Molecular basis of Q-length selectivity for the MW1 antibody–huntingtin interaction

Abstract

Huntington’s disease is caused by a polyglutamine (polyQ) expansion in the huntingtin protein. Huntingtin exon 1 (Httex1), as well as other naturally occurring N-terminal huntingtin fragments with expanded polyQ are prone to aggregation, forming potentially cytotoxic oligomers and fibrils. Antibodies and other N-terminal huntingtin binders are widely explored as biomarkers and possible aggregation-inhibiting therapeutics. A monoclonal antibody, MW1, is known to preferentially bind to huntingtin fragments with expanded polyQ lengths, but the molecular basis of the polyQ length specificity remains poorly understood. Using solution NMR, electron paramagnetic resonance, and other biophysical methods, we investigated the structural features of the Httex1–MW1 interaction. Rather than recognizing residual α-helical structure, which is promoted by expanded Q-lengths, MW1 caused the formation of a new, non-native, conformation in which the entire polyQ is largely extended. This non-native polyQ structure allowed the formation of large mixed Httex1-MW1 multimers (600–2900 kD), when Httex1 with pathogenic Q-length (Q46) was used. We propose that these multivalent, entropically favored interactions, are available only to proteins with longer Q-lengths and represent a major factor governing the Q-length preference of MW1. The present study reveals that it is possible to target proteins with longer Q-lengths without having to stabilize a natively favored conformation. Such mechanisms could be exploited in the design of other Q-length specific binders.

Huntington’s disease is an autosomal-dominant neurodegenerative disorder caused by a polyglutamine (polyQ) expansion in Huntingtin exon 1 (Httex1) of the mutant huntingtin protein (mHtt). N-terminal Htt fragments, including Httex1, are commonly observed in vivo and increasing Q-lengths make them increasingly aggregation prone (1, 2). Httex1 with polyQ expansions beyond 36 residues is sufficient to cause toxicity in cells and animal models, and it is therefore widely used for studying Huntington’s disease biology and pathology (2, 3, 4).

Httex1 can take up several conformational states in vivo and in vitro, including monomers, various oligomers, protofibrils, as well as different types of fibrils (5, 6, 7, 8, 9, 10, 11, 12, 13). While the precise nature of the toxic species is still a matter of debate, most studies have concluded that large inclusions are protective and that oligomers, protofibrils, or smaller, less bundled fibrils represent the primary toxic species (6, 14, 15, 16, 17). In addition, it has also been suggested that polyQ expansions could potentially alter the polyQ structure of the monomer, thereby introducing toxic conformations (18). Antibodies and other conformationally specific binders could become important to further our understanding of the toxic species and to potentially interfere with the aggregation pathways (19, 20, 21). One of the Htt antibodies that has received significant attention is MW1 (22). MW1 is a monoclonal antibody that recognizes expanded polyQ in monomeric Httex1 and strongly prefers longer Q-lengths (22, 23, 24). Initially, this preferential binding, which was also observed for the 1C2 antibody, was explained by the toxic threshold model (25). According to this model, the longer polyQ in the monomer adopts a toxic conformation that is different from that of normal length polyQ.

MW1 has also been used in time-resolved fluorescence readouts (TR-FRET). In these experiments, MW1 and an additional antibody, 2B7, were used as FRET pairs, allowing for a highly sensitive detection of soluble huntingtin in biological samples (26). TR-FRET can distinguish between different Q-lengths because shorter Q-lengths have a FRET signal that strongly decreases with temperature, while the corresponding signal from huntingtin with longer Q-lengths has only a minor temperature dependence (27). This has led to the suggestion that Httex1 with longer Q-lengths might have reduced flexibility that resists temperature-dependent structural changes, and compounds have been screened to counteract this putative conformational property (28). However, the notion that polyQ antibodies recognize a different, potentially toxic conformation specific to long polyQ lengths has been challenged. Several studies concluded that MW1 can bind to Httex1 with a large range of Q-lengths (23, 24, 29). Any differences in binding affinity were thought to be mainly caused by a larger number of MW1 binding sites available in proteins with longer Q-lengths. According to this linear lattice model (23), differential MW1 binding to Httex1 may therefore not depend on a particular structure formed by Httex1, but the number of available Gln residues.

More recently, the structures of Httex1 with different Q-lengths have been investigated in several biophysical studies (8, 10, 30, 31). Httex1 has three different domains, an N-terminal N17, a central polyQ, and a C-terminal proline-rich domain (PRD). It was found that the PRD contains a mixture of polyproline two helical and random coil structure regardless of Q-lengths. The N17 has partial α-helical structure, which extends into the N-terminal portion of the polyQ (8, 10, 31). While the structures of Httex1 with different Q-lengths are relatively similar, it was noted that longer Q-lengths shift the equilibrium between random coil and α-helical structure in the polyQ region toward α-helical structure (9). However, this increase in α-helical structure does not provide a good explanation for the enhanced binding of MW1 to expanded Q-lengths, because crystallographic studies of MW1 Fab fragments bound to polyQ peptides revealed an extended polyQ structure (32). Thus, unless MW1 is somehow also able to recognize α-helical polyQ regions, these results are inconsistent with the notion that MW1 specifically recognizes a structure that is preferentially present in Httex1 with expanded Q-length.

