Huntington's disease (HD) is an autosomal dominant, neurodegenerative disorder that can be characterized by the presence of protein inclusions containing mutant huntingtin within a subset of neurons in the brain. Since their discovery, the relevance of inclusions to disease pathology has been controversial. We show using super-resolution fluorescence imaging and Förster resonance energy transfer (FRET) in live cells, that mutant huntingtin fragments can form two morphologically and conformationally distinct inclusion types. Using fluorescence recovery after photobleaching (FRAP), we demonstrate that the two huntingtin inclusion types have unique dynamic properties. The ability to form one or the other type of inclusion can be influenced by the phosphorylation state of serine residues at amino acid positions 13 and 16 within the huntingtin protein. We can define two types of inclusions: fibrillar, which are tightly packed, do not exchange protein with the soluble phase, and result from phospho-modification at serines 13 and 16 of the N17 domain, and globular, which are loosely packed, can readily exchange with the soluble phase, and are not phosphorylated in N17. We hypothesize that the protective effect of N17 phosphorylation or phospho-mimicry seen in animal models, at the level of protein inclusions with elevated huntingtin levels, is to induce a conformation of the huntingtin amino-terminus that causes fragments to form tightly packed inclusions that do not exit the insoluble phase, and hence exert less toxicity. The identification of these sub-types of huntingtin inclusions could allow for drug discovery to promote protective inclusions of mutant huntingtin protein in HD.
Huntington's disease (HD) is a progressive, neurodegenerative disease caused by a CAG trinucleotide expansion within the Htt gene that codes for a polymorphic polyglutamine tract near the amino-terminus of the 350 kDa huntingtin protein (1). Individuals having polyglutamine tracts with 4–36 repeats do not develop disease, whereas those with tracts exceeding the critical threshold of 37 glutamines develop HD pathology with an inverse correlation between age-onset and CAG expansion length (2,3). The huntingtin protein is ubiquitously expressed in every human cell, yet neurodegeneration in early HD is selectively restricted to the basal ganglia and cerebral cortex of the brain. Huntingtin is a highly conserved protein across vertebrates and is involved in a variety of cellular functions including roles in vesicular transport (4–6), transcriptional regulation (7–10) and cytoskeletal dynamics (11,12). The diverse functions of huntingtin likely stem from its ability to promote molecular interactions by behaving as a scaffold protein (13). The polyglutamine expanded mutant huntingtin protein disrupts many of these critical cellular functions. Notably, mutant huntingtin can aggregate to form inclusion bodies, where a cellular hallmark of HD is the presence of mutant huntingtin containing inclusions within neurons and glia of human patient brains (14).
The ability of mutant huntingtin fragments to form cytoplasmic and nuclear inclusions was initially described in the R6/2 transgenic HD mouse model, which expresses human huntingtin exon1 with a CAG expansion (15,16). Brain slices from the R6/2 mice revealed the presence of numerous inclusion bodies composed of mutant huntingtin fragments that stained positive for ubiquitin (16). Similar inclusions reported in human HD patient brains also contained amino-terminal fragments of huntingtin and stained positive for ubiquitin (14).
Many studies have demonstrated that full-length huntingtin can be proteolytically cleaved to produce several prominent amino-terminal fragments, the smallest being exon1 (huntingtin 1–81) (17–22). Furthermore, aberrant splicing of the huntingtin mRNA transcript can lead to the translation of the pathogenic exon1 huntingtin protein in HD mouse models and HD human fibroblasts (23). Therefore, it is hypothesized that small fragments of mutant huntingtin occur naturally in HD brains where they self-associate to form inclusions.
Protein aggregation occurs in most neurodegenerative disorders including Alzheimer's disease (AD), Parkinson's disease, amyotrophic lateral sclerosis and the polyglutamine disorders. In these diseases, aggregated protein commonly forms either amyloid or amorphous aggregates. Amyloid aggregates are highly insoluble and have a rigid β-sheet structure, whereas amorphous aggregates are unstructured (24). The classic amyloid hypothesis defines the deposition of misfolded amyloid-β protein into insoluble plaques within the brain as one of the primary pathogenic causes of neurodegeneration in AD (25,26). The amyloid hypothesis has since been revised and adapted to describe a common pathogenic mechanism for most neurodegenerative diseases (27). In HD, this hypothesis is often referred to as the toxic fragment hypothesis (17,28,29). It postulates that the proteolytic cleavage of mutant huntingtin generates toxic amino-terminal fragments that can misfold and accumulate in neuronal inclusions, leading to the neurodegeneration associated with HD. For the sake of nomenclature consistency, this manuscript will refer to aggregated mutant huntingtin in concentrated puncta as inclusion bodies.
