An Expanded CAG Repeat in Huntingtin Causes +1 Frameshifting*

Maintenance of triplet decoding is crucial for the expression of functional protein because deviations either into the −1 or +1 reading frames are often non-functional. We report here that expression of huntingtin (Htt) exon 1 with expanded CAG repeats, implicated in Huntington pathology, undergoes a sporadic +1 frameshift to generate from the CAG repeat a trans-frame AGC repeat-encoded product. This +1 recoding is exclusively detected in pathological Htt variants, i.e. those with expanded repeats with more than 35 consecutive CAG codons. An atypical +1 shift site, UUC C at the 5′ end of CAG repeats, which has some resemblance to the influenza A virus shift site, triggers the +1 frameshifting and is enhanced by the increased propensity of the expanded CAG repeats to form a stem-loop structure. The +1 trans-frame-encoded product can directly influence the aggregation of the parental Htt exon 1.

Homopolymeric amino acid runs are frequent in the human genome (Ͼ2000 protein-coding genes contain at least one repeat) (1). These runs, although indispensable as structural elements for the functional dynamics of protein domains, are highly susceptible to genomic rearrangements and mutational expansions that are causative for several human pathologies, including triplet repeat expansion diseases (2), lupus antigenic afflictions, and leukemia (1). Many of these mutation-expanded repeats are often bidirectionally transcribed, e.g. their translation involves non-canonical repeat-associated non-ATG initiation (3) or frameshifting (that is, a shift of the reading frame backwards (Ϫ1 frameshifting) or forward (ϩ1 frameshifting)), which generate diverse additional trans-frame products (4). These various products substantially alter aggregation of the zero-frame product (5,6) and can even be a central component of toxicity for some neurodegenerative disorders (6,7). Repeat sequence frameshifting also occurs in human DNA tumor viruses, Kaposi sarcoma-associated herpesvirus, and Epstein-Barr virus (8) or at mutation-expanded G strings in herpesvirus (9), suggesting that recoding events might also be frequent in the expression of other genes with repeat runs.
Expanded CAG repeat stretches, implicated in CAG repeat or polyglutamine (polyQ) 4 diseases (2), mediate Ϫ1 translational frameshifting, although the underlying mechanisms differ depending on the context surrounding repeat runs (5,6,10). Our earlier work shows that pathological expansion of the CAG repeats (Ͼ35 consecutive CAG codons) in huntingtin (Htt) exon 1, implicated in Huntington disease, involves stochastic Ϫ1 frameshifting within the CAG stretch that is triggered by limitation of the charged cognate glutaminyl-tRNA Gln CUG, although uncharged tRNA Gln CUG is plentiful (5). This is conceptually similar to the frameshifting characterized earlier at tandem rare codons in Escherichia coli that is caused by cognate tRNA depletion (11,12). In partial contrast, for the stochastic Ϫ1 frameshifting in decoding the expanded CAG run in the SCA3 gene associated with spinocerebellar ataxia type 3, an mRNA structure formed by the CAG repeats has been proposed as a major stimulatory component of the Ϫ1 frameshifting. A depletion of glutaminyl-tRNA Gln CUG is likely to be involved, as alternating bicodon CAA CAG repeats read by different tRNAs but with not too distant propensity to form secondary structure do not exhibit Ϫ1 frameshifting (6,10).
Post-mortem analyses of Huntington disease-affected brain tissues revealed traces of both polyalanine (polyA) or polyserine (polyS) proteins in the CAG-encoded polyQ aggregates of Htt (13). Although our early work has shown that polyA proteins derive from Ϫ1 frameshifting in the expanded CAG run of Htt exon 1 (5), the origin of the polyS proteins remained undefined. ϩ1 shift in the decoding frame of the polyCAG repeat would result in a polyACG stretch encoding serines. Here we set to define whether decoding of the Htt exon 1 also involves a ϩ1 frameshift in the zero CAG frame, which would result in synthesis of ϩ1 trans-frame encoded polyS segments. We report that, unlike the Ϫ1 frameshifting that occurs at any position, stochastically along the CAG repeat in Htt exon 1 with expanded pathological repeats (5), ϩ1 frameshifting is initiated at a sequence (UCC UUC C) upstream of the first CAG codon and produces only one polyS-containing species. The ϩ1 trans-frame polyS protein alters the aggregation of the zero-frame CAG protein and may contribute to the aggregation heterogeneity.

