Cerebellar axonopathy in Shivers horses identified by spatial transcriptomic and proteomic analyses

Abstract

Background

Shivers in horses is characterized by abnormal hindlimb movement when walking backward and is proposed to be caused by a Purkinje cell (PC) axonopathy based on histopathology.

Objectives

Define region‐specific differences in gene expression within the lateral cerebellar hemisphere and compare cerebellar protein expression between Shivers horses and controls.

Animals

Case‐control study of 5 Shivers and 4 control geldings ≥16.2 hands in height.

Methods

Using spatial transcriptomics, gene expression was compared between Shivers and control horses in PC soma and lateral cerebellar hemisphere white matter, consisting primarily of axons. Tandem‐mass‐tag (TMT‐11) proteomic analysis was performed on lateral cerebellar hemisphere homogenates.

Results

Differences in gene expression between Shivers and control horses were evident in principal component analysis of axon‐containing white matter but not PC soma. In white matter, there were 455/1846 differentially expressed genes (DEG; 350 ↓DEG, 105 ↑DEG) between Shivers and controls, with significant gene set enrichment of the Toll‐Like Receptor 4 (TLR4) cascade, highlighting neuroinflammation. There were 50/936 differentially expressed proteins (DEP). The 27 ↓DEP highlighted loss of axonal proteins including intermediate filaments (5), myelin (3), cytoskeleton (2), neurite outgrowth (2), and Na/K ATPase (1). The 23 ↑DEP were involved in the extracellular matrix (7), cytoskeleton (7), redox balance (2), neurite outgrowth (1), signal transduction (1), and others.

Conclusion and Clinical Importance

Our findings support axonal degeneration as a characteristic feature of Shivers. Combined with histopathology, these findings are consistent with the known distinctive response of PC to injury where axonal changes occur without a substantial impact on PC soma.

Abbreviations

PC

Purkinje cells

1. INTRODUCTION

Shivers is a neuromuscular disorder of horses characterized by trembling, or shivering of the tail and thigh muscles and a characteristic hindlimb posture initiated by specific movements. ¹ , ² , ³ By 7 years of age, walking backward is difficult for Shivers horses because of either fixed hyperflexed abducted hindlimbs or rigid hindlimb extension. ² , ⁴ In addition, when a handler attempts to pick up a hindlimb, Shivers horses develop a hyperflexed abducted hindlimb posture that impairs cleaning and trimming of their hooves. ¹ , ² Interestingly, forward gaits such as walk, trot, and canter are normal. Over time, 50% of horses with Shivers slowly progress to a point at which they intermittently exhibit hindlimb hyperflexion during forward walking. ² , ³

An immunohistochemical investigation of the central and peripheral nervous system of Warmblood horses with Shivers identified selective degeneration of Purkinje cell (PC) axon projections in the lateral deep cerebellar nuclei, typified by an 80‐fold increase in calbindin and glutamic acid decarboxylase‐positive spheroids. ⁵ No other neuropathological lesions specific to horses with Shivers were identified in that study, suggesting that Shivers was caused by a cerebellar PC axonopathy. ⁵ The cerebellum coordinates locomotor behavior by integrating sensory information from the spinal cord and feedback from the motor cortex, basal ganglia, and thalamus and sending efferent inhibitory output to motor pathways via PC. Walking backward, compared with walking forward, is more specifically influenced by cerebellar activities because the cerebellum exerts anticipatory postural adjustments in the absence of the ability to visually monitor steps. ⁶ , ⁷ Consistent with the cerebellum's function in motor coordination, loss of the temporal precision of muscle firing and an increased level of motor unit recruitment was identified during backward walking in hindlimb muscles of horses with Shivers by surface electromyography (sEMG). ⁸ Thus, there is physiological and histopathological support for a cerebellar origin for Shivers. However, a cerebellar origin for Shivers has been questioned based on the lack of cerebellar signs such as proprioceptive deficits, truncal sway, and intention tremor. ⁴ , ⁵ , ⁸ , ⁹ , ¹⁰ Additional molecular genomic approaches could provide needed information to further identify a cerebellar origin for Shivers.

We hypothesized that the cerebellar white matter, and potentially PC soma, of horses with Shivers would have a distinct pattern of gene expression compared with control horses. To investigate this hypothesis, we leveraged spatial transcriptomics to define region‐specific gene expression differences in the lateral cerebellum. At the protein level, we hypothesized that the lateral cerebellar hemispheres of horses with Shivers would have a distinct pattern of protein expression.

2. MATERIALS AND METHODS

2.1. Horses

Five geldings >16.3 hands tall, of Belgian draft (3), Thoroughbred (1), and Appendix Quarter Horse (1) breeds and ranging in age from 7 to 13 years (median, 10 years) were donated to the University because of Shivers (Table 1). Horses exhibited consistent hyperflexion of both hindlimbs when walking backwards, reluctance to manually lift a hindlimb, hyperflexion once manually lifted, and no gait abnormalities at a walk or trot. Neurologic examination that included evaluation of demeanor, cranial nerves, muscle mass and symmetry, and proprioception was normal in all horses apart from the hindlimb movement disorder and symmetric muscle atrophy in 1 Belgian.