Several pathogenic mechanisms have been proposed to describe the toxicity of mutant huntingtin inclusions in HD. Inclusions have been implicated in neuronal death by sequestering critical cellular proteins leading to their functional loss (30), physically occluding active vesicle trafficking between the nucleus and neuronal extremities (31), and by impairing ubiquitin-dependent proteolysis of misfolded proteins by the proteasome (32). Alternative hypotheses view inclusion formation as being either benign or even neuroprotective (33). Live cell imaging in neurons reveals that the presence of huntingtin exon1 fragment inclusions correlates with enhanced survival, relative to neurons expressing exon1 without inclusions (33,34). It has been proposed that the response to mutant huntingtin-induced cell stress is the formation of these inclusions, which accumulate and sequester mutant protein. These studies identify the soluble monomeric or oligomeric forms of mutant huntingtin as being the cytotoxic species. This is observed by the protective effect of the polyglutamine-binding peptide (QBP1) in cultured cells, which preferentially binds soluble mutant huntingtin (35). Using high-throughput screens, compounds have been identified that can paradoxically reduce cellular toxicity by either inhibiting (36) or promoting (37) the formation of mutant huntingtin inclusions. Studies like these have necessitated the need to revisit the toxic fragment hypothesis for HD (38).
HD is one of nine CAG trinucleotide repeat disorders that are all caused by expanded polyglutamine tract lengths within different cellular proteins (39). Despite the commonality between the polyglutamine disorders, each of these diseases typically affect only a specific subset of neurons within the brain (39). Furthermore, the polyglutamine thresholds for disease pathology in most of the CAG trinucleotide diseases differ from the 37 repeats required for HD, which suggests that the context of polyglutamine within the pathogenic protein is important. Thus, some studies have focused on the importance of the sequences flanking the polyglutamine tract in mediating the toxicity of mutant huntingtin as well as its ability to form inclusions (40,41). The N17 domain of huntingtin comprises the first 17 residues of the protein prior to the polyglutamine tract. N17 adopts an amphipathic alpha-helical structure that allows it to interact with various proteins in the cytoskeleton and to associate with membranes (42,43). Recent studies performed by our group and others have shown that introducing serine 13 and 16 mutations within N17 can influence the ability of mutant huntingtin to form inclusions, the rate of inclusion formation, and also the morphology of the huntingtin inclusions (44,45). Additionally, promoting phosphorylation at residues 13 and 16 of the N17 domain has been shown to alleviate mutant huntingtin toxicity in an animal model of HD (46). Flanking the polyglutamine tract on the carboxyl-terminus is a region with two pure proline tracts separated by a proline-rich region. This domain has also been shown to interact with a variety of proteins that directly impact the toxicity of mutant huntingtin and its ability to form inclusions (47,48). Therefore, these studies strongly implicate the importance of flanking sequences to the polyglutamine tract in modulating toxicity and inclusion formation.
Previously, our group developed a Förster resonance energy transfer (FRET) sensor to demonstrate that the polyglutamine tract of huntingtin can behave as a hinge, allowing the N17 domain to fold back onto the distal polyproline region of huntingtin (41). FRET involves the non-radiative transfer of energy between a donor and an acceptor molecule, which allows for the high spatial resolution of dynamic molecular interactions and conformational changes in live cells (12,49,50). The exon1 FRET sensor was used to measure the intramolecular interactions between N17 and the polyproline domain as an indication of the conformation of soluble huntingtin in live cells. Here, we have applied the mutant huntingtin exon1 FRET sensor to observe the organization of polyglutamine expanded huntingtin within protein inclusions, measuring the intermolecular interaction between individual huntingtin fragments. We demonstrate using FRET, fluorescence recovery after photobleaching (FRAP) and super-resolution fluorescence imaging that mutant huntingtin fragments can form two morphologically and dynamically distinct inclusion types. We also show that the morphology of inclusions can be influenced by altering the phosphorylation state of serines 13 and 16 of N17. The definition of two distinct inclusion types, and the ability to identify them could lead to new insights into the controversy of the role of protein inclusions in HD and other polyglutamine diseases.
Huntingtin fragments can form morphologically unique inclusion types
Using fluorescence microscopy imaging, three dimensional (3D) deconvolution and iso-surface rendering, we were able to identify two morphologically distinct types of inclusions formed by the overexpression of mutant huntingtin fragments in striatal neuron derived STHdhQ7/Q7 cells (Fig. 1). Image deconvolution involves capturing multiple images in the z plane and using algorithms to restore the out-of focus light to one focal point in a quantitative manner, increasing the signal-to-noise ratio and resolution. Two different mutant huntingtin (Q138) fragments were generated by fusing a fluoro-phore at the carboxyl-terminus to generate huntingtin exon1 Q138-YFP (Ex1 Q138-YFP) and huntingtin amino acids 1–171 Q138-YFP (1–171 Q138-YFP). The first type of inclusion, which we termed fibrillar, appears to lack a defined shape and is composed of mutant huntingtin fibres organized in an astral morphology (Fig. 1A). The second type, which we termed globular, tends to be spherically shaped with discrete and well-defined edges (Fig. 1B). To further distinguish the morphologies of these inclusions, super-resolution structured illumination microscopy (SR-SIM) was used, a method that improves lateral resolution to ∼100 nm by illuminating the sample with a series of excitation light patterns (51). Imaging revealed very distinct morphologies between the globular inclusions (Fig. 1C) and fibrillar inclusions (Fig. 1D), highlighting a cytoskeletal element in the globular inclusion.
