Intrinsic unfoldase/foldase activity of the chaperonin GroEL directly demonstrated using multinuclear relaxation-based NMR

Abstract

The prototypical chaperonin GroEL assists protein folding through an ATP-dependent encapsulation mechanism. The details of how GroEL folds proteins remain elusive, particularly because encapsulation is not an absolute requirement for successful re/folding. Here we make use of a metastable model protein substrate, comprising a triple mutant of Fyn SH3, to directly demonstrate, by simultaneous analysis of three complementary NMR-based relaxation experiments (lifetime line broadening, dark state exchange saturation transfer, and Carr-Purcell-Meinboom-Gill relaxation dispersion), that apo GroEL accelerates the overall interconversion rate between the native state and a well-defined folding intermediate by about 20-fold, under conditions where the "invisible" GroEL-bound states have occupancies below 1%. This is largely achieved through a 500-fold acceleration in the folded-to-intermediate transition of the protein substrate. Catalysis is modulated by a kinetic deuterium isotope effect that reduces the overall interconversion rate between the GroEL-bound species by about 3-fold, indicative of a significant hydrophobic contribution. The location of the GroEL binding site on the folding intermediate, mapped from (15)N, (1)HN, and (13)Cmethyl relaxation dispersion experiments, is composed of a prominent, surface-exposed hydrophobic patch.

Keywords: chaperonins; dark state exchange saturation transfer; invisible states; lifetime line broadening; relaxation dispersion.

Fig. 1.

¹⁵N lifetime line broadening of SH3 in the presence of GroEL. (A)¹⁵N-ΔR₂ profiles for 100 μM ¹⁵N-labeled SH3^Mut (Top) and SH3^WT (Bottom) in the presence of 120 μM (in subunits) GroEL at 900 (red) MHz and 600 (blue) MHz with the experimental data shown as circles and the calculated values as lines. Also shown as controls are the ¹⁵N-ΔR₂ profiles obtained upon addition of 19 μM acid-denatured Rubisco (green squares), a substrate known to bind to the internal cavity of GroEL with nanomolar affinity. This concentration corresponds to a stoichiometry of one Rubisco molecule per GroEL cavity. (B) Dependence of ¹⁵N-ΔR₂ on concentration of GroEL for SH3^Mut (red) and SH3^WT (blue). (C) Correlation plot of observed and calculated ¹⁵N-ΔR₂ values for SH3^Mut (red) and SH3^WT (blue) at 900 MHz (Left) and 600 MHz (Right). Experiments were conducted at 10 °C. Error bars = 1 SD.

Fig. 2.

¹⁵N CPMG relaxation dispersion and DEST on SH3^Mut in the presence of GroEL. (A) Examples of experimental (circles) and best-fit (lines) ¹⁵N CPMG relaxation profiles for SH3^Mut (100 μM) in the presence (red) and absence (blue) of 120 μM GroEL at 900 MHz (Top row) and 600 MHz (Bottom row). (B) Examples of experimental (circles) and best-fit (lines) ¹⁵N-DEST profiles for SH3^Mut in the presence of GroEL at two radiofrequency field strengths (500 Hz in orange and 750 Hz in purple) for ¹⁵N saturation. All experiments were conducted at 10 °C.

Fig. 3.

Four-state exchange model for the interaction of SH3^Mut with apo GroEL. F and I represent the free folded and intermediate states, respectively, of SH3^Mut whereas F-G and I-G are the corresponding GroEL-bound states. The values of the rate constants and populations obtained for the ¹⁵N/protonated SH3 samples from the simultaneous fits to the ¹⁵N lifetime-broadening, DEST, and CPMG relaxation dispersion data obtained at 10 °C are indicated (further details in Table 2 and SI Theory). The corresponding values obtained for the [¹⁵N/¹H_N/AILV-¹³CH₃]/deuterated SH3 samples are provided in Table 2. Top, ribbon diagrams of the free F [Protein Data Bank (PDB) ID: 2LP5] and I (PDB ID: 2L2P) states of SH3^Mut together with the side chain of Phe4 (15); Bottom, top view of one cylinder of GroEL (32) shown as a ribbon with the seven subunits indicated by different colors and a surface representation of the SH3 domain (red) placed in the cavity, showing that there is ample room within the cavity to accommodate one SH3 molecule.

Fig. 4.

