with different types of diseases, such as cancer and neurodegenerative disorders (5), have recently sparked a momentous interest in IDPs, which calls for a better understanding of their structural, thermodynamic, and kinetic properties (6,7).
The disordered state of IDPs has been suggested to give them certain advantages, such as the possibility of having numerous binding partners. In fact, for many IDPs, distinct structures are adopted when bound to different targets (4, 8 -10). It has also been theorized (11) that IDPs have a larger capture radius than ordered proteins, which would then allow for a higher association rate.
One of the most comprehensive studies on the kinetics of IDP-target interactions was conducted by Wright and colleagues (12) a few years ago, where they used NMR relaxation dispersion experiments to investigate the coupled folding and binding of the intrinsically disordered pKID from the cAMPregulated transcription factor (CREB) to the three-helix bundle KIX domain from the co-activator CREB-binding protein. They showed that pKID binds KIX through an induced fit mechanism and were able to identify and characterize intermediate states along the binding reaction pathway, providing key insights into the mechanisms of molecular recognition. However, despite the growing identification of proteins that are intrinsically disordered (13), remarkably few experimental studies on the binding kinetics involving IDPs have been reported that would provide answers on the mechanisms that these proteins utilize in the interaction with their targets.
In this work, we have investigated the kinetics of the specific interaction between the nuclear co-activator binding domain (NCBD) of CREB-binding protein and the activation domain from the p160 transcriptional co-activator for thyroid hormone and retinoid receptors (ACTR) (see Fig. 1). Both are intrinsically disordered in the free state and synergistically fold upon complex formation to form a well folded structure (8) with a nanomolar dissociation constant (K d ).
NCBD has many of the characteristics of a molten globule, whereas ACTR is completely disordered (8, 14 -16). A backbone NMR relaxation study (15) showed that although both ACTR and NCBD exhibit substantial flexibility on the pico-to nanosecond time scale, both proteins displayed restricted backbone motions in the bound state. This results in a significant unfavorable conformational entropy change for binding, which is also reflected in the total entropy change upon complex formation, obtained from isothermal titration calorimetry (15). Clearly, disorder is important in modulating the binding free energy. However, characterization of the binding kinetics is an essential part in the elucidation of the binding mechanism of the interaction between NCBD and ACTR. Therefore, to shed light on the binding mechanism, we have performed fluorescence-based binding kinetic experiments. We show that the initial association between NCBD and ACTR is fast but that subsequent slow conformational changes are necessary to reach the most stable bound ground state.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The DNA sequence of human ACTR (residues 1018 -1088) was purchased from GENEART (Germany), whereas human NCBD (2058 -2116) (8) was PCR-amplified using a human brain cDNA library as template and inserted into a modified pRSET vector (Invitrogen). The final construct was made up of an N-terminal hexahistidine-tagged lipoyl fusion protein followed by a thrombin cleavage site (LVPRGS) and finally the ACTR or NCBD sequence. Mutants were generated by inverted PCR. Escherichia coli BL21(DE3) pLysS cells (Invitrogen) were grown in 2ϫ TY medium at 37°C and then induced with 1 mM isopropyl-␤-D-thiogalactopyranoside when A 600 reached 0.7-0.8 to overexpress the fusion protein at 18°C overnight. Cells were lysed by sonication followed by centrifugation at 4°C, after which the supernatant was passed through a 0.2-m filter (Sarstedt), and then loaded onto a nickel-Sepharose fast flow (GE Healthcare) column. After washing the column with binding buffer (40 mM Tris pH ϭ 8.0, 500 mM NaCl, 20 mM imidazole), the His-tagged fusion protein was eluted using buffer containing 250 mM imidazole. The fusion protein was then dialyzed against 20 mM Tris pH ϭ 8.0, 120 mM NaCl, after which the lipoyl protein was cleaved off using thrombin (GE Healthcare) and separated from ACTR or NCBD by loading the solution onto the nickel-Sepharose fast flow column. The flow-through, containing ACTR or NCBD, was subjected to a reversed phase chromatography step, using C-8 (ACTR or NCBD) or C-18 (NCBD only) columns (Grace Davison Discovery Sciences). The identity of purified ACTR or NCBD was verified by MALDI-TOF mass spectrometry.
