Cytoskeleton stability is essential for the integrity of the cerebellum and its motor- and affective-related behaviors

The cerebellum plays a key role in motor tasks, but its involvement in cognition is still being considered. Although there is an association of different psychiatric and cognitive disorders with cerebellar impairments, the lack of time-course studies has hindered the understanding of the involvement of cerebellum in cognitive and non-motor functions. Such association was here studied using the Purkinje Cell Degeneration mutant mouse, a model of selective and progressive cerebellar degeneration that lacks the cytosolic carboxypeptidase 1 (CCP1). The effects of the absence of this enzyme on the cerebellum of mutant mice were analyzed both in vitro and in vivo. These analyses were carried out longitudinally (throughout both the pre-neurodegenerative and neurodegenerative stages) and different motor and non-motor tests were performed. We demonstrate that the lack of CCP1 affects microtubule dynamics and flexibility, defects that contribute to the morphological alterations of the Purkinje cells (PCs), and to progressive cerebellar breakdown. Moreover, this degeneration led not only to motor defects but also to gradual cognitive impairments, directly related to the progression of cellular damage. Our findings confirm the cerebellar implication in non-motor tasks, where the formation of the healthy, typical PCs structure is necessary for normal cognitive and affective behavior.


Results
Effect of pcd mutation on microtubule dynamics and structure. The lack of CCP1 induces an over-polyglutamylation of microtubules in the brain 24 , which may influence their interaction with different proteins including severing and motor proteins 35,36 . However, the effect on microtubule dynamics is poorly understood. To address this, we studied the microtubules of mouse embryonic fibroblasts (MEFs) lacking the CCP1 enzyme, which we named PCD microtubules (Fig. 1). Results showed ( Table 1) that "growing rate" (p = 0.022) and "catastrophe frequency" (p = 0.012) were significantly increased in the PCD microtubules ( Fig. 1A-C). Regarding microtubule curvature, PCD microtubules were more curved than the controls (p = 0.003; Fig. 1D-F). In parallel, the curvature of trajectories was also increased in PCD microtubules (p = 0.024; Fig. 1G-I). The lack of the CCP1 enzyme increased the instability of microtubules and influenced microtubule flexibility, thus affecting their trajectories.

Morphological changes in the main dendrite of PCs begin during the pre-neurodegeneration.
Microtubules are essential for maintaining neuronal shape 13,37 . Therefore, to understand the impact of these microtubular defects, we analyzed the morphology of PCs ( Fig. 2A; Table 2) during different stages of PCD postnatal development: (1) P7, where no defects have been previously described, (2) P15 and P17, to characterize the pre-neurodegenerative stage and (3) P22 and P30, to characterize the neurodegenerative stage, when PCs disappear. Results showed that morphological changes of the main dendrite length (P7, p = 0.483; P15, p = 0.393; P17, p = 0.028; P22, p = 0.043; P30, p = 0.027) and width (P7, p = 0.379; P15, p = 0.007; P17, p = 0.017; P22, p = 0.695; P30, p = 0.020) began during pre-neurodegeneration (Fig. 2B,C), both parameters decreased in PCD mice. By contrast, a reduction of the dendritic arbor length -molecular layer thickness -(P7, p = 0.361; P15, p = 0.064; P17, p = 0.704; P22, p = 0.041; P30, p = 0.030) and of the soma size (P7, p = 0.484; P15, p = 0.466; P17, p = 0.085; P22, p = 0.008; P30, p = 0.001) was observed later, during the neurodegenerative stage (Fig. 2D,E). Hence, microtubule instability may induce morphological alterations of the PCs throughout late postnatal stages of cerebellar development, before the beginning of neurodegeneration. TUNEL analysis revealed general neuronal death in the cerebellum of PCD mice. The Nna1/ Ccp1 gene mutation is known to affect all neuronal populations of the cerebellum 24,25,27 . However, apart from PC loss, the progression of cell death of other cerebellar neuronal populations of the PCD mouse has not been previously characterized. Therefore, we monitored the death of the three main neuronal populations of the cerebellum: the molecular layer cells, the PCs and the granular layer cells (Fig. 3). Results confirmed that the PCs began to die in the PCD mouse from P22 and onwards, resulting in a process of cerebellar degeneration (P7, p = 0.276; P15, p = 0.073; P17, p = 0.444; P22, p = 0.004; P30, p = 0.012; Fig. 3D. We also found an increase in cellular death during the same period in both the molecular (P7, p = 0.168; P15, p = 0.411; P17, p = 0.302; P22, p < 0.001; P30, p = 0.010; Fig. 3C) and granular layers (P7, p = 0.581; P15, p = 0.960; P17, p = 0.346; P22, p = 0.029; P30, p < 0.001; Fig. 3E). The neuronal nature of apoptotic cells in the three cerebellar layers was confirmed using TUNEL/NeuN and TUNEL/PV immunofluorescence colocalization (Fig. 3A,B). In summary, the initial symptoms of the breakdown of cerebellar structure are seen in the main dendrite of PCs. Following on, the SCIeNTIfIC REPoRTs | (2018) 8:3072 | DOI:10.1038/s41598-018-21470-2 morphological defects extend to the whole neuron, altering both the general dendritic arbor and soma. At the same time, there is a progressive and general neuronal loss in the other two layers resulting in a general breakdown of the cerebellum.  Motor coordination is only affected during the neurodegeneration. Data from the rota-rod test ( Fig. 4A,B) showed that motor performance of the PCD mouse was not affected during pre-neurodegeneration (P15, p = 0.677 and P17, p = 0.398; Fig. 4B), but impaired during neurodegeneration; i.e. at P22 (p < 0.001) and P30 (p < 0.001; Fig. 4B). Thus, motor coordination is affected only during PC loss, but not prior to this.
