Quantitative Exchange NMR-Based Analysis of Huntingtin–SH3 Interactions Suggests an Allosteric Mechanism of Inhibition of Huntingtin Aggregation

Abstract

Huntingtin polypeptides (htt^ex1), encoded by exon 1 of the htt gene and containing an expanded polyglutamine tract, form fibrils that accumulate within neuronal inclusion bodies, resulting in the fatal neurodegenerative condition known as Huntington’s disease. Htt^ex1 comprises three regions: a 16-residue N-terminal amphiphilic domain (NT), a polyglutamine tract of variable length (Q_n), and a polyproline-rich domain containing two polyproline tracts. The NT region of htt^ex1 undergoes prenucleation transient oligomerization on the sub-millisecond time scale, resulting in a productive tetramer that promotes self-association and nucleation of the polyglutamine tracts. Here we show that binding of Fyn SH3, a small intracellular proline-binding domain, to the first polyproline tract of htt^ex1Q₃₅ inhibits fibril formation by both NMR and a thioflavin T fluorescence assay. The interaction of Fyn SH3 with htt^ex1Q₇ was investigated using NMR experiments designed to probe kinetics and equilibria at atomic resolution, including relaxation dispersion, and concentration-dependent exchange-induced chemical shifts and transverse relaxation in the rotating frame. Sub-millisecond exchange between four species is demonstrated: two major states comprising free (P) and SH3-bound (PL) monomeric htt^ex1Q₇, and two sparsely populated dimers in which either both subunits (P₂L₂) or only a single subunit (P₂L) is bound to SH3. Binding of SH3 increases the helical propensity of the NT domain, resulting in a 25-fold stabilization of the P₂L₂ dimer relative to the unliganded P₂ dimer. The P₂L₂ dimer, in contrast to P₂, does not undergo any detectable oligomerization to a tetramer, thereby explaining the allosteric inhibition of htt^ex1 fibril formation by Fyn SH3.

Graphical Abstract

INTRODUCTION

CAG (encoding for glutamine) expansion, in excess of 35 repeats, within exon-1 of the huntingtin (htt) gene, is responsible for Huntington’s disease, a fatal, autosomal dominant, neurodegenerative condition.^1–3 Polypeptide fragments of huntingtin exon-1 (htt^ex1), generated by either proteolysis⁴ or incomplete mRNA splicing,⁵ form aggregates and fibrils that accumulate in neuronal inclusion bodies.^3,6 Htt^ex1 comprises three regions:^7–11 a N-terminal 16-residue amphiphilic region (NT), a polyglutamine repeat (Q_n), and a proline-rich domain (PRD) containing two polyproline tracts (P₁₁ and P₁₀). The NT region followed by as few as 7 glutamines undergoes transient oligomerization to a tetramer.¹² The initial prenucleation oligomerization events occur on the micro- to millisecond time scale and consist of two branches: a productive pathway leading to the formation of a sparsely populated helical coiled-coil tetramer formed by a dimer of dimers of the NT region, and a nonproductive pathway that yields an amorphous ensemble of partially helical dimers that do not undergo further oligomerization.^12–14 The NT tetramer serves to increase the effective local concentration of polyglutamine tracts, thereby promoting polyglutamine self-association and subsequent fiber formation.¹² Fibrillization of htt^ex1 can be prevented by inhibiting formation of the productive tetramer by a variety of mechanisms: oxidation of Met7 located at the center of the interface between the two dimers of the tetramer¹² spontaneously or catalytically by the use of light-activated titanium oxide nanoparticles;¹⁵ sequestration of monomer by chaperones, such as the chaperonin Hsp60, thereby effectively increasing the critical concentration required for nucleation;¹⁶ and binding of profilin to one or both polyproline tracts in the PRD, which indirectly blocks the productive tetramerization pathway, while leaving the nonproductive pathway unperturbed.¹³

Src homology 3 (SH3) domains of a number of intracellular proteins involved in signal transduction have been shown to interact with the PRD region of huntingtin both in vitro and in vivo.^17–21 Here we first show, using both NMR and a thioflavin T fluorescence assay, that binding of Fyn SH3 to htt^ex1Q₃₅, a construct bearing a 35-residue polyglutamine repeat, inhibits fibril formation. Then, using NMR techniques designed to probe exchange kinetics at atomic resolution,^22–25 we show, using a shorter htt^ex1Q₇ construct, that, in contrast to profilin, binding of Fyn SH3 to the PRD allosterically enhances the formation of the excited state, productive helical coiled-coil dimer but inhibits subsequent tetramerization, which is required to nucleate self-association of the polyglutamine tract.

RESULTS AND DISCUSSION

Fyn SH3 Prevents Aggregation and Fibril Formation of htt^ex1Q₃₅.

Figure 1 illustrates the effect of Fyn SH3 on the aggregation of htt^ex1Q₃₅, a construct containing a 35-residue polyglutamine tract. In the absence of SH3, 100 μM monomeric htt^ex1Q₃₅ self-associates to form large NMR invisible aggregates and fibrils with a t_1/2 of ~40 h at 5 °C, as monitored by the decrease in intensity of the amide proton envelope as the low molecular weight NMR observable species are converted to high molecular weight species line-broadened beyond the limit of detection (Figure 1A). Negative stain transmission electron microscopy (EM) images taken after 70 h reveal the presence of htt^ex1Q₃₅ fibrils (Figure 1B and SI Figure S1). In the presence of 0.3 mM SH3, however, a loss of only ~10% of the monomeric htt^ex1Q₃₅ signal is observed after 70 h (Figure 1A).

