|Home | About | Journals | Submit | Contact Us | Français|
Abnormal *polyglutamine (polyQ) tracts are the only common feature in nine proteins that each cause a dominant neurodegenerative disorder. In Huntington’s disease (HD), tracts longer than 36 glutamines in the protein huntingtin (htt) cause degeneration. In situ, monoclonal antibody 3B5H10 binds to different htt fragments in neurons in proportion to their toxicity. Here, we determined the structure of the 3B5H10 Fab to 1.9Å by x-ray crystallography. Modeling demonstrates that the paratope forms a groove suitable for binding two β-rich polyQ strands. Using small-angle x-ray scattering (SAXS), we confirmed that the polyQ epitope recognized by 3B5H10 is a compact, two-stranded hairpin within monomeric htt and is abundant in htt fragments unbound to antibody. Thus, disease-associated polyQ stretches preferentially adopt compact conformations. Since 3B5H10 binding predicts degeneration, this compact polyQ structure may be neurotoxic.
An abnormal expansion of the polyglutamine (polyQ) tract in huntingtin (htt) results in a self-aggregating protein and neurodegeneration. Understanding which structures of polyQ in mutant htt and other polyQ-expanded proteins are most closely linked to pathogenesis has important implications for mechanisms of neurotoxicity. For example, unaggregated expanded polyQ has been suggested to mediate toxicity through aberrant recruitment of cellular proteins1. If such “recruitment-competent” polyQ is structured, then mutant htt may act as a structural mimic/competitor for the recruited protein’s normal binding partner(s). Alternatively, if “recruitment-competent” polyQ is unstructured, then expanded polyQ may simply provide a longer, more accessible recruitment site than wild-type (wt) stretches of polyQ.
These alternate pathogenic scenarios lead to potentially divergent therapeutic strategies. In the case of a structured polyQ epitope, screens for therapeutics that disrupt toxic structure formation may be warranted. In the case of an unstructured polyQ epitope, therapeutic strategies may instead focus on covalently linking together multiple copies of a molecule that recognizes a short, linear array of polyQ2,3,4,5.
While the structure of aggregated polyQ peptides takes on the classical cross-β-strand structure of amyloid fibrils6, the exact structures and pathogenic significance of the range of putative monomeric and very small oligomeric species of mutant htt are unknown. The first exon of wt htt was recently crystallized revealing multiple conformations of the polyQ stretch7. However, studies on expanded (mutant) polyQ suggest that the unaggregated forms are largely unordered, adopting secondary structure only upon aggregation5,8,9. These data led to the concept of a linear sequence of unaggregated glutamine without secondary structure as “beads on a string.” In this “linear lattice” model, toxicity of unaggregated, expanded polyQ is caused by increased accessibility of the polyQ region to cellular ligands2,5. Further, if a particular cellular ligand has two polyQ binding sites, then the ligand will exhibit a strong preference for expanded over wt stretches of polyQ driven by increased avidity. Indeed, at least two IgG antibodies (MW1 and 1C2), which inherently have two identical epitope binding sites, have been thought to preferentially bind expanded polyQ, at least in part, by the avidity mechanism implied in the “linear lattice” model2,10.
In contrast to the “linear lattice” model, some in vitro experimental evidence indicates that expanded polyQ induces a global change in conformation in unaggregated htt1,11,12. However, direct experimental evidence that disease-associated polyQ stretches adopt an emergent conformation in situ is lacking. Further, whether any particular emergent conformation formed in situ is toxic is also unknown.
We identified a monoclonal antibody, 3B5H10, that recognizes a species of htt in neurons that predicts neurodegeneration better than all other α-htt antibodies tested, including the “linear lattice”-recognizing antibody MW113. Here, we show that 3B5H10 recognizes a compact, two-stranded conformation of polyQ in monomeric htt that emerges when the polyQ stretch expands. Our data show that a specific, compact conformation of expanded polyQ forms in unaggregated htt in situ and that this compact conformation has particular pathological significance.
