Cerebellar defects are a primary pathology in mouse models of spinal muscular atrophy

Abstract

Spinal muscular atrophy (SMA), a leading genetic cause of infant mortality worldwide, is caused by reduced levels of the ubiquitous survival motor neuron (SMN) protein in SMA patients. Despite significant advancement in recent research and clinical treatments, the cellular pathologies that underlie SMA disease manifestations are not well characterized beyond those of spinal motor neurons (MNs). We previously reported cerebellar abnormalities in an SMA mouse model at the late stage of the disease, including volumetric deficits and lobule‐selective structural changes with Purkinje cell degeneration, with colocalized astrocytic reactivity. However, when these cerebellar defects arise and whether they are a consequence of MN degeneration remain unknown. We used magnetic resonance imaging, immunohistochemistry, and electrophysiology to characterize cerebellar pathology in early‐stage symptomatic SMNΔ7 mice and late‐stage SMA mice with transgenic rescue of SMN in MNs. We found disproportionate structural and lobule‐specific surface area deficits, as well as abnormal functional properties in the cerebella of early symptomatic SMA mice, suggesting that cerebellar pathologies may be a primary contributor to murine SMA phenotypes. Moreover, cerebellar pathologies were not ameliorated in SMA mice with MN rescue, suggesting that cerebellar neurons are independently vulnerable to reduced SMN expression. Overall, our study shows that cerebellar defects are a primary pathology in SMA mouse models and that therapies targeting cerebellar neurons in SMA patients may be needed for optimal treatment outcomes.

Purkinje cell (PC) degeneration is localized to posterior lobules in the cerebellum, and rescue of survival motor neuron protein expression levels in motor neurons does not ameliorate this effect. Representative images of sagittal cerebellar sections stained with anti‐calbindin in the vermis and hemisphere at P12 for wild type, ChATCre+ rescue (Rescue), and spinal muscular atrophy (SMA) cerebella are shown on the left. Lobules are labeled; ANCr, ansiform crus; COP, copula pyramidis. Arrows point to equivalent loci across genotypes where neurodegeneration is most prominent. PC perimeter density was calculated by dividing the PCs by the PC layer length. Scale bar = 0.5 mm.

1. INTRODUCTION

Spinal muscular atrophy (SMA) is one of the leading genetic causes of infant mortality worldwide and affects 1 in 6000–10,000 individuals [1, 2]. SMA patients harbor mutations of the telomeric survival motor neuron 1 (SMN1) gene and rely on expression from the centromeric gene, survival motor neuron 2 (SMN2) [3]. Both genes encode the survival motor neuron protein (SMN), a ubiquitously expressed protein with many critical roles in pre‐mRNA splicing, axon and dendrite growth, RNA metabolism, and more (for review, see [4]). However, only about 15% of SMN2 transcripts are full length and encode functional SMN protein [3, 5]. This makes the severity of patient phenotypes largely dependent upon the copy number of the SMN2 gene [2, 6].

Pathology studies of SMA have focused on degeneration of spinal cord motor neurons (MNs) in autopsy studies of SMA patients [7] and in mouse models [8, 9, 10]. In more recent years, the specific abnormalities of the neuromuscular junction (NMJ), such as the morphological defects, immaturity, denervation, as well as changes in transcriptome and neurotransmitter release have also been well characterized in mouse models [11, 12, 13, 14, 15, 16, 17, 18, 19] and to a lesser degree in patients [20, 21, 22]. In addition to MN abnormalities, several studies have posited that muscle pathology is an independent contributor to the SMA phenotype [23, 24, 25]. Most recently, an array of brain abnormalities has been identified in humans and in mouse models [21, 26, 27, 28, 29, 30]. As more widespread pathologies emerge, the conceptualization of SMA has expanded to that of a multi‐system disease [27, 30, 31, 32, 33, 34, 35, 36], and this has important implications for disease treatment. Although there have been monumental strides in the development of SMA therapeutics, disease manifestations persist in most treated patients. Some of these deficits might be caused by the consequences of SMN deficiency in nervous system regions outside of spinal cord MNs.

The cerebellum in particular has been understudied in SMA, even though it has extensive connectivity with the spinal cord via the ventral and dorsal spinocerebellar tracts—pathways that are responsible for processing sensory information and compensating for perturbations during rhythmic motor activity [37, 38]. Moreover, SMN is highly expressed in cerebellar cell types such as deep cerebellar nuclei (DCN) neurons, granule cells, and Purkinje cells (PCs) [39, 40, 41, 42]. One of the earliest histochemical case studies in the cerebellum of an SMA patient found abnormal deposits in the cytoplasm and axons of granule cells and PCs, which were also abnormally clustered and displaced [43]. Another autopsy examination of a type 0 SMA patient identified precursors of apoptosis in the cerebellum [44]. Studies using magnetic resonance imaging (MRI) of SMA patient cerebella have found reduced volume and selective degeneration in the posterior lobules [45, 46]. We previously investigated the cerebella of SMNΔ7 mice at the late stage of postnatal day (P) 12 and found diminished spinal cord connectivity, broad functional deficits, structural degeneration, PC degeneration, and increased glial reaction colocalized to posterior lobules [47, 48]. Our findings indicated that the cerebellum was affected by SMA but also opened questions as to its contribution to SMA pathology. When do these pathologies present in mice? Does the low level of functional SMN protein in cerebellar neurons directly cause these abnormalities? Or are cerebellar pathologies a downstream consequence of MN degeneration?

To answer these questions, we characterized cerebellar pathology in two cohorts of SMA mice using an array of techniques, including MRI, immunohistochemistry (IHC), and physiological multi‐electrode array (MEA) recordings, which enabled direct comparison to our previously published data. We first studied the cerebellar defects from a cohort of early symptomatic SMNΔ7 mice to determine whether cerebellar defects progress and contribute to phenotypes of SMA in a manner similar to what has been reported in MNs. Next, we studied the cerebellum of late‐stage (P12) ChAT ^Cre+ rescue SMA mice. This model has the same genetic background as SMNΔ7 mice, and the age of onset and phenotypes of SMA mice are similar. However, while the rescue mice show decreased MN degeneration, phenotypes such as body weight, survival, and motor capability are only marginally improved from SMA. This allows us to investigate how the modulation of SMN expression in MNs will impact pathologies in the cerebellum [25]. We report that early symptomatic SMA mice already exhibit volumetric, lobule‐specific surface area, and functional deficits in the cerebellum. Although ChAT ^Cre+ rescue mice in late disease stages have a whole brain volume greater than SMA mice, there is no change in cerebellar volume nor cerebellar surface area deficits or PC neurodegeneration. This study illustrates that cerebellar defects are a primary pathology in SMA mouse models.

2. METHODS

2.1. Animals

All animal procedures were approved by the Institutional Animal Care and Use Committee of Delaware State University. Mice were maintained under a 14/10 h light/dark photoperiod with PMI rodent diet (Animal Specialties and Provisions) and water available ad libitum. Heterozygote male and female mice of the FVB.Cg‐Grm7 ^{Tg(SMN2)89Ahmb} Smn1 ^tm1Msd Tg(SMN2*delta7)4299Ahmb/J strain obtained from Jackson Laboratory (stock#: 005025, Bar Harbor, ME) were mated to produce pups for experiments. The day of birth was considered P0, and mice were investigated at P3. Both sexes were used for experiments. Mouse genotyping was done by TransnetYX (Germantown, TN) after experimentation and data analyses, but before performing statistical tests for significance. For the early symptomatic cohort of SMN∆7 mice, wild type (WT) and heterozygous pups were pooled as controls (CTL), and pups with the homozygous mutation formed the SMA group as described previously [48].