The concentration of NCBD Y2108W and NCBD WT was determined by measuring the Trp and Tyr absorbance at 280 nm, respectively. For wild-type ACTR (ACTR WT ), which does not contain any Trp or Tyr, the concentration was determined by measuring the absorbance at 205 nm using an extinction coefficient obtained from amino acid analysis.
Stopped-flow Measurements-The kinetics of NCBD/ACTR association was characterized using an upgraded SX-17MV stopped-flow spectrometer (Applied Photophysics, Leatherhead, UK). Measurements were performed at T ϭ 277 or 283 K, in 20 mM sodium phosphate (pH ϭ 7.4), 150 mM NaCl. Stopped-flow experiments at high salt conditions were carried out in 20 mM sodium phosphate (pH ϭ 7.4), 0.93 M NaCl, whereas binding kinetic measurements at high trimethylamine N-oxide (TMAO) concentrations were performed in 20 mM sodium phosphate (pH ϭ 7.4), 150 mM NaCl, 1 M TMAO. Excitation was at 280 nm, and the change in fluorescence upon binding was monitored using a 320-nm long-pass cutoff filter. Association rate constants (k on Circular Dichroism Spectroscopy-CD spectra were recorded using a JASCO-810 spectropolarimeter equipped with a Peltier temperature control system. A cuvette with a path length of 1 mm was used, and far-UV spectra were recorded at T ϭ 298 K, from 260 to 200 nm with a scan speed of 50 nm/min, and a 2-s response time. Sample conditions were 10 -23 M protein in 20 mM sodium phosphate (pH ϭ 7.4), 150 mM NaCl. All spectra were corrected for the contribution from the buffer. CD-monitored thermal denaturation was performed by following the signal at 222 nm, using a scan speed of 1 K/min.
Equilibrium Fluorescence Measurements-Equilibrium measurements were carried out on an SLM 4800 spectrofluorometer (SLM instruments). Experiments were performed in 20 mM sodium phosphate (pH ϭ 7.4), 150 mM NaCl. For binding experiments between NCBD and ACTR, Trp excitation was at 280 nm, and emission spectra were recorded from 300 to 400 nm, whereas for 8-anilino-1-naphthalenesulfonic acid (Sigma-Aldrich) fluorescence, experiments were performed at T ϭ 298 K, excitation was at 350 nm, and the fluorescence emission was recorded from 400 to 662.5 nm.

Design and Validation of Tryptophan Variants of NCBD and
ACTR-Trp residues greatly facilitate the use of fluorescencebased methods to study the kinetics of binding with high sensitivity. However, neither NCBD nor ACTR contain any Trp. Early attempts were made to see whether the fluorescence of the sole tyrosine, Tyr-2108, in NCBD could be used to monitor the binding to ACTR WT . However, only a small fluorescence change could be observed with the stopped-flow technique, and a high concentration of NCBD (Ͼ10 M) was needed to obtain reliable observed rate constants. We therefore performed a screening where we made single amino acid substitutions, replacing a certain residue with a Trp at different locations in NCBD. These Trp variants where then evaluated on the basis of binding kinetics and of their free and bound state behavior using CD spectroscopy and equilibrium fluorescence measurements to determine which of these engineered Trp variants would be most suitable as a model for wild-type NCBD (NCB-D WT ). Out of these, the replacement of Tyr with Trp at position 2108 (NCBD Y2108W ) resulted in an NCBD variant that exhibited the largest fluorescence change upon binding of ACTR WT (Figs. 1 and 2A). The apparent dissociation rate constant, k off app , as well as its temperature dependence (data not shown), as determined by displacement experiments, was the same for NCBD Y2108W /ACTR WT and NCBD WT /ACTR WT ( Table 1). The CD spectrum of NCBD Y2108W was very similar to that of NCBD WT in terms of both shape and magnitude of the CD signal (Fig. 3A). The high similarity of the CD properties is also extended to the NCBD Y2108W /ACTR WT and NCBD WT / ACTR WT complexes (Fig. 3B). Furthermore, CD-monitored thermal denaturation of NCBD Y2108W showed an apparent noncooperative transition similar to that of NCBD WT (Fig. 3C) and in good agreement with previous studies (14,16). The thermal denaturations of the bimolecular complexes, NCBD Y2108W /ACTR WT and NCBD WT /ACTR WT , respectively, show that both have a clear transition, with a melting temperature of around 52°C, reflecting well folded structures of the complexes (Fig. 3D). In addition, both NCBD Y2108W and NCB-D WT bind the fluorescent hydrophobic probe, 8-anilino-1-naphthalenesulfonic acid, resulting in an increase in fluorescence intensity and with a blue-shifted emission, which is characteristic for proteins with molten globular properties. Finally, kinetic binding experiments using high concentrations of NCBD WT and monitoring Tyr fluorescence suffered from a low signal-to-noise ratio, but were consistent with those of NCBD Y2108W . Taken together, NCBD Y2108W proved to be a good pseudo-wild-type of NCBD WT and was therefore subjected to a detailed kinetic study.