Figure 1.

Inhibition of aggregation and fibril formation of htt^ex1Q₃₅ by Fyn SH3. (A) Time course of the integrated intensity of the amide proton envelope of 0.1 mM htt^ex1Q₃₅ in the absence (red) and presence (blue) of 3 mM Fyn SH3, obtained from the first increment of serial ¹H–¹⁵N correlation experiments recorded at 600 MHz and 5 °C. The inset is a picture of the NMR tubes after 70 h in the absence (left) and presence (right) of SH3. (B) Negative stain EM of the NMR sample after 70 h in the presence of SH3. (C) Concentration dependence of aggregation kinetics of htt^ex1Q₃₅ monitored by ThT fluorescence at 37 °C. (D) Effect of SH3 on aggregation of 50 μM htt^ex1Q₃₅ monitored by ThT fluorescence at 37 °C.

The formation of cross-β structure containing fibrils of htt^ex1Q₃₅ was also monitored at 37 °C by fluorescence using a thioflavin T (ThT) binding assay²⁶ (Figure 1C and D). The kinetics of aggregation display a sigmoidal pattern with an initial lag phase dependent upon the concentration of htt^ex1Q₃₅ (Figure 1C). Addition of 0.5 mM SH3 to 50 μM htt^ex1Q₃₅ substantially prolongs the lag phase (from ~1 to 3 h) and flattens the growth phase, indicative of partial inhibition of the primary nucleation and elongation processes,^27,28 while addition of 3 mM SH3 completely inhibits aggregation (Figure 1D).

Elucidation of Equilibria and Kinetics of Binding of Fyn SH3 to htt^ex1Q₇ by NMR.

To understand the mechanism of inhibition of fibril formation by Fyn SH3, we proceeded to investigate by NMR the interaction of SH3 with htt^ex1Q₇, a construct with seven glutamines, that undergoes transient tetramerization but remains predominantly monomeric over a period of several weeks.¹³

Addition of htt^ex1Q₇ at natural isotope abundance to ¹⁵N-labeled Fyn SH3 results in ¹H_N/¹⁵N chemical shift perturbations (Δδ_H/N) in the canonical polyproline binding region comprising the RT loop, the N-terminal end of strand β₃, and the 3¹⁰ helix located between stands β₄ and β₅ (Figure 2A and B).²⁹ Addition of unlabeled Fyn SH3 to ¹⁵N/¹³C-labeled htt^ex1Q₇ results in line broadening of the three ¹Hα/¹³Cα proline cross-peaks (corresponding to the N- and C-terminal prolines with the remainder completely overlapped in a single cross-peak¹³), substantial Δδ_H/N shifts within the polyglutamine tract immediately N-terminal to the first polyproline repeat (P₁₁), smaller Δδ_H/N shifts immediately C-terminal to the P₁₁ tract, and minimal Δδ_H/N shifts either N- or C-terminal to the second polyproline repeat (P₁₀) (Figure 2C). We therefore conclude that Fyn SH3 binds exclusively to the first polyproline tract, P₁₁. Surprisingly, substantial Δδ_H/N shifts are also observed within the NT region that were not observed upon profilin binding to the polyproline tracts. These are not due to binding of Fyn SH3 to the NT region but arise from a long-range, indirect effect, as no significant Δδ_H/N shifts are observed upon addition of Fyn SH3 to either ¹⁵N-labeled htt^NT or htt^NTQ₇, neither of which contains the polyproline tracts (Figure S2).

Figure 2.

Chemical shift perturbation mapping of interaction sites between htt^ex1Q₇ and Fyn SH3. (A) Weighted ¹H_N/¹⁵N perturbation profile (Δδ_H/N = {[ΔδN²/5² + ΔδH²]/2}^1/2)³⁰ obtained for 0.16 mM ¹⁵N-labeled SH3 in the presence of 0.56 mM unlabeled htt^ex1Q₇. The inset shows the Δδ_H/N values for the side chain nitrogen atoms of W36 and N53. (B) Model of polyproline P₁₁ binding to Fyn SH3 derived from the NMR structure of a SH3-polyproline-containing peptide complex (PDB code 1AZG);²⁹ side chains of residues with Δδ_H/N > 0.22 ppm are colored in red (see Experimental Section for details of modeling). (C) Δδ_H/N profile for 0.1 mM ¹⁵N/¹³C-labeled htt^ex1Q₇ in the presence of 0.4 mM SH3. The domain organization of htt^ex1Q₇ is depicted above the plot, and the inset shows progressive line broadening of ¹³C cross sections through the ¹H/¹³Cα cross-peaks of the prolines upon addition of SH3. All NMR data were recorded at 5 °C at either 600 (A) or 900 (C) MHz.

To characterize the kinetics and equilibria of the interaction of htt^ex1Q₇ with Fyn SH3 in detail, we carried out the following experiments shown in Figure 3: exchange-induced ¹⁵N/¹³Cα shift (δ_ex) and on-resonance ¹⁵N-R_1ρ (at two spin-lock fields; ¹⁵N-R_2,eff) measurements on 0.1 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ as a function of Fyn SH3 concentration;^{14 15}N³¹ and ¹³Cα³² Carr–Purcell–Meiboom–Gill (CPMG) relaxation dispersion measurements on 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ in the presence of 0.3 and 0.6 mM Fyn SH3; and ¹H_N/¹⁵N δ_ex measurements on 160 μM ¹⁵N-labeled Fyn SH3 as a function of htt^ex1Q₇ concentration.