Here, we investigated the structural features that allow MW1 to be sensitive to polyQ lengths. Special emphasis was placed on any relationship between more pronounced α-helical structure at longer Q-lengths (8) and the nature of the MW1-bound structure. Toward this end, we used the native, bidentate MW1 rather than its monodentate Fab fragments that were mainly utilized in biophysical studies. Conceptually, the bidentate nature of MW1 might also be able to contribute to the recognition of longer Q-lengths proteins through entropy saving. The structural features of Httex1 with different Q-lengths were studied using electron paramagnetic resonance (EPR) spin-label mobility and distance analysis as well as protein NMR spectroscopy. The obtained structural information was supplemented by gel filtration analysis to determine the sizes of the formed complexes as well as calorimetry measurements to determine complex affinities.

Results

MW1 binds monomeric and not fibrillar Httex1

Prior studies suggested that MW1 preferentially interacts with monomeric rather than aggregated Httex1 (23, 29, 33). In agreement with this notion, we found that MW1 has a diffuse cytoplasmic staining pattern in HEK293T cells expressing Httex1(Q72)RFP (Fig. 1A). Importantly, MW1 did not stain large Httex1 inclusions. Inasmuch as Httex1 can form different aggregates that range in size from the inert large aggregates containing bundled fibrils to the more toxic, short, and unbundled protofibrils, we next sought to test whether MW1 might be able to bind to some fibril types. MW1 reactivity to different Httex1 conformers was therefore tested in dot blots (Fig. 1B). While reactivity to Httex1(Q46) monomer was strong, little or no binding could be detected regardless of whether protofibrils (P-fibrils), unbundled (T-fibrils), or bundled (N-fibrils) fibrils were used (6). Thus, MW1 does not seem to recognize polyQ regions that have undergone conversions into β-sheet structure.

Figure 1.

MW1 preferentially binds to monomeric Httex1.A, HEK293T cells were transfected with Httex1Q72RFP and stained with the MW1 antibody. Twenty-four hours after transfection, bright and round Q72RFP (red) aggregates were seen in some cells (asterisks, middle panel). These aggregates showed very low reactivity with MW1 (green, see left panel). Other cells with lower Q72RFP expression showed a diffuse pattern within the cytoplasm. These cells showed more overlap with MW1, as can be seen in the orange signal in the right panel. B, the conformational specificity of MW1 binding was tested using dot blotting. Monomeric Httex1(46Q) or different fibril types of Httex1(Q46) as indicated in the figure were spotted on nitrocellulose membrane (5 ng each). Fluorescently labeled antibody (IRDye-800 anti-mouse IgG) was used as secondary antibody to detect MW1. The top row shows binding in the presence of both the primary and the secondary antibody while the bottom row shows the control in the presence of the secondary antibody alone. No binding was observed in the presence of the secondary antibody alone. Httex1, Huntingtin exon 1.

To ensure the use of monomeric Httex1 in our biophysical and biochemical experiments, we fused thioredoxin to the N terminus of Httex1 and employed protein concentrations in the low micromolar range. This strategy was previously employed to characterize monomeric Httex1 in solution (8) and its interaction with MW1 (29). To examine the influence of polyQ lengths on MW1 binding, we studied the thioredoxin fusion of Httex1 with 25 and 46 glutamines, referred to as Httex1(Q25) and Httex1(Q46), respectively.

MW1 binds to a central portion of polyQ in Httex1(Q25) and Httex1(Q46)

