Dissecting allosteric effects of activator–coactivator complexes using a covalent small molecule ligand

Significance

The master coactivator CREB binding protein (CBP)/p300 utilizes several conformationally dynamic protein–protein interaction motifs to regulate the transcriptional status of essential genes. The kinase-inducible domain interacting (KIX) domain within CBP/p300 forms complexes with multiple transcriptional activators, using two allosterically connected binding surfaces. Here transient kinetic experiments of KIX–ligand complexes were used to dissect the underlying mechanism of the allosteric alterations in ternary complex formation. For positive cooperativity, the stabilization of protein–ligand complex as indicated by a decrease in k_off is the primary determinant. In contrast, negative cooperativity was derived by a decrease in k_on and an increase in k_off. As KIX–activator complexes are important models for intrinsically disordered protein complexes, these data provide a broader mechanistic framework for positive and negative cooperativity.

Abstract

Allosteric binding events play a critical role in the formation and stability of transcriptional activator–coactivator complexes, perhaps in part due to the often intrinsically disordered nature of one or more of the constituent partners. The kinase-inducible domain interacting (KIX) domain of the master coactivator CREB binding protein/p300 is a conformationally dynamic domain that complexes with transcriptional activators at two discrete binding sites in allosteric communication. The complexation of KIX with the transcriptional activation domain of mixed-lineage leukemia protein leads to an enhancement of binding by the activation domain of CREB (phosphorylated kinase-inducible domain of CREB) to the second site. A transient kinetic analysis of the ternary complex formation aided by small molecule ligands that induce positive or negative cooperative binding reveals that positive cooperativity is largely governed by stabilization of the bound complex as indicated by a decrease in k_off. Thus, this suggests the increased binding affinity for the second ligand is not due to an allosteric creation of a more favorable binding interface by the first ligand. This is consistent with data from us and from others indicating that the on rates of conformationally dynamic proteins approach the limits of diffusion. In contrast, negative cooperativity is manifested by alterations in both k_on and k_off, suggesting stabilization of the binary complex.

Protein–protein interactions (PPIs) underscore all cellular processes, and the mechanistic dissection of PPI networks is thus a high priority (1–4). A particular challenge is defining the mechanism of PPI formation between intrinsically disordered proteins (IDPs) where allosteric changes play a substantive role (5–7). Allosteric communication between binding sites is vital for the proper function of feedback and regulatory circuits in the cell (8, 9). For example, the conformationally dynamic kinase-inducible domain interacting (KIX) domain of the master transcriptional coactivator CREB binding protein (CBP) undergoes a structural shift and stabilization upon binding to a cognate ligand such as the intrinsically disordered transcriptional activation domain (TAD) of mixed-lineage leukemia protein (MLL) (Fig. 1A) (10). The interaction of a second ligand, phosphorylated kinase-inducible domain (pKID) of CREB, is enhanced (up to twofold) by the presence of MLL and several elegant studies have documented allosteric communication between the two binding sites (11–15). However, the relatively modest affinities (micromolar) of the native complexes and the aggregation propensities and the promiscuous binding profiles of the transcriptional activation domains have hampered kinetic dissection of this and other complexes (16, 17).

Fig. 1.

(A) Overlay of KIX structures: simulated free KIX derived from 2AGH (30) and 1–10–KIX L664C [Protein Data Bank (PDB) ID: 4I9O] with MLL (PDB ID: 2LXS) and with both pKID and MLL (PDB ID: 2LXT). Note the structural shift in the loop and terminal regions upon ligands binding to KIX. (B) The binding site of MLL (purple) and pKID (yellow) on KIX (PDB ID: 2LXT), with 1–10 depicted in teal overlaid from structure of 1–10–KIX L664C (PDB ID: 4I9O).

The coactivator CBP exists across metazoans (18–20) and is a transcription hub that interacts with numerous transcriptional activators, using several discrete domains (21, 22). The KIX domain within CBP is a 90-residue motif that consists of a three-helix bundle along with two 3₁₀ helices (12). The KIX domain interacts with more than 10 distinct transcriptional activators (23) via two different binding sites, with the resulting complexes leading to regulation of key functions such as hematopoiesis, cell cycle progression, and memory formation (Fig. 1B) (12, 13, 24–26). Many of the KIX-binding TADs from activators such as pKID, MLL, and c-Myb are intrinsically disordered proteins that assume a helical structure only upon binding to interaction partners such as KIX (25, 27, 28). Conformational changes in both the TADs and KIX occur upon binding, and NMR solution studies have underscored that each TAD•KIX complex is conformationally unique (10, 12–14, 25). However, the transient-state kinetic mechanism of two ligands binding cooperatively to the KIX domain has yet to be defined, and there is little information on the mechanism of how different protein ligands elicit differential degrees of allosteric effects (11, 15). Answers to these questions are crucial for defining the communication pathways within CBP protein interaction networks.

