Rapid Onset of Motor Deficits in a Mouse Model of Spinocerebellar Ataxia Type 6 Precedes Late Cerebellar Degeneration1,2,3

Abstract Spinocerebellar ataxia type 6 (SCA6) is an autosomal-dominant cerebellar ataxia that has been associated with loss of cerebellar Purkinje cells. Disease onset is typically at midlife, although it can vary widely from late teens to old age in SCA6 patients. Our study focused on an SCA6 knock-in mouse model with a hyper-expanded (84X) CAG repeat expansion that displays midlife-onset motor deficits at ∼7 months old, reminiscent of midlife-onset symptoms in SCA6 patients, although a detailed phenotypic analysis of these mice has not yet been reported. Here, we characterize the onset of motor deficits in SCA684Q mice using a battery of behavioral assays to test for impairments in motor coordination, balance, and gait. We found that these mice performed normally on these assays up to and including at 6 months, but motor impairment was detected at 7 months with all motor coordination assays used, suggesting that motor deficits emerge rapidly during a narrow age window in SCA684Q mice. In contrast to what is seen in SCA6 patients, the decrease in motor coordination was observed without alterations in gait. No loss of cerebellar Purkinje cells or striatal neurons were observed at 7 months, the age at which motor deficits were first detected, but significant Purkinje cell loss was observed in 2-year-old SCA684Q mice, arguing that Purkinje cell death does not significantly contribute to the early stages of SCA6.


Introduction
Spinocerebellar ataxia type 6 (SCA6) is an autosomaldominant neurodegenerative disease that leads to pro-gressive ataxia of the limbs and gait abnormalities, and is one of the most common of the spinocerebellar ataxias (Ashizawa et al., 2013). SCA6 is caused by a CAG-repeat

Significance Statement
We confirm that disease onset in an 84Q-hyperexpanded polyglutamine mouse model of spinocerebellar ataxia type 6 (SCA6) occurs at 7 months of age, which is in agreement with a previous study by Watase et al. (2008). We characterize disease onset more precisely using a barrage of behavioral tests at multiple ages, and identify that motor coordination abnormalities emerge in a narrow time window between 6 and 7 months, in contrast with the variable age of onset observed in human patients. We find that Purkinje cell degeneration occurs in this SCA6 mouse model at 2 years, nearly 1.5 years after the onset of motor deficits, demonstrating that Purkinje cell loss is not necessary for early SCA6 disease symptoms. expansion in the gene CACNA1A encoding the ␣1Asubunit of voltage-dependent P/Q-type calcium channel, causing a polyglutamine (poly-Q) expansion (Zhuchenko et al., 1997). P/Q channels are widely expressed in the brain, including in cerebellar Purkinje cells (Westenbroek et al., 1995;Craig et al., 1998), which undergo degeneration in SCA6 (Yang et al., 2000). In patients, SCA6 symptoms typically present in midlife, with an average onset of ataxic symptoms at ϳ40-50 years of age (Matsumura et al., 1997;van de Warrenburg et al., 2002;Ashizawa et al., 2013), although disease onset has been observed across a wide range of ages, from late teens to old age (Yabe et al., 1998).
The size of the repeat expansion that gives rise to SCA6 is short compared with other triplet-repeat diseases (Gatchel and Zoghbi, 2005): unaffected individuals have Ͻ20 repeats, while pathological repeat length is 20-33 (Yabe et al., 1998;van de Warrenburg et al., 2002;Gatchel and Zoghbi, 2005). Consistent with several other triplet-repeat diseases, there is an inverse relationship between CAG repeat expansion length and age of onset in SCA6: longer repeats are correlated with earlier onset of symptoms (Matsumura et al., 1997;van de Warrenburg et al., 2002;Ashizawa et al., 2013). However, the relationship of repeat length with the age of disease onset is estimated to account for only 52% of the variance in the age of onset of SCA6 (van de Warrenburg et al., 2002), meaning that individuals who have the same repeat length can differ in the age at which they are first affected by SCA6 by decades. We wondered whether similar variability is observed in animal models of SCA6, since this may give insight into the origin of variability of disease onset in human patients.
Several mouse models have been developed for SCA6 that show a broadly similar relationship between repeat length and gene dosage on disease onset and severity as that observed for human patients, although typically shifted toward longer repeat lengths than those observed in human patients. Mice with human-length triplet repeats (SCA6 30Q ) have not been observed to develop motor deficiencies (Watase et al., 2008), while a homozygous knock-in mouse model that harbors a hyperexpanded 84-CAG repeat in the encoding region of the P/Q channel subunit (SCA6 84Q ) displays late-onset motor symptoms similar to human patients: homozygous mice show no motor abnormalities at 3 months but exhibit motor deficits at 7 months (Watase et al., 2008). Furthermore, a mouse with an even longer CAG repeat (SCA6 118Q ) displays motor impairment as early as 6 weeks old (Unno et al., 2012). While Purkinje cell loss has been reported to rapidly follow motor deficits in the SCA6 118Q transgenic mouse (detected at 10 weeks; Unno et al., 2012), no Purkinje cell loss has been reported to date in the late onset SCA6 84Q mouse (Watase et al., 2008). More recently, mice overexpressing P/Q-type calcium channel C-terminal fragments containing human-length triplet-repeat insertions have been developed that display motor phenotype (Du et al., 2013;Mark et al., 2015).
Since the variable onset of disease symptoms in SCA6 patients is only partially explained by differences in repeat length (van de Warrenburg et al., 2002), we wondered whether variability existed in disease onset in a mouse model of SCA6 as well. Since the SCA6 84Q mouse best recapitulates the midlife onset observed in human SCA6 (Watase et al., 2008), we chose to study the onset of motor coordination symptoms in more detail in the SCA6 84Q mouse in order to pinpoint the age of onset of disease symptoms more accurately. We assayed motor coordination of SCA6 84Q mice at multiple postnatal ages using several motor coordination assays including rotarod, elevated beam, and swimming. We found that motor deficits were detected simultaneously with all motor coordination assays, suggesting that there is a narrow and rapid age of onset in this SCA6 84Q mouse model, which is strikingly different from the high variability in the age of onset observed in human patients. Motor coordination deficits occurred in 7-month-old mice without any observable difference in gait or changes in Purkinje cell number or morphology, and gait abnormalities were not found even in 2-year-old mice. Although Purkinje cell degeneration was not observed at 7 months in SCA6 84Q/84Q mice, these mice have fewer Purkinje cells than wild-type (WT) mice at 2 years, arguing that although Purkinje cell death may contribute to disease progression in SCA6, it does not significantly contribute to early stages of SCA6.