TABLE 1.

The breed and age of the geldings included in the study.

Horse	Phenotype	Breed	Age	Proteomics	Transcriptomics	Calbindin IHC
S1	Shivers	Belgian draft	7	✓		✓
S2	Shivers	Belgian draft	12	✓		✓
S3	Shivers	Belgian draft	13	✓
S4	Shivers	Thoroughbred	10	✓	✓	✓
S5	Shivers	Thoroughbred × QH	9	✓	✓	✓
C1	Control	Thoroughbred	20	✓	✓	✓
C2	Control	Thoroughbred	6	✓	✓
C3	Control	Irish Sport Horse	8	✓		✓
C4	Control	Thoroughbred	10	✓

Note: Checkmarks indicate which horses that were included in the various analyses.

Abbreviations: IHC, immunohistochemistry; QH, Quarter horse.

Control horses consisted of 4 geldings (≥16.2 hands tall, 3 Thoroughbreds, and 1 Irish Sport Horse), ranging in age from 6 to 20 years (median, 9 years; Table 1). Control horses were donated to the University because of persistent lameness (2), a sinus tumor (1), or chronic back pain (1). They showed no evidence of neurologic disease, had normal forward and backward walking, and willingly lifted the hind hooves.

Horses were euthanized by IV administration of a barbiturate with informed client consent. The study was approved by the Institutional Animal Care and Use Committees of the University of Minnesota 1310‐30989A and Michigan State University PROTO201900038.

2.2. Tissue collection

The brain was removed within 2 hours of euthanasia and an approximately 2 $\times$ 4 $\times$ 4 cm section of the lateral cerebellar hemisphere was removed, immediately frozen in liquid nitrogen, and stored at −80°C until further analysis (Figure 1).

FIGURE 1.

Image of the cerebellum of a control horse. (A) Before removing samples of the lateral cerebellar hemispheres for analysis. (B) After removal of tissue from both lateral cerebellar hemispheres. A portion of the removed sample is shown by the arrow. (C) Fixed cerebellar hemisphere with the vermis removed and in the process of sectioning to reveal the left deep cerebellar nuclei. For more detail see Data S1. (D) The section containing the deep cerebellar nuclei (white arrow) that was removed for immunohistochemical evaluation.

The remaining cerebellum was placed in 10% neutral buffered formalin. After 14 days, the region containing the deep cerebellar nuclei was isolated (Figure 1C,D, Data S1). Calbindin immunohistochemical staining was performed on 4 horses with Shivers and 2 control horses as previously described using Swant 300 IgG1 calbindin antibody (Burgdorf CH). ⁵ This antibody became unavailable after 2020 and subsequent immunohistochemical (IHC) staining of the cerebellum was attempted using 2 additional calbindin antibodies (calbindin EP3478 Abcam, Cambridge, UK; calbindin D‐4, Santa Cruz, CA).

2.3. Spatial transcriptomic analysis

2.3.1. Sample preparation

A portion of the frozen lateral cerebellar hemisphere was excised from the 2 horses with Shivers and 2 control horses with Thoroughbred bloodlines. The frozen tissue samples were placed in waterproof containers in a −20°C freezer overnight. The next day, the samples were fixed in 10% neutral buffered formalin. After 4 days at room temperature in formalin, samples were paraffin embedded. Sections 4‐μm thick from paired Shivers and control horses were mounted side‐by‐side (1 Shivers, 1 control horse per slide) on positively charged slides using a RNA‐free protocol, and then air dried at room temperature overnight. Slides were placed in a slide box with a desiccant and shipped overnight to NanoString (Seattle, WA).

Sample preparation was performed as described in the NanoString GeoMx RNA‐next generation sequencing (NGS) slide preparation manuals and in a previous study ¹¹ on a Leica Bond RX or RXm automated stainer (Leica Biosystems, Cincinnati, OH). Slides were baked, deparaffinized, washed in ethanol, and washed in phosphate‐buffered saline (PBS) or Leica BondWash Solution. Targets were retrieved in Tris‐EDTA pH 9.0 in a Leica BOND Epitope Retrieval Solution for 20 minutes at 100°C and washed in PBS or BondWash Solution. Samples were digested using 0.1 μg/mL Proteinase K for 15 minutes at 37°C and washed with PBS. All samples were incubated overnight at 37°C with human whole transcriptome atlases (WTA) following the NanoString GeoMx RNA‐NGS slide preparation manual at a probe concentration of 4 nM per probe in ×2 saline/sodium citrate buffer (SSC) with 2.5% dextran sulfate, 0.2% bovine serum albumin (BSA), 100 μg/mL salmon sperm DNA, and 40% formamide. During incubation, slides were covered with HybriSlip hybridization covers (Grace BioLabs, Bend OR). After incubation, coverslips were removed by soaking in ×2 SCC buffer and 0.1% Tween‐20. Two 25‐minute stringent washes were performed in 50% formamide in ×2 SSC at 37°C to remove unbound probes, and samples were washed in ×2 SSC. For antibody morphology marker staining, samples were incubated in blocking buffer for 30 minutes at room temperature in a humidity chamber, and then incubated with 500 nm SYTO13 (Thermofisher S11364, Ann Arbor MI) and fluorescently conjugated antibodies for PC for 1 to 2 hours as well as overnight at 1:50 dilution using 3 different antibodies (calbindin EP3478 Abcam, calbindin D‐4 Santa Cruz, and Parvalbumin NB120‐11427). Samples were washed in ×2 SSC and loaded on the GeoMx digital spacial profiling (DSP) instrument.

