Serine phosphorylation suppresses huntingtin amyloid accumulation by altering protein aggregation properties

Abstract

Aggregation of expanded polyglutamine (polyQ) repeat containing fragments of the huntingtin (htt) protein may play a key role in Huntington’s disease (HD). Consistent with this hypothesis, two Ser to Asp mutations in the 17 amino acid N-terminal htt^NT segment abrogate both visible brain aggregates and disease symptoms in a full length Q₉₇ htt mouse model while compromising aggregation kinetics and aggregate morphology in a htt fragment in vitro (Gu et al., Neuron 64, 828–840 (2009)). The htt^NT segment has been shown to play a critical role in facilitating nucleation of amyloid formation in htt N-terminal exon1 fragments. We show here how these Ser to Asp mutations dramatically affect aggregation kinetics and aggregate structural integrity. First, these negatively charged Ser replacements impair the assembly of the α-helical oligomers that play a critical role in htt amyloid nucleation, thus providing an explanation for reduced amyloid formation rates. Second, these sequence modifications alter aggregate morphology, decrease aggregate stability, and enhance the steric accessibility of the htt^NT segment within the aggregates. Together these changes make the sequence-modified peptides kinetically and thermodynamically less likely to aggregate and more susceptible, if they do, to post-translational modifications and degradation. These effects also show how phosphorylation of a protein might achieve cellular effects via direct impacts on the protein’s aggregation properties. In fact, preliminary studies on exon1-like molecules containing phosphoryl-Ser residues at positions 13 and 16 show they reduce aggregation rates and generate atypical aggregate morphologies similar to the effects of the Ser to Asp mutants.

Huntington’s disease ^{1, 2} is one of ten ^{3, 4} CAG repeat diseases characterized by the expansion of a polyglutamine (polyQ) stretch in a particular protein. In HD, this polyQ stretch begins at amino acid 18 of the ~3200 amino acid long huntingtin (htt) protein. A hallmark feature of this disease is the presence of ubiquitinylated, aggregated forms of htt fragments in neurons of disease victims and in animal and cell models ^{5, 6}. Additional evidence of a disease role for protein aggregation and/or misfolding comes from observations that disease risk ³ and in vitro aggregation rates ^{7, 8} both increase with polyQ repeat expansion, as well as from the temporal association of aggregate formation and disease symptoms in some model organisms ¹. The precise identities and immediate cellular targets of these implicated toxic, aberrantly folded forms of htt have not been identified.

A dramatic example of the association of brain aggregates with HD symptoms was described by Yang and co-workers ⁹. This group previously constructed a tg mouse containing a Q₉₇ form of full length htt that exhibited accumulation of cortical and striatal aggregates and HD-like neurological deficits ¹⁰. In subsequent work, an identical Q₉₇ construct containing phosphomimetic Ser->Asp point mutations at positions 13 and 16 exhibited no disease symptoms and no accumulated visible aggregates ⁹. Although these effects might be accounted for via possible cell targeting functions of the N-terminal htt^NT sequence ^11–15 (Fig. 1), we speculated ⁹ that these Ser to Asp mutations could also alter the physical properties of cellular htt fragments, especially in the context of misfolding and aggregation. Consistent with this, we showed in that paper that the same mutations in a polyQ-containing htt fragment model peptide slow in vitro aggregation rates and yield significantly different aggregate morphologies ⁹. How the conversion of these two Ser residues to Asp might influence the folding and aggregation properties of such htt fragments, and how these changes might translate into the observed loss of aggregate accumulation in the mouse model, is the main subject of this report.

Figure 1.

Amino acid sequences of the peptides studied.

The nucleation mechanisms of polyQ amyloid formation have been analyzed in some detail, both in isolation ^{16, 17} and in the context of the sequences flanking polyQ in the htt protein ^18–22, and both mechanisms are operative in polyQ-containing fragments of htt ²³ (Fig. 2). Under most circumstances, the mechanism found in simple expanded polyQ repeat peptides, in which rare amyloid core nuclei are formed directly from within the monomer pool (Fig. 2B), is largely outcompeted in htt fragment aggregation by an alternative mechanism featuring the key involvement of the htt^NT sequence. In this much more efficient mechanism (Fig. 2A), htt^NT segments self-associate into α-helix rich tetramers and higher oligomers, which serve to locally concentrate and/or align polyQ segments and facilitate the formation of amyloid nuclei. Interference with the formation of α-helical oligomers suppresses the normally dominant A pathway, allowing enhanced contributions from the B pathway ²³.

Figure 2.

Schematic model of htt fragment aggregation mechanisms. PolyQ-containing htt fragments (a; htt^NT, green; polyQ, orange; Pro-rich, black) can undergo nucleation of polyQ amyloid formation in two ways. In path A, htt^NT segments undergo reversible formation of α-helical tetramers (b) and higher oligomers (c,d) whose disordered polyQ elements can interact to nucleate amyloid structure (e) and grow by monomer addition (f). In path B, the polyQ segment of monomeric htt fragments (a) can directly undergo reversible assembly into a critical nucleus (g) followed by elongation into amyloid (h). The conformation of htt^NT within the B pathway aggregates has not been determined.

In this paper we show that these Ser to Asp mutations within htt^NT influence aggregation rates, not by abrogating the α-helical oligomer pathway, but by decreasing its efficiency. We show that, in spite of ssNMR data showing aggregate secondary structures similar to fragments with WT htt^NT, Ser-modified htt fragment aggregates exhibit substantial differences from WT in EM morphologies, in decreased stabilities toward dissociation, and in increased susceptibility of their htt^NT segments to trypsin cleavage. These differences suggest several possible explanations for the dramatic phenotypes of the S13D/S16D tg mice. We also show that authentic phosphoryl-serine residues at positions 13 and 16 of polyQ-containing htt fragments exhibit aggregation properties similar to those of the corresponding Ser->Asp mutants, suggesting that this commonly used phosphomimetic mutation faithfully captures the aggregation-modifying effects of htt^NT phosphorylation.

RESULTS

Phosphorylation of htt^NT serine residues impairs aggregation

To assess the importance of the negative charges introduced in the Ser to Asp mutations, as well as the ability of these mutations to mimic the effects of Ser phosphorylation, we obtained htt^NTQ₃₇P₁₀K₂ (F17W) peptides with one or both of the Ser residues at positions 13 and 16 phosphorylated (Fig. 1) and assessed their effects on overall aggregation kinetics (Fig. 3a). We found that, like the S13D/S16D mutant (●), the doubly phosphorylated S13pS/S16pS peptide (●) exhibits substantially reduced aggregation compared to the WT peptide (●), with a greater rate suppression effect than that of the S13D/S16D mutant (Fig. 3a). Furthermore, overall aggregation rates of the single phosphoryl-Ser peptides S13pS (○) and S16pS (○) are also slower than the WT rate (●) and similar to S13D/S16D rate (●). Phosphorylation of these Ser residues also had effects similar to that of the S13D/S16D mutations ⁹ on aggregate morphologies in the EM. Like the S13D/S16D mutant peptides (Fig. 3 d–g), the intermediate and final aggregates of the double (Fig. 3, h–j) and single (Fig. 3, k–n) phosphoryl-Ser peptides resemble the protofibrillar and oligomeric intermediates observed on incubation of polyQ-containing htt fragments with WT htt^NT (Fig. 3c), and do not resemble mature fibrils (Fig. 3b). The results show that the Ser->Asp mutations are, in fact, good mimics of the disrupting effects of Ser phosphorylation on the ability of the htt^NT sequence to stimulate polyQ aggregation, producing reduced aggregation kinetics and final morphologies that resemble those normally seen for reaction intermediates. This emphasizes the role of the introduced negative charges on these modified properties.

Figure 3.

Aggregation of phosphoryl-Ser derivatives of htt^NTQ₃₇P₁₀K₂ peptides. (a) Sedimentation analysis of htt^NTQ₃₇P₁₀K₂ peptides in PBSA at 37°C by sedimentation analysis. WT (F17W) (●, 19 µM); S13D/S16D (●, 34 µM); S13p( ○, 21 µM); S16p (○, 20 µM); S13pS/S16pS (●, 19.5 µM). (b–r) Electron micrographs of the aggregates of different htt^NTQ₃₇P₁₀K₂ peptides isolated at different time-points. Aggregates of: WT (F17W) at 47 hrs (b) and after 1 hr (c); S13D/S16D after 100 hrs (d–g); S13pS/S16pS after 163 hrs (hd–j); S13p (kd–l) and S16p (md–n), both at 33 hrs; S13D/S16D aggregates at 1.5 hrs (od–p) and S13pS/S16pS aggregates at 2 hrs (qd–r). Scale bar = 50 nm.

Serine modifications affect oligomerization rates but not monomer conformations

One hypothesis for the molecular basis of HD posits that polyQ expansion leads to an alternative folded state within monomers that is responsible for pathology ²². To probe for effects of htt^NT serine modifications on the conformations of polyQ-containing htt fragments, we conducted solution spectroscopy experiments, working at low concentrations that disfavor helical bundle association and oligomerization ²⁰ (Fig. 4a). We found no significant differences in the CD spectra of freshly disaggregated (Methods) solutions of the WT (––––), S13D/S16D (––––) and S13pS/S16pS (––––) versions of htt^NTQ₃₇P₁₀K₂ (Fig. 4a). Furthermore, using FRET probes at positions -1 and 17 (Fig. 1) in the htt^NTQ₃₇P₁₀K₂ background, at 30 µM in PBS at pH 7.4, we found no difference in the estimates of average -1 to 17 distance for the WT (31.9 ± 0.1 Å) and S13D/S16D (31.8 ± 0.3 Å) peptides. Thus, there is no indication from these low resolution methods that these Ser modifications within htt^NT influence monomer conformation.

Figure 4.

Structural consequences of Ser modification during aggregation. (a) CD spectra of htt^NTQ₃₇P₁₀K₂ peptides (WT (black), S13D/S16D (red) and S13pS/S16pS (blue)) at ~ 35 µM in 10 mM Tris.HCl, pH 7.4; inset, secondary structure composition estimates from the CD curves (Methods). (b) Concentration dependence of α-helix structure in isolated htt^NT peptides of WT (●) and S13D/S16D (♦) sequence; inset, helical wheel analysis of htt^NT, with alanine residues in gray, hydrophobic amino acids in yellow, acidic residues in red, basic residues in blue, and neutral hydrophilic residues in violet. (c) Aggregation kinetics of htt^NTQ₈K₂ peptides by sedimentation (WT (●), S13D/S16D (♦)) and by Trp fluorescence of isolated aggregates (WT (○), S13D/S16D (◊)). (d) FTIR spectra of the aggregates of the htt^NTQ₈ peptides isolated at different times: WT at 48hrs (-----) and 840 hrs (––––); S13D/S16D at 48hrs (----) and 720 hrs (––––).

In contrast, we found that negatively charged residues at positions 13 and 16 impart a thermodynamic barrier onto the first step in the amyloid formation mechanism (Fig. 2A), oligomer formation. While the WT htt^NT sequence (●) undergoes concentration dependent, reversible assembly into α-helix rich oligomers with a mid-point of about 0.5 mM, the S13D/S16D mutant (♦) exhibits a much less cooperative assembly curve with a midpoint of about 0.9 mM (Fig. 4b). Furthermore, in peptides of the sequence htt^NTQ₈K₂ (Fig. 1), a background that supports very slow htt^NT-mediated nucleation of polyQ amyloid formation ²⁰, the S13D/S16D mutation (♦) greatly diminishes the time-dependent formation of sedimentable oligomeric intermediates compared to WT htt^NT (●) (Fig. 4c).

At first glance, the abilities of negatively charged groups at positions 13 and 16 to reduce the formation of htt^NT α-helix-rich oligomers is somewhat puzzling. The helical wheel diagram of htt^{NT 12, 13} (Fig. 4b inset) shows a strong amphipathic nature, and recent results suggest that the hydrophobic face of this helix plays a major role in oligomer stabilization (R. Mishra and R. Wetzel, Ms. in preparation). However, Ser13 and Ser16 are on the hydrophilic face of the α-helix, which is predicted to be solvent exposed and therefore not expected to play a major role in oligomer formation. One possible explanation of the surprising ability of these mutations to suppress α-helical oligomer formation is that this hydrophilic face may be involved in packing of α-helical tetramers into higher order oligomers that might be required for stabilizing oligomers and ultimately for amyloid nucleation. Alternatively, charge repulsion between positions 13 and 16 in the neighboring turns of the helix might destabilize α-helix formation and hence affect overall oligomer stability.

For both WT and S13D/S16D httNTQ8K2 sequences, the oligomers formed are α-helix-rich by FTIR (Fig. 4d). In spite of the much less favorable and slower formation of α-helix rich aggregates, the S13D/S16D htt^NTQ₈K₂ peptide is able to undergo nucleation of β-rich, amyloid-like structure with similar kinetics to WT htt^NT. This is shown by the previously described ^{19, 20} transition, within the aggregate pool, of the Trp at position 17 from a solvent exposed (λ_emm ~351–353nm) to a solvent-excluded (λ_emm ~345nm) environment (Fig. 4c, compare WT (○) to S13D/S16D (◊)), and in development of FTIR bands, including β-sheet (1625–1630 cm⁻¹), characteristic of polyQ amyloid structure (Fig. 4d, compare WT (––––) to S13D/S16D (––––)).

Serine modifications and polyQ amyloid nucleation and elongation

Collection of time dependent spectroscopic changes, as described above, is a useful approach for gauging aggregate structure in short polyQ-containing htt fragments, but is problematic for rapidly aggregating, long polyQ versions due to the difficulty in collecting useful amounts of early pre-nucleation aggregates. We therefore used other probes to assess the mechanistic impact of serine modifications in fragments with disease-relevant repeat lengths. The concentration dependence of initial aggregation rate has proved a good gauge of nucleation mechanism in this system. Thus, peptides undergoing nucleation without help from the htt^NT flanking sequence (i.e., Fig. 2B) typically exhibit log-log slopes in the range of 3 or higher ^{16–18, 21, 24, 25}, while peptides undergoing nucleation facilitated by a flanking htt^NT sequence (i.e., Fig. 2A) exhibit slopes in the range of ~1 ^{19, 21}. We found that the slope of an analogous log-log plot for the S13D/S16D mutant of htt^NTQ₃₇P₁₀K₂ is ~ 1.3 (Fig. 5a), consistent with a nucleation mechanism that is predominantly htt^NT-mediated (Fig. 2A).

Figure 5.

Aggregation kinetics consequences of S13D/S16D mutation in htt^NTQ₃₇P₁₀K₂. (a) Concentration dependence of initial phase of the aggregation reaction. (b) Inhibition of aggregation by 1:1 ratio of htt^NTQ₃ to peptide. WT alone (●, 50 µM) and with inhibitor (○, 48 µM) ; S13D/S16D alone (♦, 43 µM) and with inhibitor (◊, 68 µM). (c) Inhibition of aggregation by ~ 5-fold excess of elongation inhibitor PGQ₉P ^1,2,3 K₈ compared to peptide. WT alone (●, 22 µM) and with inhibitor (○, 26 µM); S13D/S16D alone (♦, 43 µM) and with inhibitor (◊, 34 µM).

Another useful probe is the response of the aggregation reaction to mechanism-specific aggregation inhibitors. Recently we reported that peptides consisting of htt^NT with few if any attached Gln residues are able to transiently inhibit htt^NT mediated amyloid nucleation by co-assembling with polyQ-containing htt fragments into a mixed α-helix-rich oligomer, thereby reducing the local polyQ concentration in the oligomers and hence the amyloid nucleation efficiency ²¹. We find that, in fact, the S13D/S16D mutant of htt^NTQ₃₇P₁₀K₂ responds to inhibition by an htt^NTQ₃ peptide with a similar retardation of aggregation onset as the WT peptide (Fig. 5b). Thus, consistent with the slope of the log-log plot (Fig. 5a), sensitivity to htt^NT inhibition indicates a major role for the htt^NT domain in facilitating the aggregation of the S13D/S16D mutant. We also previously described inhibitors such as PGQ₉P^1,2,3K₈ (Fig. 1). Such peptides inhibit the elongation of amyloid-like aggregates of simple polyQ and, because of the importance of elongation in the nucleation process, also suppress spontaneous amyloid formation by simple polyQ peptides ^{22, 26}. Presumably because the nucleation of htt^NT containing polyQ peptides is dominated by htt^NT-mediated oligomer formation, we find that the polyQ amyloid elongation inhibitor PGQ₉P^1,2,3K₈ gives no detectible inhibition of spontaneous aggregation of WT htt^NTQ₃₇P₁₀K₂ (Fig. 5c). Interestingly, however, this inhibitor does inhibit spontaneous aggregation of the S13D/S16D mutant (Fig. 5c). Further studies will be required to elucidate the mechanistic basis of this effect. However, the different responses of the WT and S13D/S16D peptides to the PGQ₉P^1,2,3K₈ inhibitor provides further evidence for the mechanistic impact of the negatively charged residues in htt^NT on aggregation.

Overall these mechanistic probes support the view that Ser modifications in htt^NT do not alter the fundamental, htt^NT-mediated nucleation mechanism of htt N-terminal fragments (Fig. 2A), but rather influence the efficiency of this mechanism. The reductions in overall aggregation rate previously reported for the S13D/S16D mutant ⁹, and reported here for the phosphoryl-Ser peptides (Fig. 3a), appear to be primarily due to less favorable formation of the α-helix-rich oligomeric intermediates that are the springboard for amyloid nucleation (Fig. 2A).

Serine modifications and aggregate structure and stability

Previously we reported magic angle spinning (MAS) ssNMR data on mature amyloid fibrils of htt^NTQ₃₀P₁₀K₂ peptides that confirmed the expected β-structure in many of the polyQ Gln residues while surprisingly showing that residues 4–11 of the htt^NT reside in stable α-helix ²⁷. We acquired the corresponding S13D/S16D peptide containing U-¹³C,¹⁵N-labeling of particular residues within the α-helix, the polyQ segment and the intervening linker region, then prepared mature fibrils and collected MAS ssNMR data. We used chemical shift indexing (CSI) relative to data on proteins of known structure to infer secondary structure at the labeled positions ²⁸. The results (Fig. 6) show that the labeled sites in S13D/S16D fibrils have chemical shifts, and thus structures, that are remarkably similar to that of the WT htt^NT peptide. Thus, the Lys6 and Ala10 residues of S13D/S16D exhibit the same kind of strong α-helical signal as found in the WT peptide aggregates. For Leu14, despite a slightly increased indication of α-helicity for the Cα shift alone, the overall chemical shift pattern for this residue in the S13D/S16D protein, as in the WT protein, is not indicative of a well-defined secondary structure. Note that electrostatic or structural effects due to the mutation of the neighboring S13 may affect the observed chemical shifts and in part explain these chemical shift changes. Finally, as in the previously reported aggregates with WT htt^NT, a specifically-labeled Gln18 is found in two chemically distinct environments, both of which are β-conformations. The S13D/S16D aggregates exhibit an essentially identical split β profile. Thus, in spite of significant differences in EM morphology, the S13D/S16D peptide aggregates are indistinguishable at the level of secondary structure from the WT htt^NT aggregates, at least at the labeled residues. This is consistent with the above kinetics data suggesting an unchanged aggregation mechanism compared with the WT sequence.

Figure 6.

Structural consequences of Ser modification in mature htt^NTQ₃₀P₁₀K₂ aggregates. Graphical representation of the ¹³C secondary chemical shifts (Δδ) obtained for indicated labeled residues, based on MAS ssNMR on mature aggregates from S13D/S16D (top) and WT (bottom) peptides. Δδ values were determined as differences from random coil chemical shifts for CO, Cα, and Cβ sites. The local secondary structure inferred from those values is shown. Red bars represent α-helicity, whereas blue bars β-sheet structure. Black bars indicate undetermined secondary structure. Split bars for Gln18 indicate two sets of observed NMR signals associated with two conformations, both β.

To complement these ssNMR results, we performed other tests probing for changes in aggregate structure and stability introduced by serine modifications. We exposed final aggregates of various htt^NTQ₃₇P₁₀K₂ peptides, after fractionation from any non-aggregated monomers by centrifugation, to limited trypsin proteolysis - a sensitive, if low resolution, probe of amyloid structure ²⁹, and analyzed the results by LC-MS. We found that mature WT aggregates are relatively stable to cleavage within the htt^NT segment, with only about 10% of the peptides being cleaved, all at Lys6 (Fig. 7a, black bars). In contrast, about 30% of the peptides in S13D/S16D aggregates (red bars), and nearly 50% of the peptides in S13pS/S16pS aggregates (blue bars), are cleaved by trypsin exposure under the same conditions. Interestingly, the single phosphoryl-Ser mutants, S13pS and S16pS, are both about 30% cleaved, the same level as the S13D/S16D aggregates; thus the sensitivity of these modified Ser aggregates to trypsin cleavage rank very similarly to their relative formation rates in the kinetics analysis (Fig. 3a). We also examined the trypsin cleavage sensitivities of the aggregates of some of these peptides isolated relatively early (30–35% completion) in the aggregation reactions. We observed trypsin susceptibilities in these early/intermediate aggregates very similar to those in the mature aggregates (Fig. 7a).