Animals
Transgenic SCA6 84Q mice were purchased from The Jackson Laboratory (strain B6.129S7-Cacna1atm3Hzo/J), and heterozygous mice were bred in order to produce litter-matched male and female transgenic SCA6 84Q (homozygous SCA6 84Q/84Q , and heterozygous SCA6 84Q/ϩ ) and WT mice. At each age, behavioral assays were performed on naive animals with no prior exposure to the assays during a period of 5 consecutive days (for animal numbers at each age, see Table 1), and all data were acquired blind to genotype. Animals were moved from the housing room to the experiment room and allowed 30 min to acclimatize before beginning experiments on each day of testing. Assays were performed in same order, as follows: (1)  1-4; (3) swimming; and (4) gait was tested last on day 5 (D5) of testing for 3-to 7-month-old mice. For 1-and 2-year-old mice, gait was tested first on the first day of testing prior to the rotarod assay.

Rotarod assay
Animals were placed on a rotarod (Stoelting Europe) using a standard 10-min-long accelerating assay where the rod accelerates from 4 to 40 rpm in the first 5 min and then continues to rotate at 40 rpm for the last 5 min (Carter et al., 2001;Watase et al., 2007;Fig. 1A; Movie 1, 2). The latency to fall was recorded for each mouse as a measure of cerebellum-related motor coordination (Watase et al., 2007). Mice performed four trials (T1-T4) per day, and had at least a 15 min resting period between trials, over 5 consecutive days of testing (D1-D5).

Elevated beam assay
Animals walked along a custom-built apparatus consisting of raised round wooden beams (100 cm long), toward a dark escape box, as previously described (Carter et al., 2001; Fig. 2A; Movie 3). Bright light shining on the starting point was used as an aversive stimulus to encourage mice to traverse the beam. D1 and D2 were training days, during which mice were trained to cross a beam 22 mm in diameter. On D3 and D4, corresponding to testing D1 and D2, each mouse performed a trial on beams with diameters of 22, 18, 15, and 12 mm, totaling four trials per day. The time taken to traverse 80 cm was recorded, and the number of times the foot of the mouse slipped while crossing the beams was counted during post hoc video analysis (for an example of a mouse whose foot slips three times during the assay, see Movie 4).

Swimming assay
Animals were trained to swim across a custom-built Plexiglas swimming tank (100 cm long by 6 cm wide) toward a dry, boxed-in escape platform (Carter et al., 1999;Fig. 3A;Movie 5-7). Bright light at the starting location was used as an aversive stimulus to encourage swimming across the tank. The mice were initially trained to swim across the swim tank toward the escape platform for two trials per day. D1 and D2 were considered training days, while testing days correspond to D3ϪD5. Mice were videotaped, and latency to traverse a 60 cm distance was recorded. The number of hindlimb kicks to cross the tank was counted during post hoc video analysis. See Movie 6,7; after the assay, mice were towel dried and monitored in their home cage for 20 min after the assay.

Gait analysis
Gait was analyzed as previously described (Carter et al., 2001). The forelimbs and hindlimbs of each mouse were coated with distinct colors using nontoxic paint. Mice were prompted to walk across a white sheet affixed to an elevated platform (10 cm high by 10 cm wide) toward a custom-built dark escape box (Fig. 4A), leaving a trace of their paw prints on the sheet (Fig. 4B). Stride length (the distance between subsequent left and right forelimb and hindlimbs; Fig. 5E,F) and stance width (the distance between forelimbs and hindlimbs; Fig. 5G,H) were measured for four to six consecutive strides (measured between the centers of footprints). The coefficient of variation (CV) between stride lengths, as well as the degree of overlap between forelimb and hindlimb footprints were recorded from six consecutive strides. This assay was performed in a single trial on either D1 (at 1 and 2 years) or D5 of testing (at 4, 6, and 7 months).