To understand the MW1 interactions, we first mapped the MW1 binding sites on Httex1(Q25) and Httex1(Q46) using continuous wave EPR. The detection of binding using EPR is based on the reduction in mobility that occurs in regions where the rather dynamic Httex1 monomer becomes ordered upon interaction with MW1. This is illustrated with the EPR spectra of Httex1(Q46) and Httex1(Q25) spin-labeled at position 35 (35R1) in the presence of increasing amounts of MW1 (Fig. 2A). Dose-dependent spectral changes were observed for 35R1 in both Q-lengths. In the absence of MW1, 35R1 gave rise to sharp lines for both Q-lengths, consistent with the substantial disorder of this region in the unbound monomer (8). Upon MW1 interaction, the intensity of the sharp lines decreased, while additional spectral components of low mobility appeared (see arrows in Fig. 2A). The effect of MW1 was nearly complete at a molar Httex1(Q46):MW1 ratio of 1:2, while a ratio of 1:3 was required for Httex1(Q25) for the concentrations used. In contrast, no significant spectral changes were observed for control sites in N17 or PRD (Fig. 2A). To localize the MW1 binding sites throughout Httex1(Q46) and Httex1(Q25), we performed analogous experiments with additional spin-labeled derivatives under saturating conditions of MW1. As shown in Fig. S1, the largest spectral changes were observed for sites in the polyQ region, indicating that this region engages in important contacts with MW1. Next, we quantified the local effect of MW1 binding by determining the MW1 dependent amplitude change for each derivative. For Httex1(Q46), residues 25 to 60 had the largest amplitude reduction upon MW1 binding (Fig. 2B). This stretch of ∼35 amino acids located in the central portion of the polyQ region therefore represents the primary binding region for MW1. Prior EPR studies found that mobility effects caused by local binding interactions or structural ordering can dampen the dynamics of the ∼10 neighboring amino acids (34, 35). In agreement with these studies, an analogous mobility gradient was observed here (Fig. 2B), as EPR amplitude effects gradually tapered off from residues 25 to 15 and from residues 60 to 70. The involvement of these “transition” regions therefore appears to be indirect by virtue of the nearby binding sites. Outside of these transition regions, little or no spectral changes were observed. It should be emphasized that seven residues in the N-terminal transition region and three residues in the C-terminal transition region are glutamines. Thus, MW1 does not seem to have an equal affinity for all Gln residues, preferring the central ∼35 residues of the polyQ.

Figure 2.

Mapping of the MW1 binding sites in Httex1(Q46 and Q25) using EPR.A, representative EPR spectra for sites in the N17, polyQ, and PRD of Httex1(Q46), and Httex1(Q25) in the absence or presence of varying amounts of MW1 antibody. 10 μM of Httex1 were used with in the presence of 0 (gray), 10 (blue), 20 (green), and 30 (red) μM MW1. Arrows indicate the immobile components that appear in the spectra of the 35R1 derivatives in the presence of MW1. B and C, the changes in the EPR central line amplitude are expressed by the ratio of the amplitude of the spin labeled Httex1Q46 (B) or Httex1Q25 (C) in the presence or absence of MW1. The x-axis indicates the specific labeling position. Spectra are provided in Figures 1A and S1. EPR, electron paramagnetic resonance; Httex1, Huntingtin exon 1; polyQ, polyglutamine; PRD, proline-rich domain.

Next, we performed analogous experiments with Httex1(Q25). Again, we found that binding was strongest in the central portion of the polyQ while the N- and C-terminal regions of the polyQ were in a transition region. The N17 and PRD did not show significant binding effects (Figs. S1B and 2C). Thus, the binding preference for the central portion of the polyQ seems independent of Q-length. To further verify that the reduced affinity for the N-terminal portion of the polyQ was not caused by the N-terminal Trx fusion partner, we generated tag-free Httex1(Q25) with a label at position 21. As shown in Fig. S1C, the resulting spectrum of this derivative bound to MW1 was indistinguishable from that of the fusion protein labeled at the same position. Thus, the lack of a strong binding interaction in the N-terminal region of the polyQ was not caused by the Trx fusion partner.

Approximately 27 polyQ residues appear directly MW1-bound at a Httex1:MW1 ratio of 1:2

The EPR spin labels allowed sequence-specific information. Owing to the large number of structurally equivalent Gln residues in polyQ, it was also possible to observe them by solution NMR spectroscopy at a relatively low protein concentration of 14 μM (Fig. 3A). When bound to MW1 (∼150 kDa), Gln resonances are not observable any more by solution NMR, because of the large size of the complex. At a 14 μM concentration of Trx-Httex1(Q46) and Httex1:MW1 ratio of 1:2, complex formation was essentially complete (>99%) based on the affinity of this interaction (vide infra), and the ratio of signal intensities between bound and free Gln states, termed I/I₀, provides an estimate of the number of MW1-bound Gln residues within the polyQ region. We first note that for a Trx-Httex1(Q46) linker residue, I/I₀ was reduced by only 10% (Fig. 3B), validating a high decoupling of the Trx fusion protein from Httex1(Q46). Binding of MW1 reduced I/I₀ of the Gln side-chain H₂^N signals by approximately 60% (Fig. 3B). That is, approximately 27 polyQ residues were nominally bound. At the backbone level, the H^N resonances were attenuated by 90% (Fig. 3B). This difference illustrates that, perhaps as to-be-expected, backbone motion is significantly dampened not only for the approximately 27 Gln residues in direct side chain contact with MW1, but also for adjacent residues. Consequently, we regard the number of 27 as more representative of the number of directly bound polyQ residues at Httex1:MW = 1:2. Taken together with the EPR data, these results indicate that the overall MW1 binding region is on the order of 35 Gln residues long and that approximately 27 of these residues are in direct contact with the antibody.

Figure 3.