Results
A Slippery Site Upstream of the CAG Repeat Initiates ϩ1 Frameshifting in Expanded Htt Exon 1-To investigate the formation of ϩ1 frame-encoded protein(s), we fused the coding sequence of YFP in the ϩ1 reading frame 3Ј of the Htt exon 1 with a CAG repeat in the pathological (Htt51Q(ϩ1)YFP), nonpathological polyQ repeat length (Htt22Q or Htt7Q(ϩ1)YFP) and with Ala repeat Htt51A(ϩ1)YFP (Fig. 1A). HttQ(ϩ1)YFP FIGURE 1. The ؉1 frame product is the result of a ؉1 translational frameshift. A, nucleotide and amino acid sequence of the HttQ(ϩ1)YFP reporter with Gln 7 , Gln 22 , or Gln 51 and YFP in the ϩ1 frame. The amino acid sequence of the putative ϩ1 trans-frame product is depicted under the zero-frame product. B, Western blot analysis (IB) of N2a cells transiently expressing Htt51Q(ϩ1)YFP. Mock denotes cells transfected with empty plasmid. Anti-GFP antibodies also recognize the YFP moiety. C, immunoprecipitation (IP) of Htt51Q(ϩ1)YFP expressed in N2a/Htt103Q-CFP cells with anti-GFP antibodies and subsequent detection by immunoblotting with anti-HA antibodies. The input used for the immunoprecipitation is immunostained with anti-GPF antibodies. The asterisk indicates the ϩ1 trans-frame product, and # designates Htt103-CFP, whose CFP moiety is also recognized by anti-GFP antibodies. The expression of the positive, in-frame control Htt51S-YFP in N2a/Htt103Q-CFP is detected by immunostaining with anti-HA antibody. D, MALDI-TOF analysis in linear mode of Htt51Q(ϩ1)YFP expressed in N2a/Htt103-CFP for 24 h (blue spectrum) compared with the reference spectra of Htt51S-YFP (green) and YFP (red). Inset, peaks marked with an asterisk designate the periodic difference of 17 Da resulting from dissociation of a hydroxyl group from serine side chains. The table contains the theoretical amino acid sequence of the identified peaks 1-3. Note that peak 2 represents an incomplete peptide derived from 1 and peak 3 a typical peptide resulting from tryptic digestion of YFP. a.i., arbitrary intensity. E, amino acid sequence of the detected peptides in reflector mode. constructs were N-terminally tagged with an HA tag to monitor the total expression with the start in the zero-frame (Fig. 1B). Transient expression of Htt51Q(ϩ1)YFP in murine neuroblastoma cells (N2a) revealed a single band that was YFP-positive on immunoblots (Fig. 1B), i.e. corresponds to ϩ1 frame protein(s). ϩ1 frame products were immunoprecipitated using their YFP moiety and immunostained with anti-HA antibodies (Fig. 1C). The HA-detected product migrates as a 46-kDa protein (Fig. 1C), the same size as the YFP-immunodetected protein (Fig. 1B), suggesting that this product has both a zero frame-encoded N terminus and a ϩ1 frame-encoded C terminus, i.e. trans-frame encoded. Note that this approach detects products with a start in the zero frame; possible repeat-associated non-ATG-initiated products remain undetected by this approach.