Figure 3.

Quantitative analysis of the kinetics and equilibria of Fyn SH3 binding to htt^ex1Q₇ and subsequent dimerization. (A) Examples of the dependence of δ_ex and R_2,eff (at two spin lock fields, 750 and 3000 Hz) measured for 0.1 mM ¹⁵N/¹³C-labeled htt^ex1Q₇ on the concentration of added unlabeled SH3. The data were acquired at 900 MHz. (B) Examples of ¹⁵N (800 MHz) and ¹³Cα (600 and 800 MHz) CPMG relaxation dispersion profiles for 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ in the presence of 0.3 mM (¹⁵N and ¹³Cα) and 0.6 (¹⁵N) mM unlabeled SH3. (C) ¹⁵N-δ_ex observed at 800 MHz for 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ in the presence of 0.3 mM unlabeled SH3. (D) Examples of the dependence of ¹⁵N- and ¹H_N-δ_ex measured at 600 MHz for 0.16 mM ¹⁵N-labeled SH3 on the concentration of added unlabeled htt^ex1Q₇. The experimental data are shown as circles in panels A, B, and D and as filled-in red circles in panel C; the global best-fit curves to the scheme shown in Figure 4 are displayed as continuous lines in panels A, B and D, and as filled-in blue circles in panel C. All data were recorded at 5 °C. The complete experimental data sets and best-fit curves are provided in Figures S3–S6.

The kinetic model that accounts for all the NMR data simultaneously is described by the four-state scheme shown in Figure 4. Note that the binding scheme in Figure 4 does not incorporate the transient oligomeric species sampled in the absence of Fyn SH3, as their total fractional population never exceeds 0.4% and 1% at the concentrations employed for the δ_ex/R_2,eff (0.1 mM htt^ex1Q₇) and CPMG relaxation dispersion (0.3 mM htt^ex1Q₇) experiments, respectively (Figure S7).

Figure 4.

Scheme for the binding of Fyn SH3 to htt^ex1Q₇ and subsequent dimerization viewed from the perspective of (A) htt^ex1Q₇ and (B) Fyn SH3. The bold font is used to denote magnetization of the isotopically labeled species. Δω_i are the differences in chemical shift to the major NMR observable, either monomeric htt^ex1Q₇ (P) or free SH3 (L). ki and k_–i are second-order association and first-order dissociation rate constants, respectively, and k_–4 = (k₁k₄k_–3k_–2)/(k_–1k₂k₃). The prefactors in front of the rate constants relate to the differential equations (eq 1) describing the time-dependence of magnetizations. The species populations, shown in parentheses in panel A, relate to 0.1 mM htt^ex1Q₇ in the presence of 0.3 mM SH3.

Viewed from the perspective of htt^ex1Q₇ (P, where the bold font is used to denote magnetization of the isotopically labeled species), Fyn SH3 (L) binds to htt^ex1Q₇ (P) to form a binary complex PL; PL self-associates to form a dimer P₂L₂. Dissociation of L from the P₂L₂ dimer generates the singly bound dimeric species P₂L, which is included in the scheme for completeness (Figure 4A). From the NMR perspective, P₂L comprises two different magnetic states, P₂L^B and P₂L^U, since the ligand-bound and ligand-free subunits of dimeric htt^ex1Q₇, respectively, are generally not magnetically equivalent. The time dependence of the magnetizations M_i for the htt^ex1Q₇ species is given by^33,34

$$\frac{\text{d}}{\text{d}t}\left[\begin{array}{c}{M}_{\text{P}}\\ M_{\text{PL}}\\ M_{{\text{P}}_2\text{L}}^{\text{B}}\\ M_{{\text{P}}_2\text{L}}^{\text{U}}\\ M_{{\text{P}}_2{\text{L}}_2}^{\text{B}}\end{array}\right]=\left[\begin{array}{ccccc}-\left\{k_1[\text{L}]+2k_2[\text{PL}]\right\}& k_{-1}& k_{-2}/2& k_{-2}/2& 0\\ k_1[\text{L}]& -\left\{k_{-1}+2k_2[\text{P}]\right\}& k_{-2}/2& k_{-2}/2& k_{-4}\\ 0& 2k_2[\text{P}]& -\left\{k_{-2}+k_3[\text{L}]\right\}& 0& k_{-3}\\ 2k_2[\text{PL}]& 0& 0& -\left\{k_{-2}+k_3[\text{L}]\right\}& k_{-3}\\ 0& 2k_4[\text{PL}]& k_3[\text{L}]& k_3[\text{L}]& -\left\{2k_{-3}+k_{-4}\right\}\end{array}\right]\left[\begin{array}{c}{M}_{\text{P}}\\ M_{\text{PL}}\\ M_{{\text{P}}_2\text{L}}^{\text{B}}\\ M_{{\text{P}}_2\text{L}}^{\text{U}}\\ M_{{\text{P}}_2{\text{L}}_2}^{\text{B}}\end{array}\right]$$ (1)