Previously we described the use of stopped-flow transient-state kinetic studies to examine complexes between activators and coactivators (29). Here we hypothesized that a similar strategy could be applied to dissect the allosteric communication between the MLL and pKID binding sites of native KIX. As a complementary tool in this effort we use a covalent small molecule ligand of the KIX domain, fragment 1–10, that quantitatively forms a stable complex at the MLL binding site and in doing so alters the conformational dynamics of KIX. This small molecule can influence binding at the second site either positively or negatively, depending on the site of covalent attachment, thus facilitating transient kinetic studies of KIX in complex with native binding partners to examine both positive and negative cooperativity. We find that the change in the dissociation rate of the TAD•KIX complex is the main driving force for changes in overall binding affinities in positive cooperative binding events. These results suggest that KIX allosterically regulates its binding events not simply through altering interaction interfaces, but also through stabilizing the energetic state of the ternary complex. Thus, in this IDP protein complex, dissociation rate mainly governs the binding equilibrium.

Results

The goal in this study was to kinetically dissect the continuum of cooperativity in KIX ternary complex formation. Toward that end, in addition to wild-type KIX domain in complex with one or both of the native transcriptional activation domains of MLL and pKID, we used several KIX mutants as well as a cysteine-reactive small molecule ligand for KIX, compound 1–10 (Fig. 1A). Replacement of isoleucine 660 with a valine has been shown to disrupt allosteric communication within KIX, providing a valuable control for these experiments (10, 14). The two additional KIX mutants are those in which a cysteine residue replaces either N627 or L664, both of which border the MLL binding site. Both the N627C and L664C mutants maintain affinity for MLL and for pKID (30, 31). Previously, we described the discovery of compound 1–10 through a tethering screen of KIX L664C (30). This small molecule forms a selective covalent complex through formation of a disulfide bond with the cysteine residue and, in doing so, occupies the MLL binding site. Furthermore, equilibrium binding experiments with the 1–10–KIX L664C complex demonstrated that 1–10 tethered to a cysteine residue at position 664 allosterically inhibits pKID binding. Parallel to the KIX L664C screen we carried out a tethering screen of N627C and compound 1–10 emerged as a molecule that selectively tethers to that cysteine with high efficiency as well (31); the binding properties of the 1–10–KIX N627C complex are defined below. An advantage of 1–10–KIX L664C and 1–10–KIX N627C complexes is that they can be generated quantitatively in a disulfide exchange reaction and then purified, enabling a single species to be used in kinetic assays.

Equilibrium Binding Measurements Define Extent of Cooperativity for Each Complex.

Dissociation constant measurements of wild-type KIX and mutants either in complex with excess MLL or fully labeled with small molecule 1–10 were carried out to define the extent of cooperativity for each of the complexes (Fig. 2 and Figs. S1 and S2). As expected, KIX I660V precomplexed with excess MLL exhibited no cooperative binding to pKID (32). However, KIX N627C either precomplexed with excess MLL or fully labeled with small molecule 1–10 recapitulated the cooperative binding behavior observed in the wild-type protein. In contrast, when 1–10 was fully tethered to KIX L664C, negative cooperativity was observed, with a more than twofold loss of pKID affinity.

Fig. 2.

Bar graph of relative K_D values of pKID•KIX complexes normalized to the K_D of pKID and free KIX constructs, obtained from fluorescence anisotropy assays performed in triplicate. The KIX structure (PDB ID: 4I9O) with the corresponding mutated residue depicted in a gray sphere is shown under each KIX construct. Bars below the dotted line reflect positive cooperativity whereas bars above the line reflect negative cooperativity. Errors reflect the standard error (SE) of nonlinear fits in GraphPad Prism propagated for normalization to free KIX constructs. *0.01 < P < 0.05, **P < 0.01; P values are calculated by GraphPad Prism. K_D values are in Table S1.

Identification of KIX Residues Affected in Positive and Negative Cooperativity Modes.

