Introduction

Multiple Sclerosis (MS) is a chronic neuroinflammatory and neurodegenerative disease with a largely unknown etiology, secondary to an autoimmune demyelination in the central nervous system (CNS). Histopathologically, it is characterized by gliosis, oligodendrocyte death, and neuroaxonal loss1. Worldwide, approximately 2.8 million patients with MS deal with significant levels of disability2.

Several studies have shown that the autoimmune response seen in MS mostly derives from a dysfunctional autoreactive CD4+ T cell compartment3,4,5, although other pro-inflammatory cells, such as B cells and myeloid cells, seem to be implicated as well6. Recently, mitochondria have also been shown to play a role in driving MS disease activity and progression. In the CNS, mitochondrial dysfunction has been found to be a critical trigger for neuroaxonal loss and brain atrophy7,8. In the peripheral immune compartment, CD4+ T cells from patients with MS show oxidative phosphorylation (OxPhos) insufficiency9,10,11, which, in animal models, has been linked to an exacerbation of CNS autoimmune-mediated inflammation12,13.

Mitochondrial DNA (mtDNA) is responsible for encoding 22 transfer RNAs (tRNAs), two ribosomal RNAs (rRNAs), and 13 proteins of the OxPhos chain, where multiple mtDNA molecules coexist, allowing for the existence of multiple genotypes at various variant levels (VLs) within the same cell—heteroplasmy14.

Interestingly, mtDNA mutations, as well as particular haplogroups, which are inherited mutational patterns that may be classified into phylogenetic clusters15, have been associated with an increased risk of MS, albeit not consistently16,17,18,19,20. Additionally, mtDNA polymorphisms have been shown to modulate both metabolism and immunity21,22, and mtDNA variants have significant tissue-specificity, including in T cells23. Nonetheless, whether the aforementioned CD4+ T cell OxPhos dysfunction in patients with MS is an inherent consequence of the particular mitochondrial genotype of this immune subset remains unknown.

We hypothesized that mitochondrial genotype in CD4+ T cells might influence MS disease activity and progression. Thus, we aimed to explore the differences in mtDNA extracted from CD4+ T cells between patients with a recent diagnosis of either Clinically Isolated Syndrome (CIS) or Relapsing–Remitting MS (RRMS) and healthy controls (HC). We also analyzed longitudinal mtDNA changes in patients with CIS/RRMS (PwMS).

Methods

Cohort description

We performed an observational retrospective evaluation of prospectively collected data within the Berlin CIS-Cohort (reference: NCT01371071)24 at the Charité—Universitätsmedizin Berlin, Berlin, Germany. Berlin CIS-Cohort’s inclusion criteria are: (I) age ≥ 18 years and (II) diagnosis of either Clinically Isolated Syndrome (CIS) within 6 months from symptom onset or of Relapsing–Remitting Multiple Sclerosis (RRMS) within two years from symptom onset, according to the 2017 revisions of the McDonald criteria25.

To address the disease activity and progression of patients with CIS/RRMS (PwMS), the following variables were assessed on each clinical visit: number of relapses and time to last relapse; expanded disability status scale (EDSS) and Multiple Sclerosis functional composite (MSFC) scores; brain MRI: T2 hyperintense lesions and T1 gadolinium-enhancing lesions; and optical coherence tomography (OCT): peripapillary retinal nerve fiber layer (RNFL) thickness and ganglion cell-inner plexiform layer (GCIPL) volume. Additionally, we assessed if patients fulfilled the criteria for no evidence of disease activity (NEDA)-3, namely, absence of new relapses, on MRI: no T1 gadolinium-enhancing lesions and no new or enlarging T2 hyperintense lesions and the absence of EDSS worsening. Details of MRI, OCT, and NEDA-3 were described earlier26. Finally, blood samples were collected and, subsequently, peripheral blood mononuclear cells (PBMCs) were isolated. Further details on the methodology used for collection and analysis of clinical data may be found in Sect. 1 of the Supplementary Information.

PwMS were included in this study if biological samples and a clinical assessment were available for two clinical visits: 6 months (VIS1) and 36 months (VIS2) after disease onset. Recruitment of HC without a family history of MS was finalized in May 2019; HC were matched to PwMS at VIS1 in a 1:1 ratio according to sex and a maximum age difference of five years.