We also inserted Trp residues in ACTR, just prior to helix 1 (ACTR Q1042W ) or in the C-terminal helix (ACTR L1076W ), to probe different regions of the ACTR/NCBD complex. The determined apparent association rate constants for ACTR Q1042W /NCBD WT and ACTR L1076W /NCBD WT , at 277 K, were very similar to those of NCBD Y2108W /ACTR WT (Table 1), which further validates the use of NCBD Y2108W as a representation of NCBD WT .
Binding Kinetics for NCBD and ACTR-We measured the kinetics of association of NCBD/ACTR using the stopped-flow technique and by varying the concentration of ACTR WT and monitoring the fluorescence change of NCBD Y2108W . Alternatively, NCBD WT was varied at a constant concentration of ACTR Q1042W or ACTR L1076W , respectively. Because of the fast binding kinetics, the experiments were performed at low temperatures, 283 and 277 K, to accurately determine the observed rate constant, k obs .
The binding kinetics of NCBD Y2108W /ACTR WT was biphasic with a fast phase with positive amplitude and a slow phase with negative amplitude (Fig. 4A) at 20 mM phosphate (pH ϭ 7.4), 150 mM NaCl. The concentration dependences of both kinetic phases were analyzed (Fig. 4B). The fast phase increased linearly with ACTR WT concentration with a slope of 3 ϫ 10 7 M Ϫ1 s Ϫ1 , which is the apparent association rate constant at 277 K ( Table 1). The slow phase remained rather constant with ACTR WT concentration with a rate constant ϳ1.2 s Ϫ1 .
The apparent overall dissociation constant, k off app , was 2.6 s Ϫ1 at 277 K (Fig. 5), as determined in a displacement experiment. In a multistep binding reaction, this experimental parameter is a function of all first order rate constants, and it is normally equal to or smaller than the lowest microscopic off-rate constant on the reaction pathway. Biphasic dissociation kinetics was observed for the complex between ACTR L1076W and NCBD WT , i.e. when the Trp was placed in the C-terminal helix of ACTR, with k off app ϭ 3.3 and 0.5 s Ϫ1 , respectively.
The initial association of NCBD and ACTR is thus very rapid, and the apparent k on app is among the fastest that has been determined so far for an IDP system and the first to be characterized for a system where both binding partners are IDPs. A . NCBD is shown in cyan and ACTR is in green, with side chains of NCBD Y2108 , ACTR L1076 , and ACTR Q1042 shown in red. These residues were mutated to Trp in this study. Also shown is the buried salt bridge formed between ACTR D1068 (red spheres) and NCBD R2104 (blue spheres).
simple extrapolation using the values of k on app , determined at 277 and 283 K to the physiological temperature 310 K, shows that the binding reaction approaches the diffusion-controlled limit, with a k on ϳ10 9 M Ϫ1 s Ϫ1 . The temperature dependences of k off app for NCBD Y2108W /ACTR WT , NCBD WT /ACTR WT , NCBD WT /ACTR L1076W , and NCBD WT /ACTR Q1042W are all similar. The k off app at physiological temperature (310 K) was determined to be around 130 s Ϫ1 for NCBD WT /ACTR WT .
We also performed measurements at high salt concentration to investigate the role of electrostatics in the binding reaction. At 0.9 M NaCl, an additional kinetic phase was detected. This phase was rather constant with ACTR WT concentration (within the range it could be accurately fitted) with a value of ϳ40 s Ϫ1 (Fig. 6). Interestingly, both the fast and the slow phases were not significantly affected by the addition of salt (2 ϫ 10 7 M Ϫ1 s Ϫ1 and 0.9 s Ϫ1 , respectively). This suggested that the salt affected a particular step in the binding reaction that was too fast to detect at low salt. We note, however, that we could not detect the 40 s Ϫ1 phase when an excess of NCBD Y2108W was mixed rapidly with ACTR WT . The reason might be that the higher total fluorescence in this experiment obscured the phase.
