Снижение митохондриальной активности в демиелинизирующей мозжечке прогрессивного множественного склероза и хронической EAE способствует потере клеток Purkinje.

Multiple sclerosis (MS) affects 2.3 million individuals globally, characterized by autoimmune-driven inflammation leading to central nervous system (CNS) demyelination and chronic disability. About 80% of MS patients experience cerebellar inflammation-associated demyelination, manifesting as tremors, impaired motor control, and coordination loss throughout their lives ( 1 – 3 ). Cerebellar gray matter atrophy occurs in later MS disease stages ( 4 ). However, studies suggest a correlation between cerebellar cortex demyelination and earlier mortality in primary progressive MS (PPMS) and secondary progressive MS (SPMS) patients ( 2 , 5 ). Despite cerebellar dysfunction being prevalent in MS, the underlying mechanisms remain incompletely understood.

Purkinje cells (PCs), myelinated gamma-aminobutyric acid (GABA)ergic neurons, are the sole output neurons of the cerebellar cortex, integrating signals from the brain and spinal cord to modulate motor activity ( 6 ). PCs receive input from climbing fibers (from the inferior olive) and parallel fibers (from granule cells) and send inhibitory signals to the deep cerebellar nuclei, which then influence motor pathways ( 7 ). PCs, with their extensive dendritic arbors and glutamatergic inputs, necessitate high metabolic activity for efficient signal transmission to deep cerebellar nuclei ( 8 ). Postmortem analysis reveals inflammation, demyelination, and PC loss within cerebellar MS lesions ( 5 , 9 , 10 ), alongside evidence of PC fusion and heterokaryon formation possibly for neuronal repair initiation ( 11 ). MS-afflicted PCs, influenced by demyelination and increased energy demands, also exhibit abnormal expression of Na v 1.8, a voltage-gated Na + channel typically found in unmyelinated nociceptive fibers, potentially leading to increased spontaneous firing and disruption of CNS circuitry ( 12 ). This also likely contributes to exacerbated MS-associated cerebellar dysfunction ( 13 ).

Mitochondria, crucial for energy production via oxidative phosphorylation, are implicated in MS neurodegeneration ( 14 ). The CNS’s high metabolic rate makes it particularly vulnerable to even slight aberrations in energy supply. Under pathological conditions such as oxidative stress, excessive Ca 2+ influx can trigger mitochondrial fission and/or fusion, yielding fragmented/swollen and dysfunctional mitochondria and contributing to cell apoptosis and subsequent neurodegeneration ( 15 , 16 ). PCs in the cerebellum are among the most energy-demanding cells in the human body, with each PC using approximately 1.24 × 10 10 adenosine triphosphate (ATP) molecules per second ( 17 ), making them exceptionally susceptible to mitochondrial dysfunction. Indeed, PC demyelination, PC loss, and ataxia are prominent in primary mitochondrial diseases ( 18 , 19 ).

Postmortem studies of MS CNS tissue demonstrated a compromised neuronal ATP production by oxidative phosphorylation and suppressed mitochondrial complex IV (COXIV) activity in demyelinated axons and neurons ( 20 – 22 ). Given that neuronal mitochondrial dysfunction is implicated in MS pathogenesis ( 22 – 25 ), understanding mitochondrial dynamics, especially in PCs, could be pivotal in deciphering cerebellar symptoms in MS.

Previous research from our group and others have demonstrated inflammatory demyelination in the cerebellum and PC loss during the late stages of experimental autoimmune encephalomyelitis (EAE), a mouse model of MS ( 14 , 26 ). Mitochondrial dysfunction has been observed as early as the asymptomatic phase of EAE, preceding the onset of clinical symptoms ( 27 , 28 ). Evidence suggests that mitochondrial stress is closely associated with neuronal loss ( 28 ). Additionally, during CNS inflammation in EAE, metabolic disruption arises due to increased energy demands that demyelinated neurons cannot meet. This is attributed to altered mitochondrial morphology, reduced mitochondrial activity, and changes in the expression of genes involved in mitochondrial function ( 20 , 29 ).

Our hypothesis suggests that changes in mitochondrial dynamics within the cerebellum play a role in progressive PC loss and worsening cerebellar dysfunction in MS and its experimental model, EAE. To explore mitochondrial dysfunction in the MS cerebellum, we conducted a pathological analysis of cerebellar tissue, primarily from postmortem SPMS patients and healthy controls, obtained from the NeuroBioBank and the Cleveland Clinic. To further investigate the mechanisms and progression of cerebellar dysfunction, we utilized the chronic EAE model, which mirrors critical aspects of MS, such as inflammation, demyelination, and neurodegeneration ( 26 , 30 ). At various stages of EAE progression, we assessed behavior, pathology, mitochondrial activity, and mitochondrial protein expression and performed NanoString ribonucleic acid (RNA) analysis. Findings from both MS and EAE studies revealed significant pathological changes and PC death. Notably, the loss of COXIV in demyelinated PC cells suggests a link between PC death and mitochondrial dysfunction. Unraveling the pathophysiology of MS symptoms holds promise for developing improved treatments and enhancing the quality of life for individuals living with MS.