The ChAT ^Cre rescue mouse line was developed by the Sumner laboratory at Johns Hopkins University with an intricate genetic manipulation [25]. ChAT ^Cre (Jackson Laboratory Stock #006410) was bred to SMA mice expressing a Smn Cre‐inducible allele (Smn ^Res ) [49] and two copies of human SMN2 and SMNΔ7 alleles. WT mice were homozygous for the murine Smn allele (Smn ^+/+ / SMN2 ^+/+ / SMNΔ7 ^+/+ ) in the absence of Cre. The ChAT ^Cre+ rescue experimental group consisted of mice lacking the murine Smn allele that were homozygous for the Cre‐inducible Smn allele (Smn ^Res/Res /SMN2 ^+/+/SMNΔ7 ^+/+ ) and either homozygous (ChAT ^Cre+/+ ) or heterozygous (ChAT ^Cre+/− ). Since we did not find a significant difference between homozygous ChAT ^Cre+/+ rescue and heterozygous ChAT ^Cre+/− rescue among all our measurements, the two genotypes were combined as ChAT ^Cre+ rescue mice in this study. SMA mice with a mutated murine Smn allele (Smn ^−/− /SMN2 ^+/+/SMNΔ7 ^+/+ ) were also homozygous for the Cre‐inducible Smn allele (Smn ^Res/Res /SMN2 ^+/+/SMNΔ7 ^+/+ ) in the absence of Cre. Mouse breeding pairs used for these experiments were provided by the Sumner laboratory. Pups of both sexes and all genotypes were euthanized at P12, and the brains were collected for experiments. The tails were sent to TransnetYX (Germantown, TN) for genotyping.

2.2. T2‐weighted magnetic resonance imaging and volumetric measurements

Magnetic resonance (MR) images were obtained using a 9.4T Bruker Biospec 94/20 small animal MR system (Bruker BioSpec MRI, Ettlingen Germany) in the Center for Biomedical and Brain Imaging at the University of Delaware. SMNΔ7 mouse brains were extracted at P4 and ChAT ^Cre rescue mouse brains were extracted at P12. They were immediately fixed in 4% paraformaldehyde and stored at 4°C. Mouse brains were fixed for at least 2 weeks before being submerged in fomblin oil inside a 15 mL conical tube with a 3D‐printed brace for ex vivo MRI. We collected high‐quality T2‐weighted anatomical imaging sets using the turbo rapid acquisition with relaxation enhancement (turboRARE) pulse sequence and diffusion tensor images. For the SMNΔ7 mice at P4, T2‐weighted images were obtained with the following parameters: a repetition time of 1.5 s, an echo time of 40 ms, 1 average, isotropic 100 μm³ voxel resolution, and a rare factor of 12. The scan time was 25 min for each brain. For the P12 ChAT ^Cre rescue mouse brains, the parameters for T2‐weighted imaging were as follows: a repetition time of 2.5 s, an echo time of 40 ms, 1 average, isotropic 100 μm³ voxel resolution, and a rare factor of 12. The scan time was 28 min.

We acquired volumetric measurements of whole brains and cerebella via segmentation in ITK‐SNAP (itksnap.org). We used T2‐weighted images to inform the manual segmentation of the cerebella and semi‐automatic segmentation of the whole brain. From these segmentations, we generated 3D renditions and volumetric measurements within the program. Regions that were lost during the extraction or scanning process were extrapolated using the Allen Brain Atlas (mouse.brain-map.org). In P4 mice, we acquired morphological measurements of cerebellar width, as well as vermis and hemisphere heights and depths using T2‐weighted scans and a ruler tool in DSI‐Studio (dsi-studio.labsolver.org). Proportional measurements of cerebellar size were calculated as a percentage of the whole brain volume (i.e., relative cerebellar volume). These measurements were used for comparison between CTL and SMA mice in SMNΔ7 model, as well as between SMA and ChAT ^Cre+ rescue mice in the ChAT ^Cre rescue mouse line. Multiple unpaired two‐tailed t‐tests with a false discovery rate of 1% were used for statistical comparisons between SMA and CTL, and ordinary one‐way analysis of variance (ANOVAs) with Tukey's correction for multiple comparisons between SMA and ChAT^Cre+ .

2.3. IHC, imaging, and analysis

IHC staining with low‐ and high‐magnification imaging was performed at the Histochemistry and Tissue Processing Core Laboratory at Nemours A.I. duPont Children's Hospital in Wilmington, Delaware. Mouse brains were bisected along the sagittal plane and processed for 1 h on an ASP300S tissue processor (Leica, Buffalo Grove, IL). For the early‐stage mice in SMNΔ7 model, four out of six brains within the SMA and CTL groups were used for both IHC and MRI scanning. In the P12 ChAT ^Cre rescue mouse line, all brains were used for both IHC and MRI scanning. We chose this approach to enhance comparative power across gross structural and cellular information. Each brain was bisected along the sagittal plane and paraffin‐embedded on a single block; one half was embedded with the medial side down, while the other half was embedded with the lateral side down. Then for each brain, two slices of 5 μm thickness were floated onto one Superfrost Plus slide (Fisher Scientific, Pittsburg, PA). This was done three times with each pair of sections on an individual slide; each slide was stained with one of the following antibodies: anti‐calbindin antibody (Abcam ab229915) for PC staining, anti‐glial fibrillary acidic protein (GFAP) antibody (Abcam ab207165) for astrocyte staining, and anti‐NeuN antibody (Abcam ab177487) for granule cell staining. For each brain, this was done twice in both hemisphere and vermis regions of the cerebellum. Slides were incubated for 30 min in one of the three antibodies, incubated in a Peroxide Block, developed in 3,3′‐diaminobenzidine (DAB), and counter‐stained with hematoxylin to visualize the nucleus of the cells. The slides were dried with ethanol, then xylene, and coverslipped with Tissue‐Tek® Prisma™ mounting media (Sakura, Torrance, CA). Imaging of the cerebellar slices was performed with an Olympus Bx51 Microscope (Olympus, Waltham, MA) for all three stains. The three antibodies combined with hematoxylin counterstaining enabled morphological analyses. Measurements of immunoreactivity were performed using ImageJ (imagej.net/ij) and cell detection was performed with the QuPath (qupath.github.io) brightfield DAB positive cell detection function.

We counted cells and acquired morphological measurements as previously described [47]. Using the 4X NeuN‐stained images, single‐lobule region of interest (ROIs) were acquired in accordance with the Allen Brain Atlas (atlas.brain-map.org) within the vermis (lobules III–X) and hemisphere (simple lobule: SIM, ansiform crus lobule I and II: ANCr1, ANCr2, and paramedian lobule: PRM). The total cerebellar and lobular surface areas were measured and the lobule proportions were calculated by dividing the total area by the area of each lobule. The cerebellar perimeter and total length of the PC layer were measured, and the latter was divided by the former to calculate foliation index (F _i) [50]. Folia lengths in the vermis and hemisphere were also measured. Using calbindin‐stained slices, PC count per slice and per lobule were acquired manually by two researchers who were blinded to mouse genotype. PC perimeter density was then calculated by dividing the number of PCs (either within lobules or across the entire slice) by the (lobular or total) PC layer lengths. For all the measures above, statistical comparisons between WT, ChAT ^Cre+ rescue SMA and SMA mice in the ChAT ^Cre rescue model, as well as between CTL and SMA mice in P4 SMNΔ7 mouse model, were performed using ordinary one‐way ANOVAs with Šídák's correction for multiple comparisons. All immunoreactive pixel density analyses were performed using QuPath version 0.3.0 (qupath.github.io) brightfield H‐DAB cell detection function with a Gaussian sigma of 2 px, negative threshold of 0.1 optical density (OD) units and positive threshold of 0.11 OD units.

