Formation and Toxicity of Soluble Polyglutamine Oligomers in Living Cells

Background Aggregation and cytotoxicity of mutant proteins containing an expanded number of polyglutamine (polyQ) repeats is a hallmark of several diseases, including Huntington's disease (HD). Within cells, mutant Huntingtin (mHtt) and other polyglutamine expansion mutant proteins exist as monomers, soluble oligomers, and insoluble inclusion bodies (IBs). Determining which of these forms constitute a toxic species has proven difficult. Recent studies support a role for IBs as a cellular coping mechanism to sequester levels of potentially toxic soluble monomeric and oligomeric species of mHtt. Methodology/Principal Findings When fused to a fluorescent reporter (GFP) and expressed in cells, the soluble monomeric and oligomeric polyglutamine species are visually indistinguishable. Here, we describe two complementary biophysical fluorescence microscopy techniques to directly detect soluble polyglutamine oligomers (using Htt exon 1 or Httex1) and monitor their fates in live cells. Photobleaching analyses revealed a significant reduction in the mobilities of mHttex1 variants consistent with their incorporation into soluble microcomplexes. Similarly, when fused to split-GFP constructs, both wildtype and mHttex1 formed oligomers, as evidenced by the formation of a fluorescent reporter. Only the mHttex1 split-GFP oligomers assembled into IBs. Both FRAP and split-GFP approaches confirmed the ability of mHttex1 to bind and incorporate wildtype Htt into soluble oligomers. We exploited the irreversible binding of split-GFP fragments to forcibly increase levels of soluble oligomeric mHttex1. A corresponding increase in the rate of IBs formation and the number formed was observed. Importantly, higher levels of soluble mHttex1 oligomers significantly correlated with increased mutant cytotoxicity, independent of the presence of IBs. Conclusions/Significance Our study describes powerful and sensitive tools for investigating soluble oligomeric forms of expanded polyglutamine proteins, and their impact on cell viability. Moreover, these methods should be applicable for the detection of soluble oligomers of a wide variety of aggregation prone proteins.


Introduction
Numerous cellular proteins are regulated by interconversion between monomeric and oligomeric states. Distinguishing the different forms in live cells is challenging and requires the use of biophysical fluorescence techniques, including Förster Resonance Energy Transfer (FRET) and Fluorescence Correlation Spectroscopy (FCS) [1,2]or enzymatic reporters that amplify a signal from weak protein-protein interactions, such as yeast two hybrid or split luciferase systems [3,4]. While these methods can detect relative levels of oligomers, the methods do not distinguish the functional importance of monomeric and oligomeric species, which often coexist in cells. For example, if a cytotoxic protein can form soluble oligomers, is the oligomeric species necessarily cytotoxic? The problem is especially relevant in Huntington's disease (HD), other polyglutamine expansion diseases, and aggregation-prone protein diseases in general. Knowing whether monomers, soluble oligomers or both forms are cytotoxic will impact therapeutic strategies-i.e. which form of the protein to target.
HD is an autosomal dominant neurodegenerative disease correlated with the expression of a mutant form of the huntingtin protein (Htt). This ubiquitously expressed large protein contains 3144 amino acids and an uninterrupted series of CAG repeats that are translated into a polyglutamine (polyQ) tract. Fewer than 36 CAG repeats produces no phenotype. In contrast, individuals with a copy of mutant Htt (mHtt) containing a pathogenic length of 36 to 150 CAG repeats have HD [5]. The age of pathology onset correlates inversely with the number of repeats [6]. HD is one of nine described polyQ diseases. Expression of exon 1 of mHtt is sufficient to promote development of severe disease symptoms in mice, resembling the ones observed in HD [7]. Exon 1 includes the first 67 amino acids of full length Htt with an internal stretch of a variable number of glutamines. Many HD studies suggest a key feature distinguishing wildtype (wt) and mHtt N-terminal fragments (including exon 1) is the tendency of the mutant protein to irreversibly aggregate into SDS-insoluble cytoplasmic amyloid-like fibrils termed inclusion bodies (IBs) [8,9]. The role of the IBs remains controversial. While IBs have been associated with neuronal cell death [10,11,12,13], other studies find cells die without ever forming IBs and correlate IBs with increased cell survival [14,15,16,17,18,19]. IBs may act as a cellular coping mechanism to sequester and detoxify mHtt.
The increased tendency of mHtt to aggregate may represent an exaggerated behavior of an inherent property of wt Htt. Some biochemical data suggest wt Htt undergoes physiologic oligomerization in a regulatable manner [20]. For example, homooligomerization of wt Htt fragments can be increased by overexpression of p21 activated kinase (Pak1) [21]. Other live cell studies did not detect significant oligomerization of wt Htt ex1 , even with sensitive biophysical techniques such as FRET [18]. The functional consequences of oligomerization of wt Htt remain unclear.
Oligomerization of mutant polyQ proteins appears to be detrimental to cells [18]. Several studies have employed inhibitory molecules and antibodies to block polyQ oligomer formation and observed a corresponding decrease in cell death [22,23,24,25,26,27,28,29,30,31,32]. Thus, the soluble oligomers forming prior to IBs may represent the toxic species in HD [33]. Yet, other studies have raised the potential for the toxicity of conformational variants of the monomeric species of polyQ proteins [26,34]. The conformational hypothesis suggests reagents that block oligomerization may also be sterically obscuring toxic protein domains on monomers [33].
To better define the relevant species for polyglutamine toxicity, it would be useful to directly monitor both monomeric and oligomeric species, preferably in intact individual cells that could also be assayed for susceptibility to death. In addition, a method for increasing the cellular levels of soluble oligomers could more directly address the relative toxicities of Htt ex1 species. It was demonstrated that various mutant polyQ proteins (including mHtt ex1 ) can form soluble oligomers before IBs appear [18]. FRET approaches have successfully quantified interactions of mHtt ex1 in IBs in cells [18,35]. However, detection of FRET signals for the soluble mHtt ex1 fraction has yielded conflicting results [18,35]. FCS performed on lysates of cells expressing expanded polyQ-GFP also detected formation of soluble oligomers [30]. Detecting and characterizing the contributions of mutant polyglutamine monomers and oligomers in cells is fundamental to the understanding of polyglutamine diseases. Therefore, we sought to develop new approaches to monitor polyglutamine oligomer formation in living cells.