Results

SPMS Postmortem Cerebellar Tissue Demonstrates Demyelination, Inflammation, and Axon Loss.

The cerebellar cortex is a major site for demyelination with nearly 38.7% of the cerebellar cortex being affected in a cohort of PPMS and SPMS patients ( 5 , 9 , 10 ). To establish MS-induced cerebellar pathology, a cohort of MS cerebellar sections from primarily a SPMS cohort with an expanded disability status scale (EDSS) of 6 to 9 (details in SI Appendix , Table S1 ) was compared to “normal” control cerebellar sections using immunohistochemistry (IHC) ( Fig. 1 A , i and SI Appendix , Table S2A ). The MS cerebellar sections ( Fig. 1 A ) showed significant white matter (WM) demyelination using proteolipid protein (PLP) immunostaining as compared to normal control sections (t = 8.774, df = 14, P = 0.0001) ( Fig. 1 A , ii – iv ). Additionally, increased inflammatory lesions which extended from the WM into the adjacent granule cell layer (GCL), molecular layer (ML), and the meningeal layers were identified with the presence of Glial Fibrillary Acidic Protein (GFAP) + activated astrocytes (WM+GCL: t = 4.295, df = 13, P = 0.0009; ML: t = 3.250, df = 13, P = 0.0063) and leukocytes (CD45) (WM+GCL: t = 2.601, df = 13, P = 0.0219; ML: t = 2.188, df = 13, P = 0.0476) ( Fig. 1 B , i – iii ). Parallel fibers, the axons of granule cells, run perpendicular to PC dendrites. They serve as a major pathway for transmitting sensory and motor information to PCs. A significant loss of parallel fibers (immunostained with NF200) in MS is accompanied by axon damage (t = 2.617, df = 14, P = 0.0101, Fig. 1 C , i – iii ).

Neuropathological analysis of the human postmortem cerebellum. ( A , i ) Table of MS brain samples used for IHC. ( ii ) Confocal image of normal cerebellum stained for NF200 (red) and PLP (green); dashed rectangle marks PLP WM analysis region. ( iii ) PLP staining in normal and MS cerebellum. ( iv ) PLP expression significantly decreases in MS WM. ( B , i ) Confocal image of MS cerebellum stained for CD45 (red) and GFAP (green); dashed area marks lesion regions. ( ii ) High-magnification images of GFAP+ and CD45+ cells in lesions. ( iii ) GFAP and CD45 expression significantly increase in ML, GCL, and WM of MS tissue. ( C , i ) NF200 (red) and DAPI (blue) staining show NF200 variability in MS and normal sections. ( ii ) Representative parallel fibers in ML. ( iii ) NF200+ parallel fiber expression significantly decreases in MS cerebellum. n = 4 normal, n = 11 MS tissue.

MS Postmortem Cerebellar Tissue Demonstrates PC Loss and Decreased COXIV Expression.

Past studies have shown significant PC loss in lesioned areas compared with normal areas ( 5 , 9 ). To determine whether PC counts are altered in the MS cerebellum, IHC for Calbindin, a calcium-binding protein expressed in PCs, was performed. There was a significant decrease in Calbindin + PCs in the MS postmortem tissue compared to the control postmortem tissue in every lobule ( Fig. 2 A , i and ii ; t = 2.882, df = 13, P = 0.0128). When the MS tissue cohort was split separately by sex, both male and female tissue revealed decreased Calbindin + PCs ( Fig. 2 A , iii ; Female: t = 3.243, df = 7, P = 0.007; Male: t = 2.932, df = 8, P = 0.0095). COXIV catalyzes the final step in the mitochondrial electron transfer chain and is regarded as one of the major regulation sites for oxidative phosphorylation. To assess whether this PC loss coincided with mitochondrial dysfunction, IHC was performed with COXIV and Calbindin antibodies. COXIV puncta were quantified in individual delineated PC somas ( Fig. 2 B ). There was a significant decrease in COXIV expression in the PCs of MS tissue compared to control tissue ( Fig. 2 B , i and ii ; t = 3.403, df = 28, P = 0.0020). When the MS postmortem tissue was split separately by sex, the decrease in COXIV expression was consistent in both sexes [ Fig. 2 B , iii ; Female: F(2, 27) = 6.151, P = 0.022; Male: F (2, 27) = 6.151, P = 0.004].

( A , i ) Representative normal (NM) and MS cerebellar sections showing Calbindin (green), in the PCL and ML of the cerebellum. ( ii ) There was a significant decrease in Calbindin cell counts in the MS PCL and ML as compared to the normal. ( iii ) There was a significant decrease in PC counts in both the MS female and MS male samples. ( B , i ) Representative images from postmortem normal and MS ML and PC layers showing COXIV (red), Calbindin (green), and DAPI (blue) staining. ( ii ) There was a significant decrease in COXIV puncta in the PC somas of MS cerebellum. ( iii ) There was a significant decrease in COXIV puncta in both females and males compared to the normal group. n = 4 normal tissue, n = 11 MS tissue. * P < 0.05, ** P < 0.01 with Student’s t test.