To confirm that these different inclusion types were not just an artifact of the fluorescent protein fusion, we also generated huntingtin constructs with a small hemagglutinin (HA) tag and expressed them in STHdhQ7/Q7 cells. Immunofluorescence (IF) with additional antigen retrieval revealed that both huntingtin exon1 Q138-HA (Ex1 Q138-HA) and 1–171 Q138-HA fragments can form both fibrillar and globular inclusion types (Supplementary Material, Fig. S1). Thus, these morphologies were validated as fluorescent protein fusions, allowing further observations in live cells.
In addition to characterizing the inclusions by their morphology, the fibrillar type of inclusion can be distinguished from the globular type using a thioflavin-T staining assay. Thioflavin-T is commonly used to stain amyloid fibrils with a detectable β-sheet structure (52). Fibrillar inclusions were thioflavin-T positive (Fig. 2A), while globular inclusions did not show thioflavin-T specific staining (Fig. 2B). As a positive control, STHdhQ7/Q7 cells were transfected with amyloid-beta 1–42-mRFP (Aβ 1–42-mRFP), the amyloid fibril-forming cleavage product of the amyloid precursor protein in AD. Inclusions formed from the overexpression of the amyloid-β construct stained positive for thioflavin-T (Fig. 2C).
Immunofluorescence against ubiquitin was performed to test whether one or both types of inclusions consisted of misfolded protein. Both types of inclusions formed in STHdhQ7/Q7 cells transfected with Ex1 Q138-HA revealed ubiquitinated protein; however, no difference in ubiquitination between types was detected (Supplementary Material, Fig. S2).
Fluorescence lifetime imaging microscopy-FRET reveals two conformationally distinct inclusion types formed by mutant huntingtin fragments
Despite their unique phenotypes, visually distinguishing between the two types of inclusions was challenging due to the high intensity and diffraction-limited spatial resolution of these inclusions using standard microscopy. To overcome these limitations, we implemented FRET-based techniques to study these two types of inclusions at nanometer resolution in live cells. FRET is a well-established technique used to measure molecular interactions and conformational changes in live cells (50). As donor and acceptor probes for FRET, we chose a well-established FRET pair consisting of a cyan fluorescent protein variant, mCerulean (mCer), and an enhanced yellow fluorescent protein (eYFP) (53,54). The most accurate method of measuring FRET is using fluorescence lifetime imaging microscopy (FLIM), where fluorescence lifetime refers to the amount of time a valence electron remains in the excited state prior to returning to ground state and emitting a photon (55,56). The lifetime of a fluorophore can be directly affected by the biochemical and biophysical properties of the surrounding microenvironment; notably, FRET between two molecules causes a decrease in the donor fluorophore lifetime (55). All controls for FLIM were performed to validate the use of mCer and eYFP as a FRET pair in our live cell system (Supplementary Material, Fig. S3).
To determine if we could detect FRET changes when huntingtin fragments were organized into higher-order inclusion structures within live STHdhQ7/Q7 cells, we tested the huntingtin exon1 Q138 sensor tagged at the amino-terminus with mCer and at the carboxyl-terminus with eYFP (mCer-Ex1 Q138-eYFP). Using FLIM to measure FRET, we were able to accurately measure intra- and intermolecular interactions between individual huntingtin molecules during the nucleation and maturation of inclusions, allowing us to differentiate between the two inclusion types based on overall structure. The fibrillar type of inclusion consistently had significantly higher percent FRET efficiency values relative to the globular type of inclusion (Fig. 3A, B and D). This suggested a higher degree of interaction within or between huntingtin molecules in the fibrillar inclusions compared with the globular type. To determine whether the FRET we measured was a result of intra- or intermolecular FRET, we co-expressed mCer-Ex1 Q138 and Ex1 Q138-eYFP constructs on separate plasmids and measured fluorescence lifetime at both fibrillar (Fig. 3C) and globular (data not shown) inclusions. Since the lifetimes at fibrillar and globular inclusions were similar to those found using the FRET sensor, we concluded that the majority of the FRET measured at each type of inclusion was a result of intermolecular FRET between huntingtin fragments.
In order to acquire higher spatial resolution of FRET values within each inclusion type, we collected z-stacks and performed deconvolution on both fibrillar and globular inclusions formed by the huntingtin exon1 FRET sensor. FRET efficiency was calculated using sensitized emission FRET (seFRET) as an alternative to FLIM-FRET. seFRET is a technique in which the excitation of the donor leads to the non-radiative transfer of energy between the probes, causing an increase in fluorescence intensity of the acceptor (sensitized emission) that can be quantified to generate ratiometric FRET images (57). Consistent with the data collected using FLIM, we observed that the FRET within fibrillar inclusions was dramatically higher than that in the globular type of inclusions (Fig. 3E and F). Additionally, we noted a heterogeneous distribution of FRET efficiency values within the fibrillar inclusions; ranging from high FRET in the centre to progressively lower FRET towards the edges of the inclusion (Fig. 3F). This suggested tighter, more densely packed huntingtin molecules at the core of fibrillar inclusions relative to the edges, and overall more loosely packed protein in globular inclusions (Fig. 3E), throughout the entire volume of the inclusion.