Experimental design
Two different approaches were employed to quantitate global changes in the monomeric mHtt pool and detect soluble oligomers in live cells. To study changes in the total population of mHtt ex1 , we measured changes in protein mobility, and by extension molecular size, with Fluorescence Recovery after photobleaching (FRAP). For FRAP, GFP-tagged proteins are expressed in live cells and imaged with a scanning confocal microscope. A discrete region of interest (ROI) in a cell is irreversibly photobleached with high intensity laser light. Movement of unbleached fluorescent molecules into the ROI is quantitated over time and analyzed to determine the diffusion coefficient (D) of the fluorescent molecule [36]. D (mm 2 /s) is inversely proportional to environmental viscosity and the size of the molecule (hydrodynamic radius or R h ) or an associated molecular complex [37]. Previously, FRAP analysis revealed that polyQ proteins, when incorporated into IBs, exhibit exceptionally low mobility [38]. However, these results provide no indication of the mobility of the polyQ protein prior its incorporation into IBs. While FRAP can report changes in molecular size, it provides no information on the composition of molecular complexes. Decreased mobility of mHtt proteins could represent mHtt ex1 oligomers, assemblies with other cellular proteins, such as chaperones or a mixture of both [39].
To specifically detect Htt oligomers, we employed split-GFP technology, also termed bimolecular fluorescence complementation (BiFC). Two non-fluorescent fragments of green fluorescent protein (GFP) can associate to form a single fluorescent GFP, but only when they are fused to two proteins capable of interacting with each other [40,41,42,43]. Split-GFP is sufficiently sensitive to report weak or transient interactions even for small populations of interacting proteins that could be otherwise obscured in sensitized emission FRET experiments [41]. The irreversible nature of the split-GFP interaction also allows trapping of otherwise transient interactions [44]. Therefore, this feature can be used to enrich the pool of oligomeric Htt ex1 in living cells [45]. Using FRAP and split-GFP, we sought to detect changes in soluble Htt ex1 oligomer levels in live cells and correlate oligomer formation with mHtt ex1 toxicity.

mHtt polyQ expansion and formation of IBs in neuronal cells
To study mHtt ex1 aggregation in living cells, Htt ex1 sequences containing 23, 73 or 145 CAG repeats were fused to monomeric GFP (Figs. 1A and B). When transfected into differentiated N2a cells, Q73 and 145 mHtt ex1 -GFP exhibited diffuse cytoplasmic fluorescence and gradually developed visible IBs. In contrast, the wt form, Q23 Htt ex1 -GFP, only exhibited a diffuse cytoplasmic distribution, similar to previous reports ( Fig. 1C and D) [46]. The Q145 construct developed IBs more quickly and in substantially more cells than the Q73 construct (Fig. 1D). Interestingly, the fluorescence intensity of the soluble Q145-GFP measured in cells not presenting IBs was significantly lower than the Q73 mutant and the wt Q23 Htt ex1 -GFP at multiple times post-transfection (Fig. 1E). This difference may reflect an increased propensity of the Q145 protein to be degraded by the proteasome or other degradation pathways, such as autophagy [47,48,49]. The fluorescence result is consistent with the observed comparatively low levels of Q145-GFP in the whole cell lysate immunoblot (Fig. 1B).
Potentially pathogenic soluble cytoplasmic forms of Q73 and Q145 mHtt ex1 -GFP constructs can also be detected using the 3B5H10 antibody, which exclusively stains soluble cytoplasmic mHtt, but not IBs (Fig. 1F) [34,50]. The 3B5H10 antibody staining intensity was significantly higher for mutant Q73 and Q145 constructs relative to the Q23 construct ( Fig. 1G) Recently, this antibody was shown to specifically recognize a potentially toxic conformation of mHtt monomers [33,34]. Interestingly, even when cells expressing Q145 Htt ex1 -GFP presented relatively low cytoplasmic GFP intensity, these same cells exhibited robust 3B5H10 staining, comparable with Q73 expressing cells (Fig. 1G). The poor correlation between GFP intensity and 3B5H10 staining likely reflects the affinity of the antibody for extended polyglutamine stretches or for a specific conformation of soluble mHtt constructs.