Slice images stained with anti‐GFAP were used to assess astrocytic reactivity across different regions in QuPath version 0.3.0. The positive pixel density, which is acquired by dividing the number of positive pixel detections in each ROI by its area, was calculated across whole sagittal sections, DCN, and the lobules of the vermis or hemisphere. In P4 SMNΔ7 mice, positive pixel density measurements by lobule were compared between SMA and CTLs. In the ChAT ^Cre+ rescue mice, relative GFAP expression in the cortices was calculated by dividing the percentage of positive pixels in the affected lobules (hemisphere: ANCr2, vermis: lobule VI/VII) by the percentage of positive pixels in the unaffected lobules (hemisphere: SIM, vermis: lobule III). Affected and unaffected lobules were chosen based on results from our previous study [47], as well as prior data in this study. All length and area measurements were acquired in ImageJ (imagej.net) and converted to μm or mm and μm² or mm², respectively. Statistical tests for significance were done using ordinary one‐way ANOVAs with Tukey's correction for multiple comparisons.

2.4. Microelectrode array recordings and analysis

A P3–4 mouse was anesthetized in a sealed chamber containing the inhalation anesthetic isoflurane in an upward flowing BSL2 cabinet. After decapitation, the brain was quickly removed and immersed in the ice‐cold oxygenated slicing solution contained (in mM): 2.5 KCl, 10 MgSO₄, 0.5 CaCl₂, 1.25 NaH₂PO₄, 234 sucrose, 11 glucose, and 26 NaHCO₃ (pH 7.2, 310–330 mOsm). Three hundred micrometer sagittal slices within the cerebellum were sectioned using a LEICA VT1200s tissue slicer (Leica Microsystems, Wentzler, Germany) while submerged in ice‐cold oxygenated slicing solution. Slices were incubated at 34°C for 30 min in oxygenated artificial cerebral spinal fluid (ACSF) solution contained (in mM): 126 NaCl, 3 KCl, 2 MgCl₂, 2 CaCl₂, 1.25 NaH₂PO₄, 10 glucose, and 26 NaHCO₃ (pH 7.2, 310–330 mOsm). Then the slices were kept at room temperature for more than 60 min before electrophysiology recordings.

Extracellular recordings were performed with a MEA (MEA 2100 system, MultiChannel Systems, Germany). The system consists of a peristaltic perfusion system pump for ACSF perfusion, a TC02 temperature controller, and a video microscope table to monitor the slice orientation. The microelectrode arrays (MEAs) for brain slice recording have 120 channels in a roughly 12 × 12 planar grid of titanium nitride (TiN) electrodes with 30 μm diameter and 200 μm spacing. Slice position and contact with electrodes of the MEA were ensured by slice anchors (Warner Instruments, USA). Data acquisition was performed using the Multi‐Channel Experimenter program with a data sample rate of 20 kHz at 37°C. To ensure comparability and enhance efficiency, a pair of mice from the same litter—one CTL‐like and one SMA‐like mouse (based on phenotype determined by weight and righting reflex) were recorded on the same day. The recording time was 3 min for each slice. The slice preparation and recording sequence for the CTL‐like and SMA‐like mice were alternated every day to avoid any difference caused by the timing or the preparation procedure. For data analysis, pictures of the cerebellum orientation on the MEA were saved for identifying electrodes that were in contact with the cerebellum during recording.

Analysis was done using Multichannel Analyzer to detect spontaneous spikes by Spike Detector with a high pass filter at 200 Hz. The spiking and bursting activities were then analyzed with Spike Analyzer and Burst Analyzer accordingly and exported to a spreadsheet for further analysis. Only the data of channels from the electrodes in contact with the cerebellum were included in the analysis based on the slice orientation picture taken during MEA recording. The total number of channels, active channel number, active channel percentage, total spikes, spikes/active channel, and average spike frequency were analyzed in Excel. An active channel was defined as recording >5 spikes in a 3 min recording. Statistical significance between CTL and SMA for all measurements was tested by student t‐test and chi‐square test with Prism (10.3.1, GraphPad Software, Boston, MA). Statistical significance was set at p < 0.05. The data shown represent mean ± standard error of the mean.

3. RESULTS

3.1. Cerebellar structural abnormalities are present in early symptomatic SMA mice

We used MRI scans of mouse brains to generate 3D renderings of the whole brain and the cerebellum in SMNΔ7 mice at P4 to detect possible cerebellar abnormalities at this age. The 3D renderings were done using semi‐automatic and manual segmentation of T2‐weighted images of the whole brain and cerebellum, respectively. The whole brain and cerebellar volumes themselves were not significantly smaller (Figure 1A–C) in SMA mice (n = 6) compared to CTL (n = 5). However, the cerebellar volume relative to whole brain volume was significantly lower in SMA compared to CTL (Figure 1D; p = 0.037). This indicates that the volume of the cerebellum is disproportionately smaller relative to the whole brain even at the early symptomatic stage in SMA mice.

FIGURE 1.

Spinal muscular atrophy (SMA) cerebella were disproportionately smaller and display morphological abnormalities at the early symptomatic stage. (A) Representative 3D segmentations from magnetic resonance imaging scanning for a control (CTL) and SMA cerebellum at P4. Scale bar = 1 mm. (B and C) Volumetric measurements of 3D renditions for control and SMA whole brains (B) and cerebella (C). (D) Cerebellar volume ratio, represented as a percentage of whole brain volume. (E) Point‐to‐point length measurements of various cerebellar dimensions. Inset: Representative cerebellum displaying where length measurements were acquired. x: cerebellar width, y: vermis height, y1: left hemisphere height, y2: right hemisphere height, z: vermis depth, z1: left hemisphere depth, z2: right hemisphere depth. ⊗ denotes the location of the depth measurements. *p < 0.05, **p < 0.01 for SMA (n = 6) compared to control (n = 5), unpaired two‐tailed Student's t‐test.

To understand which cerebellar regions may be accounting for this disproportionate loss in volume, we compared the dimensional morphometry of the cerebellum between SMA and CTL as well. This included measurements of cerebellar width, as well as depth and height for the vermis, left hemisphere, and right hemisphere. The only significant decrease observed was in the SMA cerebellar vermis height (denoted by y; p < 0.005). We also calculated the hemisphere ratios to test if one side or the other was particularly affected and found no significant differences, indicating that the cerebellum was otherwise symmetrically smaller (Figure 1E). These results suggest that the structure of the cerebellum is diffusely affected by SMA in mice, but the effect in the vermis is stronger than in the hemispheres.