Since the epitope recognized by mAb 3B5H10 predicted neurodegeneration better than the epitope recognized by MW113, we considered the possibility that 3B5H10 recognizes an epitope formed preferentially by mutant htt rather than a repeated epitope envisioned by the linear lattice model. We reasoned that a conformation that preferentially forms in mutant htt should be stable at disease-associated polyQ lengths, unstable at near-threshold lengths, and relatively unformed at short lengths.
To test this putative difference in stability, we probed the effects of the denaturant SDS on 3B5H10 binding to mutant, threshold, and wt versions of htt. Specifically, we chose to test three different polyQ stretches (Q17, Q25, Q40) based on the frequency with which the corresponding CAG codon stretches are found in the htt gene within humans. A stretch of Q17 is among the most common alleles found in the normal population14,15,16,17 whereas a stretch of Q40 is relatively common among HD patients and is fully penetrant15,17. Htt alleles with Q23–34 are relatively rare but correspond to a transition zone between the most common normal and disease-associated alleles and, therefore, may have particularly interesting biochemical properties15,17.
Cells were transfected with N-terminal 171–amino acid fragments of htt containing polyQ stretches of 17, 25, or 40 and hemagglutin (HA) and FLAG epitope tags fused, respectively, to the N- and C-termini of htt. Cells were lysed under native conditions 48 h after transfection, and the lysates were immunoprecipitated with α-HA epitope or 3B5H10 antibodies, subjected to SDS-PAGE, and blotted with α-HA epitope or 3B5H10 antibodies. When lysates were subjected to immunoprecipitation (IP) and blotting with an α-HA antibody, the three versions of htt showed equal immunoblotting intensities, signifying that all were equally available for IP (Fig. 1a). In contrast, IP with 3B5H10 and blotting with α-HA antibody revealed a band for the Q40 and Q25 versions of HA-171-Htt-FLAG but not for the Q17 version. This finding confirms that 3B5H10 preferentially binds to versions of htt near-threshold or higher. Interestingly, when the three versions of htt were immunoprecipitated with 3B5H10 and then blotted with 3B5H10, only HA-171-Q40-FLAG was detected. This suggests that the epitope on HA-171-Q25-FLAG recognized by 3B5H10 was unstable and disappeared upon SDS exposure, while the HA-171-Q40-FLAG epitope recognized by 3B5H10 remained. Since some proteins demonstrably retain or regain substantial secondary structure on a nitrocellulose membrane after SDS-PAGE18,19,20,21,22,23,24,25,26, immunoreactivity to HA-171-Q40-FLAG after SDS exposure does not exclude the possibility that 3B5H10 recognizes a protein fold. These results deviate from simple predictions of the “linear lattice” model and suggest 3B5H10 recognizes an epitope that is sensitive to SDS denaturation and whose sensitivity varies in a polyQ-length-dependent manner.
Since our data suggested that 3B5H10 recognizes a conformation of polyQ that emerges with longer polyQ stretches, we sought to characterize the structure of this conformation. First, we visualized 3B5H10’s epitope-binding groove by purifying26 and crystallizing27 its Fab fragment. We determined its x-ray crystal structure at 1.9Å by single isomorphous replacement with anomalous scattering and molecular replacement (PDB: 3S96) (Fig. S1, Tables S1–3). During the process of solving the structure, we noticed several similarities between 3B5H10 and MW1. For example, both antibodies have a lambda light chain, which is found in only 5% of the mouse antibody repertoire28. Further, a sequence comparison of 3B5H10 and MW1 reveal nearly identical light chain complementarity-determining regions (CDR) and highly similar heavy chain CDRs that are particularly enriched in aromatic residues (Fig. 1b), many of which are solvent accessible (yellow coloring, Fig. 1c and d). Consistent with aromatic moieties facilitating glutamine binding3,29,30, 33% and 30% of the residues in the heavy chain CDRs of MW1 and 3B5H10, respectively, are aromatic, in contrast to the 9% of residues for the heavy chain CDRs of the closest 100 homologues.
