Множественный вариант жесткости склероза в DYSF-ZNF638 связует с нейронной потерей и воспалением.

Introduction

Multiple sclerosis (MS) is an inflammatory disease of the central nervous system (CNS). Clinically, MS is characterized by progressive accumulation of neurological disability together with inflammatory attacks resulting in neurological deficits which may be transient or can contribute to disability accumulation. 1 Pathologically, MS is partly characterized by focal, demyelinating lesions throughout the CNS. 2 These lesions typically display neuroinflammatory characteristics, such as activated microglia/macrophages and the presence of (resident) lymphocytes. 2 , 3 , 4 , 5 MS lesions may also display neurodegenerative components, 6 such as neuro-axonal loss and acute axonal stress, the extent of which is associated with the presence of myelin laden foamy microglia. 7

To date, over 200 genetic variants have been associated with increased susceptibility for MS. Most of these associated loci are related to immunological pathways. 8 In line with this, immune-modulating and -depleting therapies are efficacious in preventing early attacks of MS. However, especially in advanced disease, there is an unmet need for therapies capable of halting and reversing progression of the disease due to limited knowledge of underlying neurodegenerative processes driving disability progression. 9

Recently, a genetic locus was reported that affects disease severity. A genome-wide association study (GWAS) of 12,584 people with MS identified homozygous carriership of the A allele at variant rs10191329 in the DYSF - ZNF638 locus, hereafter referred to as rs10191329 AA , to associate with a 3.7-year shorter median time to require a walking aid. Homozygous carriers in the MS cohort of the Netherlands Brain Bank (NBB) showed a more extensive demyelinating disease as evidenced by an increased brainstem lesion load and cortical lesion rate. 10 This indicates the much closer distance between genetics and pathology as compared to clinical phenotype and provided early validation that the association of rs10191329 with MS severity is biologically relevant. This is reinforced by earlier investigation which showed that people with rs10191329 AA also presented with increased brain atrophy on magnetic resonance imaging (MRI). 11 Importantly, a better understanding of the biological implications of this variant on MS pathology could reveal essential underlying mechanisms of therapy-resistant disease progression in MS. Here, we analyzed the NBB MS cohort using immunohistochemistry and gene expression analysis to further characterize the pathological features associated with rs10191329 and potentially uncover biological mechanisms driving disease progression.

Results

rs10191329 AA variant associates with foamy microglia morphology

Within the NBB MS autopsy cohort ( n  = 290), we identified 6 (2.1%) homozygous rs10191329 A allele carriers, which is in line with a reported minor allele frequency of 0.17 in people with MS. 10 Clinical and pathological characteristics between rs10191329 AA carriers, homozygous non-carriers (hereafter rs10191329 CC ), and heterozygous NBB-MS donors are presented in Table 1 . Although not significant, point estimates were consistently suggesting a more severe disease course in rs10191329 AA donors compared to all other donors, including on average a 6.16-year younger age of death ( p  = 0.22), a 9.11-year shorter disease duration ( p  = 0.17), a 6.18-year shorter period from onset of symptoms to reach expanded disability status scale (EDSS)6 ( p  = 0.18), and a 1.78-year shorter period from EDSS6 to EDSS8 ( p  = 0.42).

Table 1 Donor demographics, clinical characteristics, and pathological characteristics of variant rs10191329 in the Netherlands Brain Bank MS cohort C:C A:C A:A p -value Donor demographics (C:C n  = 155, A:C n  = 68, A:A n  = 6) Age (years) 62.83 (12.75) 64.63 (13.62) 56.67 (12.03) 0.22 Sex (F/M) 97/58 43/25 6/0 0.15 PMD (hours) 8:38 (2:45) 7:45 (2:08) 9:36 (1:03) 0.25 pH of CSF 6.47 (0.29) 6.52 (0.35) 6.50 (0.48) 0.92 Weight of brain (g) 1183.76 (8.05) 1193.65 (130.14) 1138.60 (119.72) 0.94 Clinical characteristics (C:C n  = 155, A:C n  = 68, A:A n  = 5) MS type (R/PP/SP) 10/47/95 4/20/38 0/1/5 0.45 Duration of disease (years) 30.44 (16.53) 31.86 (14.94) 21.33 (10.76) 0.17 Age at onset (years) 28.54 (19.57) 31.36 (19.68) 35.33 (7.87) 0.81 Years from onset to EDSS6 15.38 (11.62) 16.61 (12.70) 9.20 (5.63) 0.18 Years from EDSS6 to EDSS8 6.78 (5.26) 8.38 (6.64) 5.00 (2.45) 0.42 Pathological characteristics (C:C n  = 146, A:C n  = 64, A:A n  = 6) Lesion load 13.29 (11.55) 10.29 (8.42) 21.67 (10.84) 0.01 Reactive site load 2.14 (3.30) 3.12 (6.92) 2.17 (2.71) 0.97 Proportion active 0.20 (0.21) 0.20 (0.22) 0.26 (0.20) 0.63 Proportion mixed 0.29 (0.24) 0.28 (0.26) 0.40 (0.22) 0.20 Proportion active & mixed with ramified microglia 0.23 (0.23) 0.21 (0.18) 0.17 (0.12) 0.01 Proportion active & mixed with foamy microglia 0.12 (0.18) 0.13 (0.20) 0.29 (0.25) 0.04 Proportion inactive 0.32 (0.28) 0.32 (0.27) 0.26 (0.18) 0.46 Proportion remyelinated 0.32 (0.24) 0.32 (0.27) 0.20 (0.18) 0.29 Cortical lesion rate 12.17 (16.80) 9.11 (12.71) 49.67 (40.69) 0.003 F, female; M, male; PP, primary progressive; R, relapsing; SP, secondary progressive. Significance tested between rs10191329 CC and rs10191329 AC/CA versus rs10191329 AA donors. Proportions were tested for significance using a quasi-binomial regression, categorical data using a chi-square test, other data using a quasi-Poisson regression.

Donor demographics, clinical characteristics, and pathological characteristics of variant rs10191329 in the Netherlands Brain Bank MS cohort

F, female; M, male; PP, primary progressive; R, relapsing; SP, secondary progressive. Significance tested between rs10191329 CC and rs10191329 AC/CA versus rs10191329 AA donors. Proportions were tested for significance using a quasi-binomial regression, categorical data using a chi-square test, other data using a quasi-Poisson regression.

rs10191329 AA donors skew toward a 1.30-times higher proportion of active ( p  = 0.63) and a 1.40-times higher proportion of mixed lesions ( p  = 0.20) compared to rs10191329 CC donors, although not significant. The average proportion of inactive ( p  = 0.46) and remyelinated ( p  = 0.29) lesions were comparable between groups. Most prominently, there was a 2.42-times higher abundance of active and mixed lesions with foamy microglia/macrophages in rs10191329 AA donors ( p  = 0.04) and, correspondingly, a 1.35-times lower proportion of lesions with ramified microglia/macrophages morphology ( p  = 0.01). In accordance with a more severe clinical disease and more extensive pathology earlier associated with rs10191329, these results indicate that there is more extensive phagocytosis by myeloid cells in carriers with a rs10191329 AA genotype.