The DSP experiments were performed according to the NanoString GeoMx‐NGS DSP Instrument manual and as previously described. ¹² Briefly, slides were imaged in 4 fluorescence channels (calbindin EP3478/525 nm, Styo83/532 nm, calbindin D‐4/594 nm, and parvalbumin/647 nm) to visualize PC and white matter. Because the 3 immunofluorescent markers failed to specifically identify PC, PC soma were identified in the PC layer based on their distinctive morphology. In 3 separate areas per horse, the soma of each PC was manually encircled as a region of interest (ROI; Figure 2A,B). In 3 separate areas per horse, white matter below the granule cell was manually outlined (Figure 2C). The ROIs were illuminated and released tags were collected into 96‐well plates as previously described. ¹¹ , ¹²

FIGURE 2.

Immunofluorescent images of the lateral cerebellar hemisphere of Horse S5. The molecular layer stains magenta, granule cell layer green and white matter pink. (A) Manually encircled Purkinje cells soma (white) representing 1 of 3 regions of interest (ROI) analyzed per horse in the spatial transcriptomic analysis. (B) Purkinje cells (arrows). (C) Two ROIs (white) in the cerebellar white matter. Three ROIs were evaluated per horse.

2.3.2. Transcriptomic analysis

Library preparation was performed according to the NanoString GeoMx‐NGS Readout Library Prep manual (NanoString, Seattle, WA). Briefly, the DSP aspirate was dried and resuspended in 10 μL diethyl pyrocarbonate (DEPC)‐treated water, and 4 μL was used in a PCR reaction. NanoString SeqCode primers were used to amplify the tags and add Illumina adaptor sequences and sample demultiplexing barcodes. The PCR products were pooled either in equal volumes or in proportion relative to ROI size, depending on the experiment, and purified with 2 rounds of AMPure XP beads (Beckman Coulter, Indianapolis, IN). Libraries were sequenced on an Illumina (San Diego, CA) NextSeq 550, NextSeq 2000, or NovaSeq 6000 according to the manufacturer's instructions, with at least 27 × 27 paired end reads. The FASTQ files were processed using the NanoString GeoMx NGS Pipeline v2.3.4 Briefly, reads were trimmed to remove low quality bases and adapter sequences. Paired end reads were stitched and aligned, and the barcode and sequences with unique molecular identifiers (UMI) were extracted. Barcodes were matched to known probe barcodes with a maximum of 1 mismatch allowed. Reads matching the same barcode were deduplicated by UMI. Count data were processed and normalized using either the NanoString DSPDA software v2.2 or v2.3, the GeoMxTools R package v1.0 (https://bioconductor.org/packages/release/bioc/html/GeomxTools.html) or an equivalent development version. Sequencing quality by area was inspected to ensure sufficient saturation and high sensitivity of low expressors. Data then was normalized to the third quartile to account for differences in cellularity and ROI size. All statistical analyses and data visualizations were performed in R (R Core Team 2021) or using the DSPDA software v2.3. Differential expression was performed using a linear mixed effect model with slide and DSP instrument as random effect variables, and P‐values were corrected for multiple hypothesis testing using a false discovery rate (FDR) of P < .01. Gene set enrichment analysis (GSEA) was performed using the GSVA R package. ¹³

2.4. Proteomic analysis

2.4.1. Sample preparation and liquid chromatography mass spectrometry (LC/MS/MS)

Protein isolation was performed on frozen cerebellar tissue samples of all horses utilizing radioimmunoprecipitation assay lysis buffer (Thermo Scientific, Waltham, MA) with protease inhibitor (Roche cOmplete, Mini, EDTA‐free, Thomas Scientific, Swedesboro, NJ). Protein concentration was measured by standard bicinchoninic acid assay (Pierce Biotechnology, Rockford, IL) and Coomassie‐stained sodium dodecyl‐sulfate gel.

In brief, 120 μg of protein of each sample was subjected to proteolytic digestion using Trypsin/LysC enzyme mix (Promega, Madison, WI) at 1:100 (enzyme: protein) by volume. After enzymatic digestion, the samples were incubated with agitation. The samples then were acidified (2% trifluoroacetic acid), purified with c18 SepPaks (Waters, www.waters.com) and dried by vacuum centrifugation.