FRAP demonstrates that the fibrillar and globular inclusion types have distinct exchange dynamics
Next, we used FRAP to gain insight into the recruitment dynamics of mutant huntingtin entering both fibrillar and globular inclusions within live STHdhQ7/Q7 cells. FRAP uses photolysis to permanently destroy the fluorophore of a fluorescent protein with high-intensity light. Any signal seen in a region of interest (ROI) over time is the result of unbleached molecules entering this space, thus giving a read-out of protein dynamics (58,59). We precisely photobleached both types of inclusions and measured the recovery of fluorescence to these ROI over time. The fluorescence recovery to the ROI in this assay represented the recruitment of soluble polyglutamine expanded huntingtin into inclusions. We observed that the fluorescence recovery within the globular type of inclusion occurred rapidly relative to the fibrillar inclusion type (Fig. 4A and B). The recovery of fluorescence to each inclusion type was temporally quantified for both Ex1 Q138-YFP and 1–171 Q138-YFP. The globular inclusions recovered significantly faster than the fibrillar types for both constructs (Fig. 4C and D).
To test whether inclusion formation could recruit and sequester soluble huntingtin protein, we monitored inclusion formation temporally within STHdhQ7/Q7 cells over 24 h periods using an environmentally controlled microscope system (33°C, 5% CO2). We observed that the formation of the globular inclusions did not affect the fluorescence of the soluble mutant huntingtin within the cells (Fig. 4E). Conversely, the formation of fibrillar inclusions progressively caused all the soluble mutant huntingtin in the cell to be absorbed and sequestered to the inclusion (Fig. 4F). We quantified the recruitment of soluble huntingtin into inclusions by temporally measuring the loss of fluorescence from a specific ROI in the cytoplasm of cells forming inclusions following expression of either Ex1 Q138-YFP (Fig. 4G) or 1–171 Q138-YFP (Fig. 4H). We measured significantly more fluorescence loss in cells forming fibrillar compared with globular inclusions at multiple time points (Fig. 4G and H).
Mutant huntingtin is actively recruited to globular inclusions by microtubules
In order to investigate whether mutant huntingtin inclusions localized to components of the cytoskeleton, IF was performed against β-tubulin, actin and vimentin. Globular inclusions were found to localize along microtubule filaments (Fig. 5A) whereas fibrillar inclusions did not (Fig. 5B). To test if mutant huntingtin was being actively recruited to either inclusion type along cytoskeletal structures, we treated cells expressing the mutant exon1 fragment with compounds that would inhibit microtubule or actin polymerization. We noted that live cells expressing mutant exon1 for ∼10 h treated with low concentrations of nocodazole, a potent inhibitor of microtubule polymerization, greatly reduced the formation of large globular inclusions and caused many smaller inclusions to be dispersed throughout the cytoplasm after ∼3 h of treatment (Fig. 5C). These results complement our FRAP data which demonstrated that fluorescence within the globular inclusions recovered rapidly as if recruitment of mutant huntingtin was driven via an active process. This was further demonstrated by temporal observation of the effect of nocodazole on fully formed globular inclusions (Fig. 5E) (supplemental Video 1). Cells expressing mutant exon1 for 24 h revealed that large globular inclusions could break up into smaller inclusions due to the disruption of the active recruitment of mutant huntingtin. Notably, treatment with the highest concentrations of nocodazole had no effect on the formation of fibrillar inclusions within the cell (Fig. 5D). Treatment of live cells with phalloidin, a potent inhibitor of actin polymerization, had no effect on the formation or size of either fibrillar or globular inclusion types (data not shown). Therefore, these data support the hypothesis that mutant huntingtin is being shuttled into globular inclusions via molecular motors on microtubules, whereas mutant huntingtin is being recruited to fibrillar inclusions by passive diffusion.
Temporal FRET reveals distinct formation and maturation dynamics of fibrillar and globular inclusions
To gain further insight into the structure of both inclusion types during their formation and maturation in live cells, we used the huntingtin exon1 sensor to generate temporal FRET videos using the seFRET technique. For these temporal experiments, we substituted mCerulean for mTurquoise2 (mTq2) due to its increased brightness and photostability over other cyan fluorescent protein variants (60). seFRET is an appropriate method in this context because the donor and acceptor fluorophores have approximately the same quantum yield, and expression levels are identical due to the fixed 1:1 ratio of the sensor. This method allows rapid and continuous temporal measurements of FRET. The FRET images were pseudocoloured with a lookup table (LUT) where FRET efficiency values correspond to the color ramp presented below the images. Following expression of mTq2-Ex1 Q138-eYFP in STHdhQ7/Q7 cells for ∼24 h, we measured the changes in FRET efficiency at 2 min intervals for 60 min. The formation of fibrillar inclusions caused a dramatic increase in relative FRET efficiency over the period of observation (Fig. 6A and B). These high FRET values measured in fibrillar inclusions using seFRET were consistent with the relative values measured using FLIM, and thus validated the use of seFRET in our system with the huntingtin sensor. Conversely, the formation and maturation of globular inclusions did not cause any significant increase in FRET efficiency over time, despite having comparable fluorescence intensities to the fibrillar inclusions (Fig. 6C and D). As a control to normalize FRET values between inclusion types, we captured a cell that formed both fibrillar and globular inclusions (Fig. 6E and F, white arrows F, G). As seen in cells forming only one type inclusion, fibrillar formation caused a drastic increase in FRET efficiency whereas globular formation caused little to no FRET changes above background values (Fig. 6E and F).