Our primary outcomes were a mitochondrial genotype comparison between: (I) PwMS and HC in the cross-sectional analysis and (II) different diagnostic evolutions from VIS1 to VIS2 in the longitudinal analysis, namely, CIS in both clinical visits, a conversion from CIS to RRMS, and RRMS in both clinical visits.

Since mtDNA whole genome sequencing (WGS) in CD4+ T cells from PwMS was unreported in the literature, a pilot study with 20 triplets (20 PwMS at VIS1&VIS2 and 20 age- and sex-matched HC) was performed to determine the appropriateness of the sample size through the Dupont method27.

Ethical approval

The institutional review board (IRB)’s approval was obtained by the Ethics Committee of the Charité—Universitätsmedizin Berlin (application number: EA1/182/10), informed consents were given by every subject, and the study followed the standards of the Declaration of Helsinki28.

Sample processing

CD4 + T cell enrichment and flow cytometry analysis

PBMC sample processing was performed at the same time for each triplet (PwMS at VIS1&VIS2 and HC), to minimize differences within processing. The MojoSort™ Human CD4 T Cell Isolation Kit (#480010, BioLegend, San Diego, CA, USA) was used for CD4+ T cell enrichment, in accordance with the manufacturer’s instructions.

To assess whether CD4+ T cell enrichment was achieved, flow cytometry was performed in a BD LSRFortessa™ X-20 Cell Analyzer (BD Biosciences, Franklin Lakes, NJ, USA) on a subset of samples (before and after magnetic enrichment). The following antibodies were used: CD14-eFluor450 (#48014941, clone 61D3, eBioscience™, Thermo Fisher Scientific, Waltham, MA, USA); CD19-Alexa Fluor 647 (#302222, clone HIB19, BioLegend, San Diego, CA, USA); CD3-PerCP-Cy5.5 (#45003741, clone OKT3, eBioscience™, Thermo Fisher Scientific, Waltham, MA, USA); CD4-BV711 (#317439, clone OKT4, BioLegend, San Diego, CA, USA); and CD56-PE-Cy7 (#25056741, clone CMSSB/NCAM, eBioscience™, Thermo Fisher Scientific, Waltham, MA, USA).

Regarding the gating strategy, CD4+ T cells were defined as CD3+CD4+CD19-CD56- single cells (see Supplementary Figs. S6, 7 online). A detailed protocol and additional details regarding the methodology used for CD4+ T cell enrichment and flow cytometry may be found in Sect. 2 of the Supplementary Information.

DNA extraction and mtDNA sequencing, including bioinformatic processing and data analysis

Similarly, DNA extraction was performed at the same time for each triplet with the QIAamp® DNA Blood Midi Kit (QIAGEN GmbH, Hilden, Germany), according to the manufacturer’s instructions.

Samples were sequenced with the Applied Biosystems™ Precision ID mtDNA Whole Genome Panel (Thermo Fisher Scientific, Waltham, MA, USA) at IPATIMUP—Instituto de Patologia e Imunologia Molecular da Universidade do Porto (Porto, Portugal). Libraries were prepared using the Ion Chef™ automated protocol and samples were run on 530™ chips with the Ion Torrent™ Ion S5™ (Thermo Fisher Scientific, Waltham, MA, USA). Samples with a coverage uniformity < 85% and/or a mean coverage < 1500 reads were resequenced.

Regarding bioinformatic processing, the PrecisionCallerPipeline (PCP) pipeline29 was used, with mutserve v230,31 for variant calling, HaploGrep v2.4.015 for haplogroup calling, and Haplocheck v1.3.331 for a contamination check. To account for possible false positives, variants with a variant level (VL) ≥ 10% were only accepted if they were found with both the PCP pipeline and the Ion Torrent Suite™ Software (TSS), while variants below TSS’s limit of detection (10%) and solely present with PCP where filtered in accordance with: (I) sequencing indicators for variant reliability (normalized coverage, coverage ratio, mean value of reported nuclear insertions of mitochondrial DNA [NUMTs], and the distance to the amplicon’s edge); and (II) previous reports of the same variants in curated mtDNA databases. Variants only present in TSS were excluded29.