Chemical shift perturbation experiments of 1–10 tethered to KIX L664C (30) or KIX N627C revealed distinct differences in the binding modes. In this experiment uniformly ¹⁵N-labeled KIX N627C was subsequently reacted with sufficient 1–10 to enable complete conversion to tethered product and purified. Heteronuclear single quantum correlation (HSQC) spectra of ¹H-¹⁵N of the 1–10-labeled and free KIX N627C construct were then compared with chemical shift perturbation data of 1–10–KIX L664C from an earlier study (Fig. 3 and Fig. S3) (30). Not surprisingly, 1–10 bound to either mutant affects the residues within and proximal to the pKID binding site. However, 1–10 has a larger overall effect on KIX L664C residues (blue and green), with perturbations distributed not only at both MLL (between helices α₂ and α₃) and pKID (between helices α₁ and α₃) binding interfaces but also at the loop region, which has been implicated to play an important role in the conformational dynamics of KIX during binding events (12, 33, 34). The perturbations exhibited at the N terminus of α₃ and the two termini of α₁ are consistent with those observed upon MLL binding to KIX (12).

Fig. 3.

Results of chemical shift perturbation experiments superimposed upon the KIX structure (PDB ID: 4I9O). Residues displaying chemical shift perturbation greater than 1 SD upon 1–10 tethering are depicted in spheres: The perturbed residues of KIX L664C (from a previous study) (30) resulting in negative cooperativity are shown in blue, those of KIX N627C resulting in positive cooperativity are shown in yellow, and residues that are perturbed in both KIX L664C and KIX N627C are shown in green. Residues 664 and 627, the sites of 1–10 tethering, are depicted as black spheres (PDB ID: 4I9O). Chemical shift data are in SI Text.

Transient-State Kinetic Analysis of KIX Complexes.

Subsequently, we examined the transient-state binding kinetics between KIX and pKID by fluorescence stopped flow in the presence and absence of the peptide and small molecule ligand at room temperature (schematic shown in Fig. 4A). The association rate (see k_on in Fig. 6A) is determined from the changes in fluorescence intensity after rapidly mixing fluorescently labeled pKID with KIX complexes (Fig. S4A) and the dissociation rate (see k_off in Fig. 6A) is measured from changes in fluorescence intensity when the labeled-pKID•KIX complexes are mixed with unlabeled pKID peptide (Fig. S4B). Fluorescein is tagged at the N terminus of the pKID peptide, a construct that has been used in previous studies (35–37) to show allosteric effects consistent with results from isothermal calorimetry (ITC) assays using unlabeled pKID (11). One rapid association phase was observed upon mixing pKID with varying concentrations of KIX complexes and the observed association rates exhibit a linear dependence on the KIX complex concentration (Fig. 4B). The slope value from the linear regression fit reflects the rate constant (k_on) of the association step. The value of k_on for pKID binding to free KIX WT (k_on = 14 ± 1 μM⁻¹⋅s⁻¹) is within the same magnitude as values obtained by NMR relaxation dispersion in the literature (6 μM⁻¹⋅s⁻¹) (25), confirming that this is a viable method for studying the transient-state interaction between pKID and KIX. We did not observe an additional rapid conformational change step (∼200 s⁻¹) reported in the literature due to time resolution constraints of the stopped-flow instrument (dead time ∼2 ms). The calculated association and dissociation rate constants measured for the various pKID•KIX complexes are summarized in Table 1, Fig. 5B, and Fig. S5A. (Representative spectra are in Fig. 4A and SI Text.)

Fig. 4.

(A) A schematic of the stopped-flow setup for measuring association. Excess KIX complex of varying concentrations is rapidly mixed with a constant amount of FITC-pKID peptide (final concentration 25 nM) at 25 °C, and the change in fluorescence intensity is measured in real time. (B) The dependence of k_obs on the concentration of excess KIX for formation of pKID•KIX complexes (KIX WT, KIX I660V, KIX N627C, and KIX L664C), which suggests this association step is a bimolecular interaction The linear regression of each slope approximates the association rate constant (k_on) of this interaction. Each k_obs value is an average of two separate experiments; each experiment is an average of five to eight traces. The error bars represent the SD of the two k_obs values.

Fig. 6.