FRAP analysis reveals the incorporation of mHtt ex1 into soluble complexes
To determine whether the various constructs were suitable for FRAP analysis, we first established whether the GFP-tagged mHtt ex1 constructs were mobile in the cytoplasm and, thus, capable of diffusing. In Fluorescence Loss in Photobleaching (FLIP) an ROI is repeatedly photobleached and alternately imaged. All fluorescence within the cell will be depleted if the fluorescent molecules outside of the ROI are mobile and can pass through the ROI during the time course of the experiment [36]. FLIP experiments revealed that only the soluble pool of mHtt is mobile while the IBs appear to be very stable structures (Fig. 2), as previously described for mHtt and other polyQ proteins [51,52,53]. Similar mobility results were observed in photoactivation experiments, in which mHtt ex1 was fused to photoactivatable GFP and optically highlighted (Fig. S1).
Next, we used FRAP to determine whether we could detect quantitative differences in the mobilities of soluble polyglutamine constructs with increasing numbers of polyQ repeats. No detectable fluorescence recovery was observed when IBs were photobleached during the brief time course of the experiment (data not shown). We predicted all of the soluble Htt ex1 -GFP constructs would be sufficiently large to diffuse more slowly than cytoplasmic GFP alone. It was unclear whether the constructs would diffuse significantly differently from each other. Purified polyQ rapidly oligomerizes into 180 nm+ diameter aggregates in solution [54]. Such large structures would likely exhibit low, if any, mobility in cells [55]. Interestingly, full-length wt Htt with 15 polyQ repeats has a reported large diameter of 19 nm [56]. Soluble mHtt ex1 oligomers from cells are 4-50 nm in diameter [57]. Molecules larger than 20-30 nm are not expected to diffuse significantly within the cytoplasm [58]. In fact, some proteins as small as 14.4 nm appear effectively immobile in the cytoplasm [59]. Consistent with FLIP and photoactivation experiments, all of our Htt ex1 -GFP constructs were mobile in the cytoplasm (Fig. 3A). However, the Q145 mHtt ex1 -GFP exhibited significantly lower D values than the Q23 and Q73 Htt ex1 -GFP proteins (Fig. 3B). Similar results were observed in two additional cell types that were transiently transfected and analyzed by FRAP (Fig. S2). No significant correlation between D (mm 2 /s) values and the expression level of Htt ex1 -GFP (as measured by microscope detector gain) was observed (Fig. 3C).
While the sizes of the constructs could account for the mobility differences, diffusion is also inversely proportional to the environmental viscosity. Overexpression of Htt-GFP could potentially alter the cytoplasmic viscosity and confound interpretation of the changes in mHtt ex1 -GFP D (mm 2 /s) values. To control for viscosity changes, we co-transfected N2a cells with the various Htt ex1 -GFP vectors and an inert cytoplasmic reporter, mCherry. As in Figure 3B, we detected a reduction in the Q145 mHtt ex1 -GFP mobility relative to the other constructs. However, no significant change occurred for the mCherry when co-expressed with any of the constructs (Fig. 3D). Thus, the observed change in mobility for Q145 mHtt ex1 -GFP was not due to changes in cytoplasm viscosity. Therefore, we could calculate apparent size differences for the different constructs, using the Stokes-Einstein equation [37]. Assuming GFP has an R h of 2.3 nm [60], the GFP fusions to Q23, Q75, and Q145 had R h values of ,3.15, 3.4, and 3.7 nm, respectively. When factoring in the size of GFP, Q145 is 1.4 fold larger than the Q23 construct based on primary sequence. With this in mind, the 1.2 fold difference in R h suggests the constructs are monomeric or not assembled into larger complexes at early times post-transfection.
What happens to the sizes of the soluble constructs as cells begin to accumulate IBs? Previously, FCS analysis of cell lysates containing pure polyQ-GFP proteins revealed polyQ protein oligomers form and increase in size up to 8.5 fold over 72 h [30]. Similarly we observed the majority of Q73 and Q145 mHtt ex1 -GFP protein mobilities decreased in mobility in cells while the Q23 Htt ex1 -GFP D (mm 2 /s) values remain unchanged over 72 h (Fig. 3E). Thus, pathogenic polyglutamine expansion length constructs incorporate into mobile soluble complexes, distinct from IBs.