Data analysis was performed with R version 4.1.132 and Excel 2016 (Microsoft Corporation, Redmond, WA, USA).

The level of statistical significance was set at a two-sided p-value < 0.05 for all tests and multiple comparison testing was adjusted with false discovery rate (FDR). Whenever data was missing, it was censored from the analysis. Reporting followed the standards from STROBE33 and its extension STREGA34.

A detailed description of the methodology regarding DNA extraction, bioinformatic processing, and data analysis may be found in Sect. 3 of the Supplementary Information.

Results

Cohort description

Overall, 61 PwMS were included in this study cohort (Fig. 1, Table 1). Most patients presented with RRMS at both clinical visits and approximately one third of the PwMS (N = 20) suffered a relapse between VIS1 and VIS2, while PwMS under MS immunomodulatory treatment increased from 26.23% to 47.54%.

Figure 1
figure 1

Clinical cohort flowchart. 61 newly diagnosed PwMS were included in this study cohort and reassessed after 30.50 ± 1.27 (mean ± SD) months. Additionally, HC without a family history of MS were matched to PwMS at cohort entry in a 1:1 ratio according to sex and age, with an average age difference of 2.06 ± 1.31 (mean ± SD) years. Mean time from disease onset is depicted followed by the minimum and maximum values between brackets. Figures and arrows were adapted from OpenMoji. License: CC BY-SA 4.0. HC healthy control(s); MS Multiple Sclerosis; N number; PwMS patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis; V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1; V2 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 2.

Full size image
Table 1 Cohort characteristics.
Full size table

As mentioned previously, a pilot study with 20 triplets (20 PwMS at VIS1&VIS2 and 20 age- and sex-matched HC) was performed to determine the appropriateness of the sample size through the Dupont method27. Following WGS data analysis of this subset, a 35% discordance in prevalence was detected for deleterious variants in Complex I—55% for PwMS at VIS1 (discordant prevalence: 45%) vs. 20% for HC (discordant prevalence: 10%). Hence, the sample size was adequate according to this endpoint.

CD4 + T cell enrichment, DNA extraction, and WGS quality

Following the magnetic enrichment in CD4+ T cells, we obtained a wide range in the number of cells and cell mortality, for all subject types (Table 2 and online Supplementary Table S6). In comparison with HC, PwMS at VIS1 showed a lower number of cells (mean difference of 0.65 million, 95% confidence interval [CI] 0.16–1.14; paired t-test adjusted with FDR). In comparison with VIS2, cells from PwMS at VIS1 had lower mortality (mean difference of 2.77%, 95% CI 0.56–4.97%; paired t-test adjusted with FDR).

Table 2 Magnetic enrichment and DNA extraction: Summary.
Full size table

Nonetheless, a significant increase in CD4+ T cells (CD3+ CD4+ CD56- CD19-) was achieved (Fig. 2a and online Supplementary Table S4: mean fold change in paired samples of 2.59, 95% CI 2.31–2.87; paired t-test, N = 49), regardless of subject type (see Supplementary Table S5 online).

Figure 2
figure 2

CD4+ T cell enrichment and mitochondrial DNA whole genome sequencing. (a) Percentage of CD4+ T cells (CD3+ CD4+ CD56- CD19-) before and after magnetic enrichment, as assessed with flow cytometry—lines connect the same sample; (b) Mean mappability, per subject type and per definition—error bars denote minimum and maximum values; (c) Mean mtDNA coverage, per subject type; (d) Median mtDNA coverage, per subject type. HC healthy control(s); mtDNA mitochondrial DNA; V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1; V2 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 2.

Full size image

Subsequently, after the extraction of DNA, variable amounts of DNA were achieved (Table 2 and online Supplementary Table S6). In comparison with PwMS at VIS1, HC had higher yields of extracted DNA (mean difference of 4.45 ng/μL for the measurement with Synergy HTX, 95% CI 1.06–7.84, and mean difference of 3.35 ng/μL for the measurement with Qubit®, 95% CI 0.80–5.90; paired t-tests adjusted with FDR).