Immunocytochemistry
Mice were deeply anesthetized and perfused intracardially with 4% PFA (EMS). The brain was extracted and stored in PFA at 4°C for 24 h, then transferred to PBS with 0.5% sodium azide. The cerebellar vermis was sliced into 100-m-thick parasagittal slices, and the striatum was sliced into 100-m-thick coronal slices on a Leica Vibratome 3000 plus. Staining was performed in a blocking solution consisting of 5% BSA, 0.05% sodium azide, and 0.4% Triton X in 0.01 M PBS. The primary antibodies used were rabbit anti-Calbindin D-28k (Swant) at a dilution of 1:1000, and mouse-anti NeuN (MAB377, Millipore) tagged with Alexa Fluor 488 at a dilution of 1:500, and slices were incubated with this at room temperature on a rotary shaker at 70 rpm for 72 h. Slices were then rinsed three times in a solution of 0.4% Triton X in 0.01 M PBS, and for calbindin staining, a secondary antibody (Alexa Fluor 594 anti-rabbit, Jackson ImmunoResearch) was used at a dilution of 1:1000 in blocking solution and incubated for 90 min at room temperature while shaking. Sections were then rinsed and immediately mounted onto slides with Prolong gold antifade mounting solution (Life Technologies), and stored in the dark at 4°C. Slices were imaged with a custom-built two-photon microscope with a Ti: Sapphire laser (MaiTai, SpectraPhysics) tuned to 775 nm. Image acquisition was performed using ScanImage (Pologruto et al., 2003) running in Matlab (Mathworks). Purkinje cell numbers were counted in anterior (lobule 3) or posterior (lobule 9) vermis at a density of 100 m Purkinje cells/layer, and molecular layer thickness was measured from the distance between the Purkinje cell layer at multiple evenly spaced locations in the lobule. Chemicals were purchased from Sigma unless otherwise indicated.

Data analysis and statistics
All data were analyzed blind to genotype. For each behavioral assay, mouse performance was compared among the three genotypes using one-way ANOVA; when significance was found, this was followed by Tukey's HSD post hoc test using JMP Software (SAS Institute). Purkinje cell density and molecular layer height were similarly compared with a oneway ANOVA followed by Tukey's HSD post hoc test, and imaging data were acquired and analyzed blind to condition. Striatum cell counts were compared with Student's t test.
Data are reported as the mean Ϯ SEM.