Transverse (R₂) relaxation rates and chemical shifts for the intermediate (I-G) state of SH3^Mut bound to GroEL. (A) Calculated ${}^{15}\text{N}\mathrm{-}{R}_2^{\text{bound}}$ profiles for the I-G state at 900 (red) MHz and 600 (blue) MHz (note the lines in A simply serve to connect the calculated ${}^{15}\text{N}\mathrm{-}{R}_2^{\text{bound}}$ values). The corresponding ${}^{15}\text{N}\mathrm{-}{R}_2^{\text{bound}}$ profiles for the F-G state are shown in Fig. S4A. Error bars: 1 SD. (B) Mapping of chemical shift differences between the I-G and I states on a surface representation of the I state (Left) and for reference the F state (Right) (note the views are identical to those of the ribbon diagrams in Fig. 3, Top). Only residues that have a lower bound of |Δϖ^I,I-G| > 1.0 ppm (E5, T14, I50, S52, L55, and A56), 0.4 ppm (T2, L3, F4, E5, L7, I28, W37, S52, L55, and A56), and/or 1.0 ppm (A6, I28, L55, and A56) for ¹⁵N, ¹H_N, and/or ¹³C_methyl nuclei, respectively, are colored according to residue type (green, hydrophobic; light blue, polar; and red, negatively charged). Note the reduced extent of the hydrophobic patch in the F state compared with the I state, accounting for the substantially reduced affinity of the F state for GroEL.

Fig. S1.

In the absence of exchange between the F-G and I-G states, relaxation dispersion and ΔR₂ cannot be accounted for simultaneously in the presence of GroEL. Each panel shows the same simulations for different values of Δϖ^F,I = Δϖ^F,I-G (i.e., Δϖ^I,I-G = 0; note that in the absence of a fast exchange process connecting I-G to F via F-G, Δϖ^I,I-G has minimal impact on the relaxation dispersion profiles). Black curves show relaxation dispersions calculated for a two-state system with the interconversion between F and I states with k_FI and k_IF set to the experimentally determined values of 7.5 s⁻¹ and 310 s⁻¹, respectively (compare with Table 1 in main text). Red curves show relaxation dispersions calculated using the four-state model and parameters (rate constants and R₂ rates) shown in Fig. 2 of the main text (compare with Tables 2 and 3 in main text). Blue curves show relaxation dispersions for a four-state model where the F-G and I-G states do not directly interconvert with one another (i.e., $k_{\text{FI}}^{\text{G}}=k_{\text{IF}}^{\text{G}}=0$), $k_{\text{on}}^{\text{app}}=6{\hspace{0.17em}\text{s}}^{-1},60\hspace{0.5em}{\text{s}}^{-1},600\hspace{0.5em}{\text{s}}^{-1}$, and 6,000 s⁻¹, and the population of the I-G state I, p_I-G, is kept constant by scaling the k_off proportionately (i.e., k_off = 20 s⁻¹, 200 s⁻¹, 2,000 s⁻¹, and 2 × 10⁴ s⁻¹), and the remaining rate constants and R₂ values are set to the same values as in the red curves. Experimentally, the dispersion profiles in the presence of GroEL are shifted by the ΔR₂ value while retaining similar shapes (compare black vs. red simulated curves). In the absence of interconversion between the F-G and I-G states (blue curves), varying the association rate constant $k_{\text{on}}^{\text{app}}$ for the binding of F and I to GroEL does not reproduce the experimental observations, indicating that interconversion between F-G and I-G states is required.

Fig. S2.

Chemical shift difference comparisons between the free F and I states. Comparison of absolute chemical shift differences between the free F and I states (|Δϖ^F,Ι|) is determined here from the global fits (x axis) and those reported in the literature by Neudecker et al. (14) (y axis) for (A) ¹⁵N, (B) ¹H_N, and (C) ¹³C_methyl nuclei. Also shown in D is the correlation between the |¹⁵N-Δϖ^F,Ι| values obtained in the present work for the ¹⁵N/protonated (x axis) and [¹⁵N/¹H_N/AILV-¹³CH₃]/deuterated (y axis) SH3^Mut samples. The Pierson correlation coefficient R is also listed.

Fig. S3.

The experimental relaxation dispersion profiles and ΔR₂ values in the presence of GroEL are consistent with an acceleration in the interconversion rates between the F and I states only upon binding GroEL. The four panels show comparisons of experimental and simulated ¹⁵N-CPMG relaxation dispersion profiles at a spectrometer frequency of 900 MHz for the same residues (Asp9, Ser19, Trp36, and Leu55) shown in Fig. 2 of the main text. The experimental data for SH3^Mut in the absence and presence of GroEL are displayed as blue and red circles, respectively. The corresponding calculated curves obtained from the global best-fitting procedure to kinetic scheme 1 (Eq. S1) with the optimized values of the rate constants (Table 2 of main text; the optimized values of the interconversion rate constants between the F and I states when bound to GroEL are $k_{\text{FI}}^{\text{G}}=\mathrm{4,000}\hspace{0.5em}{\text{s}}^{-1}$ and $k_{\text{FI}}^{\text{G}}=\mathrm{1,300}\hspace{0.5em}{\text{s}}^{-1}$), bound R₂ values, and Δω values (SI Theory) are shown as solid blue and red lines, respectively. Simulated CPMG profiles are shown as green dashed lines for values of $k_{\text{FI}}^{\text{G}}$ ranging from 50 s⁻¹ to 50,000 s⁻¹, as indicated at Top Left. The $k_{\text{FI}}^{\text{G}}/k_{\text{IF}}^{\text{G}}$ ratio is kept constant at a value of 3.077 (the value obtained from the global best fit) to maintain detailed balance (see Eq. S2). The shape of the profiles, the R_ex values (the difference in R₂ at 0-Hz and 1,000-Hz CPMG field), and the ΔR₂ values (the difference in R₂ in the presence and absence of GroEL at a CPMG field of 1,000 Hz where exchange contributions to R_2,eff are suppressed) are reproduced only with the values of $k_{\text{FI}}^{\text{G}}$ and $k_{\text{IF}}^{\text{G}}$ provided in Table 2. For values of $k_{\text{FI}}^{\text{G}}$ and $k_{\text{IF}}^{\text{G}}$ smaller than the optimized ones, the calculated values of both R_ex and ΔR₂ are too small; for values of $k_{\text{FI}}^{\text{G}}$ and $k_{\text{IF}}^{\text{G}}$ larger than the optimized ones, the calculated values of ΔR₂ are too large.