From the perspective of Fyn SH3, only a three-state scheme needs to be considered comprising L, PL, and P₂L₂ (Figure 4B) since one can assume that the shifts of bound Fyn SH3 are independent of the dimerization state of htt^ex1Q₇ given that the site of dimerization (NT domain) is distant from the SH3-binding polyproline tract (Figure 2C). The time dependence of the magnetizations M_i for the Fyn SH3 species is given by

$$\frac{\text{d}}{\text{d}t}\left[\begin{array}{c}{M}_{\text{L}}\\ M_{\text{PL}}\\ M_{{\text{P}}_2{\text{L}}_2}\end{array}\right]=\left[\begin{array}{ccc}-\left\{k_1[\text{P}]+k_3\left[{\text{P}}_2\text{L}\right]\right\}& k_{-1}& k_{-3}\\ k_1[\text{P}]& -k_{-1}& 0\\ k_3\left[{\text{P}}_2\text{L}\right]& 0& -k_{-3}\end{array}\right]\left[\begin{array}{c}{M}_{\text{L}}\\ M_{\text{PL}}\\ M_{{\text{P}}_2{\text{L}}_2}\end{array}\right]$$ (2)

It should be noted that the magnetization ($M_{{\text{P}}_2\text{L}}$) of the species P₂L does not enter directly into the scheme of Figure 4B and eq 2, as $M_{{\text{P}}_2\text{L}}$ does not enter into the differential equation describing the evolution of M_L (the second equation in eq S7 of the SI, where only [P₂L] is present). This is a direct consequence of the omission from our analysis of the process of dimerization of free htt^ex1, P, to form a ligand-free dimer P₂, P ↔ P₂ (Figure 4A), which is fully justified at the low concentrations of htt^ex1 used in this study (see above and Figure S7).

The full derivation of the time dependence of the magnetizations of the various species from the corresponding differential equations describing the time dependence of species concentrations is provided in the SI.

The complexity of the kinetic binding scheme shown in Figure 4A warrants invoking a number of simplifying assumptions. We note that when all seven independent rate constants (with k_–4 being expressed in terms of the other seven as k_–4 = k₁k₄k_–3k_–2/k_–1k₂k₃ since the overall scheme is a closed thermodynamic cycle) are optimized, the values of k₃ and k_–3 are ill-determined. Given that the polyproline stretch (the binding site for Fyn SH3) is well separated from the N-terminal portion of the htt^ex1Q₇ polypeptide chain, involved in helix formation-mediated oligomerization,¹² we considered it logical to assume that the association rate constant for the binding of Fyn SH3 to htt^ex1Q₇ is independent of the oligomerization state of the latter (k₁ = k₃). Likewise, the association rate constant of the htt^ex1Q₇ protomers P to form dimers was assumed to be independent of the ligation state of a protomer, that is, independent of whether htt^ex1Q₇ protomer is free or bound to Fyn SH3 (k₂ = k₄). These assumptions are all the more reasonable considering that the relevant association rate constants are expected to be largely diffusion-controlled. These assumptions reduce the number of variable global parameters (rate constants) in the fit from 7 to 5 (with k_–4 given by k_–2k_–3/k_–1). Further, the changes in chemical shifts of all species in the scheme of Figure 4A are expressed through (1) Δω_B = Δω_PL, representing the change in chemical shift exclusively due to ligand binding, and (2) $\Delta {\omega }_{\text{D}}=\Delta {\omega }_{{\text{P}}_2\text{L}}{}^{\text{U}}$, describing the change in chemical shift arising solely from dimerization of P, where the chemical shifts from the two processes are assumed to be additive. This assumption effectively decouples the process of dimerization from that of ligand binding with respect to the changes in chemical shifts. The assumed independence of the two processes from each other is supported by the amino acid sequence separation between the first polyproline tract and the NT oligomerization domain of htt^ex1Q₇.

Examples of the global fit to the experimental data are shown in Figure 3, and the complete data sets and corresponding best-fit curves are provided in Figures S3–S6. The values of the global optimized rate (k_i) and derived equilibrium dissociation (K_D) constants are given in Table 1, and the values of the optimized residue-specific chemical shift and transverse relaxation parameters (Δω and R₂, respectively) are provided in Tables S1–S4. The chemical shift differences relative to free htt^ex1Q₇ attributable to SH3 binding (Δω_B) and dimerization (Δω_D) are plotted in Figure 5. The values of Δω_D for both ¹⁵N and ¹³Cα are ~7- to 10-fold larger than those of Δω_B. The average ¹³Cα–Δω_B values within the NT region are small and positive (~0.25 ppm), indicative of a small increase in helical propensity upon ligand binding. The much larger average ¹³Cα–Δω_D values (~+2.4 ppm) are fully consistent with the formation of a coiled-coil helical dimer.

Table 1.

Optimized Values of Rate Constants and Derived Equilibrium Dissociation Constants^a

Rate Constants
k1 = k3 (M−1 s−1)	6.4 (± 0.7) × 107
k−1 (s−1)	1.6 (± 0.3) × 104
k2 = k4 (M−1 s−1)	4.5 (± 0.5) × 106
k−2 (s−1)	2.9 (± 0.3) × 104
k−3 (s−1)	7.2 (±1.9) × 103
k−4 (s−1)	1.4 (± 0.5) × 104
Equilibrium Dissociation Constants
KD1 (mM)	0.25 ± 0.05
KD2 (mM)	3.3 ± 1.0
KD3 (mM)	0.30 ± 0.03
KD4 (mM)	3.0 ± 0.7

^aThe equilibrium dissociation constants, expressed in terms of the rate constants, are given as follows: K_D1 = k_–1/k₁; K_D2 = k_–2/2k₂; K_D3 = 2k_–3/k₃; K_D4 = k_–4/k₄; and K_D1K_D3=K_D2K_D4 (see SI for the derivation).

