Loss of oligodendrocyte transcription factor 2 protein expression in metabolically stressed oligodendrocytes

Abstract

Oligodendrocytes are essential for myelin production, maintenance, and repair, and their dysfunction contributes to the pathogenesis of demyelinating diseases such as multiple sclerosis (MS). Here, we identify an early stress-associated oligodendrocyte state characterized by a rapid, post-transcriptional loss of the lineage-defining transcription factor Oligodendrocyte Transcription Factor 2 (OLIG2). Using the cuprizone model of toxic demyelination, we observed an early appearance of OLIG2^low expressing oligodendrocytes, followed by the emergence of OLIG2-negative oligodendrocytes at later stages. This observation was particularly pronounced among cells expressing the integrated stress response marker Activating Transcription Factor 3 (ATF3). Transcriptomic analysis, quantitative PCR, and combined in situ hybridization–immunohistochemistry confirmed that these changes occurred without a corresponding reduction in Olig2 mRNA levels, indicating that OLIG2 protein loss is a stress-induced, post-transcriptional event not captured by RNA-level profiling. A similar phenotype was observed in a reversible metabolic stress paradigm (i.e., chronic starvation model) and in post-mortem MS lesions, where stressed oligodendrocytes showed reduced or absent OLIG2 protein expression. Pharmacological intervention with the sphingosine-1-phosphate receptor modulator siponimod during cuprizone intoxication attenuated OLIG2 protein loss, indicating that this stress-induced state is pharmacologically modifiable. These findings reveal a transient and potentially reversible phenotype in stressed oligodendrocytes that may precede overt cell loss or demyelination. Thus, OLIG2 protein loss may serve as an early indicator of oligodendrocyte stress and a possible therapeutic target for preserving myelin integrity in demyelinating disorders. These findings have additional methodological implications as stressed oligodendrocytes may evade detection using OLIG2-based lineage markers.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00401-026-03000-x.

Introduction

Oligodendrocytes, the myelinating glia of the central nervous system (CNS) enable rapid saltatory conduction and provide essential metabolic support to axons [8]. The progression from oligodendrocyte progenitor cells (OPC) to myelinating oligodendrocytes is governed by lineage-defining transcription factors. Among these, Oligodendrocyte Transcription Factor 2 (OLIG2), a basic helix–loop–helix protein, has a central role in specification, survival, and early differentiation [47, 61, 62]. Due to its lineage-specific expression in both rodents and humans, and its sustained expression into adulthood, OLIG2 is widely used as a pan-oligodendrocyte marker in developmental and pathological studies [57]. While genetic manipulations have underscored the importance of OLIG2 for differentiation and myelination [53, 54], the stability and contribution of constitutive OLIG2 protein expression in mature oligodendrocytes under pathological conditions remain poorly defined.

Mature oligodendrocytes are particularly sensitive to metabolic and oxidative stress. Continuous myelin lipid synthesis imposes high ATP and NADPH demands [25], yet oligodendrocytes possess a comparatively limited antioxidant repertoire [14, 18, 56]. As a consequence, stressors that neighboring astrocytes and microglia can withstand often precipitate rapid oligodendrocyte dysfunction and cell death [45]. This selective vulnerability makes oligodendrocytes a key target of metabolic insults in both experimental models and neurological disease. In mice, cuprizone intoxication rapidly induces metabolic oligodendrocyte stress, culminating in a severe and selective oligodendrocyte loss and robust demyelination [7, 10, 33, 52]. Similarly, systemic metabolic imbalances, such as in chronic starvation models, can induce pathologic alterations in glial cells [63]. Notably, metabolic stress and oligodendrocyte vulnerability are also prominent features of multiple sclerosis (MS) lesions. Active and chronic MS lesions show an accumulation of mitochondrial-derived reactive oxygen species [36], ferroptotic lipid peroxides [51] and display profound oligodendrocyte energy failure [41]. While the susceptibility of oligodendrocytes to these insults is well established, it is unknown whether such metabolic stress affects the stability of lineage-defining transcription factors like OLIG2. Since cellular stress responses often involve a global attenuation of protein translation [19, 40], we hypothesized that the abundance of OLIG2 protein is compromised or dynamically regulated in stressed oligodendrocytes.

Clarifying this point is critical: persistent depletion of lineage-defining transcription factors could indicate destabilization of the oligodendrocyte gene-regulatory network, possibly impairing myelin integrity and remyelination capability. Moreover, OLIG2 expression is routinely used to define oligodendroglial lineage cells in histological and molecular analyses. If metabolic stress reduces OLIG2 protein expression, conventional methods could underestimate oligodendrocyte densities. Consequently, stress-induced alterations in OLIG2 protein expression levels carry important methodological implications for reliably identifying oligodendrocytes in diseased or stressed tissues.

To investigate whether lineage-defining transcription factors can be stabilized under metabolic stress, we utilized pharmacological modulation of sphingosine-1-phosphate receptor (S1PR) signaling. Siponimod, a CNS-penetrant S1PR modulator approved for treatment of secondary progressive MS [30], has been shown to promote oligodendrocyte survival in animal models [5, 28]. In the cuprizone model, siponimod’s protective effects likely reflect a combination of direct actions on oligodendrocytes and indirect effects mediated through glial crosstalk. We reasoned that if OLIG2 protein expression is stress-sensitive, oligodendrocyte protective treatments like siponimod should prevent its loss. Preserving OLIG2 expression could suggest that improving the cell’s stress state can maintain lineage identity and the gene networks required for myelination [60].

The in vivo relevance and specific impact of stress on OLIG2 protein expression remain largely unexplored. Here, we test the hypothesis that OLIG2 protein expression is dynamically regulated in response to cellular stress. We quantify OLIG2 protein expression across complementary contexts—cuprizone intoxication, chronic starvation, and post-mortem MS brain tissues—using chromogenic immunohistochemistry and immunofluorescence. We further employ in situ hybridization and quantitative PCR to directly compare protein abundance with Olig2 transcript levels. Finally, to link mechanism to intervention, we examine whether pharmacological modulation by siponimod can mitigate stress-associated OLIG2 protein loss during metabolic challenge.

