Abnormalities in synaptic dynamics during development in a mouse model of spinocerebellar ataxia type 1

Late-onset neurodegenerative diseases are characterized by neurological symptoms and progressive neuronal death. Accumulating evidence suggests that neuronal dysfunction, rather than neuronal death, causes the symptoms of neurodegenerative diseases. However, the mechanisms underlying the dysfunction that occurs prior to cell death remain unclear. To investigate the synaptic basis of this dysfunction, we employed in vivo two-photon imaging to analyse excitatory postsynaptic dendritic protrusions. We used Sca1154Q/2Q mice, an established knock-in mouse model of the polyglutamine disease spinocerebellar ataxia type 1 (SCA1), which replicates human SCA1 features including ataxia, cognitive impairment, and neuronal death. We found that Sca1154Q/2Q mice exhibited greater synaptic instability than controls, without synaptic loss, in the cerebral cortex, where obvious neuronal death is not observed, even before the onset of distinct symptoms. Interestingly, this abnormal synaptic instability was evident in Sca1154Q/2Q mice from the synaptic developmental stage, and persisted into adulthood. Expression of synaptic scaffolding proteins was also lower in Sca1154Q/2Q mice than controls before synaptic maturation. As symptoms progressed, synaptic loss became evident. These results indicate that aberrant synaptic instability, accompanied by decreased expression of scaffolding proteins during synaptic development, is a very early pathology that precedes distinct neurological symptoms and neuronal cell death in SCA1.

Scientific RepoRts | 5:16102 | DOi: 10.1038/srep16102 Sca1 154Q/2Q mice develop motor learning impairment before any obvious Purkinje cell death occurs or nuclear inclusions form in the cerebellum 5 . In the limbic area, Sca1 154Q/2Q mice show nuclear inclusions in pyramidal neurons, and cognitive deficits are observed without evident neuronal loss 5 . Clinical studies have demonstrated that neuronal death is most prominent in the cerebellum, whereas little occurs in the cerebral cortex and hippocampus, despite the presence of cognitive impairments in patients with SCA1 6 . These lines of evidence suggest that neuronal dysfunction, preceding cell death, causes subsequent behavioural impairments in the pathogenesis of SCA1; however, the mechanisms underlying the dysfunction remain unclear.
In the present study, we focused on SCA1 as a genetic model of neurodegenerative disease, and used Sca1 154Q/2Q knock-in mice to elucidate the synaptic basis of neuronal dysfunction. We analysed the dynamics, morphology, and density of dendritic protrusions, which are excitatory postsynaptic structures classified into mature 'spines' and immature 'filopodia' . These features are strongly associated with synaptic development 9 , plasticity 10 , and various pathologies 11 . Using two-photon laser-scanning microscopy, we investigated the synaptic pathologies of Sca1 154Q/2Q knock-in mice in vivo, maintaining contributions from peripheral tissues and non-neuronal cells expressing mutant ataxin-1, as well as neurons. To evaluate neuronal dysfunction while excluding the effects of neuronal death, we focused on the cerebral cortex and hippocampus, in which apparent neuronal death does not occur despite the presence of cognitive dysfunction in both Sca1 154Q/2Q mice and human SCA1 patients 5,7 . Our findings demonstrate that aberrant synaptic instability accompanied by a reduction in the expression of scaffolding proteins in affected neurons appears during synaptic development in SCA1 mice. These results suggest that deficits in neuronal circuitry development may underlie subsequent behavioural and neurological impairments in late-onset neurodegenerative diseases.