A surface representation of the 3B5H10 Fv reveals a potential polyQ epitope binding groove studded with regularly spaced aromatic residues that stretches diagonally across the antibody’s CDRs (Fig. 1c). The diagonal orientation of the potential epitope-binding groove, which is also seen in the crystal structure of MW1 Fv (Fig. 1d), is unique among 48 other antibody-peptide structures surveyed from the Protein Data Bank2. Previous studies of a GQ10G peptide complexed to MW1 Fv (PDB: 2OTU)2 revealed that the polyQ peptide follows the path of solvent-accessible aromatic residues in the diagonal epitope-binding groove (Fig. 1e). Upon peptide binding, the groove straightens through a 3° rotation between the antibody’s heavy and light variable chains (domain-rotation) to accommodate the uncurved epitope.
The strong sequence and structural similarities between MW1 and 3B5H10 suggested that polyQ binds to 3B5H10 along the analogous diagonal groove, taking advantage of regularly spaced aromatic residues lining the groove. Further, the strong similarities suggest that the 3° domain rotation seen in MW1 upon epitope binding also occurs in 3B5H10. Such domain rotations are observed for other antibody/epitope interactions and are well documented in the literature31. Indeed, alternative crystal forms of the 3B5H10 Fab (PDB: 4DCQ) were consistent with this domain rotation (Fig. S2e and Supplementary Discussion). Analogous to MW1, this domain rotation straightens the epitope-binding groove of the Fv of 3B5H10 while producing minimal steric clash. Of all residues at the heavy chain/light chain interface after domain rotation, only seven needed to adopt an alternate side-chain rotamer involving at least a 1Å shift for at least one atom in the residue. No atom shifted by greater than 1.6Å (Fig. S2 and Supplementary Discussion and Methods).
While the MW1 Fv and 3B5H10 Fv appear sequentially and structurally similar (Fig. 1), several key differences are observed when the crystal forms are overlaid on each other (Fig. 2a). The 3B5H10 Fab has unusually long β-strands in CDR3 of the light chain (L-CDR3) and CDR2 of the heavy chain (H-CDR2), unlike the corresponding CDRs of the Fv fragment of MW1 (Fig. 2a). Since MW1 and 3B5H10 share nearly identical light chains (Fig. 1b), differences in paratope conformation are determined by the heavy chain, particularly the packing interactions between H-CDR2 and L-CDR3 and between H-CDR2 and H-CDR1 (Fig. 2b–d). While the structure of L-CDR3 is loop-like in MW1, it forms an extended β-hairpin that is observed in all crystal forms of 3B5H10 Fab, regardless of the location of the crystal cell contacts. Although stable rigid secondary structures in CDRs are unusual, similar β-strands have been found in the CDRs of other Fabs32.
In MW1, hydrogen bonds between L-CDR3 and H-CDR2 appear to stabilize the extended loop-like structure of L-CDR3 and bring L-CDR3 and H-CDR2 closer together to form part of the wall of the epitope-binding groove (Fig. 2b and c). Because of sequence differences in H-CDR2 between MW1 and 3B5H10, these hydrogen bonds do not exist in 3B5H10. The overall effect of these differences is to convert a steep wall midway through the epitope-binding groove in MW1 into a shallow, less rigid wall for 3B5H10 (Fig. 2b and c). This serves to make the diameter wider midway through 3B5H10’s epitope-binding groove, as compared to MW1.
The extended β-hairpin in H-CDR2 of 3B5H10 also prevents significant packing between H-CDR2 and H-CDR1, thereby contributing to the widening of the epitope-binding groove at one end of the 3B5H10 paratope (Fig. 2b and d). In contrast, the more loop-like structure of MW1 H-CDR2 packs tightly against H-CDR1, contributing to a wall that significantly narrows the corresponding portion of the paratope valley in MW1 (Fig. 2b and d) and allows for only a single polyQ strand to bind. Indeed, the narrowest diameter of the peptide-binding groove in MW1 spans from G101 in H-CDR3 to Y33 in H-CDR1 and measures 5.5Å. In contrast, the narrowest diameter of the peptide-binding groove in 3B5H10 spans from Y103 in H-CDR3 to N52 in H-CDR2 and measures 12.15Å. Thus, the extended β-hairpin structure of H-CDR2 and the interaction it forms with H-CDR1 and LCDR-3 in 3B5H10 results in a significantly wider groove across the length of 3B5H10’s paratope, as compared to MW1.