Binding Kinetics of Salt Bridge Mutants-The importance of a highly conserved and buried salt bridge in the NCBD/ACTR complex, formed between Asp-1068 in ACTR and Arg-2104 in NCBD, has been the subject of an experimental mutational analysis by Wright and colleagues (14), where both Asp-1068 and Arg-2104 were mutated to leucine, thus replacing the salt bridge with a hydrophobic interaction. Further, in a recent molecular dynamics simulation study (17), it was concluded that this salt bridge stabilizes an on-pathway intermediate toward the bound state.
is the fluorescence signal, B is its total amplitude, C is its intercept value, and [ACTR] 0 and [NCBD] 0 are the respective total concentrations of the ACTR and NCBD variants. Fitting was performed using KaleidaGraph (Synergy Software). As seen in A, the binding between NCBD Y2108W and ACTR WT is stoichiometric, which precludes a reliable and accurate determination of the dissociation binding constant, K d . This is in good agreement with a previous study, which determined the K d to be 34 nM at 304 K, using ITC (8).  To test whether the 40 s Ϫ1 phase detected at high salt concentration corresponded to formation of this salt bridge, we made the mutant D1068L in ACTR WT and R2104L in NCBD Y2108W and performed binding experiments at high and low salt. All three kinetic phases were affected by the D1068L/ R2104L mutations. In addition, the fluorescence properties were modulated such that the sign of the amplitude for the fast phase turned negative (Fig. 7). At low salt, the concentration dependence of the first phase yielded a k on app value of 1.5 ϫ 10 6 M Ϫ1 s Ϫ1 , i.e. 20-fold lower than that of the wild-type domains ( Table 1). The second phase 2 was now clearly hyperbolic, with a 2 max value of around 20 s Ϫ1 (Fig. 7 and Table 2). The presence of the slow phase 3 was not clear. It showed a decrease (0.2 s Ϫ1 ), along with a small amplitude, such that it approached a phase related to photobleaching. Further, fitting of all three phases simultaneously yielded a k Ϫ3 value that was much lower than k off app from a separate displacement experiment (3 s Ϫ1 ). On the other hand, fitting of a simpler two-step scheme to the salt bridge mutant gave parameters that were consistent with both k off app and overall K d value ( Table 2, Figs. 2B and 7). At 0.9    3 . The dependence of the three observed rate constants on ACTR WT concentration was globally fitted to a four-state sequential binding mechanism model (as illustrated in Fig. 8) to estimate the microscopic rate constants, using the Prism software (GraphPad). The analytical solution to the four-state model is known and has been described in detail by Chemes et al. (26). A, data fitted to a model that involves initial binding and two on-pathway intermediate states (Fig. 8, Scheme 1). Best fit gives k 1 ϭ 1.
, data fitted to a model involving the selection of a binding-competent species followed by binding and an intermediate state (Fig. 8, Scheme 2). Best fit gives The standard errors from the fits are large for several of the rate constants, in particular k 3 and k Ϫ3 , due to the lack of concentration dependence of the slower phases, 2 and 3 . Data were fitted to a double exponential function, and the residuals from the fit are shown below the trace. Global fit of the dependence of the two experimentally observed rate constants on ACTR D1068L concentration was performed to obtain the microscopic rate constants. In C, a model describing a two-step induced fit mechanism was employed to fit the data (panel B, Scheme 1). D, a conformational selection model (panel B, Scheme 2) was used to fit the data. The fitting was restricted by reducing the number of free variables through the use of a dissociation binding constant K d ϭ 5.4 M, which was determined in a separate equilibrium binding experiment (Fig. 2B). See Table 2 for the estimated microscopic rate constants for the two models. The fitting was performed using the Prism software (GraphPad).
M NaCl, the amplitude of the fast phase 1 decreased such that a quantitative analysis was difficult. However, we estimated the slope of the phase, k on app , to 0.2 ϫ 10 6 M Ϫ1 s Ϫ1 , showing that other residues than Asp-1068 in ACTR WT and Arg-2104 in NCBD Y2108W influence binding electrostatics.