In our previous study, we found that posterior cerebellar lobules in mice at the late stage of SMA are particularly vulnerable to the disease [47]. Sagittal sections within the cerebellar vermis and the hemispheres were stained with anti‐calbindin in SMA (n = 5) and CTL (n = 4) mice to assess regional cerebellar pathology in early symptomatic mice (Figure 2A,B), which allows for observations of structural integrity. We first measured the total surface area of the whole cerebella in both regions and found no significant differences (Figure 2C). Upon closer examination of the lobules, we found that the surface area of lobule VI/VII in the vermis was significantly smaller in SMA mice compared to CTLs (Figure 2D; p = 0.004). In the hemisphere sections, SMA mice had smaller total and hemisphere lobule surface areas (PRM; ANCr1/2), but the difference did not reach significance (Figure 2C,E). We measured the F _i, which is defined as PC layer length (red line) divided by cerebellar perimeter (blue line; Figure S1A). F _i has previously been used to assess cerebellar development, and alterations in this index are associated with abnormal development of PCs [50, 51, 52]. We indeed found aberrations in foliation within the cerebellar vermis, which suggests the foliation of the cerebellum is disrupted in SMA mice during development (Figure S1A–D; p = 0.003). These data confirm our previous finding of posterior vulnerability in the cerebellum and suggest that structural deformations are present in early symptomatic SMA mice.

FIGURE 2.

Posterior lobules in the cerebellar vermis are proportionally smaller in spinal muscular atrophy (SMA) mice compared to controls. (A and B) Representative images of control (CTL; n = 4, two technical replicates each mouse) and SMA (n = 5, two technical replicates for each mouse) cerebellar sagittal sections with the lobules labeled in the (A) vermis and (B) hemisphere at P4. Scale bars = 0.5 mm. (C) Total surface area of a sagittal cerebellar section. (D and E) Normalized lobule proportion, displayed as a percentage of its total slice area, for lobules in the cerebellar vermis (D) and hemisphere (E). ANCr1/2, ansiform crus I and II; PRM, paramedian. Unpaired two‐tailed Student's t‐test in (C), ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons in (D) and (E); **p < 0.01.

3.2. Early symptomatic SMA cerebella do not exhibit PC degeneration or reactive astrocytes

To determine whether the lobule‐selective PC degeneration we previously observed in late‐stage SMA mice is present at the early symptomatic stage, we stained cerebellar slices with the PC marker anti‐calbindin [53] (Figure 3A,B). Two blinded researchers manually counted PCs by lobule, summed, and averaged them for total slice numbers and found no significant difference between CTL and SMA (Figure 3C). We compared PC counts by lobule and found no significant differences in either the vermis or the hemisphere of the cerebellum between CTL and SMA mice (Figure 3E,G, respectively).

FIGURE 3.

Purkinje cell (PC) degeneration is not yet present in the cerebella of early symptomatic spinal muscular atrophy (SMA) mice. (A and B) Representative images of a sagittal section of the cerebellar vermis (A) and hemisphere (B) at P4, stained with anti‐calbindin. Respective lobules are labeled, and the arrow indicates the Purkinje cell layer. Scale bar = 0.5 mm. (C and D) Total PC counts (C) and PC perimeter density (D) across sagittal sections in the cerebellar vermis and hemisphere. (E and F) PC counts (E) and PC perimeter densities (F) by lobule in the hemisphere. (G and H) PC counts (G) and PC perimeter densities (H) by lobule in the vermis; control (CTL) n = 4, SMA n = 6, two technical replicates for each mouse. ANCr, ansiform crus; PRM, paramedian; SIM, simple. Ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons across SMA and control within each lobule of each cerebellar region.

Total PC layer lengths were not different in either the vermis or the hemisphere of SMA cerebella compared to CTL (not shown). By dividing PC counts by PC layer lengths, we calculated a 2D cell density, which we refer to as PC perimeter density. For whole sagittal sections, there was no significant difference in PC perimeter density between CTL and SMA (Figure 3D). We acquired the PC layer lengths within each lobule as well, and there was no significant difference found in any lobule between SMA and CTL mice (not shown). We then used these lobule length values to calculate the PC perimeter density in each lobule, and there were also no significant differences in PC perimeter density in the vermis or hemisphere lobules between SMA and CTL cerebella (Figure 3F,H). These results indicate that the lobule selective PC degeneration observed at the late stage, specifically in the ANCr2 and PRM lobules [47], is not yet present at the early symptomatic stage in SMA mice.

We previously identified PC degeneration accompanied by increased GFAP expression in astrocytes in the late‐stage cerebellum, so we also investigated GFAP immunoreactivity as a marker for astrocytic reactivity and gliosis [54, 55, 56, 57]. The total astrocytic reaction across entire sections of the vermis and hemisphere in SMA mice was not significantly different from CTLs (Figure S2A). We also observed similarity across SMA and CTLs in the DCN and the lobules, regardless of whether we were looking in the vermis or the hemisphere (Figure S2B–D). These results suggest that astrocytic activity is normal at the early symptomatic age and support the notion that glial and PC pathologies may be related to each other.

3.3. Functional abnormalities in early symptomatic SMA cerebella

SMA MNs demonstrate functional deficits prior to the onset of degeneration [9]. We have previously reported a reduction of spontaneous cerebellar activity in SMA mice at the late disease stage [48]. Given that cerebellar pathology is present, but not prominent at the early disease stage, we investigated the spontaneous spike and burst properties of the cerebellum using MEA recordings at P4 (Figure 4A,B). There were approximately 20% fewer channels occupied by SMA (n = 6) cerebella compared to CTLs (n = 6; Figure 4C), coinciding with the size disparities we observed in our MRI and IHC data.

FIGURE 4.

Early symptomatic spinal muscular atrophy (SMA) cerebella exhibit functional deficits. Representative multi‐electrode array images of sagittal slices and sample channel recordings are displayed in (A) control and (B) SMA cerebella at P4. Each channel is spaced 200 μm from the next. (C) Total number of channels contacting cerebellar sections. (D) Percentage of the total channels displaying activity. (E) Total number of spikes within a slice over the recording duration, in thousands. (F) Number of spikes per active channel, in thousands. (G) Percentage of slices that displayed bursting activity. (H) Average number of spikes within a single burst. N = 6, three to four technical replicates per mouse for both control and SMA groups. *p < 0.05, **p < 0.01, unpaired two‐tailed Student's t‐test in (C–F) and (H), Chi‐square test in (G). CTL, control.

Among the channels in contact with the cerebellum, the percentage of active channels that recorded at least five spikes during the 3‐min recording was similar between SMA and CTLs (Figure 4D). However, in SMA cerebella, the total number of spikes per slice was 79% less (Figure 4E) and the number of spikes per active channel was 75% less compared to CTLs (Figure 4F; p = 0.046). We also investigated neuronal bursting properties across the cerebella and found bursting activity in 50% of CTL slices and in 38% of SMA slices (Figure 4G; p = 0.06). The number of bursts and the burst durations were similar (not shown), but there were 20% fewer spikes within each burst in SMA cerebella compared to CTLs (Figure 4H; p = 0.012). These results suggest that functional abnormalities in the neuronal network precede PC degeneration in cerebella of early symptomatic SMA mice.