The width of 3B5H10’s epitope-binding groove, as compared to MW1, suggests that it binds to a larger/wider epitope than the single polyQ strand that binds MW1’s paratope (see Fig. S3). Since β-strands are believed to be the fundamental secondary structural unit of compact polyQ structures33, we used flexible docking in CLUSPRO34 and Z-DOCK35 to explore the compatibility of single- and double-stranded β-rich polyQ conformations with 3B5H10 and MW1 (Supplementary Methods). This procedure replicated the crystal structure of a GQ10G peptide bound to MW12 and demonstrated that MW1 recognizes various polyQ models containing β-strands (Fig. S3 and S4). For all multistranded polyQ models complementary to MW1, only a single strand was inserted into the antibody’s peptide binding groove, following the path of the bound GQ10G peptide (for example, see Fig. S5). Thus, MW1 may bind additional conformations of polyQ besides the random coil conformation described2,5 (Supplementary Discussion).
In contrast, only two-stranded, β-hairpin models customized to fit 3B5H10’s paratope docked with 3B5H10 (Supplementary Discussion and Methods). Attempts to dock single- and double-stranded models position a glutamine directly over two solvent-accessible tryptophans in the paratope groove (Fig. 3a), consistent with a 5-nm red shift in tryptophan fluorescence when 3B5H10 binds a thioredoxin hexahistidine fusion protein of exon 1 of htt with 39 glutamines (Thio-Httex1-Q39-His6) (Fig. 3b). However, single-stranded models fail to completely fill the binding surface, resulting in significantly fewer hydrogen bonds per square angstrom than in the MW1:GQ10G crystal structure (Fig. S3), double-stranded models (Fig. S4), or complexes of homologous Fabs and their antigens (Fig. S3). For double-stranded docking, we utilized the paratope from the 3B5H10 crystal structure in Fig. 1c and the paratope from the domain-rotated version of 3B5H10 discussed above. Docking results revealed that a β-rich hairpin model is complementary to the paratope of 3B5H10 observed in the crystal structure (Fig. 3c), whereas the domain-rotated paratope accommodates a flatter β-rich hairpin model (Fig. 3d). Each model aligned both β-rich strands roughly parallel to the binding surface, burying 950–1100Å2, and forming 12–16 hydrogen bonds, well within the range for complexes of homologous Fabs and their protein antigens (Fig. S4). Thus, the epitope of 3B5H10 is a conformation of polyQ similar to a two-stranded, β-rich hairpin.
Our docking studies also suggested that 3B5H10 recognizes fewer polyQ conformations than MW1. 3B5H10 appears to only bind compact conformations with two β-rich strands, but MW1 recognizes a wide range of exposed polyQ in soluble httex1. Since 3B5H10 binding to httex1 predicts neurodegeneration significantly better than MW1 binding13, its prognostic value may reflect its higher specificity for a toxic conformer or closely related species.
To experimentally test our docking predictions that 3B5H10 binds to a compact, double-stranded conformation of polyQ, we used SAXS to determine the shape of htt fragments in solution and in complex with 3B5H10 Fab. Htt fragments were expressed within six fusion proteins. The protein tags within the fusion proteins served two critical functions. First, as has been established, these tags help to prevent htt from aggregating too quickly to analyze substantially non-aggregated material5,36. Second, a series of protein tags, each of which has an established crystal structure, allows us to accurately deduce the volume occupied by the polyQ stretch within the SAXS molecular envelope of htt or Fab:htt complexes. We reasoned that if six different htt fusion constructs all demonstrated a similar polyQ conformation, the chances that the protein tags were artificially altering the htt polyQ conformation was low.