Results
Rapid onset of motor coordination abnormalities at 7 months old An accelerating rotarod assay was used to test the motor coordination of mice at several postnatal ages, ranging from 3 to 7 months (Fig. 1A). The performance of mice on In agreement with an earlier study by Watase et al. (2008), we found that SCA6 84Q/84Q mice exhibited no motor abnormalities at 3 months, but displayed significant motor deficits compared with SCA6 84Q/ϩ and WT mice at 7 months (Fig. 1B). We wondered whether the onset of motor deficits occurred between these ages, but found no differences in rotarod performance across genotypes at 4, 5, or 6 months (for sample size at each age, see Table 1; Fig. 1B), suggesting that the onset of motor abnormalities in SCA6 84Q/84Q mice does not occur earlier than 7 months, Figure 2. Increased latency on elevated beam at 7 months in SCA6 84Q/84Q mice. A, Schematic of experimental design for elevated beam assay. Two days of training were followed by 2 d of testing (D1 and D2 in B؊E). B؊E, Latency to cross the beam was measured for each genotype at each age (3, 4, 5, 6, and 7 months) over D1 and D2. SCA6 84Q/84Q mice were significantly slower at traversing the beam at 7 months on D2 for the following diameters: B, 22 mm (F (2,17) ϭ 7.36; p ϭ 0.005); C, 18 mm (F (2,17) ϭ 7.46; p ϭ 0.005); D, 15 mm (F (2,17) ϭ 4.34; p ϭ 0.03); and E, 12 mm (F (2,17) ϭ 5.27; p ϭ 0.017). SCA6 84Q/ϩ mice were indistinguishable from WT mice. ‫ء‬p Ͻ 0.05, ‫‪p‬ءء‬ Ͻ 0.01, ‫‪p‬ءءء‬ Ͻ 0.005; p Ͼ 0.05 where not indicated, one-way ANOVA followed by post hoc Tukey's test; N ϭ 8 Ϫ 10 SCA6 84Q/84Q mice depending on age, 5 Ϫ 9 SCA6 84Q/ϩ mice, and 6 Ϫ 9 WT mice (consult Table 1 for sample size at each age). and that disease onset is relatively rapid between 6 and 7 months of age. Since some studies have found that other assays are more sensitive than rotarod for detecting early motor abnormalities (Stroobants et al., 2013;Larivière et al., 2015), we chose to test motor coordination with additional assays as well.
We next conducted an elevated beam assay to gain insight into motor coordination and balance in SCA6 84Q mice ( Fig. 2A). Using beams of varying diameters, we measured the latency of the mice to cross the beam as an assay of motor coordination. Wide beams are typically easier for mice to walk across than narrow beams, and we reasoned that a range of beam sizes might capture subtle motor deficits of fine motor coordination that were not detectable with rotarod. In contrast to what we found for rotarod, older mice (6 months) performed better by crossing the beams faster than younger mice (3 months) for all genotypes tested, although this was true only for wide beams (Age: 22 mm beam, F (4,105) ϭ 4.72; p ϭ 0.003; 18 mm beam, F (4,105) ϭ 2.36; p ϭ 0.04; Fig. 2B,C), while no age dependence was observed for narrow-diameter beams (15 mm, p ϭ 0.38; 12 mm, p ϭ 0.14; Fig. 2D,E). SCA6 84Q/84Q mice took longer to traverse the elevated beam at 7 months in comparison with WT and SCA6 84Q/ϩ mice for the majority of beams used, while their performance was indistinguishable at earlier months ( Fig. 2B-E). To look in more detail at the elevated beam phenotype, we measured the number of times the hindlimb feet of each mouse slipped during beam crossings (Movies 3, 4). For most trials, the majority of mice crossed the beam without any footslips, irrespective of genotype, age, or beam width (p Ͼ 0.05 for all beams and ages excluding the 12 mm beam at 7 months; Fig. 3). However, at 7 months, the majority of SCA6 84Q/84Q mice experienced footslips when crossing the narrowest beam, and this was significantly different from WT or SCA6 84Q/ϩ mice (12 mm beam at 7 months: F (2,37) ϭ 4.19; p ϭ 0.02; Fig. 3D). Thus, the increased latency to cross most beams for 7-monthold SCA6 84Q/84Q mice likely reflects motor coordination and/or balance abnormalities. SCA6 84Q/ϩ mice showed no significant differences compared with WT mice for any age or beam diameter (p Ͼ 0.05; Figs. 2, 3). In summary, like rotarod, the elevated beam assay found motor coordination and balance abnormalities in SCA6 84Q/84Q mice at 7 months and no earlier, suggesting that the two assays are broadly similar in their ability to detect SCA6 motor abnormalities.
While rotarod and elevated beam are standard assays for motor coordination and balance deficits, we wanted to explore whether less standard motor assays might be useful to detect a motor phenotype in SCA6 mice. Swimming assays have been shown to detect subtle motor deficits at an earlier age than both rotarod and the elevated beam assay in Huntington's disease (HD) mice (Carter et al., 1999), and we wondered whether this might be similar in SCA6. We used a swimming assay to further characterize motor performance in SCA6 84Q mice (Fig.  4A). There were no obvious visual differences in the coordination of the limbs when SCA6 84Q mice swam. Unlike with rotarod and the elevated beam assay, swimming Movie 3. Elevated beam assay. A 7-month-old WT mouse crosses an elevated beam.
Movie 6. Swimming assay-SCA6 84Q/84Q mouse (side view). A 7-month-old SCA6 84Q/84Q mouse swims across the tank. Asterisks indicate right hindlimb kicks; 19 kicks were counted. tank. SCA6 84Q/ϩ and WT mice had a similar number of kicks across all ages (p Ͼ 0.05; Fig. 4C). However, SCA6 84Q/84Q mice produced a small but significant increase in hindlimb kicks at 7 months on the third day of testing (Fig. 4C,D). These results strengthen our findings from rotarod and elevated beam assays that SCA6 84Q/84Q mice have normal motor ability and coordination up until 6 months, and significant motor deficits are detected 1 month later at 7 months of age, when more hindlimb kicks are required to traverse the swim tank.
No gait abnormalities observed in SCA6 84Q/84Q mice Gait abnormalities have been reported recently for presymptomatic SCA6 patients (Rochester et al., 2014) and have also been observed in an SCA6 mouse model with an even longer 118Q expansion repeat (Unno et al., 2012). While no differences in gait were observed by eye, we examined gait in SCA6 84Q/84Q mice using footprint analysis (Fig. 5A,B) and found no differences across genotypes for stride length (Fig. 5C-F) or stance width (Fig. 5G,H), and no differences across ages 3 Ϫ 7 months for all genotypes (SCA6 84Q/84Q , SCA6 84Q/ϩ , and WT; p Ͼ 0.05 for all measurements).
To further test for possible changes in gait, we looked at the variance of stride lengths and paw overlap since stride lengths of mice are known to be very precise with minimal variation (low CV; Carter et al., 1999). We measured the CV of interstride distances from six consecutive strides to detect whether changes in this variance could be observed in SCA6 84Q mice. We found that the CV of hindlimb and forelimb strides was low and not significantly different across genotypes (hindlimb stride: Genotype, F (2,63) ϭ 0.27, p ϭ 0.77; forelimb stride: Genotype, F (2,63) ϭ 0.10, p ϭ 0.91; data not shown). We next compared the variance in paw overlap from forelimb and hindlimb paws, but found that, consistent with our other gait analyses, no significant differences in the CV of interpaw overlap was observed across genotypes (Genotype: F (2,63) ϭ 0.44; p ϭ 0.56; data not shown). Furthermore, there were no significant age-dependent changes for any genotype (p Ͼ 0.05 for each measure; Table 2). These results suggest that Movie 7. Swimming assay-WT mouse (side view). A 7-monthold WT mouse swims across the tank. Asterisks indicate right hindlimb kicks; 15 kicks were counted.  Table 1 for N values). C, In contrast with latency, there was an increase in the number of hindlimb kicks performed to cross the tank at 7 months of age in SCA6 84Q/84Q but not SCA6 84Q/ϩ compared with WT mice (Age ϫ Genotype ϫ Days: F (16,210) ϭ 1.81; p ϭ 0.03). N ϭ 8 Ϫ 10 SCA6 84Q/84Q mice depending on age, 5 Ϫ 9 SCA6 84Q/ϩ mice, and 6 Ϫ 9 WT mice (consult Table 1 for sample size at each age). D, Summary data showing the number of kicks on Day 3 at 7 months old for the different genotypes. ‫ء‬p Ͻ 0.05 one-way ANOVA followed by Tukey's post hoc test; N ϭ 6 WT mice, 5 SCA6 84Q/ϩ mice, and 9 SCA6 84Q/84Q mice.