Materials and methods

Animals

Male and female C57BL/6 mice (Janvier Labs, Le Genest-Saint-Isle, France) were housed under standard laboratory conditions with ad libitum access to food and water, at a maximum of five animals per cage. Cages were cleaned weekly, and microbiological status was monitored in accordance with the Federation of European Laboratory Animal Science Associations (FELASA) guidelines. Animals were acclimatized for at least 1 week prior to experimental procedures. All experimental procedures were approved by the review boards for the care of animal subjects of the district government (Cuprizone and siponimod experiments: government of upper Bavaria, reference no. 55.2-154-2532-73-15 and government of Mecklenburg-Western Pomerania, reference no. 7221.3-1-001/19; Starvation and refeeding experiments: government of Mecklenburg-Western Pomerania reference no. 7221.3-1-005/21). An overview of all experimental groups in this study can be found in Table 1.

Table 1.

Overview of the experimental animal groups involved in this study

							Sex	Analysis
Cuprizone course experiments
Control (n = 5)	1 week cuprizone (n = 5)	3 weeks cuprizone (n = 5)			5 weeks cuprizone (n = 5)		♂ C57BL/6	IHC, ISH
1 week cuprizone + siponimod
Control (n = 3)	1 week cuprizone + vehicle (n = 7)			1 week cuprizone + siponimod (n = 7)			♀ C57BL/6	IHC
Control (n = 6)	1 week cuprizone + vehicle (n = 4)			1 week cuprizone + siponimod (n = 4)			♂ C57BL/6	RT-PCR
3 weeks cuprizone + siponimod
Control (n = 4)	3 weeks cuprizone + vehicle (n = 4)			3 weeks cuprizone + siponimod (n = 4)			♀ C57BL/6	IHC, RNA seq
Starvation and refeeding experiments
Control starvation (n = 5)	Starvation (n = 6)		Control refeeding (n = 5)			Refeeding (n = 5)	♀ C57BL/6	IHC

Cuprizone intoxication and siponimod application

Cuprizone-induced demyelination was induced as previously described [29]. C57BL/6 mice received a diet containing 0.25% cuprizone (bis(cyclohexanone)oxaldihydrazone), which was precisely weighed and thoroughly mixed into ground standard chow using a commercial food processor (Kult X, WMF Group, Geislingen an der Steige, Germany) at maximum speed with intermittent manual agitation. The cuprizone-containing chow was provided ad libitum in two separate plastic Petri dishes per cage. Exclusion criteria included rapid weight loss (> 10% within 24 h), severe neurological symptoms (e.g., reduced locomotion, seizures, stupor), or signs of infection. Siponimod (provided as a powder formulation by Novartis Pharma AG, Basel, Switzerland; MTA: #PLSMTA19AUG177) was dissolved in a 1% carboxymethylcellulose sodium solution (CMCS, Sigma-Aldrich, #4888-500G) and administered by oral gavage of 100 µl at a dose of approximately 3.1 mg/kg as described previously [5]. Mice were treated daily with either siponimod or a drug-free vehicle solution (CMCS) for 1 week or 3 weeks while receiving a diet containing 0.25% cuprizone. Two mice in the 1-week treatment group (IHC group) had to be euthanized prematurely due to stupor and apathy following oral gavage with the vehicle solution and were excluded from the analysis. Control mice received food ad libitum throughout the entire experiment.

Starvation and refeeding paradigm

The starvation model was performed as previously described by Staffeld et al. [46]. Female C57BL/6J mice (4 weeks old) underwent a 10-day acclimatization period with ad libitum access to food and water. Body weight, food intake, and estrous cycle status were recorded daily at 1 PM. The acute starvation phase was defined as a 1-week phase, in which the mice received 40% of their baseline food intake. To model chronic starvation, daily food allotments were individually adjusted to achieve and maintain a 25% reduction in body weight over a 2-week period, corresponding to approximately 45–70% of baseline intake. During this phase, animals had unrestricted access to their allocated food (Ssniff, Soest, Germany). Following the starvation period, animals entered a 21-day refeeding phase with ad libitum access to standard chow. Control mice received food ad libitum throughout the entire experiment. Mice were removed from the study if humane endpoint criteria were met, including an additional weight loss > 10% within 24 h, seizures, paralysis, respiratory distress, or labored breathing.

Tissue preparation and tissue embedding

For histological and immunohistochemical analyses, mice were anesthetized with intraperitoneal injections of ketamine (100 mg/kg) and xylazine (10 mg/kg), followed by transcardial perfusion with 20 ml of ice-cold phosphate-buffered saline (PBS) and subsequently 50 ml of 3.7% paraformaldehyde (PFA, pH 7.4). Brains were post-fixed overnight in the same fixative, then dissected, paraffin-embedded, and sectioned coronally at 5 µm thickness. Immunohistochemistry and in situ hybridization were performed on these paraffin sections. All analyses were conducted between brain level 265 and 295, as defined by the Sidman mouse brain atlas (http://www.hms.harvard.edu/research/brain/atlas.html). This corresponds to Bregma −1.01 according to the stereotaxic coordinates of Franklin and Paxinos [15]. For mRNA expression analyses, mice were transcardially perfused with ice-cold PBS, and brains were removed and manually dissected under a stereomicroscope. Tissue samples comprising the corpus callosum and overlying neocortex were collected between levels 215 and 355 according to Sidman et al. and snap-frozen for subsequent processing.

Multiple sclerosis post-mortem tissue

Paraffin-embedded post-mortem brain tissues were obtained through a rapid autopsy protocol from donors with progressive MS in collaboration with the Netherlands Brain Bank, Amsterdam. The institutional ethics review board approved the study, and all donors or their relatives provided written consent to use brain tissues and clinical information for research purposes. For this study, active and chronic-active lesions from seven donors were included (see Table 2). The average age at death of the donors was 62.0 ± 15.4 years (mean ± standard deviation). The average postmortem delay was 8.67 ± 2.44 h. Staging of lesions was performed as reported previously [50], using consecutive immunostaining for myelin proteolipid protein (PLP) and the human leukocyte antigen [HLA]-DR (clone LN3). Normal appearing white matter (NAWM) regions were defined as white matter areas located on the same tissue section but at the maximum possible distance from the active lesion border. These regions were histologically verified to be free of demyelination (intact PLP staining) and inflammatory infiltrates (absence of HLA-DR clusters) on consecutive sections, following established neuropathological criteria [52].

Table 2.