Detection of Htt ex1 oligomers using split-GFP in IBs free cells
Our FRAP results revealed incorporation of mHtt into increasingly larger complexes of unknown composition. To determine if at least some of these complexes contain mHtt oligomers, we turned to split-GFP technology to directly visualize soluble polyglutamine oligomer formation in living cells [61,62,63,64,65]. All three (Q23, Q73 and Q145 Htt ex1 ) constructs were fused at their COOH termini to split-GFP variants (1-157 and 158-238, we termed s157 and s238) [65] of superfolder GFP (Fig. 4A) [66]. A myc tag was also added to the NH 2 terminus of Htt ex1 constructs inserted in the 158-258 SFGFP vector. When expressed separately in N2a cells, split-GFP vectors displayed no detectable GFP fluorescence (Fig. 4B). The proteins could be differentially labeled by immunofluorescence to monitor overall expression levels in the absence of GFP fluorescence (Fig. 4B). Htt s157-GFP is preferentially recognized by an anti-GFP antibody [67], while Htt s238-GFP is recognized by the anti-myc antibody (Fig. 4B). When expressing both pieces of split-GFP fused to the various Htt ex1 constructs, all three pairs exhibited GFP fluorescence (Fig. 4C). All three constructs displayed GFP signals higher than the background levels detected for transfection of Q23 Htt ex1 s238 co-transfected with the negative control cytoplasmic protein, Nalp1B tagged to s157 ( Fig. 4C and D).
GFP fluorescence depended on the position of the split GFP. Co-transfection of cells with Q73 mHtt ex1 split-GFP constructs in which a COOH-terminal S157 Q73 construct was cotransfected with a Q73 construct with the s238 fused to the NH 2 terminus resulted in no fluorescence (Fig. 4C). Our results are consistent with the FRET results of Takahashi et al., who reported only detecting FRET when the fluorescent proteins were placed on the same end of the two interacting mHtt ex1 proteins [18]. Thus, mHtt ex1 proteins must assemble in parallel to form oligomers.
As we also observed for the intact Q145 mHtt ex1 -GFP fusion, the split-GFP constructs mean cytoplasmic GFP intensity measured in cells not presenting IBs was substantially lower than for Q73 ( Fig. 4C and D). This may reflect increased targeting of Q145 for degradation. Remarkably, we detected a robust GFP intensity for the wt Q23 constructs, demonstrating the wt Htt ex1 can oligomerize [21]. We discuss the implications of this observation in the next section. When compared to the expression level of the whole intact GFP construct, Q73 mHtt ex1 split-GFP fluorescence was significantly lower, even though the total level of soluble toxic mHtt is similar as measured with 3B5H10 levels by immunofluorescence (Fig. S3). These data indicate mHtt ex1 oligomers represent a minor population of the total mHtt ex1 pool. In theory, this is to be expected. Since 50% of Htt molecules fused to the same split-GFP moiety could form dimers and the resulting homodimers would be nonfluorescent, the split-GFP signal should only represent, ideally, 50% of the total pool of Htt oligomers. Furthermore, if the same total amount of split-GFP constructs relative to intact Q73 GFP is expressed, then the maximal possible split-GFP fluorescence signal is only 50% of intact GFP. Together, these two qualifications should make split-GFP cells only 25% as bright as intact GFP cells, at best. This assumes all of the split-GFP constructs interact to form fluorophores.
Next, we compared the mobilities of intact and split GFP Q73 constructs 16h after transfection, in cells without IBs. Q73 mHtt ex1 -split-GFP exhibited a four fold decreased mobility relative to intact Q73-GFP (mean D6 standard errors of 22.860.7 vs. 6.560.5) (Fig. 4E). This difference is significantly larger than would be anticipated for simply dimerizing the two proteins. Even if the molecules were arranged end to end, the predicted increase in R h would only be two fold, with a corresponding two-fold decrease in D. Therefore, the split-GFP Q73 construct oligomer must contain multiple dimers or associate with cytoplasmic chaperones. Furthermore, as FRAP is an ensemble measurement, our results suggest a mean increase in oligomer complex size, which could easily include the 8.5 fold larger complexes reported in the FCS study [30].

Incorporation of wt Htt ex1 into mHtt ex1 oligomers
Several lines of evidence suggest HD pathology is due to a gain of function of the mHtt as a result of the extended polyQ stretch [68]. However, an alternative hypothesis suggests the HD phenotype is due to loss of function of the wt Htt protein [69]. Consistent with this hypothesis, mHtt ex1 protein can recruit wt Htt ex1 into insoluble aggregates both in vitro and in vivo [70,71,72].
It remains unclear whether wt binds directly to IBs or can interact with mHtt prior to incorporation into IBs. We investigated whether wt/mHtt ex1 soluble oligomers format in living cells. First, we confirmed our mHtt ex1 could sequester wtHtt ex1 into IBs. Q23 Htt ex1 -GFP and Q145 mHtt ex1 -mCherry were co-expressed in N2a cells. Wt Q23 was readily detectable with Q145 in the IBs (Fig. 5A). Next, we used FRAP to determine whether coexpression of wt and mHtt affected diffusion of both proteins (Fig. 5B). Importantly, we found that expression of Q145 mHtt ex1 significantly decreased Q23 Htt ex1 -GFP mobility (Fig. 5C). Data from Figure 3D established oligomer formation does not generally slow the mobility of other proteins in the cytoplasm. Thus, the FRAP data in Fig. 5B suggest wt protein incorporates into larger soluble complexes in the presence of mHtt ex1 . To test whether the complexes include oligomers of wt and mHtt ex1 , we used the split-GFP approach to directly visualize hetero-oligomer formation between wt and mHtt. The robust fluorescent signal revealed the ability of mHtt to recruit wt protein into soluble oligomers (Fig. 5D). These results were subsequently validated with immunoprecipitations. When coexpressed in N2a cells, Q23 Htt ex1 -GFP was able to pull down Q145 mHtt ex1 -mCherry to a level similar to cotransfected Q145 Htt ex1 -GFP with Q145 mHtt ex1 -mCherry (Fig. 5E). Experiments were performed at 16 h posttransfection to minimize the number of IBs.