To evaluate protrusion stabilization that occurs with neuronal circuitry development and its disruption in SCA1 mice, we investigated the 1 h turnover rate of spines and filopodia in SCA1 mice at various ages during development (Fig. 3a). In Sca1 154Q/2Q mice, the spine turnover rate was significantly higher than that in Sca1 2Q/2Q mice throughout synaptic maturation ( Fig. 3b; 4 wks: p = 0.0009; 6 wks: p < 0.0001; 8 wks: p < 0.0001; two-way ANOVA followed by Bonferroni test), whereas filopodium turnover rate was higher in 6-and 8-week-old Sca1 154Q/2Q mice ( Fig. 3c; 4 wks: p = 0.2299; 6 wks: p = 0.0123; 8 wks: p = 0.0042; two-way ANOVA followed by Bonferroni test). No significant difference was observed in filopodium turnover rate at 4 weeks of age. This may be because of the high turnover rate of filopodia during early synaptic development, even in the control group. These results indicate that the normal development of dendritic protrusions, particularly filopodium stabilisation, is disrupted in Sca1 154Q/2Q mice. SCA1 mice exhibit progressive impairments in the density and morphology of dendritic protrusions in the hippocampus. Sca1 154Q/2Q mice show spatial memory deficits at 8 weeks of age 5 , but there is little neuronal death in the hippocampus, a region associated with spatial learning 6 . To elucidate the neuronal dysfunction associated with cognitive deficit in the absence of neuronal loss, we focused on hippocampal CA1 dendrites and investigated the density and morphology of dendritic protrusions in SCA1 mice. We performed confocal laser-scanning microscopy on slices of fixed brain samples from 5-and 12-week-old Sca1 2Q/2Q and Sca1 154Q/2Q mice (Fig. 4a,e). We chose this technique because the hippocampus is deep within the brain, precluding the non-invasive use of in vivo two-photon imaging. No difference in protrusion density was observed between Sca1 2Q/2Q and Sca1 154Q/2Q mice at 5 weeks of age (Fig. 4b) [Sca1 2Q/2Q (1.66 ± 0.08/μ m) vs Sca1 154Q/2Q (1.53 ± 0.06/μ m), p = 0.2135; unpaired t-test], whereas at 12 weeks, dendritic protrusion density was significantly lower in Sca1 154Q/2Q mice than in Sca1 2Q/2Q mice (Fig. 4f) [Sca1 2Q/2Q (1.53 ± 0.06/μ m) vs Sca1 154Q/2Q (1.28 ± 0.04/μ m), p = 0.0011; unpaired t-test]. An abnormal frequency distribution of dendritic protrusion width was observed in Sca1 154Q/2Q mice, particularly at 12 weeks of age, when the distribution curve shifted to the left (Fig. 4c,g; 5 wks: p = 0.0107; 12 wks: p < 0.0001; Kolmogorov-Smirnov test). The mean dendritic protrusion width in 12-week-old Sca1 154Q/2Q mice was lower than that in Sca1 2Q/2Q mice ( Fig. 4i; 5 wks: p = 0.8637; 12 wks: p < 0.0001; two-way ANOVA followed by Bonferroni test). These results show that the dendritic protrusion width in Sca1 154Q/2Q mice decreased as SCA1 symptoms developed. An abnormal frequency distribution of protrusion length in Sca1 154Q/2Q mice was evident at 5 weeks of age, and the frequency distribution  (a,f) In vivo two-photon imaging of dendrites in 6-and 8-week-old Sca1 2Q/2Q mice (n = 14 dendrites from five mice and n = 12 dendrites from four mice, respectively) and Sca1 154Q/2Q mice (n = 14 dendrites from five mice and n = 10 dendrites from five mice, respectively). Repeated imaging of the same dendrites over 1 h in each group showed increased formation (filled arrowhead) and elimination (open arrowheads) of dendritic protrusions in Sca1 154Q/2Q mice at 6 and 8 weeks of age. (b,g) Dendritic protrusion density in 6-and 8-week-old Sca1 2Q/2Q and Sca1 154Q/2Q mice. Sca1 154Q/2Q mice exhibited decreased spine density at both 6 (b) and 8 (g) weeks of age. (c,h) Spines and filopodia as a percentage of total protrusions in Sca1 2Q/2Q and Sca1 154Q/2Q mice at 6 (c) and 8 (h) weeks of age. The percentage of spines was lower, and filopodia higher, in 8-week-old Sca1 154Q/2Q mice than in Sca1 2Q/2Q mice, whereas no differences were observed in 6-week-old Sca1 154Q/2Q mice. (d,i) Percentage of total spines formed and eliminated. Sca1 154Q/2Q mice showed higher formation and elimination rates of spines than Sca1 2Q/2Q mice at 6 (d) and 8 (i) weeks of age. (e,j) Percentage of total filopodia formed and eliminated. Formation and elimination rates were greater in 6-(f) and 8-week-old Sca1 154Q/2Q mice than in Sca1 2Q/2Q mice (k). Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, Student t-test. Scale bar, 5 μ m.
curve was shifted to the right (Fig. 4d,h; 5 wks: p < 0.0001; 12 wks: p = 0.0441; Kolmogorov-Smirnov test). No difference was observed in the mean length of dendritic protrusions between Sca1 154Q/2Q and Sca1 2Q/2Q mice at either age ( Fig. 4j; 5 wks: p = 0.5210; 12 wks: p > 0.9999; two-way ANOVA followed by Bonferroni test). These results indicate that the protrusion lengths in Sca1 154Q/2Q mice differ little from those in Sca1 2Q/2Q mice from early life through to adulthood. In summary, SCA1 mice demonstrated progressive deficits in dendritic protrusions in the hippocampus from adult ages.