The six fusion proteins used for our SAXS analysis included Thio-Httex1-Q39-His6, described earlier. MBP-Httex1-Q46-Cer is a fusion protein of httex1 with a stretch of 46 glutamines, Cerulean fluorescent protein (Cer) on its C-terminus, and maltose binding protein (MBP) on its N-terminus. MBP-Htt1–17-Q46-Cer is similar but lacks the portion of httex1 after the polyQ stretch, which includes the proline-rich region. N-terminal tags from MBP-Httex1-Q46-Cer, MBP-Htt1–17-Q46-Cer, and Thio-Httex1-Q39-His6 were removed proteolytically to generate the three remaining fusion constructs: Httex1-Q46-Cer, Htt1–17-Q46-Cer, and Httex1-Q39-His6, respectively. By comparing SAXS data from fusion proteins with the same htt fragment but different protein tags in the presence or absence of 3B5H10 Fab, we could assign regions of the molecular envelope to the tags, 3B5H10 Fab, and htt (Supplementary Methods) (Table S4).
Unlike 3B5H10 Fab:htt complexes containing MBP-Htt1–17-Q46-Cer or MBP-Httex1-Q46-Cer, the Fab:htt complex containing Thio-Httex1-Q39-His6 is dimeric at the high concentrations used in SAXS13. The polyQ in the Thio-Httex1-Q39-His6 complex SAXS envelope can be unambiguously assigned, but there are two possibilities for the path of the proline-rich region (Fig. S6). We favor one model (Fig. 4a) because it is consistent with the location of the proline-rich and Htt1–17 regions in monomeric httex1-Q46-Cer complexes and the SAXS-derived structure of Thio-Httex1-Q39-His6 in solution.
For all Fab:htt complexes tested, the htt fragment bends in the polyQ region, and the polyQ epitope in the complex is cylindrical, 24Å in diameter and 40–55Å long. This compact volume is only consistent with a two-stranded, hairpin structure, and most consistent specifically with 16–24 glutamines in direct contact with the paratope of the Fab and a turn of 4–8 glutamine residues (Fig. 5a–c). The varying number of glutamines associated with the variable domain of the Fab–24 for MBP-Htt1–17-Q46-Cer and 29 for MBP-Httex1-Q39-His6 and MBP-Httex1-Q46-Cer—may reflect an influence of the proline-rich domain on the polyQ region37. However, SAXS analysis of Fab:htt complexes in which the proline-rich region or the N-terminal or C-terminal tags were removed also revealed a compact, two-stranded, hairpin-like structure for the polyQ region of htt (Table S4). Thus, compact hairpins appear to be an intrinsic property of the mutant polyQ stretch.
Having identified the structure of a polyQ conformation in htt that strongly predicts neurotoxicity, we sought to understand how abundant this conformation was in the absence of 3B5H10 with SAXS, circular dichroism (CD), and chemical cross-linking. SAXS and CD provide the population average of conformations sampled by a protein in solution, while cross-linking “freezes” the set of conformations a protein is sampling at a given time. Thus, if the conformation 3B5H10 recognizes is abundant in solution, SAXS and CD analysis of mutant htt unbound to antibody should reveal a structure similar to the structure of mutant htt bound to antibody. Additionally, cross-linking of mutant htt should preserve strong subsequent binding by 3B5H10.
SAXS analysis of the htt fusion proteins unbound to the 3B5H10 Fab revealed that the population average of polyQ conformers is best represented by a compact, two-stranded, hairpin structure highly similar to the conformation of polyQ bound to 3B5H10 (compare Fig. 5a–c and 5d–f). These findings are consistent with structural studies of smaller polyQ-containing peptides38. Further, CD spectra of htt fusion proteins complexed to the 3B5H10 Fab represent the simple sum of the CD spectra of the htt fusion proteins alone and the CD spectra of the Fab alone (Fig. 5g; secondary structure analysis is shown in Fig. S7; Supplementary Methods). Thus, CD results demonstrate the lack of significant secondary structure change in htt upon 3B5H10 binding and confirm that 3B5H10 recognizes a secondary structure of mutant htt that represents the population average of polyQ conformers in unbound htt. In agreement with the SAXS and CD data, strongly cross-linking mutant htt expressed in primary neurons with 4% paraformadehyde and 0.2% glutaraldehyde, followed by 2% glutaraldehyde, results in robust subsequent antibody binding, as assessed by electron microscopy (Fig. S8). Collectively, these results are consistent with data demonstrating the ability of some proteins to sample several native conformations39 and with 3B5H10 recognizing a conformation among those sampled by unaggregated htt40. Thus, we conclude that 3B5H10 recognizes a compact, two-stranded, hairpin conformation of polyQ that is abundant in solution and that is strongly predictive of neurodegeneration.