3.4. Decreased SMA mouse cerebellar volume is not restored by increased SMN in MNs

We next examined the cerebellum in late‐stage ChAT ^Cre+ rescue mice. It has been previously shown that increased SMN expression in MNs in these mice prevents MN loss and synaptic dysfunction, improves motor behavior, and increases survival [9, 25, 58, 59]. To assess whether rescue of MNs restores the cerebellar volumetric deficits that we reported in late‐stage SMA mice [47], we acquired T2‐MRI scans of brains in WT and SMA mice, as well as SMA mice with ChAT ^Cre+ and homozygous Cre‐inducible Smn allele (referred to hereafter as ChAT ^Cre+ rescue mice). Volumetric measurements were made from semi‐automatic segmentation of the whole mouse brain and the cerebellum for WT (n = 5), ChAT ^Cre+ rescue (n = 11) and SMA (n = 7) brains (Figure 5A). Confirming our previous results [47], we found that the whole brain and the cerebellum volumes were significantly smaller in SMA compared to CTL (Figure 5B,C; p < 0.0001). Furthermore, we again found that the relative cerebellar volume is significantly smaller in SMA compared to CTL as well (Figure 5D; p = 0.002). The cerebellar volumes of the ChAT ^Cre+ rescue mice were also significantly smaller compared to CTLs and not significantly different from SMA (Figure 5C; p < 0.0001 and p = 0.064, respectively). The relative cerebellar volume of ChAT ^Cre+ rescue mice was also not significantly different from SMA mice (Figure 5D; p = 0.54). These results suggest that ChAT ^Cre+ rescue does not restore volumetric deficits in the cerebellum. However, there is a slight restorative effect for the volume of the whole brain (Figure 5B; p = 0.006). This may be caused by the ChAT ^Cre+ expression within the cholinergic neurons of the brainstem and cortex [60].

FIGURE 5.

Defects in spinal muscular atrophy (SMA) cerebellar volume and proportional size are not spared in ChAT ^Cre+ rescue mice. (A) Representative 3D renderings of cerebella from wild type (WT) (left), ChAT ^Cre+ rescue (Res; middle), and SMA (right) mice at P12. (B and C) Volumetric measurements from magnetic resonance imaging segmentations of the (B) cerebella and (C) whole brains for wild type (WT), ChAT ^Cre+ rescue, and SMA mice. (D) Cerebellar volume ratio, represented as a percentage of whole brain volume. WT: n = 6, ChAT ^Cre+ rescue: n = 14, and SMA: n = 8. *p < 0.05, **p < 0.01, ****p < 0.001, ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons.

3.5. Increased SMN in MNs does not prevent structural deficits in the cerebellum of SMA mice

IHC staining of sagittal sections of the cerebellum was used to determine the effect of SMN rescue in MNs on structural pathologies in the cerebellum. To do this, two sagittal slices in each cerebellar region (hemisphere and vermis) were stained with anti‐calbindin, and counterstained with hematoxylin. We found that neither ChAT ^Cre+ rescue (n = 11) nor SMA (n = 5) mice have abnormal foliation in the cerebellar hemisphere compared to CTLs (n = 5; Figure S1E; p = 0.137 and p = 0.163, respectively). However, the F _i in the vermis is lower in all SMA mice, with or without SMN rescue in MNs. In fact, the decrease in F _i compared to WT within the vermis was larger for the ChAT ^Cre+ rescue mice than for the SMA mice (Figure S1F; p = 0.002 and p = 0.080, respectively). These data suggest that the distorted foliation in SMA mice is not prevented in ChAT ^Cre+ rescue mice.

We quantified the cerebellar structural defects further by measuring total and lobular surface area. We observed significant structural abnormalities across nearly all lobules of both the cerebellar vermis and hemisphere in SMA mice (Figure 6). The total surface area of the cerebellar vermis was significantly smaller for all SMA mice compared to WT, regardless of the rescue of MNs (Figure 6B). The lobule surface area of nearly all the lobules of vermis was significantly smaller in the SMA and ChAT ^Cre+ rescue mice compared to WT, except for lobule II. Lobule surface area deficits were not significantly ameliorated except for lobule VI in ChAT ^Cre+ rescue mice (Figure 6C). In the hemisphere, the total surface area was significantly smaller in SMA and ChAT ^Cre+ rescue mice compared to WT. However, there were significant differences between SMA and the rescue mice, indicating that there was a limited recovery of total surface area (Figure 6E). A closer look at each lobule within the hemisphere identified a finding similar to the vermis. The SIM, ANCr1, ANCr2, PRM, and copyramidal (COP) lobules were all significantly smaller in the SMA and ChAT ^Cre+ rescue mice compared to WT, but not the culmen 4/5 lobule (CUL4/5). There was no amelioration of SMA‐caused surface area deficits in any of the lobules in ChAT ^Cre+ rescue mice (Figure 6F). p Values are indicated by the figure legend.

FIGURE 6.

MN rescue in spinal muscular atrophy (SMA) mice has a minimal ameliorative effect on surface area in cerebellar lobules. Surface area analysis of sagittal sections within the cerebellar vermis (A–C) and hemisphere (D–F). Representative images of (A) vermis and (D) hemisphere sections at P12 stained with anti‐calbindin. Surface area measurements are compared across wild type (WT), ChAT ^Cre+ rescue (Rescue), and SMA for total surface area and for each lobule in the (B, C) vermis and (E, F) hemisphere, respectively. WT: n = 6, ChAT ^Cre+ rescue: n = 14, and SMA: n = 8, two technical replicates for each mouse. *p < 0.05, **p < 0.01, ***p < 0.001, ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons. ANCr1/2, ansiform crus I/II lobule; COP, copyramidal lobule; CUL4/5, culmen lobule; PRM, paramedian lobule; SIM, simple lobule.

3.6. The lobule selective relationship between PC degeneration and reactive astrocytes in SMA mice is not prevented by SMN rescue in MNs

A key finding in our previous study was that lobule‐specific PC degeneration colocalized with increased astrocytic reactivity [47]. We investigated whether MN rescue ameliorates this pathology by staining cerebellar sections with anti‐calbindin (Figure 7A,B) and anti‐GFAP (Figure 8A). In the vermis, the structure of lobules VI and VII stands out as particularly vulnerable to SMA pathology in both SMA (n = 5) and ChAT ^Cre+ rescue (n = 5) mice compared to WT (n = 10; Figure 6C). Next, to directly assess PC neurodegeneration, two blinded researchers manually counted the PCs within each lobule. In the posterior lobules of the vermis (lobules VI, VII, and VII), there were significantly fewer PCs in both SMA and ChAT ^Cre+ rescue mice compared to WT mice. Interestingly, there was a partial ameliorative effect in the rescue mice for PC count in lobule VI, resulting in a significantly more PCs in ChAT ^Cre+ rescue mice compared to SMA. These lobular counts were totaled in vermis slices, showing that SMA and ChAT ^Cre+ rescue mice had approximately 20% and 22% fewer PCs, respectively, compared to WT mice (not shown). The average PC counts were divided by PC layer length to get PC perimeter density (Figure 7C). This method can control for biological variability in lobule size across samples [61]. We got similar results as PC counts. Across the vermis of WT mice, there is a consistent PC density. However, there is a significant drop‐off in cell density within the posterior lobules VI and VII in ChAT ^Cre+ rescue and SMA mice (Figure 7C), which confirmed our previous report [47].

FIGURE 7.

Purkinje cell (PC) degeneration in the cerebellum is not ameliorated by survival motor neuron rescue in the motor neurons of spinal muscular atrophy (SMA) mice. (A and B) Representative images of sagittal cerebellar sections stained with anti‐calbindin in the (A) vermis and (B) hemisphere at P12. Examples of wild type (WT), ChAT ^Cre+ rescue (Rescue), and SMA cerebella are shown, with labeled lobules and arrows pointing to equivalent loci across genotypes. Scale bar = 0.5 mm. (C and D) PC perimeter density is shown for each lobule in the (C) vermis and (D) hemisphere. Wild type (WT): n = 6, ChAT ^Cre+ rescue: n = 14, and SMA: n= 8, two technical replicates for each mouse. *p < 0.05, **p < 0.01, ***p < 0.001, ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons. ANCr1/2, ansiform crus I/II lobule; COP, copyramidal lobule; CUL4/5, culmen lobule; PRM, paramedian lobule; SIM, simple lobule.