Clinical characteristics of MS tissue donors

Sex	Age at death (years)	MS disease duration (years)	Date of birth	Date MS diagnosis	PMD (h)	MS type	Cause of death
MS donors
♂	44	21	1965	1988	10.25	PPMS	End stage MS
♂	51	Unknown	1958	< 1990	11	SPMS	Infection
♀	60	10	1950	2003	10.66	SPMS	Assisted dying
♂	54	15	1957	1999	8.25	PPMS	Assisted dying
♀	66	32	1945	1988	9.5	PPMS	Assisted dying
♀	62	24	1949	1990	12.5	PPMS	Cachexia
♀	35	10	1979	2004	10.33	SPMS	Assisted dying
Controls (non-neurologically impaired)
♂	87	Not applicable	1918	Not applicable	6.5	Not applicable	Pneumonia
♀	73		1929		4.0		Lung fibrosis
♀	86		1923		6.5		Multiple myeloma
♀	60		1950		7.5		Infection
♀	66		1941		7.0		Heart failure

♂, male; ♀, female; MS, multiple sclerosis; PMD, post-mortem delay; PPMS, primary progressive multiple sclerosis; SPMS, secondary progressive multiple sclerosis

Immunohistochemistry and evaluation procedure

Immunohistochemistry was performed following previously established protocols [59]. Briefly, tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval, when necessary, by heat-induced epitope unmasking using either citrate buffer (pH 6.0) or Tris/EDTA buffer (pH 9.0). After rinsing in PBS, sections were incubated for 1 h in blocking solution containing 5% normal serum from the species in which the secondary antibodies were raised. Following removal of the blocking solution, sections were incubated overnight at 4 °C with primary antibodies diluted in the same blocking solution. A complete list of primary antibodies used is provided in Table 3. Negative controls, including omission of primary antibodies, were processed in parallel. On the following day, after rinsing in PBS, sections were incubated in 0.35% hydrogen peroxide (H₂O₂) in PBS for 30 min to quench endogenous peroxidase activity. After additional PBS washes, sections were incubated for 1 h at room temperature with biotinylated secondary antibodies, followed by a 1 h incubation with peroxidase-conjugated avidin–biotin complexes (ABC kit; Vector Laboratories, Peterborough, UK). Immunoreactivity was visualized using 3,3′-diaminobenzidine (DAKO, Hamburg, Germany) and H₂O₂, resulting in a brown precipitate at antigenic sites. To quantify cell/particle density, the stained and processed sections were digitalized using the OCUS^®20 microscope slide scanner (Grundium, Finland) to produce scans of the respective sections and regions of interest (ROIs). The scans were analyzed using the open-source software QuPath (version 0.5.1) [4]. The area of the respective ROI was manually outlined. Cell nuclei were identified using QuPath’s integrated cell detection algorithm. Detection parameters were individually optimized for each immunohistochemical labeling to account for differences in staining patterns and background intensity. Detected cells were subsequently classified as positive or negative using a random tree-based object classifier. For anti-OLIG2 immunohistochemical labeling, positively classified cells were further subdivided into intensely stained (OLIG2^high) and faintly stained (OLIG2^low) populations. The OLIG2^low phenotype was defined operationally as OLIG2-positive nuclei classified as ‘faintly stained’ by the supervised random tree classifier trained separately for each immunohistochemical staining. Classifier training was performed on randomly selected regions from the analyzed scans. An independent, blinded evaluator manually classified a subset of detected cells within the training images to provide ‘ground truth’ annotations. The random tree classifier utilized all per-cell measurements generated by QuPath. This approach was necessitated by the inherent variability of immunohistochemistry, where staining intensity is influenced by section thickness, antigen retrieval efficiency, and background variability related to tissue myelination status. Consequently, OLIG2^low and OLIG2^high classification was performed based on the training images evaluated by a blinded evaluator within the same staining batch rather than using a universal cutoff for optical density. All scripts and parameters are provided in the Supplementary Information.

Table 3.

List of antibodies used for immunohistochemistry and immunofluorescence

Antigen	Species	Dilution	Retrieval method	Purchase number	RRID	Supplier
Primary antibodies
ASPA	Rabbit	1:2000	Tris/EDTA	ab223269	AB_3662876	Abcam, UK
ATF3	Rabbit	1:200	Tris/EDTA	ab254268	AB_2910214	Abcam, UK
Cleaved caspase 3	Rabbit	1:50	Tris/EDTA	9661	AB_2341188	Cell signaling, UK
CNPase	Mouse	1:2000	Tris/EDTA	ab6319	AB_2082593	Abcam, UK
DDIT3/CHOP	Mouse	1:200	Tris/EDTA	ab11419	AB_298023	Abcam, UK
GFAP	Chicken	1:1000	Tris/EDTA	ab4674	AB_304558	Abcam, UK
IBA1	Rabbit	1:5000	Tris/EDTA	019-19741	AB_839504	Wako, USA
IBA1	Goat	1:200	Tris/EDTA	ab48004	AB_870576	Abcam, UK
MHC-II, HLA-DR (clone LN3)	Mouse	1:1500	Citrate	MA5-11966	AB_10979984	Thermo fisher, USA
NeuN	Mouse	1:1000	Tris/EDTA	MAB377	AB_2298772	Merck Millipore, USA
NG2	Rabbit	1:200	Tris/EDTA	AB5320	AB_91789	Merck Millipore, USA
OLIG1	Rabbit	1:500	Tris/EDTA	15849S	AB_3713255	Cell signaling, UK
OLIG2	Goat	1:200	Tris/EDTA	AF2418	AB_2157554	R&D Systems, USA
OLIG2	Rabbit	1:1000	Tris/EDTA	AB9610	AB_570666	Merck Millipore, USA
OLIG2	Mouse	1:100	Tris/EDTA	MABN50	AB_10807410	Merck Millipore, USA
PLP	Mouse	1:5000	None	MCA839G	AB_2237198	BioRAD, USA
Phospho-Histone H2A.X (Ser139)	Rabbit	1:300	Tris/EDTA	9718	AB_2118009	Cell signaling, UK
Serpina3n	Goat	1:2500	Tris/EDTA	AF4709	AB_2270116	R&D Systems, USA
SOX10	Rabbit	1:500	Tris/EDTA	ab180862	AB_2721184	Abcam, UK
SOX10	Mouse	1:100	Tris/EDTA	ab218522	AB_3683672	Abcam, UK
Secondary antibodies
Anti-rabbit IgG	Goat	1:200	Not applicable	BA-1000	AB_2313606	Vector, USA
Anti-mouse IgG	Goat	1:200		BA-9200	AB_2336171	Vector, USA
Anti-goat IgG	Rabbit	1:200		BA-5000	AB_2336126	Vector, USA
Anti-rabbit Alexa Fluor™ 488	Donkey	1:250		ab150065	AB_2860569	Abcam, UK
Anti-mouse Alexa Fluor™ 488	Donkey	1:250		ab150109	AB_2571721	Abcam, UK
Anti-goat Alexa Fluor™ 488	Donkey	1:250		A11055	AB_2534102	Invitrogen, USA
Anti-chicken Alexa Fluor™ 488	Donkey	1:250		A78948	AB_2921070	Invitrogen, USA
Anti-goat Alexa Fluor™ 594	Donkey	1:250		ab150136	AB_2782994	Abcam, UK
Anti-rabbit Alexa Fluor™ 594	Donkey	1:250		A21207	AB_141637	Invitrogen, USA