Discussion
Understanding the mechanisms of the neuronal dysfunction that precedes behavioural impairments and neuronal death is a longstanding challenge in neurodegenerative diseases such as SCA1. We found that Sca1 154Q/2Q mice showed abnormal synaptic instability in the cerebral cortex during the development of neuronal circuitry, when apparent nuclear inclusions, neuronal death, or behavioural impairments are not yet observed. Synaptic instability in the cerebral cortex of Sca1 154Q/2Q mice persisted into adulthood in the cerebral cortex, and subsequent deficits in the number and morphology of dendritic protrusions became evident as symptoms developed. We also observed progressive deficits of dendritic protrusions in Sca1 154Q/2Q hippocampus, a region implicated in cognitive dysfunction in SCA1. Furthermore, compared with Sca1 2Q/2Q mice, Sca1 154Q/2Q mice showed lower expression levels of the postsynaptic scaffolding proteins Homer and Shank, even before synaptic maturation, when increased synaptic instability was observed. These results suggest that one of the mechanisms underlying neuronal dysfunction in SCA1 involves the association of synaptic instability, abnormal protrusion morphology during synaptic development, and a decline in scaffolding protein expression. We therefore hypothesized that impaired synaptic development triggers subsequent neurological symptoms and pathological abnormalities. Sca1 154Q/2Q mice demonstrated higher spine turnover rates throughout synaptic development than Sca1 2Q/2Q mice. (c) Compiled turnover rates of filopodia at different ages (in weeks). Sca1 154Q/2Q mice showed higher filopodium turnover rates than Sca1 2Q/2Q mice from 6 weeks of age. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, Student t-test (b) or two-way ANOVA followed by Bonferroni test (c).

Figure 4. SCA1 mice show progressive deficits in the density and morphology of hippocampal dendritic protrusions. (a-h) Analysis of dendrites in the hippocampal CA1 stratum radiatum of 5-(a-d) and
12-week-old (e-h) Sca1 2Q/2Q mice (n = 15 dendrites from three animals and n = 20 dendrites from four animals, respectively) and Sca1 154Q/2Q mice (n = 15 dendrites from three animals and n = 25 dendrites from five animals, respectively). (a,e) Confocal images of dendrites in 5-and 12-week-old Sca1 2Q/2Q and Sca1 154Q/2Q mice. Images are best projections (3-7 optical sections, 0.43 μ m apart). (b,f) Dendritic protrusion density in Sca1 2Q/2Q and Sca1 154Q/2Q mice. Sca1 154Q/2Q mice showed a lower protrusion density than Sca1 2Q/2Q mice at 12 weeks of age. (c,g) Cumulative frequency distribution of protrusion width in Sca1 2Q/2Q and Sca1 154Q/2Q mice. The distribution of protrusion width was abnormal in Sca1 154Q/2Q mice, particularly at 12 weeks of age. (d,h) Cumulative frequency distribution of protrusion length in Sca1 2Q/2Q and Sca1 154Q/2Q mice. The distribution of protrusion length was abnormal in Sca1 154Q/2Q mice at both ages. (i) Mean protrusion width at 5 and 12 weeks of age. Sca1 154Q/2Q mice had narrower protrusions than Sca1 2Q/2Q mice at 12 weeks of age. (j) Mean protrusion length at 5 and 12 weeks of age. Sca1 154Q/2Q mice had normal protrusion lengths at both ages. Data are presented as the mean ± SEM (b,f,i,j). *p < 0.05, **p < 0.01, ***p < 0.001, Student t-test (b,f), Kolmogorov-Smirnov test (c,d,g,h), or two-way ANOVA followed by Bonferroni test (i,j). Scale bar, 5 μ m. We used an established knock-in mouse model of SCA1, which expresses mutant ataxin-1 at endogenous levels in the normal spatial and temporal pattern, and accurately replicates pathological features observed in the human disease 5,13 . Sca1 154Q/2Q mice show motor learning impairment by approximately 5 weeks of age, which is followed by the development of nuclear inclusions, cognitive deficits, and Purkinje cell death. Other studies have also demonstrated motor learning impairment in 5-week-old mice, before neuronal death in the cerebellum, which occurs only in the late stages of the disease, using Sca1 transgenic mice expressing full-length human ATXN1 cDNAs with 82 CAG repeats specific to Purkinje cells 14,15 . Therefore, the abnormal synaptic instability detected in 4-week-old Sca1 154Q/2Q mice in the present study is one of the earliest pathological signs observed in SCA1 mouse models.