However, while, as expected, the two DNA measurements showed a correlation between each other (adjusted R2 of 0.88 and a p-value < 2.2 × 10–16, linear regression model) and the number of cells used for DNA extraction correlated with DNA concentration (adjusted R2 of 0.77 and a p-value < 2.2 × 10–16, linear regression model), DNA concentration had no effect on mtDNA coverage, with most variability appearing to arise from each sequencing run (see Supplementary Fig. S10 online). Correspondingly, no differences in mtDNA coverage and mappability were found between different subject types (Fig. 2b–d and online Supplementary Fig. S10).

Taking into account all samples analyzed through WGS (see Supplementary Table S8 online), no contamination was detected and all haplogroups corresponded to European lineages. PwMS had the same haplogroup with two minor exceptions (Triplets #5 and #48) at both visits, albeit without changing their simplified haplogroup (see Supplementary Table S8 online).

Cross-sectional comparison

The total number of variants was similar between different subject types, i.e., between HC and PwMS at VIS1 and between VIS1 and VIS2 for PwMS (Fig. 3 and online Supplementary Table S8: paired t-tests adjusted with FDR). Variant distribution was also bimodal in all subject types (Fig. 3a), likely arising from each sample’s haplogroup (Fig. 3b–c), since the mean number of variants in HC and PwMS at VIS1 was significantly different in haplogroups with at least three samples (p-value 3.34 × 10–16; Kruskal–Wallis test). Haplogroups H and HV had significantly fewer variants (see Supplementary Table S9 online), which is consistent with the revised Cambridge Reference Sequence (rCRS)’s own H haplogroup35. Nonetheless, haplogroup distribution was independent from subject type (HC and PwMS at VIS1; Fisher’s exact test with Monte Carlo simulations).

Figure 3
figure 3

Total number of variants: Cross-sectional comparison (subject types and haplogroups). (a) Number of variants, per subject type; (b) Number of variants, per subject type and per haplogroup; (c) Expansion of (b) for better visualization. HC healthy control(s); V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1; V2 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 2.

Full size image

When comparing HC and PwMS at VIS1, we observed that the total number of variants did not vary according to age (linear regression model). The difference in the number of mutations between HC and PwMS at VIS1 was also not influenced by sex (two sample t-test, χ2 test of independence, and two-proportions z-test) (see Supplementary Fig. S11 online).

When looking at the differences between PwMS at VIS1 and HC in various mtDNA regions, the highest discordancy in prevalence was in MT-ND3 for PwMS at VIS1 and in MT-RNR2 for HC, while the highest discordant adjusted mutational rate, which is the sum of all VLs in a specific region divided by its region size, was in MT-TA for PwMS at VIS1 and in MT-TT for HC. However, there were no significant differences for discordant prevalence (McNemar’s tests adjusted with FDR); nor for discordant mutational rates (paired t-tests adjusted with FDR) (see Supplementary Fig. S12 online).

Regarding macro regions in mtDNA, the most commonly discordant affected region for PwMS at VIS1 was Other, which refers to positions in rCRS with an overlap between the two strands or left unannotated, and, for HC, transfer RNA (tRNA), while the region with the highest discordant adjusted mutational rate was Complex III for PwMS at VIS1 and Other for HC. Nonetheless, there were no significant differences for discordant prevalence (McNemar’s tests adjusted with FDR); nor for discordant mutational rates (paired t-tests adjusted with FDR) (see Supplementary Fig. S12 online).

We further took all variants into account (see Supplementary Table S10 online), to see if individual variants differed between HC and PwMS at VIS1. However, no significant differences were found (McNemar’s tests adjusted with FDR). Interestingly, a single PwMS had a pathogenic variant, namely, m.11778G > A in the MT-ND4 gene, which is associated with Leber’s hereditary optic neuropathy (LHON) and neuropathy, ataxia, and retinitis pigmentosa (NARP), albeit at low VLs: 5.8% and 7.3% for VIS1 and VIS2, respectively; well below the VLs usually required for pathogenesis36. The patient in question initially presented with pyramidal and sensory complaints, maintaining the latter throughout their disease course, without any visual changes.