Effect of MW1 binding on the solution NMR spectrum of the Httex1(Q46) polyQ region.A, two-dimensional ¹⁵N-¹H^N correlation spectra of 14 μM ¹⁵N-labeled Trx-Httex1(Q46) were recorded in the absence (top panel) or presence (bottom panel) of 28 μM MW1 at 10 °C and 700 MHz. Spectral regions centering on the H^N backbone and H₂^N side-chain signals of polyQ are shown. B, the intensity ratios of the depicted signals in the presence or absence of MW1, termed I/I₀, were quantified for molar Httex1(Q46):MW1 ratios of 1:0 and 1:2. Control refers to a signal from the linker connecting Trx and Httex1(Q46). Httex1, Huntingtin exon 1; polyQ, polyglutamine.

MW1 binding induces a coil to extended structure transition in the polyQ region of Httex1(Q46)

To determine whether MW1 recognizes a particular structure present in Httex1 or whether it alters Httex1 structure, we performed distance measurements between pairs of spin-labeled residues in the polyQ region. In the presence of MW1, pulsed EPR (DEER) revealed distances for the 25R1/35R1, 35R1/45R1, and 40R1/50R1 pairs that were within 2 Å of the distances expected for a fully extended structure (number of amino acids × 3.5 Å/amino acid, Fig. 4 and Table 1). Much shorter distances with broad distance distributions were observed in the absence of MW1 (Table 1). This is consistent with the more coiled structure previously described for the polyQ region of the monomer, which is in a dynamic equilibrium between α-helical and random coil structure (8, 10, 31). Thus, when bound to MW1, the polyQ of Httex1 takes up a largely extended structure. This structure is induced by MW1, and it is not dominant in native Httex1(Q46).

Figure 4.

Spin-label intramolecular distances from four-pulse DEER experiments.A, C, and E, represent dipolar evolution data for Httex1(Q46) 25R1-35R1, 35R1-45R1, and 40R1-50R1 monomers. B, D, and F, represent the respective data for Httex1(Q46) 25R1-35R1, 35R1-45R1, and 40R1-50R1 monomers in the presence of the antibody MW1. The black traces in the left panels show the experimental data, while the red curves are from fits obtained based on Gaussian distributions. The gray traces in the panels on the left, A, C, and E, represent raw data without background correction. No background correction was required for the data obtained in the presence of MW1, indicating that the spin labeled sites are kept far apart from each other (on the EPR scale) by the antibodies. DEER, pulsed EPR; EPR, electron paramagnetic resonance; Httex1, Huntingtin exon 1.

Table 1.

Comparison of DEER distances between Httex1(Q46) monomers alone, Httex1(Q46) monomers in presence of MW1, and theoretical distance for an extended structure

Double labeled Httex1(Q46) positions	Distance between spin labels in Httex1(Q46) [Å]	Distance between spin labels in Httex1(Q46) with MW1 [Å]	Distance in extended structure [Å]
25R1-35R1	∼16	33	35
35R1-45R1	∼23	33	35
40R1-50R1	∼16	34	35

In addition, the DEER data also revealed a more subtle point about MW1 binding. At the concentrations used in DEER experiments, it is common that all proteins, including proteins that do not self-interact, have background signal due to intermolecular spin interactions. This is caused by a fraction of molecules that are in close enough proximity to give rise to a DEER signal. While the typical background subtraction was performed for the monomeric Httex1 in solution (Fig. 4, A, C, and E, gray), no background subtraction needed to be performed for the MW1-bound proteins (Fig. 4, B, D, and F). This unusual feature indicated that MW1 binding prevented a proximity between Httex1 molecules close enough to be detected by DEER. In other words, MW1 acted as a spacer to keep Httex1 molecules apart from each other (>60 Å between labels).

MW1 binding causes oligomerization for proteins with expanded Q-lengths

Next, we performed gel filtration experiments to estimate the sizes of MW1 complexes with Httex1 of different Q-lengths. As shown in Figure 5, MW1 alone eluted at an approximate molecular weight of ∼150 kD, while the individual Httex1 fusion proteins had an apparent molecular weight of just under 50 kD, consistent with prior reports (29). Upon incubation of Httex1(Q16) with MW1 (1:1 ratio), two main peaks were observed corresponding to unbound antibody (150 kD) as well as a peak with a molecular weight between 200 and 300 kD, consistent with MW1 bound to one or two Httex1(Q16) molecules. When the same experiments (at 1:1 ratios) were performed with Trx-Httex1(Q25) and Trx-Httex1(Q46), progressively larger complexes were formed. For Trx-Httex1(Q25), we observed a relatively broad peak ranging from ∼200 to 400 kD. While the smaller molecular weights are consistent with one MW1 bound to one or two Httex1 molecules, the larger molecular weights likely require at least two MW1 molecules. This effect was even more pronounced for Trx-Httex1(Q46), where peaks around 600 to 700, 1200 to 1400, and ∼2900 kD were observed, indicating that some complexes can be larger, containing several MW1 molecules.