We performed in vivo imaging to rigorously evaluate the synaptic pathology of SCA1, because non-neuronal cells expressing mutant ataxin-1 are also involved in the pathogenesis of SCA1 models [16][17][18] . Furthermore, we believe that analysis of protrusion dynamics in living animals, in addition to morphology, enabled us to detect potential synaptic lesions. There have been cases in which changes in protrusion dynamics were observed, without any alterations in protrusion density, upon changes in synaptic plasticity and pathology 10,19 . Indeed, we also demonstrated a higher turnover rate of dendritic protrusions in 4-week-old Sca1 154Q/2Q mice than in Sca1 2Q/2Q mice, in the absence of any differences in protrusion density. We applied the thinned-skull method for in vivo imaging, which allows excitation and emission lights to penetrate the skull without eliciting any microglial inflammatory responses 20 , because many neurodegenerative diseases are associated with inflammation of the brain 21,22 .
A recent study using the rotarod test demonstrated that motor skill acquisition and coordination require the activation of neurons in the secondary motor cortex, which receives inputs from the somatosensory cortex 23 . Here, we demonstrated synaptic instability and dendritic spine loss in the somatosensory cortex of Sca1 154Q/2Q mice, which also show impaired rotarod performance 5 . These results suggest that Sca1 154Q/2Q mice have deficits in somatosensory and sensorimotor function. In the cerebellum of Sca1 154Q/2Q mice, however, no synaptic dysfunction, neuronal cell death or nuclear inclusions of ataxin-1 protein are observed when motor incoordination develops at 5 weeks of age 5 . Sca1 154Q/2Q mice do not show cerebellar neurodegeneration until 16 weeks and nuclear inclusions until 21 weeks. Therefore, the involvement of cerebellar dysfunction in the early symptomatic stage in Sca1 154Q/2Q mice remains unclear.
Previous studies, using conditional Sca1 transgenic mice that stage-specifically express a full-length human ATXN1 cDNA with 82 CAG repeats in the cerebellum, have demonstrated that the suppression of mutant ataxin-1 expression during the first 14 postnatal weeks inhibits subsequent motor dysfunction and dendritic atrophy 24,25 . Suppressing the expression of mutant ataxin-1, even for the first 5 postnatal weeks, can also inhibit impairments in synaptic transmission in the adult cerebellum 26 . Neuronal circuitry development occurs during the first few postnatal weeks, with a decrease in the turnover of dendritic protrusions and in the ratio of filopodia to total protrusions 9 . We identified enhanced synaptic instability in Sca1 154Q/2Q mice at 4 weeks of age compared with controls. These results suggest that the stage at which synaptic development occurs is a critical period in SCA1 pathogenesis, and that dendritic protrusions are excessively unstable in SCA1 mice and do not stabilise with maturation. Our present findings can be interpreted as developmental impairment in the synapses of SCA1 mice. This is a conceptually novel finding that implies that it is not only neurodevelopmental disorders, such as fragile X syndrome, autism spectrum disorder, and schizophrenia that involve deficits in synaptic development, but also SCA1, a late-onset neurodegenerative disease. It should be noted that synaptic instability in SCA1 mice is commonly observed in animal models of these neurodevelopmental disorders 19,[27][28][29] . Interestingly, Shank genes, associated with autism, were also downregulated in SCA1 mice, and Shank is required for the maintenance of the density and morphology of dendritic spines 30,31 . Other studies using mouse models of neurodegenerative diseases such as Alzheimer's and Huntington's have also demonstrated instability and progressive loss of dendritic protrusions similar to that observed in Sca1 154Q/2Q mice 32,33 ; however, these studies did not investigate the dynamics of dendritic protrusions at 4 weeks of age, i.e., during synaptic development. In contrast, we detected abnormal synaptic instability in 4-week-old Sca1 154Q/2Q mice. It is possible that the Alzheimer's and Huntington's disease models both show synaptic instability during development of neuronal circuitry, due to the similarities in synapse pathologies and subsequent progression of symptoms among these neurodegenerative disease models 34,35 . In Huntington's disease models in particular, there are many similarities to Sca1 154Q/2Q mice: both Huntington's and SCA1 are polyglutamine diseases; Huntington's mouse models (R6/1 and R6/2) and Sca1 154Q/2Q mice have the same extent of expanded CAG repeats; and both models show synaptopathy [36][37][38][39][40] . We can therefore hypothesize that many neurodegenerative diseases share latent deficits in neuronal circuitry development, which precede the onset of symptoms.