In this study, we showed that polyQ stretches within monomeric mutant htt fragments form compact, two-stranded, hairpin structures in solution. This conformation, which represents the population average of conformations that mutant htt polyQ samples in solution, is recognized by a newly developed monoclonal antibody 3B5H10. Since the epitope recognized by 3B5H10 predicts neurodegeneration better than all other epitopes tested13, we propose that this polyQ structure has particular pathologic significance.
That 3B5H10 binds a compact, two-stranded, hairpin structure of polyQ is supported by multiple lines of evidence. First, the epitope-binding groove of 3B5H10 is at least twice as wide as the otherwise highly similar paratope on MW1, an antibody which accommodates only a single-strand of polyQ. Molecular docking studies of polyQ structures into the paratope of MW1 versus 3B5H10 revealed that only compact, two-stranded, β-hairpin-rich structures of polyQ fit into 3B5H10. In contrast, MW1 accommodated a wide range of single-stranded polyQ structures, including the random coil structure of the polyQ peptide that was co-crystallized with the antibody2. However, the MW1 paratope was unable to accommodate double-stranded polyQ structures, except when one strand was placed above and outside the paratope.
In further support that 3B5H10 binds polyQ in a beta-strand conformation, others have introduced amino-acid substitutions in the polyQ tract of htt whose nature and location were designed to disrupt beta structure. There was a correlation between predicted disruption of beta structure and reduced 3B5H10 binding41. Our SAXS data on a range of Fab:htt complexes confirmed predictions that 3B5H10 binds a compact, two-stranded hairpin structure of polyQ. Further, SAXS analysis, CD analysis, and chemical cross-linking experiments showed that the conformation recognized by 3B5H10 represents the population average of unbound htt in solution40. While our modeling suggests a predominately beta secondary structure to this compact conformation, molecular dynamics simulations and single molecule force-clamp experiments indicate that more disordered hairpin secondary structure in polyQ can also result in a compact ensemble of structures42,43.
Interestingly, MW1 and 3B5H10 both bind expanded polyQ stretches better than wt stretches, but their respective epitopes carry different prognostic values. Our docking studies of polyQ peptides into the MW1 paratope suggest that the paratope accommodates a wide-range of single-stranded polyQ structures. Thus, in agreement with previous studies, we conclude that MW1 is a polyQ-binding protein rather than a conformation-specific protein. The antibody prefers longer polyQ stretches because they enable more avid binding: the bivalent antibody dramatically increases binding when the short, unstructured polyQ epitope it recognizes appears in tandem2,5.
In contrast, 3B5H10 is a conformation-specific antibody and therefore recognizes a much smaller range of htt species. Its preference for longer polyQ stretches is based on recognition of a specific compact, two-stranded, hairpin conformation that emerges as the polyQ stretch expands. This specific conformation carries greater prognostic significance than the broader range of polyQ epitopes recognized by MW1. Interestingly, MW1 and 1C2, another α-polyQ antibody that may bind polyQ according to the “linear lattice” model, were both raised against wt stretches of polyQ in different protein contexts. MW1 was raised against a 53–amino acid fragment of wt atrophin-1 containing Q1944 [wt atrophin-1 contains <Q4945]. 1C2 was raised against the full-length, wt TATA-binding protein (TBP) containing Q38, the most common polyQ allelic variant in the human population46,47,48 [wt TBP contains <Q4745]. In contrast, 3B5H10 was raised against a 171–amino acid fragment of mutant htt (Q65) prepared under native conditions [wt htt contains <Q3645]. Thus, the immunogens for MW1 and 1C2 would be expected to form the varied extended conformations characteristic of wt stretches of polyQ. In contrast, the immunogen for 3B5H10 may form the more compact, two-stranded, hairpin conformation of mutant polyQ seen in our SAXS experiments on mutant htt fusion proteins unbound to antibody.