HIER, heat-induced epitope retrieval; RRID, research resource identifiers [3]

Analysis of optical density in anti-2′,3′-cyclic-nucleotide 3′-phosphodiesterase (CNPase) immunohistochemistry (Fig. 3a, c) was performed in QuPath using a pixel classifier. For classifier training, CNPase-positive structures were manually annotated in a randomly assembled training image. The CNPase-positive area was quantified as the percentage of CNPase-positive pixels relative to the total area of the ROI.

Fig. 3.

Anti-CNPase and anti-OLIG2 co-labeling reveals a subset of mature oligodendrocytes that lose OLIG2 protein expression during cuprizone intoxication. a, c Anti-CNPase immunohistochemistry reveals a significant decrease in cortical CNPase⁺ area after 3 and 5 weeks of cuprizone intoxication, consistent with cortical demyelination in the cuprizone model. a, insets Despite the overall decrease in CNPase density, oligodendrocytes remained identifiable by a strong membranous staining. b, d A significant reduction in the density of CNPase-positive oligodendrocytes in the cortex was only observed after 5 weeks of cuprizone intoxication. b, e Double labeling for CNPase and OLIG2 revealed the emergence of a substantial subset of CNPase⁺ oligodendrocytes lacking OLIG2 expression, especially after 3 and 5 weeks of cuprizone intoxication. f In parallel, the proportion of OLIG2⁺CNPase^⁻ cells increased during the course of cuprizone intoxication, likely reflecting a migration of oligodendrocyte progenitor cells. c, d, f one-way ANOVA followed by Tukey’s multiple comparisons test, e Kruskal–Wallis test followed by Dunn’s multiple comparisons test. All groups n = 5. Scale bar: a 250 µm, (insets a) 20 µm, b 150 µm, (insets b) 25 µm

Immunofluorescence labeling and evaluation procedure

Paraffin-embedded tissue sections were deparaffinized, rehydrated, and, if required, subjected to antigen retrieval as described previously. Following rehydration, sections were blocked using normal serum from the species in which the secondary antibodies were raised. Sections were then incubated overnight at 4 °C with the indicated combinations of primary antibodies diluted in blocking solution. After thorough washing in PBS, sections were incubated for 2 h at room temperature with the appropriate combinations of fluorescently labeled secondary antibodies. Following a final PBS wash, slides were mounted using Fluoroshield™ with DAPI to counterstain cell nuclei. To control for non-specific binding, two sets of negative controls were included: (i) sections were incubated with primary antibodies followed by mismatched secondary antibodies, and (ii) sections were incubated with secondary antibodies alone to assess background staining. Stained sections were imaged using an epifluorescence microscope (Leica DM6 B) equipped with a K5 sCMOS camera, and image acquisition was performed using the Leica Application Suite X software (version 3.7.0.20979, 2019, Germany). The scans were analyzed using the open-source software QuPath (version 0.5.1) [4]. The area of the respective ROI was manually outlined. Cells were analyzed either (i) manually, by a blinded evaluator, for all scans involving anti-CNPase immunofluorescence labeling, or (ii) automatically. For automated analyses, cell nuclei were detected using QuPath’s integrated cell detection algorithm. Detected cells were then classified as positive or negative using a random tree-based object classifier. In anti-OLIG2 immunofluorescent labeling, positively classified cells were further subdivided into intensely (OLIG2^high) and faintly stained (OLIG2^low) populations. Specifically, the OLIG2^low expressing phenotype in immunofluorescence images was defined using the same supervised machine learning workflow as for the chromogenic immunohistochemistry. Briefly, a random tree classifier was trained on manually annotated ground-truth cells (classified as ‘OLIG2^high’ or ‘OLIG2^low’ by a blinded evaluator) within each experimental batch. This classifier utilized multiple cellular features (mean intensity, texture, morphology) to robustly distinguish OLIG2^low from OLIG2^high nuclei in a background of varying fluorescence intensity.

When classification based on standard measurements proved challenging, additional features were incorporated. Specifically, we calculated Haralick texture features [22] or quantified the proportion of positively stained pixels within each cell, as determined by a separate pixel classifier. Classification was performed sequentially for both antigens of interest. Scripts and parameters for cell detection are provided in the supplementary information.

To evaluate signal intensity at the single-cell level, the mean nuclear fluorescence intensity of each cell was normalized to the median intensity of cells classified as double negative within the same animal. Per-animal normalization was applied to account for differences in background intensity, which vary particularly with the degree of myelination in the respective ROI. An example R script is provided in the Supplementary Information.

Gene array analysis, next-generation sequencing and real-time qPCR

We reanalyzed previously published gene expression datasets generated using Affymetrix GeneChip microarrays (Affymetrix, Santa Clara, CA, USA) and next-generation sequencing (NGS) experiments performed as described before [5, 32].

Normalized, log2-scale gene expression values were analyzed for the comparison. For each gene we computed the log2 fold-change (log2FC) as the difference of group means on the log2 scale. Two-sided Welch’s t tests were performed on the log2 values, and the resulting p values were adjusted for multiple testing using the Benjamini–Hochberg false-discovery-rate (FDR) procedure. Genes with FDR q < 0.05 and |log2FC| ≥ 1 were considered differentially expressed. Genes with zero within-group variance or missing values were treated as not significant (p set to 1). Volcano plots were created in GraphPad Prism 8 (XY scatter; X = log₂FC, Y = − log₁₀q), with reference lines at log₂FC = ± 1 and − log₁₀q = 1.3; upregulated genes (q < 0.05, log₂FC ≥ 1) and downregulated genes (q < 0.05, log₂FC ≤ − 1) were colored red and blue, respectively, and all others gray.