Full-length, endogenous huntingtin is actively recruited to globular inclusions and sequestered within fibrillar inclusions
In order to validate the physiological relevance of studying huntingtin exon1 inclusions, we wanted to test whether endogenous huntingtin could be recruited and sequestered to sites of mutant huntingtin aggregation. Most standard methods of cell fixation for IF fall within two categories: the crosslinking (aldehydes) and the denaturing (alcohol) fixatives. Therefore, we tried both fixation methods with a variety of permeabilization techniques to determine which would be optimal to assay the presence of endogenous huntingtin within exon1 inclusions. To label endogenous huntingtin, we chose validated huntingtin monoclonal antibodies 1HU4C8 (MAB2166) which recognizes an epitope between amino acids 181–810 of huntingtin and HDC8A4 which recognizes amino acids 2703–2911. These antibodies were generated to huntingtin epitopes downstream of exon1 (amino acids 1–81) and therefore do not recognize the overexpressed Ex1 Q138-YFP. To control for spectral bleed through between channels due to the high fluorescence intensity of the inclusions, all primary antibodies were indirectly labelled with a Cy5 conjugated secondary antibody that is spectrally distinct from YFP. Fixation with paraformaldehyde (PFA) followed by antigen retrieval using formic acid and permeabilization with a detergent allowed us to detect full length, endogenous huntingtin at fibrillar (Fig. 7A and C) but not globular (Fig. 7B and D) inclusions using both 1HU4C8 (Fig. 7A and B) and HDC8A4 (Fig. 7C and D) antibodies. Alternatively, fixation and permeabilization with methanol, which denatures cellular proteins by disrupting hydrophobic interactions, caused loss of the fluorescence at the globular but not the fibrillar type of inclusion (Supplementary Material Fig. S4a). Notably, 1HU4C8 (Supplementary Material, Fig. S4b), MAB2168 (Supplementary Material, Fig. S4c) and HDC8A4 (Supplementary Material, Fig. S4d) antibodies detected full-length huntingtin within the fibrillar inclusions under these conditions. Despite attempting every permutation of fixative and permeabilization agents to perform IF, we were never able to identify increased endogenous huntingtin at globular inclusions. This result was consistent with the fragment FRET and FRAP data, where soluble huntingtin had a low residence time in globular inclusions.
N17 phospho-mutants influence huntingtin inclusion morphology
Phospho-modifications of serine residues 13 and 16 within N17 have been shown to affect the localization and toxicity of huntingtin within the cell (42,44,45). Previous work by others with synthetic huntingtin 1–50 peptides have shown that phospho-mimicry at serines 13 and 16 alters inclusion morphology in vitro (45). In order to assay the effects of phospho-mutations on the type of inclusions formed in live cells, we generated mutant huntingtin Ex1 Q142-YFP constructs with serines 13 and 16 mutated to glutamic acids (S13ES16E) to mimic phosphorylation or to alanines (S13AS16A) to render N17 resistant to phosphorylation. A polyglutamine repeat length of 142 was considered to be similar to the 138 polyglutamine length, so Ex1 Q138-YFP without serine mutations was used as a control. When overexpressed in STHdhQ7/Q7 cells, the huntingtin phospho-mimetic (S13ES16E) mutant predominantly formed the fibrillar type of inclusion (Fig. 8A). Conversely, the alanine mutation skewed the inclusion population towards the globular type of inclusion (Fig. 8A). As expected, the polyglutamine expanded N17 wild type constructs produced a mixed phenotype of both fibrillar and globular types of inclusions, as residues S13 and S16 can exist in both phospho-states (Fig. 8A). We also observed the effect on inclusion morphology of fusing YFP to the amino-terminus of the exon1 mutants (Fig. 8B). Consistent with the carboxyl-terminus tagged constructs, phospho-mimetic mutants formed more fibrillar and phospho-resistant mutants formed more globular inclusions (Fig. 8B). This suggested that the location of the fluorescent tag on huntingtin fragments can alter inclusion formation properties, but does not affect the type of inclusion formed by altering the phosphorylation status.
Kinase inhibitors influence huntingtin inclusion morphology
In order to study the effects of true phospho-modulation on the type of inclusion formed by amino-terminal huntingtin fragments, we used kinase inhibitors known to either inhibit or promote phosphorylation of huntingtin at serine residues 13 and 16 of N17. Phosphorylation at N17 can be inhibited by casein kinase 2 (CK2) inhibitors DMAT and quinalizarin (44). Conversely, N17 phosphorylation can be promoted by treatments with IKK inhibitors, the ATP analog Bay 11-7082, and the allosteric inhibitors BMS-345541 and IMD-0354 (44). STHdhQ7/Q7 cells overexpressing Ex1 Q138-YFP treated with CK2 inhibitors had little effect on the type of inclusion formed in the cell, whereas the IKK inhibitors skewed the cellular inclusion population towards the fibrillar type (Fig. 8C). These results suggest that the fate of inclusions that form in the cell can be influenced by kinase inhibitors or other small molecules that alter the phosphorylation state of serines 13 and 16 of N17.