Home-cage behavior is altered in PCD mice before PC death. Results of the home-cage behavior test showed no differences between genotypes regarding the time spent displacing, during any of the developmental stages, which indicated that the pcd mutation did not affect general movement (Fig. 4D). Concerning grooming time, no differences were found during pre-neurodegeneration (Fig. 4E), but it did decrease during However, at P30 no differences were detected in the PCD mice regarding the percentage of time spent exploring both types of objects (p = 0.497), suggesting a deficit in object memory recognition at this age. The number of visits to both objects at P30 was further analyzed to avoid possible biases derived from motor alterations in PCD mice. Results did not show differences between the genotypes (Fig. 5E), thus discarding any effects of ataxia in PCD mice, with respect to moving within the plastic box and exploring the objects, and validating the previous result.
Social preference is affected in PCD mice because of cerebellar pre-neurodegeneration. The results of the social preference test revealed that WT mice at all ages spent a larger percentage of time exploring the chamber containing other mouse rather than the chamber containing an object (P15, p = 0.028; P17, p = 0.042; P22, p = 0.001; P30, p < 0.001; Fig. 6), indicating a preference for social contact. However, PCD mice spent the same percentage of time exploring both chambers at all ages (P15, p = 0.305; P17, p = 0.328; P22, p = 0.272; P30, p = 0.812; Fig. 6). Thus, the pre-neurodegenerative changes observed prior to the breakdown of cerebellar structure seem to be enough to induce social deficits.
In conclusion, the behavioral analysis of PCD mice strongly suggests a close relationship between cerebellar function and cognition. Moreover, the progression of the cerebellar alterations increasingly impaired the performance of cognitive and affective tasks.

Discussion
One of the main obstacles to establishing a relationship between the cerebellum and cognitive and social processes is the lack of studies analyzing the impact of progressive cerebellar defects on non-motor behavior. Also, this association has only been studied over short time frames. Here, to allow for a better understanding of this relationship, our results will be discussed chronologically based on their appearance throughout postnatal development in normal and PCD mice. The relevant order includes the first week of postnatal development (P7), then during the pre-neurodegenerative stage (P15-P17), and finally during the neurodegenerative stage (P22-P30). Thus, a temporal pattern is presented, characterizing the microtubule alterations observed in the PCD mouse, the progression of the cerebellar defects, and the influence of these alterations on motor and non-motor behaviors.