Increased neuro-axonal damage and lymphocyte infiltration in rs10191329 AA donors

We next further characterized the impact of rs10191329 on MS-associated neuro-axonal damage and on lymphocyte infiltration in normal appearing white matter (NAWM) in rs10191329 AA and matched rs10191329 CC donors ( Figure 1 A). Although axonal density, measured with neurofilament (SMI312), was comparable ( p  = 0.50), there was a higher frequency of amyloid precursor protein (APP) + bulbs/axons in pyramidal tract NAWM of rs10191329 AA donors compared to rs10191329 CC donors, reflecting acute axonal stress ( p  < 0.001). The density of CD3 + T cells was similar in rs10191329 AA and rs10191329 CC donors ( p  = 0.25), whereas CD79A + B cell presence was increased ( p  < 0.001). Lastly, in medial-frontal gyrus NAWM, the density of SOX10 + oligodendrocytes was comparable between rs10191329 AA and rs10191329 CC donors ( p = 0.79). Next, we compared the extent of acute axonal damage and infiltration of lymphocytes in mixed lesions ( Figure 1 B). Similar to NAWM, in rs10191329 AA compared to rs10191329 CC donors, mixed active/inactive (mixed) lesions displayed more normalized APP + bulbs/axons ( p  = 0.01). Accordingly, although non-significant, rs10191329 AA donors showed a higher level of neurofilament light chain (NfL) Z scores in cerebrospinal fluid (CSF) samples ( Figure S1 A; p  = 0.339), which was strongly positively correlated with the proportion of foamy (active and mixed) lesions ( Figure S1 B; R 2  = 0.68, p  = 0.043). There was a higher abundance of CD3 + T cells ( p  = 0.03), but not clearly of CD79A + B cells ( p  = 0.27) in rs10191329 AA donors compared to rs10191329 CC donors, although with substantial variation. In normal appearing gray matter (NAGM) ( Figure 1 C), rs10191329 AA donors had a lower neuronal density measured with NeuN compared to rs10191329 CC donors, although not significant in all layers ( p = 0.09), this was a robust observation in layer 2 specifically ( p  = 0.002). The density of oligodendrocytes measured with SOX10 was comparable ( p  = 0.40). Taken together, rs10191329 AA donors had a higher propensity of neuro-axonal damage in NAWM, mixed lesions and NAGM. This is in line with the higher lesion rate and higher abundance of foamy myeloid cells in lesions.

Homozygous carriers of rs10191329 AA have more severe neuro-axonal pathology and increased lymphocyte infiltration

Panels show quantifications stratified for matched rs10191329 AA compared to rs10191329 CC donors.

(A) Left to right show distribution in NAWM of axonal density as measured by surface area of SMI312 (mm 2 /mm 2 ), acute axonal stress as measured by APP + bulbs/axons (n/mm 2 ), lymphocyte density of CD3 + T cells and CD79A + B cells (cells/mm 2 ), and density of SOX10 + oligodendrocytes (cells/mm 2 ). Scale bars indicate 50 μm

(B) Left to right show distribution in mixed lesions of acute axonal stress as measured by APP + bulbs/axons normalized to NAWM (n/mm 2 ), density of CD3 + T cells and CD79A + B cells (cells/mm 2 ). Scale bars indicate 50 μm

(C) Left to right show distribution in NAGM of neuronal density (NeuN + cells/mm 2 ), and SOX10 + oligodendrocytes (cells/mm 2 ). Scale bars indicate 100 μm and 50 μm, respectively. Data are represented as individual data points in a violin plot with a bar indicating the mean. SMI312 stainings were tested for significance with a quasi-binomial regression. Quantifications in mixed lesions were tested for significance using a mixed generalized model with a quasi-Poisson distribution. Other quantifications were tested for significance with a quasi-Poisson regression. p values are shown in the plots.

See also Figure S1 .

Flanking regions of rs10191329 associate with oligodendrocyte presence in MS

rs10191329 is a methylation quantitative trait locus (QTL) in the promotor region of Dysferlin ( DYSF ), and showed a correlation with other expression and splicing QTLs for Zinc Finger Protein 638 ( ZNF638 ). 10 , 12 Thus, we explored the association of these genes and their products with MS white matter pathology. 13 In active lesions, expression of DYSF and ZNF638 was lower compared to NAWM from the same sample ( p  < 0.001 and p  = 0.02, respectively; Figure 2 A). Through deconvolution, we determined the cell composition per sequenced sample. The expression of DYSF and ZNF638 strongly correlated positively with the proportion of oligodendrocytes ( DYSF: R = 0.8, p  < 0.001 ; ZNF638: R = 0.71, p  < 0.001), and negatively with microglia ( DYSF : R = −0.7, p  < 0.001; ZNF638 : R = −0.64, p  < 0.001), astrocyte ( DYSF : R = −0.73, p  < 0.001; ZNF638 : R = −0.7, p  < 0.001), and endothelial cell ( DYSF : R = −0.34, p  = 0.003; ZNF638 : R = −0.2, p  = 0.1) proportion ( Figure 2 B). Accordingly, previous single-nucleus RNA sequencing (RNA-seq) study of post-mortem brain from MS patients and controls also showed the expression of ZNF638 and DYSF to be highest in oligodendrocytes and, to a lesser extent, in neurons ( Figure S2 ). 14 To confirm that RNA expression translated to the protein level, we performed immunohistochemistry in NAWM and verified that SOX10+ oligodendrocytes expressed DYSF and ZNF638 proteins ( Figure 2 C). Additionally, in NAGM, DYSF, and ZNF638 proteins also colocalized with NeuN ( Figure 2 D). This confirms the protein expression of DYSF and ZNF638 in oligodendrocytes and neurons.

Expression of ZNF638 and DYSF in lesions and normal-appearing tissue

(A) DYSF and ZNF638 gene expression in active lesions compared to NAWM. Data are represented as individual data points in a violin plot with a bar indicating the mean.

(B) Correlations between cell type presence assessed by cell type deconvolution and gene expression of DYSF and ZNF638 . DYSF and ZNF638 associates with oligodendrocytes. Data are represented as individual data points with a linear regression line and 95% confidence interval. Data were acquired from Chen et al., Brain 2025. 13

(C and D) Immunofluorescence image of DYSF and ZNF638 double labeled (C) with SOX10, a pan-lineage oligodendrocyte marker, in NAWM and (D) with NeuN, a neuronal cell marker, in NAGM. All scale bars indicate 15 μm. Changes in gene expression between NAWM and active lesions were tested using t tests. The relationships between gene expression and cell type deconvolution were tested using a Pearson correlation. See also Figure S2 .