The immunoprecipitated ϩ1 trans-frame products using their YFP moiety (Fig. 1C) were also subjected to mass spectrometry analysis (Fig. 1D). For detection reasons, we expressed Htt51Q(ϩ1)YFP in N2a/Htt103Q-CFP cells (14), as they produce larger amounts of the ϩ1 trans-frame-encoded protein than N2a cells (Fig. 1, B and C). Tryptic digested product was analyzed by MALDI-TOF in linear and reflector mode. Repeat proteins commonly deliver poor spectra, as they dwell poorly in the gas phase; some of them, polyQ proteins, are undetectable by this approach. MALDI-TOF analysis revealed the products SFS 49 NSR and S 49 NSR (linear mode; Fig. 1D, inset and table) in which the CAG repeat encoding polyQ is completely converted to a ϩ1-frame-recoded polyS product. The k.AFESLK.s peptide (Fig. 1E) originated from the N-terminal region of the intact Htt exon 1 in zero frame and was identified in reflector mode (Fig.  1A). We did not detect any product of ϩ1-frame decoding in the region 5Ј upstream of the CAG stretch. Also, the two amino acids (SF) in the SFS 49 NSR peptide specified by zero-frame decoding of the two codons adjacent to the polyQ stretch imply that the site of frame transition is immediately 5Ј of the repeat CAG codons. The periodical peak pattern in the spectrum is typical for fragments with multiple amino acids containing hydroxyl groups (Fig. 1D, inset).
To mimic a frameshifting event prior to the CAG repeat, an "in-frame" construct was made with a single nucleotide deletion 5Ј adjacent to the CAG repeat to bring serine codons into the zero frame (Htt51S-YFP). The product of Htt51S-YFP comigrated with the trans-frame product of Htt51Q(ϩ1)YFP (Fig. 1C), and MALDI-TOF analysis yielded the same pattern of tryptic digested peaks, resulting from Htt51S-YFP being more pronounced (Fig. 1D).
We next sought to determine the frequency of frameshifting. In most studies of reading frame transitions, the ratio of transframe-encoded product is compared with the product of standard zero-frame decoding (mainly by electrophoresis-based approaches) to calculate the frameshift frequency. However, the full-length zero-frame product of Htt51Q is highly insoluble and readily forms SDS-insoluble aggregates (5,15) that cannot be separated electrophoretically to monomers, thus precluding reliable quantitation. We adopted an approach used previously to determine Ϫ1 frameshifting of other polyQ-containing aggregation-prone proteins, e.g. ataxin-3 (10) and Htt (5). We calculated the fraction of YFP-positive cells (which reports on the presence of ϩ1 trans-frame product) from all HA-Htt51Q-expressing cells (i.e. immunostained positively for the HA tag). Note that this approach does not report on the fraction of trans-frame product from the zero-frame protein.
Is the Identity of the A Site Codon Significant for Frameshifting?-As deduced from the mass spectrometry analysis (Fig. 1, D and E), the site of frame transition is 5Ј adjacent to the repeat CAG codons run, i.e. at the sequence UUC CAG CAGC 50 (last zero frame, UUC, and first ϩ1 codon, AGC). UUC C resembles to some extent the ϩ1 slippery site UUU C in influenza A virus (16,17). However, in the virus context, the UUU C mutation to UUC C (which would be similar to the sequence in Htt) mediates less ribosomal frameshifting than UUU C (16). To explore whether a similar effect occurs in the Htt context, the wild-type UUC CAG of HttQ was changed to UUU CAG (underlining denotes a mutated nucleotide). This mutation increased the frequency of ϩ1 frameshifting from 10% for wild-type Htt51Q with the UUC CAG sequence to 40% with UUU CAG (Fig. 2A).
The frameshifting at the UUU C slippery sequence in influenza A virus is stimulated by a rare codon in the ribosomal A site, and its slow-to-decode nature is important for ϩ1 frameshifting (16). The counterpart A site codon in Htt51Q is the first CAG of the CAG repeat. Although CAG is a common codon, translation of expanded CAG runs depletes the cognate charged glutaminyl-tRNA Gln CUG, rendering CAG slow to decode (5). Increasing the amount of simultaneously translated CAG codons by expressing Htt51Q(ϩ1)YFP in N2a/Htt103Q-CFP cells increased the number of cells with ϩ1 frameshifting by 2-fold compared with expression in N2a cells (Fig. 1E). (N2a/ Htt103Q-CFP cells stably express Htt103Q, in which glutamines are encoded by alternating CAG CAA codons (14), i.e. 51 CAG codons per mRNA copy.) To investigate whether increased consumption of glutaminyl-tRNA Gln CUG by decoding consecutive CAG codons correlates with ϩ1 frameshifting, we down-regulated tRNA Gln CUG using a specific shRNA probe. Notably, partial knockdown of tRNA Gln CUG did not change the frameshifting frequency in N2a/Htt103Q-CFP cells expressing the Htt51Q(ϩ1)YFP reporter and was comparable with the control with scrambled shRNA (Fig. 2B). This suggests that ϩ1 frameshifting in Htt51Q is independent of tRNA Gln CUG.
Further proof that an induced hungry codon downstream of the slippery site did not play a role in ϩ1 frameshifting came from the following experiment. Variants with mutated UUC CAG (Gln) to UUC GAG (Glu) or UUC AAG (Lys) yielded a level of ϩ1 frame product comparable with that of the wildtype UUC CAG sequence in both N2a or N2a/Htt103Q-CFP cells ( Fig. 2A). Also, we detected a similar approximate increase of ϩ1 frameshifting with UUC AAG or UUC GAG in N2a/ Htt103Q-CFP compared with N2a, which we observed with the wild-type UUC CAG sequence in N2a/Htt103Q-CFP compared with N2a ( Fig. 2A).
Furthermore, we reasoned that the higher amount of transframe product in N2a/Htt103Q-CFP cells compared with N2a cells (Fig. 1F) might be due to differences in degrading it in both cells. To prove this, we expressed Htt51S-YFP with serine codons into the zero frame (Htt51S-YFP) in N2a cells and when co-expressed with a polyQ-protein in N2a/Htt103Q-CFP cells. Notably, Htt51S-YFP accumulated in detectably higher amounts in N2a/Htt103Q-CFP than in N2a (Fig. 2C, top panels, compare the 4-and 6-h expression time points). Homopolymeric polyQ peptides impair proteasome function (18), which most likely results in higher amount of Htt51S-YFP in N2a/ Htt103Q-CFP cells (Fig. 2C, bottom panels), although the Htt51S-YFP mRNA amount is equal in both N2a and N2a/ Htt103Q-CFP cells (Fig. 2E). We further determined the expression level of Htt51S-YFP in both cell lines in the presence of the proteasomal inhibitor MG-132 (Fig. 2C, right panels). The rationale behind this comparison is that if Htt103Q-CFP protein has already inhibited the proteasomal function, then the proteasomal inhibitor MG-132 would not additionally stabilize the Htt51S-YFP in N2a/Htt103Q-CFP cells. Indeed, MG-132 stabilized Htt51S-YFP only in N2a cells and marginally in N2a/Htt103Q-CFP cells (Fig. 2C, compare ϮMG-132 for each cell line). Together, these results suggest that the higher frameshifting frequency in N2a/Htt103Q-CFP cells is a result of the higher stability of the ϩ1 trans-frame product in these cells.
CAG Expansion Acts as a Stimulator for ϩ1 Frameshifting-mRNA structure 3Ј of productively utilized ϩ1 frameshift sites can stimulate frameshifting (19,20). Experimental evidence shows that elongated CAG runs exhibit a high propensity to form elongated stem-loop structures (21). This raised the ques-tion of whether the structural propensity of the CAG repeat influences the ϩ1 frameshifting in Htt exon 1. Thus, we generated two additional variants with shorter repeat lengths in the non-pathological range, Htt22Q(ϩ1)YFP and Htt7Q(ϩ1)YFP, for which secondary structures are predicted (Fig. 3, A-C), but their stability is much lower and they do not persist as experimentally shown earlier (21). For both Htt22Q(ϩ1)YFP and Htt7Q(ϩ1)YFP, we did not detect any ϩ1 trans-frame products when expressed in either N2a cells or N2a/Htt103Q-CFP cells (Fig. 2D), although they expressed at comparable levels (Fig. 3E). Moreover, no YFP-positive cells were detected even when the more shift-prone site UUU C was introduced in Htt22Q(ϩ1)YFP (Fig. 2D), suggesting that the stability of the 22-CAG run is not enough to stimulate ϩ1 frameshifting.