Figure 5.

Gel filtration chromatography for mixtures of MW1 and Httex1 with different Q-lengths. The complexes were incubated at 1:1 for Httex1Q16, 0.5:1 and 1:1 for both constructs Httex1Q25 and Httex1Q46. The concentration for Httex1 was 10 μM for all experiments. The vertical lines indicate the approximate molecular weights estimated based on molecular weight standards. Httex1, Huntingtin exon 1.

When the gel filtration experiments were repeated at a Httex1:MW1 ratio of 1:0.5, the elution profile for Httex1(Q25) gave a main peak around 200 to 300 kD, consistent with the binding of the bidentate MW1 antibody to one or two Httex1 molecules. In contrast, negligible amounts of this complex formed in the case of Httex1(Q46), which instead gave rise to larger oligomeric complexes, also detected at the 1:1 ratio. However, some Httex1(Q46) remained unbound. This suggests that the large complexes are favored over the interaction of one MW1 with two Httex1(Q46) molecules, which might have otherwise been expected as a main product based on the Httex1(Q46)-MW1 stoichiometry of 1:0.5. Considering that these larger complexes are more favorable than the small complexes, oligomerization must be energetically quite favorable, and at least be partially responsible for the enhanced affinity of MW1 for proteins with longer Q-lengths.

MW1 binding affinity is polyQ-length dependent

To gain a better understanding of the Httex1-MW1 binding energetics, we performed isothermal calorimetry (ITC) titrations of 0.4 μM Httex1(Q25) and Httex1(Q46) with MW1. For Httex1(Q46), a sigmoidal binding curve could be obtained (Fig. 6A) that indicated a Httex1:MW1 binding stoichiometry of 1:2.2 and affinity of (2.6 ± 0.7)·10⁷ M⁻¹ (Table 2). Although the binding epitope of Httex1(Q25) will be highly similar for at least the binding of the first Fab arm of MW1, no binding saturation and, hence no stoichiometry information, could be obtained (Fig. 6B). Crudely assuming a Httex1(Q25):MW1 binding stoichiometry of 1:1 based on approximately half the number of Gln residues as Httex1(Q46), the apparent binding affinity was 150-fold lower than obtained for Httex1(Q46) (Table 2). Despite offering the highly similar polyQ epitopes to MW1, the difference in apparent binding affinities further highlights that entropic and avidity effects play a major role in MW1 binding and polyQ-length differentiation. This circumstance is also visible in the initial decrease of enthalpy changes for Httex1(Q46) up to Httex1:MW = 1:1 (Fig. 6A). Such a behavior violates a truly sigmoidal binding curve (Fig. 6A) and shows that the underlying assumption of identical and independent binding sites is not quite accurate. Thus, ITC further shows the Httex1–MW1 interaction to be stoichiometry dependent.

Figure 6.

Isothermal titration calorimetry of Trx-Httex1 with MW1.A and B, Trx-Httex1(Q46) (A) and Trx-Httex1(Q25) (B) at concentrations of 0.4 μM were titrated with MW1 at 25 °C in 20 mM NaH₂PO₄/Na₂HPO₄ pH 7.4, 150 mM NaCl solution. Httex1, Huntingtin exon 1.

Table 2.

Thermodynamic parameter of the Trx-Httex1-MW1 interaction^a

Httex1 variant	Ka (M−1)	ΔH° (kcal/mol)	TΔS° (kcal/mol)	ΔG° (kcal/mol)	Stoichiometry
Trx-Httex1(Q46)	(260 ± 68)∗105	−73.5 ± 2.9	−63.4 ± 2.9	−10.1 ± 0.2	2.17
Trx-Httex1(Q25)	(1.69 ± 0.37)∗105	−404.6 ± 71.4	−397.5 ± 71.4	−7.1 ± 0.1	1b

^aAssuming independent and identical MW1-binding sites for Trx-Httex1.

^bThe binding stoichiometry was fixed at 1:1.