Increased expression of flanking genes DYSF and ZNF683 in rs10191329 AA donors

Subsequently, we investigated whether there was a difference between rs10191329 AA and rs10191329 CC donors in expression of flanking genes. In NAWM, the density of cells expressing DYSF protein was comparable ( p  = 0.57). In rs10191329 AA compared to rs10191329 CC donors, we observed non-significant, but numerically higher density of ZNF638 + cells ( p  = 0.072; Figure 3 A), which was validated by an increased ZNF638-positivity in SOX10 + oligodendrocytes ( p  = 0.02; Figure 3 B). In mixed lesions, the density of DYSF + cells and ZNF638 + cells were comparable between rs10191329 AA and rs10191329 CC donors ( p  = 0.85, p  = 0.57, respectively; Figure 3 C). Contrastingly, in NAGM, the density of DYSF + cells was higher in rs10191329 AA compared to rs10191329 CC donors ( p  = 0.01), and the density of ZNF638 + cells was comparable between groups ( p  = 0.34; Figure 3 D). Increased number of DYSF + cells was not associated with an increased percentage of NeuN + neuronal cells nor SOX10 + oligodendrocytes expressing DYSF ( p  = 0.88, p  = 0.42, respectively; Figure 3 E). In cortical GM lesions, the density of DYSF + cells and of ZNF638 + cells was comparable between donors ( p  = 0.78, p  = 0.633, respectively; Figure 3 F). The altered abundance of flanking genes implicates ZNF638 and DYSF as potential molecular mediators.

Expression of DYSF and ZNF638 by oligodendrocytes and neurons

Panels show quantifications stratified for matched rs10191329 AA compared to rs10191329 CC donors.

(A and B) (A) Left to right show distribution in NAWM of DYSF + and ZNF638 + cell density (cells/mm 2 ) and (B) percentage ZNF638 + of SOX10 + oligodendrocytes. Arrowhead indicates a ZNF638 + SOX10 + cell.

(C) Left to right show distribution in mixed lesions of DYSF + and ZNF638 + cell density (cells/mm 2 ).

(D) Left to right show distribution in NAWM of DYSF + and ZNF638 + cell density (cells/mm 2 ).

(E) Percentage of DYSF + cells among SOX10 + oligodendrocytes or NeuN + neurons.

(F) Left to right of distribution in cortical lesions of DYSF + and ZNF638 + cell density (cells/mm 2 ). All data are represented as individual data points in a violin plot with a bar indicating the mean. Immunohistochemistry quantifications were tested for significance using a quasi-binomial generalized linear model. Scale bars of images indicate 50 μm or 15 μm (fluorescent magnifications). p values are shown in the plots.

Mitochondrial changes in rs10191329 AA donors

To further investigate potential driving factors of the more severe disease course, we isolated and sequenced Olig2 + and NeuN + nuclei from normal-appearing superior temporal gyrus derived white and gray matter of rs10191329 AA ( n  = 5) and rs10191329 CC ( n  = 6) donors ( Figures 4 A, 4B, and S3 A–S3C). Principal-component analysis (PCA) shows that the most variation was explained by cell type ( Figure 4 C). First, we explored expression of genes flanking rs10191329 in NeuN + neuronal and Olig2 + oligodendrocytic nuclei, and found no apparent differences in DYSF and ZNF638 transcript abundance between rs10191329 AA and rs10191329 CC donors ( Figure 4 D). Repeating the PCA on each nuclear subset separately, most variation was explained by genotype for rs10191329 ( Figures 4 E, 4G, S3 D, and S3E). Yet, little significant differential gene expression between rs10191329 AA and rs10191329 CC donors was detected. In Olig2+ nuclei, TOGARAM2 and RMRP were significantly downregulated in rs10191329 AA donors, while MT-ND4 was upregulated ( Table S1 ); differential gene expression analysis in NeuN+ nuclei showed no significantly differentially expressed genes after multiple testing correction. Pathway enrichment analysis showed a significant enrichment for mitochondrial genes in both Olig2 + (false discovery rate [FDR] <0.001) and NeuN + nuclei (FDR <0.001), and specifically genes related to Leber’s hereditary optic neuropathy (LHON) in rs10191329 AA donors (Olig2: FDR = 0.019; NeuN: FDR = 0.093; Figures 4 F, 4H, S3 F, S3G, and Table S1 ).

Whole transcriptome profiling of NeuN + and Olig2 + nuclei rs10191329 AA and rs10191329 AA donors

Bulk RNA-seq was performed on isolated nuclei from normal appearing superior temporal gyrus of n  = 5 rs10191329 AA and n  = 8 rs10191329 AA donors.

(A) Representative dot plot showing flow cytometry gating strategy used for isolating nuclear subsets. Numbers indicate the percentage of events within the shown gates.

(B) Expression of genes on which was gated during isolation ( RBFOX3 /NeuN and OLIG2 ). Data are represented as individual data points with a bar indicating the mean.

(C) PCA plot of all samples, where the first principal component correlates with cell type.

(D) Gene expression of flanking genes ( DYSF and ZNF638) of rs10191329. Data are represented as individual data points with a bar indicating the mean.

(E) PCA plot of Olig2 + nuclei, where the first principal component correlates with rs10191329 genotype.

(F) Volcano plot of Olig2 + nuclei of rs10191329 AA versus rs10191329 CC donors, showing gene set enrichment analysis (GSEA) for mitochondrial genes and genes related to LHON. Points in red are part of the respective gene set. Significant genes not present in these gene sets are shown in blue.

(G) PCA plot of NeuN + nuclei, where the first principal component correlates with rs10191329 genotype.

(H) Volcano plot of NeuN + nuclei of rs10191329 AA vs. rs10191329 CC donors, showing GSEA for mitochondrial genes and genes related to Leber’s hereditary optic neuropathy. Points in red are part of the respective gene set. PCA was performed on counts-per-million that were corrected for number of nuclei input and estimated proportional cell type abundance using deconvolution. GSEA was performed using “CAMERA” from the limma R package. CPM = counts-per million; PC = principal component.

See also Figure S4 .

Discussion

Here, we explored the impact of rs10191329, a variant associated with MS progression, in a nested case-control study comparing homozygous rs10191329 AA carriers to rs10191329 CC carriers. After previous observation of an increased presence of focal demyelinating lesions in carriers of rs10191329 AA , 10 we now report (1) more immune activation in rs10191329 AA donors as indicated by an increased B cell presence in NAWM, and an increased frequency of foamy macrophages and T cells in mixed lesions; (2) more prominent neurodegeneration in rs10191329 AA donors as characterized by lower neuronal density in cortical layer 2, more acute axonal stress, and by CSF NfL levels correlating with foamy macrophage presence; (3) a positive correlation of flanking genes DYSF and ZNF638 with oligodendrocyte presence in MS WM lesions; and (4) a higher percentage of oligodendrocytes expressing ZNF638 in NAWM, and a higher density of DYSF + cells in NAGM in rs10191329 AA donors. Nuclear RNA sequencing (5) did not show any changes in DYSF or ZNF638 expression but rather an upregulation of mitochondrial genes in rs10191329 AA donors. These data suggest that an increased susceptibility to tissue damage and inflammation could drive the association between the MS severity-associated variant rs10191329 and increased myeloid cell activation. Although its molecular mediators remain to be further studied, DYSF , ZNF638 , and mitochondrial pathways are proposed as candidates affecting MS severity.