One hundred micrograms of each sample was resuspended in 100 μL of 100 mM triethylamonium bicarbonate. The peptides then were tagged with TMT11 or TMT6 reagents (Thermo Scientific, Waltham, MA) per manufacturer protocol. Labeled peptides were mixed in equal portions and reverse phase C18 SepPaks were used to de‐salt the combined sample.

One control sample was run in duplicate as an internal assay control. Tagged peptides were resuspended, washed, and eluted with the Thermo Acclaim PepMap RSLC 0.1 mm × 20 mm C18 trapping column over 125 minutes at a constant flow rate (300 nL/min). The resulting eluted peptides were sprayed into a ThermoScientific Q‐Exactive HF‐X mass spectrometer (Thermo Scientific, Waltham, MA) using a FlexSpray spray ion source. The top 15 ions in each survey scan (Orbi trap 120 000 resolution at m/z 200) were subjected to higher energy collision induced dissociation with fragment spectra acquired at 45000 resolution. The resulting MS/MS spectra were processed using Proteome Discoverer v2.2 (Thermo Scientific, Waltham, MA) to generate peak lists. Peak lists were searched against the EquCab3.0 UniProt:UP000002281 protein database appended with common laboratory contaminants (cRAP project) using Mascot v2.6 (Matrix Science, London, UK; version Mascot in Proteome Discoverer 2.2.0.388). The output then was analyzed using Scaffold, v4.8.7 (www.proteomesoftware.com) to probabilistically validate protein identifications with 1% false discovery rate confidence considered true. Mass spectrometry proteomic data are available at the ProteomeXchange with identifier PXD042370.

2.4.2. Quantitative data analysis

Scaffold Q+ (v5.0.1; Proteome Software Inc., Portland, OR) was used to quantitate TMT‐11 plex‐labeled peptide and to probabilistically validate protein identifications. Peptide identifications were accepted if they could be established at >10.0% probability to achieve an FDR <0.1%. Probabilities generated by Mascot were assigned by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at >99% probability, as assigned by the Prophet algorithm, ¹⁴ and contained at least 2 identified peptides. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters. Channels underwent matrix correction as reported by i‐Tracker. ¹⁵

Normalization was performed both across samples and spectra as previously reported. ¹⁶ Spectra data were log‐transformed, pruned of those matched to multiple proteins, and weighted by an adaptive intensity weighting algorithm. Differentially expressed proteins (DEP) between controls and horses with Shivers horses were determined using a permutation test and corrected using the Benjamini‐Hochberg procedure (P < .002).

3. RESULTS

3.1. Immunohistochemistry

Calbindin‐positive spheroids in distal PC axons were identified in the deep cerebellar nuclei of the 4 horses with Shivers examined and were absent in 2 control horses examined using the Swant 300 calbindin antibody previously validated (Figure 3). ⁵ The deep cerebellar nuclei of the remaining Shivers (1) and control (2) horses could not be evaluated because the Swant 300 calbindin antibody was no longer available and the newer calbindin antibodies failed to identify PC.

FIGURE 3.

The deep cerebellar nuclei of a horse with Shivers (S5) used for spatial transcriptomic and proteomic analyses stained with calbindin (Swant 300). Calbindin positive spheroids (arrows) are present. N, neuronal cell bodies.

3.2. Spatial transcriptomics

There were 18 676 total targets with 9468 genes normalized in the 3rd quartile and expressed above the limit of quantitation in at least 5% of ROIs. Assay sensitivity was assessed across total targets and 1846 genes were expressed in 50% of ROIs.

3.2.1. Principle component analysis (PCA) plots

The first principal component (19.1% variance) separated samples by region (PC soma vs white matter ROIs; Figure 4). The second principal component (9.5% variance) separated white matter ROIs of horses with Shivers from control horse ROIs (Figure 4).

FIGURE 4.

Principal component (PCA) analysis of transcriptomic data for white matter (labeled axons) and PC soma for 2 Shivers and 2 control horses. Each symbol represents 1 ROI, where 3 ROI were examined per horse for white matter and 3 for PC soma.

3.2.2. Differential gene expression

In the cerebellar white matter, 455 significant DEG genes were found in Shivers vs control horses, with 350 downregulated and 105 upregulated (Table S1; Figure 5). There were no DEG between Shivers and control horses in PC soma (FDR P < .01).

FIGURE 5.

Volcano plot depicting the probability of observing the estimated change in gene expression (−log₁₀ scale) on the y axis and the degree of fold change differences (log₂ scale) on the x axis. Cerebellar white matter analysis identified 455 DEGs (FDR <0.01; 350 downregulated in dark blue and 105 upregulated in light yellow) between Shivers and control horses.

Gene set enrichment analysis identified 4 significant pathways (adjusted P < .05) that were altered in the cerebellar white matter regions between Shivers and control horses. The pathways and their normalized enrichment scores (NES) included Toll‐like receptor 4 (TLR4) cascade (NES, 1.65; Padj = .05), Toll‐like receptor (TLR) cascades (NES, 1.61; Padj = .05), SARS‐CoV2 (NES, 1.60; Padj = .001), and SARS‐COV infection (NES, 1.58; Padj = .01).