Figure 5.
No abnormalities observed in gait in SCA6 84Q mice before or at the onset of motor coordination deficits. A, Schematic of painted footprint experiment used to study gait: forelimbs were painted blue and hindlimbs were painted red. B, Representative footprints from mice in each genotype reveal no significant differences in gait. C-H, The distance between subsequent limb placements (stride length) at 4, 6, and 7 months of age were not significantly different across phenotypes for the following: C, left hindlimb (F (2,63) ϭ 0.14; p ϭ 0.87); D, left forelimb (F (2,63) ϭ 0.25; p ϭ 0.78); E, right hindlimb (F (2,63) ϭ 0.46; p ϭ 0.64); and, F, right forelimb (F (2,63) ϭ 0.08; p ϭ 0.91). Likewise, no significant differences were observed for stance (distance between left and right limb placements) of the following: G, hindlimbs (F (2,63) ϭ 1.23; p ϭ 0.30); and H, forelimbs (F (2,63) ϭ 0.53; p ϭ 0.60) at 4, 6, or 7 months. One-way ANOVA; N ϭ 8 Ϫ 10 SCA6 84Q/84Q mice depending on age, 5 Ϫ 9 SCA6 84Q/ϩ mice, and 6 Ϫ 9 WT mice (consult Table 1 for sample size at each age).  (Fig. 1), elevated beam (Figs. 2, 3), and swimming assays (Fig. 4). We next tested gait in older SCA6 84Q/84Q mice to determine whether gait abnormalities emerged as the disease progressed, as has been observed in other mouse models (Unno et al., 2012). Using the same analyses as we performed at younger ages, we observed a difference comparing 1-and 2-year-old mice on several gait measurements across genotypes (Age: p Ͻ 0.005 for left and right forelimb and hindlimb stride, left and right stance, and CV of interpaw overlap; ANOVA followed by post hoc Tukey's test; Fig. 6), although not all measures showed significant changes (no significant difference in the CV of left or right forelimb or hindlimb strides, p Ͼ 0.05; Fig.  6). These data suggest that there is a general agerelated alteration in gait in aging mice. Surprisingly, there were no significant differences in any measure of gait in 1-or 2-year-old SCA6 84Q/84Q mice compared with age-matched WT mice (Fig. 6). This result strongly argues that gait abnormalities are not observed in SCA6 84Q/84Q mice throughout the majority of their lifespan, although they, like WT mice, experience agingrelated gait alterations.
To assess disease progression in the same older mice, we examined rotarod performance in 1-and 2-year-old mice. Motor coordination abnormalities were observed in    Figure 6. Disease progression marked by no gait abnormalities, but worsening motor coordination. A-F, Gait was examined in aging animals that were 1 and 2 years old to determine whether differences in gait emerged as SCA6 progressed. The stride lengths at 1 and 2 years were not significantly different across SCA6 84Q/84Q and WT mice for the following: 1-and 2-year-old SCA6 84Q/84Q mice on rotarod (Fig. 6), and these deficits progressively worsened compared with deficits observed at 7 months old (Age ϫ Genotype: F (2,40) ϭ 3.67; p ϭ 0.034; Figs. 1, 6). Together, our data illustrate that motor coordination deficits have a rapid midlife onset in a narrow time window in SCA6 84Q/84Q mice, and that these symptoms progressively worsen without any alterations in gait.