To understand the importance of the CCP1 enzyme in cerebellar development, the effect of the lack of this enzyme on microtubule dynamics and the structure of fibroblasts was initially studied, as the in vitro model available was able to provide insight into the structure and dynamics of this cellular component. While previous results have demonstrated that the lack of CCP1 enzyme increases microtubule polyglutamylation 24 , here it is shown that the lack of this de-polyglutamylase affects microtubule stability and flexibility. Microtubules are more unstable at the growth cones, which consequently increases their catastrophe frequency, an essential process for normal neuronal development 38,39 . Additionally, certain factors related to neuronal migration and development preferentially bind to curved microtubules [40][41][42][43] . Therefore, microtubule dynamics and curvature seem to have a different pattern during neuronal development and maturation, being more dynamic and curved during development and stable and rigid during maturation. The microtubules of PCD mice are likely very similar to those found in developing neurons, in terms of dynamics and structure. In WT mice, CCP1 expression increases at P15 in the cerebellum 29 , that is, at the beginning of the pre-neurodegeneration stage of the PCD mouse 26,29 . Precisely, during this short time frame, the growth of the dendritic arbor of PCs is completed and the process of maturation and remodeling takes place 44,45 . Thus, microtubule polyglutamylation seems to be essential during cerebellar development, and a reduction of this post-translational modification seems to be necessary at the end of neuronal growth. Then, the microtubule over-polyglutamylation occurring in the PCD mouse triggers the pre-neurodegeneration and the loss of the PCs from P18 and onwards 25,26,29 . Conversely, at P7 no differences were observed in the morphology of the PCs between genotypes, the general structure of the cerebellum was apparently unaffected. Since neuronal migration and growth occur in the normal cerebellar cortex during this first postnatal week 46 , it appears that the pcd mutation (i.e., microtubule over-polyglutamylation) does not affect the initial perinatal stages of cerebellar cortex formation. CCP1 activity is extremely low during the first two postnatal weeks (it significantly increases at P15) 29 thus, our data support the idea that microtubule polyglutamylation is essential during early cerebellar development, but deleterious for establishing a mature pattern. Therefore, the fine-tuning of microtubule polyglutamylation has a major role in cerebellar postnatal development.
In order to understand the influence of the cerebellum on cognitive and social behavior, different authors have analyzed the effects of different cerebellar alterations in animal models 14,15 . However, to our knowledge, there are no studies that have attempted to analyze progressive cerebellar alterations throughout postnatal development to understand the normal cerebellar function.
Morphological defects in the PCs of different psychiatric and neurodevelopmental disorders have been recently described [47][48][49] , but the direct effect of alterations in the morphology of PCs on non-motor behavior is not known. We demonstrate that prior to the breakdown of cerebellar structure (during pre-neurodegeneration, when no motor defects were detected), morphological alterations in the main dendrite of PCs and social and environmental exploratory behavioral disabilities occur simultaneously (both at P15 and P17). In psychiatric and neurodevelopmental disorders, such as autism or schizophrenia that can often exhibit severe symptoms of social disability, have been related to cerebellar alterations [8][9][10][11][12]15 . Furthermore, PC loss is the most common anatomopathological change found in autistic patients 47 , and even in those cases where no loss of PCs was observed, morphological defects were still found 48 . Also, in accordance with our results, environmental exploratory defects have been previously associated with cerebellar alterations [50][51][52] . A possible explanation for these impairments is that alterations in dendritic morphology influence the electrophysiological activity of neurons, including the PCs 53,54 . More precisely, defects in the PCs axial resistance, which directly depends on dendritic length and width, disrupt the synaptic integration in PCs 53 , as previously hypothesized 55,56 . Hence, it is plausible to relate the progressive extinction of Purkinje cells in PCD mice to the progressive alteration of the behaviors analyzed (see below).
During the cerebellar degeneration that occurs in PCD mice, neuronal death takes place throughout the three cerebellar layers. Previous studies assumed that in the PCD animal, the loss of neuronal populations of the granular and molecular layers is residual and delayed in time, in comparison to the death of PCs 25 ; however, the progression of their death has never been characterized. Our results demonstrate that this neuronal loss progresses at the same time as the loss of PCs, indicating a simultaneous decline in all three layers in the cerebellum of the PCD mouse.
During neurodegeneration, the morphological alterations of the PCs worsen, with the most evident changes linked to the length and width of the main dendrite, the soma area and the dendritic arbor length. In parallel, the defects observed in social preference and environmental exploratory behavior also persist during this stage in the PCD mouse. Alterations in these behaviors have been described in both animal and human studies with cerebellar impairments 8-10,57-59 . It was observed that these alterations may be due both to neuronal loss and to PC malfunction. Interestingly, we saw that ataxia did not bias the results of the cognitive/social tasks performed by the PCD mice, and during neurodegeneration grooming was found to be reduced. In contrast, this behavior has been previously shown to increase in animal models of specific PC loss 15 , being more so in models presenting a general alteration of the cerebellum 58 . Accordingly, during grooming behavior there is an increase in the activity of both granule cells 60 and PCs 59 . Consequently, grooming may be affected differently depending on whether the cerebellar alterations affect the PCs, granule cells or both. Thus, the severity of cerebellar defects helps to understand the relationship between the cerebellum and grooming. Although grooming is a very complex motor task, we cannot discard the possibility that the motor defects caused by cerebellar alterations have influenced the observed behavioral changes.