Fig. S4.

Transverse relaxation rates of SH3 bound to GroEL. (A) Calculated ${}^{15}\text{N}‐{\mathit{R}}_2^{\text{bound}}$ profile for SH3^WT bound to GroEL [state F-G at 900 (red) MHz and 600 (blue) MHz]. (B) Correlation between calculated ${}^{15}\text{N}‐{\mathit{R}}_2^{\text{bound}}$ values for the I-G state and the measured ¹⁵N-ΔR₂ values for SH3^Mut. Experiments were conducted at 10 °C. Error bars: 1 SD.

Fig. S5.

CPMG relaxation dispersion data for the [¹⁵N/¹H_N/AILV-¹³CH₃]/deuterated-SH3^Mut sample in the presence and absence of GroEL. (A–C) Examples of experimental (circles) and best-fit (lines) CPMG relaxation dispersion profiles for (A) ¹⁵N, (B) ¹H_N, and (C) ¹³C_methyl as a function of CPMG field strength (ν_CPMG) in the presence (red, 120 μM in subunits) and absence (blue) of GroEL at 900 MHz (800 MHz for ¹³C_methyl) (A–C, Top rows) and 600 MHz (A–C, Bottom rows). All experiments were conducted at 10 °C.

Fig. S6.

Comparison of observed and calculated ΔR₂ values for [¹⁵N/¹H_N/AILV-¹³C_methyl]/deuterated-SH3^Mut in the presence of GroEL. (A) ¹⁵N-ΔR₂ at 900 MHz (Left) and 600 MHz (Right). (B) ¹³C_methyl-ΔR₂ at 800 MHz (Left) and 600 MHz (Right).

Fig. S7.

¹⁵N- and ¹³C_methyl-DEST profiles recorded on [¹⁵N/¹H_N/AILV-¹³CH₃]/deuterated-SH3^Mut in the presence of GroEL. (A and B) Examples of experimental (circles) and best-fit (lines) DEST profiles for (A) ¹⁵N and (B) ¹³C_methyl as a function of frequency offset from the ¹⁵N (119.5 ppm) and ¹³C (18 ppm) carrier, respectively. Data were recorded at two RF field strengths: 500 Hz (orange) and 750 Hz (purple) for ¹⁵N and 750 Hz (orange) and 1,500 Hz (purple) for ¹³C. All experiments were conducted at 10 °C.

Fig. S8.

Effect of the value of the rate constant $k_{\text{FI}}^{\text{G}}$ on ΔR₂. Shown are plots of simulated ¹⁵N-ΔR₂ as a function of $k_{\text{FI}}^{\text{G}}$ for $k_{\text{on}}^{\text{app}}$ = 6 s⁻¹, 60 s⁻¹, 600 s⁻¹, and 6,000 s⁻¹. The population of the GroEL-bound I-G state, p_I-G, is kept constant by scaling the rate constant k_off proportionately (k_off = 20 s⁻¹, 200 s⁻¹, 2,000 s⁻¹, and 2 × 10⁴ s⁻¹), and the remaining rate constants and bound R₂ values are as listed in Tables 2 and 3 of the main text, respectively. ΔΔR₂ represents the decrease in ΔR₂ between the protonated (“H”) sample and its deuterated counterpart (“D”).

Fig. S9.

Effect of the sign of Δϖ^F,I-G on the ¹⁵N relaxation dispersion profiles. ¹⁵N profiles are simulated for the four-state exchanging system shown in Fig. 3 of the main text (as well as in Eq. S1 of SI Theory), using the rate constants and populations for ¹⁵N-SH3^Mut listed in Table 2 of the main text for Δϖ^F,I = 1 ppm, 2 ppm, 3 ppm, and 4 ppm at a ¹H spectrometer frequency of 900 MHz. In each case, the profiles shown with black circles are calculated for Δϖ^F,I-G = 0 ppm; the profiles shown with blue circles and asterisks are calculated for Δϖ^F,I-G = +2 ppm and −2 ppm, respectively; and the profiles shown with red circles and asterisks are calculated for Δϖ^F,I-G = +4 ppm and −4 ppm, respectively.