Progressive Demyelination, Inflammation, and Axon Damage Start Early and Persist throughout EAE.

Longitudinal EAE was used to examine demyelination, inflammation, and PC numbers from peak disease to the chronic stage. Mice induced with EAE were grouped and perfused at timepoints EAE D21, EAE D40, and EAE D60, along with age-matched normal (only injected with Freuds adjuvant) controls. Sagittal slices of the cerebellum were processed for myelin basic protein (MBP) IHC ( Fig. 3 B , i ). Images were then taken at lobule IV/V, specifically at the GCL and WM layers. There was a decrease in MBP expression in both the WM [H(4) = 16.41, P < 0.0001, EAE D21: P = 0.0002, EAE D40: P = 0.0008, EAE D60: P < 0.0001] and GCL [H(4) = 13.66, P = 0.0034, EAE D21: P = 0.0057, EAE D40: P = 0.0108, EAE D60: P = 0.0003] compared to the normal mice ( Fig. 3 B , ii – iv ).

Experimental design and EAE scores. ( A , i ) Schematic of EAE induction and longitudinal assessment of cerebellar and PC mitochondrial pathology. ( ii ) Three representative EAE experiments are shown. EAE mice (red) show disease onset around day 10. ( B , i ) MBP (green) and DAPI (blue) staining in a normal cerebellum. ( ii ) MBP analysis in WM and GCL across groups. ( iii and iv ) MBP expressions significantly decrease in WM and GCL at EAE D21, D40, and D60. ( C , i and ii ) Thy1 (green) and Iba-1 (red) staining with DAPI (blue); analysis regions marked. ( iii and iv ) Iba-1 expression significantly increases in WM and GM at EAE D21, D40, and D60. n = 8 mice/group for IHC.

Inflammation was investigated by quantifying the expression of a marker for microglia and macrophages, ionized calcium binding adaptor molecule 1 (Iba-1), in normal and EAE-induced Thymus Cell Antigen (Thy1)-YFP mice. There was an increase in Iba-1 expression in both the WM and GCL of EAE mice [H(4) = 21.98, P < 0.0001, EAE D21: P = 0.0133, EAE D40: P = 0.0013, EAE D60: P < 0.0001] and GCL [H(4) = 7.35, P = 0.0006, EAE D21: P = 0.0070, EAE D40: P = 0.0048, EAE D60: P < 0.0001] ( Fig. 3 C , i – iv ). Axon damage during EAE in Thy1-YFP axons was visualized in the WM. In addition, changes in parallel fibers were measured using NF200 IHC. Fig. 4 A , i shows robust Thy1 fluorescence in the cerebellum GCL and WM. Thy1-YFP+ axon blebs are visible at EAE D21 with swollen axons apparent at EAE D40 and EAE D60 ( Fig. 4 A , ii ). Similarly, decreased NF200 parallel fiber expression in the ML at EAE D40 and EAE D60 [ Fig. 4 B , i and ii , H(4) = 17.51, P = 0.0006; EAE D40: P = 0.0033; EAE D60: P = 0.0002] was also observed.

WM axon damage and parallel fiber loss in EAE. ( A , i ) Thy1-YFP cerebellum image highlighting the analyzed axon blebbing region. ( ii ) Axon blebbing increases at EAE D21 and persists, with swollen, discontinuous axons at EAE D40 and D60. ( B , i ) NF200 (red) and DAPI (blue) staining show PC bodies and parallel fibers. ( ii ) NF200+ parallel fibers significantly decrease at EAE D40 and D60. n = 6 mice/group.

A Decrease in COXIV Protein Expression, Swollen Mitochondria, and PC Loss Are Evident in the EAE Cerebellum.