Late Purkinje cell degeneration in SCA6 84Q/84Q mice long after the onset of motor coordination deficits
The transgenic SCA6 84Q/84Q mice that we used in this study have previously been reported to exhibit no Purkinje cells degeneration at 20 months old (Watase et al., 2008), which is in contrast to the degeneration observed early in 118Q hyperexpanded mice (Unno et al., 2012) and postmortem in human SCA6 patient data (Yang et al., 2000). We first examined Purkinje cell density in 7-month-old mice and found 5.2 Ϯ 0.17 cells/100 m Purkinje cell layer in WT mice, with no significant differences in SCA6 84Q/84Q mice (WT mice, 496 cells measured in 9.7 mm of the Purkinje cell layer from N ϭ 3 animals; SCA6 84Q/84Q mice, 490 cells in 9.7 mm of the Purkinje cell layer from N ϭ 3 animals; Fig. 7A,B), which is consistent with previous reports (Watase et al., 2008). We wondered whether subtle changes in Purkinje cell numbers or morphology might be restricted to only part of the cerebellum, since Purkinje cell degeneration has been reported to be more prevalent in anterior lobules of cerebellar vermis in some human patients (Gierga et al., 2009;Nanri et al., 2010). To address whether changes might be localized to subregions of the cerebellar vermis, we measured Purkinje cell density in anterior and posterior lobules, but observed no significant differences in both WT and SCA6 84Q/84Q mice (data not shown). To look in more detail at Purkinje cell , while molecular layer thickness is reduced at 2 years in SCA6 84Q/84Q mice compared with WT mice (F (1,189) ϭ 33.12, p Ͻ 0.0001; right). N ϭ 3-4 animals for each genotype at each age; at least 10 mm of the Purkinje cell layer was measured for each comparison; one-way ANOVA with post hoc Tukey's test. ‫‪p‬ءءء‬ Ͻ 0.0001, ‫‪p‬ءء‬ Ͻ 0.01, ‫ء‬p Ͻ 0.05; p Ͼ 0.05, where not indicated. morphology at disease onset, we measured the height of the molecular layer (Fig. 7A) as an estimate of the height of Purkinje cell dendritic trees and found no significant differences between WT and SCA6 84Q/84Q mice at 7 months (WT mice, 291.8 Ϯ 6.8 m; SCA6 84Q/84Q mice, 301.2 Ϯ 8.1 m; Fig. 7A,D). Thus, the onset of disease symptoms in SCA6 84Q mice is not associated with alterations in Purkinje cell number or gross dendritic morphology. Since no changes in Purkinje cell numbers have been reported in 20-month-old SCA6 84Q/84Q mice (Watase et al., 2008), we looked for degeneration in older mice at 2 years. We observed no significant reduction in Purkinje cell density in 2-year-old WT mice compared with 7-month-old mice, suggesting that little degeneration has occurred at this age in WT mice (p ϭ 0.99; Fig. 7B,D). However, we observed a reduction in the density of Purkinje cells in SCA6 84Q/84Q mice at 2 years compared with 7 months (p Ͻ 0.0001; Fig. 7B,C). Consistent with this, SCA6 84Q/84Q mice had ϳ22% fewer Purkinje cells than their litter-matched WT siblings at 2 years (WT mice: 5.09 Ϯ 0.18 cells/100 m, 538 cells measured in 10.3 mm of Purkinje cell layer from N ϭ 3 mice; SCA6 84Q/84Q mice: 4.01 Ϯ 0.16 cells/100 m, 385 cells in 9.8 mm of Purkinje cell layer from N ϭ 3 mice; Fig. 7D). To address whether these changes were localized across the vermis, we compared the cell density in anterior and posterior lobules in SCA6 84Q/84Q mice, and found no significant differences at 2 years (anterior, 3.7 Ϯ 0.19 cells/100 m; posterior, 4.2 Ϯ 0.25 cells/100 m; p ϭ 0.51). Rather, we observed a reduced Purkinje cell density in both anterior and posterior lobules of SCA6 84Q/84Q mice at 2 years compared with age-and litter-matched WT mice (anterior: F (3,92) ϭ 10.16; p Ͻ 0.0001; posterior: F (3,106) ϭ 3.37; p ϭ 0.02; data not shown), in contrast to the predominantly anterior Purkinje cell degeneration observed in some human SCA6 patients (Gierga et al., 2009;Nanri et al., 2010).
Since Purkinje cell degeneration has been associated with both a reduction in cell number as well as structural changes in Purkinje cell dendrites (Yang et al., 2000), we looked at the height of the molecular layer as a readout of Purkinje cell dendritic alterations. We found that the mean molecular layer height in 2-year-old WT mice was 299.5 Ϯ 4.9 m, which was not significantly different from younger (7-month-old) WT mice (p ϭ 0.72; Fig. 7A,C,D). Together with the cell count data for 2-year-old WT mice (Fig. 7B), this suggests that very little Purkinje cell degeneration has occurred in the vermis in aged WT mice. However, the average molecular layer thickness of SCA6 84Q/84Q mice at 2 years was 257.8 Ϯ 5.3 m, a ϳ15% reduction from the molecular layer height in 7-month-old SCA6 84Q/84Q mice (p Ͻ 0.0001). We found that the molecular layer height in 2-year-old SCA6 84Q/84Q mice was significantly reduced compared with age-matched WT mice (Fig. 7C,D). Both our cell count and molecular layer data suggest that there is significant Purkinje cell degeneration by 2 years of age in SCA6 84Q/84Q mice.
Although SCA6 has been considered to be an example of a pure cerebellar ataxia (Solodkin and Gomez, 2012), noncerebellar symptoms are present in some patients, with up to 25% of affected individuals having signs of basal ganglia-related symptoms (Solodkin and Gomez, 2012). Since several recent studies have reported degeneration in the striatum of patients with other SCAs, including SCA2 and SCA3 (Schöls et al., 2015), and SCA17 (Brockmann et al., 2012), we looked at the number of striatal neurons in SCA6 84Q/84Q mice at 7 months to determine whether degeneration in the striatum was associated with the onset of motor abnormalities. We found that there were no significant differences between the density of cells in the striatum of SCA6 84Q/84Q and WT mice at 7 months (SCA6 84Q/84Q mice: 1191 Ϯ 36 cells/ mm 2 striatum, total of 3332 cells counted from N ϭ 4 mice; WT mice: 1212 Ϯ 48 cells/mm 2 , total of 2586 cells counted from N ϭ 3 mice; not significantly different, p ϭ 0.72; Fig. 8).

