Hippocampal mitochondrial dysfunction and psychiatric-relevant behavioral deficits in spinocerebellar ataxia 1 mouse model

Spinocerebellar ataxia 1 (SCA1) is a devastating neurodegenerative disease associated with cerebellar degeneration and motor deficits. However, many patients also exhibit neuropsychiatric impairments such as depression and apathy; nevertheless, the existence of a causal link between the psychiatric symptoms and SCA1 neuropathology remains controversial. This study aimed to explore behavioral deficits in a knock-in mouse SCA1 (SCA1154Q/2Q) model and to identify the underlying neuropathology. We found that the SCA1 mice exhibit previously undescribed behavioral impairments such as increased anxiety- and depressive-like behavior and reduced prepulse inhibition and cognitive flexibility. Surprisingly, non-motor deficits characterize the early SCA1 stage in mice better than does ataxia. Moreover, the SCA1 mice exhibit significant hippocampal atrophy with decreased plasticity-related markers and markedly impaired neurogenesis. Interestingly, the hippocampal atrophy commences earlier than the cerebellar degeneration and directly reflects the individual severity of some of the behavioral deficits. Finally, mitochondrial respirometry suggests profound mitochondrial dysfunction in the hippocampus, but not in the cerebellum of the young SCA1 mice. These findings imply the essential role of hippocampal impairments, associated with profound mitochondrial dysfunction, in SCA1 behavioral deficits. Moreover, they underline the view of SCA1 as a complex neurodegenerative disease and suggest new avenues in the search for novel SCA1 therapies.

response amplitudes after the predictable and relative to the unpredictable stimuli, thus reflecting the prepulse inhibition (PPI; %), and the average time between the unpredictable stimulus and the maximum startle amplitude (startle latency; ms). Since the mice were unable to walk during the procedure, the test was not affected by abnormal mobility of the SCA1 mice. A number of individuals from the youngest cohort (1 WT and 2 SCA1) that showed persistently frozen responses with no movement were removed from the ASR analysis.
Gait: Gait characteristics were evaluated using the DigiGait device (Mouse Specifics, Inc., MA) with continuously running belt, forcing the animals to walk at a speed set by the experimenter. We used belt speeds of 12 and 18 cm/s. For all mice and both speeds, we aimed to get 5 records with > 10 steps. Animals that were not able perform the experiment (no continuous walk with ≥ 10 steps for each of the speeds) were removed from the gait data analysis (6 weeks of age: 1 WT mouse; 10 weeks of age: 1 WT and 4 SCA1 mice; 17 weeks of age: 2 SCA1 mice; 26 weeks of age: 2 SCA1 mice). Gait parameters from multiple measurements and from the left and right legs were averaged for each individual. See Suppl. Table S1 for list of evaluated gait parameters.
Rotarod: We used the RotaRod Advanced (TSE Systems GmbH, Germany) device, with rod diameter of 3.5 cm and slow acceleration from 0 to 60 RPM within 8 minutes. Mice underwent the rotarod test on 5 consecutive days with 5 measurements per day and inter-trial interval of 20 minutes.
We used day-average values of latency to fall from the rotarod (rotarod latency, s).
Morris water maze test (MWM): Mice learnt to locate the hidden platform (the center of the north-western quadrant) using intra-maze distant cues (pictures and symbols on arena wall) for 7 consecutive days. The mice underwent 4 trials per day, each with a different starting position and with 8-minute-long inter-trial intervals. Next, the animals underwent the probe trial (one session with starting position in the south; D8) with absent escape platform. Thereafter, mice underwent a test of navigation to the visually marked platform (two days D8 and D9) to verify that the potential learning deficit is not caused by a visual deficit or lack of motivation. The pool had a diameter of 1 meter, with water depth of 21 cm and the platform 0.5-1 cm under the water surface. Each session lasted a maximum of 1 minute. Unsuccessful mice were slowly put on to the platform and forced to stay there for 30 seconds. Water was made opaque by white non-toxic (food) coloring and was at a temperature of 26(±1)°C. We primarily measured latency to locate the platform averaged per day (MWM latency, s) and the proportion of time when the animal was in a non-moving state (swimming speed < 1.75 cm/s; MWM non-moving, %).
Water T-maze (WTM): the WTM test was performed as described elsewhere 3 , with a number of modifications. The mice learned to navigate to the hidden platform (0.5 cm under the surface of the water) placed in one of the two arms of the T-shaped arena (arm width: 7 cm; arm lengths: 38, 30 and 30 cm). The arena was filled with 15 cm of opaque water at 25(±1)°C. The test proceeded on 4 consecutive days (D) with 3 sessions (S) per day (2 on D1). Each session consisted of 10 trials. The time between the sessions was at least 1 hour. Prior to the experiment, the platform was removed and the mice were left to turn to one of the sides (repeated 5 times) so as to allow for the evaluation of the side preference. The platform was then placed on the non-preferred side for S1-S7 and reversed to the opposite side for S8-S11. After reaching the platform, the mice were left there for 15 seconds. We recorded the error rate manually (error = the mouse turned to the wrong arm first). The two sessions at the start of the experiment (S1-S2) and following the relocation of the platform (S8-S9) were considered training sessions, whereas the rest were considered testing sessions (S3-S7 and S10-S11).
After turning to one of the arena arms, the mice were not allowed to return to the starting arm. If the mice demonstrated immobile behavior, they were motivated to move via noise or the gentle touching/pinching of the tail, if necessary, until the mice reached the hidden platform. We assessed the overall error rate averaged across all the testing sessions (T-maze errors, %, S3-S8 and S10-S11), the error rate in the testing sessions specifically during the learning phase (T-maze learning e., %, S3-S8) and the error rate in the testing sessions following the relocation of the platform (T-maze inflexibility, %, S10-S11).
Forced swimming test (FST): The mice were placed in a cylindrical container (28 cm high, 18 cm diameter) filled with 15 cm of water (25±1 °C) and left there for 6 minutes. We recorded the duration of immobility across the whole of the test (FST immobility, %). Since the genotype-related difference in the immobility was more pronounced at the first half of the test in some age cohorts (Suppl. Fig. S1), we also evaluated immobility specifically during the 1st half of the test (FST initial immobility, %). Immobility was classified automatically according to a change in the pixels that reflected the area shape of the bodies of the mice (<5% pixel change averaged over 5 seconds).
The behavior in the EPM, OF, OLM, MWM and FST tests was automatically tracked and evaluated by means of EthoVision® XT 7.1 (Noldus Information Technology b.v., Netherlands). The preprocessed data obtained from the behavioral characterization are shown in Suppl. Data1, including those measured parameters that were not statistically evaluated.
Beside the basic characterization, 2 independent cohorts underwent FST (11 animals per group, aged 6 weeks) as described above or sucrose preference test (21 WT and 7 SCA1 mice, aged 11-13 weeks). The sucrose preference was performed as described elsewhere 4 . At first, separately housed mice were being habituated to 2 bottles in the cage (for 5 days). Then, one bottle was filled by 1% sucrose whereas the second contained water. Subsequently, fluids intake was measured daily for 4 days. Positions of the bottles were switched daily as well. The consumptions of both fluids were averaged over the 4 days and sucrose preference was expressed as relative sucrose consumption over the total fluid intake (%).

Selection of the most sensitive tests of functional impairments
Out of all evaluated indicators from behavioral and motor characterization (Suppl . Table S1), we chose those which were highly sensitive towards the SCA1 genotype in at least two consecutive age cohorts.
The criteria were defined by statistical significance of difference between WT and SCA1 mice (P < 0.01) and by Cliff's delta effect size (> 0.6 or < -0.6). If there were several strongly correlating indicators from the same test, we chose the one with the highest sensitivity and/or better interpretability.

Histology
Firstly, we performed preliminary analysis in the oldest age cohort (8 animals per group). Next, the brain regions identified as genotype-sensitive were evaluated across all age cohorts (N = 8 WT and 10 SCA1 mice or 9 animals per group in case of the youngest cohort).
To assess cerebellar volume, we measured the volume of VIII, IX and Copula pyramidalis lobules, separately for molecular and granular layers. We used every 4th slice and grid density of 200 µm. As the caudal edge of the cerebellum was sometimes damaged (cohort aged 22 weeks at the day of death and several WT mice of the two youngest cohort), we started from the posterior part where both molecular and granular layers were present and without damage. For evaluation of hippocampal subregions' volumes, we used every 8th slice for preliminary analysis and every 4th slice for follow-up analysis. Grid density was set at 200 µm in the case of Cornu ammonis (CA) subregions and molecular layer of dentate gyrus (DG) volume. A grid density of 66.7 µm was used for DG polymorph and DG granular layers' volume estimation. The sampling density that we used has already been shown to be sufficient to estimate the volume of different DG subregions in mice 5 . Parietal cortex thickness was directly measured from the 2nd to the 6th cortical layers. The 1st cortical layer was ignored, as it was often damaged. The thickness was measured by 5 evenly distributed measurements per each side and slice. The measured region was defined as the cortex above the hippocampal DG where hippocampal CA extends under DG, but does not reach the lower edge of the brain. For hypoglossal nucleus volume estimation, we used every 4th slice with grid density of 200 µm. All histological data are shown in Suppl. Data1 along with data from behavioural characterization. See Suppl. Fig. S4 for representative images.
In case of hippocampal volumes, we firstly compared absolute volumes. In the subsequent analysis, we used the brain weight as a covariate in the general linear model (supplemented by bootstrapping of partial regression coefficients) to obtain partial effect of the genotype on the hippocampal volume, adjusted for the confounding effect of the brain weight.
For evaluation of brain-behavior direct associations, we focused on the volume of the most genotype-sensitive brain regions (Cb-ML, CA-SRLM and DG-ML) and total brain weight (to investigate the association between the behavior and non-specific brain atrophy). We then evaluated the association between volumes of these brain measures and sensitive indicators from the behavioral characterization. Moreover, because cognitive flexibility was more impaired than initial learning during the water T-maze in young SCA1 mice and because the neurobiological substrate of the flexibility could substantially differ from the initial learning, we also explored the association of the brain measures with T-maze inflexibility.
NeuN staining was performed using mouse anti-NeuN primary antibodies (1/500, Millipore, MAB377) and Alexa Fluor® 488 anti-mouse secondary antibodies (Abcam, ab150117). With respect to the DCX and PSA-NCAM double staining, we used rabbit anti-DCX (1/4000, Abcam, ab18723) with mouse anti-PSA-NCAM (1/400, Bioscience, 14-9118-82) as the primary antibodies and Alexa Fluor® 594 goat anti-rabbit pre-adsorbed (Abcam, ab150084) and goat anti-mouse IgM cross-adsorbed (Thermofisher, A-21042) as the secondary antibodies. 2 or 4 brains were stained at once (i.e. in 1 of 3 blocks), with equal numbers of WT and SCA1 mice in each of the blocks. Only the brains of mice that had not previously been exposed to behavioral experiments (13-15 weeks of age; 4 males and 1 female per group) were used for immunofluorescent staining. We used every 6th slice for each animal and staining, commencing from the 12 th slice containing the hippocampus (from the frontal part); a total of 4 slices per animal. Imaging was performed using a fluorescent Olympus BX51 microscope and an Olympus DP70 digital camera (Olympus, Japan). Detailed images were acquired by means of an Olympus IX83 spinning disk confocal microscope (Olympus, Germany). The images were acquired using a standardized setting. The analysis was performed using Fiji software. The NeuN staining was analyzed by means of measuring the color intensity while subtracting the immunofluorescent intensity in neighboring sites without the presence of neuronal bodies (the CA strata oriens; Suppl. lacunosum-moleculare (CA1-SLM) layers (Suppl. Fig. S5).

Mitochondrial respiration protocol
The substrate-uncoupler-inhibitor titration (SUIT) protocol was performed by sequential addition of the sample and the following chemicals:   The oxygen concentration in the chambers was kept high enough to avoid oxygen limitation of respiration.
We evaluated respiration in steps 5-8 (states P I, P I+II, E I+II and E II; all adjusted for ROX) and complex IV capacity (the decrease of respiration between steps 10 and 11). All these parameters were adjusted for 1 mg of tissue. When respiration of one sample distinctly differed from another three samples of the same animal across at least three parameters of respiration, it was removed as outlier (a total of 4 samples for hippocampal and 2 for cerebellar data).

Measuring citrate synthase activity
The procedure is described elsewhere in detail 6 . Briefly, the assay medium consisted of 0.1 mmol/l 5,5dithio-bis-(2-nitrobenzoic) acid, 0.25% Triton-X, 0.5 mmol/l oxalacetate, 0.31 mmol/l acetyl coenzyme A, 5 µmo/l EDTA, 5 mmol/l triethanolamine hydrochloride, and 0.1 mol/l Tris-HCl, pH 8.1. 100 µl of the mixed and homogenized chamber content were added to 900 µl of the medium. Rate of absorbance change (RAC) was measured spectrophotometrically at 412 nm and 30 °C over 200 s. The RAC was used to calculate specific citrate synthase activity in a given sample (mIU/mg of tissue). As the specific citrate synthase activity showed skewed distribution, we used log-transformed values (log [mIU/mg]).

Statistical analyses
All statistical analyses were extended by permutational or bootstrapping techniques. These approaches are less sensitive to small sample size, do not rely on assumptions of parametric methods and are more robust toward outlier values 7 . Although results from fully parametric analyses are also shown in result tables, we considered the permutational (the effect of genotype) or bootstrappingbased results (models containing a continuous predictor[s]) as more relevant.
Generally, comparisons between SCA1 and WT mice were performed by permutation t-test (20 000 Monte Carlo permutations). Data with distant outliers or highly unequal variances were compared by permutation test of differences in medians (20 000 Monte Carlo permutations). All tests were two-tailed, except for testing specific hypotheses suggested by prior experiments (FST data from experiment following basic behavioral characterization). To determine effect size of the SCA genotype, Cliff´s d (as non-parametric indicator of effect size), standardized regression coefficient (ẞ) and their 95% confidence intervals (based on bias-corrected and accelerated bootstrap) were computed, using effsize 8 and boot 9 packages in R. To obtain standardized regression coefficient (ẞ), we fitted the linear model with genotype factor (i.e. presence of SCA1; 0-1) as a predictor and the given parameter as response variable. Before the model fit, both variables were scaled (zero mean and unit variance).
For serial data with > 4 timepoints, autoregressive 1 (AR1) variance-covariance structure was included in the model if it decreased the Bayesian information criterion value (measure of model parsimony; the lower BIC, the more parsimonious the model is). The analyses were performed using nlme 10 and predictmeans 11 R packages. When needed, post-hoc tests were performed by permutation t-test or paired permutation t-test (within-subject comparisons), followed by False Discovery Rate correction for multiple comparisons 12 .
When interested in the difference of general patterns rather than the difference in a single variable (e.g. analysis of gait parameters), we used the permutational multivariate analysis of variance (PERMANOVA). All evaluated parameters were scaled to have unit variance and standardized mean before the PERMANOVA execution. The multidimensional data were visualized using non-metric multidimensional analysis (NMDA). Both PERMANOVA and NMDA were performed using the vegan 13 R package.
In order to identify those behavioral measures that might be affected by signs of ataxia in the young SCA1 mice, we examined whether the sensitive indicators correlated to abnormal gait. In order to reduce the multidimensionality of the gait data, we extracted the principal components (PCs; using the vegan R package) and subjected them to principal component regression (used for Suppl. Table   S13). The PCs were based on those gait parameters that differed between the genotypes significantly in at least 1 of the young cohorts (aged ≤ 15 weeks) and the direction of the genotype-related difference was stable across both cohorts. 4 parameters were chosen: 3 describing the stride lengths and 1 representing the coefficient of variance in the stride length (hind leg, 18 cm/s). The PC1 axis correlated principally with the stride length (explaining 48% of the variance), whereas the PC2 correlated with the variance coefficient of the stride length (hind leg, 18 cm/s; explaining 27% of the variance). Only the axis with a higher effect on the given parameter is presented in Suppl. Table S13.
To visualize and compare the ability of given parameters to discriminate between WT and SCA1 mice, we visualized the ROC curve and computed the area under the ROC curve (ROC-AUC). The ROC-AUC ranges from 0.5 (i.e. the parameter does not give any clue about the genotype) to 1 (no overlap between WT and SCA1 mice). 95% CI of the ROC-AUC and the statistical difference between ROC-AUCs of given parameter vs. rotarod latency were computed using the bootstrap method. All were performed by using the pROC 14 To dissociate partial effects of several predictors on a single response variable, we used a linear model extended by bootstrapping of partial regression coefficients. If some of the covariates showed a non-linear effect, we used the generalized additive model (GAM), using the mgcv package in R.
Smoothness of non-linear effect was limited by a maximum number of 3 knots. All 95% CIs for partial effects were computed by the bias-corrected and accelerated (BCa) bootstrap method 15 (10 000 resamplings) using the boot 9 package in R. The (partial) effect was considered as statistically significant if the range of 95% CI for the standardized ẞ coefficient did not cross zero.
The standardized respiration parameters were compared by linear mixed-effect model (LME) supplemented by a permutation test. Next, to obtain standardized ẞ, its 95% confidence interval and P-value based on percentile bootstrap, we scaled the fixed-effect predictor (genotype) and standardized respiration to have zero mean and unit variance. Next, we established the LME model using these (scaled) values. We used the bootMer function from the lme4 R package 16 to obtain confidence intervals and P-values based on percentile bootstrap. See representative R code for analysis of hippocampal mitochondrial complex IV respiration: ### Fitting LME model c_IV_model<-lme(C_IV~genotype,random=~1|subject,data=hp) ### Permutation test of the LME model perm.model<-permmodels(c_IV_model,data=hp,nsim=10000) ### Scaling the variables to have zero mean and unit variance hp$genotype_numeric<-scale(as.numeric(hp$genotype)) hp$ c_IV_scaled<-scale(hp$c_IV) ### New LME model with variables scaled to zero mean and unit variance to obtain standardized ẞ and its 95 CI c_IV_model_2<-lmer(C_IV_scaled~genotype_numeric+(1|subject),data=hp) summary(c_IV_model_2) ### Bootstrapping of LME using bootMer function boot_LME<-bootMer(x=c_IV_model_2,FUN=fixef,nsim=10000) ### 95% CI boot.ci(boot_LME,type="perc",index=2) (a) Association between thigmotaxis and locomotion during the open field test. The P-value is based on the general additive model and reflects the significance of the partial effect of the distance moved adjusted for genotype. See Methods or Suppl. Table S1 for the numbers of animals in each of the experimental groups (a-d). (d) Maximum startle amplitude (left) and latency to maximum startle response following a startle sound not preceded by a prepulse (right).
(e) Relative time in the immobility state during the forced swimming test (FST), shown per each minute of the test (mean ± SEM). Mice aged 6 weeks were not exposed to any other behavioral tests (11 mice per group; P-values for the immobility = 0.2 [averaged over the 6 minutes] and 0.046 [averaged over 1st half of the experiment]).
Box-whisker plots (b-g) indicating the inter-quartile (IQ) intervals (box), 1.5*IQ range (whiskers) and medians (middle line). * P < 0.05, ** P < 0.01, *** P < 0.001. n.s. = not significant. Permutational t-test (b, d-e) or permutational test of difference in medians (c). Exact p-values are shown in Suppl. Table S1 (c-e). Day-specific latencies to find platform during the Morris water maze test. Gray area indicates phase of the test with visually marked platform. Mean ± SEM is visualized. See Methods for the numbers of animals in each of the experimental groups. Figure S3. Non-metric multidimensional scaling (NMDS) showing the similarity of the functional impairments of SCA1 mice from differing age cohorts in multidimensional space. Each point represents 1 SCA1 mouse. The closer the points, the more similar were the mice in terms of their functional impairments (sensitive indicators). The significances of the effect of the age cohort on the sensitive indicators (based on the permutational analysis of variance, PERMANOVA) were as follows: P < 0.001 when all the data were used, P = 0.65 for the dataset that included specifically mice ≤ 14 weeks of age and P = 0.21 for the dataset that included specifically mice ≥ 17 weeks of age. See Suppl. Table S12 for the detailed results. See Methods for the numbers of animals in each of the experimental groups. Figure S4. Representative images of the brain regions (Nissl-stained) that were histologically evaluated. (d) Visualization of partial effects of genotypes on DG-ML (left) or CA-SRLM (right) volume after adjustments for the effect of total brain weight (partial residuals scaled to 400 mg), derived from general linear models (Suppl. Tables S15 and S16). The visualized values could be interpreted as expected DG-ML or CA-SRLM volume if the total brain weight of the brain was 400 mg. Box-whisker plots (d) indicating the inter-quartile (IQ) intervals (box), 1.5*IQ range (whiskers) and medians (middle line). Each point = 1 animal. N = 8 (WT) and 10 (SCA1) animals (9 per group in case of the youngest cohort). * P < 0.05, ** P < 0.01, *** P < 0.001. n.s. = not significant (permutational linear models; see Suppl. Tables S15 and S16 for detailed results).  Tables   Table S1. List of indicators from characterization experiment. N = Number of successfully evaluated animals per group in given test.  = Cliff´s non-parametric measure of effect size. ẞ = standardized regression coefficient for SCA1 genotype. L,U = limits of 95% confidence interval. EPM = elevated plus maze test. OF = open field test. PPI = prepulse inhibition. MWM = Morris water maze. FST = Forced swimming test. P-values are based on permutation t-test. Gait parameters were evaluated separately for hind (H) and fore (F) legs and using 2 belt speed (12 or 18 cm/s). Rotarod latency is average over all 5 days of experiment. MWM-hidden latency is average from days 2-7 of the experiment. MWM-visible latency is average from days 8 and 9. MWM non-moving is average from days 1-7. T-maze errors is average over all testing sessions (S3-S7 and S10-S11). T-maze learning e. is average error rate from all testing sessions of learning phase (S3-S7), whereas inflexibility reflects errors during testing sessions in reversal phase (S10-S11). FST initial immobility is relative time in immobility state during first 3 minutes of FST.  Table S13. Linear models describing an effect of indicators potentially sensitive to motor deficits (rotarod latency and gait) or propensity to inactivity (distance moved in the OF and MWM non-moving) on the sensitive indicators in young (≤ 14 weeks of age) SCA1 mice. In case of gait, we extracted the principal components (PCs) and subjected them to principal component regression. The PCs were based on those gait parameters that differed between the genotypes significantly in at least 1 of the young cohorts and the direction of the genotype-related difference was stable across both cohorts. 4 parameters were chosen: 3 describing the stride lengths and 1 representing the coefficient of variance in the stride length (hind leg, 18 cm/s). The PC1 axis correlated principally with the stride length (explaining 48% of the variance), whereas the PC2 correlated with the variance coefficient of the stride length (hind leg, 18 cm/s; explaining 27% of the variance). Only the PC with a higher effect on the given parameter is presented. See Suppl. Methods for details.
d.f. = degrees of freedom. ẞ = standardized regression coefficient. CI = limits for 95% confidence intervals (CI-L: lower limit, CI-U: upper limit) based on bias-corrected and accelerated (BCa) bootstrap. P = significance based on parametric approach. boot. P = significance based on percentile bootstrap.  Table S25. Results of linear mixed-effect models describing (partial) effect of the SCA1 genotype on density of NeuN + neurons (a, b) and NeuN immunofluorescence intensity (c, d) in CA1 (a, c) and CA2/3 (b, d) hippocampal pyramidal layer. The fluorescence intensity was subtracted by fluorescence of the background (see Suppl. Methods and Suppl. Fig. S5). Because adding a grouping factor (block) improved model fit (measured by BIC) in case of NeuN signal in CA2/3, it was included into model as a covariate. N = 5 mice per group with 4 evaluated sections per mouse. ẞ = standardized partial regression coefficient. CI = limits for 95% confidence intervals of ẞ (CI-L: lower limit, CI-U: upper limit) based on percentile bootstrap. P = significance based on parametric approach. boot. P = significance based on percentile bootstrap. perm. P = significance based on permutation test of LME.  Table S26. Results of linear mixed-effect models describing the effect of the genotype on immunofluorescent markers related neurogenesis or neuroplasticity. The DCX + neuronal dendrites were quantified by means of the number of crossing lines (at least 200 µm per evaluated DG) located in three positions: i) the border between the DG granular and molecular layers (M/G); ii) the inner half of the DG-ML (closer to the granular layer; M-in) and iii) the outer half of the DG-ML (M-out). The PSA-NCAM immunofluorescent signal was measured in the DG hilus (DG-PL), DG-ML, CA4 pyramidal and CA1 stratum lacunosum-moleculare (CA1-SLM) layers (Suppl. Fig. S5). N = 5 mice per group with 4 measurement (i.e. evaluated hippocampi) per mouse. Other shortages are the same as in Suppl.  Table S28. Results of linear mixed-effect models describing partial effects of SCA1 genotype and specific citrate synthase activity (mIU/mg) on mitochondrial respiration (pmolO2/s/mg) in cerebellar (a-e) and hippocampal (f-i) tissue. Following states of mitochondrial respiration were evaluated: 1) complex I OXPHOS capacity in the ADP-activated state of oxidative phosphorylation (P I). 2) Complex I + II OXPHOS capacity (P I+II). 3) Maximum capacity for electron transport (E I+II). 4) Complex II uncoupled capacity (E II). 5) Complex IV capacity. N = 5 WT and 6 SCA1 animals (in quadruplicates).
Shortages are the same as in Suppl.