Figure 5.

¹⁵N and ¹³Cα Δω profiles within the NT region of htt^ex1Q₇ attributable to binding of Fyn SH3 (Δω_B) to the first polyproline tract of the PRD and to dimerization of the NT (Δω_D) obtained from the global fit to the experimental NMR data. The Δω_B and Δω_D values are displayed as gray and black circles, respectively. Note that Gln19 exhibits no observable ¹³Cα exchange-induced shifts or ¹³Cα-CPMG relaxation dispersion.

A plot of the fractional populations of the P, PL, P₂L, and P₂L₂ species as a function of Fyn SH3 concentration is shown in Figure 6. At a total concentration of 0.1 mM htt^ex1Q₇, the population of free monomeric htt^ex1Q₇ (P) is depleted by ~70% and ~93% at SH3 concentrations of 0.6 and 3 mM, respectively; these values are consistent with the concentration dependence of inhibition of htt^ex1Q₃₅ aggregation by SH3 observed by ThT fluorescence shown in Figure 1D. Under the same conditions the populations of the monomeric SH3-bound state of htt^ex1Q₇ (PL) reach values of ~65% and ~90% at 0.6 and 3 mM SH3, respectively, while the corresponding populations of the P₂L₂ dimer are 2% and 4%, respectively. The population of the single SH3-bound dimer, P₂L, reaches a maximum of ~0.5% at 0.2–0.3 mM SH3 and decays thereafter to a value of only ~0.2% at 3 mM SH3.

Figure 6.

Simulation of the populations of free monomer htt^ex1Q₇ (P) and monomeric (PL) and dimeric (P₂L and P₂L₂) states of SH3-bound htt^ex1Q₇ as a function of total Fyn SH3 concentration. (A) [P]_total = 0.1 or 0.3 mM with [SH3]_total up to 0.6 mM. (B) [P]_total = 0.1 mM with [SH3]_total up to 3 mM. Note the population scale on the left panels ranges from 0 to 100%, while that on the right panels is from 0% to 6%. The equilibrium dissociation constants for the four-state binding scheme are given in Table 1.

While a two-state binding scheme (P ↔ PL) can account for the exchange-induced shift data, such a model fails to fit either the concentration-dependent R_2,eff data or the CPMG relaxation dispersion profiles (Figure S8). This is due to the ~10-fold larger chemical shift differences attributable to dimerization (Δω_D) versus SH3 binding (Δω_B). As a result, in the current sub-millisecond exchange regime, δ_ex is only slightly impacted by the sparsely populated dimeric states (P₂L₂ and P₂L), while relaxation dispersion (CPMG and R_1ρ) is highly sensitive to such species owing to the large chemical shift differences with respect to the observable species (P and PL) (Figure S9). It is also worth noting that a two-state fit to the δ_ex data for htt^ex1Q₇ yields a K_D value (210 ± 90 μM) that is the same within uncertainty as that for K_D1 (250 ± 50 μM) obtained from the full four-state fit to all the experimental NMR data, but the resulting values of Δω_B are overestimated by about 20%. We further note that a “full” kinetic model that includes dimerization of free P, P ↔ P₂, and the concomitant association of P₂ with L to form P₂L, P₂ + L ↔ P₂L (involving an additional dimeric state P₂ and three additional rate constants) was considered. However, the low concentrations of P (htt^ex1Q₇) used here make the acquired NMR data completely insensitive to these additional processes.

The association rate constant for SH3 binding (assumed to be equal for binding monomer or dimer, k₁ = k₃) is ~6 × 10⁷ M⁻¹ s⁻¹, as expected for a diffusion-limited process.³⁵ The association rate constant for SH3-bound htt^ex1Q₇ monomer (PL) to form the P₂L₂ dimer (k₂ = k₄) is an order of magnitude slower (~6 × 10⁶ M⁻¹ s⁻¹), which is also consistent with a diffusion-limited process involving higher molecular weight species.³⁶ Interestingly, k₂ is an order of magnitude larger than the association rate constant for self-association of free htt^ex1Q₇ (P) into productive dimer P₂,^13,14 which presumably can be attributed to the increased helical propensity of the NT region of PL relative to P. In addition, the dissociation rate constant of P₂L₂ into PL is ~3-fold slower than that for P₂ into P.^13,14 As a result, the stability of the P₂L₂ dimer is enhanced ~25-fold relative to the P₂ dimer.

The P₂L₂ Dimer Does Not Undergo Any Detectable Self-Association to a Tetramer.