Interestingly, general changes in cerebellar structure in the PCD mouse, which disrupt recognition memory, only occurred at P30 when most of the PCs are lost. The cerebellum has been associated with procedural learning and memory 61,62 , and current neuroimaging studies have also described activation of the cerebellum during non-procedural memory tasks [63][64][65] . These works, however, did not study the impact of the progression of the breakdown of cerebellar structure on recognition memory. According to our results, cerebellar alterations lead to the impairment of recognition memory in PCD mice, and during neurodegeneration both the cerebellar and behavioral defects become worse. Therefore, these results suggest a correlation between the progression of the breakdown of the cerebellar structure and the severity of the defects related to performing cognitive and affective tasks. Then, a healthy, typical cerebellar structure is necessary to execute specific cognitive and affective tasks. In conclusion, polyglutamylation affects microtubules dynamics and structure, and its modulation is essential throughout normal cerebellar development. During the first two postnatal weeks, when the PCs are still migrating and growing, microtubules instability is crucial. However, a pathologically extended instability of microtubules (triggered by the pcd mutation) affects the morphology and survival of PCs, which are vital for the correct functioning of the cerebellum. In this sense, cerebellum plays an important role in cognitive and affective functions, since the progressive breakdown of the cerebellar cortex in PCD mice is directly related to the increasing severity of cognitive and affective impairments.

Material and Methods
Animals. Both WT and PCD mice from the C57/DBA strain were used and purchased from The Jackson Laboratory, Maine, USA. For in vitro studies, MEFs were prepared from both WT and CCP1-KO embryos (see below). The embryos were separated into groups depending on their genotype (WT or PCD; n = 3 embryos per group). For in vivo studies, animals were separated into groups depending on their genotype and age at the time of analysis: P7, P15, P17, P22 and P30 (n = 4 per age and genotype) for the immunohistochemical analyses; and P15, P17, P22 and P30 (n = 8-9 per genotype) for the behavioral analyses.
Since PCD and CCP1-KO animals are not fertile, heterozygous mice were mated, and the embryos or newborn mice were genotyped (see below and 28 ).
Animals were housed, handled and sacrificed per the guidelines established by European (2010/63/UE) and national legislations (Spanish RD53/2013 and Law 32/2007; French permit n°38 07 11). Every effort was made to  (Fig. 7). The linearized construct was electroporated in 129S2/SvPas mouse embryonic stem (ES) cells. After selection, targeted clones were identified by PCR using external primers and further confirmed by Southern blot with 5′ and 3′ external probes. Two positive ES clones were injected into blastocysts, and derived male chimaeras gave germline transmission. The excision of the neomycin-resistance cassette was performed in vivo by breeding the chimeras with a Flp deleter line (C57BL/6 N genetic background FLP under ACTB promoter). The Flp transgene was segregated by breeding the first germ line mice with a wild type C57BL/6 N animal. For generation ccp1 KO, ccp1 floxed mice were crossed with transgenic mice expressing Cre recombinase under the control of a CMV promoter.
Genomic DNA isolated from mouse tail snip was analyzed by PCR. Mice were genotyped by PCR according to MCI protocols using GoTag polymerase (Promega) and 33 amplification cycles. The three primer pairs listed below were used to define the genotypes.  previous studies 68,69 . To ensure the accuracy of the analysis, we compared the growing speed data obtained both automatically and manually. The microtubule curvature analysis was performed using the first image of each video time-lapse with m-cherry α-tubulin. Additionally, with the GFP-EB3 labeling experiment we obtained the maximum projection of the entire video time-lapse, allowing us to reconstruct the trajectories of the microtubules, whose curvature was also analyzed. Thus, the curvature of microtubules was analyzed both in a static way and within their trajectory of movement.
Curvature (k) was determined using the Mathematica software (Wolfram Research Europe, UK). After obtaining the equation of each microtubule with a spline cubic interpolation, the general formula of curvature 70 was employed to analyze each microtubule (equation 1): where (equations 2, 3, 4 and 5) i.e., the first ( ′ ′ x y , ) and second ( ′′ ′ x y , ) derivatives at each point.

Tissue extraction and preparation for in vivo studies.
To analyze the effect of the pcd mutation on the PCs morphology and the general cerebellar structure in vivo, animals were anesthetized and intracardially perfused with Somogyi's fixative without glutaraldehyde (5 ml/g body weight), as previously described 26 Cerebella were sectioned sagittally at 30 µm thick using a freezing microtome (Leica Jung SM 2000, Nussloch, Germany), as previously described [71][72][73] . The histological analyses were focused on the vermis using three sections per animal (separated by 180 µm each), with a total of five animals per genotype and age. All values represent the mean count of the three sections analyzed.