Discussion
We have performed an in-depth analysis of motor coordination and gait in a late-onset mouse model (84Q repeat length) of SCA6 in order to better understand the onset and progression of the SCA6 phenotype. We confirm that homozygous SCA6 84Q/84Q mice display motor coordination deficits at 7 months, and that deficits were detectable simultaneously in all motor coordination assays tested, including rotarod, elevated beam, and swimming. Prior to 7 months, the behavior of SCA6 84Q/84Q , SCA6 84Q/ϩ , and WT mice were indistinguishable, arguing that motor coordination and performance in SCA6 84Q mice are normal prior to disease onset, and that motor deficits appear rapidly between 6 and 7 months in SCA6 84Q/84Q mice. Although Saegusa et al. (2007) report that heterozygous mice expressing a human-length (28Q) repeat show enhanced motor coordination compared with WT mice, we observe no significant differences between heterozygous SCA6 84Q/ϩ and WT mice at any ages tested, which is in agreement with previous reports (Watase et al., 2008). Despite observing motor coordination deficits in 7-monthold SCA6 84Q/84Q mice, we observe no changes in any measures of gait, suggesting that this mouse model does not reproduce the progression of gait abnormalities typically observed for the human disease. Indeed, we were unable to detect changes in gait up to 2 years old, when motor coordination deficits had worsened, arguing that gait and motor coordination deficits can present independently. Finally, the onset of disease symptoms in SCA6 is not accompanied by morphological changes or the survival of cerebellar Purkinje cells, although Purkinje cell degeneration, reflecting both reductions in dendritic height and cell number, is observed nearly 1.5 years later in 2-year-old SCA6 84Q/84Q mice. The long time delay between disease onset and Purkinje cell degeneration suggests that Purkinje cell loss does not significantly contribute to early stages of SCA6 pathophysiology, and suggests that the potential for therapeutic intervention before cell death occurs might be a promising avenue of future study.

Comparison of motor assays for detection of SCA6 and motor abnormalities
We studied a range of motor coordination assays in SCA6 84Q mice as it was unknown whether one motor assay would be more sensitive than others to subtle changes in motor coordination in SCA6. With rotarod, a mouse has the opportunity to slip only one time per trial, upon which it typically falls, which means that graded differences in motor performance may be underestimated. We reasoned that other motor coordination assays, like the elevated beam in which mice can slip multiple times during the completion of the task, might provide a more nuanced readout of motor deficits, as has been observed in some animal models of ataxia (Larivière et al., 2015), although there are animal models, like the SCA3 transgenic mouse, where deficits are detected with rotarod before elevated beam (Switonski et al., 2015). We also used a swimming assay that has been shown to detect motor abnormalities in an HD transgenic mouse earlier than the elevated beam (Carter et al., 1999). Since we observe similar results with three different motor coordination assays, we argue that these assays have com-parable power to detect motor coordination deficit onset in SCA6. However, in our hands rotarod is simpler and easier to administer than the elevated beam or swimming assays, making it our preferred assay for SCA6 detection. It is possible, however, that even more sensitive assays exist to detect SCA6 motor abnormalities in rodents that we have not tested (Jarrahi et al., 2015;Vinueza Veloz et al., 2015).
We found robust motor deficits at 7 months without any concomitant changes in gait, and, indeed, gait abnormalities were not observed even in 2-year-old SCA6 84Q/84Q mice. This may at first appear surprising given that gait abnormalities are some of the first changes to be observed in SCA6 patients (Rochester et al., 2014), and gait abnormalities have been observed in an SCA6 mouse model with an even longer poly-Q expansion repeat (Unno et al., 2012). However, although motor coordination deficits and gait abnormalities often present together in mouse models Swarnkar et al., 2015), this is not always the case: there are some ataxic models where gait abnormalities and motor coordination deficits are not temporally correlated (Clark et al., 1997;Simon et al., 2004;Larivière et al., 2015), while in other transgenic mouse models, gait abnormalities have even been shown to accompany enhanced performance on motor coordination assays (Nakatani et al., 2009;Piochon et al., 2014). The absence of gait abnormalities with motor coordination changes in SCA6 84Q/84Q mice highlights a limitation of this model in faithfully recapitulating human SCA6 symptoms (Rochester et al., 2014).
In addition to the reduced motor coordination observed with each motor assay for SCA6 84Q/84Q mice at 7 months, we also observed age-dependent changes in performance across genotypes that were strikingly different for the different motor assays we used: rotarod showed decreased motor performance with age, the elevated beam assay showed increased motor performance with age for wider beams, while swimming showed no apparent agerelated differences in performance. While we and others have found decreased rotarod performance in aged WT mice (Ͼ18 months, Barreto et al., 2010; 2 years, Fig. 6D), which has been posited to arise because of neurodegeneration, we saw no significant reduction in WT Purkinje cell density at 2 years compared with 7 months (Fig. 7). In any case, degeneration cannot explain the agedependent decreases in performance seen in 7-monthold mice of all genotypes with rotarod ( Fig. 1). All three genotypes showed reduced performance with rotarod at 7 months compared with earlier performance; this suggests that there is a natural decline in performance at ϳ7 months that is exacerbated in SCA6 84Q/84Q mice. Understanding the mechanism of aging-dependent decline at 7 months in all genotypes may provide insight into disease onset in SCA6.
Another difference between rotarod and the elevated beam assay is that for a given age, performance tends to improve on successive trials per day (data not shown) as well as over experimental days with rotarod ( Fig. 1), while on the elevated beam mice tend to perform worse on the second day of training on most beam diameters (Fig. 2). In our experience, this day-on-day declining performance reflects at least in part the waning motivation of the mouse to cross the beam, which could confound the evaluation of motor coordination with this assay, while the day-onday improvement observed with rotarod may involve cerebellar learning (Ly et al., 2013). For the swimming assay, there are few day-on-day or age-dependent changes in performance, suggesting that this assay may not be best suited to measure some aspects of motor performance. For these reasons, we find that rotarod is our best assay to detect motor learning alterations in mouse models of SCA6.