The NMR experiments carried out under conditions where the concentration of htt^ex1Q₇ is kept constant and the concentration of SH3 is varied do not permit one to ascertain whether the P₂L₂ dimer undergoes further self-association to a P₄L₄ tetramer. Tetramerization can be readily investigated by examining the dependence of ¹⁵N-δ_ex on the concentration of ¹⁵N-labeled htt^ex1Q₇ under conditions where the first polyproline tract of htt^ex1Q₇ is close to fully saturated with SH3 (i.e., where the population of free htt^ex1Q₇ is 5% or less). Under such conditions, the population of the P₂L₂ dimer will be linearly dependent upon the concentration of PL, while that of a putative P₄L₄ tetramer will be dependent upon the cube of the concentration of PL (Figure 7A). If tetramerization does not occur at detectable levels, δ_ex will exhibit an approximately linear dependence on the total concentration of htt^ex1Q₇; however, if any detectable P₄L₄ tetramer is formed, the dependence of δ_ex on the total concentration of htt^ex1Q₇ will be curved. The experimental δ_ex data for ¹⁵N-labeled htt^ex1 Q₇ in the presence of 5 mM SH3 (corresponding to ~95% binding site saturation) is linear (Figure 7B). The expected concentration dependence of δ_ex, assuming the ¹⁵N chemical shifts for the dimer and tetramer are the same (given that these are largely influenced by secondary structure)¹² for various values (ranging from 15 μM to ≥1 mM) of the equilibrium dissociation constant, K_D5, for the putative equilibrium between P₄L₄ tetramer and P₂L₂ dimer are also shown in Figure 7B. From these data, a lower limit of ~1 mM is established for K_D5 with an upper bound of ~1% on the population of P₄L₄ at the highest concentration of htt^ex1Q₇ employed (0.4 mM). By way of comparison, the corresponding equilibrium dissociation constant for the dissociation of unliganded P₄ tetramer into unliganded P₂ dimer is ~25 μM.¹³ Thus, the stability of the putative P₄L₄ tetramer is reduced by a factor ≥40 relative to that of the P₄ tetramer.

Figure 7.

Probing putative tetramerization of htt^ex1Q₇ in the presence of Fyn SH3. (A) Scheme for the oligomerization of htt^ex1Q₇ upon binding of SH3 that considers the existence of a putative tetramer P₄L₄ formed by self-association of the P₂L₂ dimer and characterized by the equilibrium dissociation constant K_D5. The expression for the dependence of the population of P₄L₄ on the population of the bound monomer PL is provided. For simplicity the singly bound dimer, P₂L, shown in the scheme of Figure 4A has been omitted, as its equilibrium population is ~20 times smaller than that of P₂L₂. (B) Examples of the dependence of ¹⁵N-δ_ex on the concentration of ¹⁵N/¹³C-labeled htt^ex1Q₇ in the presence of 5 mM SH3. The experimental data, recorded at 800 MHz and 5 °C, are displayed as red circles, and simulations of ¹⁵N-δ_ex for a range of values of K_D5 are shown as dashed lines. The calculated curve for K_D5 ≥ 1 mM (red continuous line) is indistinguishable from that without tetramerization. Simulations were performed using the values of the rate constants provided in Table 1 and the values of Δω_B and Δω_D reported in Table S1. The ¹⁵N chemical shifts within the NT region of the P₂L₂ dimer and P₄L₄ tetramer were assumed to be the same (i.e., $\Delta {\omega }_{{\text{P}}_4{\text{L}}_4}=\Delta {\omega }_{{\text{P}}_2{\text{L}}_2}=\Delta {\omega }_{\text{B}}+\Delta {\omega }_{\text{D}}$). The populations indicated above the species in panel A correspond to those at the highest concentration of htt^ex1Q₇ (0.4 mM) employed in the experiments with a SH3 concentration of 5 mM. Under these conditions the population of free htt^ex1Q₇ (P) is only ~5%.

CONCLUSIONS

In summary, binding of Fyn SH3 to the first polyproline tract (P₁₁) of htt^ex1Q_n increases the helical propensity of the monomeric NT domain via a long-range allosteric effect, which in turn greatly enhances self-association of the NT domain into a coiled-coil helical dimer. The dimeric SH3-bound species (P₂L₂) is stabilized by over an order of magnitude relative to the transient P₂ dimer. However, whereas the P₂ dimer undergoes further self-association on the sub-millisecond time scale to form a thermodynamically quite stable tetramer (P₄; K_D ≈ 25 μM),¹³ which serves to nucleate fibril formation of the polyglutamine tracts by increasing their effective local concentration, further self-association of P₂L₂ is not detectable in our current experiments and the lower limit of the K_D for any putative P₄L₄ tetramer is ~1 mM.

In effect, binding of SH3 to the first polyproline tract of htt^ex1 results in the formation of monomeric and dimeric SH3-bound species that are nonproductive in terms of nucleation of polyglutamine tracts and subsequent fibril formation. The net result is that the binding of SH3 depletes the concentration of free htt^ex1 by over an order of magnitude, and hence of the productive P₄ tetramer, which constitutes the prequel to self-association and nucleation of polyglutamine tracts and subsequent fibril formation. One might speculate that further self-association of the P₂L₂ dimer is inhibited by steric interference of the SH3-bound P₁₁ polyproline tract, effectively masking the tetramer interface. Finally, the current work may suggest possible future avenues for the design of small-molecule allosteric inhibitors of htt^ex1 tetramerization and subsequent polyglutamine nucleation and fibril formation.