To determine whether these pathological changes were coupled with alterations in mitochondria activity, overall cerebellar COXIV expression was assessed ( Fig. 5 A ). There was a regional change in the distribution of COXIV during EAE. Compared to normal, EAE D21 showed a significant decrease in overall expression, specifically in the GCL as compared to the PCL [F(3, 16) = 30.27, P < 0.0001] ( Fig. 5 A , i and iii ). Interestingly, in EAE D40, COXIV staining appeared to rebound, and by EAE D60, there was no significant difference as compared to the normal group ( Fig. 5 A , iii ). This change in COXIV expression was not in the PC layer but the GCL and WM of the cerebellar lobules ( Fig. 5 A , i and iii ). Next, the number of PCs was counted and COXIV puncta were quantified in individual delineated PC somas ( Fig. 5 B , i and ii ). Analysis revealed a significant decrease in PC numbers at EAE D40 and EAE D60 compared to normal [ Fig. 5 B , iii , F(3, 32) = 4.084, EAE D40: P = 0.0144; EAE D60: P = 0.0143]. This also coincided with a decrease in COXIV puncta at EAE D60 compared to normal [ Fig. 5 B , ii – iv , F(3, 113) = 3.906, EAE D60: P = 0.0031]. The decrease in PC COXIV puncta was contrary to the overall increase in cerebellar COXIV expression at EAE D60 in Fig. 5 A , i and iii and indicated dysfunctional mitochondria in PCs. To investigate PCs and axonal mitochondrial pathology at a higher resolution, we conducted transmission electron microscopy (TEM) analysis of normal and EAE D40 cerebellar sections ( Fig. 5 C ). As compared to normal, the cerebellar WM of EAE D40 sections indicated loss of myelin ( Fig. 5 C , i ). Swollen morphologically distinct mitochondria were observed in the PC axons and PC bodies of EAE D40 as compared to normal ( Fig. 5 C , i ). Quantification of WM axons revealed a significant increase in axon diameter and G-ratio in the EAE D40 group, indicative of axon demyelination ( Fig. 5 C , i – iv , axon diameter: t = 4.945, df = 398, P < 0.0001; g-ratio: t = 8.590, df = 398, P < 0.0001) along with rounder, swollen, and increased mitochondria size in demyelinated axons. In contrast, elongated mitochondria evenly distributed in myelinated axons were observed in normal controls ( Fig. 5 C , i and v ).

EAE-associated PC loss and mitochondrial pathology. ( A , i ) GFAP (green) and COXIV (red) staining show COXIV reduction in the GCL at EAE D21, later decreasing in the PCL and increasing in the GCL at EAE D40 and D60. ( ii ) Higher magnification highlights COXIV loss in the PCL at EAE D60. ( iii ) Quantification confirms a significant COXIV decrease at EAE D21. ( B , i ) Calbindin+ PCs (green) exhibit loss and altered dendritic arborization at EAE D60. ( ii ) Calbindin and COXIV costaining show COXIV reduction in PCs. ( iii ) PC counts significantly decline at EAE D40 and D60. ( iv ) COXIV puncta in PC somas significantly decrease at EAE D60. ( C , i ) EM high magnification images of cerebellar WM, PC bodies, and mitochondria (mito) at EAE D40 compared to normal. ( ii ) Axon diameter and G-ratio comparison. ( iii and iv ) EAE D40 exhibits increased axon diameter and G-ratio. ( v ) Mitochondrial size vs. count in EAE D40 and normal groups. n = 4 to 5 mice/group for EM analysis.

EAE Mice Have a Decrease in Coupled Respiration at Peak EAE Disease.

To characterize overall mitochondrial dynamics and function, mitochondria were isolated from the cerebellum and subjected to the Seahorse XFp Analyzer Mito Stress test (Agilent, San Diego, CA). The Mito Stress test evaluates the functionality of the mitochondrial respiratory chain by real-time monitoring of the oxygen consumption rate (OCR). The average traces between the EAE D21, D60, and normal mice are demonstrated in Fig. 6 A . Compared to the normal group, no difference in basal respiration, uncoupled respiration, and spare respiratory capacity (SRC) was observed at EAE D21. However, a significant decrease in coupled respiration (State 3) ( Fig. 6 A , i , t = 2.486, df = 7, P = 0.0418) was observed. Compared to normal groups, no differences in basal respiration, coupled respiration, uncoupled respiration, and spare respiratory capacity (SRC) were observed between EAE D60 and normal mice ( Fig. 6 A , ii ).

Decrease in coupled respiration at peak EAE disease. ( A ) Seahorse assay of cerebellar mitochondrial OCR. ( i ) EAE D21 showed reduced coupled respiration compared to normal after ADP, Oligomycin, FCCP, and Rotenone/Antimycin A addition. ( ii ) No significant OCR differences were observed between normal and EAE D60. ( B , i ) Western blot of mitochondrial homogenates. ( ii ) No differences in COXI, COXII, COXIII, or ATP Synthase across groups, but CIV-MTCO1 (COXIV) and syntaphilin were significantly reduced in EAE D40 and D60. n = 4 to 6 mice/group.

EAE Mice Have Alterations in the Electron Transport Chain.

To further assess mitochondrial changes in the cerebellum, a comprehensive analysis of oxidative phosphorylation was performed ( Fig. 6 B and SI Appendix , Table S2B ). Mitochondria homogenates of normal, EAE D21, EAE D40, and EAE D60 groups were compared with subunits of all 5 components of the electron transport chain (OXPHOS), using voltage-dependent anion channel (VDAC) as a loading control. Decreased protein levels of the mitochondrially encoded cytochrome c oxidase I (MTC01) subunit of COXIV in cerebellar mitochondria at EAE D40 and EAE D60 was observed compared to normal mice [ Fig. 6 B , i and ii , H(4) = 7.138, P = 0.0303, EAE D40: P = 0.0313, EAE D60: P = 0.0116]. There were no significant differences in any of the other complexes between normal and EAE groups. To determine whether mitochondrial transport is altered in EAE longitudinally, we measured syntaphilin, an axonal protein that is used as a measure of anchored mitochondria ( 31 ) ( Fig. 6 B , ii ). EAE D40 and EAE D60 mice had decreased syntaphilin expression compared to normal mice [H(4) = 14.50, P = 0.0023, EAE D40: P = 0.0088, EAE D60: P = 0.0067].