Implications of SCA6 84Q Mouse Model for Human SCA6
While motor coordination abnormalities have been observed in SCA6 84Q/84Q mice at 7 months in a previous study (Watase et al., 2008), and because only a few time points were studied with a single motor assay in this previous report, the onset of disease symptoms was not well understood. Here we have characterized a rapid disease onset between 6 and 7 months that is detected with multiple behavioral assays, suggesting that disease onset is relatively strong since it can be detected by assays of varying sensitivity. This characterization of SCA6 84Q/84Q mice helps to strengthen this transgenic mouse as a model system for studying disease onset in SCA6. Our result of a narrow age of onset of disease symptoms is in contrast with observations in human patients, where individuals with a given repeat length can differ in age of onset by decades (van de Warrenburg et al., 2002). What might account for the discrepancy in disease onset variability between human patients and this transgenic mouse model? Transgenic SCA6 84Q mice are genetically homologous and live in a controlled environment, thus both the genetic and environmental diversity that exist in SCA6 patients may be absent in our study. Future enquiry is required to understand the contribution of epigenetic, environmental, and/or epistatic influences on SCA6 disease onset.
Since Purkinje cell loss is a common attribute of SCA6 (Yang et al., 2000), one of the limitations of the SCA6 84Q transgenic mouse model in the past has been the absence of reported Purkinje cell degeneration (Watase et al., 2008), unlike in other mouse models where Purkinje cell death is detected within weeks of disease onset (Unno et al., 2012). We observe Purkinje cell degeneration in 2-year-old SCA6 84Q/84Q mice, nearly 1.5 years after the onset of motor deficits, which argues that Purkinje cell death does not contribute to the early onset of motor abnormalities, and supports findings from other SCA6 mouse models suggesting that early symptoms of SCA6 arise from cellular alterations. Mechanisms that might contribute to the onset of motor dysfunction might be cellular inclusions in Purkinje cells (Watase et al., 2008;Mark et al., 2015), although Watase et al. (2008) show immunocytochemical evidence for these inclusions only at later ages (22 months), long after the onset of motor dysfunction in these mice. Other mechanisms contribut-ing to early motor dysfunction might be due to a C-terminal fragment encoded by the CACNA1A gene, which may act as a transcription factor (Du et al., 2013) and/or may have deleterious action in the cytoplasm of Purkinje cells (Mark et al., 2015). Changes in synaptic input to Purkinje cells may also be involved (Mark et al., 2015). Our demonstration of Purkinje cell degeneration, a key feature of human SCA6 (Yang et al., 2000) in aged SCA6 84Q/84Q mice supports these mice as a good model system of SCA6. From a therapeutic perspective, these results are promising, as they suggest that a window for therapeutic intervention might exist where motor function could be ameliorated by the rescue of cellular abnormalities before cell death occurs. Rescue of motor function after cell death at later stages of the disease may be more challenging and may require different therapeutic approaches.
Several recent studies on SCAs have found that neural degeneration is not limited to cerebellum. For instance, the degeneration of neurons in the striatum has been observed in SCA2 and SCA3 (Schöls et al., 2015), and SCA17 (Brockmann et al., 2012). Interestingly, significant degeneration has been observed in the striatum without motor symptoms in both HD (Cowan and Raymond, 2006), and in several SCAs (Brockmann et al., 2012;Schöls et al., 2015), which made us wonder whether changes in the striatum might be involved in the onset of motor symptoms in SCA6. However, we found no significant cell loss in the stratum of SCA6 84Q/84Q mice at 7 months, suggesting that striatal degeneration is unlikely to contribute to early stages of disease onset in SCA6.
In summary, homozygous SCA6 84Q/84Q mice display motor coordination deficits that arise rapidly between 6 and 7 months of age without gait abnormalities or Purkinje cell degeneration. Motor coordination deficits progress, and Purkinje cell degeneration is observed in 2-year-old SCA6 84Q/84Q mice, confirming that these mice display this hallmark feature of human SCA6. The temporal lag between disease onset and neuronal degeneration argues that degeneration plays a role only in later stages of SCA6. These results are important as they suggest that a wide therapeutic window may exist after SCA6 disease onset before cell death occurs.