It was shown previously that NCBD R2104L displays a more cooperative urea denaturation than NCBD WT (14), with an unfolding transition that occurs at about 1 M higher urea concentration. Furthermore, the magnitude of the CD signal at 222 nm of the unbound NCBD R2104L corresponded to a 25% increase in helix content when compared with NCBD WT , a result that was corroborated by NMR experiments (14); thus, the ground state structures are different for NCBD WT and NCBD R2104L , which complicates the kinetic analysis. Although the kinetic analysis of the D1068L/R2104L double mutation is complex, it is clear that these charged residues contribute to the high association rate constant observed for the binding of ACTR and NCBD. The most likely explanation for the appearance of the middle phase 2 is slowing down of an intramolecular step that is facilitated by formation of the salt bridge because the phase appears both at high salt (with wild-type domains) as well as upon mutation of the salt bridge.
The Order of Events-The fact that we observe more than one phase in stopped-flow experiments means that the interaction between NCBD and ACTR is not a simple one-step reaction. In a multistep reversible reaction, such as the one under study, each k obs value is a complex function of all microscopic rate constants. Therefore, kinetic phases often cannot be directly assigned to a certain step. For example, the slow phase, 3 , is dependent not only on k 3 and k Ϫ3 but also on the other first order rate constants in Fig. 8, Scheme 1. Nevertheless, the order of events may be inferred or demonstrated using different techniques.
Our data are consistent with a binding mechanism that involves at least two intermediate states, as schematically shown in Fig. 8, Scheme 1. However, the observed rate constants are also consistent with an initial conformational change (Fig. 8, Scheme 2, and see also Fig. 6). If we consider the slow step as the initial one (Scheme 2), the best fit gives rate constants of ϳ1.1 and 0.5 s Ϫ1 for k 1 and k Ϫ1 , respectively, resulting in equilibrium concentrations of 31% of NCBD and 69% as NCBD*. The amplitude of the slow phase (negative) is 10 -20% of the fast phase. Thus, if the slow phase is due to an initial conformational transition between NCBD and NCBD*, then the fluorescence yield of NCBD must be much larger than the fluorescence yield of NCBD*. We rule out an initial slow step (Scheme 2) because (i) the large expected difference in fluorescence yield between the NCBD conformers is unlikely. Further, in kinetic experiments where NCBD was mixed with TMAO, we observed no slow phase, which would be expected if Scheme 2 applies; (ii) mutations in both ACTR and NCBD modulate the magnitude of the slow phase; (iii) signals in the heteronuclear single quantum coherence spectra are significantly broadened at lower temperatures (15,16), suggesting exchange between NCBD conformers on the s-ms time scale, much more rapid than the observed slow phase; and (iv) the biphasic dissociation of ACTR L1076W /NCBD WT suggests that there are two conformational transitions in the complex because ACTR is completely disordered.
The 40 s Ϫ1 phase detected at 0.9 M NaCl is likely related to formation of the buried salt bridge. At low salt, this step would be rapid and not visible. With the D1068L/R2104L mutations, the 40 s Ϫ1 phase is replaced by a hyperbolic phase saturating at 20 s Ϫ1 , in agreement with the experiment in 0.9 M NaCl. On a kinetic basis, this phase may be related to a conformational transition occurring before or after binding. However, for the wild-type domains, the presence of 1 M TMAO affects the offrate constant significantly and only marginally affects k on app ( Table 3). If the conformational changes occurred before binding, TMAO should affect the fast equilibrium and thus the association rate constant because TMAO selectively stabilizes more ordered structures. The lack of effect on k on also suggests that native interactions are formed late in the binding reaction (18,19). Thus, we propose that the two non-concentration-dependent phases we observe are related to steps occurring after initial binding (induced fit).

DISCUSSION
IDPs may assume different structures when interacting with different ligands (20). A clear example of this structural plastic- FIGURE 8. Schemes that were used to quantitatively describe our kinetic data. Scheme 1 involves two productive on-pathway intermediates along the binding reaction, whereas in Scheme 2, there is an initial selection of a binding-competent species of NCBD followed by the formation of an on-pathway intermediate. PDB accession codes 2KKJ and 1KBH were used for the structural models of unbound NCBD WT and the NCBD WT /ACTR WT complex, respectively. ACTR WT in the free state is completely disordered, and the schematic model of ACTR WT is shown only to visualize such a state.  ity is in fact NCBD, which adopts a three-dimensional structure in its interaction with interferon regulatory factor 3 (IRF-3) (9) that has a different topology when compared with the structure when bound to ACTR (8). There are also other examples where IDPs adopt distinct structures with different ligands (4,10). Further, in a recent NMR study, Poulsen and colleagues (16) were able to determine the three-dimensional structure of free state NCBD and showed that the conformer, which was longlived enough to be structurally characterized by NOEs, is very similar to that of NCBD when bound to ACTR or to the transactivation domain of p53 (21), suggesting that a conformational selection mechanism is taking place. These experiments raise the question about what binding mechanisms that are involved in the recognition processes for IDPs.