Messenger RNA Transcripts Related to Oxidative Phosphorylation Are Significantly Downregulated in EAE.

To quantify the expression of genes involved in core metabolic processes and immunometabolism in normal and EAE D40 cerebellum, NanoString gene analysis was performed ( Fig. 7 ). The nCounter “Metabolic Pathways Panel” quantifies 768 genes that are involved in essential pathways related to cell metabolism, signaling, and stress ( Dataset S1 ). Raw data and calculated fold changes of all genes from normal (n = 3) and EAE D40 (n = 4) are presented in Dataset S2 . A heat map shows changes in all genes from the Metabolic Pathways panel ( Fig. 7 A , ii ). A volcano plot was used to demonstrate genes that are differentially expressed during EAE ( Fig. 7 A , iii ). Then, 104 genes were significantly upregulated, and 66 genes were found to be significantly downregulated in EAE D40 using predefined cut-offs for selection ( P < 0.05 with a log2 fold-change) compared to normal. Genes that were significantly upregulated in EAE D40 such as Col4a1 and Col4a2 are components of type IV collagen and important in the PI3K/Akt signaling pathway. Other upregulated genes such as Aspa , Nat8l , and Slc16a6 have been shown to be important in amino acid synthesis and transport. Genes that were significantly downregulated in EAE D40 were related to different components of the mitochondrial oxidative phosphorylation (OXPHOS) system, such as COXI, COXIV, and COXV. Past studies have shown significant changes in associated genes during EAE in the CNS ( 32 , 33 ). NADH:ubiquinone oxidoreductase, also known as COXI, starts the electron transport chain by catalyzing the transfer of two electrons from NADH to ubiquinone ( Fig. 7 B , i ). Subunits of NADH:ubiquinone oxidoreductase such as Ndufa12 (t = 5.774, df = 5, P = 0.0011), Ndufa3 (t = 2.883, df = 5, P = 0.0172), Ndufa6 (t = 2.885, df = 5, P = 0.0172), Ndufb10 (t = 3.701, df = 5, P = 0.0070), Ndufb8 (t = 5.243, df = 5, P = 0.0017), and Ndufs8 (t = 3.391, df = 5, P = 0.0097) were significantly downregulated in the cerebellum of the EAE D40 mice compared to the normal mice ( Fig. 7 B , i ). Subunits of COXIV such as Cox4i1 (t = 4.833, df = 5, P = 0.0024) , Cox5a (t = 5.734, df = 5, P = 0.0011) , Cox5b (t = 3.771, df = 5, P = 0.0065) , Cox6a1 (t = 4.708, df = 5, P = 0.0026) , Cox6b1 (t = 6.302, df = 5, P = 0.0007) , Cox7c (t = 5.280, df = 5, P = 0.0016), and Cox8a (t = 3.988, df = 5, P = 0.0052) were also significantly downregulated in EAE D40 mice compared to normal ( Fig. 7 B , ii ). In addition, subunits of COXV such as Atp5k (t = 3.118, df = 5, P = 0.0132) and Atp6v1f (t = 6.431, df = 5, P = 0.0007) were significantly downregulated in the EAE D40 mice compared to normal ( Fig. 7 B , iii ).

Impaired mitochondrial oxidative phosphorylation gene expression in late-stage EAE cerebellum. ( A , i ) NanoString analysis of gene expression in EAE D40 and normal mouse cerebellum (n = 3-4/group). Image credit: Created in BioRender. Feri, M. (2025) https://BioRender.com/9669j9h , ( ii ) Heatmaps show expression levels (low: red, high: green), and ( iii ) a volcano plot highlights differentially expressed genes (fold change, P -value < 0.50). ( B ) Heatmaps display downregulated genes in EAE D40 related to oxidative phosphorylation: ( i ) Complex I (Ndufa12, Ndufa3, Ndufa6, Ndufb10, Ndufb8, Ndufs8), ( ii ) Complex IV (Cox4i1, Cox5a, Cox5b, Cox6a1, Cox6b1, Cox7c, Cox8a), and ( iii ) Complex V (Atp5k, Atp6v1f). ( C ) Schematic summarizing affected OXPHOS genes. Image credit: Created in Biorender. Feri, M. (2025) https://BioRender.com/a87e2gt .

Mice Show Alterations in Gait during Early EAE That Persist through Chronic EAE.

Cerebellar pathology in MS can lead to various forms of ataxia, but directly linking symptoms to specific cerebellar damage, such as PC loss, remains challenging. Postmortem cerebellar tissue analyzed in this study came from patients with severe motor disability (EDSS 6 to 9). Similarly, EAE mice exhibited significant motor impairment, as indicated by high EAE scores and rotarod performance deficits ( 34 , 35 ). Gait analysis revealed a persistent decrease in stride length from early EAE (D13) through the chronic stage, while other gait parameters remained unchanged ( SI Appendix , Fig. S1 ) ( 36 , 37 ). Although not solely indicative of cerebellar dysfunction, the reduced stride length suggests impairment within cerebellar circuitry in EAE mice.