(A) A mechanism scheme for pKID association with KIX complexed with peptide (in purple, MLL) or small molecule (in cyan, 1–10) ligand, where the association step (k_on) is a bimolecular process. The orange star on pKID depicts the position of the fluorescein label, which is not at the primary site of interaction. (B) Cooperativity correlation plot of relative k_on vs. K_D (Left) and k_off vs. K_D values (Right) (normalized to values for free KIX binding to pKID for each KIX construct). These plots illustrate that there is little correlation between k_on and K_D values whereas k_off values linearly correlate with K_D values exhibiting an R² value of 0.8. The k_on value for 1–10–KIX L664C (negative cooperativity) is not included in this plot. Full details are in SI Text. (C) A schematic energy diagram of the proposed mechanism of positive cooperativity through stabilizing the final ternary complex (resulting in increase of k_off). (D) A schematic energy diagram of the proposed mechanism of negative cooperativity through stabilizing the initial binary complex (resulting in both decrease in k_on and increase in k_off).

Table 1.

Association and dissociation rate constants of pKID•KIX complexes

Construct	Rate constants	Free	With 4× MLL	With 1–10
KIX WT	kon, μM−1⋅s−1	14 ± 1	14 ± 1	NA
	koff, s−1	7.0 ± 0.4	5.0 ± 1.3	NA
KIX I660V	kon, μM−1⋅s−1	14 ± 1	11 ± 1	NA
	koff, s−1	5.3 ± 0.5	5.4 ± 0.3	NA
KIX N627C	kon, μM−1⋅s−1	15 ± 1	14 ± 1	17 ± 2
	koff, s−1	7.3 ± 1.3	4.1 ± 2.5	4.0 ± 0.8
KIX L664C	kon, μM−1⋅s−1	23 ± 2	24 ± 2	6 ± 1
	koff, s−1	8.7 ± 0.4	7.7 ± 0.3	13.8 ± 0.6

Association rate constants are slopes of linear regression fits of Fig. 4B. Errors reflect the SE of nonlinear fits in GrapPad Prism. For dissociation rate constants, errors are SDs of the average of two separate values.

Fig. 5.

(A) Bar graph of equilibrium dissociation constants between pKID and KIX complexes. The error bars reflect the SE of nonlinear fits in GraphPad Prism. (B) Bar graph of dissociation rate constants of pKID dissociating from KIX complexes. The error bars represent SD of the average of two values from separate experiments. *0.01 < P < 0.05, **P < 0.01; P values are calculated by GraphPad Prism.

The kinetics for the binding of pKID to the various KIX complexes show a positive correlation between the change in the affinity of KIX for pKID (K_D) and the dissociation rate of pKID from the KIX•pKID complex (k_off) (Figs. 5 and 6B). Specifically, N627C shows positive cooperativity upon addition of either MLL or 1–10, increasing the binding affinity of pKID to KIX by nearly twofold, whereas the value of k_off similarly decreases almost twofold from 7.3 s⁻¹ to 4.1 s⁻¹ with no observed effect on the k_on, which is in a constant range between 14 μM⁻¹⋅s⁻¹ and 17 μM⁻¹⋅s⁻¹. Therefore, in this system, positive cooperativity is primarily determined by k_off. KIX L664C previously showed negative cooperativity upon tethering to 1–10 (30). Whereas an increase in k_off contributes to this allosteric change as well (from 8.7 s⁻¹ to 13.8 s⁻¹), there is also a decrease in k_on values from 23 μM⁻¹⋅s⁻¹ to 6 μM⁻¹⋅s⁻¹. The control KIX variant, KIX I660V, in which the allosteric network is interrupted, shows no significant change in either the association or dissociation rate constants with and without the presence of MLL.

Discussion

Taken together, these data provide insight into the mechanism of ternary activator–coactivator complex formation (Fig. 6A). In the case of positive cooperativity, the consistent k_on values and the correlation of relative k_off and K_D values (Fig. 6B and Fig. S4B) suggest that this cooperativity is achieved not by lowering the energy barrier for association but by increasing the energy barrier for dissociation. This indicates a scenario where the binding of one ligand to KIX does not significantly change the interaction interface for the second ligand, but when both ligands bind to KIX, this ternary complex occupies an energy minimum that is more stable than that of the binary complex. Therefore, the major effect of positive cooperativity is to stabilize the ternary complex (Fig. 6C). That k_off plays a major role in the cooperative binding of two activators to KIX is reasonable, considering that our measured k_on values (0.6–2.4 × 10⁷ M⁻¹⋅s⁻¹, Table 1) are close to the predicted maximal diffusion rate of 10⁷–10⁹ M⁻¹⋅s⁻¹ for proteins in a cellular environment (28, 38–40). Thus, when cellular regulation requires an allosteric increase in protein–protein binding affinity, there is little room for increase in the association rate constant (k_on). Alternatively, the dissociation rates (k_off) are not governed by diffusion limits and can be significantly altered, providing access to a large range of binding affinities. The dissociation rates of transcriptional activators from DNA or small molecule ligands from proteins have been shown to govern the residence time of proteins and ligands in particular functional states and/or locations (41, 42). It is possible that in this case the residence time of various TADs with the coactivator CBP is used to regulate the degree of transcriptional activation and/or histone acetyl transferase activity in cellular pathways.