FIGURE 8.

Cerebellar astrocytic reaction is unaltered by motor neurons rescue. (A) Representative image of wild type (WT) and spinal muscular atrophy (SMA) cerebellar hemispheres at P12 stained using anti‐glial fibrillary acidic protein (GFAP), a marker for glial reactivity. Region of interest selections were made in the deep cerebellar nuclei (DCN; blue), the crown of a lobule shown to be unaffected (dark green) and affected (light green) in SMA cerebella, as well as the arbor vitae within lobules shown to be unaffected (dark purple) and affected (light purple) in SMA cerebella. Scale bar = 0.5 mm. (B, C, E, and F) Relative GFAP expression was acquired by dividing the pixel density in lobules affected by SMA by that of lobules typically unaffected by SMA. This was done in the cortex of the hemisphere (B) and vermis (C), and the arbor vitae of the hemisphere (E) and vermis (F). (D) The positive pixel percentage for the DCN is also shown. WT: n = 6, ChAT ^Cre+ rescue: n = 14 and SMA: n = 8, two technical replicates for each mouse. *p < 0.05, **p < 0.01, ***p < 0.001, ordinary one‐way analysis of variance with Tukey's correction for multiple comparisons.

We performed the same PC analysis in the cerebellar hemisphere and again found that the posterior lobules were significantly altered by SMA, regardless of MN rescue. In the posterior lobules of the hemisphere—ANCr1, ANCr2, PRM, and COP, there were significantly fewer PCs in the rescue and SMA mice as compared to WT. The total PC counts within the hemisphere slices of the ChAT ^Cre+ rescue mice and SMA mice showed approximately 15% and 25% fewer PCs, respectively, compared to WT (not shown). By calculating PC perimeter densities to control for variation in lobule size, we found a significantly lower PC density in the ChAT ^Cre+ rescue and SMA mice compared to WT mice for the four most posterior lobules: ANCr1, ANCr2, PRM, and COP (Figure 7D). There was only a restoration of PC numbers in the PRM lobule of the hemispheres in the ChAT ^Cre+ rescue mice (Figure 7D). p Values are indicated by the figure legend. These results support our previous findings of lobule selective structural defects and PC degeneration. Our data also suggest that MN rescue provides minimal restoration of PC degeneration in the SMA cerebellum.

Lastly, we assessed whether MN rescue affects the colocalization of the astrocytic reaction and PC degeneration in the late stage of SMA mice that was demonstrated in our previous report [47]. To do this, we quantified positive pixels within IHC slices in the hemisphere and vermis regions of the cerebellum that were stained with anti‐GFAP. Figure 8 shows our quantification of GFAP expression in the cortex (enclosed in green shapes) and arbor vitae (enclosed in purple shapes) of the cerebellar vermis and hemisphere, and the DCN (enclosed in blue shapes). For the cerebellar cortex, relative GFAP expression was calculated by dividing the percentage of positive pixels in the affected lobules (hemisphere: ANCr 2, vermis: lobule VI/VII) by the percentage of positive pixels in the unaffected lobules (hemisphere: SIM, vermis: lobule III). The classification of the lobules was determined by our previous findings and effectively controls for technical variability across slices. At the crown of the cortices within the hemisphere and the vermis, we found increased GFAP expression in SMA (n = 4) and ChAT ^Cre+ rescue (n = 10) mice compared to WT (n = 4; p < 0.05), with no difference between the two groups (Figure 8B,C; p > 0.20). We also quantified the relative glial activation in the arbor vitae, which was calculated by dividing the percentage of positive pixels in the arbor vitae of the affected lobules: in the hemisphere, the ANCr 2 and COP lobules, and in the vermis the SIM and CUL4/5 lobules. We observed significantly greater GFAP immunoreactivity in the hemisphere of SMA cerebella compared to WT (p = 0.025), but the difference was not significant for ChAT ^Cre+ rescue mice (Figure 8E; p > 0.19). In the vermis no significant differences were observed among CTL, SMA or ChAT ^Cre+ rescue mice (Figure 8F; p > 0.91).

For the DCN, we compared the positive pixel percentage among our mouse groups. SMA and ChAT ^Cre+ rescue mice had significantly higher GFAP immunoreactivity compared to WT mice (p = 0.003 and p = 0.010, respectively), but there was no difference between SMA and ChAT ^Cre+ rescue mice (Figure 8D; p = 0.501). These results confirmed our previous finding of lobule‐selective astrocytic reaction correlated with PC degeneration in the late stage of SMA mice and added that rescue of SMN in MNs does not ameliorate glial pathology in the cerebellum.

4. DISCUSSION

Our previous study in late‐stage SMA mice raised important questions about the cause of cerebellar defects and their contribution to the overall pathology of SMA. The data presented here describe functional and structural abnormalities in the cerebella of early symptomatic SMA mice prior to neuronal loss or astrocytic reactivity. Importantly, we also found that increasing SMN expression specifically in the MNs of late‐stage SMA mice has minimal ameliorating effects on cerebellar structural defects and neurodegeneration, suggesting that cerebellar defects arise independently of MN pathology and are a primary pathology in SMA mouse models.

4.1. Cerebellar abnormalities primarily contribute to SMA pathology

Spinal MNs are selectively vulnerable to SMN protein deficiency. Over the past two decades, the pathology of this cell type as well as its target skeletal muscle has been thoroughly characterized in human studies and various animal models [23, 30, 62, 63]. However, SMA is increasingly considered a multi‐system disease with a variety of complications throughout the body [27, 30, 31, 32, 33, 34, 35, 36]. As our understanding of SMA develops, it is important to elucidate the developmental sequence of other pathologies, whether they contribute to disease phenotypes, and how they relate to canonical MN degeneration. SMA pathologies in the cerebellum stand out for several reasons: the severity of the cerebellar pathologies in patients [43, 44, 45, 46], the high dependence on SMN protein function that is implied by its dense expression [40, 42], and the high level of connectivity of the cerebellum to the motor and sensorimotor circuits of the spinal cord. The proprioceptive signals necessary for motor refinement originate from muscle spindles and tendon organs, ascend to the spinal cord, and project to the cerebellum through inferior peduncles via pathways referred to as dorsal and ventral spinocerebellar tracts [37, 64].