Discussion

Ongoing research seeks to better understand the mechanisms by which MS affects the reduction in PC numbers, including the roles of specific inflammatory pathways, mitochondrial dynamics, and neurodegeneration. Identifying these mechanisms may help develop targeted therapies to protect cerebellar function and alleviate the motor deficits associated with MS.

After 10 to 15 y of MS, most patients transition to SPMS, characterized by gradual brain and spinal cord atrophy due to myelin loss and cumulative neuronal damage. In the cerebellum, PCs and other neurons are progressively damaged, contributing to symptoms like ataxia, dysmetria, intention tremors, and nystagmus ( 3 , 5 ). PCs are particularly vulnerable due to MS-related oxidative stress, demyelination, and inflammation, which disrupt mitochondrial function. This dysfunction leads to energy deficits, increased reactive oxygen species production, and impaired calcium handling, all of which can negatively affect PC survival ( 22 ).

Previous studies have shown significant cerebellar inflammation and demyelination in MS, but findings on PC loss vary. Howell et al. used 14 SPMS female samples and quantification revealed no significant difference in the PC density between controls and the SPMS samples ( 10 ). Similarly, Redondo et al. ( 9 ) used 6 MS 40 to 50-y-old female samples and compared them to 80-y-old normal male tissue. Their results showed PC pathology in MS but did not show an overall loss in PC numbers throughout the cerebellum. PC loss was specifically found in the lesioned areas similar to what was observed earlier by Kutzelnigg et al. ( 5 ). However, Kemp et al. ( 11 ) found a significant loss of PCs using varied samples from eight patients with MS that included 1 RRMS, 1 PMS, and 5 SPMS (mostly female patients) compared to six young, control male patients with no pathology. Significant PC loss was observed in some MS cases, including the Charcot and Marburg variants, linking it to cerebellar demyelination ( 38 ). Interestingly, PC loss is also seen in essential tremor ( 39 ), suggesting a potential common pathological mechanism underlying ataxia.

Here, we used a cohort of mostly SPMS (n = 9), along with RRMS n = 1 and PMS n = 1 cerebellar tissue controlled for age, sex, duration of disease, and EDSS scores, and compared them to normal (n = 5) cerebellar tissue. Our study findings revealed significant demyelination, inflammation, and parallel fiber loss, accompanied by a significant reduction in PCs and a decrease in COXIV puncta within PC bodies in the MS cerebellum as compared to the normal group. These results unequivocally suggest the connection between inflammatory demyelination and mitochondrial dysfunction in PCs.

Our study is the first to quantify a reduction in mitochondrial function (COXIV IHC) specifically in PCs somas in the MS brain. Previous research has reported decreased COXIV expression in pattern III MS lesions, affecting axons, astrocytes, and oligodendrocytes ( 40 ). Additionally, COXIV expression was reduced in axons at the rim of chronic active lesions, whereas axons in inactive lesions showed increased COXIV activity ( 22 ), indicating distinct mitochondrial alterations across different lesion types. Other studies have also reported an overall reduction in COXIV expression in postmortem MS brains ( 22 ). Further investigations, including genomic spatial profiling of MS postmortem tissue, are needed to correlate decreased COXIV in PCs with demyelinating white matter lesions, inflammation, and broader cerebellar pathology ( 41 ).

The second phase of our study examined whether the inflammatory demyelination, PC loss, and mitochondrial changes observed in SPMS patients could be studied over time in the EAE mouse model. Human postmortem studies are limited by factors like patient variability and treatment history, making animal models crucial for understanding disease progression and the molecular mechanisms driving mitochondrial dysfunction and PC degeneration.

Pathological changes observed in SPMS postmortem tissue were similarly assessed in EAE mice at peak (D21), late (D40), and chronic (D60) stages. Previous studies have reported significant cerebellar pathology in EAE, including inflammation, axon damage ( 42 , 43 ), and PC loss ( 26 ). In this study, Iba-1+ cell increases were noted at all EAE stages, consistent with findings in other CNS regions ( 30 , 44 , 45 ). While Iba-1 is not a definitive inflammation marker, further analysis of immune cell involvement will enhance comparisons with human MS pathology. Inflammation and demyelination were observed as early as EAE D14 ( 26 , 42 , 43 , 46 ), with axon damage appearing as a hallmark feature of both EAE and MS ( 47 , 48 ). Some studies suggest that axon degeneration may precede demyelination due to oxidative stress ( 24 ) and sodium channel alterations ( 49 ). This study confirms white matter axon damage and reduced parallel fibers in EAE mice, further supporting its relevance as a model for MS-related cerebellar pathology.