Since other studies could not replicate the association of this genetic variant to MS severity, 15 , 16 possibly due to issues with power, 17 investigating the underlying mechanisms of this genetic variant is essential. Previously, an increased abundance of foamy microglia/macrophages has been associated with increased axonal damage and acute axonal stress in MS lesions in brain donors as measured with CSF NfL chain levels. 7 Our current replication of this observation in rs10191329 AA/CC donors reiterates the importance of rs10191329 as a variant relevant for MS severity. Moreover, the increase in axonal damage and decreased neuronal density in itself illustrates the skewing toward a more severe disease course in rs10191329 AA donors. It is known that clearance of myelin by anti-inflammatory microglia/macrophages is a prerequisite for remyelination. 18 However, excessive myelin uptake, as is the case in foamy microglia, can be pro-inflammatory. 3 , 19 A more pro-inflammatory phenotype of these brain-resident foamy phagocytes in rs10191329 AA donors is likely, as we found a higher lesion load, no difference in proportion of remyelinated lesions, and a positive correlation of CSF NfL levels with a higher proportion of foamy microglia/macrophages. This confirms previous reports of increased MRI brain atrophy and serum NfL-levels in living rs10191329 AA subjects. 11 , 20 In addition, an increased amount of NAWM CD79A + B cells and CD3 + T cells in mixed lesions of rs10191329 AA donors also indicates a more pro-inflammatory environment in rs10191329 AA donors. Taken together, rs10191329 associates with a higher propensity for tissue damage and increased immune activation.

Better understanding of the mechanism underlying this association is vital to better understand and treat disability progression in MS. Regarding the flanking genes, we have confirmed the gene and protein expression of DYSF and ZNF638 in neurons and oligodendrocytes. We also report an increase of these proteins in normal-appearing tissue of rs10191329 AA donors. Increased DYSF expression in NAGM could not be attributed to a specific cell type, whereas increased expression of ZNF638 in NAWM of rs10191329 AA donors is associated with an increase in ZNF638 + oligodendrocytes. These observations suggest that both molecules could be relevant to better understand the impact of rs10191329 on MS severity. The association of these molecules with neurons and oligodendrocytes primarily suggests an impact on susceptibility of these cell types to MS-inflicted tissue damage, although immune-regulatory properties have also been attributed to oligodendrocytes. 21 We have not explored whether rs10191329 can have effects more distal than its flanking genes or other genes within the locus.

DYSF is well known for its expression in muscle cells, where a deficiency may cause muscular dystrophy. 22 As a general function, DYSF is a universal mediator of membrane repair in response to damage. 23 , 24 , 25 We do not note an alteration of DYSF expression in NAWM, but only general increased expression in NAGM without specificity for neurons. As a parallel, DYSF aggregates are seen in association with neuritic plaques in Alzheimer disease, 26 and these aggregates are hypothesized to form due to the inability of neurons to repair membrane damage. Contrastingly, in the highly inflammatory experimental autoimmune encephalomyelitis (EAE) model of neuroinflammation, presence or absence of DYSF did not affect clinical course. 27 However, this model is likely to trigger qualitatively and quantitatively different mediators of tissue damage compared to the compartmentalized inflammation as is seen in progressive MS, and may therefore not address a similar biological role of DYSF in progressive MS.

The other flanking gene, ZNF638 , is a transcription factor implicated in transcriptional silencing in association with the human silencing hub (HUSH) complex. 28 , 29 Up to now, this has been functionally described in retroviruses and adeno-associated viruses. This could be an important mechanism of action of the severity variant, especially in light of the involvement of Epstein-Barr virus (EBV) in MS pathogenesis. 30 , 31 That we do not see our observed increase in protein level of ZNF638 at the transcript level make our findings less certain. Still, it has to be noted that there are a couple of factors obscuring the RNA-seq data. First, oligodendrocyte nuclei were sampled from both NAWM and NAGM, whereas an effect was only to be seen in NAWM at protein level. Furthermore, nuclear RNA mostly reflects transient and ongoing transcription and only reflects a small proportion of all RNA within a cell. 32 Possible changes in abundance or function of ZNF638 can still indicate disease relevant consequences: (1) a higher need for viral silencing, although this would not be a logical consequence of a genetic variant or (2) a failure to silence viral DNA implicated in MS. In light of rs10191329 correlation with sQTL for ZNF638 , this could possibly be due to alternative splicing. Our data suggest that functional differences of the variant might present in oligodendrocytes, and the mechanism of action needs further investigation.

Another possibility is that effects of rs10191329 are not a consequence of changes in abundance or function of the proximal genes of this variant. Bulk nuclear RNA sequencing suggests mitochondrial gene expression as differentiating rs10191329 AA and rs10191329 CC donors. Frequently in RNA-seq studies, high levels of mitochondrial RNA are associated with sample quality or stressed cells. Due to the randomized and balanced way in which nuclei and RNA isolations as well as sequencing were performed, and that quality metrics, such as DV200, were comparable between groups, this likely reflects a biological effect. Since we performed nuclear RNA-seq, this means that in rs10191329 AA donors there could be more translocation of mtRNA to the nucleus. Indeed, it has been described in endothelial cells of human diabetic donors that mtRNAs translocate to the nucleus in a cell stress- and disease-dependent manner. 33 When mitochondrial genes are simply upregulated due to a higher abundance of mitochondria, this could indicate a higher energy demand. Alternatively, mitochondrial donation of microglia to neurons has recently been identified to occur in disease burdened neurons in order to attenuate oxidative stress. 34 Generically higher activity of mitochondrial complex I and III can generate increased reactive oxygen species (ROS) production which is damaging to cells and tissues. 35 As such, mitochondrial changes have been described to associate with MS pathology. 36 For instance, changes in axonal mitochondrial frequency positively correlates with the amount of active and phagocytic microglia in the optic nerve of people with MS. 37 Furthermore, in acute human MS lesions there are signs of focal axonal degeneration, often associated with mitochondrial pathology. 36 In the EAE model, this was reversible by neutralization of reactive oxygen and nitrogen species. 38 More evidence from the EAE model shows that (1) at onset of neurological symptoms, respiratory chain complex I activity is compromised in the spinal cord, and (2) mitochondrial dysfunction coincides with the presence of CD45 + perivascular macrophages. 39 In line with our other findings, the increased presence of mitochondrial genes describes an worsened MS disease course in rs10191329 AA donors.

Of special interest is the finding that mitochondrial genes related to LHON are upregulated in rs10191329 AA donors. LHON is a mitochondrial genetic disease characterized by focal degeneration of the retinal ganglion cell layer and the optic nerve. 40 In LHON, causative genetic variants cause mitochondrial dysfunction. 41 , 42 The selectivity for optic nerve degeneration is hypothesized to exist because the retinal ganglion cells have the highest demand for mitochondrially produced energy. 43 Clinically, case series of people with both MS and LHON have been published. Although the co-occurrence of these two distinct diseases has been described to likely be due to chance, MS tends to develop more severely in people with both MS and LHON. 44 Moreover, LHON sometimes presents with an MS-like (LMS) disease with associated white matter lesions. These LMS-lesions present more often in females with LHON and are not distinguishable to MS white matter lesions on MRI. 45 Overall, this implies a mechanistic interaction between the MS and LHON. Since mitochondrial RNA is increased in cells affected by LHON, 46 this gene-association could indicate a similar mitochondrial dysfunction in MS, where an unmet energy demand could cause a more rapid decline in people with MS. Overall, this supports the theory that rs10191329 AA carriers of the severity variant would be more vulnerable to neurodegeneration coinciding with either the inability to provide enough energy to neurons or an abundance of ROS production. Whether these mitochondrial aberrancies are a cause or effect of accelerated disease course remains to be determined.