Discussion

The preferential binding of MW1 to huntingtin proteins with expanded, pathological Q-lengths makes a deeper understanding of the polyQ–MW1 interaction of significant interest to polyQ research. Using ITC, we quantified the increase in MW1 affinity for monomeric Httex1(Q46) over monomeric Httex1(Q25) to a factor of approximately 150. To better understand the underlying basis of this effect, we investigated the structural features of MW1-bound Httex1 as well as complex stoichiometry. EPR analysis of distances between spin labels in the polyQ region of Httex1Q46 indicated that MW1 binding brings about a conformational change from a coiled structure to a nearly fully extended backbone conformation. Further, EPR data revealed that the MW1 binding sites in Httex1(Q46) are in the central portion of the polyQ region, approximately containing residues 25 to 60. The DEER data showed an extended structure across this entire binding region. The ∼35 residues long (25–60) extended structure seen here is significantly longer than the extended, ten residue-long structure of a polyQ peptide bound to a single MW1 Fab fragment (32). Thus, the concerted binding of multiple MW1 molecules results in a nearly linear alignment of individual binding sites in which the polyQ is extended (Fig. 7, C and D). The transition from bound to unbound regions in the polyQ is gradual, which was particularly apparent in the different reductions of backbone and side-chain Gln NMR signals upon MW1 binding. At the side-chain level, 27 residues were nominally bound. When comparing this number to the ∼35-residue binding region detected by EPR, the difference suggests that not all residues in the binding region interact with MW1 at the same time. Given the repetitive polyQ sequence, it is quite likely that the 27 out of 35 MW1-bound residues are not always the same residues, rather we expect that there will be a degree of “sliding” of MW1 along the polyQ backbone. Interestingly, the N-terminal region of polyQ is not strongly, if at all, bound by MW1 in Httex1(Q46) or Httex1(Q25). In monomeric Httex1, this polyQ stretch and the preceding N17 region become more helical with increasing polyQ lengths (8). However, our EPR distance measurements and prior crystallographic studies of MW1 in complex with a polyQ peptide (32) show extended conformations when bound to MW1, which is incompatible with α-helical structure. This difference in secondary structure preference likely contributes to the strongly reduced binding in the N-terminal polyQ region.

Figure 7.

Schematic Model illustrating different modes of binding and binding stoichiometries for Httex1 of different Q-lengths in the presence of MW1.A, in case of short Q-length proteins, only one MW1 protein (red) can bind to the polyQ region (green). Due to the bidentate nature of the MW1 antibody, one antibody can bind up to two polyQ regions. B–D, longer Q-lengths can accommodate additional MW1 antibodies, causing the formation of larger complexes that are favored entropically. Httex1, Huntingtin exon 1; polyQ, polyglutamine.

If MW1 does not recognize a polyQ conformation specific for Httex1 with longer Q-lengths and potentially even slides along polyQ, then what is the origin of its higher affinity for long Q-lengths? Prior work using Fab fragments of MW1 showed that the large number of Gln residues in Httex1 with expanded polyQ lengths provide more binding sites, thereby leading to higher binding affinities by virtue of mass action (32). The present study, which uses the full-length MW1, finds that additional factors are at play. Gel filtration detects pronounced hetero-oligomerization into larger complexes of varying sizes for Httex1 with longer Q-lengths. This was not observed in analogous experiments with MW1 Fab fragments (29) and must arise from the ability of the bidendate MW1 to effectively cross-link multiple antibodies and Httex1 molecules (Fig. 7). Crosslinking of Fab regions giving rise to adjacent binding sites has been shown to enhance binding, albeit to a smaller degree to that observed here (32). MW1 binding to exon 1 is likely different since the bidentate binding sites in IgG2b isotypes are about 130 Å apart (36). Even in the extreme case of a perfectly extended polyQ linker, it would require a minimum of 37 amino acids to bridge this distance (130 Å/3.5 (Å/amino acid) = ∼37 amino acids), which is more than the entire 35 amino acids in Httex1(Q46) found to bind MW1. Further, 20 of the 35 residues would be needed for binding to two Fabs, leaving only 15 amino acids for the linker. The shortness of the linker likely prevents binding of one MW1 to a single Httex1 molecule using two Fab sites and explains why larger multimers are favored over the 1:1 complex of MW1 and Httex1(Q46) (Fig. 5). In case of a short Q-length, e.g., Httex1(Q16), gel filtration indicates that the antibody complexes contain only one MW1 molecule. Thus, after engaging the first Httex1 molecule, the bidentate MW1 can only complex another Httex1 molecule in a separate binding event that is not cooperative (Fig. 7A). This is different for longer Q-lengths, where additional binding sites are available. In that case, subsequently bound MW1 molecules can simultaneously interact with two Httex1 molecules (Fig. 7C). Such binding interactions depend on only one and not two separate molecular encounters, making them entropically favored compared to the binding of the first MW1 molecule (Fig. 7B). The number of MW1 molecules that can bind to Httex1 depends on polyQ length, making it likely that several entropically favorable interactions can occur upon binding to two Httex1 molecules with long Q-length. In addition, larger complexes with more than two MW1 molecules can be envisioned (Fig. 7D). Based on the missing background in the DEER data, we expect that individual Httex1(Q46) molecules are held apart from each other and that all contacts occur via the bridging MW1 molecules. While future work will be required to understand the exact nature of the complexes, we propose that the hetero-oligomerization of MW1 and Httex1 with larger Q lengths likely makes a substantial contribution to the ∼150-fold higher binding affinity of MW1 for Httex1(Q46) compared to Httex1(Q25).