EXPERIMENTAL SECTION

Expression and Purification of htt^ex1Q₇ and htt^ex1Q₃₅ Proteins.

htt^ex1Q₇ and htt^ex1Q₃₅ were expressed in Escherichia coli (E. coli) as fusion proteins with the N-terminal immunoglobulin-binding domain of streptococcal protein G (GB1) and a factor Xa cleavage site. Protein expression, purification, and uniform isotope labeling (¹⁵N/¹³C and ¹⁵N/¹³Cα) as well as the cleavage of factor Xa to remove GB1 were carried out as described previously.¹³ Uniform ¹⁵N and ¹³C labeling was achieved by using ¹⁵NH₄Cl and ¹³C₆-d-glucose as the sole nitrogen and carbon sources, respectively. Fractional ¹³Cα labeling (with all other carbons at natural isotopic abundance) was achieved using [2-¹³C]-d-glucose as the sole carbon source; this labeling strategy produces proteins selectively fractionally ¹³Cα-labeled in all residues except leucine (where the Cα position is labeled at less than 10%) and isoleucines and valines (where ¹³Cα¹²Cβ and ¹³Cα¹³Cβ pairs are in a ratio of 1.2:1).³⁷ Following the final step of purification by reverse phase high-performance liquid chromatography (HPLC) using a preparative-scale C4 column (Waters), the lyophilized fractions of htt^ex1 were dissolved in a 1:1 trifluoroacetic acid (TFA)/hexafluoroisopropanol (HFIP) solvent mixture to remove any pre-existing aggregates that can promote the aggregation of monomeric polypeptides. Complete removal of the TFA/HFIP solvent mixture was carried out by directing a stream of N₂ gas into the flask through a glass pipette overnight. The peptide film was then dissolved in 0.1 mM TFA, lyophilized, and stored at −30 °C. The identity and purity of htt^ex1Q₇ and htt^ex1Q₃₅ were confirmed by electrospray ionization mass spectrometry (ESI-MS).

Expression and Purification of Wild-Type Fyn SH3 Domain.

Expression and purification of unlabeled and ¹⁵N-labeled Gallus gallus tyrosine kinase Fyn SH3 domain followed the procedure described previously.^38,39 The codon-optimized vector for expression of Fyn SH3 included an N-terminal His₆-tag and a tobacco etch virus (TEV) protease cleavage site.³⁹ Briefly, E. coli cells containing the plasmid for Fyn SH3 overexpression were grown at 37 °C to an OD₆₀₀ ≈ 0.6, induced with 1 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG) and grown for a further 6 h at 30 °C. Upon harvesting via centrifugation, the cells were resuspended in a denaturing buffer containing 6 M urea and lysed by sonication. His-tagged Fyn SH3 was initially isolated by passing the cleared lysate through a 5 mL HisTrap HP column installed on an AKTA Explorer FPLC system (GE Healthcare). Refolding of Fyn SH3 was achieved by step dialysis and removal of the His₆ tag by addition of TEV protease. Following proteolytic cleavage with TEV overnight, the protein was further purified using Ni²⁺ affinity chromatography. The flow-through containing Fyn SH3 was then concentrated and subsequently exchanged into a NMR buffer (20 mM monobasic sodium phosphate, pH 6.5, 50 mM NaCl, and 10% D₂O/90% H₂O v/v) by gel filtration using a Superdex75 (GE Healthcare) column. Purified Fyn SH3 was stored at −30 °C. The molecular weight of purified Fyn SH3 samples was confirmed by mass spectrometry.

Samples for NMR and Fluorescence Measurements.

Samples of htt^ex1Q₇ and htt^ex1Q₃₅ for NMR and fluorescence measurements were prepared by first dissolving an aliquot of the purified peptides in a 13.8 mM monobasic sodium phosphate buffer, pH 4.6, containing 50 mM NaCl and 10% D₂O/90% H₂O (v/v). Protein concentration was adjusted to a maximum of 3 mM for htt^ex1Q₇, and less than 0.2 mM for htt^ex1Q₃₅. The pH of the buffer was subsequently adjusted to 6.5 by adding dibasic sodium phosphate for a final sodium phosphate concentration of 20 mM. Peptide concentrations were determined by UV absorbance at 205 nm (neither peptide contains tryptophan or tyrosine residues).⁴⁰ Samples of Fyn SH3 were dissolved in a 20 mM monobasic sodium phosphate buffer, pH 6.5, containing 50 mM NaCl and 10% D₂O/90% H₂O (v/v).

Visualization of Fibrils by Electron Microscopy.

A sample containing 0.1 mM htt^ex1Q₃₅ was allowed to fibrillize in a Shigemi NMR tube at 5 °C for ~70 h. Following aggregation, htt^ex1Q₃₅ (comprising both insoluble and soluble fractions) was recovered from the NMR tube, and an aliquot of ~10 μL of the sample was blotted onto carbon-coated grids (ultrathin carbon film/holey carbon; Ted Pella) and incubated for 1 min. The grids were subsequently rinsed with deionized H₂O and stained with 2% uranyl acetate for 20 s. EM images were obtained using a FEI Tecnai T12 electron microscope equipped with a Gatan US 1000 CCD camera at 120 kV.

Thioflavin T (ThT) Fluorescence Assays.

The kinetics of aggregation was monitored by ThT fluorescence on samples of htt^ex1Q₃₅ with protein concentrations ranging from 5 to 100 μM, in the absence and presence of Fyn SH3 (0.5 and 3 mM). Each sample was distributed among three wells with a total volume of 150 μL. All experiments were performed in an assay buffer (20 mM sodium phosphate, 50 mM NaCl, pH 6.5) with 10 μM ThT in a 96-well microplate (Microplate Greiner). ThT fluorescence was measured every 5 min using a bottom read-out with constant, orbital 3 mm amplitude shaking. Fluorescence measurements were performed at 37 °C using an Infinite 200 PRO plate reader (Tecan Group Ltd., Switzerland) with a 440 nm excitation filter and a 480 nm emission filter. The background signal from the buffer with 10 μM ThT was subtracted from all data.

NMR Spectroscopy.