3.3. Proteomic analysis

Of 12 854 spectra in the experiment at the given thresholds, 11 184 (87%) were included in quantitation. If a protein was not expressed in all samples, it was not used for downstream statistical analyses. This left 936 identified proteins expressed in all horses.

There were 50 DEP in Shivers compared with control horses: 27 had decreased (↓) expression (Table 2) and 23 had increased (↑) expression (Table 3). The magnitude of change in protein expression was relatively small, with a range of log₂‐fold change (FC) of −0.04 to 0.36. Numerous proteins common to axons had decreased expression in horses with Shivers, including the 5 Intermediate filaments – light chain neurofilament (NEFL), medium chain neurofilament (NEFM), heavy chain neurofilament (NEFH), α internexin (INA), and glial fibrillary acidic protein (GFAP). Additional proteins common to axons included 3 myelin‐related proteins, myelin basic protein (MBP), 2′,3′‐cyclic‐nucleotide 3′‐phosphodiesterase (CNP) and breast carcinoma‐amplified sequence (BCAS1), 2 cytoskeletal proteins, reticulon‐4 (RTN4) and microtubule‐associated protein 1A (MAP1A), and the Na/K ATPase (ATP1B1; Table 2). ¹⁷

TABLE 2.

The gene identification, protein, name log₂ fold change (FC) and adjusted P value for significantly differentially expressed downregulated proteins in the lateral cerebellar hemispheres of Shivers vs control horses.

Gene ID	Protein	Log2 FC	AdjP < value
Intermediate filaments
NEFL	Neurofilament light polypeptide	−0.31	<.0001
NEFM	Neurofilament medium polypeptide	−0.28	<.0001
NEFH	Neurofilament heavy polypeptide isoform X1	−0.21	<.0001
INA	Alpha‐internexin	−0.16	<.0001
GFAP	Glial fibrillary acidic protein	−0.13	<.0001
Microtubules
RTN4	Reticulon‐4 isoform X1	−0.08	<.0001
MAP1A	Microtubule‐associated protein 1A isoform X1	−0.04	.001
Myelin
MBP	Myelin basic protein isoform X3 [Equus caballus]	−0.33	<.0001
CNP	2′,3′‐cyclic‐nucleotide 3′‐phosphodiesterase isoform X1	−0.28	<.0001
BCAS1	Breast carcinoma‐amplified sequence 1 isoform X4	−0.23	.0004
Calcium signaling
CALB2	Calbindin 2	−0.36	.0001
PCP4	Calmodulin regulator protein	−0.17	.001
PVALB	Parvalbumin alpha isoform X1	−0.12	.0001
Mitochondria
SIR2	NAD‐dependent protein deacetylase sirtuin‐2 isoform X1	−0.18	.001
ES1	ES1 protein homolog, mitochondrial isoform X1	−0.16	<.0001
ATP5IF1	ATPase inhibitor, mitochondrial isoform X1	−0.15	.001
COX5B	Cytochrome c oxidase subunit 5B, mitochondrial	−0.1	.001
HSPD1	60 kDa heat shock protein, mitochondrial	−0.07	<.0001
ACO2	Aconitate hydratase, mitochondrial	−0.07	<.0001
HSPA9	Stress‐70 protein, mitochondrial	−0.06	.002
Cytosolic energy metabolism
MDH1	Malate dehydrogenase, cytoplasmic	−0.17	.001
ALDOCC	Fructose‐bisphosphate aldolase C	−0.08	<.0001
CKB	Cluster of creatine kinase B‐type	−0.08	.002
Neurite outgrowth
CMPK1	UMP‐CMP kinase	−0.16	<.0001
GPRIN1	G protein‐regulated inducer of neurite outgrowth 1	−0.12	.0002
Ion channel
ATP1B1	Sodium/potassium‐transporting ATPase subunit beta‐1	−0.15	<.0001
Protein synthesis
RPS19	40S ribosomal protein S19	−0.09	.0008

TABLE 3.

The gene identification, protein, name log₂ fold change (FC) and adjusted P value for significantly differentially expressed upregulated proteins in the lateral cerebellar hemispheres of Shivers vs control horses.

Gene ID	Proteins	Log2 FC	AdjP value
Cytoskeleton
VCL	Vinculin isoform X1	0.26	<.0001
VIM	Cluster of vimentin	0.36	<.0001
DES	Desmin	0.29	<.0001
PFN1	Profilin‐1	0.21	.001
FLNA	Cluster of filamin‐A isoform X1	0.29	<.0001
CALD1	Caldesmon isoform X2	0.26	.002
TAGLN	Transgelin	0.3	<.0001
Extracellular matrix
COL6A2	Collagen alpha‐2(VI) chain isoform X1	0.12	.002
COL6A3	Collagen alpha‐3(VI) chain isoform X3	0.27	<.0001
OGN	Mimecan isoform X1	0.23	<.0001
DCN	Decorin precursor	0.23	.0005
BGN	Biglycan precursor	0.33	.0001
PRELP	Prolargin	0.3	<.0001
TLN1	Talin‐1 isoform X3	0.22	.0001
Redox balance
PGD	6‐phosphogluconate dehydrogenase, decarboxylating	0.09	.001
TXNDC17	Thioredoxin domain‐containing protein 17	0.18	.002
Neurite outgrowth
AHNAK	Neuroblast differentiation‐associated protein AHNAK isoform X1	0.21	.0004
Signal transduction
ANXA2	Annexin A2	0.32	<.0001
Lipid metabolism
APOB	Apolipoprotein A‐I	0.2	<.0001
Protein synthesis
EET2	Elongation factor 2	0.09	.0005
Blood‐borne
HBB	Cluster of hemoglobin subunit beta	0.36	.002
ALB	Serum albumin precursor	0.23	<.0001
A2M	Alpha‐2‐macroglobulin	0.22	<.0001

Additional proteins with decreased expression in Shivers vs control horses included 3 calcium signaling or buffering proteins: calmodulin (CALM2), calretinin (CALB2), and parvalbumin (PVAL; Table 2). There were 7 significant ↓DEP mitochondrial proteins, including 2 heat shock proteins (HSPD1, HSPA9), sirtuin (SIR2, which orchestrates mitochondrial metabolism and adaptations to stress ¹⁸ ), a cytochrome c oxidase subunit (COX5B), an ATPase inhibitor (ATP51F1), a tricarboxylic acid (TCA) cycle protein (ACO2) and ES1 protein homolog, a glutamine amidotransferase that contributes to the stable functionality of mitochondria. ¹⁹ Three cytosolic energy metabolic proteins malate dehydrogenase, (MDH1), fructose‐bisphosphate aldolase C (ALDOCC) and creatine kinase B‐type (CKB), 2 neurite outgrowth factors UMP‐CMP kinase (CMPK1) and G protein‐regulated inducer of neurite outgrowth 1 (GPRIN1) as well as 1 ribosomal protein (RPS19) also exhibited significant ↓DEP in Shivers vs control horses (Table 2).

Among the 23 proteins with significantly increased DEP were 7 extracellular matrix proteins including 4 proteoglycans mimican (OGN), decorin precursor (DCN), biglycan (BGN), profilin (PNF1), 2 collagen proteins collagen alpha‐2 (VI; COL6A2), collagen alpha‐3 (VI; COL6A3), 1 anchoring protein (PRELP), and 1 cytoskeletal linker protein talin‐1 (TLN1; Table 3). Seven cytoskeletal proteins vinculin (VCL), desmin (DES), vimentin (VIM), PFN1, filamin A (FLNA), caldesmin (CALD1), taglin (TAGLN), and 2 redox proteins 6‐phosphogluconate dehydrogenase, decarboxylating (PGD), and thioredoxin domain‐containing protein 17 (TXNDC17) had significantly increased expression (Table 3).

4. DISCUSSION

Our study showed that, within the lateral cerebellar hemispheres, both spatial gene and protein expression differ significantly between Shivers and control horses. Spatial transcriptomic analysis of 2 Shivers and 2 control horses localized differences to the white matter of the lateral cerebellar hemispheres, with significant DEG found in white matter containing axons but not PC soma. Upregulation of the TLR4 cascade, which is involved in the induction of neuroinflammation in neurodegenerative diseases ²⁰ was evident in gene set enrichment analysis of cerebellar white matter. Proteomic analysis highlighted a loss of proteins associated with axonal loss or damage in horses with Shivers, as well as increased expression of proteins involved in axonal degeneration and regeneration. Histopathologic findings of a PC axonopathy evident in the deep cerebellar nuclei in our study and previous studies suggest that axonal damage likely is associated with PC axons. ⁵ , ⁸ Our findings are consistent with the known distinctive response of PC to injury in which axonal changes occur without substantial impact on PC soma. ²¹ , ²²

In spatial transcriptomics, there were over 450 DEG in Shivers vs control horses within axon‐rich white matter, supporting cerebellar axons as the site of molecular pathology in Shivers. Although our study and 2 previous histopathologic studies identified lesions in distal PC axons in horses with Shivers, no significant DEG were found in PC soma between Shivers and control horses. ⁵ , ⁸ The usual neuronal cell body response to axon injury consists of initial reactive or compensatory attempts that eventually become regressive leading to atrophy or cell death. ²³ Notably, in PC, axonal injury results in the development of so‐called torpedoes (previously identified in Shivers horses), and hypertrophy of both the initial neuritic segment and recurrent collateral branches, producing arciform fibers. ⁵ , ²¹ In essence, the PC soma is transformed into a local interneuron drawing trophic support from the cortical milieu and conveying its output information through nearby uninjured neurons insuring survival and functionality. ²¹ This novel PC reaction to injury could explain the lack of DEG in PC soma in horses with Shivers despite axonal damage. The resistance of the PC soma to injury and the focal nature of PC axonal damage also could account for the lack of diffuse signs of ataxia and tremor in horses with Shivers. PC axon terminals are restricted to a precise terminal domain, reflecting the topographical arrangement of the cerebellar network. ²⁴