Mitochondrial dysfunction, fragmentation, and impaired trafficking are more pronounced in spinal cord axons near perivascular inflammatory cells and are associated with increased phosphofructokinase-2 expression in activated astrocytes and reduced mitochondrial complex I activity, preceding demyelination or axonal degeneration ( 28 ). ATP production is essential for neurotransmission and synaptic plasticity, and demyelination increases neuronal energy demands, requiring mitochondria to redistribute accordingly ( 50 , 51 ). PCs have a high mitochondrial density to support their elevated energy needs ( 52 , 53 ). In this study, EAE D60 mice exhibited decreased COXIV levels in PC somas and increased PC loss, suggesting that COXIV reduction may precede PC degeneration. A prior study of MS autopsy tissue found a positive correlation between mitochondrial content in demyelinated axons and COXIV deficiency in dorsal root ganglia neuronal cell bodies ( 51 ). COXIV expression in the EAE cerebellum fluctuated between EAE D21 and D60, likely due to mitochondrial redistribution from cell bodies to axons. High COXIV intensity in granule cells and PCs made quantification in cerebellar white matter challenging. Future studies using two-photon microscopy in MITOtag-YFP mice will further investigate these mitochondrial dynamics. Ultrastructural analysis revealed increased mitochondrial number and size in demyelinating PC axons and morphologically distinct mitochondria in PC somas at EAE D40.

The decrease in coupled respiration at EAE D21 may result from inflammation, active demyelination, and cellular stress affecting oligodendrocytes, neurons, and their progenitors. This decline aligns with the reduction in COXIV observed via IHC at peak disease. By EAE D60, COXIV levels (except in PC somas) rebounded, corresponding with restored coupled respiration. However, Western blot analysis revealed an overall decline in COXIV protein in late and chronic EAE, mirroring the COXIV decrease in PCs. While a link between reduced COXIV in PCs and impaired respiration is possible, further cell-specific studies are needed. Previous research has shown that deleting mitochondrial ATPase in PCs leads to PC loss and decreased COXI and COXIV expression ( 54 ). Although no significant difference in ATP5A protein expression was found in EAE D21 mice, nuclear-encoded ATP Synthase genes were downregulated in SPMS postmortem tissue ( 23 ) and EAE D40 RNA analysis. The decline in COXIV protein in late-stage EAE suggests that early disease events influence long-term cellular health.

Syntaphilin is crucial for anchoring mitochondria to microtubules, ensuring proper distribution within axons to meet local energy demands. Its role becomes especially important in demyelinated regions, where axons experience increased metabolic stress due to myelin loss. In MS, elevated syntaphilin levels in chronic spinal cord lesions suggest that mitochondria in demyelinated axons become relatively immobile, potentially contributing to disease pathology ( 22 , 31 , 55 ). This reduced mitochondrial mobility may contribute to the accumulation of mitochondria within demyelinated axons, potentially playing a role in the pathophysiology of chronic MS lesions ( 22 ). However, in EAE, syntaphilin levels in the cerebellum decrease significantly at EAE D40 and D60, indicating impaired mitochondrial anchoring, which may promote neuronal degeneration. Alternatively, this reduction could reflect changes in mitochondrial positioning within demyelinated axons. In a healthy CNS, syntaphilin supports mitochondrial function, but during demyelination, increased syntaphilin may have harmful effects. This is supported by findings in shiverer mice, where syntaphilin deletion improved survival, preserved PCs, and reduced axonal damage ( 55 ). These results suggest syntaphilin as a potential therapeutic target for demyelinating diseases.

During EAE D40, several mitochondrial genes were significantly downregulated, including members of the NADH:ubiquinone oxidoreductase family, such as Ndufa6 , which is essential for COXI activity and assembly ( 56 ). Similar gene downregulation ( Ndufa6, Ndufb8 ) has been observed in the MS motor cortex ( 23 ). In EAE cerebellum, reductions were also noted in cytochrome c oxidase genes ( Cox4i1, Cox5a, Cox5b, Cox6a1, Cox7c ) and ATP synthase genes ( Atp5k, Atp6v1f ), mirroring changes in the MS motor cortex ( 23 , 33 ). However, these mitochondrial alterations are complex and influenced by the diverse cellular environment of the cerebellum, including astrocytes, microglia, oligodendrocytes, and neurons. The interplay of astrogliosis, microgliosis, neurodegeneration, and immune infiltration in EAE and MS suggests that mitochondrial changes occur in a cell-specific manner. Future research will investigate early MS (RRMS) to assess initial PC responses and compare findings to SPMS. Additionally, region-specific mitochondrial studies using brain slice punches and Seahorse XFp analysis ( 57 ), as well as cell-specific investigations with Mitotag transgenic models ( 58 ), will be crucial for understanding mitochondrial dynamics in demyelinating diseases.

In the current study, we highlight MS cerebellar pathology in a new largely SPMS population that have minimal postmortem autolysis duration controlled by age, sex, and EDSS scores. Here, we focused on longitudinal EAE-induced cerebellar pathology with a focus on mitochondria pathology, function, and morphology and their relationship to PC health. The EDSS is the most used tool to assess disability in MS, ranging from 0 to 10 in 0.5-unit increments that indicate increasing levels of disability, based on a subjective examination by a neurologist. EDSS scores from 1.0 to 4.5 apply to individuals with MS who can walk without assistance. The postmortem cerebellar brain sections analyzed in our study had EDSS scores between 6 and 9, indicating a loss of ambulatory ability.