Figure 7.

Structural consequences of Ser modification in mature htt^NTQ₃₇P₁₀K₂ aggregates. (a) Fraction of molecules cleaved within htt^NT after exposure to trypsin in various mature aggregates (WT, black bars; S13pS/S16pS, blue bars; S13D/S16D, red bars; S13pS, red hatched bar; S16pS, blue hatched bar). (b) Sedimentation assays of aggregation kinetics and plateau levels of monomer for WT (●), S13D/S16D (♦) and S13pS/S16pS (▲) peptides in PBS at 37 °C. Inset: Dissociation of S13D/S16D aggregates in PBS at 37 °C by sedimentation assay.

In addition to a greater amount of cleaved material, we also observed “deeper” cleavages in the digests of fibrils from the htt^NT mutated peptides. Thus, while the small amount of cleavage of WT fibrils was all at Lys6, cleavage of fibrils from mutated htt^NT peptides occurred at both Lys6 and Lys9. About equal amounts of fragments from cleavage at positions 6 and 9 were obtained on digestion of the S13D/S16D mutant fibrils, while the position 9 fragment predominated in the digest of the S13pS/S16pS fibrils.

The changes in aggregate stabilities suggested by the limited proteolysis experiments are consistent with a robust measure of global stability ^{30, 31}, the concentration of monomeric peptide remaining when the aggregation reaction reaches equilibrium. This residual monomer concentration, which typically can be reached both in the aggregate association and dissociation directions ^{17, 30}, is essentially the critical concentration, or C_r, for the polymerization reaction, and hence is linked to the thermodynamic favorability of aggregate elongation ^{30, 31}. We found a surprisingly dramatic effect of the Ser modifications within htt^NT on this measure of aggregate stability. We found that the aggregation reaction of WT htt^NTQ₃₇P₁₀K₂ reaches equilibrium, with a C_r in the 0.3 µM range after about 50 hrs, with no further change up to 200 hrs (Fig. 7b, ●). In comparison, the corresponding S13D/S16D peptide aggregation reaction reaches a plateau in monomer concentration in the 2.5 – 3.0 µM range after about 100 hrs, with no further change out to 1400 hrs (Fig. 7b, ♦). A similar monomer concentration of 2 µM is reached when S13D/S16D aggregates are diluted and given the opportunity to dissociate toward equilibrium (Fig. 7b inset), showing the robustness of the measurement as a quantity associated with a dynamic equilibrium. The S13pS/S16pS peptide requires about 200 hrs to reach equilibrium where it indicates a C_r value in the 4.5 – 6.0 µM range (Fig. 7b, ▲). Based on these values, we calculated the aggregation destabilization (ΔΔG_ag) ³¹ attributable to the negative charges introduced at the Ser residues to be ~1.3–1.4 kcal/mol for the S13D/S16D mutation and ~1.7–1.8 kcal/mol for the S13pS/S16pS peptide. Since the C_r value is the concentration below which aggregation cannot occur, these C_r values may have important practical consequences for cellular aggregate accumulation.

DISCUSSION

Protein phosphorylation and dephosphorylation play an enormous role in cell regulation and cell pathway dysfunction linked to disease ³². Phosphorylation of the protein huntingtin ^{33, 34}, especially at Ser421 ₃₅ and within the N-terminal htt^NT sequence ^{14, 15, 36, 37}, has been implicated in controlling a variety of this protein’s cellular activities. Phosphorylation at serines 13 and 16 has been implicated in enhanced degradation of htt by the proteasome in cell models ¹⁴, in suppressed accumulation of htt aggregates in a mouse model ⁹, and in enhanced htt transport into the nucleus ^{14, 15, 37}. The dramatic effects on aggregate accumulation and HD phenotype of the phosphomimetic S13D and S16D mutations in the BACHD mouse model suggest that persistent phosphorylation of these residues in the cell could play an equally strong role in abrogating the effects of an expanded polyQ sequence in huntingtin. The possible impact of partial and/or transient phosphorylation is more difficult to assess from the phosphomimetic results.

Regardless of the implications for a strong role for phosphorylation, the dramatic effects of the S13D/S16D mutations in mice offer an opportunity to explore elements of the HD disease mechanism. That is, it is clear that mutation of these Ser residues to Asp, at only two positions out of over 3,000 residues in full length htt, dramatically neutralizes the aggregation effects of a Q₉₇ repeat in this system ⁹. We therefore focused our efforts on understanding how this double mutation influences various aspects of htt aggregation. Our previous results showed that replacement of Ser13 and Ser16 with Asp significantly reduced the rate of amyloid formation in a htt exon1-like peptide and dramatically changed the morphology of the aggregates ⁹. In that paper we speculated that diminution of the aggregation rate could affect the accumulation of aggregates by allowing the cellular mechanisms of aggregate eradication ^{38, 39} to better keep up with aggregation. Here we describe the underlying molecular basis of this diminution in aggregation rate, as well as some potentially equally important differences conferred on the product fibrils. We also show that the Ser to Asp mutation is in fact a reasonable approximation for the effect of stable, 100% phosphorylation of these two Ser residues on amyloid formation rates and amyloid structure.