Activation of a neuroinflammatory cascade involving Toll‐like receptors (TLRs) was highlighted by pathway analysis within the cerebellar white matter. The TLRs are involved in the induction of neuroinflammation in neurodegenerative diseases by regulating the production of proinflammatory cytokines that contribute to further neuronal damage. ²⁵ The most significantly disrupted molecular pathway in our study, the TLR4 cascade, has an important causal relationship with motor dysfunction in neurodegenerative conditions. Motor dysfunction occurs through suppression of inhibitory gamma aminobutyric acid (GABA) receptor activities at postsynaptic sites, decreasing GABA synthesis at presynaptic sites and causing a decrease in the number of PC. ²⁰ , ²⁵ Although no genes were significantly dysregulated in PC soma from horses with Shivers, the gene with the lowest adjusted P value, arginine methyltransferase 8 (PRMT8; P = .16) was also ↓DEG within cerebellar white matter. Prmt8‐null mice display abnormal PC dendritic arborization and abnormal motor behaviors, including hindlimb clasping and hyperactivity. ²⁶ Thus, our spatial transcriptomic analysis identified abnormalities in cerebellar white matter in horses with Shivers that were consistent with degenerative changes that could affect PC and their function.

Proteomic analyses also supported cerebellar axonal involvement in the pathogenesis of Shivers. Cerebellar homogenates had decreased expression of proteins found extensively in axons. These included 5 intermediate filament proteins (NEFL, NEFM, NEFH, INA, GFAP) and 2 microtubular proteins, (RTN4, MAP1A). ¹⁷ , ²⁷ Neurofilaments are highly associated with axons and are essential for the radial growth and structural stability of myelinated axons as well as for achieving optimal conduction velocity of electrical impulses along axons. ¹⁷ , ²⁸ , ²⁹ , ³⁰ Decreased expression of neurofilaments occurs in response to axonal injury and results in axon atrophy and decreased conduction velocities. ¹⁷ , ²⁹

We also found decreased expression of proteins involved in myelinating axons in horses with Shivers. These included MBP and CNP and BCAS1 that are involved in early myelinogenesis. ³¹ , ³² Furthermore, the beta 1 subunit of the Na/K ATPase (ATP1B1) which is expressed in PC, axons and basket cells and the activity of which controls the intrinsic firing mode of cerebellar PC, also had significantly decreased expression in horses with Shivers. ³³ , ³⁴ The α‐subunit of SCNA10, the voltage‐gated sodium channel present in axons, was a top downregulated gene. Altered expression of this channel causes abnormal firing patterns in PC and motor coordination deficits in the absence of obvious signs of ataxia in mouse models. ³⁵ Thus, our finding of altered expression of neurofilaments, myelin, and ion pumps and channels in the cerebellum of horses with Shivers implicates an abnormality in axons in horses with Shivers that could impact the cerebellum's ability to coordinate movement.

Horses with Shivers had increased expression of the calcium‐dependent phospholipid binding protein annexin A2 (ANAX2) which, in the brain, is thought to be largely associated with pathological conditions such as inflammation and neurodegeneration. ³⁶ Proteomic evidence for a potential stress response in the cerebellar hemispheres of horses with Shivers included increased expression of sirtuin 2 (SIRT2) and heat shock proteins (HSPD1, HSPA9). Sirtuin 2 orchestrates the oxidative stress response and has been linked to neuroinflammation, synaptic dysfunction, and abnormal metabolism. ¹⁸ Two oxidoreductive proteins also had increased expression in horses with Shivers, PGD and TXNDC17.

Evidence of regeneration within the cerebellum of horses with Shivers included increased expression of a neuroblast differentiation‐associated protein (AHNAK) as well as vimentin (VIM). ³⁷ , ³⁸ Vimentin, which had the highest differential expression of all cerebellar proteins, is a cytoskeletal protein present in regenerating but not mature axons. ³⁸ Damaged axons develop a growth cone replete in actin‐rich structures at its tip. CALD1, FLNA, TGLN, and PFN1 were upregulated actin‐associated proteins that had increased expression in horses with Shivers. ³⁹ The growth cone attempts to navigate back to its target through essential links to the extracellular matrix. ³⁹ There were 7 extracellular matrix proteins that had increased expression in horses with Shivers (TLN, COL6A2, COL6A3, OGN, DCN, BGN, PRELP, and TLN). Talin is crucial for axonal regeneration because it serves as a linker protein to connect the extracellular matrix to the regenerating cytoskeleton. Increased expression of collagens indicate a potential increase in extracellular matrix of horses with Shivers. ³⁹ Migration of the growth cone is impeded by the development of a glial scar that contains proteoglycans. ⁴⁰ Three proteoglycan‐associated proteins were upregulated in horses with Shivers including OGN, DCN, and BGN. Thus, our proteomic results suggest that there may be an attempt at a regenerative process within the axons in the lateral cerebellar hemispheres of horses with Shivers. Notably, the PC response to stress is not typical of that seen in most neurons and represents a partly degenerative and partly compensatory response characterized by upregulation of injury‐ and growth‐associated genes. ²¹ , ⁴¹ , ⁴²