Together, these data imply that pathological expansion of the CAG repeat stabilizes the secondary structure (21) so that it acts as a ϩ1 frameshifting stimulatory element. To further address the implication that the secondary structure is a major determinant of ϩ1 frameshifting, we tested another construct without glutamine codons, Htt51A(ϩ1)YFP, which displays very similar secondary structure stability as Htt51Q(ϩ1)YFP. Its slippery sequence, UUC C, is intact (Fig. 3D). Our reasoning was to use a construct with identical nucleotide sequence to the polyCAG tract to ensure the same stability of the hairpin, but, unlike Htt51Q, it has a lower aggregation propensity (22,23). Indeed, the expression Htt51A(ϩ1)YFP displayed a similar frequency of cells with ϩ1 trans-frame products ( Fig. 2A). Cumulatively, our results suggest that the secondary structure is likely the major determinant of ϩ1 frameshifting.
The ϩ1 trans-Frame Product Predominantly Localizes in the Nucleus and Alters polyQ Aggregation-To determine whether the frameshift-derived product with 51 serines, Htt51S, inter- acts with the parental Htt51Q protein and thus potentially modulates its aggregation, we first analyzed the intracellular localization of Htt51S inclusions by confocal fluorescence microscopy. The YFP-positive trans-frame product localized almost exclusively in the nucleus; only very rarely (less than 5%) did we find a cell with additional cytoplasmic inclusions (Fig.  4A). The Htt51S-YFP control also exclusively localized in the nucleus (Fig. 4B). Nuclear ϩ1 trans-frame protein was diffusively distributed with small fluorescently denser loci, whereas the rare cytoplasmic counterpart formed larger, ring-shaped structures surrounding the core aggregates composed of Htt103Q-CFP (Fig. 4A). The mobility of nuclear and cytoplasmic ϩ1 trans-frame products differed markedly (Fig. 4C). The nuclear loci were composed of mobile species, as their fluorescence rapidly recovered after photobleaching, whereas the cytoplasmic structures contained species with reduced mobility, as they displayed much slower recovery (Fig. 4C). The nuclear ϩ1 trans-frame products seem to not alter the nuclear membranes; the continuous lamin B1 rims are indicative of an intact nuclear envelope (Fig. 4, A and B).
The co-localization of Htt51S with Htt103Q in the rare cytoplasmic aggregates raised the question of whether the polyScontaining product derived from ϩ1 frameshifting can influence the aggregation of the poly-glutamine protein. To test this, we added Htt51S in different concentrations at the onset of Htt52Q aggregation and monitored aggregation by light scattering at 90°and a filter retardation assay; the latter captures only large detergent (SDS)-resistant aggregates. Htt51S accelerated Htt52Q aggregation only at high concentration (Fig. 5A). The lag phase of Htt51Q aggregation was markedly shortened, suggesting an impact on the initial aggregation phase, whereas the exponential stage of aggregates growth remained unaltered (Fig. 5A). Notably, the SDS-resistant Htt52Q aggregates contained Htt51S (Fig. 5B), indicating that the two proteins co-aggregated. When individually subjected to aggregation, Htt51S exhibited no clear sigmoidal aggregation signature as observed for Htt52Q (Fig. 5C) and, unlike Htt52Q, formed no SDS-resistant aggregates within the time frame of the experiments (Fig.  5D). Importantly, the lack of SDS-resistance of Htt51S aggregates alone correlated with the observed motility of the nuclear  AUGUST 26, 2016 • VOLUME 291 • NUMBER 35 aggregates composed solely of Htt51AGC-YFP. Detergent-labile inclusions are highly mobile, whereas amyloid SDS-resistant aggregates display restricted mobility (24). Most likely, the cytoplasmic aggregates with restricted mobility (Fig. 4C) are amyloid Htt51Q/Htt103Q-CFP assemblies that can recruit Htt51S-YFP (monomer and/or preformed aggregates).

؉1 Frameshifting in Huntingtin
Together, these results suggest that the ϩ1 trans-frame Htt51S protein aggregated by itself and formed rather detergent-labile aggregates. When present in high concentration, it also co-aggregated with the amyloid Htt52Q aggregates, although it is unclear whether Htt51S was stably incorporated into the Htt52Q fibrils or rather co-assembles at the edges of Htt52Q aggregates.