In conclusion, we show that the rs10191329 variant is associated with more extensive pathological changes in MS, such as an increased susceptibility to tissue damage and increased myeloid cell activation and disclose both DYSF and ZNF638, as well as mitochondrial pathways as potential candidates underlying these effects. In sum, we have further biologically validated the findings of the recently discovered MS-severity variant rs10191329. These findings may open new avenues for ameliorating MS progression.

Limitations of the study

There are some limitations to this study. Due to the low minor allele frequency of 0.17 and the scarce available tissue, this study generally lacks power to investigate all components of MS pathology extensively. Nevertheless, investigation within the homogeneously sampled and processed Netherlands Brain Bank tissue collection limits confounding factors and thereby increases the sensitivity of our study. Additionally, investigating biological processes closely related to the pathophysiology of MS allows assessment of the impact of rs10191329 in a smaller sample size compared to the impact of this genetic variant on a more generic endpoint such as the age of using a walking aid. Another limitation is that the isolation of nuclei yielded low abundance of RNA, resulting in limited library sizes. Overall, the characterization as performed by the NBB 2 and the availability of genotyped high-quality brain tissue make this dataset unique and provide insight into the association of rs10191329 with MS severity and the pathophysiology of MS.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, dr. Joost Smolders ( j.j.f.m.smolders@erasmusmc.nl ).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Nucleus RNA-seq data deposited at GEO and are publicly available as of the date of publication. The Accession numbers of nuclei RNA sequencing is GSE273954 . Tissue RNA-seq data (Fig. 2AB) is available under accession number GSE283092 . This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We are grateful to the brain donors and their families for their commitment to the Netherlands Brain Bank donor program. We are grateful to dr. Aldo Jongejan for sharing his knowledge on RNAseq analysis. Funding for this research was obtained from the Blom-de Wagt Foundation (MS Research: 23–1208), MoveS (MS Research: 17–975), and MS Research (19–1079).

Author contributions

H.J.E., A.M.R.v.d.B., I.H., J.H., and J.S. contributed to the study concept; H.J.E., A.M.R.v.d.B., J.Q.A.C., C.-C.H., and M.-J.M. acquired and analyzed data; H.J.E. and A.M.R.v.d.B. drafted the figures; H.J.E., J.H., and J.S. verified and interpreted the underlying data; H.J.E., A.H., I.H., J.H., and J.S. wrote the manuscript; J.H. and J.S. supervised the project team; all authors reviewed and approved the manuscript for final publication.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples Human brain tissue and CSF Netherlands Brain Bank https://www.brainbank.nl Antibodies and secondary reagents Rabbit-anti-human Olig2 – Alexa Fluor® 647 Abcam Cat# ab225100 RRID: - Donkey-anti-mouse IgG – Cy3 Jackson ImmunoResearch Cat# 715-165-150 RRID: AB_2340813 Donkey-anti-rabbit IgG – Cy3 Jackson ImmunoResearch Cat# 711-165-152 RRID: AB_2307443 FcR-Blocking Reagent Miltenyi Biotec Cat# 130-059-901 RRID: AB_2892112 Goat-anti-human SOX10 R&D Systems Cat# AF2864 RRID: AB_442208 Horse-anti-goat-IgG – biotin Vector Laboratories Cat# BA-9500 RRID: AB_2336123 Horse-anti-mouse-IgG – biotin Vector Laboratories Cat# BA-2000 RRID: AB_2313581 Horse-anti-rabbit-IgG – biotin Vector Laboratories Cat# BA-1100 RRID: AB_2336201 Mouse-anti-human APP A4 Sigma-Aldrich Cat# MAB348 RRID: AB_94882 Mouse-anti-human CD79A DAKO Cat# M705001-2 RRID: - Mouse-anti-human Dysferlin Leica Cat# HAMLET-CE RRID: - Mouse-anti-human NeuN Sigma-Aldrich Cat# MAB377 RRID: AB_2298772 Mouse-anti-human NeuN – Alexa Fluor®488 Sigma-Aldrich Cat# MAB377x RRID: AB_2149209 Mouse-anti-human neurofilament marker BioLegend Cat# 837901 RRID: - Rabbit-anti-human CD3 DAKO Cat# A045201 RRID: - Rabbit-anti-human ZNF638 Sigma-Aldrich Cat# HPA036784 RRID: AB_10672478 Streptavidin – Alexa Fluor®488 Jackson ImmunoResearch Cat# 016-540-084 RRID: AB_2337249 Chemicals, peptides, and recombinant proteins 3,3′-diaminobenzidine (DAB) DAKO Cat# K5007 Biotinylated tyramide Sigma-Aldrich Cat# SML2135 Bovine serum albumin (BSA) Sigma-Aldrich Cat# A7906 cOmplete ULTRA tablets, Protease Inhibitor Cocktail Merck Cat# 5892953001 Dulbecco’s phosphate-buffered saline (DPBS) Thermo Fisher Scientific Cat# 14190094 Dithiothreitol (DTT) Thermo Fisher Scientific Cat# R0861 Entallan Sigma-Aldrich Cat# 1.07961 Haematoxylin Merck Cat# 4305 Hoechst 33342 Thermo Fisher Scientific Cat# H3570 Hydrogen peroxide Merck Cat# 107209 MgCl 2 Thermo Fisher Scientific Cat# AM9530G Mowiol Merkc Cat# 4705904 Sucrose Sigma-Aldrich Cat# 16104 Triton X-100 Sigma-Aldrich Cat# X-100 UltraPure™ 0.5M EDTA, pH8.0 Thermo Fisher Scientific Cat# 15575020 UltraPure™ 1M Tris-HCI, pH8.0 Thermo Fisher Scientific Cat# 15568025 Critical commercial assays DNeasy Blood & Tissue Kit Qiagen Cat# 69506 NF-light ELISA CE kit for CSF Uman Diagnostics Cat# 10-7001 RNEasy micro kit Qiagen Cat# 74004 SMARTer Stranded Total RNA-Seq Kit v3 - Pico Input Mammalian Library Prep Kit Takara Bio Cat# 634485 Vector Elite ABC kit Vector Laboratories Cat# PK-6100 RRID: AB_2336819 Deposited data Nuclei bulk RNAseq data This paper GEO: GSE273954 Tissue bulk RNAseq data Chen et al. 13 GEO: GSE283092 Software and algorithms Application Suite X Leica N/A Bioconductor (v3.13) Huber et al. 47 https://bioconductor.org/ biomaRt package (v.2.58.2) Durinck et al. 48 Bioconductor dtangle (v.2.0.9) Hunt et al. 49 https://cran.r-project.org/package=dtangle dupRadar (v1.12.1) Sayols et al. 50 Bioconductor edgeR package (v3.34.1, 3.42.4) Robinson and Oshlack 51 Bioconductor Emmeans package (v1.10.2) CRAN https://cran.r-project.org/package=emmeans fastp (v0.23.4), Chen et al. 52 https://github.com/OpenGene/fastp FastQC (v0.11.9 & v0.12.1) Babraham Institute https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ featureCounts (v2.0.6) Liao et al. 53 https://subread.sourceforge.net/featureCounts.html FlowJo (v10.10.0) BD Biosciences https://www.flowjo.com/ gencode.v45.primary_assembly.annotation.gtf Gencode https://www.gencodegenes.org/human/ ggplot2 package (v3.5.1) Wickham 54 https://cran.r-project.org/package=ggplot2 glmmTMB package CRAN https://CRAN.R-project.org/package=glmmTMB GRCh38.105 Ensembl https://www.ensembl.org/index.html GRCh38.primary_assembly.genome.fa (v.45) Gencode https://www.gencodegenes.org/human/ HISAT2 (v2.2.1) Kim et al. 55 http://daehwankimlab.github.io/hisat2/ Homo_sapiens.GRCh38.105.gtf Ensembl https://www.ensembl.org/index.html HTseq (v1.99.2) Anders et al. 56 https://htseq.readthedocs.io/en/master/ Fiji for ImageJ (v1.54f) Schindelin et al. 57 https://imagej.net/software/fiji/ limma package (v3.56.2) Ritchie et al. 58 Bioconductor lme4 package CRAN https://cran.r-project.org/package=lme4 PCAtools package (v2.14.0). N/A Bioconductor Qupath 0.4.0 Bankhead et al. 59 https://qupath.github.io/ R (v4.1.0 & v4.3.1) The R Project https://www.r-project.org/ Samtools (v1.20) Danecek et al. 60 https://www.htslib.org/ STAR2 (v2.7.11b) Dobin et al. 61 https://github.com/alexdobin/STAR Trimmomatic (v0.39) Bolger et al. 62 http://www.usadellab.org/cms/index.php?page=trimmomatic Other Axioscan Z1 Zeiss N/A Ensembl (v112) EMBL-EBI https://www.ensembl.org/index.html FACSAria II cell sorter BD Bioscience N/A Fragment Analyzer Advanced Analytical Technologies at GenomeScan B.V. N/A KASP genotyping platform LGC Genomics N/A Molecular Signatures Database (MSigDB, version 2023.2) Subramanian et al. 63 https://www.gsea-msigdb.org/gsea/msigdb/index.jsp NovaSeq6000 Illumina at GenomeScan B.V. N/A STEDYCON microscope Abberior Instruments N/A TCS SP8 confocal microscope Leica N/A

Experimental model and study participant details

Characterization of MS lesions and genotyping

Following informed consent, brain donors with pathologically confirmed MS that were recruited to the NBB since 1990 were included in this study ( Table 1 ). Autopsy procedures were approved by the Ethics Committee of the VU medical center in Amsterdam, the Netherlands. Of all donors, detailed clinical summaries were available.

As previously described, 64 blocks were dissected at standardized CNS locations (including the brainstem), with additional blocks targeted to MS lesions using macroscopic and post-mortem magnetic resonance imaging (MRI) assessment. Sections were stained for proteolipid protein (PLP) and human leukocyte antigen (HLA) for lesion characterization. 65 Each MS lesion was characterized blind to genotype status. Of each donor, the reactive site load and lesion load were calculated in the standardized CNS locations, and the proportions of active, mixed lesions, inactive lesions, and remyelinated lesions, the proportion of active and mixed lesions with foamy or ramified microglia, and the cortical lesion rate were calculated as described previously. 2 Shortly, all tissue blocks taken at autopsy were characterized for lesion type, and in case of active and mixed active/inactive lesions, the dominant microglia/macrophage morphology was registered.

Nested case-control cohort

For further investigation, all 6 rs10191329 AA donors were matched with 12 rs10191329 CC donors for age, sex, post-mortem delay, and pH of CSF for a nested case-control cohort ( Table 2 ). Further, donors were selected such that differences in clinical characteristics were as small as possible, where the only characteristic approaching significance is time between EDSS6 and EDSS8. Although unfortunate, this conveys the close relationship between genotype and disease severity. As there were no clinical and cohort-pathological effects of heterozygosity for rs10191329, these donors were not included in the cohort.

Table 2 Donor demographics, clinical characteristics, and pathological characteristics of the nested case-control design study C:C ( n  = 12) A:A ( n  = 6) p value Donor demographics Age (years) 62.33 (15.26) 56.67 (12.03) 0.41 Sex (F/M) 12/0 6/0 1.00 PMD (hours) 9:35 (3:36) 9:36 (1:03) 0.99 pH of CSF 6.32 (0.31) 6.50 (0.48) 0.46 Weight of brain (g) 1193.33 (115.19) 1138.60 (119.72) 0.41 Clinical characteristics MS type (PP/SP) 5/7 1/5 0.60 Duration of disease (years) 27.17 (9.80) 21.33 (10.76) 0.29 Age at onset (years) 35.17 (14.68) 35.33 (7.87) 0.98 Years from onset to EDSS6 12.83 (10.93) 9.20 (5.63) 0.38 Years from EDSS6 to EDSS8 10.67 (7.84) 5.00 (2.45) 0.08 Pathological characteristics Lesion load 17.58 (10.00) 21.67 (10.84) 0.46 Reactive site load 5.00 (4.31) 2.17 (2.71) 0.11 Proportion active lesions 0.30 (0.21) 0.26 (0.20) 0.70 Proportion mixed lesions 0.34 (0.21) 0.40 (0.22) 0.56 Proportion active & mixed lesions with ramified microglia 0.32 (0.16) 0.17 (0.12) 0.06 Proportion active & mixed lesions with foamy microglia 0.13 (0.12) 0.29 (0.25) 0.05 Proportion inactive lesions 0.23 (0.15) 0.26 (0.18) 0.76 Proportion remyelinated lesions 0.36 (0.20) 0.20 (0.18) 0.23 Cortical lesion rate 8.33 (11.16) 49.67 (40.69) 0.05 Proportions were tested for significance using a quasi-binomial regression, categorical data using a chi-square test, other data using a quasi-Poisson regression. F, female; M, male; PP, primary progressive; SP, secondary progressive.

Donor demographics, clinical characteristics, and pathological characteristics of the nested case-control design study

Proportions were tested for significance using a quasi-binomial regression, categorical data using a chi-square test, other data using a quasi-Poisson regression. F, female; M, male; PP, primary progressive; SP, secondary progressive.

Of selected donors, we collected formalin-fixed, paraffin-embedded (FFPE) tissue blocks from (1) the brainstem containing the pyramidal tract longitudinally (rs10191329 AA , n  = 5; rs10191329 CC , n  = 11), (2) mixed lesions (rs10191329 AA , n  = 5; rs10191329 CC , n  = 6) and (3) cortical lesions (rs10191329 AA , n  = 5; rs10191329 CC , n  = 5) from varying supratentorial regions, (4) NAWM (rs10191329 AA , n  = 5; rs10191329 CC , n  = 10) and NAGM (rs10191329 AA , n  = 5; rs10191329 CC , n  = 9) from a widely available standardly dissected region (the medial frontal gyrus) and (5) fresh-frozen tissue from NAWM and NAGM from another widely available standardly dissected region (the superior temporal gyrus; rs10191329 AA , n  = 6; rs10191329 CC , n  = 9).

Method details

Genotyping

Of 290 MS brain donors from the NBB, DNA was extracted from whole blood or frozen cerebellar tissue using the DNeasy Blood & Tissue Kit (QIAGEN), or, alternatively, from formalin-fixed, paraffin-embedded (FFPE) cerebellar tissue. Genotyping for rs10191329 was performed using the KASP genotyping platform (LGC Genomics).

Quantitative immunohistochemistry

Neuronal damage, immune components, and flanking regions of rs10191329 were assessed through immunohistochemistry (IHC) on 8-μm FFPE or fresh-frozen sections. After rehydration, antigen retrieval was accomplished with microwave treatment at 700 W. Sections were incubated in 3% H 2 O 2 in PBS for 15 minutes and subsequently incubated with a primary antibody overnight in blocking buffer (PBS + 10% normal horse serum + 1% bovine serum albumin + 0.5% TritonX) at 4°C ( Table 3 ). Sections were incubated with appropriate biotinylated secondary antibody, followed by avidin-biotin-horseradish peroxidase (HRP) complex (Vector Elite ABC kit; Vector Laboratories), before visualization with 3,3′-diaminobenzidine (DAB) and hematoxylin counterstaining.

Table 3 Antibody overview for immunohistochemistry Protein Company (catalog #) Host Clone RRID Dilution Fixation Antigen retrieval SMI312 Biolegend (837901) Mouse SMI312 NA 1:2,000 FFPE CB, pH6.0 APP Millipore (MAB348) Mouse A4 AB_94882 1:2,000 FFPE CB, pH6.0 NeuN Sigma (MAB377) Mouse A60 AB_2298772 1:100 FFPE PBS, pH7.6 CD3 DAKO (A045201) Rabbit Polyclonal NA 1:100 (FFPE) 1:250 (Frozen) FFPE 4% PFA CB, pH6.0 – CD79A DAKO (M705001-2) Mouse Polyclonal NA 1:200 FFPE CB, pH6.0 DYSF Leica (HAMLET-CE) Mouse HAM1/7B6 NA 1:50 FFPE CA, pH7.6 ZNF638 Sigma (HPA036784) Rabbit Polyclonal AB_10672478 1:100 FFPE CA, pH7.6 SOX10 R&D (AF2864) Goat Polyclonal AB_442208 1:200 FFPE CA, pH7.6 NeuN Sigma (MAB377X) Mouse A60 AB_2149209 1:500 FFPE CA, pH7.6 CA: citraconic anhydride; CB: citrate buffer; FFPE, formalin-fixed, paraffin-embedded; PBS, phosphate-buffered saline; PFA: paraformaldehyde; RRID, research resource identifier.

Antibody overview for immunohistochemistry

CA: citraconic anhydride; CB: citrate buffer; FFPE, formalin-fixed, paraffin-embedded; PBS, phosphate-buffered saline; PFA: paraformaldehyde; RRID, research resource identifier.

For immunofluorescence (IF) staining, sections were incubated with Cy3-conjugated secondary antibodies. NeuN was directly labelled with Alexa Fluor 488. For SOX10, sections were incubated with biotinylated secondary antibodies, biotinylated tyramide (Sigma-Aldrich), and streptavidin conjugated with Alexa Fluor 488. Sections were incubated with Hoechst 33342 (Thermo Fisher Scientific).

Brightfield images were taken using the Axioscan Z1 (Zeiss) at 20x magnification. For IF, confocal imaging was performed using a STED microscope (STEDYCON; Abberior Instruments) at 40x magnification. All images were analyzed using QuPath 0.4.0. 59 Images were excluded if quality was not sufficient for quantification. For example because of weak staining due to probable over-fixation, high amounts of irremovable background or difficulties annotating a lesioned area precisely. Pyramidal tracts were outlined based on anatomical location. Lesions were outlined based on HLA-PLP stains. For SMI312, a thresholder was used. APP + bulbs/axons were manually counted in lesions and NAWM of the pyramidal tract. In mixed lesions, APP + bulbs/axons were normalized to the number of APP + bulbs/axons in surrounding NAWM. A positive cell thresholder was used for quantification of NeuN, CD3, CD79A, DYSF, and ZNF638 stainings. Thresholds were set based on visual inspection. For SOX10, NeuN, ZNF638, and DYSF, positive cells were detected using a random trees-based classifier. NeuN + cells were counted in medial frontal gyrus NAGM or alternatively in fresh frozen sections from the superior temporal gyrus (rs10191329 AA , n  = 1; rs10191329 CC , n  = 1; no outliers). The number of CD3 + , a pan T-cell marker, and CD79A + , a pan B-cell marker, cells were quantified in lesions and pyramidal tract NAWM. DYSF and ZNF638 were quantified in lesions and in medial frontal gyrus NAWM and NAGM.

NfL ELISA

NfL was measured in CSF using the NF-light ELISA CE kit for CSF from Uman Diagnostics (Umea, Sweden, 10-7001). ELISA was performed according to the datasheet. Concentrations were calculated using the standard curve range 50-5,000 pg/ml. Donors with neuropathological signs of a concurrent neurodegenerative disorder or a cerebrovascular accident within a year prior to autopsy were excluded.

Tissue RNAseq analysis

Data was adapted from previous work, which isolated white matter lesions for bulk RNA sequencing. 13 Active lesions and adjacent NAWM were collected from n  = 96 frozen tissue blocks. RNA was extracted using the RNeasy Micro kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. Bulk RNA sequencing was performed on a NovaSeq6000 platform by GenomeScan (Leiden, the Netherlands).