Increased oligomerization correlates with increased cell death
In light of our findings, we asked if increased mHtt ex1 oligomerization affected IBs formation and mHtt ex1 toxicity. In their FRET study, Takashi and colleagues reported the soluble form of mHtt ex1 represents the toxic species in HD [18] and IBs are protective [14]. However, they did not observe a significant correlation between mHtt ex1 oligomer formation (FRET positive) and cell death when compared to FRET negative cells. The result suggests both mHtt oligomers and b-strand polyQ monomers are potentially toxic to cells. In addition, the conformation of mHtt ex1 may be a determinant of its toxicity [76].
To test whether or not incorporation of mHtt into potentially oligomeric complexes impacts cell viability, we used our FRAP assay to determine if mHtt mobility correlates with neuronal cell survival. N2a cells were transfected with the live cell apoptosis reporter ER-DEVD [77]. The reporter localizes to the endoplasmic reticulum (ER). Upon induction of apoptosis with staurosporine, activated caspases cleave the DEVD peptide in the reporter, releasing it from the ER membrane, followed by translocation to the nucleus (Fig. S4). N2a cells were cotransfected with ER-DEVD-tdTomato and then Q145 mHtt ex1 -GFP and analyzed by FRAP 48 h posttransfection. FRAP was performed on cells without (no caspase activity) and with nuclear localized reporter (+caspase activity) and Q145 mHtt ex1 -GFP mobility was calculated. Consistent with the mHtt FRET reporter results, no relationship was observed between mHtt mobility and cell death (Fig. 6A) [18]. The FRAP results do not resolve the relative contributions of monomeric and oligomeric mHtt ex1 , each of which may be equally toxic.
To investigate the cellular consequences of increased levels of mHtt ex1 oligomers, we exploited the irreversible nature of the split-GFP complementation system. We hypothesized the split GFP would increase the concentration of oligomers in two ways. First, the irreversibility of split-GFP makes it a powerful trap for normally transient interactions [44]. Second, the nonproductive dimers formed by constructs containing the same split-GFP fragment will not be trapped and, should, associate at some rate with a complementary split-GFP fragment to be trapped. Thus, split-GFP should increase the fraction of detectable (by GFP fluorescence) and irreversible dimers of mHtt ex1 . Indeed, when run on a native polyacrylamide gel, Q23 Htt ex1 fused to split-GFP constructs exhibited formation of higher molecular weight complexes when compared to the Htt ex1 fused to intact GFP migrated much more rapidly as a small, presumably monomeric species (Fig. S5). Interestingly, Q23-s157 was unable to induce the formation of higher oligomers when expressed with Q23 fused to intact GFP (Fig S5). This result confirmed the ability of split-GFP to stabilize otherwise transient interactions. Based on these data, we then asked if trapping mHtt ex1 -mHtt ex1 interactions would increase IBs formation. Fusion of mHtt ex1 to split-GFP enhanced both the rate of IBs formation and the number of cells with IBs, relative to mHtt fusions to intact GFP (Fig. 6B). Our data do not exclude the possibility that monomeric mHtt ex1 may also be recruited to IBs [33].
Next, we asked whether increased levels of soluble mHtt oligomers enhance cytotoxicity of polyglutamine proteins. Following transfection, levels of plasmid expressed proteins should steadily increase. Higher protein concentrations should increase the probability of protein oligomerization and aggregation. Cells expressing mHtt ex1 fused to intact GFP exhibited increasing levels of caspase activity with increasing time post-transfection (Fig. 6C). No caspase activity was observed for cells expressing Q23-GFP (Fig. 6C). When Htt-GFP expressing cells were categorized as having or lacking IBs, a significant trend was revealed. Consistent with previous reports, cells had reduced apoptotic activity when IBs were present compared to cells with diffuse cytoplasmic mHtt, [14,18] (Fig. 6C). The percent of cells with caspase activity remained low over 48h for Q73-GFP expressing cells containing IBs, but increased about 50% for cells expressing Q145-GFP (Fig. 6D). These data suggest the presence of IBs could help protect cells from mHtt cytotoxicity and are consistent with previous reports [14,18]. As with the FRAP data, it remained unclear whether soluble monomers or oligomers preferentially affected mHtt-GFP cytotoxicity.
With the split-GFP constructs, two trends immediately stand out. First, mHtt split-GFP expressing cells have substantially higher levels of caspase activity and this occurs in a much higher percentage of cells compared to cells expressing intact GFP constructs (Figs. 6C and D). Interestingly, the dramatically increased caspase activity was observed earlier than for cells expressing intact GFP constructs (Fig. 6C). Second, fusion of mHtt ex1 to split-GFP significantly increased the average caspase activity per cell regardless of the presence of IBs compared to cells expressing intact GFP fusions to mHtt ex1 (Figs. 6C and D). Soluble mutant polyglutamine expression levels of both the intact or split-GFP correlated with increased caspase activity (Fig. 7). Importantly, higher levels of soluble oligomers strongly correlate with increased cytotoxicity (Fig. 7). Caspase activity typically only occurred above some threshold level of any of the mHtt ex1 constructs (Fig. 7). That is, mHtt ex1 must achieve a critical concentration in cells to induce caspase activation. Taken together, our results strongly implicate soluble polyglutamine oligomers as an important (if not the predominant) toxic species in polyglutamine disease pathogenesis.
The intact and split-GFP construct results indicate IBs are neither required for toxicity nor do IBs appear to enhance polyglutamine toxicity. Notably, a high percentage of cells expressing mHtt ex1 split-GFP contained IBs and exhibited caspase activity (Fig. 6D). This contrasts with the report of Arrasate et al. [14,18] describing how the presence of IBs formed from intact GFP mHtt ex1 constructs correlated with enhanced cell survival.
Could split-GFP, itself, cause cytotoxicity? We observe Q23 split-GFP also forms oligomers, but with no apparent cellular toxicity (Figs. 4C, D, and 6C). Split-GFP alone caused no apparent toxicity. As the only differences between Q23 and the Q73 and Q145 constructs are the numbers of polyglutamine repeats, we suggest cytotoxicity is due to oligomerization of extended polyglutamine proteins. Also, the use of split-GFP does not alter the trend of Htt ex1 toxicity, since mHtt ex1 fragments tagged to regular GFP are also toxic. However, increased mHtt ex1 oligomerization due to split-GFP does, in fact, result in an increase in the absolute toxicity of mHtt ex1 intermediate oligomers.
It is possible the split-GFP IBs could represent a novel cytotoxic species or could lack cytoprotective properties of intact GFP IBs. We consider it more likely that the rates of oligomer formation and IBs formation represent the relevant parameters for predicting cell survival outcomes. While we observed higher numbers of cells with split-GFP IBs and the rate of IBs formation was faster (Fig. 6B), mean caspase activity was not significantly higher than in cells with soluble cytoplasmic split-GFP extended polyglutamine constructs (Fig. 6C). This would suggest IBs could form at the time of or even after caspase activation. Furthermore, the constitutive production of split-GFP polyglutamine constructs with a high tendency to oligomerize could substantially increase the rate of soluble polyglutamine oligomer formation regardless of the presence of IBs, possibly overwhelming oligomer sequestration by the IBs formation pathway.
Formation of IBs and related aggresome structures is a cell regulated process. Relatively small oligomers or aggregates are specifically trafficked by the motor dynein along microtubules to the perinuclear microtubule organizing center [78]. There, cellular proteins, including p62, stimulate autophagic degradation of IBs [78]. We hypothesize that either sequestration of oligomers into IBs or the rate of autophagic clearance may not be sufficiently accelerated in the split-GFP system to cope with higher soluble oligomer levels. The split-GFP system now provides a powerful tool for investigating the kinetics and responsiveness of the IBsautophagosome pathway to exceptionally high levels of soluble oligomers.