In principle, there are a number of mechanistic steps that might be targeted by htt^NT sequence changes to explain observed aggregation rate reductions. These include control of monomer conformation, oligomer formation, amyloid nucleation and amyloid elongation. Our studies show that only one of these mechanisms is consistent with the data. We found that the S13D/S16D mutation does not appreciably affect the ensemble monomeric conformation of the htt^NTQ₃₇P₁₀K₂ peptides based on CD (Fig. 4a) and FRET (Results). (In contrast, Truant and colleagues described CD experiments showing significant reductions of α-helix in htt^NT peptides modified on the Ser residues, which they interpreted as evidence for altered monomer conformations ¹⁵. The different results may be due to sample disaggregation procedures, peptide sequence background used, or peptide concentrations used.) The htt^NT Ser modifications also do not appear to change the fundamental amyloid nucleation mechanism, which is equally dependent, for both WT and S13D/S16D peptides, on the initial formation of htt^NT-mediated α-helix rich oligomers (Fig. 5 a,b). Likewise, by using short polyQ versions of exon1-like peptides we show that the S13D/S16D mutations have no great effect on the amyloid nucleation and elongation reactions once α-helix-rich oligomers have formed (Fig. 4 c,d). In contrast to these negative findings, we observed that htt^NT segments containing the S13D/S16D mutations are thermodynamically impaired in their assembly into α-helix-rich oligomers (Fig. 4 b,c). Thus, the most likely source of the observed diminished spontaneous aggregation rates is a reduced ability to form the α-helix-rich tetramers and higher aggregates that serve as the species within which polyQ amyloid structure is nucleated ^{19, 20, 22} (Fig. 2).

These results for Ser to Asp mutations within htt^NT on the solution structure and reactions of exon1-like monomers are replicated in peptides in which one or both of the Ser residues are phosphorylated. Thus, the approximate secondary structural distribution within the monomer ensemble of the S13pS/S16pS peptide is indistinguishable from both the WT htt^NT sequence and the S13D/S16D double mutant (Fig. 4a). Likewise, aggregation of htt^NTQ₃₇P₁₀K₂ peptides in which one or both of the htt^NT Ser residues are phosphorylated is slowed dramatically (Fig. 3a); in fact the S13pS/S16pS peptide aggregates substantially less rapidly than the S13D/S16D peptide, presumably due to the more highly charged nature of the phosphoryl groups compared with carboxylate groups.

Thus, we obtained consistent data suggesting that the homogeneous, stable phosphorylation of Ser13 and Ser16, or their replacement with Asp residues, is expected to lead to slower kinetics of formation of amyloid fibrils due to the reduced stability of the transient oligomers that are required for amyloid nucleation. By slowing amyloid formation rates, these sequence alterations might allow cellular processes to better manage aggregate formation and therefore suppress aggregate accumulation. It also seemed possible, however, that alterations in the structures and properties of the product fibrils once formed might contribute to the tg mouse phenotype. We therefore compared aggregate structure and properties for WT peptides and peptides with modifications at residues 13 and 16.

We examined the final amyloid-like aggregates grown from the S13D/S16D htt^NTQ₃₇P₁₀K₂ peptide by ssNMR and found that, just as in WT fibrils ²⁷, residues 4–11 exist in α-helix, residue 14 is in irregular structure, and Gln18 is in β-sheet (Fig. 6). This result confirms the importance of α-helix formation by htt^NT in the aggregation mechanism as well as showing no major difference in residue specific secondary structural preferences between WT and mutant aggregates. In spite of these similar secondary structural profiles, however, further examination showed significant differences in structure and properties between WT and htt^NT-mutated exon1 fibrils.

For example, electron micrographs clearly show a difference in aggregate morphology ⁹ (Fig. 3). Thus, while WT htt^NT exon1 fibrils preparations are essentially homogeneous, exhibiting long, single fibrils with rough edges (Fig. 3b), final aggregates from peptides containing either Ser to Asp mutations or phosphoryl-Ser in the htt^NT segment are a mixture of morphologies including large bundles of thin filaments (Fig. 3 d,h), amorphous aggregates (Fig. 3 e,j), and short, protofibril-like structures (Fig. 3 f,g,i) reminiscent of an early reaction intermediate in the WT reaction (Fig. 3c).

Additionally, the final aggregates from htt^NTQ₃₇P₁₀K₂ peptides containing either S13D/S16D or S13pS/S16pS replacements are dramatically more susceptible to trypsin (Fig. 7a). Thus, WT fibrils subjected to limited trypsin digestion suffer less than 10% cleavage, all of which occurs at residue 6, the most distal Lys residue from the polyQ amyloid core. In contrast, S13D/S16D final aggregates under the same conditions are cleaved in about 30% of the constituent molecules, with a distribution of about equal portions of cleavage at Lys6 and Lys 9. Finally, final aggregates of the S13pS/S16pS peptide are cleaved within htt^NT in almost 50% of the constituent peptides, and most of the cleavages occur at Lys9 nearer the polyQ amyloid core. This suggests significantly greater access by trypsin, and therefore perhaps other enzymes, to the htt^NT segments within the aggregates containing negative charges at the normal Ser positions. This suggests that these aggregates may be more susceptible to in vivo proteolysis than WT aggregates. In addition, this indication of different degrees of steric access might be relevant to the known ability of htt fragments to be ubiquitinylated at the N-terminus ^{11, 14} and to suspicions that ubiquitin-targeted proteasomal degradation plays a large role in the cell’s ability to remove htt aggregates ¹. If ubiquitinylating enzymes have greater access to the Lys residues that are the targets of ubiquitin modification in the aggregates of phosphomimetic and phosphorylated peptides, compared to WT, this might suggest a greater ability of the ubiquitin system to target mutant aggregates for proteasome degradation, consistent with the lack of inclusion accumulation in the S13D/S16D BACHD mice.