Immunofluorescence. Sections were incubated with the primary antisera at 4 °C during 72 h, as previously described 74 . The antibodies employed were mouse anti-calbindin (Cb-28k 1:1,000; Swant, Switzerland), mouse anti-parvalbumin (PV 1:1,000; Swant, Switzerland) and mouse anti-NeuN (1:8,000; Merck Millipore, Darmstadt, Germany). Appropriate secondary antibodies conjugated with Cy2 or Cy3 (1:500; Jackson Laboratories, West Grove, PA, USA) were used. Sections were counter-stained with DAPI (1:30,000; Sigma Aldrich) to identify the cell nuclei. Antibodies were selected based on their selectivity to identify different neuronal populations. Calbindin is a well-known calcium-binding protein that label the whole PCs 46,75,76 , allowing us to analyze the PCs morphology. In addition, parvalbumin stains both PCs and interneurons in the molecular layer 46 . Finally, NeuN was used to counterstain whole cerebellar nuclei during cell counting, ensuring no biases were made due to the lack of immunolabeling during degeneration.

Terminal Deoxynucleotidyl Transferase-Mediated Fluorescein dUTP Nick-End Labeling (TUNEL) detection.
To study apoptosis throughout cerebellar postnatal development, the TUNEL technique was used as previously described 29,74 . To determine the neuronal nature of apoptotic cells, the TUNEL was combined with immunohistochemistry against NeuN or PV. Finally, both TUNEL and immunohistochemistry were revealed using Cy2-streptavidin and Cy3-secondary antibodies, respectively.
Morphological analysis of PCs. Morphological analysis of the PCs was made using sections immunostained for Cb-28k from P7 to P30. To avoid possible biases, only those Purkinje cells containing the soma and a clear dendritic arbor were analyzed. Also, all Purkinje cells fitting this criterion were analyzed in three cerebellar sections per animal (separated by 180 µm each). The final value for each animal used to compare was the mean of all Purkinje cells analyzed in these three sections. The morphological parameters studied were selected in basis to its meaning for PCs function 77,78 : (1) the length and width of the main dendrite; (2) the length of the dendritic arbor of the PCs (measured indirectly through the molecular layer thickness); and (3) the soma size.
Morphological analyses were carried out with the Neurolucida (MBF Bioscience, Williston, Vermont, USA) and ImageJ (NIH, USA) software, as previously described 77 . Behavioral analyses. We characterized the effects of progressive cerebellar defects on motor, cognitive, and social processes. Behavioral analyses were performed at P15, P17, P22 and P30, encompassing both the pre-neurodegenerative (P15-P18) and neurodegenerative (P18-P30) stages. All devices employed were cleaned with 96% (v/v) ethanol before and after each session.
The rota-rod test was used to characterize motor coordination. It was performed as previously described 79 , with an acceleration of 0.6 rpm/s, from 4 to 40 rpm, in 10 min.
The home-cage behavior analysis was used to characterize general behavior. Animals were placed each day in the home-cage for 10 min and allowed to freely explore. Afterwards, the animals' behavior was observed for 10 min, during which time the following parameters were recorded manually: (1) grooming time (stereotyped behavior); (2) number of rearings (environmental exploratory behavior) and (3) time moving (general movement). Each animal was assigned to an individual home-cage to avoid the influence of odor from the other animals.
The novel object recognition test was used to analyze recognition memory. The test was performed in the same home-cage used described above to help the animals to become habituated. On the first day, at P15, animals were placed in the home-cage with two identical objects (A and A) for 10 min. Then, for each of the following sessions (at P17, P22 and P30) the animals were placed in the home-cage with: (1) the familiar object (A) and (2) a new object (B, C, and D, respectively for each of the different ages; see Fig. 5). The objects were placed in opposite corners of the home-cage, the time interacting with each object was measured and their ratio was calculated.
The social preference test was performed in a white Plexiglas ® box (50 × 29 cm) divided into three chambers.
The social preference test was performed as previously described [80][81][82] . On each test day, a mouse could explore the box for 10 min. Then, the animal was removed from the box and a different mouse, of the same age and sex, was placed in one of the lateral chambers under a drilled pencil cup. On the other side of the box, an object was placed in the other lateral chamber, also under a drilled pencil cup. The original mouse was then returned to the box and the percentage of time spent exploring each lateral chamber and each pencil cup was analyzed.
Statistical analysis. For data analysis, the Student's t-Test was performed. For histological analysis, both genotypes were compared for each age. For behavioral analysis, the parameters analyzed and the different confrontation variables are described in each behavioral procedure. All analyses were performed with the SPSS statistical package software (IBM, NY, USA).