Real-time qPCR (Bio-Rad, Germany) was performed using 2 × Sensi Mix SYBR and Fluorescein (Bioline/Meridian) based on a standardized protocol as previously described [32]. Relative quantification was performed using internal standard curves and Hprt1 as the housekeeping reference gene. Melting curves of the PCR products were routinely performed to determine the specificity of the PCR reactions (data not shown). The primer sequences and the respective annealing temperatures are given in Table 4.

Table 4.

Sequence of primers used in the study

Gene	Forward primer	Reverse primer	Ta (°C)	Product size ( BP)
Atf3	ACAGAGTGCCTGCAGAAAGAGT	CCATTCTGAGCCCGGACGAT	59	149
Cnp	CTCCAGGTGTGCTGCACTGTA	GATCTCTTCACCACCTCCTGCT	62	94
Hprt1	TCAGTCAACGGGGGACATAAA	GGGGCTGTACTGCTTAACCAG	61	142
Olig2	TGAAGCGATGGAGAGATGCG	CCCAGACCCTTGGAGTGTTC	64	71
Plp1	TGGCGACTACAAGACCACCA	GACACACCCGCTCCAAAGAA	64	116
Sox10	AGGTTGCTGAACGAAAGTGAC	CCGAGGTTGGTACTTGTAGTCC	64	102

T_a, annealing temperature; BP, base pairs

Combined in situ hybridization and immunohistochemistry

A combined protocol was established to enable the simultaneous detection of Olig2 mRNA and OLIG2 protein within the same tissue section. Deparaffinized brain sections from control animals (n = 4) and 1-week cuprizone-intoxicated animals (n = 4) were processed for immunohistochemistry as previously described. Immediately after deparaffinization, sections were washed in PBS and incubated for 1 h in 5% normal rabbit serum to block non-specific binding, without prior heat-induced epitope retrieval. Following blocking, sections were incubated overnight at 4 °C with a goat anti-OLIG2 primary antibody (AF2418; AB_2157554; R&D Systems) diluted 1:100 in 5% normal rabbit serum. The next day, sections were incubated for 1 h at room temperature with a biotinylated anti-goat secondary antibody (1:200, Vector BA-5000), followed by a 1-h incubation with alkaline phosphatase-conjugated avidin–biotin complexes (ABC-AP kit; AK-5000; Vector Laboratories, Peterborough, UK). OLIG2 protein expression was visualized using the ImmPACT Vector Red substrate (VEC-SK-5105; Vector Laboratories), producing a red precipitate at sites of antigenicity. Immediately thereafter, Olig2 mRNA was visualized in the same sections using a chromogenic in situ hybridization (ISH) protocol (RNAscope™ 2.5 HD Detection Reagents-BROWN, Cat. 322310; ACD^®), along with a specific Olig2 RNA probe (Cat. 447091; ACD^®). To enhance the ISH signal, tyramide signal amplification was applied (Akoya Biosciences, NEL700A001KT). Sections were digitized as described above and analyzed in a manual, blinded manner.

Statistical analysis

All evaluations were performed by investigators blinded to the experimental groups during both data collection and analysis. To avoid pseudo-replication in single-cell analysis (i.e., fluorescence intensity measurements in Fig. 2e, Fig. 7d, Supplementary Figure S4c, d), data collected from individual cells were aggregated to calculate a mean value per biological replicate (i.e., per animal). Statistical comparisons were strictly performed using these biological replicates (n = number of mice). Data are presented as arithmetic means ± standard error of the mean (SEM). Statistical analysis was conducted using Prism software (version 8.0.2; GraphPad Software Inc., San Diego, CA, USA) with a significance threshold set at p < 0.05. The following symbols denote levels of statistical significance: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001; ‘ns’ indicates no significant difference. No data points were excluded as outliers. Normality of data distribution was assessed using the Shapiro–Wilk test. Specific statistical tests applied are indicated in the corresponding figure legends.

Fig. 2.

The OLIG2^low phenotype is associated with expression of the ISR marker ATF3, particularly during early cuprizone intoxication. a Immunofluorescent co-labeling of OLIG2 and ATF3 in the corpus callosum during cuprizone intoxication. Note that OLIG2⁺ATF3⁺ double-positive cells show a lower OLIG2 signal (arrow) compared to OLIG2⁺ATF3⁻ cells (arrowhead). b, c The absolute and relative densities of double-labeled cells are significantly increased after 1 week of cuprizone (1 week Cup) compared to controls (Ctrl) and decrease again with prolonged intoxication. d ATF3⁺ oligodendrocytes show a significantly higher proportion of the OLIG2^low phenotype compared to ATF3⁻OLIG2⁺ cells. e OLIG2⁺ATF3⁺ cells show a significant decrease in normalized nuclear OLIG2 signal intensities when compared to OLIG2⁺ATF3⁻ cells. Data points represent independent biological replicates (animals), calculated as the mean of all analyzed cells per animal. b, c Kruskal–Wallis test followed by Dunn’s multiple comparisons test, d, e Mann–Whitney test. All groups n = 5. Scale bar: 50 µm, scale bar insets: 15 µm

Fig. 7.

Siponimod attenuates OLIG2⁺ cell loss and the emergence of OLIG2^low expressing oligodendrocytes during early cuprizone intoxication. a Representative anti-OLIG2 and ATF3 double-labeling after 1 week of cuprizone intoxication and treatment with either siponimod or vehicle. b Cuprizone intoxication causes a marked reduction in overall OLIG2⁺ cell density compared to control, which is significantly less pronounced in siponimod-treated animals. In both the vehicle and the siponimod group, the density of OLIG2⁺ATF3⁺ cells increases after cuprizone intoxication. c In the vehicle group, ATF3⁺ oligodendrocytes exhibit a lower OLIG2 signal intensity (arrows) compared to ATF3^⁻ oligodendroglia (arrowhead). This relative increase in OLIG2^low expressing cells is significant in the vehicle group but absent in siponimod-treated mice. d The normalized nuclear OLIG2 intensity is significantly reduced in ATF3⁺ oligodendrocytes after 1 week of cuprizone intoxication in the vehicle group. No OLIG2 intensity differences can be observed between ATF3^⁻ and ATF3⁺ oligodendrocytes in siponimod-treated animals, indicating that siponimod treatment preserves higher nuclear OLIG2 protein levels in stressed (i.e., ATF3⁺) oligodendrocytes. Data points represent independent biological replicates (animals), calculated as the mean of all analyzed cells per animal. b OLIG2⁺ cell density: one-way ANOVA followed by Sidak’s post hoc test; OLIG2⁺ATF3⁺ cell density: Kruskal–Wallis test followed by Dunn’s multiple comparisons test. c, d Mann–Whitney test. Control group n=3, Cuprizone+Vehicle n=5, Cuprizone+Siponimod n=7. Scale bar: 75 µm, scale bar insets: 25 µm.

Results

Temporal dynamics of OLIG2 protein loss during cuprizone intoxication

We first examined the temporal dynamics of OLIG2 expression during cuprizone intoxication by OLIG2 immunohistochemistry, a well-established approach in this model [27] and in human pathology [34]. As shown in Fig. 1a and consistent with previous studies [24], loss of OLIG2-positive cells was evident in the corpus callosum after 1–3 weeks of cuprizone intoxication (Fig. 1a, b) and in the cortex after 3–5 weeks (Fig. 1a, c). A subsequent increase in OLIG2-positive cells in the corpus callosum after 5 weeks of cuprizone intoxication can be attributed to the recruitment and proliferation of oligodendrocyte progenitor cells as is known in this model [42].

Fig. 1.

Oligodendrocyte loss is accompanied by expression of integrated stress response markers and a relative increase of Oligodendrocyte transcription factor 2 (OLIG2)^low expressing cells in the cuprizone model. a OLIG2 immunohistochemistry during the course of cuprizone intoxication in the corpus callosum (b) and cortex (c). Consecutive brain sections were processed for anti-ATF3 (d) and anti-DDIT3 (e) immunohistochemistry, both downstream markers of the integrated stress response pathway. Note the decline in total OLIG2⁺ cell densities with accompanying relative increase in OLIG2^low expressing cells (arrowheads) beside strongly expressing OLIG2 cells (arrows) in the corpus callosum and cortex (a–c). Simultaneously, the density of ATF3⁺ and DDIT3⁺ cells increased, especially after initial cuprizone intoxication (d–g). Differences between cuprizone-intoxicated and control groups were determined using one-way ANOVA followed by Tukey’s multiple comparisons test if normal distribution of data was given; otherwise, the Kruskal–Wallis test followed by Dunn’s multiple comparisons test was used. All groups n = 5. Scale bar: 50 µm, insets: 25 µm

Notably, in both the corpus callosum and the cortex, a subset of OLIG2-positive cells exhibited markedly reduced nuclear staining during the early phase of cuprizone intoxication (Fig. 1b corpus callosum: Ctrl 12.63% ± 3.34; 1 week Cup 32.93% ± 2.78; 3 weeks Cup 24.29% ± 1.75; 5 weeks Cup 25.42% ± 0.65; 1c cortex: Ctrl 17.51% ± 1.51; 1 week Cup 34.29% ± 1.74; 3 weeks Cup 34.78% ± 3.63; 5 weeks Cup 33.16% ± 2.39). We hereafter refer to this phenotype as OLIG2^low expressing cells. The loss of OLIG2-positive cells and the relative increase in OLIG2^low cells were accompanied by a significant rise in the density of ATF3⁺ (Fig. 1d, f) and DDIT3⁺ cells (Fig. 1e, g) in the corpus callosum and cortex after 1 week of cuprizone intoxication—both established markers of the activated integrated stress response (ISR) in oligodendrocytes [13, 19]. The ISR is a conserved stress-signaling pathway in which kinases such as PERK phosphorylate eIF2α, resulting in translational attenuation and induction of stress-associated transcription factors including ATF4, ATF3, and DDIT3/CHOP. ISR activation in oligodendrocytes has been reported as early as 4 days after cuprizone exposure [19] and is associated with reduced expression of myelin-related transcripts under chronic stress conditions [17, 38, 58].

To verify these findings, we repeated anti-OLIG2 immunohistochemistry on consecutive brain sections with two additional anti-OLIG2 primary antibodies of different species and clonality. Both confirmed a comparable absolute decrease in OLIG2-positive cell densities over the course of cuprizone intoxication, as well as the relative increase in the proportion of OLIG2^low-expressing cells in the cortex and corpus callosum (Supplementary Figure S1).

OLIG2^low phenotype associates with integrated stress response activation in the cuprizone model

If oligodendrocytes exhibit reduced OLIG2 protein expression in response to stress (i.e., become OLIG2^low), the observed decrease in OLIG2-positive cells during cuprizone intoxication could reflect either (i) an actual loss of oligodendrocytes or (ii) a reduction of OLIG2 protein expression to levels too low to be detected by immunohistochemistry. To determine whether stressed oligodendrocytes exhibit reduced OLIG2 expression on an individual cell level, we performed immunofluorescence double-labeling with the ISR marker ATF3. This revealed a significant increase in OLIG2⁺ATF3⁺ cell density in the corpus callosum after 1 week of cuprizone intoxication, which declined as cuprizone intoxication progressed (Fig. 2a, b).

After 1 week of cuprizone intoxication, OLIG2⁺ cells constituted the majority of ATF3⁺ cells (Fig. 2c: Ctrl 0.00%; 1 week Cup 76.61% ± 1.36). During this early phase of oligodendrocyte intoxication, we next tested whether stressed oligodendrocytes (i.e., OLIG2⁺ATF3⁺) express less OLIG2 than non-stressed cells (i.e., OLIG2⁺ATF3⁻). Oligodendrocytes expressing ATF3 more frequently exhibited an OLIG2^low phenotype than oligodendrocytes negative for ATF3 (Fig. 2d: proportion of OLIG2^low within OLIG2⁺ cells: ATF3⁻ 57.46% ± 3.27; ATF3⁺ 89.86% ± 1.42). Single-cell quantification of nuclear OLIG2 fluorescence intensity confirmed a reduced OLIG2 signal in ATF3⁺ relative to ATF3⁻ oligodendrocytes. ATF3⁺ oligodendrocytes displayed a significantly lower OLIG2 signal intensity than ATF3⁻ counterparts (Fig. 2e: OLIG2⁺ATF3⁻ 3.96 AU ± 0.29; OLIG2⁺ATF3⁺ 2.29 AU ± 0.08).

Because ATF3 is a dynamic ISR readout that may miss stressed cells once expression returns to baseline, we additionally assessed Serpina3n, a recently described marker of metabolically stressed oligodendrocytes [26]. Serpina3n was largely absent at 1 week despite the emergence of OLIG2^low cells, but was robustly induced after 3 weeks of cuprizone intoxication, when Serpina3n was largely observed in cells with low to undetectable OLIG2 signal (Supplementary Figure S2). Together, these data indicate that OLIG2 protein loss can precede the emergence of a broader metabolic stress signature captured by Serpina3n.