We show here that, in the cerebellum of SMA mice, functional abnormalities precede structural and neuronal degeneration. At the early symptomatic age of P4, neural activities in the cerebellum are abnormal, while cerebellar size, structure, and PC integrity are comparable to CTLs. Interestingly, we also observed increased folding in the P4 SMA cerebellar vermis, which suggests that there are functional defects in PCs [50]. At this stage, we found that the total brain volume of SMA mice was comparable to CTLs, but cerebellar volume showed a minor volumetric deficit, with the most concentrated effect being observed in the vermis. Notably, the vermis is important in postural control [65], which is a compromised motor function in SMA patients [66, 67]. At P4, cerebellar glia also appeared normal in the SMA mice studied here. The lack of distinct neurodegeneration in the cerebellum at this stage, even while it is disproportionately smaller, combined with the functional deficiencies we observed, suggests that critical mechanisms during cerebellar development are affected when SMN protein levels are diminished. In our previous study on SMA mice at P7, days after the acceleration of SMA phenotypes [68, 69], we found that the same functional defects are evident. Additionally, there were also defects identified in the membrane properties and excitability of PCs, which progressively worsened through P11–12 [48]. At this late stage of SMA, the PCs and lobular structure of the mice studied here exhibit severe degeneration localized to posterior lobules. This was paired with a colocalized increase in GFAP expression that suggests increased astrocytic reactivity and gliosis. Our previous work has shown DCN neuronal function is also severely affected, displaying altered spontaneous firing and excitability [47]. DCN neuronal output activity is strongly modulated by direct inhibitory inputs from PCs [37], emphasizing their involvement in SMA pathology.

The progression of SMA pathologies in the cerebellum of SMA mice is sequentially and temporally analogous to that of MNs in the spinal cord and at the NMJ. Among these neurons and structures, functional deficits are a precursor to the degeneration of synapses and structures in SMA mice [70]. Recently published findings showed that at birth or shortly after, the most vulnerable MNs within lumbar segment (L) 1 of SMN∆7 mice exhibit abnormal spiking properties and increased excitability, but no signs of degeneration. At P4–5, when phenotype retrogression begins to accelerate, functional deficits worsen and coincide with a decrease in proprioceptive synapses as well as a significant decrease in MN number [9, 14, 68, 69]. At this age, a vulnerable subpopulation of NMJs also become significantly denervated, and by the late stage of P12–14, functional deficits and neurodegeneration in the spinal cord and at the NMJs are severe [9, 14, 15, 16]. Moreover, microglial activity surrounding MNs is increased at this late stage, but not at P7 [16, 71]. Our current results and those of our previous study show that the progression of glial response surrounding PCs in the cerebellum mirrors that of the spinal cord, which starts after P4 but before P12 in SMNΔ7 mice [47]. Other studies using the Smn ^−/− ;SMN2 mouse model of SMA have found a more accelerated progression of symptoms than SMN∆7 mice, but a similar sequence. At P1, NMJs are fully innervated in hindlimb muscles and there is no MN degeneration. But at the late stage of P5, there are disruptions in muscle innervation and axonal integrity indicating neuronal degeneration [17, 72]. The similarity across the progression of cerebellar and spinal cord pathologies suggests the former also makes a primary contribution to the phenotypes of SMA mouse models, rather than just being a downstream effect of the latter.

As shown here and in our previous study, neurodegeneration is not ubiquitous in the cerebellum of late‐stage SMA mice; posterior lobules exhibit the greatest degree of PC loss and structural deformities, while anterior and nodular lobules are unaffected at the late stage [47]. Interestingly, studies in SMA mice that differentiated MNs by their spinal cord segments have also found a pattern of selective vulnerability, where some segments have MNs that are more vulnerable to neurodegeneration and glial reactivity than others [17]. In L1, there are disruptions in functional properties and significant MN loss observed at P4, but not in L4–5 [9, 14]. Moreover, cervical segments of the spine are smaller but show no signs of degeneration until the late stage of the disease [73]. The selectively vulnerable nature of MNs is indicative of their differential dependence upon SMN protein and thus, their primary nature in SMA pathology. We show here and in our previous study that the cerebellum displays a similar pattern of selective vulnerability, suggesting a primary role in the pathology of SMA mouse models as well.

4.2. Cerebellar defects are independent of MN pathology

In order to isolate cerebellar pathology from MN pathology, we employed the ChAT ^Cre+ SMA mouse line, which mainly rescues SMN expression in spinal cord cholinergic MNs [25, 49, 74]. In contrast to the spinal motor system, the cerebellum has a much less prominent cholinergic system, and overall levels of choline acetyltransferase (ChAT) expression are much lower [75, 76, 77, 78, 79]. The predominant cholinergic fibers project to the molecular and granule cell layers of lobules IX and X, and to a lesser extent in the DCN. Notably, cholinergic inputs to lobules VI and VII, the most affected lobules in the vermis region, are minimal or nonexistent [75, 76, 77, 78, 79]. The only neurons in the cerebellum that exhibit intrinsic cholinergic activity are a few Golgi cells (<5% of the total cerebellar Golgi population) and some DCN neurons [80]. Thus, recovery of SMN levels via ChAT‐driven expression in the cerebellum is expected to be minimal, and the lack of phenotypic or pathological recovery that we observed in that region of the Cre+ mouse cerebellum suggests that the pathologies there arise independently of MNs. Indeed, almost none of the cerebellar defects investigated in this study were ameliorated in the ChAT ^Cre+ rescue mice.

For whole brain volumes, we found that SMA and ChAT ^Cre+ mice were both significantly smaller than WT; this is consistent with our previous results and with reports of smaller overall body mass and brain volumes in SMA mice [8, 68]. However, the whole brain volume was larger in ChAT ^Cre+ rescue mice compared to SMA mice, suggesting that SMN expression recovery throughout the cholinergic system of the brain may contribute to a restoration in volume [25, 49, 81]. In contrast, for the cerebellum, we found that the volume and volume ratios of ChAT ^Cre+ rescue mice were significantly smaller than WT, and not significantly different from SMA. This is consistent with the less prominent cholinergic system in the cerebellum, suggesting that the restoration of MN pathology in the spinal cord has no effect on the volumetric deficit in the SMA cerebellum. While it remains unclear how SMA causes a volumetric deficit in the cerebellum, these findings do confirm that the region is disproportionately affected by SMA and that this effect is independent from MN pathology.

Cerebellar F _i is a measure of the extent of cortical folding and will typically increase to counteract constrained volume by decreasing inter‐neuronal distances and supporting connectivity [50, 82, 83]. But this is not what we observed in the SMA and ChAT ^Cre+ mice cerebella, which both had lower foliation indices than WT and were not different from each other, suggesting that widespread impairments in cerebellar functionality were responsible for this effect [84, 85, 86, 87, 88, 89]. Furthermore, lower cerebellar foliation functionally translates to lower PC density [50]. Since having fewer neurons translates to decreased functional capacity in many brain regions [90], a loss in functional capability caused by the PC neurodegeneration likely imparts a causal effect on cerebellar foliation in SMA mice. PCs are inhibitory gamma‐aminobutyric acid (GABAergic) neurons with dense SMN expression [40, 41, 42]. Importantly, they are almost completely isolated from the cholinergic system in the central nervous system. They receive little to no cholinergic innervation [77], and moreover, the functional relationship between PCs and mossy fibers are unaffected by cholinergic antagonists [76, 78, 91]. We found that ChAT ^Cre+ rescue mice showed little to no reduction of lobule‐specific PC degeneration compared to SMA mice, with signs of marginal recovery within only a single lobule. The same is true for GFAP immunoreactivity surrounding PCs and in the rest of the cerebellum. Recent work on the microglia surrounding MNs in SMA mice and MNs derived from SMA stem cells found that they exhibited a heightened reactivity and inflammatory nature [71, 92, 93, 94, 95, 96, 97]. In types I–III SMA patients, increased astrocytic reactivity and gliosis are also observed surrounding spinal cord MNs as well [96, 98, 99, 100]. Since PC pathology and astrocytic reactivity are minimally and in no way ameliorated in ChAT ^Cre+ rescue, respectively, we conclude that these pathologies are independent from MN degeneration in the spinal cord.

