Tailoring Relaxation Dispersion Experiments for Fast-Associating Protein Complexes

Abstract

NMR relaxation dispersion spectroscopy is a powerful technique to elucidate the mechanism of protein-protein binding reactions. However, it is difficult to optimize the concentration ratios that give relaxation dispersions of appropriate amplitude to determine accurate kinetic and thermodynamic parameters, especially in cases of very tight binding. In this study, we have obtained ¹⁵N R₂ dispersions of Asn803-hydroxylated hypoxia-inducible factor-1α (HIF-OH) in the presence of sub-stoichiometric amount of its target protein, the transcriptional adapter zinc-binding (TAZ1) domain of CREB binding protein, whereas no R₂ dispersion was observed for the bound state of HIF-OH at 1:1 concentration ratio because the binding is too tight. Although the R₂ dispersions were measured for the free peaks of HIF-OH, they enabled us to quantitate the kinetic and thermodynamic parameters of HIF-OH/TAZ1 binding process. Simulations of effective R₂ rates revealed that the association rate is the key factor to determine the amplitude of R₂ dispersions. By careful optimization of the concentration ratio, the R₂ dispersion method should be generally applicable for studying a wide range of protein-protein, protein-nucleic acid, and protein-small molecule interactions.

NMR relaxation dispersion spectroscopy^1–3 is a very powerful technique to characterize chemical and conformational exchange processes in proteins and other biologically important molecules on μs–ms timescales.⁴ Many of these processes proceed through weakly populated higher energy states that are difficult to detect experimentally. A major advantage of the relaxation dispersion method is that it can provide information on the structure of invisible high energy intermediates by analyzing the relaxation rates of an observable state that interconverts with them.⁵

We recently developed a method combining ¹⁵N R₂ relaxation dispersion spectroscopy and NMR titration to elucidate the mechanism of protein-protein binding reactions, and in particular the coupled folding and binding processes that occur when an intrinsically disordered protein binds to its target.⁶ The method was applied to study binding of the phosphorylated kinase inducible domain (pKID) of the transcription factor CREB to the KIX domain of the CREB binding protein (CBP). By measuring ¹⁵N R₂ dispersions of pKID in the bound form as a function of pKID:KIX concentration ratio, we showed that pKID forms weak non-specific encounter complexes with KIX that evolve via a folding intermediate to the fully bound state. Although our method is potentially applicable to a broad spectrum of biological interactions, it can be difficult to optimize the concentration ratios that give relaxation dispersions of sufficient amplitude to determine accurate kinetic and thermodynamic parameters. In this Communication, we describe methods for measuring relaxation dispersion in cases of very tight binding or loss of resonances in the bound state due to intermediate exchange.

As an example, we consider binding of the Asn803-hydroxylated hypoxia-inducible factor-1α (HIF-OH) to the transcriptional adapter zinc finger (TAZ1) domain of CBP.^7–9 Although HIF-OH binds to TAZ1 by a coupled folding and binding mechanism (Supplementary Figure S1) similar to the binding of pKID to KIX,⁶ we observe no ¹⁵N R₂ dispersion for HIF-OH in the fully bound state, i.e. at concentration ratios close to 1:1 that were appropriate for study of the lower-affinity pKID:KIX complex (Supplementary Figure S2).

We therefore analyzed ¹⁵N R₂ dispersions of the free HIF-OH resonances in samples containing sub-stoichiometric amounts of TAZ1; no dispersion was observed in the absence of TAZ1. In contrast to experiments performed on the bound state of HIF-OH in the presence of equimolar TAZ1, high quality R₂ dispersion curves were obtained (Figure 1) and could be fitted well to a two-site exchange model:

$$\text{HIF}-\text{OH}+\text{TAZ}1\underset{k_{\text{off}}}{\overset{[\text{TAZ}1]k_{\text{on}}}{\rightleftarrows }}\text{HIF}-\text{OH}:\text{TAZ}1$$

where the fitting parameters are the chemical shift difference between the free and bound states, Δω_FB, the population-average intrinsic relaxation rate, R₂⁰, the association and dissociation rate constants, k_on and k_off, and the total concentrations of HIF-OH and TAZ1 (denoted [HIF-OH]₀ and [TAZ1]₀, respectively). The free TAZ1 concentration, [TAZ1], is calculated from K_D, [HIF-OH]₀ and [TAZ1]₀.

Figure 1.

¹⁵N R₂ dispersion profiles for Arg820 of HIF-OH recorded for 510 μM [¹⁵N]-HIF-OH in the presence of sub-stoichiometric amounts of TAZ1. Data were acquired at 900 MHz (filled circles) and 600 MHz (open circles) using relaxation compensated Carr-Purcell-Meiboom-Gill (CPMG) pulse sequences.¹,¹² Curves for all observable dispersions are shown in Supplementary Figure S3.

For simplicity, we focused on the C-terminal α-helix (residues 816-825), which forms a distinctive binding element in the HIF-1α-TAZ1 complex.^7,8 and fitted the dispersion curves to a global exchange process with fixed k_on and k_off values for all residues. The chemical shift changes Δω_FB determined from the R₂ dispersion data correlate well with equilibrium chemical shift changes measured from NMR titration of HIF-OH with TAZ1 (Supplementary Figure S4). The linear correlation with slope near 1 indicates that the weakly populated state detected by the R₂ dispersion experiment is the TAZ1-bound form.

Since the k_on obtained from these measurements (Table 1) is on the same order as translational diffusion,¹⁰ we infer that HIF-OH binding to TAZ1 is diffusion limited.