While the absolute densities of OLIG2⁺ and ATF3⁺ cells each decreased to a similar extent by roughly one-half between 1 and 3 weeks of cuprizone intoxication (Fig. 1b, f), the density and proportion of OLIG2⁺ATF3⁺ cells declined much more sharply (Fig. 2c: 3 weeks Cup 8.46% ± 1.27; 5 weeks Cup 10.30% ± 1.04). This disproportionate decrease in OLIG2⁺ATF3⁺ cells could reflect (i) reactive ATF3 expression in non-oligodendrocyte glia cells during the course of cuprizone intoxication (i.e., OLIG2⁻ATF3⁺), (ii) the proliferation of OLIG2-expressing OPCs that are less affected by cuprizone intoxication [6] and therefore remain ATF3⁻ (i.e., OLIG2⁺ATF3⁻ OPC) or, (iii) a complete loss of OLIG2 protein in stressed oligodendrocytes as part of the metabolic stress response (i.e., OLIG2⁻ATF3⁺ oligodendrocytes).

To address the cellular specificity of ATF3 expression (explanation i), we performed immunofluorescence double-labeling of ATF3 with GFAP (i.e., a marker for astrocytes) and IBA1 (i.e., a microglia/macrophage marker). ATF3 did not co-label with GFAP at any stage of the cuprizone intoxication (data not shown). Co-expression with IBA1 emerged only from week three of cuprizone intoxication (Supplementary Figure S3). Thus, up to and including week one, ATF3 serves as a highly specific marker of oligodendrocyte stress in this model.

To test explanation (ii), namely whether the expanding OLIG2⁺ but ATF3⁻ OPC pool contributes to the disproportionate decline in OLIG2⁺ATF3⁺ cells at later cuprizone time points, we performed double immunofluorescence for OLIG2 and the OPC marker NG2 (Supplementary Figure S4). Indeed, the NG2⁺OLIG2⁺ OPC population expanded robustly at weeks three and five of cuprizone intoxication (Supplementary Figure S4b). Importantly, after 1 week of cuprizone intoxication, the proportion of OLIG2^high cells was markedly reduced among NG2^– cells, whereas NG2⁺ OPC largely retained high OLIG2 levels (Supplementary Figure S4c, d). Notably, even under control conditions, NG2⁺ OPC exhibited a significantly higher nuclear OLIG2 intensity than NG2⁻ oligodendrocytes (Supplementary Figure S4c, d). Together, these data indicate that the reduced OLIG2/ATF3 co-localization at later stages is at least partly influenced by a lineage shift toward an expanded OLIG2⁺ OPC pool.

Complete loss of OLIG2 protein in mature oligodendrocytes in the cuprizone model

To test whether metabolic stress can lead to a complete loss of OLIG2 protein expression in oligodendrocytes (explanation iii), we assessed OLIG2 expression relative to an independent marker of mature oligodendrocytes: 2′,3′-cyclic-nucleotide 3′-phosphodiesterase (CNPase), which is a myelin-associated enzyme and oligodendrocyte-restricted in the vertebrate CNS [31]. Because abundant myelin complicates single-cell resolution, these analyses were performed in the cortex to reliably identify individual CNPase⁺ oligodendrocytes. CNPase optical density (CNPase⁺ area) decreased significantly at 3 and 5 weeks of cuprizone intoxication, consistent with cortical demyelination (Fig. 3a, c). Despite the overall reduction in CNPase⁺ area, mature oligodendrocytes remained identifiable by their characteristic membranous CNPase⁺ staining pattern (Fig. 3a, insets), and a significant decrease in the density of CNPase⁺ cells was observed only after 5 weeks of cuprizone intoxication (Fig. 3d). Next, we investigated whether metabolic stress in oligodendrocytes can lead to a complete loss of OLIG2 (i.e., CNPase⁺OLIG2⁻). To this end, we performed immunofluorescence double-labeling with CNPase and OLIG2 over the course of cuprizone intoxication (Fig. 3b). Interestingly, there was a significant increase in CNPase⁺ oligodendrocytes that did not express OLIG2 after 3 and 5 weeks of cuprizone intoxication (Fig. 3e: Ctrl 0.0%; 1 week Cup 6.48% ± 1.17; 3 weeks Cup 53.86% ± 2.38; 5 weeks Cup 32.79% ± 6.58), indicating a complete loss of OLIG2 protein in a substantial subset of mature oligodendrocytes. In parallel, we noted an increase in OLIG2⁺CNPase⁻ cells over the course of cuprizone pathology (Fig. 3f), consistent with increased densities of OPC (relate to Supplementary Figure S4) that have not yet started to express CNPase in the cortex [21].

Temporal dynamics of DNA damage in oligodendrocytes in the cuprizone model

To gain a better understanding of the temporal course of oligodendrocyte cell stress in the cortex, we performed double-immunofluorescence labeling with CNPase and phosphorylated histone H2AX (γ-H2AX), which is a sensitive marker of double-strand DNA breaks. Previous evidence suggests that DNA damage contributes to oligodendrocyte degeneration following cuprizone intoxication [11]. We first quantified the density of all γ-H2AX-positive cells in the cortex over the course of the cuprizone intoxication. A significant increase in the density of γ-H2AX-positive cells was observed after 3 weeks of cuprizone intoxication (Fig. 4b: Ctrl 0.68 ± 0.36 cells/mm²; 1 week Cup 11.24 ± 1.88 cells/mm²; 3 weeks Cup 28.90 ± 5.16 cells/mm²; 5 weeks Cup 9.50 ± 2.43 cells/mm²).

Fig. 4.