4.3. From cerebellar pathologies to SMA phenotypes

Our data support the primary contribution of cerebellar defects in SMA mouse model pathology, but what is their contribution to SMA phenotypes? The cerebellum has a high level of connectivity to the motor and sensorimotor circuits of the spinal cord with functional implications in sensory input processing [37, 38, 65]. Degenerative ataxias of the cerebellum disrupt this function and often result in disruptions in gait, postural control, and locomotive timing [65]. These phenotypes are often experienced by SMA patients and are a primary effect of degeneration and dysfunction in spinal cord MNs [66, 67]. In mice, certain spinal segments, such as L1, are selectively vulnerable to SMA and have displayed degenerative morphology from P1 followed by MN loss and axonal loss at the NMJ of paraspinal and quadratus lumborum muscles [14]. These segments innervate the proximal muscles of the lower trunk, which show the most severe dysfunction and degeneration in SMNΔ7 mice [8, 14, 16, 17, 69, 101]. In contrast, other spinal segments do not display pathology until the late stage, if at all [9, 14, 17, 73]. Although the exact cause for the selective vulnerability of MNs remains unclear [101], one explanation for the vulnerability of these MNs could be the variation in gene expression profiles, such as full‐length and truncated SMN protein levels [102, 103, 104].

We show here that PCs within the posterior lobules of the cerebellum are selectively vulnerable to SMA in a very similar manner to what has been observed for spinal MNs. Interestingly, there are other conditions where both PCs and MNs are individually vulnerable. For example, two early autopsies of SMA patients identified severe cerebellar hypoplasia [105, 106], which is an SMA‐like disease where patients experience motor impairment caused by spinal cord MN degeneration [107]. Also, loss of function mutations in the gene for GEMIN5, a critical associating protein in the SMN complex, cause cerebellar ataxia and atrophy [108]. Like MNs, variations in PC gene expression profiles may be relevant to their selective vulnerability, which indeed differ across the regions and lobules of the cerebellum [109, 110, 111, 112]. In particular, PCs show transverse bands of gene expression across the mediolateral axis. For example, zebrin II is a metabolic enzyme that has been the most documented molecular marker in the cerebellum to date [113, 114], and there are bands of zebrin II+ and zebrin II− PCs across the cerebellum [109, 115, 116]. Moreover, there are a variety of non‐overlapping markers that co‐express in either zebrin II+ cells or zebrin II− cells [109]. Interestingly, there is a mediolateral strip of zebrin II+ cells that follows the same pattern as the PC degeneration that we observe in SMA cerebella, suggesting the potential for a functional relationship between SMN and zebrin II or the gene products of zebrin II+ co‐expressors [109].

The cerebellum modulates sensory inputs from motor systems in the body [37, 38, 65], and has complex functional relationships between regions of the cerebellum and the rest of the body. This has been demonstrated in rodents, connecting specific regions of the cerebellum to the CTL of facial muscles, limbs, and digits [111, 117]. Specific lobular somatotopy is not well documented in rodents, but in humans, the ANCr1, ANCr2, and PRM lobules of the cerebellar hemisphere map to eye, tongue, and hand function, while in the cerebellar vermis, lobules VI–X map to hand and foot function [118, 119]. Afferent fibers through the inferior peduncle project to granule cells mainly by way of mossy fibers and some climbing fibers [37]. Granule cell axons split into parallel fibers in the molecular layer and innervate PCs with glutamatergic synapses [64].

The specific impact that cerebellar pathology has on motor function is beyond this study, but other PC disorders may provide some insight. For example, spinocerebellar ataxias are a group of neurodegenerative diseases with a primary pathology of PC dysfunction and degeneration [120, 121, 122]. Clinical features vary depending on the subtype of spinocerebellar ataxia but typically consist of disrupted gait and incoordination, muscle fasciculations, difficulty speaking, and involuntary eye movements [123, 124]. Intermediate SMA (types II and III) patients experience grasping deficiencies, hand tremors, tongue fasciculations, and lower limb weakness—functions that are implicated by cerebellar somatotopy. They also experience pelvic weakness and abnormal gait and muscle fatigue—functions also similar to spinocerebellar ataxia [66, 67]. Our data and the studies described above suggest that cerebellar pathologies may contribute to the manual, lingual, gait, and voluntary gaze deficits observed in SMA patients. Moreover, it is very interesting that the lobules found to be most vulnerable in SMA mice from our study are the lobules that primarily contribute to non‐motor functionalities of the cerebellum, such as language, attention, and executive processes [125, 126]. This may explain why intellectual impairments and autism‐like phenotypes are emerging after SMN modifying treatments in SMA patients [127, 128].

5. CONCLUSIONS

SMA has been traditionally understood as a MN disease of the spinal cord. However, the multi‐system nature of SMA has become widely recognized in recent years, with defects being observed across the central and peripheral nervous systems. In our previous study, we reported widespread cerebellar pathology caused by SMA, including volumetric deficits, compromised structural integrity, and posterior PC degeneration colocalized with increased astrocytic reactivity. In consideration of recent findings indicating that SMA muscle tissue pathology is a primary pathology independent of MN degeneration in mouse models, we aim here to determine whether the same is true for the cerebellum.

This work builds on our previous findings and shows that cerebellar pathology follows a similar progression as spinal cord MN pathology. Furthermore, even in SMA mice where SMN expression was rescued in spinal cord MNs, we demonstrate that cerebellar pathology persists. Our findings strongly suggest that cerebellar vulnerability is a primary pathology contributing to symptoms in SMA mouse models, independent from MN pathology. The cerebellum has become a critical region for study in SMA research. Clinically, it is likely that cerebellar pathology contributes to the phenotypes seen in SMA patients, such as gait instability, tongue fasciculations, hand and digit motion, and muscle weakness. This is particularly important considering that the levels of SMN expression remain low in the cerebellum after treatment with either onasemnogene abeparvovec or nusinersen [41, 129]. One study found that risdiplam induced increased SMN expression in the brain, but the study did not isolate the cerebellum [130]. Low SMN recovery in the cerebellum may indeed contribute to why some SMA patients do not respond to those treatments. Future studies should focus on the molecular changes related to low SMN expression in cerebellar PCs and how this brain region could be addressed in the development of SMA treatments. This study contributes to the effort to complete the characterization of SMA pathology in the hopes that SMA treatments and therapy can become more comprehensive, and a cure can be developed.

AUTHOR CONTRIBUTIONS

Conception and design of the study: JS, MAH, NCC. Acquisition and analysis of data: NCC, JS, CJS, LK, MHC ‐C, CJC, MD, NN, CG. Drafting a significant portion of the manuscript or figures: NCC, JS, MAH. All authors read and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

Charlotte J. Sumner has or currently receives grant support from Roche Ltd., Biogen, and Actio Biosciences and has served as a paid advisor, consultant, and/or speaker to Biogen, Roche/Genentech, and Novartis. These arrangements have been reviewed and approved by Johns Hopkins University in accordance with its conflict‐of‐interest policies.

ETHICS STATEMENT

All animal procedures were approved by the Institutional Animal Care and Use Committee of Delaware State University.

Supporting information

Data S1. Supporting Information.

Cottam NC, Dowling M, Kong L, Chan‐Cortés MH, Charvet CJ, Norzeron N, et al. Cerebellar defects are a primary pathology in mouse models of spinal muscular atrophy. Brain Pathology. 2025;35(6):e70025. 10.1111/bpa.70025

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.