All NMR experiments were recorded at 5 °C on 600, 800, and 900 MHz Bruker Avance-III spectrometers equipped with TCI triple-resonance z-axis gradient cryogenic probes. All NMR data were processed using the NMRDraw/NMRPipe programs and associated software.⁴¹ In the case of exchange-induced chemical shift measurements, the time domain in the indirect (¹⁵N or ¹³C) dimension was extended 2-fold through the application of sparse multidimensional iterative line-shape-enhanced (SMILE) reconstruction.⁴²

Measurements of ¹³Cα and ¹⁵N Exchange-Induced Chemical Shifts (δ_ex).

The changes in ¹³Cα and ¹⁵N chemical shifts measured on 0.1 mM ¹⁵N/¹³C-labeled htt^ex1Q₇ in the presence of increasing amounts of unlabeled Fyn SH3 were obtained from 2D constant-time ¹H–¹³C and ¹H–¹⁵N heteronuclear single quantum coherence (HSQC) spectra recorded at 900 MHz as described previously.^12,13 The ¹⁵N/¹³Cα exchange-induced chemical shifts (δ_ex) were calculated as δ_ex(i) = δ_obs(i) – δ_ref, where δ_obs(i) is the chemical shift observed at Fyn SH3 concentration i, and δ_ref is the chemical shift of 0.1 mM htt^ex1Q₇ in the absence of Fyn SH3. The changes in ¹⁵N chemical shifts of 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ upon addition of 0.3 mM unlabeled Fyn SH3 were measured at 800 MHz.

The changes in ¹⁵N and ¹H_N chemical shifts of 0.16 mM ¹⁵N-labeled Fyn SH3 upon addition of increasing amounts of unlabeled htt^ex1Q₇ were obtained from ¹H–¹⁵N HSQC spectra recorded at 600 MHz. The values of ¹⁵N-δ_ex and ¹H_N-δ_ex for Fyn SH3 were obtained in the same manner as for htt^ex1Q₇ but using the chemical shifts of 0.16 mM Fyn SH3 as the reference (δ_ref).

Measurements of ¹⁵N Transverse Spin Relaxation rates (¹⁵N-R_2,eff).

The values of ¹⁵N-R_2,eff for 0.1 mM ¹⁵N/¹³C-labeled htt^ex1Q₇ in the absence and presence of increasing amounts of Fyn SH3 were obtained from on-resonance ¹⁵N-R_1ρ and ¹⁵N-R₁ data measured at 900 MHz using the pulse schemes and procedures described previously.^39,43 R_2,eff is calculated as (R_1ρ − R₁ cos² θ)/sin²θ, where θ is the angle between the direction of the effective spin-lock field and the z-axis of the laboratory frame (external magnetic field B₀). At each concentration of Fyn SH3, on-resonance ¹⁵N-R_1ρ relaxation rates were measured with RF field strengths of 750 and 3000 kHz using a two-time-point measurement.¹⁴

¹⁵N/¹³Cα-CPMG Relaxation Dispersion Experiments.

¹⁵N-CPMG relaxation dispersion experiments were recorded at 800 MHz on samples of 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇ in the absence and presence of 0.3 mM and 0.6 mM Fyn SH3 using a pulse scheme that quantifies the relaxation rates of in-phase ¹⁵N coherences.³¹ The relaxation period was set to 60 ms, and the following CPMG field strengths (ν_CPMG) were used: 0 (reference experiment), 17, 33, 67, 100, 133, 167, 200, 267, 300, 367, 433, 500, 583, 667, 750, 833, and 1000 Hz. Continuous wave (CW) decoupling of amide protons during the relaxation period was applied with an RF field strength of 11 kHz.

¹³Cα-CPMG relaxation dispersion profiles were acquired using a previously described pulse scheme³² with ¹Hα CW decoupling (RF field strength of 12 kHz) during the relaxation period. ¹³Cα-CPMG relaxation dispersion experiments on 0.3 mM ¹⁵N/¹³Cα-labeled htt^ex1Q₇, in the presence of 0.3 and 0.6 mM unlabeled Fyn SH3, were recorded at 600 and 800 MHz using a relaxation period of 20 ms and CPMG field strengths of 0 (reference experiment), 50, 100, 150, 200, 250, 300, 400, 500, 600, 700, 900, 1000, 1200, 1300, 1500, 1750, and 2000 Hz.

Docking of P₁₁ Polyproline Tract to Fyn SH3.

The binding of the P₁₁ polyproline tract to Fyn SH3 was modeled from the NMR structure of the complex of human Fyn SH3 bound to a synthetic proline-rich peptide derived from residues 91–104 of the P85 subunit of PI3-kinase (PDB code 1AZG)²⁹ as follows. First the NMR coordinates of native Gallus gallus SH3 (PDB code 2LP5)⁴⁴ were superimposed on those of human Fyn SH3 (PDB code 1AZG)²⁹ using the “align” command in the PyMol Molecular Graphics System (version 2.0, Schröder LLC). Then the Val39, Pro53, and Val55 mutations in the Gallus gallus Fyn SH3 coordinates were replaced in PyMol by the wild-type residues Ala, Asn, and Val, respectively. Finally, all non-proline residues in the bound peptide sequence of the human Fyn SH3-peptide complex were substituted with proline (in the trans conformation) and the length of the bound peptide was reduced to 11 residues by truncating the three C-terminal residues of the peptide. The ribbon diagrams in Figure 2B were created in PyMol.