Conclusions
Using two different and complementary approaches, we have followed changes in the oligomeric state of soluble mHtt ex1 in cells. The FRAP approach revealed increasing tendencies of pathogenic polyglutamine expansions to incorporate into progressively larger soluble complexes. These complexes are likely to include chaperones and potentially other proteins [75,79,80]. However, FRAP does not report whether complexes contain more than one mHtt ex1 molecule or even the general composition of complexes. Our FRAP results are consistent with the native gel results of Legleiter et al. [33] and Takahashi et al. [30], who observed greater than 50% of pathogenic polyQ constructs incorporate into complexes significantly larger than a nonpathogenic construct.
Split-GFP constructs revealed the capacity of all of the constructs to form oligomers of at least two mHtt ex1 molecules. However, the trapping ability of split-GFP prevents us from quantitating the fraction of dimer or higher oligomers that normally would exist at steady state. A recent study by Takahashi et al [30] reported detecting soluble complexes containing at least 3.5 monomers. Similarly, Ossato et al [81] measured oligomers of 5-15 mHtt ex1 molecules in cells, using a ''number and brightness method.'' Soluble oligomers never constituted more than 20% of the total soluble mHtt ex1 in the Ossato et al. study. However, our data suggest the majority ($50%) of mHtt ex1 -GFP reporters must be incorporated into complexes of significantly increased size to cause the observed decrease in D for FRAP and the slowed migration in native gels. This is because native gels and FRAP experiments are ensemble measurements, which detect relatively large changes for the entire population of proteins.
Previously, increasing levels of soluble polyglutamine oligomers in cells has required either expressing higher levels of polyglutamine proteins or regulating a co-factor of polyglutamine oligomerization. For example, Luo et al. stimulated oligomerization of mHtt fragments by overexpression of Pak1 [21]. Increased IBs formation and cell death followed, though the assay used did not directly investigate cell fates relative to the presence or absence of IBs. Split-GFP trapping of polyglutamine oligomers enabled us, for the first time, to directly demonstrate the higher propensity of soluble polyglutamine oligomers to form IBs and the increased toxicity of soluble oligomers in living cells without manipulating other cellular parameters. However, by artificially trapping monomer interactions and accelerating IBs formation, the split-GFP method increases the normal rate of conversion of monomers into oligomers and the flux of oligomer integration into IBs. Thus, the split-GFP method underemphasizes any potential role monomer levels play in cell fate. While we cannot exclude the potential toxicity of mHtt ex1 monomers, our data establish that driving oligomerization clearly enhances the cytotoxic potential of mHtt ex1 .
Taken together, an emerging picture of mHtt oligomerization in cells suggests all Htt ex1 constructs, including the nonpathogenic Q23, can form transient oligomers. Pathogenic polyQ constructs progressively incorporate into large complexes in cells, some of which include additional copies of polyQ constructs. Formation of complexes containing two or more copies of pathogenic polyQ molecules substantially impairs cell viability. Future studies will need to tease apart the relative toxicities of different soluble oligomers, i.e. do increasingly larger numbers of polyQ molecules in complexes increase cytotoxicities. Htt ex1 fused to GFP or split-GFP and ER-DEVD tdTomato for 48 h. GFP or split-GFP intensities and the nuclear/ER ratio of the apoptosis reporter were quantified for individual cells not presenting IBs. In each plot, the cell with the brightest GFP intensity was defined as having a mean intensity of 100 arbitrary units and other cell intensities were converted to this scale. Thus intensities in one plot are not directly comparable to intensities in another plot. Each square represents a single cell. Increasing mHtt ex1 , but not wt Htt ex1 , levels above some threshold were associated with increased cell death. The existence of apparent thresholds correlating with activation of the caspase activity suggests mHtt ex1 toxicity is sharply concentration dependent and may depend on titration of one or more key cellular factors. doi:10.1371/journal.pone.0015245.g007 Our findings support the hypothesis that mHtt oligomers are indeed the toxic species in HD. Given the increasing body of evidence against a pathogenic role for IBs in polyglutamine diseases, we suggest therapeutics that block formation of soluble polyglutamine oligomers merit investigation. The split-GFP methodology is well suited for a noninvasive high throughput screen to identify compounds capable of inhibiting polyglutamine oligomerization in cells.

Constructs and transfection
Htt ex1 Q23, Q73 and Q145 constructs were obtained from the Coriell Institute for Medical Research and the CHDI Foundation. Htt ex1 fragments were amplified by PCR using the following primers: forward primer GATCAGATCTGCCACCATGGCGACC-CTGGAAAAG reverse primer GATCACCGGTCCTGGTCGGTGCAGC-GG and subcloned into the BglII and Age1 site of monomeric pEGFP-N1 vector.
The ER DEVD tdTomato vector was generated by PCR amplification of tdTomato [82] using the following primers: forward primer GATCACCGGTATGGTGAGCAAGGGC reverse primer GGTTCAGGGGGAGGTGT and subcloned the resulting fragment into ER-DEVD-mStrawberry [77] using the Age1 and BsrG1 sites.
The Nalp1b allele 2 split-GFP was generated by PCR by amplifying Nalp1b using the forward primer GATCGCTAGCGCCACCATGGAACAA-TCTCAG reverse primer GATCCCCGGGCACCGGTACGCGTAGA The resulting fragment was cloned into split-GFP-N1 vectors using the Nhe1 and Xma1 sites.
DNA constructs were transfected into N2a cells using Lipofectamine 2000 transfection reagent (Invitrogen) according to manufacturer instructions.

Aggregation assay and quantification
N2a cells were plated and differentiated for two days prior transfection with Htt ex1 -GFP or split-GFP constructs. Cells were imaged at various time points following transfection. Imaged were collected using fluorescence microscopy with a widefield microscope, Axiovert 200, (Carl Zeiss Microimaging Inc., Thornwood, NY) 636 oil NA 1.4 objective, 450-490 excitation/500-550 emission bandpass filter) and a Retiga 2000R camera (QImaging; Surrey, BC). The percentage of cells containing at least one or more IBs was quantified.

FRAP and FLIP
Cells were imaged in phenol red-free RPMI supplemented with 10mM HEPES and 10% FBS. Live cells were imaged on a 37uC environmentally controlled chamber of a Duoscan confocal microscope system (Carl Zeiss Microimaging) with a 636 NA 1.4 oil objective and a 489 nm 100 mW diode laser with a 500-550 nm bandpass filter for GFP. FRAP and FLIP experiments were performed by photobleaching a region of interest at full laser power of the 489 nm line and monitoring fluorescence loss or recovery over time. No photobleaching of the adjacent cells during the processes was observed. Diffusion coefficient (D) measurements were calculated as described previously [36,83].

Photoactivation
N2a cells were transfected with Htt ex1 -PA-GFP constructs for 24h and image on confocal microscope as previously described. Photoactivation experiments were performed by bleaching a small ROI within the cells with a 405 nm laser while constantly imaging cells with a 489 nm laser to detect activated state of PA-GFP.

Quantitative fluorescence microscopy
Cells were fixed with freshly diluted 3.7% formaldehyde in PBS for 15 min at RT, permeabilized with PBS with 0.1% Triton X-100. Blocking was performed in 10% fetal bovine serum in 16 PBS. Subsequently, cells were labeled with anti-monomeric polyglutamine antibody 3B5H10 (Sigma Aldrich), followed by Alexa 555-conjugated anti-rabbit IgG secondary antibodies. Cells were imaged using fluorescence microscopy with an Axiovert 200 widefield microscope (Carl Zeiss Microimaging Inc.) with a 636 oil NA 1.4 objective, 450-490 excitation/500-550 emission bandpass filter. Image analysis was performed with ImageJ (National Institutes of Health; Bethesda, MD).
An important technical note is that the high fluorescence intensity of IBs saturates the CCD detector in the GFP channel, obscures the fluorescent signal of soluble GFP construct levels, and thus limits our ability to quantitate soluble GFP construct levels in IBs containing cells. Therefore, most of our study focuses on cells prior to the appearance of IBs. Alternatively, cells containing IBs are scored in general terms of having or not having IBs, without any attempt to quantitate absolute GFP intensity levels in those cells.