Here, we directly address the reaction mechanism for the interaction between NCBD and ACTR. Our kinetic data on NCBD/ACTR demonstrate that slow conformational transitions occur after an initial rapid binding, in agreement with NMR studies and simulation on another IDP system, pKID/ KIX (12,22). Such binding mechanism is consistent with the observed structural plasticity of IDP complexes, where disordered regions search the most stable conformation with specific interactions after the initial encounter event. The presence of intermediate states is also in agreement with recent molecular dynamics simulation studies (17,23) on NCBD/ACTR, where it was suggested that productive on-pathway intermediates may arise through two different pathways. One intermediate was formed by docking of the C-terminal helices and stabilized by the highly conserved and buried salt bridge between Asp-1068 in ACTR and Arg-2104 in NCBD, whereas formation of the second intermediate was found to be initiated by interactions between the N-terminal helices.
Although it is very difficult to experimentally distinguish binding reaction mechanisms involving several steps, we can say that the simplest mechanism that is consistent with the observed kinetics is one with three consecutive steps, most probably with two conformational changes occurring after binding, as depicted in Fig. 8, Scheme 1. The parallel pathways for initial binding suggested by the molecular dynamics simulations is neither confirmed nor ruled out by our data. For example, the observed small differences in the apparent k on values for variants with different Trp probes (Table 1) may reflect parallel pathways for the initial encounter but could equally well be explained by mutational effects in a consecutive binding mechanism. Our data also do not rule out very rapid conformational transitions in NCBD as suggested by NMR experiments (15,16). However, if the initial binding of ACTR is to a high energy conformer of NCBD, the microscopic k on must be higher than the observed k on (which is already very high) due to the conformational selection. Thus, it is possible that there are unbound NCBD conformers with preformed binding-competent elements that are similar to those in the bound state NCBD. The binding reaction would then involve subsequent induced fit steps, as indicated by recent molecular dynamics simulations studies (17), where the authors argued that both conformational selection and induced fit may in fact be in operation in the interaction between NCBD and ACTR, as suggested as a general mechanism (24). Although we cannot rule out a fast conformational selection in the binding reaction, we can say that observable rate-limiting step(s) follow the induced fit mechanism.
There are indications from previous NMR studies that different bound species that are in exchange could be present, as observed in the current study. A backbone NMR relaxation study (15) showed that although pico-to nanosecond and micro-to millisecond (s-ms) backbone dynamics for bound NCBD and ACTR were reduced when compared with the free state, several residues for both proteins in the bound state had chemical exchange contributions to the transverse relaxation rate R 2 , indicative of s-ms motions and possibly the result of exchange between different states. In another study (25), the authors concluded that the NMR structure of the complex between the activation domain of stereo receptor co-activator 1 (SRC1) and NCBD (SRC1 is a ACTR homolog), which they determined, was in exchange with another minor bound species, due to the presence of additional cross-peaks.
The removal of the buried salt bridge by mutation (NCBD R2104L and ACTR D1068L ) has a profound effect on the association rate constant, reducing it by a factor of 20. IDPs tend to be enriched in charged residues and depleted of bulky hydrophobic residues, making it difficult to form a hydrophobic core. For instance, NCBD has a total of 7 Arg and Lys residues, and only 1 Asp residue, whereas ACTR has 13 Glu and Asp residues and 5 Arg and Lys residues. This suggests that electrostatics may be one of the key determinants for the fast associations that have been experimentally observed for some IDPs (26). We note that in general, k on values for IDPs and intrinsically disordered regions may not be larger than those of ordered proteins (18,19,26,27).
In conclusion, although our results show that initial binding could be fast for IDPs, they also highlight the disadvantage of having multiple binding partners, namely that finding the most stable conformation in the bimolecular complex may be a relatively slow process.