To further validate the effect of small molecules on inclusion type, we treated STHdhQ7/Q7 cells expressing Ex1 Q138-YFP with either CK2 or IKK beta inhibitors and looked at the inclusion migration under non-denaturing conditions using polyacrylamide gel electrophoresis (PAGE). Native PAGE maintains the folded structure/conformation and the hydrodynamic size of proteins, thus mobility varies with changes in the biophysical properties of the huntingtin inclusions. Consistent with data that IKK inhibitors skew the cellular inclusion population towards the fibrillar type of inclusions, we noted that these inclusions migrated further since they are more tightly packed (Fig. 8D). Conversely, CK2 inhibitor treatment of cells had little effect on inclusion type (Fig. 8C) and also did not alter the migration of inclusions relative to untreated using native PAGE (Fig. 8D).
The concept that mutant huntingtin can form multiple types of inclusions has previously been proposed by others, based on observations with small huntingtin fragments in cell culture and synthetic protein in vitro at super-physiological concentrations (40,45,61). Fibrillar and heterogeneous morphology inclusions have also been noted in HD brains (14). The formation of amorphous or amyloid aggregates is initiated by the presence of abnormally folded protein. Past studies have described a pathway for the formation of amyloid aggregates where abnormally folded monomers initiate the formation of oligomeric intermediates. These intermediates have been described to be globular in shape and lead to the formation of fibrillar aggregates (62). The globular inclusions described in this study were not the same as the intermediates since we never observed the direct conversion of globular to fibrillar inclusions. These two types of inclusions were therefore identified as two distinct terminal forms of aggregated mutant huntingtin. We also concluded that these inclusions were not aggresomes, a type of cytoplasmic aggregate that forms at the centrosome, since neither inclusion type was found to localize to the centrosome or cause the characteristic redistribution of vimentin to the periphery of the inclusions (63).
This manuscript characterizes different inclusion types based on a number of biophysical properties in live cells. A thioflavin-T fluorescent staining assay shows that the fibrillar type of inclusion has a detectable amyloid fibril structure and that the globular type does not. Notably, both types of inclusions were positive for ubiquitin, a characteristic of misfolded protein. Using biophotonic techniques including FLIM-FRET, seFRET, FRAP, deconvolution and temporal imaging, we observed that the fibrillar and globular inclusion types formed by polyglutamine expanded huntingtin have distinct morphological, structural and dynamic properties in live cells. We consistently measured increased FRET efficiency of the huntingtin sensor in fibrillar inclusions relative to the globular type, regardless of the intensity and size of the inclusion formed. Notably, intermolecular FRET between multiple huntingtin molecules represented the majority of the FRET measured at both inclusion types, with a small contribution of intramolecular FRET between the N17 and polyproline domains. Therefore, the higher FRET values we observed within fibrillar inclusions represented a tighter, more compact and structured organization of huntingtin fragments, as compared with the globular inclusions.
Using FRAP, we demonstrated that globular and fibrillar inclusions formed by mutant huntingtin fragments have very distinct recruitment dynamics. FRAP studies have been performed on multiple polyglutamine protein inclusions to show heterogeneity in the dynamics of inclusions from different diseases. This implies that the context of polyglutamine is important and regions flanking the polyglutamine tracts may contribute to different dynamic properties in different polyglutamine disease proteins (64). Here, we describe that inclusion heterogeneity can exist within one polyglutamine disease protein, mediated by post-translational modifications of a flanking region. We noted that following photobleaching of inclusions, the soluble huntingtin in the cell was recruited back into globular inclusions significantly more rapidly relative to the fibrillar type of inclusions. We demonstrated that soluble huntingtin was actively being shuttled into globular inclusions along microtubules. Treatment of cells expressing mutant exon1 fragments with low concentrations of nocodazole prevented the formation of large globular inclusions and caused a redistribution of smaller inclusions throughout the cell. These results are consistent with previous research that has shown that huntingtin plays a critical role in cytoskeletal dynamics and interacts directly with microtubules (65).
Temporal experiments done with an incubated microscope showed that fibrillar inclusion formation caused all the soluble mutant huntingtin in the cell to be progressively recruited and sequestered within the inclusion. These data suggested that the huntingtin within fibrillar inclusions remained relatively static and was not being dynamically exchanged between the soluble and insoluble phases. Conversely, when the globular type formed, these inclusions quickly grew and maintained their size without affecting the fluorescence of the soluble huntingtin within the cells. This suggested that there was a constant dynamic exchange between the soluble and insoluble phases in these inclusions.