Finally, the measured C_r values for various exon1 fragments suggest that mutations introducing negative charges at positions 13 and 16 of htt^NT abrogate the positive contribution of htt^NT to fibril stability. Thus, we previously reported C_r values for K₂Q₃₆P₁₀K₂ and K₂Q₄₀P₁₀K₂ peptide fibrils in the 2–4 µM range under the same growth conditions used in the experiments described here ¹⁸. The covalent attachment of htt^NT to such molecules, in the WT peptide htt^NTQ₃₇P₁₀K₂, reduces the C_r to the 0.3 µM range (Results). However, the C_r of a similar molecule containing negative charges at positions 13 and 16 reduces stability to C_r values of ~ 2.5 µM for the S13D/S16D sequence and ~5 µM for the S13pS/S16pS sequence (Results). Thus, introduction of these negative charges destabilizes htt exon1-like fibrils to a range essentially identical to fibrils from similar sequences lacking htt^NT entirely. The implications for exon1 fibril structure are not entirely clear. Part of the stabilization contributed by htt^NT could derive from the involvement of the aromatic residue at position 17 in β-structure ²⁷, and part could derive from helix-helix interactions. In any case, it is clear that the negative residues at positions 13 and 16 somehow compromise these htt^NT contributions and that this may have consequences for aggregate stability within the cell. These results are consistent with several recent model peptide studies demonstrating a general ability of phosphorylation to modify the kinetics and product structure of amyloid formation ^40–42.

Together, the results of analysis of aggregate structures and properties suggests that placing negative charges at positions 13 and 16 decreases aggregate thermodynamic stability (both in the early htt^NT-mediated oligomers and the mature fibrils) as well as increases the accessibility of the htt^NT terminus to enzyme modification. Although these results may appear to be inconsistent with the ssNMR results showing approximately equivalent secondary structures for the WT and S13D/S16D aggregates, we suggest that the results merely reflect the different methods’ complementary insights into fibril structure, dynamics and stability. Thus, the ssNMR experiments are exquisitely sensitive to the local structure of the immobilized material (i.e., the atomic structure of the protofilaments), but lack direct information on non-local interactions such as changes in morphology as seen by EM (i.e., supramolecular assembly of smaller units into fibrils). The ssNMR experiments also detect the overall signals of the bulk sample, in particular the most rigid parts of the molecule (see also ref. ²⁷). These points are brought out by the hypothetical model shown in Figure 8.

Figure 8.

A model for the htt exon1-like fibril structure, integrating observations from ssNMR, EM and trypsin cleavage. In the figures, polyQ β-sheets are in blue, more rigid htt^NT α-helices are in red, and more mobile htt^NT α-helices are in green. Positions 13 and 16 are in a segment lacking secondary structure ²⁷. The figure shows how increased mobility (green helices) caused by charge repulsion (b) leads to greater exposure of htt^NT sequences to trypsin access compared to more stably packed WT htt^NT (a). In both WT and mutated htt^NT, the bulk of the htt^NT helices (red) remain stably involved in structure and detectible by MAS ssNMR.

The Figure 8 model (as well as the Fig. 2f model) proposes that the htt^NT segments of peptides within a filament interact with and stabilize each other. It also proposes that the htt^NT segments may be largely responsible for the filament-filament interactions required to build up the thicker fibrils seen for WT exon1 like aggregates (Fig. 3b). This model is similar to a recently proposed model for the role of htt^NT in fibril structure based on EPR data ⁴³. Since the addition of negative charges in htt^NT occurs in the ‘linker’ part of the peptide between the β-sheet and α-helix in the mature fibrils ²⁷, it will not necessarily disturb the α-helix in residues 4–11 observed by ssNMR in the final aggregates. At the same time, charge repulsion could introduce destabilization that manifests in the EM, C_r and trypsin data presented. Thus, repulsion could interfere with fibril assembly resulting in the accumulation of filaments (compare panel b to panels d and h in Figure 2). Second, charge repulsion could reduce the htt^NT contribution to fibril stability, as observed in the C_r values (Fig. 7b). Third, charge repulsion could lead to more dynamic unfolding-refolding of the α-helices, at least at certain locations such as fibril ends, thus increasing the sensitivity of htt^NT to trypsin (Fig. 7a). The model in Figure 8 may account for why only a subset of the htt^NT segments (the green helices in the figure) are trypsin-sensitive, being in an environment (fibril termini in the schematic) that is particularly sensitive to charge repulsion effects. The model also suggests that the bulk of the htt^NT segments in both WT and S13D/S16D fibrils are expected to be found in α-helix in the MAS ssNMR experiments.

Taken together, our studies on the S13D/S16D peptide indicate a number of possible biophysical explanations for the absence of inclusions in the tg mice carrying these mutations. First, the mutations decrease the tendency of the disordered htt^NT sequence to engage in α-helical tetramer formation, reducing the steady state concentration of helical oligomers and consequently the rate of amyloid nucleation and growth. Second, the mature aggregates produced are less thermodynamically stable, to the extent that there is less of a driving force for their formation in the cell. Third, the htt^NT segment in these aggregates is more accessible to trypsin, and perhaps to other enzymes such as ubiquitinylating enzymes, potentially providing a biophysical mechanism leading to increased proteasomal degradation.