Discussion
The CAG-encoded polyQ aggregates of Huntington diseaseaffected brain tissues contain traces of both polyA and polyS proteins (25). Although, in earlier work, we have shown that polyA proteins derive from Ϫ1 frameshifting in the expanded CAG repeat of Htt exon 1 (5), here we show that polyS proteins originate from ϩ1 frameshifting guided by a slippery sequence, UUC C, upstream of the CAG repeat. Both Ϫ1 and ϩ1 frameshifting are detectable only in Htt variants with pathological CAG lengths, but their underlying mechanisms clearly differ. Partial depletion of glutaminyl-tRNA Gln CUG stimulates Ϫ1  frameshifting in HttQ (5), whereas ϩ1 frameshifting is tRNA Gln CUG-independent (Fig. 2B) and is greatly enhanced by the stem-loop structure of the elongated CAG repeat.
Translational recoding with a shift of the canonical triplet periodicity movement of the ribosomes occurs more frequently than initially thought and is beneficial either for regulatory purposes and/or to expand the coding capacities of the genome (26,27). A typical recoding cassette involves a shift-prone site and a discrete enhancing signal 5Ј or 3Ј or both to the frameshift site (26 -33). The mechanisms of ϩ1 frameshifting are less defined than for Ϫ1 frameshifting, in part because Ϫ1 frameshift sites are easier to recognize, and their stimulatory elements are more idiosyncratic. Typically, Ϫ1 frameshifting is stimulated by secondary structures adjacent to the slippery site (29, 31, 34 -37). ϩ1 frameshifting in the expression of Htt with expanded CAG repeats resembles this pattern. A slippery sequence, UUC C, redirects linear readout from the 0 to the ϩ1 frame and is enhanced by the stem-loop structure of the elongated CAG repeat (21) compared with its counterparts with a CAG length in the non-pathological range. The shift site in Htt resembles that of the influenza A virus ribosomal frameshift site, UUU C, but is a rather atypical sequence with substantially lower frameshifting efficiency (16,17). The secondary structure downstream of the slippery sequence in Htt is greatly stabilized by the CAG expansion; exclusively expanded CAG repeats stimulate ϩ1 frameshifting. This secondary structure most likely pauses some of the translating ribosomes and allows for repairing of tRNA Phe AAG reading the UUC codon. It is unknown whether peptidyl-tRNA anticodon dissociates from the zero-frame codon UUC and repairs to the overlapping UCC or whether ϩ1 frameshifting proceeds without its repairing to mRNA. However, if repairing to mRNA after realignment takes place, it would involve the non-Watson-Crick rule (38) and include an A:C apposition at the central and most important position of the tRNA Phe AAG anticodon (39). It shares some similarities with the ϩ1 frameshifting utilized in yeast retrotransposon decoding: Saccharomyces cerevisiae Ty3 ϩ1 frameshifting has been proposed to occur in the absence of complete dissociation of peptidyl-tRNA anticodon loop pairing (40). Also, a hexanucleotide Ϫ1 frameshifting occurs in the apparent absence of peptidyl-tRNA repairing (41), and a low level of translational bypassing resumption can occur in the absence of peptidyl-tRNA pairing (42).
A large body of functional studies has been based on the assumption that single proteins are expressed from any gene associated with trinucleotide repeat disorders. The emerging evidence for alternative products generated through repeat-associated non-ATG translation (3,43), by translational Ϫ1 frameshifting (5, 6, 10), or, as revealed here, by translational ϩ1 frameshifting suggests that dynamic reprogramming of translation to generate alternative products from one gene is more common than previously appreciated. The toxicity of HttQ proteins is mainly attributed to nuclearly localized species (44,45). The almost exclusive localization of the ϩ1 trans-frameencoded polyS in the nucleus in our cellular model system and their higher dynamic mobility compared with the cytoplasmic aggregates suggest a likely contribution to toxicity at the cellular level. The estimation for the frequency of ϩ1 frameshifting in brain tissues is difficult, as the steady-state concentration of endogenous Htt mRNA is not known. However, the presence of polyS proteins that are the ϩ1 trans-encoded product of Htt in Huntington disease-affected brain tissues (13) provides evidence for the likelihood of ϩ1 frameshifting of endogenous Htt with an expanded CAG repeat in neuronal tissues, which may contribute to aggregation onset or even toxicity at the cellular level. Along this line, in ataxin 3, in which the CAG expansion is causative of spinocerebellar ataxia type 3, evidence has been presented suggesting that neither the expanded mRNA nor the zero-frame polyQ translation product are directly responsible for the cellular toxicity in Drosophila and mammalian neurons but, rather, that the Ϫ1-trans-frame-encoded product is a key (6).