The analysis of negative cooperative binding with KIX (1–10 tethered to KIX N664C) reveals a more complicated scenario. In this instance both k_on and k_off are affected, with k_on decreasing 4-fold and k_off increasing 1.6-fold (Table 1), suggesting stabilization of the unbound binary complex plays a major role (Fig. 6D). In a case when negative cooperativity is required in physiological regulation, there is ample room for the decrease of k_on from diffusion-controlled limits. We noted from the chemical shift perturbation data that the residues perturbed by 1–10 upon tethering to KIX L664C are similar to those altered by the binding of sekikaic acid, a natural product small molecule that also allosterically decreases pKID binding affinity (43): In the case of both 1–10 and sekikaic acid, the residues 608–611 and 629–631, which flank the loop region in helices α₁ and α₂, exhibit significant alterations. As we have identified two small molecule ligands of KIX, 1–10 and sekikaic acid, that both show negative cooperativity with pKID binding through standard screens, it is possible that negative allosteric regulation of KIX binding occurs with native ligands, a scenario currently under investigation.

Finally, the transient kinetic experiments described here were significantly facilitated by a covalent small molecule ligand for KIX. Not only do covalent ligand–protein complexes offer the advantage of an isolable, homogeneous sample for study, but also in this case we were able to access a continuum of cooperativity modes, significantly expanding the breadth of the analysis. Given the ease with which such covalent small molecules can be discovered via tethering, we anticipate the broader use of such ligands in mechanistic studies of conformationally dynamic protein complexes that comprise critical cellular functions (44–46).

Materials and Methods

HSQC NMR Experiments on ¹H-¹⁵N.

Uniformly ¹⁵N-labeled KIX N627C protein was expressed and purified as previously described (47). Aliquots of the purified ¹⁵N-labeled KIX N627C were tethered with the small molecule 1–10 as previously described (30). A 45-μM solution of ¹⁵N-labeled KIX L627C with or without conjugated small molecule was prepared in a 9:1 H₂O:D₂O 10-mM sodium phosphate buffer containing 100 mM NaCl, pH 7.2. HSQC experiments on ¹H-¹⁵N were recorded at 27 °C on an Avance Bruker 600 MHz NMR spectrometer equipped with a 5-mm cryogenic probe. HSQC data were collected with the protein alone and the protein was covalently tethered to 1–10 and the chemical shifts were compared. Data were processed using NMRpipe (48) and analyzed in Sparky (University of California, San Francisco) (www.cgl.ucsf.edu/home/sparky/). Chemical shifts of residues were identified based on previous assignments (43). Chemical shift changes for individual peaks were quantified as ((0.2 × Δδ¹⁵N) ² + (Δδ¹H) ²)^0.5, which is a weighted length of the vector from free KIX N627C to protein complexed with a small molecule (Fig. S3).

Fluorescence Anisotropy Assays.

The fluorescence anisotropy assays were performed in triplicate with a final sample volume of 10 μL in a low-volume, nonbinding, black, 384-well plate (Corning) and read using a Tecan Genios Pro plate reader with polarized excitation at 485 nm and emission intensity measured through a parallel and perpendicularly polarized 535-nm filter. Fluorescein isothiocyanate (FITC)-labeled peptides were diluted in storage buffer [10 mM sodium phosphate, 100 mM NaCl, 0.01% Nonidet P-40, 10% (vol/vol) glycerol, pH 6.8] to a concentration of 25 nM. Then 10 µL of the peptide solution was added to a series of 50-µL solutions of varying KIX concentrations in storage buffer to obtain the final concentrations of up to 20 μM. The samples were incubated for 30 min at room temperature before the degree of fluorescence anisotropy was measured (Tecan Genios Pro). Anisotropy data were corrected for the change in fluorescence intensity using Eq. 1, where “f_B”is the fraction bound, “R” is the ratio between fluorescence intensity of 100% bound ligand and free ligand, “r” is the measured anisotropy value, “r_F” is the anisotropy value of free ligand, and “r_B” is the anisotropy value of 100% bound ligand:

$$f_{\text{B}}=\frac{r-r_{\text{F}}}{R\left(r_{\text{B}}-r\right)+r-r_{\text{F}}}.$$ [1]