The hippocampus of Sca1 154Q/2Q mice show impaired CA1 synaptic plasticity and dendritic arborization by 24 weeks of age, but no differences from control mice are observed until 8 weeks of age 5,13 . Our present results indicate that an evident decrease in the density of dendritic protrusions in the hippocampus of Sca1 154Q/2Q mice occurs between 5 and 12 weeks of age. This suggests that synaptic deficits in the hippocampus of Sca1 154Q/2Q mice develop by 12 weeks, and that they are mainly due to postsynaptic impairments. In the cerebral cortex, however, Sca1 154Q/2Q mice showed synaptic instability by 4 weeks of age and a decrease in synaptic number by 6 weeks. It is possible that in the hippocampus, Sca1 154Q/2Q mice also develop deficits in dendritic protrusion dynamics during the early stages of development. Sca1 154Q/2Q knock-in mice demonstrate cerebellar abnormalities, manifesting as motor learning impairment, at 5 weeks of age, although there is little difference from controls in the electrophysiological properties of Purkinje cells at this age 5 . The association between motor behavioural impairment and synaptic dysfunction in the cerebellum of Sca1 154Q/2Q mice remains unknown, and in vivo imaging studies of the synapses of Purkinje cells in Sca1 154Q/2Q mice are warranted.
In the present study, we found that expression levels of Homer and Shank proteins were lower in Sca1 154Q/2Q mice than in Sca1 2Q/2Q mice. This decline in the expression of postsynaptic scaffolding proteins occurred before synaptic maturation. Interestingly, Shank proteins were lower in Sca1 154Q/2Q mice exclusively at 4 weeks of age. Shank1 and Shank2 mRNA expression are higher during postnatal brain development than after maturation 41 , and their temporal expression patterns are similar to that of ATXN1 mRNA 42 . Therefore, the effect of mutant ataxin-1 on the expression of Shank proteins may be strongest during postnatal development. These lines of evidence provide an insight into the molecular mechanisms of developmental impairments in the synapses of Sca1 154Q/2Q mice. It is interesting that Shank1 knock-out mice show impaired rotarod performance and decreased spine width, similarly to Sca1 154Q/2Q mice 43 . Homer and Shank, which form a polymeric network structure at postsynaptic sites, interact with glutamate receptors and regulate their downstream signalling, inducing the accumulation of inositol-1,4,5-triphosphate (IP3) receptors in protrusions [44][45][46] . Moreover, Homer and Shank proteins regulate the morphology and function of dendritic protrusions 31,45 . Because the morphology of dendritic protrusions correlates with the dynamics of these proteins 47 , Homer and Shank proteins may be involved in protrusion turnover. Our results are supported by previous studies that demonstrated a reduction in the levels of Homer3 and IP3 receptors in the Purkinje cells of Sca1 transgenic mice 48,49 . Taken together, this evidence suggests that Sca1 154Q/2Q mice have deficits in Homer-and Shank-mediated intracellular calcium release from IP3 receptors, and the deficits may cause synaptic instability and abnormal synaptic maturation.

Materials and Methods
All experimental protocols were approved by an Animal Ethics Committee at the National Institute of Neuroscience, National Center of Neurology and Psychiatry, Japan, and performed in strict accordance with institutional guidelines.