The present data also have implications for the interpretation of TR-FRET experiments that use 2B7 and MW1 to study the structure of monomeric Htt. While these experiments are highly useful in detecting Httex1 and other N-terminal Htt fragments at low concentration with different signals for long and short Q-lengths, they do not provide direct information on the underlying native solution structure or dynamics of Httex1 or other Htt fragments. First, MW1 binding strongly alters the native structure, and it is not reporting on local structure, rather it induces a new one. Second, the degree to which MW1 binding results in the formation of large oligomeric networks is strongly Q-length dependent. The different sizes and stabilities of the different complexes would be expected to result in very different temperature stabilities. Thus, the pronounced temperature-dependent TR-FRET changes for proteins of different Q-lengths could reflect the stability of the complexes, rather than intrinsic flexibility of the Httex1 or other N-terminal Htt fragments.

Collectively, the present data suggest that MW1 does not detect a specific toxic conformation favored by expanded Q-lengths and that the apparent Q-lengths rigidity differences seen by TR-FRET are unrelated to the structural properties of native Httex1 proteins. Rather, MW1 senses longer Q-lengths by virtue of mass action as well as multimerization. The lack of a defined toxic structure in Httex1(Q46) does not necessarily mean that only oligomeric or smaller fibrillar species contribute to toxicity. The markedly stronger interaction of Httex1(Q46) with MW1 compared to Httex1(Q25) represents a powerful illustration of how a repetitive sequence, such as the polyQ region, can lead to binding interactions that are orders of magnitude higher than those of the wildtype sequence. Future work might show whether monomers with long Q-lengths can undergo potentially toxic, aberrant binding interactions in the cell, especially if multivalent binding partners are involved. At the same time, a design of a multivalent polyQ binder may also prove useful for allele-specific targeting of huntingtin with longer Q-lengths in vivo.

Experimental procedures

Protein expression, labeling, and purification

The Httex1 proteins with different Q-lengths (Q16, Q25, Q46) were expressed and purified as thioredoxin fusion proteins using pET32a or pET28a vectors as described previously (8). As in our prior studies, the cysteines in thioredoxin were substituted with serines to facilitate the EPR experiments (7, 8).

Dot blot assay

A film of purified and disaggregated Httex1 was formed at the bottom of the borosilicate tube by removing 0.5% TFA/Methanol under a gentle stream of nitrogen gas. The protein film was dissolved in 20 mM Tris, 150 mM NaCl, pH 7.4 for dot blot assay. Protofibrils (P) were formed by incubation at 4 °C for 14 h. The nonbundled, toxic (T) fibrils and bundled fibrils (N-fibrils) were formed by the method as described previously (6). All fibril preparations were then centrifuged (150,000g, 60 min) to remove monomers. The concentration was estimated using a standard curve calibrated from minima at 213 nm obtained through circular dichroism measurements for disaggregated fibrils. The samples (1 μl) were blotted onto a nitrocellulose membrane (VWR) and incubated with MW1 at 1:5000 dilution. Fluorescently labeled secondary antibody (IRDye-800 anti-Mouse IgG, 1:10,000 dilution; LI-COR) was used to visualize binding via a LI-COR Odyssey Infrared Imaging system.

Immunohistochemistry using MW1

HEK293T cells were grown on poly D-lysine coated glass coverslips in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum (Gibco) and 100 U/ml penicillin-streptomycin (ThermoFisher) until 80% confluency. Media were replaced with serum-free Dulbecco's modified Eagle's medium, and cells were transfected with Httex1 Q72-RFP plasmid using Lipofectamine (ThermoFisher) for 4 h at 37 °C. Cells were returned to complete culture media after transfection and further incubated for 24 h. Cells were then rinsed in PBS, followed by PBS with 0.1% Triton X-100, fixed in 3.7% paraformaldehyde for 10 min, washed in PBS, and incubated with MW1 antibody (0.1 μg/ml) in blocking buffer (PBS with 1% bovine serum albumin) overnight. The next day, cells were rinsed in PBS, incubated with a secondary antibody (donkey anti-mouse IgG alexa fluor 405, ThermoFisher) for 2 h, rinsed in PBS, and the coverslips were mounted on glass slides using mounting media with DAPI (Vector Laboratories). Images were obtained using a LSM 780 confocal microscope.

Preparation of spin-labeled samples

Purified Httex1 thioredoxin fusion proteins were spin-labeled by incubation with a 5- to 10-fold excess of MTSL spin label (Toronto Research Chemicals, Inc). Labeled proteins were loaded onto a HiTrap Q XL column (GE Healthcare) using an AKTA FPLC system (Amersham Pharmacia Biotech) in 10 mM Tris-HCl, pH 7.4. The protein was eluted using 20 mM Tris-HCl, pH 7.4 buffer and a salt gradient from 20 mM to 1 M NaCl. The eluted protein was buffer exchanged into 20 mM sodium phosphate, pH 7.4, 150 mM NaCl using a PD10 column.