A binding isotherm that accounts for ligand depletion (assuming a 1:1 binding model of peptide to KIX) was fitted to the observed anisotropy values as a function of KIX concentration to obtain the apparent equilibrium dissociation constant, K_D,

$$y=c+\left(b-c\right)\times \frac{[\left(K_{\text{D}}+a+x\right)-\sqrt{{\left(K_{\text{D}}+a+x\right)}^2-4ax}]}{2a},$$ [2]

where “a” and “x” are the total concentrations of fluorescent peptide and KIX, respectively, “y” is the observed anisotropy at any KIX concentration, “b” is the maximum observed anisotropy value, and “c” is the minimum observed anisotropy value. Each data point in Figs. S1 and S2 is an average of three independent experiments with the indicated error (SD). Data analysis was performed using GraphPad Prism 5 software (Figs. S1 and S2).

An analogous experiment with varying molecular equivalents of MLL to KIX was also carried out to determine that at least four molecular equivalences of MLL to KIX are required to elicit a maximum degree of positive cooperativity (Fig. S1).

Fluorescence Stopped-Flow Kinetic Assays.

Stopped-flow experiments were performed on a KinTek model SF-2001 stopped-flow apparatus equipped with a 75-W Xe arc lamp in two-syringe mode. FITC was excited at 493 nm and its emission was monitored at wavelengths >510 nm, using a long-pass filter (Corion).

All kinetic traces reported are an average of five to eight independent determinations. The possibility of the traces fitting to multiple exponentials was explored by using variations of Eq. 3. However, all traces displayed a monophasic association and fit to a single exponential equation (Eq. 4). A single exponential was fitted to the transient kinetic time courses, F(t) as in Eq. 4, to obtain the fluorescence amplitude (A) and the observed rate, k_obs, for each exponential phase where F(0) is the initial fluorescence intensity and t is the time:

$$F\left(t\right)={{\displaystyle \sum }}^{\text{}}{A}_n\left[1-\exp \left(-k_{\text{obs},n}t\right)\right]+F\left(0\right)$$ [3]

$$F\left(t\right)=A\left[1-\exp \left(-k_{\text{obs}}t\right)\right]+F\left(0\right).$$ [4]

Analysis of the time courses was performed using Kintek software, and the reported errors are the asymptotic SEs. The dependence of the observed rates on KIX concentration was analyzed using GraphPad Prism 4.0 software.

Control experiments.

A total of 25 nM pKID-FITC was rapidly mixed with either KIX storage buffer (10 mM sodium phosphate, 100 mM NaCl, 0.01% Nonidet P-40, 10% glycerol, pH 6.8) (Fig. S2A) to control for potential photobleaching of the fluorophore or 0.1 mg/mL of BSA in KIX storage buffer (Fig. S2B) to control for effects of nonspecific binding and crowding of protein. Both control experiments revealed no change in fluorescence intensity over 120 s.

Association experiments.

A total of 25 nM (final concentration after mixing) pKID-FITC in KIX storage buffer was rapidly mixed with varying concentrations (0.1–5 μM, final concentration after mixing) of KIX complexes (free protein, precomplexed with 4 eq of MLL peptide, or pretethered with 1–10) in KIX storage buffer at 25 °C. The measured time domains (0.1 and 1 s or 0.05 and 0.5 s) were selected to enhance the data analysis as they were closest to the predicted best time frame by the fits in the Kintek software (Fig. S6A). The results are summarized in Table 1.

Dissociation experiments.

A total of 25 nM (final concentration after mixing) pKID-FITC in KIX storage buffer was preequilibrated with 500 nM KIX complex (final concentration after mixing) and rapidly mixed with 12.5 μM (500 M eq, final concentration after mixing) unlabeled pKID peptide in KIX storage buffer at 25 °C (Fig. S6B). The results are summarized in Table 1. The Y intercepts from Fig. 3 (on rate vs. KIX concentration) approximately correlate with the off rates fitted in the dissociation experiments. Three of the Y-intercept values had a larger error of fitting, giving them too high a degree of uncertainty to compare with experimental dissociation values.