Three of the proteins with decreased expression in Shivers horses, (calmodulin, calretinin, and parvalbumin) are involved in calcium buffering or calcium signaling, with parvalbumin found in high concentration in PC axons. ⁴³ , ⁴⁴ , ⁴⁵ Mice deficient in calretinin (Cr ^−/−) appear to have increased binding of calcium to the PC‐specific calcium binding protein calbindin, disturbed calcium homeostasis, and altered PC firing. ⁴³ Calcium binding proteins impact presynaptic neurotransmitter release and knock out mouse models for calcium binding proteins show that they produce a specific type of motor dysfunction. ⁴² , ⁴³ , ⁴⁴ , ⁴⁶ , ⁴⁷ Notably, calretinin and calbindin null mice do not display neurologic or behavioral deficits in a normal cage environment or on wide flat surfaces. However, in a context‐specific environment such as walking on a narrow runway, homozygous calretinin or calbindin null mice show impaired motor coordination. ⁴³ Heterozygotes show a similar but milder phenotype. ⁴³ Sex‐specific differences in calbindin concentrations have been identified in mouse PCs, with lower calbindin mRNA and protein expression in male mice than females. ⁴⁸ In horses with Shivers, males are 3 times more likely to develop the disease than females. ³ Thus, calcium binding proteins are good candidates to explore in diseases with motor coordination deficits without severe ataxia, particularly if there is a predilection for males to be affected. ⁴³

We chose to sample gray and white matter in the lateral cerebellar hemispheres and not the deep cerebellar nuclei to include PC soma and white matter in the same sections. The lateral cerebellar hemispheres function in motor planning and the timing, onset and coordination of movements ⁴⁹ and were therefore deemed appropriate to sample in horses with Shivers. Because the cerebellum has functional compartments that have not been mapped in horses, there could have been alternative areas to study that had more differences in gene and protein expression. ⁵⁰ To minimize any potential variation, we ensured that the same region of the lateral cerebellar hemispheres was sampled in the Shivers and control horses.

Our study had some limitations. We studied a small number of horses because of the rarity of Shivers cases and, in part, because of the need to obtain donated Shivers and control horses with no systemic diseases where the brain could be immediately removed. Spatial transcriptomics had not previously been validated in horses and therefore we chose to evaluate 2 Shivers and 2 control horses using this new technique. Despite the small numbers of horses, we were able to obtain significant results for differences in white matter between Shivers and control horses by sampling 3 ROIs per horse. It is possible that we did not have the power to detect subtle changes in the PC soma because of the limited number of horses. The value of the GeoMx Digital Spatial Profiling (DSP) method was that it combines the strengths of imaging and sequencing approaches and provides valuable information regarding gene profiles within specific regions of the equine cerebellum. It uses RNA hybridization probes labeled with photocleavable next generation sequencing indexing tags to capture specific regions in tissue sections using microscopy‐guided photocleavage and next generation sequencing. ¹² This technique has great utility because DSP works optimally on formalin‐fixed paraffin‐embedded tissue, owing to probe hybridization‐based mRNA detection. ¹²

In conclusion, our results indicate that horses with Shivers have both transcriptomic and proteomic evidence for axonal degeneration in the lateral cerebellar hemispheres that potentially could disrupt cerebellar motor coordination without producing ataxia. Combined with current and previous histopathologic findings, our study supports a PC axonopathy as a cause for Shivers in horses and suggests several potential candidate genes for future investigations.

CONFLICT OF INTEREST DECLARATION

Dr. Valberg runs a commercial muscle biopsy service and receives royalties for genetic tests, but these have no overlap with the neurologic disease presented in this article. No other authors declare a conflict of interest.

OFF‐LABEL ANTIMICROBIAL DECLARATION

Authors declare no off‐label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION

Approved by the IACUC of the University of Minnesota, 1310‐30989A and Michigan State University, PROTO201900038. Horses were euthanized by intravenous administration of a barbiturate with informed client consent.

HUMAN ETHICS APPROVAL DECLARATION

Authors declare human ethics approval was not needed for this study.

Supporting information

Table S1. Differentially expressed downregulated and upregulated genes in white matter of the lateral cerebellar hemispheres with a log₂FC >1.5 comparing Shivers to control horses.

Data S1. A detailed description of the method utilized to trim the deep cerebellar nuclei for histopathology.

Valberg SJ, Williams ZJ, Henry ML, Finno CJ. Cerebellar axonopathy in Shivers horses identified by spatial transcriptomic and proteomic analyses. J Vet Intern Med. 2023;37(4):1568‐1579. doi: 10.1111/jvim.16784