The compact conformation of mutant polyQ we found for htt fusion proteins in solution unbound to antibody may seem surprising, as the polyQ stretch in purified Thio-Httex1-Q39-His6 reportedly has an extended linear conformation5. Further, extensive stable regions of α- or β-secondary structure in polyQ peptides have been ruled out through CD and nuclear magnetic resonance analysis, suggesting the polyQ stretch adopts a random coil conformation5. Our SAXS analysis also indicates that Httex1-Q39 lacks a stable secondary structure. For example, the Kratky plot of Thio-Httex1-Q39-His6 in solution has the typical shape of a protein with an extensive unfolded region (Fig. S9). Similarly, the maximum dimension of Thio-Httex1-Q39-His6 from SAXS (140 Å ± 10) is within experimental error of the predicted length of the protein (130Å) if it were random coil49 (Table S4). Finally, the calculated molecular volume of Thio-Httex1-Q39-His6 is 1.5 ± 0.2 times the volume expected for a globular protein (Table S4). This volume is within the range typical of other proteins in which 40–50% of the residues are predicted to be unfolded50. Thus, in aggregate, our data suggest that the polyQ stretch in mutant htt samples many conformations when unbound to an antibody, failing to form any consistent, stable secondary structure. However, the population average of these transiently sampled conformations is best represented by a compact, two-stranded, hairpin conformation.
In summary, we present experimental evidence of a compact, two-stranded, hairpin conformation of polyQ in htt fragments containing polyQ stretches long enough to cause neurodegeneration. This structure, which is abundant in htt fragments unbound by antibody, is recognized by a newly developed monoclonal antibody, 3B5H10. Since 3B5H10 binding predicts neurodegeneration better than MW1 binding13, this conformation may be neurotoxic. How this structure might lead to neurodegeneration is unclear. The pair of polyQ strands could be part of a structure that interacts with certain intracellular proteins or lipids, sequestering or regulating their function in deleterious ways1. Alternatively, the polyQ conformation could alter the normal functions of the proteins that contain them (e.g., htt), resulting in toxicity. If the length of the polyQ stretch partly governs the distribution of conformers39, it could explain how versions of these proteins with wt polyQ stretches can sometimes cause neurodegeneration when overexpressed51,52. Finally, while our data demonstrate that the 3B5H10 epitope appears predominately on monomeric htt13, if the conformation is exposed and available for interaction in aggregated protein under certain unique circumstances53, the conformation may mediate pathogenesis in such circumstances as well. If inclusion body formation then reduces the specific activity of these toxic conformers by sequestration or refolding, it could explain how inclusion body formation improves survival54,55.
Purification, sequencing, and crystallization of 3B5H10 Fab are described in Supplementary Methods.
Data collection and solutions for the orthorhombic (not deposited), monoclinic (PDB: 3S96), and alternate (PDB: 4DCQ) crystal forms of 3B5H10 Fab are described in Supplementary Methods (Tables S1–3). The final Rwork and Rfree for the monoclinic model (refined from 11 Å– 1.90 Å) were 0.19 and 0.26, respectively (PDB: 3S96).
Purified fusion protein constructs used for SAXS are described in Supplementary Methods. SAXS data were collected using protein concentrations in the range of 1–5 mg/ml and an x-ray wavelength of 1.11Å at beam line 12.3.1 (Advanced Light Source) or beam line 4–2 (Stanford Synchrotron Radiation Laboratory). Samples of the running buffer from the size-exclusion columns used to purify proteins and protein complexes for SAXS analysis were used for buffer subtraction. Data were integrated with software customized for each beam line and processed with the program PRIMUS56. The program GNOM57 was used to calculate the maximum dimension and the radius of gyration and to estimate the intensity of the scattering at zero angle. The dimensional data for each sample are summarized in Table S4. No significant differences were observed in the dimensional data across the concentration ranges tested for each sample. SAXS data was analyzed as described in Supplementary Methods.