Subsequently, we took into account variants predicted in silico to be likely deleterious for proteins, defined by a mean value > 0.5 from two independent scores: (I) MutPred37,38; and (II) APOGEE39. Nevertheless, the number of deleterious variants (see Supplementary Fig. S13 online) and the cumulative deleterious burden (Fig. 4a), which corresponds to the total sum of a variant’s VL multiplied by its deleterious score per sample, did not differ significantly between the different subject types (paired t-tests adjusted with FDR).

Figure 4
figure 4

Cumulative deleterious burden: Cross-sectional comparison (subject types and haplogroups). (a) Cumulative deleterious burden, per subject type; (b) Cumulative deleterious burden, per subject type and haplogroup; (c) Expansion of (b) for better visualization. HC healthy control(s); V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1; V2 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 2.

Full size image

We observed a similar haplogroup-specific effect for both the number of deleterious mutations and cumulative deleterious burden (online Supplementary Fig. S13 and Fig. 4b,c: p-values 4.61 × 10–13 and 1.38 × 10–15, respectively; Kruskal–Wallis tests). Haplogroups J and T had significantly more deleterious variants (see Supplementary Table S11 online) and cumulative deleterious burden (see Supplementary Table S12 online).

When comparing HC and PwMS at VIS1, the number of deleterious variants did not vary according to age (Kendall rank correlation test). The difference in the number of deleterious mutations between HC and PwMS at VIS1 was also not influenced by sex (two sample t-test, χ2 test of independence, and two-proportions z-test). The same was true for cumulative deleterious burden (linear regression model for age and two sample t-test for sex) (see Supplementary Fig. S14 online).

Regarding differences between PwMS at VIS1 and HC in various mtDNA regions, the highest discordancy in prevalence was in MT-ND5 for PwMS at VIS1 and in MT-ATP6 for HC (Fig. 5a). In parallel, the highest discordant adjusted cumulative deleterious rate, which is the cumulative deleterious rate for a specific region relative to its region size, was in MT-ND3 for PwMS at VIS1 and in MT-ATP6 for HC (Fig. 5b). Regardless, no significant differences were found for discordant prevalence (Fig. 5a: McNemar’s tests adjusted with FDR); nor for discordant cumulative deleterious rates (Fig. 5b: paired t-tests adjusted with FDR).

Figure 5
figure 5

Deleterious variants: Cross-sectional comparison (regions). (a,c) Discordant prevalence of deleterious variants for each mtDNA coding region/locus and for each macro mtDNA coding region, per subject type, respectively; (b,d) Relative cumulative deleterious burden for each mtDNA coding region/locus and for each macro mtDNA coding region, per subject type, respectively. bp base pair; HC healthy control(s); V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1.

Full size image

Regarding macro regions in mtDNA, the most commonly discordant affected region was Complex I for PwMS at VIS1 and Complex V for HC (Fig. 5c), while the region with the highest adjusted cumulative deleterious rate was Complex III for PwMS at VIS1 and Complex V for HC (Fig. 5d). However, there were no significant differences for discordant prevalence (Fig. 5c: McNemar’s tests adjusted with FDR); nor for cumulative deleterious rate (Fig. 5d: paired t-tests adjusted with FDR).

Finally, we looked into tRNA mutations predicted to be pathogenic according to the MitoTIP score40: no variants were found (see Supplementary Fig. S15 online). The cumulative MitoTIP score, which corresponds to the total sum of a variant’s VL multiplied by its MitoTIP score per sample, was similar in all subject types (paired t-tests adjusted with FDR).

Nonetheless, as observed previously, haplogroup had a significant influence in the cumulative MitoTIP score (p-value 3.59 × 10–13; Kruskal–Wallis test), with haplogroups K and U having higher cumulative MitoTIP scores, whereas haplogroups H and HV had lower scores (see Supplementary Table S13 online). No influence was found regarding age (linear regression model) or sex (two sample t-test). The region with the highest discordant adjusted cumulative MitoTIP rate, which is the cumulative MitoTIP rate for a specific region relative to its region size, was MT-TA for PwMS at VIS1 and MT-TT for HC. However, no significant differences were found (paired t-tests adjusted with FDR) (see Supplementary Fig. S15 online).