Cell death
Cells were cotransfected with Htt ex1 -GFP and ER-DEVD-tdTomato constructs. Alternatively, cells were transfected with ER-DEVD-tdTomato and treated for 3 h with 5 mM staurosporine (EMD Biosciences; Gibbstown, NJ). Fluorescent images were acquired with a widefield fluorescent microscope as previously described. For the ER DEVD-tdTomato, fluorescent intensity within the nucleus and the ER were obtained with ImageJ. Caspase activity was calculated by obtaining the ratio of mean fluorescence intensity of the nucleus divided by the intensity of the ER. Increase in caspase activity in mHtt expressing cells was quantified by calculating the % of cells with Nucleus/ER ratio higher than the average value obtained for Q23 Htt ex1 -GFP.

SDS-PAGE and immunoblot
N2a cells were plated in 12 well plates, differentiated and transfected with either Htt ex1 constructs fused to GFP or split-GFP. Cells were rinse twice with PBS and cells were lysed in 50 ml of sample buffer containing 1% SDS, 0.1 M Tris, pH8.0. Lysate were run on 12% Tris-tricine gels and transferred to nitrocellulose membrane, Antibodies used included anti-GFP and anti-RFP (generous gifts from Ramanujan S. Hegde), and HRP-labelled anti-rabbit (Jackson Immunoresearch Laboratories).

Immunoprecipitations
For immunoprecipitations, N2a cells expressing Htt ex1 constructs in 6-well plates were washed twice with 16 PBS and lysed with IP buffer (1% Triton X-100, 50mM Hepes, pH 7.4, 100mM NaCl) containing EDTA-free protease inhibitor cocktail (Roche). Lysates were clarified for 10 min at maximum speed in a microcentrifuge at 4uC and incubated for 2 h at 4uC with agarose conjugated anti-GFP beads (MBL ltd.). The beads were washed four times in IP buffer, once in distilled water, eluted with SDS-PAGE sample buffer, and analyzed on 12% Tris-glycine gels, followed by blotting, staining, and development as for immunoblots.

Native gel
N2a cells were plated in 12 well plates, differentiated and transfected with either Htt ex1 constructs fused to GFP or split-GFP. Cells were rinsed twice with native buffer (20mM Hepes, 150 mM NaCl and protease inhibitor cocktail). Then cells were lysed in cold native buffer containing 0.2% triton X-100. Sample were clarified by centrifugation and loaded on a 7.5% Tris-glycine native gel. Tris-glycine buffer was used as running buffer and gels were allowed to migrate for 4 h. Proteins were subsequently transferred on nitrocellulose membrane and processed for immunoblot for anti-GFP. Figure S1 The mobility of Htt ex1 revealed by GFP photoactivation. N2a cells were transfected with Q23, 73 or 145 Htt ex1 -PA-GFP for 24 h. Photoactivation a small ROI within the cytoplasm was performed with a 405 nm laser. By 5 s postactivation, the photoactivated pool of Htt ex1 -PA-GFP has diffused throughout the entire cytoplasm. When IBs (bottom Q145 panels) were photoactivated, no significant redistribution of the mHtt ex1 -PA-GFP to the rest of the cytoplasm was observed. (TIF) Figure S2 D values (mm 2 /s) of single cells transiently transfected with Htt ex1 -GFP constructs containing 23, 73 or 145 polyQ repeats for 16 h and analyzed by FRAP in 293 and U-2 OS cells. * p,0.05 compared to Q23 Htt ex1 -GFP. (TIF) Figure S3 (A) Comparison of Htt ex1 -GFP and split-GFP fluorescence intensities. N2a cells were transfected for 24 h with Q73 mHtt ex1 -GFP or split-GFP, fixed and immunofluorescently labeled with 3B5H10 antibody. Both constructs were imaged using the same exposure time for both channels. Subsequently exposure time was increased for split-GFP to reveal the GFP signal. Bar = 20 mm. (B) Plot of D values and camera gain settings for N2a cells transiently expressing Q73 Htt ex1 -GFP or Q73 Htt ex1 -Split-GFP. Note that lower gain settings are used for brighter cells. (TIF) Figure S4 Validation of the ER-DEVD-tdTomato reporter functionality. N2a cells were transfected with ER-DEVD-tdTomato for 16 h and then treated with or without 5 mM staurosporine for 3h. The fluorescent intensity ratio of the nucleus over the ER calculated is presented in the plot. ** p,0.0001 compared to untreated cells. Bar = 20 mm. (TIF) Figure S5 Comparison of Htt ex1 -GFP and Split-GFP constructs migration on native gel. Lysates from N2a cells cotransfected with Q23, 73 or 145 Htt ex1 fused to GFP or split-GFP were run onto a native gel and process for immunoblot with anti-GFP. Htt ex1 split-GFP constructs forms higher molecular weigh complexes than those fused to intact GFP. Interestingly, when Q23s157 is coexpressed with Q23-GFP, interaction is not possible and this is reflected by its inability to form higher complexes observed when expressed with the Q23 s238. (TIF)