Proteins were prepared in PBS and allowed to incubate together for at least 30 min at room temperature before the CD spectrum was measured. The complex of Thio-Httex1-Q39-His6 and 3B5H10 Fab was formed in a 1:1 stoichiometry. Protein concentrations for the experiment were 55 µg/ml of Thio-Httex1-Q39-His6 and 95 µg/ml of 3B5H10 Fab. Experiments were also repeated using 110 µg/mL of Thio-Httex1-Q39-His6 and 190 µg/mL of Fab with the same results. Similar analysis was performed on MBP-Httex1-Q46–Cer and the Fab at two different protein concentrations (data not shown). Conclusions from the MBP-Httex1-Q46–Cer experiments mirror conclusions from the Thio-Httex1-Q39-His6 experiments. CD was performed using standard conditions on a Jasco J-810 instrument. CD data analysis is presented in Figure S7 and Supplementary Methods.
Striatal neurons transfected with Httex1-Q46 C-terminally tagged with enhanced GFP were fixed for 15 min in a solution of 4% paraformaldehyde and 0.2% glutaraldehyde. Plates were then fixed in 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4, at room temperature for 2 min, then placed on ice for 30 min, washed with 0.1 M cacodylate buffer and postfixed with 1% osmium tetraoxide for 30 min. Cells were then prepared for electron microscopy in standard fashion, sectioned into 60-nm slices, and imaged on a Zeiss EM10 electron microscope (Supplementary Methods).
We thank P. Bjorkman for the Thio-Httex1-Q39-His6 plasmid, D. Minor (UCSF) for the pET28-HMT and pTEV vectors, and Raymond Boynton (N-terminal Protein Sequencing Facility, Biogen, Inc.) for assistance with N-terminal deblocking and sequencing of the 3B5H10 heavy and light chain polypeptides. We thank members of the Finkbeiner and Weisgraber laboratories for discussions, Greg Hura, Ken Frankel, and Thomas Weiss for assistance with SAXS data collection, Stephen Ordway and Gary Howard for editorial assistance, Kelley Nelson for administrative assistance, and Margaret Sutherland and Diane Murphy for their interest and support. Primary support for this work was provided by the High Q Foundation (S.F.), the National Institute on Aging (S.F.), and the National Institute of Neurological Disorders and Stroke (S.F.). Additional support was provided by the Taube Family Foundation Program in Huntington’s Disease Research, a Therapeutics Initiative Award from the Huntington’s Disease Society of America, the National Center for Research Resources, and the J. David Gladstone Institutes (S.F.). M.A. and J.M. are supported by the Hillblom Foundation. S.M. and J.M. are supported by the NIH-NIGMS UCSF Medical Scientist Training Program. J.M is supported by a fellowship from the UC Graduate Division Achievement Rewards for College Scientists. Crystallographic data and SAXS data were collected through the general user program at the Advanced Light Source and the Stanford Synchrotron Radiation Laboratory. K.H.W. is supported by the National Institutes of Health’s Grants HL-64963 and AG-020235.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
*Abbreviations: polyQ, polyglutamine; HD, Huntington's disease; htt, huntingtin; mAb, monoclonal antibody; SAXS, small-angle X-ray scattering; SDS, sodium dodecyl sulphate; HA, hemagglutin; CDR, complementarity-determining region; Cer, Cerulean fluorescent protein; MBP, maltose binding protein; TBP, TATA-binding protein; CD, circular dichroism
The coordinates of monoclinic form of 3B5H10 presented in the main text have been deposited in the PDB databank and have been assigned PDB ID 3S96 and RCSB ID RCSB065918. The alternate crystal form of 3B5H10 (Fig. S2e) has been assigned PDB ID 4DCQ and RCSB ID RCSB070164.
Supplementary materials related to this article can be found online at doi: