For all mice, data were grouped by CGG repeat length (wildtype, Low CGG, High CGG) and analyzed across days of training using repeated measures ANCOVA with percent successful reaches as the dependent variable, CGG repeat group as the grouping factor, and day of training as a repeated within-subjects factor with total number of attempted reaches during each session as a covariate. All p values have been FDR adjusted per Benjamini et al. (2001)
. There was a main effect of CGG repeat length group (F(2,211)= 54.75, p(adj)
<.001), an effect for training day (F(14,211)=26.38, p(adj)
<.001), and there was a significant interaction between group and day (F(28,211)=1.69, p(adj)
=.02), suggesting that the longitudinal performance trajectory differed among CGG repeat groups. Total number of reaches per session did not significantly contribute to skilled forelimb reaching task acquisition (F(14, 211)=1.08, p(adj)
=.30), and did not differ among CGG repeat groups (F(28,211)=.94, p (adj)
=.56; wildtype mean 46 +/− 12 (SEM) reaches per session; Low CGG repeat group 53 +/− 19 reaches per session; High CGG repeat group 51 +/− 9 reaches per session).
To further characterize the significant interaction, a Tukey-HSD post hoc pairwise comparisons test demonstrated that no groups differed during days 1–6 of training (all p(adj)>.15), on day 7–15 the wildtype group showed a greater percentage of successful reaches than the CGG KI mouse groups (all p(adj)<.001). On day 8–15 the Low CGG group with 70–116 CGG repeats showed a greater percentage of successful reaches than the High CGG repeat group with 136–200 CGG repeats (all p(adj)<.01).
Based on the results of the paired comparisons, it appears that the three groups show differential time-courses for reaching asymptotic performance on the skilled forelimb reaching task (). To evaluate differential learning rates, the day of training at which the learning curve changed from linear to curvilinear was determined for each group, and used to define the day of acquisition during which significant improvement in performance had occurred. All groups showed a clear linear trend for days 1–6 of training (all p(adj)<.01). Beginning on day 7 the learning curve for the wildtype group became curvilinear (i.e., significant quadratic trend emerged; F(1,5)=12.04, p(adj)=.01), reflecting the marked increase in successful reaches beginning day 7. The Low CGG group did not show a curvilinear trend until day 8 (F(1,5)=8.25, p(adj)=.03), and the High CGG group did not show a curvilinear trend until day 9 (F(1,5)=7.13, p(adj)=.04). A one way ANOVA comparing the days when the trend became curvilinear across CGG groups revealed a main effect of CGG repeat grouping (F(2,16)=39.39, p(adj)<.001). A Tukey-HSD post hoc pairwise comparisons test revealed the wildtype group showed a curvilinear trend earlier than the high CGG repeat group (p(adj)<.001) and low CGG repeat group (p(adj)<.01). The low CGG repeat group showed a curvilinear trend earlier than the high CGG repeat group (p(adj)<.01).
CGG repeat length modulates skilled reaching task performance
A confirmatory analysis of differences in the learning curve among CGG groups was performed using a change point algorithm described by Gallistel and colleagues (2004)
. The first change point in the data (corresponding to the first alteration to the learning curve at a p<0.001 threshold) returned by the algorithm was selected for each mouse and compared across groups: the wildtype mice showed a significant change in the slope of the learning curve on day 7 (group mean 6.9 +/− .25 SEM), the Low CGG repeat group showed a change in slope on day 8 (group mean 8.1 +/− .33), and the High CGG repeat group showed a change in slope on day 9 (group mean 9.25 +/− .35), confirming the analysis using the curvilinear trend as the measure of learning rate.
To characterize any possible relationship between CGG repeat length and performance on the skilled forelimb reaching task in CGG KI animals with expanded CGG trinucleotide repeats, a Pearson's correlation coefficient was calculated between CGG repeat length and averaged performance for days 12–15 for CGG KI mice (). A negative association was observed between the CGG trinucleotide repeat length and the asymptotic level of skilled reaching performance in the CGG KI mice (wildtype mice were excluded from the analysis to focus on expanded CGG repeat lengths unique to CGG KI mice; corr = −.63, p(adj)=.03; R(adj)2 =.44).
Qualitative observations collected during the reaching task suggest the CGG KI mice’s reaching patterns differed from the wildtype mice. The wildtype mice reached with a linear trajectory toward the reward pellet, grasped the pellet, and returned the pellet to the mouth for consumption. The CGG KI mice, however, generally reached with a less precise, nonlinear trajectory, specifically using a more sweeping or arcing motion to reach for the reward pellets. Additionally, on the attempts when the CGG KI mice reached with a linear trajectory, they appeared to close the hand either too soon or too late, generally displacing the sucrose pellet. Both of these differences resulted in skilled reaching errors. Once the reward was grasped by a CGG KI mouse; however, the CGG KI mice did not show any tendency to drop the reward prior to consumption.