Using site directed mutagenesis, we demonstrated that S13ES16E or S13AS16A mutants in the context of polyglutamine-expanded exon1 were enough to push the cellular population of inclusions towards the fibrillar or globular type, respectively. This observation is consistent with work done by others using purified synthetic huntingtin 1–56 Q37 constructs with either phospho-mimetic or alanine mutations at serine residues 13 and 16 expressed in vitro (45). Using electron microscopy, they discovered that huntingtin peptides with S13 and S16 mutations can associate to form morphologically different inclusion types (45). Despite this, our data showed that constitutive phospho-mimicry or alanine mutations at these sites did not skew the population of inclusions entirely to one form or the other, suggesting that other factors can influence inclusion fate within the cell. We hypothesize that these auxiliary factors could be other post-translational modifications or molecular interactions with either N17 or the polyproline domain. A recent study by our group has shown that N17 phosphorylation can affect soluble huntingtin conformation (41), and we hypothesize that these conformational differences can influence the nucleation and properties of an inclusion when high levels of protein are present.
Previous work has shown that mutant huntingtin is hypo-phosphorylated at serines 13 and 16 of N17 and that increasing phosphorylation at these sites can reduce the toxicity of the mutant protein (44). Others have demonstrated that promoting phosphorylation at these residues can dramatically improve motor function in an HD mouse model (46). In this study, we show that the type of inclusion formed by mutant huntingtin can be affected by small molecule kinase inhibitors that modulate the phospho-state of serine residues 13 and 16 of N17, described by us previously (44).
The distinct properties of these huntingtin inclusions suggest that one type may represent a toxic form, whereas the other may be benign or even protective to the cell. The globular inclusions were shown to have a looser packing of mutant huntingtin, which allowed for the rapid recruitment and continuous exchange of mutant protein with the soluble phase. Additionally, S13AS16A mutations skewed the population of inclusions toward the globular type. Conversely, fibrillar inclusions were shown to be densely packed and soluble mutant huntingtin was slowly recruited and sequestered within these inclusions (see model, Fig. 9). This sequestration of the mutant protein within the fibrillar inclusions could represent a cellular stress response to cope with mutant huntingtin load.
Many groups have shown that transgenic mice expressing exon1 of huntingtin rapidly develop motor and cognitive symptoms comparable to HD (15). Despite the striking phenotypes in these transgenic animals, the question of whether this small fragment occurs naturally in the brains of HD patients has been controversial. However, recent work has elegantly shown that aberrant splicing of the huntingtin pre-mRNA leads to the translation of an exon1 fragment in a variety of HD models and human HD fibroblasts (23). This work supports the hypothesis that exon1 fragments can occur naturally in abundance within neurons and other cells. However, a caveat of our studies, and those of others using small fragment over-expression, is that the protein concentrations are either super-physiological, or only relevant to neurons in late-stage HD with a massive accumulation of mutant protein. Regardless, the characterization of inclusions in late or severe HD could provide insight into huntingtin properties in early HD or even premanifest HD, which is likely the therapeutic window in this disease.
Using FRET techniques in live cells to visualize inclusions provides a valuable tool to accurately measure the unique conformational/structural differences for each type of inclusion within a cell. Additionally, FRET offers an added level of information by providing high spatial resolution, beyond even that of super-resolution microscopy. This assay is amenable to high-content screening since it provides a reliable and robust phenotypic read-out of inclusion type. This would allow for the screening of compounds that could skew the cellular population of inclusions towards one type or the other. Furthermore, since inclusion formation is a characteristic of most neurodegenerative disorders, a FRET sensor to quantify different inclusion types could be adapted to other neurodegenerative disease proteins.
MATERIALS AND METHODS
Immortalized mouse striatal STHdhQ7/Q7 cells were grown as previously described (42).
The huntingtin exon1 Q138 sensor was generated from cDNA using forward primer GATCTCCGGAATGGCGACCCTG with a BspEI restriction site and reverse primer GATCGGTACCGGGTCGGTGCAGCGGCTC with an Acc65I site. The PCR insert was then cloned into either a YFP-N1 plasmid (Clontech) or a modified mCerulean-C1 plasmid (Clontech) with an eYFP insert cloned into BamHI and XbaI sites at the opposing end of the multiple cloning site. The HA-tagged constructs were generated using synthetic oligos (MOBIX) GATCCTACCCATACGATGTTCCAGATTACGCTT with a BamHI restriction site overhang and CTAGAAGCGTAATCTGGAACATCGTATGGGTAG with an XbaI restriction site overhang. The insert was then cloned into a huntingtin exon1 Q138 or huntingtin 1–171 Q138 vector.
Transfection of STHdhQ7/Q7 cells was done using TurboFect in vitro reagent (Fermentas, R0531) as previously described (41).
The huntingtin specific mouse monoclonal 1HU 4C8 (Millipore International, MAB2166) and HDC8A4 (Pierce Antibodies, MA1-82100) antibodies were used to perform IF of endogenous huntingtin in STHdhQ7/Q7 cells.