Longitudinal comparison

As mentioned previously, there were no significant differences between PwMS at VIS1 and at VIS2 (paired t-tests adjusted with FDR) concerning total number of variants (Fig. 3a), number of deleterious variants (see Supplementary Fig. S13 online), cumulative deleterious burden (Fig. 4a), and cumulative MitoTIP score (see Supplementary Fig. S15 online).

The mean proportion of mutations present in a single visit—transient variants—were 4.01% (range: 0.00–53.33%) for VIS1 and 3.38% (range: 0.00–37.14%) for VIS2 (Fig. 6a), with mean VLs of 8.39% (range: 2.50–100.00%) for VIS1 and 4.66% (range: 2.50–19.60%) for VIS2 (Fig. 6b); no significant differences were found between different visits (paired t-tests).

Figure 6
figure 6

Longitudinal changes in PwMS: Intrapersonal changes. (a) Proportion of transient variants, per subject and clinical visit; (b) Mean VL of transient variants, per subject and clinical visit; (c,d) Density plots of VL change in persistent variants, for all mutations and per subject, respectively. PwMS patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis; V1 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 1; V2 patient(s) with Clinically Isolated Syndrome/Relapsing–Remitting Multiple Sclerosis at visit 2; VL variant level.

Full size image

Subsequently, we considered the difference in VLs for the previously found variants between the second and the first visits—persistent variants. The mean VL change was 0.06% in all variants (Fig. 6c) and 0.09% per subject (Fig. 6d) (range: − 10.90–50.30%). Both VL changes were statistically similar to zero (one-sample t-test).

When we compared transient and persistent variants, transient variants had a significantly lower VL than persistent variants (adjusted p-value 1.80 × 10–150; two-sample t-test adjusted with FDR) and a lower coverage ratio (adjusted p-value 1.09 × 10–6; two-sample t-test adjusted with FDR). There were no significant differences for other quality control variables29 (two-sample t-tests adjusted with FDR), nor for protein deleterious score and MitoTIP score (two-sample t-tests adjusted with FDR) (see Supplementary Fig. S16 online).

Since each PwMS had a very similar mutational pattern between the two visits, we then moved on from an intrapersonal comparison to an interpersonal comparison, with a focus on exploring the origin of these transient variants and the VL change in persistent variants in the full PwMS cohort. For that purpose, we considered a PwMS as a single data point, where we performed the mean of the results from both sequencing runs.

Afterwards, we considered the various possible causes or cofactors behind differences in the proportion of transient variants, VLs of transient variants or VL change in persistent variants. Haplogroup (Kruskal–Wallis test), age (linear regression model), and sex (two sample t-test) did not influence the aforementioned WGS variables (see Supplementary Figs. S17–19 online). Interestingly, different regions had a significant effect on the proportion of transient variants, excluding patients without transient variants and regions with mutations from fewer than three paired samples (p-value 5.90 × 10–6; Kruskal–Wallis test), where MT-TI had the highest prevalence and D-loop the lowest (see Supplementary Table S14 online). This effect was absent for VLs of transient variants or VL change in persistent variants (Kruskal–Wallis test) (see Supplementary Figs. S18–19 online).

Regarding the clinical variables in the cohort, change in MS medication status, i.e., Yes vs. No (Table 1 and online Supplementary Table S15), was independent from diagnostic evolution (Fisher’s exact test with Monte Carlo simulations) (see Supplementary Fig. S20 online).

To identify associations of clinical disability with diagnostic evolution, we performed batch Kruskal–Wallis tests with diagnostic evolution as a grouping variable for each available clinical variable (see Supplementary Figs. S21, 22 online). Two different approaches were tested: (I) the mean of each clinical variable between VIS1 and VIS2; and (II) the change of each clinical variable from VIS1 to VIS2, by subtracting VIS2 and VIS1. After multiple comparison correction with FDR, only the mean values of the number and volume of T2 hyperintense lesions were associated with diagnostic evolution (see Supplementary Table S16 online).