Nuclei isolation and RNAseq

Nuclei were isolated from frozen normal appearing super temporal gyrus tissue from the case-control cohort (rs10191329 AA , n  = 5; rs10191329 CC , n  = 8), according to a previously described protocol. 66 First, 1mL ice-cold NF1 buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 5 mM MgCl2, 0.1 M sucrose, and 0.5% Triton X-100, 1x proteinase inhibitor) was added to 100 mg of brain tissue and homogenized. Afterwards, the homogenate was incubated in a total volume of 5 mL NF1 buffer for 30 minutes on ice and strained afterwards. Then, the homogenate was diluted in NF1 buffer to a total of 20 mL and a sucrose cushion (1 M sucrose solution with 10 mM Tris-HCl pH8.0, 3 mM MgCl2, and 1 mM DTT) was underlaid. The homogenate was centrifuged at 3,166g for 45 minutes at 4°C without braking, and all supernatant was carefully removed. The nuclei pellet was resuspended in 1 mL NF1 buffer and diluted to a total volume of 10 mL and centrifuged at 1,600g for 5 minutes at 4°C. Then the pellet was resuspended in 10 mL FANS buffer (DPBS, 1% BSA, 1 mM EDTA) and centrifuged 1,600g for 5 minutes at 4°C. The pellet was incubated in FcR blocking reagent for 30 minutes at 4°C and subsequently with Hoechst 33342, rabbit-anti-human-olig2-AF647 (ab225100; Abcam) and mouse-anti-human-NeuN-AF488 (MAB377x; Sigma-Aldrich) for 30 minutes. Single-positive NeuN + or Olig2 + nuclei were sorted on a BD FACSAria II cell sorter (BD Biosciences). Total RNA was isolated with the RNeasy MicroKit according to the manufacturers’ instructions. Samples were depleted for rRNA, and DV200 was measured as a quality control on a fragment analyzer (Advanced Analytical, Heidelberg, Germany). Samples were library prepped using the Takara SMARTer Stranded Total RNA-Seq Kit v3 - Pico Input Mammalian Library Prep Kit and sequenced with a depth of 60 million reads by GenomeScan (Leiden, the Netherlands) on an Illumina NovaSeq6000 platform. Visualization of flow cytometry was performed using FlowJo (v10.10.0).

Confocal imaging of nuclei

Nuclei of one pilot-sample were isolated as described above, spun down and diluted in mowiol. This was mounted on a glass slide and imaged using a Leica TCS SP8 confocal microscope and Application Suite X (Leica) ( Figure S3 B). Images were processed using ImageJ (v. 1.54f).

Quantification and statistical analysis

Clinical characteristics

Differences in clinical characteristics were assessed using linear regression, adjusted for sex, age at onset, and initial disease course. Pathological characteristics were tested for differences using a quasi-binomial regression, corrected for sex, age at onset, and initial disease course.

Immunohistochemistry

SMI312-positive area, density of DYSF + cells and ZNF638 + cells was tested using quasi-binomial regression. APP + bulbs/axons, NeuN + cells, SOX10 + cells, CD3 + cells, and CD79A + cells per mm 2 were tested for significance using quasi-Poisson or negative binomial regression, offset by area measured. Statistics were performed in R (v4.3.1) using key packages ggplot2, tidyverse, lme4, glmmTMB, and emmeans. p -values <0.05 were considered significant.

NfL measurements

We calculated z-scores for NfL using the following formula 67 , 68 : CSF NfL z − score = [ log 2 ( NfL − value ) − ( 6.661 + ( age × 0.045 ) ) ] / 0.736 . Differences between rs10191329 AA and rs10191329 CC donors were tested using a two-sided t-test. The correlation between proportion of lesions containing foamy macrophages and NfL z-scores were tested using Pearson correlation.

Tissue RNAseq

Shortly, Adapter sequences were trimmed with Trimmomatic (v0.39), and alignment was performed against human reference genome GRCh38.105 using default parameters of HISAT2 (v2.2.1). Quality control was performed using FastQC (v0.11.9) and dupRadar (v1.12.1). Counts were obtained using HTseq (v1.99.2) with the ‘Homo_sapiens.GRCh38.105.gtf’ file. Counts were analysed in R (v4.1.0) with Bioconductor (v3.13). Genes with more than two count-per-million reads (CPM) in at least six samples were kept. Count data were normalized using the trimmed mean of M-values (TMM) method 51 (edgeR package, v3.34.1) and transformed to log2-counts per million (logCPM). Differential gene expression of DYSF and ZNF638 was determined using a t-test.

Nuclei RNAseq

Reads were first clipped for adapters using fastp (v0.23.4), and mapped using STAR2 (v0.23.4) using the settings: --outFilterMismatchNmax 2 --outFilterScoreMinOverLread 0 --outFilterMatchNminOverLread 0 --outFilterMatchNmin 60. Separate fastq-files from the same sample were merged after mapping using samtools (v1.20). Quality control was performed using FastQC (v0.12.1). Counts were obtained using featureCounts (v2.0.6) with the ‘gencode.v45.primary_assambly.annotation.gtf’ file. Counts were analyzed in R (v4.3.1). One sample (Olig2 – A:A) was removed, due to having the lowest library size and no olig2 expression. The eventual median library size was 3,2 million counts. Genes with more than two count-per-million reads (CPM) in at least two samples were kept. Count data were transformed to log2-counts per million (logCPM), normalized using the trimmed mean of M-values (TMM) method 51 (edgeR package, v3.42.4) and precision weighted using voom (limma package, v3.56.2). Remaining genes were reannotated with the biomaRt package (v2.58.2), using Ensembl (v112). PCA was performed on logCPM values of the 500 most variable genes using the PCAtools package (v 2.14.0). For both Olig2 + and NeuN + nuclei, the first principal component was strongly correlated with the log10-transformed nuclei number counts and percentage of respectively oligodendrocytes and neurons based on deconvolution ( Figure S3 C). Therefore, principal component plots were made after correction for this covariate and donor variation, using the function removeBatchEffect (limma package). Proportional abundance of cell types was estimated using marker genes from a single cell study 69 with dtangle (v2.0.9) deconvolution.

Differential expression was assessed using an empirical Bayes moderated t-test within limma’s linear model framework including estimated proportion of oligodendrocytes or neurons and log10-transformed nucleus number counts as covariates [Y = 0 + experimental condition + estimated cell-type proportion + log10 (cell number)]. P -values were corrected for multiple testing using the Benjamini-Hochberg false discovery rate (FDR). Genes with FDR<0.05 were considered significantly differentially expressed. Results were visualized with the ggplot2 package (v3.5.1).

Competitive gene set enrichment analysis was performed with CAMERA with preset value of 0.01 for inter-gene correlation using the Hallmark, C1, C2, C3, C5, C6, and C7 gene set collections from the Molecular Signatures Database (MSigDB, version 2023.2; https://www.gsea-msigdb.org/gsea/msigdb/index.jsp ).

Published: April 15, 2025

Footnotes

Supplemental information can be found online at https://doi.org/10.1016/j.isci.2025.112430 .

Contributor Information

Hendrik J. Engelenburg, Email: j.engelenburg@nin.knaw.nl.

Joost Smolders, Email: j.j.f.m.smolders@erasmusmc.nl.

Supplemental information

Document S1. Figures S1–S3

Table S1. Results of differential gene expression analysis and gene set enrichment analysis of bulk nuclei sequencing, related to Figure 4

References

Associated Data

Supplementary Materials

Document S1. Figures S1–S3

Table S1. Results of differential gene expression analysis and gene set enrichment analysis of bulk nuclei sequencing, related to Figure 4

Data Availability Statement

Nucleus RNA-seq data deposited at GEO and are publicly available as of the date of publication. The Accession numbers of nuclei RNA sequencing is GSE273954 . Tissue RNA-seq data (Fig. 2AB) is available under accession number GSE283092 . This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.