Table 1. R.

₂ dispersion with two-site exchange

Parameter	value (fitting)
[HIF-OH]0 [μM]	505 (62)a
[TAZ1]0 [μM]	26.9 (3.2)a,b
kon × 109 [M−1·s−1]	1.29 (0.94)a
koff [s−1]	185 (7.6)a
KD [nM]	143c

^aValues in parentheses are standard deviations.

^bdetermined for the sample with the largest amount of TAZ1. The concentrations of TAZ1 in other samples were scaled by 0.4, 0.6 and 0.8.

^ccalculated according to the formula: K_D=k_off/k_on. This value is in reasonable agreement with the K_D of 70 nM measured by isothermal titration calorimetry (unpublished data). Fits to three-site exchange models failed to reproduce the measured K_D.

The k_on value is greater by three orders of magnitude than that for pKID-KIX complex formation (10⁶ M⁻¹·s⁻¹).⁶ The difference suggests that we were not able to observe R₂ dispersions in the bound state for the HIF-OH-TAZ1 complex because k_on is very fast. To understand how the relative concentrations of the binding partners affect the R₂ dispersion profile, we simulated¹¹ the dependence of the effective R₂ relaxation rate, R₂^eff (=R₂⁰+R_ex), on the concentration ratio, [TAZ1]₀/[HIF-OH]₀, using the parameters obtained from the relaxation dispersion analysis (Figure 2a). R_ex is the excess contribution to R₂^eff caused by the exchange process.

Figure 2.

R₂^eff rates for Arg820 of HIF-OH simulated using the parameters listed in Table 1. (a) R₂^eff rates versus the concentration ratio, [TAZ1]₀/[HIF-OH]₀. The red lines indicate R₂⁰, where the R₂⁰ rates for the free and bound states are 5 and 15 s⁻¹, respectively. The grey area indicates the concentration ratios that give R_ex rates from 3 to 25 s⁻¹. A similar plot for the pKID/KIX system⁶ is shown in Figure S5. (b) R₂ dispersion profiles at concentration ratios of 0.97, 0.98, 0.99 and 1. τ_CP indicates the delay between successive ¹⁵N 180 pulses in the CPMG sequence.

Our previous experience of R₂ dispersion curve fitting suggests that R_ex should be moderate, ~3–25 s⁻¹_, as shown by the gray area (concentration ratio of ~0.02–0.12 for the HIF-OH-TAZ1 system), in order to obtain reliable kinetic and thermodynamic parameters from the fits. The concentration ratios that gave good R₂ dispersion curves in the experiment (0.021–0.053) indeed fall into this range. When R_ex is too large, the NMR peak will be very broad (or unobservable), resulting in an R₂ dispersion curve with low signal/noise. On the other hand, R_ex is completely suppressed in the presence of excess TAZ1, because the HIF-OH:TAZ1 complex is of sufficiently high affinity (ca. 70 nM by ITC) that the population in the free state is very low. In this case, R₂^eff reaches the limiting value of R₂⁰ for the bound state and there will be no R₂ dispersion. R_ex is determined by the ratio of the exchange rate, which is a sum of the forward and backward rates([TAZ1]k_on+k_off), to the chemical shift difference. Since the forward rate, [TAZ1]k_on, is the only parameter that depends on the concentration ratio, the forward rate is the key factor that determines the amplitude of R_ex. At equimolar concentrations of HIF-OH and TAZ1, the forward rate is very fast compared to the chemical shift difference between the free and bound states ([TAZ1]k_on=10883 s⁻¹; Δω_FB=1675 rad·s⁻¹ (3.78 ppm) at ¹⁵N frequency of 60.8 MHz). This is the reason why no R₂ dispersion was observed at 1:1 concentration ratio. Figure 2a suggests that it should be possible to observe R₂ dispersions that satisfy the criterion that R_ex be in the range ~3–25 s⁻¹ for concentration ratios from 0.97 to 1. However, simulations of the resulting dispersion profiles (Figure 2b) show that R₂^eff does not converge to R₂⁰ with experimentally attainable ¹⁵N 180° pulse repetition rates (1/τ_CP) so that the derived kinetic and thermodynamic parameters will be unreliable. Therefore, we conclude that when the k_on is very fast, dispersion curves should be measured for the free state of a protein in the presence of sub-stoichiometric amounts of the binding partner.

In summary, despite the absence of observable R₂ dispersions for the bound state of HIF-OH, we have obtained R₂ dispersion profiles of HIF-OH sufficient to derive kinetic and thermodynamic parameters by using sub-stoichiometric concentrations of TAZ1. Although the relaxation rates of the free resonances were analyzed, the chemical shifts of the unobservable bound state could be determined by fitting the dispersion data. In the rather common case where bound peaks are unobservable due to intermediate exchange, relaxation dispersion measurements for free ligand resonances could provide chemical shifts and hence valuable structural information for the complex. By careful optimization of the concentration ratio, the R₂ dispersion method should be generally applicable for studying a wide range of protein-protein, protein-nucleic acid, and protein-small molecule interactions.

Supplementary Material

1si20071001_12. Supporting Information Available.

Details of sample preparation, NMR titration of [¹⁵N]-HIF-OH with TAZ1, ¹⁵N R₂ dispersion experiments, fitting of R₂ dispersion profiles and simulation of R₂ dispersion profiles. This material is available free of charge via the Internet at http://pubs.acs.org.