Similarly, we performed batch Kruskal–Wallis tests with diagnostic evolution as a grouping variable for each WGS variable explained previously (see Supplementary Fig. S23 online); no significant differences were found (see Supplementary Table S17 online). Additionally, haplogroup was independent from diagnostic evolution (Fisher’s exact test with Monte Carlo simulations) (see Supplementary Fig. S24 online). Similarly, as a post hoc analysis, dividing patients according to NEDA-3 status (Table 1 and online Supplementary Fig. S25) or cumulative deleterious burden (see Supplementary Figs. S26, 27 online) yielded null results.

Discussion

The present study investigated whether mitochondrial genotype in CD4+ T cells was associated with a diagnosis of CIS/RRMS or with longitudinal diagnostic evolution. Accordingly, we performed WGS from CD4+ T cells in a matched cohort of PwMS. Overall, however, there were no significant differences regarding number of variants, number of deleterious variants, cumulative protein deleterious burden, or cumulative MitoTIP score between PwMS and HC, nor after a mean of 30.50 months of follow-up between VIS1 and VIS2 for PwMS.

Importantly, the CIS/RRMS cohort herein described is representative of the overall population of PwMS in Germany, regarding mean age at onset, female predominance, and distribution of the diagnoses, with a vast majority of RRMS41.

According to the flow cytometry analysis performed in a subset of magnetically enriched samples, CD4+ T cell enrichment was successful with the MojoSort™ Human CD4 T Cell Isolation Kit, which was expected42. The Precision ID mtDNA Whole Genome Panel, in combination with the PCP pipeline29, proved particularly valuable, as it was able to obtain mtDNA WGS from all samples without signs of contamination, despite a wide range of the number of cells after processing the samples, and, consequently, of DNA yield, which is consistent with previous reports43.

Curiously, amongst the 21 most prevalent mutations in PwMS, five had already been associated with MS, ranging from case reports to large epidemiological studies16,18,44,45,46. Furthermore, remarkably, the high number of deleterious variants in complex I and IV matches the pattern of decreased activity in these complexes in CD4+ T cells of PwMS in previous studies9,47. Moreover, the higher number of deleterious variants and cumulative deleterious burden from haplogroups J and T is consistent with previous reports of increased MS risk for these haplogroups16,17.

Nevertheless, in our cohort, the mitochondrial genotype did not differ significantly between PwMS and HC, nor between VIS1 and VIS2 within PwMS, which is consistent with previous mtDNA WGS studies in PwMS19,20.

In contrast to the 35% discordance in prevalence for deleterious variants in Complex I registered in the pilot study, the final difference was 18%, due to an increase in prevalence of 8% for HC, while in PwMS it decreased by 9%. Thus, our sample size was incompatible with the previously-set endpoint.

A further limitation might have also been the mtDNA WGS of the very heterogeneous CD4+ T cell compartment, as well as its magnetic enrichment, as opposed to other methods aimed at achieving higher cell purity. Nonetheless, previous studies focused on CD4+ T cell OxPhos dysfunction in patients with MS employing magnetic enrichment and fewer subjects than the present cohort were still successful in defining a clear mitochondrial phenotype9,47. Furthermore, several existing disease modifying drugs have been shown to interfere with CD4+ T cell metabolism10,47,48,49,50, which substantiates the relevance of these pathways in MS pathophysiology and points towards new and intriguing possibilities of pharmaceutical development.

Finally, since we have only explored cross-sectional differences between PwMS at VIS1 and HC, we cannot exclude the possibility of having different longitudinal mtDNA mutational rates between PwMS and HC.

Despite these limitations, this study constitutes a thorough cross-sectional and longitudinal analysis of mtDNA WGS in PwMS, in addition to surveying mitochondrial genotype specifically in CD4+ T cells. Moreover, we were able to assess the longitudinal dynamics of mtDNA, which have only very recently started to be unveiled51,52.

Overall, CD4+ T cell mitochondrial genotype was not associated with a diagnosis of CIS/RRMS, nor with longitudinal diagnostic evolution. We further postulate that mitochondrial dysfunction in CD4+ T cells is unlikely to derive from mitochondrial genotype.