Immunofluorescence on STHdhQ7/Q7 cells to visualize endogenous huntingtin was done either using the antigen retrieval or the methanol method. For the antigen retrieval method, cells were fixed with 4% PFA for 20 min at 4°C. Cells were then washed 2× with PBS and treated with 10% formic acid for 20 min at room temperature. Cells were then washed again 2× with PBS and subsequently permeabilized with Triton X-100 detergent (BioShop, TRX777.100) for 10 min at room temperature. Using the methanol method, cells were fixed and permeabilized using ice-cold methanol for 12 min at −20°C. Consistent between both methods, cells were then blocked 3× with 2% FBS in PBS blocking solution. Primary antibodies were added to cells at a concentration of 1:100 in a solution of 2% FBS in PBS with 0.02% Tween 20 (Sigma, P9416). Antibodies were then labelled with the far-red Cy5 (Molecular Probes, A10524) (ex 650 nm/em 670 nm) dye at a concentration of 1:500.
To visualize overexpressed huntingtin HA-tagged constructs, IF was done using the antigen retrieval method described above. Primary antibody to the HA tag (Abcam, AB16918) was added to the cells at a concentration of 1:250 in a solution of 2% FBS in PBS with 0.02% TWEEN20. Antibody was then labeled with the AlexaFluor 488 (Molecular Probes, A21206) (ex 499 nm/em 519 nm) dye at a concentration of 1:500.
For visualization of ubiquitin, IF was done using antigen retrieval method with an anti-ubiquitin antibody (Sigma, SAB4503053) added to the cells at a concentration of 1:100 in a solution of 2% FBS in PBS with 0.02% Tween 20. For visualization of β-tubulin, IF was done with 4% PFA fixation without antigen retrieval with an anti β-tubulin antibody (University of Iowa Hybridoma Bank, E7-S) at a concentration of 1:250 in 2% FBS in PBS with 0.02% Tween 20. Both anti-ubiquitin and anti β-tubulin primary antibodies were labelled with Cy5 dye at a concentration of 1:500.
Live cell temporal videos of inclusion formation were acquired with a 40× air objective using a Lumascope 500 inverted widefield epifluorescent microscope (Etaluma Inc., Carlsbad, CA, USA) housed in an incubator regulated at 33°C with 5% CO2.
Inclusion imaging was done with a 60× oil immersion objective (PlanApo NA = 1.4) on a Nikon Eclipse Ti2000 inverted widefield epifluorescent microscope using the Orca-Flash4.0 CMOS camera (Hamamatsu, Japan). Image acquisition was done using the NIS-Elements Advanced Research version 4.10.01 64-bit acquisition software from Nikon (USA). z-Stacks of inclusions were acquired using a motorized stage (Prior Scientific, USA) with step sizes of 0.3–0.5 µm. Deconvolution was done on z-stacks using the AutoQuant blind deconvolution module as part of the NIS-Elements version 4.10.01 software package (Nikon). Iso-surface rendering on inclusions was applied using the Imaris software from Bitplane (AG).
Thioflavin-T amyloid fibril staining
After 24 h of expression of Ex1 Q138-mRFP, transfected STHdhQ7/Q7 cells were fixed with 4% PFA for 30 min and stained with 0.05% thioflavin-T (Sigma, T3516) for 8 min. Cells were washed 3× with PBS and imaged. Positive control for the assay was STHdhQ7/Q7 cells transfected with Aβ 1–42-mRFP with 24 h of expression. Cells were fixed and stained as described above.
Native polyacrylamide gel electrophoresis
Cell samples were lysed using an NP-40 lysis buffer on ice for 15 min. Supernatants were collected and the remaining pellet was resuspendend in NP-40 lysis buffer and sonicated. Protein concentrations for soluble and insoluble fractions were calculated using a Bradford assay. Samples were prepared with a non-reducing loading buffer without boiling. The 7% acrylamide gels and running buffers were prepared without SDS or any other reducing agents.
Small molecule treatments
Transfected STHdhQ7/Q7 cells were treated with compounds at concentrations optimized previously (44). Cells were treated with compounds for ∼16 h prior to experiments.
Time-Domain FLIM and analysis
Sensitized emission FRET
seFRET was performed using the FRET module as part of NIS-Elements version 4.10.01. All controls were performed as required by the FRET module. Percent FRET efficiency values were calculated by the module using equations derived from the Gordon method. A rainbow look-up table was applied to the ratiometric FRET image to show the range of FRET efficiency values.
All statistical analyses were done using the SigmaPlot software 11.0 (Systat Software Inc.). For comparisons between two groups, Student's t-tests were performed if data passed the normality assumptions. If data did not pass the normality test, it was analyzed by the Mann–Whitney method. For multiple pairwise comparisons, one-way analysis of variance (ANOVA) using the Student–Newman–Keuls method was performed if the data passed the normality test of distribution. If the data did not pass the normality assumptions, then we performed a one-way ANOVA on ranks using the Tukey test. For FLIM quantifications, every cell was represented as its own N and the box-whisker plot graph was generated using cumulative data from three-independent trials.
Conflict of Interest statement. None declared.
This work was supported with operating grants to R.T. from the Huntington Society of Canada, the Canadian Institutes of Health Research (CIHR MOP-119391) and the Krembil Family foundation.