Electron paramagnetic resonance

For continuous wave EPR, spin-labeled proteins were loaded into glass capillaries (0.6-mm inner diameter × 0.84-mm outer diameter, VitroCom). EPR spectra were obtained on a X-band Bruker EMX spectrometer (Bruker Biospin Corporation). All spectra were scanned at 100 G width. The spectra were recorded at room temperature using 12.6 mW power in a HS cavity. Modulation amplitude was 1.5 G. Spectra obtained for a given derivative in the presence or absence of antibody are shown normalized for the same number of spins to better illustrate the effect of MW1 on the spectral amplitude. The concentration of spin-labeled Httex1 was 10 μM, and the MW1 concentration was varied as indicated in the figure legends.

Four-pulse DEER experiments were done to obtain intramolecular distances between spin label pairs of 25R1/35R1, 35R1/45R1, and 40R1/50R1. The double spin-labeled proteins were used at a concentration of 20 μM with or without MW1. All samples included 10% glycerol as a cryoprotectant that was added prior to flash freezing the samples in liquid nitrogen. According to light microscopy, Httex1 does not undergo liquid–liquid phase separation in buffer containing 10% glycerol. All measurements were performed on Bruker ELEXSYS E580 X-band pulse EPR spectrometer equipped with a 3-mm split ring (MS-3) resonator, a continuous-flow cryostat (CF935; Oxford Instruments), and a temperature controller (ITC503S; Oxford Instruments) at a temperature of 78 K. Data were fitted using Gaussian regularization as implemented in DEER Analysis 2019 to obtain distance distribution for the spin-labeled pairs. The background contribution due to nonspecific interactions were subtracted for the samples without MW1. No background correction was required for spin-labeled pairs in the presence of MW1. The distances shown in the results section is the maxima in the Gaussian distance distribution.

Solution NMR spectroscopy

¹⁵N-¹H HSQC spectra of ¹⁵N-labeled Trx-Httex1(Q46) at a concentration of 14 μM were recorded on a Bruker Avance 700 spectrometer at 10 °C in the absence and presence of 28 μM MW1 in 20 mM NaH₂PO₄/Na₂HPO₄ pH 7.4, 150 mM NaCl solution. Using the previously carried out chemical shifts assignments (8) data were processed and analyzed with the nmrPipe package.

Gel filtration

Nonequilibrium protein interaction experiments were carried out on a Superdex 200 Increase 10/300 GL gel-filtration column (GE Healthcare) equilibrated in a buffer containing 20 mM phosphate (pH 7.4) and 150 mM NaCl. A final concentration of 10 μM Trx-Httex1 was used for all experiments, with MW1 concentrations varied as indicated in the text and figures. Five hundred microliter was injected and flowed through the column at 0.5 ml/min at room temperature. The absorbance of the eluent was monitored at 280 nm. Globular protein standards of known molecular weight from Bio-Rad were used to calibrate the Superdex-200 gel-filtration column, ranging from ∼50 to 1300 kD, under nondenaturing conditions, i.e., thyroglobulin, γ-globulin, ovalbumin, myoglobin, and vitamin B12.

Isothermal titration calorimetry

Utilizing a Microcal VP-ITC calorimeter, 0.4 μM Trx-Httex1(Q46) or Trx-Httex1(Q25) was titrated with MW1 antibody at 25 °C in 20 mM NaH₂PO₄/Na₂HPO₄ pH 7.4, 150 mM NaCl solution. MW1, at a concentration of 8.8 μM, was injected in 11 μl aliquots over a period of 12 s (Fig. 6). Prior to data analysis, the measurements were corrected for the heat of dilutions of Trx-Httex1 and MW1. The reaction enthalpy (ΔH°), binding stoichiometry (n), and affinity constant, K_A, were extracted from the measured heat changes, δH_i, for each injection, i, by nonlinear curve fitting, assuming independent and identical MW1-binding sites for Trx-Httex1. For Trx-Httex1(Q25), n was fixed at 1. The free energy change (ΔG°) was subsequently obtained as −RT ln K_A, where R denotes the gas constant and T the absolute temperature. The entropy change (ΔS°) equals (ΔH° − ΔG°)/T.

Data availability

Data available within the article or its Supporting information.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare no conflict of interest with the contents of this article.

Biography

José M. Bravo-Arredondo received his Ph.D. in Materials Science from Benemérita Universidad Autónoma de Puebla in México. He then joined Dr. Langen’s group at USC on a USC-CONACyT fellowship to study the structure, aggregation, and interactions of huntingtin, a protein that plays a central role in Huntington’s disease. He is currently at the Universidad Autónoma de Tlaxcala, working on synthesizing different metal-organic frameworks to detect biomolecules including huntingtin.

Reviewed by members of the JBC Editorial Board. Edited by Ursula Jakob