Conceptually, the ϩ1 frameshifting in Htt with pathological expansion of the CAG repeat shows resemblance with repeat recoding in some human DNA tumor viruses (8). The common mechanistic signature between frameshifting in CAG repeat sequences and the examples of repeat recoding in DNA viruses imply that frameshifting might be commonly associated with translation of highly repetitive sequences that are frequently found in coding sequences of eukaryotic genomes.
More than 1000 cells from several independent biological replicates were imaged, and the frameshifting frequency was determined as a fraction of cells expressing YFP (which reports on ϩ1 frameshifting) from all HA-Htt51Q-expressing cells (i.e. stained positively for the HA tag using anti-HA antibodies). Because the frameshift frequency per cell differs, we set up a threshold of YFP fluorescence to consider a cell with frameshifting over the background fluorescence. The threshold value of 0.48 was determined by averaging the integrated background fluorescence of 50 N2a cells that do not express HA-Htt51Q(ϩ1)YFP. For the siRNA experiments, cells cotransfected (1:1 ratio, 1 g of total DNA) with the plasmid for expression of Htt51Q(ϩ1)YFP and pSuper bearing the shRNA against tRNA Gln CUG or scrambled control were imaged to determine the frameshifting frequency (5). Unless stated other-wise, results are expressed as mean Ϯ S.E. of n replicates. Differences between groups were evaluated using chi-square test and considered statistically significant when p Ͻ 0.05. In the proteasome inhibition experiment, to cells expressing different constructs for 24 h, 10 M MG-132 was added, and samples were withdrawn at various times.
For fluorescence recovery after photobleaching analysis, transfected cells were grown in in 35-mm Cellview TM dishes with glass slides (Greiner), and an area of 1.13-1.37 m 2 was photobleached for 6 s with a 514-nm laser wavelength at 100% power, and single images were collected before and every 1 s after photobleaching with an interval of 100 s.
Immunoblotting and Mass Spectrometry-A total of 250,000 cells were harvested and directly dissolved in 250 l of formic acid and incubated at 37°C for 40 min to maximally dissolve aggregates, and expression was analyzed either by SDS-PAGE or spotted directly onto nitrocellulose membrane in a slot-blot manifold. The detection was carried out by immunostaining with HA or GFP antibody (Roche).
RNA Isolation and Quantitative Real-time PCR-Total RNA was isolated from cultures expressing different HttQ variants using TRI reagent (Sigma). The integrity of the RNA was analyzed using 1% formaldehyde gel electrophoresis. 0.5 g of RNA was treated with DNase I (Thermo Scientific) and used for cDNA synthesis with oligo(dT) primers and Revert Aid reverse transcriptase (Thermo Scientific). The cDNA in 1:10 dilution was used in a quantitative real-time PCR (SYBR Green-based approach, Qiagen). Each reaction was performed in duplicates, and each run included control samples containing either no template or no reverse-transcribed transcript. Actin mRNA was used for normalization. Results are represented as mean Ϯ S.E.
In Vitro Aggregation Assays-Proteins were expressed as GST fusions on the pGEX-6P-1 plasmid (GE Healthcare) in E. coli BL21(DE3) and affinity-purified with glutathione-Sepharose 4 Fast Flow (GE Healthcare) as described previously (15). Aggregation was initiated by adding PreScission protease, which removes the GST tag, and monitored by means of a filter retardation assay (15) and static 90°light scattering at 532 nm (Quanta Master 30, PTI).
Author Contributions-P. S. performed most of the in vivo experiments. F. A. conducted the mRNA expression, protein stability, and shRNA experiments. R. S. performed the in vitro aggregation experiments. P. S. and Z. I. analyzed the data, conceived the concepts, planned and designed the experiments, and wrote the manuscript. J. F. A. contributed with significant input to mutation design, interpretation of the data, and manuscript writing.