Surgical procedure for in vivo imaging. The thinned-skull cranial window technique 52 was used because it is less invasive than the open-skull method 20 . Sca1 2Q/2Q and Sca1 154Q/2Q mice expressing YFP were deeply anesthetized with intraperitoneal ketamine and xylazine (0.1 and 0.015 mg/g body weight, respectively). Body temperature was maintained at 37 °C with a heating pad during surgery and imaging. Eyes were lubricated with ointment to prevent dryness. After scalp incision, the primary somatosensory area (1.1 mm posterior to bregma and 3.4 mm lateral from the midline) was identified with stereotactic coordinates. A small metal plate with a round hole was glued onto the skull with cyanoacrylate glue and acrylic resin dental cement (Unifast; GC, Tokyo, Japan), and mice were fixed to a custom-made skull immobilization stage via the metal plate. The skull above the imaging area, located in the center of the hole in the metal plate, was thinned to approximately 20 μ m with a high-speed microdrill (UG23A/ UC210C; Urawa, Saitama, Japan) and microsurgical blade (USM-6400; Sable Industries, Vista, CA, USA). The hole in the metal plate was filled with artificial cerebrospinal fluid during surgery and imaging.
In vivo transcranial two-photon imaging. Sca1 2Q/2Q and Sca1 154Q/2Q mice were imaged under anaesthesia using a two-photon laser-scanning microscope (FV1000-MPE; Olympus, Japan) with a water-immersion objective lens (25× , NA 1.05) at 8× digital zoom, yielding high-magnification images suitable for the quantification of dendritic spines. A Ti-sapphire laser (MaiTai HP DeepSee-OL; Spectra-Physics, Mountain View, CA, USA) was tuned to 950 nm. To minimize phototoxicity, laser intensity was maintained between 10 and 30 mW at the focus. Image stacks (512 × 512 pixels; 0.124 μ m/pixels; 0.75 μ m z-step) were taken at approximately 70 μ m below the pial surface, where layer 1 dendrites of layer 5 pyramidal neurons are located. Dendrites were imaged for each experiment at time 0, and again after an interval of either 1 or 48 h. Image acquisition time in each imaging session was approximately 5 minutes. In the 1 h interval experiment, mice were maintained under anaesthesia between the two imaging sessions. In the 48 h interval experiments, mice were allowed to recover from anaesthesia after the first imaging session and were returned to their home cages until the next session.
Confocal laser-scanning microscopy for ex vivo fixed samples. YFP-labelled Sca1 2Q/2Q and Sca1 154Q/2Q mice were anesthetized and perfused transcardially with phosphate-buffered saline (pH 7.4) followed by 4% paraformaldehyde. Brains were removed and 50 μ m sections were cut with a Vibratome 3000 (Vibratome Company, St Louis, MO, USA). Image stacks (1024 × 1024 pixels; 0.069 μ m/pixel; 0.43 μ m z-step) were taken of the secondary dendrites of CA1 neurons in the dorsal hippocampus, using a confocal laser-scanning microscope (FV1000; Olympus, Japan) with a silicone-immersion objective lens (60× , NA 1.3) at a digital zoom of 3. For YFP, the excitation and emission wavelengths were 488 nm and 515 nm, respectively. Image analysis. The turnover rate, density, head width, and neck length of the postsynaptic dendritic protrusions were analysed with Neurolucida neuron tracing software (MicroBrightField, Williston, VT, USA) from three-dimensional two-photon or confocal z-stacks. Morphometric analysis of dendritic protrusions was performed in accordance with a previous report 9 . Filopodia were identified as long, thin structures (head width to neck width ratio < 1.2:1; protrusion length to neck width ratio > 3:1). Other protrusions were classified as spines. Protrusions were identified as being identical between two successive frames by their spatial relationship to adjacent landmarks (e.g., axonal and dendritic orientations) and their relative position to immediately adjacent protrusions. Protrusions were considered different between two successive frames if they were located > 0.7 μ m from their expected positions based on the first image. The formation and elimination rates of the protrusions were defined as the percentage of protrusions that appeared and disappeared, respectively, between two successive frames, relative to the total protrusion number. Protrusion turnover rate was defined as the sum of the protrusions formed and eliminated, divided by twice the total number of protrusions. Data were collected from 10-18 dendrites and 460-1,162 protrusions in four to eight mice for in vivo studies, and from 15-25 dendrites and 1,676-2,350 protrusions in three to five mice for ex vivo studies.

Statistical analysis.
To determine statistical significance, we used the Student t-test, one-way