Mutations of Ser or Thr residues to Asp or Glu residues have long been successfully used as ribosomally programmable mimics of phosphorylated residues, in spite of the dramatically different sizes and pK_as of carboxylic acid and phosphoric acid moieties. Our results show that Asp mutations can also be reasonable mimics of the effects of phosphorylation of Ser residues on protein aggregate formation rates and properties. Many substrates for phosphorylation are intrinsically disordered proteins, and the mechanisms by which phosphorylation imparts cellular effects have a strong protein folding component ⁴⁴. Our results suggest one feasible pathway by which protein phosphorylation might generally impact protein misfolding and aggregation, by directly modulating the long term viability of disordered proteins in the cell. More specifically, the data provide a consistent mechanism to explain the lack of aggregate accumulation, and potentially the lack of toxicity, of S13D/S16D htt in a mouse model of HD. Since these Ser modifications suppress amyloid formation by first suppressing oligomer formation, the results suggest that α-helix rich oligomers, while not visible in the mouse brain, also remain among the list of suspects as possible toxic agents in HD.

MATERIALS AND METHODS

Materials

Water (HPLC grade), acetonitrile (99.8% HPLC grade), HFIP (99.5%, spectrophotometric grade), and formic acid were from Acros Organics, TFA (99.5% purity) from Pierce, and thioflavin T was from Sigma. Chemically synthesized peptides (Fig. S1) were obtained from either the Keck Biotechnology Center at Yale University (http://keck.med.yale.edu/ssps/) or from GenScript, Inc. and purified on an Agilent Zorbax C3 reverse phase HPLC column before use. Most peptides contained a conservative Phe->Trp replacement at position 17 previously shown to have little effect on aggregation rates ¹⁹.

General Methods

All peptides were rigorously disaggregated as described ^{45, 46}. Peptide concentrations were determined by analytical HPLC based on standard curves ⁴⁵ from stock solutions of purified peptides calibrated as described ²⁰. For Trp fluorescence, FTIR analysis and trypsin digestion studies, aggregates were isolated from reaction mixtures at specified time points by centrifugation at 21,000 × g for 45 mins and the pellets were washed 3 times by resuspension and resedimentation with PBS to remove traces of TFA and the other solutes. Aggregate concentrations were determined directly ⁴⁵ or estimated from the measured monomer concentrations of aggregation reactions. A sedimentation/analytical HPLC assay ⁴⁵ was used to determine aggregation and aggregate dissociation ^{17, 30} kinetics. Trypsin limited proteolysis was performed as described ⁴⁶ on aggregates isolated as described above using 12 hrs incubation at a 1:25 weight ratio of trypsin to aggregate in 50 mM Tris.HCl, pH 7.0 at 37 °C. Transmission electron microscopy was on uranyl acetate negative stained samples as described ^{17, 46}.

Spectroscopic methods

Far-UV CD measurements were performed on a JASCO J-810 spectropolarimeter using a 0.1 mm or 1 mm path length cuvette. Samples for CD measurement were prepared as described ⁴⁶. CD spectra were analyzed using the CONTINLL program ⁴⁷ from the CDPro package (lamar.colostate.edu/~sreeram/CDPro) in which the SP37A reference set (ibasis 5) was used to estimate the amount of secondary structure. For FTIR analysis, aggregates isolated from 1–2 mM reactions were analyzed on an ABB Bomem FTIR instrument. Spectra were obtained by averaging a total of 400 scans collected at RT with 4 cm_-1 resolution. Residual buffer absorption was subtracted and spectral components were identified from second-derivative minima using PROTA software (Biotools, Inc.). Tryptophan fluorescence measurements on aggregates were as described ⁴⁶. Methods for acquiring ss NMR data have been described ²⁷.

Limited trypsin digestion

For quantification of trypsin digestion, aggregates were isolated by centrifugation, resuspended in buffer, and the aggregate concentration determined by dissolving an aliquot in formic acid and quantifying by the HPLC assay. These aggregates were then incubated at concentrations of 25–50 µg/ml in 50 mM Tris-HCl, pH 7.0 for 12–16 hrs at 37 °C in the presence of trypsin at a 1:25 weight ratio of trypsin (SEQUENZ-Trypsin, Worthington Biochemical Corp.) to protein. At the end of the incubation, the aggregates were isolated by sedimentation, washed twice with 50mM Tris-HCl pH 7 buffer, then dissolved in formic acid and injected onto LC-MS (Agilent 1100 electrospray). Unique elution positions were obtained for full-length peptide and for peptides cleaved at positions 6 and 9. Structures of these peptides were confirmed by MS (see Supplemental Table 1). To quantify digestion, the A215 peaks of the peaks were integrated to determine the relative amounts of intact and nicked htt^NTQ₃₇P10K₂. The ratio of the area of the fragment(s) to the total (i.e., intact plus fragment(s)) peptide area was then used to determine the fraction of the total aggregate that was cleaved.

FRET measurements

FRET efficiencies were calculated based on a ratiometric approach on the basis of measured donor (I_D) and acceptor (I_A) fluorescence intensities from the protein tagged with a pair of FRET fluorophores (see Supplemental Methods).

HIGHLIGHTS.

• S13D/S16D mutations in ~3200 amino acid huntingtin rescues mice from polyQ toxicity

• In peptide models S->D and S->pS at 13 and 16 reduce aggregation kinetics

• These mutations suppress α-helical tetramerization required for amyloid nucleation

• Within the aggregates, they weaken stability and enhance access to the N-terminus

• Serine modification may thus reduce aggregate burden by several biophysical means