There are no behavioral tests that exclusively assess cerebellar function without spinal cord and motor cortex involvement. EAE mice underwent gait testing ( 36 ), which may reflect cerebellar dysfunction alongside spinal cord and motor cortex impairments. Mice with an average EAE score of 2.5 showed decreased stride length early in disease progression, persisting through EAE D60 coinciding with cerebellar demyelination, inflammation, and neurodegeneration.

Our findings indicate that cerebellar pathology in MS may result from mitochondrial dysfunction, which, alongside significant demyelination, contributes to the loss of myelinated PC axons, nonmyelinated parallel fibers, increased inflammation, and PC loss ( Fig. 8 ). To further investigate these changes, we examined the EAE model’s potential for studying cerebellar pathology and disease progression in MS. While EAE does not fully replicate the complexity of MS, its pathological similarities, and comparable mitochondrial alterations to those observed in the MS cerebellum suggest that it serves as a valuable model. When combined with postmortem MS brain tissue analysis, EAE can help advance our understanding of MS and other neurodegenerative diseases.

Summary of PC Loss in MS/EAE: ( A ) Under normal conditions, myelinated PCs receive excitatory and inhibitory inputs. In peak EAE, demyelination, inflammation, and astrogliosis induce PC stress, leading to axon damage and initial PC loss. In chronic EAE, persistent damage results in increased axon blebbing, swelling, and further PC and parallel fiber loss. ( B ) Mitochondria redistribute to meet rising energy demands during demyelination, but sustained stress overwhelms them. Prolonged disease leads to COXIV loss, reduced energy production, and PC degeneration. Image credit: Created in BioRender. Feri, M. (2025) https://BioRender.com/t12njs8 .

Materials and Methods

Ethics Statement.

All mouse studies and mouse procedures were conducted in accordance with ARRIVE guidelines, NIH guidelines and approved by the Institutional Care and Use of Laboratory Animals Committee (IACUC) at the University of California, Riverside (UCR; approved animal welfare assurance #123). Mice were maintained in an Association for Assessment and Accreditation of Laboratory Animal Care-accredited facility under 12 h light/dark cycle and fed standard mouse chow with unrestricted access to food and water.

Human Postmortem Cerebellar Brain Section Preparation.

Human cerebellar sections were obtained from the NIH NeuroBioBank and Cleveland Clinic. Sample details are in SI Appendix , Table S1 . IHC and PC puncta counting followed published methods ( 59 , 60 ).

Mice and Experimental Fate.

Mice were divided into normal (n = 8 × 3) and MOG EAE-induced (n = 8 × 3) groups. They underwent a gait test ( 37 ) before perfusion for IHC/electron microscopy or killing for cerebellar homogenate/mitochondria isolation at EAE days 21, 40, and 60 (n = 8 each). Experimental details are in refs. 30 , 35 , 44 , and 61 and SI Appendix , Table S2A .

Cerebellar Mitochondria Isolation and Seahorse Oxygen Consumption Rate Mito Stress Test, and Western Blot.

Mitochondria isolation followed the method in ref. 62 , as outlined in SI Appendix . Complex-I-dependent mitochondrial respiration using the Seahorse XFp Analyzer and Western Blot analysis details are in ref. 35 and SI Appendix , Table S2B .

Statistics.

Graphs represent mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 using Student’s t test ( Figs. 1 , 2 , 4 C , and 7 ), unbalanced two-way ANOVA ( Fig. 3 A ), and one-way ANOVA with Bonferroni’s or Kruskal–Wallis tests ( Figs. 3 B , 4 A and B , 5 , and 6 ). Analysis details are in SI Appendix .

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Acknowledgments

Brandon Poole isolated cerebellar RNA and Alyssa Anderson assisted with the NanoString experiment. Anna Khalaj prepared samples for electron microscopy. This work was supported by NIH Grants NIH R01 NS081141- 01A1 to S.K.T.-W., National Multiple Sclerosis Society Grant RG-1901-33349, RG-2206-39734, RG-2110-38560, and the NSF graduate research fellowship to MS under Grant No. DGE-1326120.

Author contributions

K.C.A. and S.K.T.-W. designed research; K.C.A., S.D., M.F., M.T.S., M.O., S.S., W.R., B.B., and S.K.T.-W. performed research; K.C.A., S.D., M.F., M.T.S., S.N., and S.K.T.-W. analyzed data; and K.C.A., S.D., M.F., and S.K.T.-W. wrote the paper.

Competing interests

The authors declare no competing interest.

Footnotes

This article is a PNAS Direct Submission.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information .

Supporting Information

References

Associated Data

Supplementary Materials

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data Availability Statement

All study data are included in the article and/or supporting information .