Double labeling for CNPase and γ-H2AX identifies mature oligodendrocytes suffering DNA damage during cuprizone intoxication. a Representative images showing γ-H2AX immunoreactivity in CNPase⁺ oligodendrocytes in the cortex after cuprizone intoxication. The absolute density of γ-H2AX⁺ cells (b) and γ-H2AX⁺CNPase⁺ double-positive cells (c) showed a progressive increase during the early phase, reaching statistical significance at week three of cuprizone intoxication compared to controls. was significantly increased at all cuprizone timepoints compared to control and peaked after 3 weeks of cuprizone intoxication. Although levels remained elevated above baseline at 1 week and 5 weeks, these differences did not reach statistical significance. d At the 3-week cuprizone timepoint, γ-H2AX⁺CNPase⁺ cells represented a significant proportion of all mature oligodendrocytes in the cortex. e Across all time points examined, oligodendrocytes constituted the majority of cells undergoing γ-H2AX associated DNA damage. b–e Kruskal–Wallis test followed by Dunn’s multiple comparisons test. All groups n = 5. Scale bar: 200 µm, scale bar insets: 20 µm

To define the onset and temporal progression of DNA damage specifically within the oligodendrocyte lineage, we performed immunofluorescence double-labeling for CNPase and γH2AX. The density of CNPase⁺γH2AX⁺ cells showed a temporal pattern (Fig. 4c), with a significant peak at week three of cuprizone intoxication (Ctrl 0.12 ± 0.12 cells/mm²; 3 weeks Cup 25.98 ± 4.52 cells/mm²). The proportion of CNPase⁺γH2AX⁺ cells among all CNPase⁺ cells showed an elevation after 1 week of cuprizone (Fig. 4d: Ctrl 0.08% ± 0.08; 1 week Cup 10.66% ± 1.99) and reached statistical significance after 3 weeks (3 weeks Cup 38.12% ± 7.03). Importantly, across all cuprizone time points, CNPase⁺ oligodendrocytes comprised the vast majority (> 75%) of γH2AX⁺ cells in the cortex (Fig. 4e), suggesting that the observed DNA damage is predominantly restricted to the oligodendrocyte lineage. Taken together, the OLIG2^low phenotype emerges early in the cuprizone model (by week one, relate to Fig. 1) and temporally precedes the subsequent accumulation of DNA damage in oligodendrocytes (CNPase⁺γH2AX⁺ cells by week three, see Fig. 4). To assess whether this DNA damage is associated with ‘classical’ caspase-dependent apoptosis, we performed double immunofluorescence for cleaved caspase-3 and CNPase. We detected only sporadic cleaved caspase-3⁺ cells in the cortex of cuprizone-intoxicated mice, and the majority of these were CNPase-negative (Supplementary Figure S5). This paucity of ‘classical’ apoptotic markers, despite extensive oligodendrocyte loss, is consistent with caspase-independent cell death mechanisms previously reported in this model [11, 64].

Chronic starvation induces an OLIG2^low phenotype that normalizes after nutrient restoration

We next evaluated whether the OLIG2^low phenotype can be observed under a different paradigm of metabolic cellular stress. We therefore employed a setting of malnutritional stress by using a chronic starvation model. In this experimental setup, mice received 40% of their normal caloric intake for 1 week, which resulted in a reduction of body weight of 25%. This body weight was maintained for 2 weeks to mimic chronic starvation. In a separate cohort of mice, a 2-week refeeding period followed starvation, enabling discrimination between reversible and irreversible oligodendrocyte changes.

First, we quantified OLIG2⁺ cell density in the corpus callosum (brain region 265) in the starvation and refeeding cohorts relative to controls and found no significant differences (Fig. 5c, h). Quantifying OLIG2^low expressing cells using our established workflow, starvation significantly increased the density of OLIG2^low cells (Fig. 5d: Ctrl 324.6 ± 21.66 cells/mm²; Starvation 394.9 ± 17.77 cells/mm²) and their proportion among all OLIG2⁺ cells (Fig. 5e: Ctrl 20.46% ± 1.37; Starvation 26.83% ± 1.42); these changes were fully reversible, as refeeding returned OLIG2^low phenotypes to control levels (Fig. 5i, j). Together, these data identify OLIG2 protein loss as a conserved and possibly reversible hallmark of metabolically stressed oligodendrocytes across different experimental paradigms.

Fig. 5.

Starvation induces a reversible OLIG2^low state in corpus callosum oligodendrocytes. Representative anti-OLIG2 immunohistochemistry of control (a, f), starvation (b), and refeeding (g) brain sections (brain region 265). Quantification of OLIG2 cell densities in the starvation (c) and refeeding cohort (h) compared to control groups. Densities of OLIG^low expressing cells in the starvation (d) and refeeding cohort (i). Relative numbers of OLIG^low expressing cells compared to all OLIG2-positive cells in mice of the starvation (e) and refeeding cohort (j). Note the OLIG2^low expressing cells in the corpus callosum of mice after 2 weeks of starvation (arrows, inset b), compared to the strongly OLIG2-positive cells in the refeeding group (inset g). Differences between experimental groups and control groups were determined using unpaired t test. Starvation and controls n = 6; refeeding and controls n = 5. Scale bars: 75 µm, scale bar insets: 20 µm

OLIG2^low and OLIG2⁻ oligodendrocytes are present in human MS lesions

To extend these findings to human pathologies, we examined whether the stress-associated reduction in OLIG2 protein is evident in MS tissue by quantifying OLIG2 expression in active and chronic-active lesions from seven donors and comparing it with white matter from five non-neurologically impaired controls. To accurately analyze OLIG2 densities in different areas of MS lesions, we delineated the rim and core of demyelinated lesions using PLP and LN3 immunohistochemistry (Fig. 6a). Anti-OLIG2 staining revealed significantly lower OLIG2⁺ cell densities in both the lesion rim and core compared with white matter from control donors and with the normal-appearing white matter (NAWM) of MS donors (Fig. 6b: Ctrl WM 920.1 ± 35.61 cells/mm²; NAWM 1113 ± 106.50 cells/mm²; lesion rim 643.70 ± 56.91 cells/mm²; lesion core 106.20 ± 39.82 cells/mm²). Notably, in the chromogenic anti-OLIG2 immunohistochemistry, we observed cells consistent with an OLIG2^low phenotype, especially in the lesion rim (Fig. 6b). To investigate this more closely, we identified oligodendrocytes in human tissue using aspartoacylase (ASPA) and co-labeled OLIG2 by double immunofluorescence. To verify ASPA specificity for oligodendrocytes, we conducted ASPA double immunofluorescence alongside GFAP, IBA1, and NeuN on MS brain sections (Supplementary Figure S6). ASPA showed no co-expression with astrocytes (i.e., GFAP⁺ cells), microglia (i.e., IBA1⁺ cells), or neurons (i.e., NeuN⁺ cells) in white or gray matter, and co-localized only with OLIG2-positive cells. Thus, we conclude that ASPA serves as a specific marker of oligodendrocytes in human CNS tissue.

Fig. 6.

Active MS lesions contain ASPA⁺ oligodendrocytes with low or undetectable OLIG2 signal. a Representative anti-PLP and LN3 stained sections to determine MS lesion configuration and inflammatory activity. b A chromogenic anti-OLIG2 immunohistochemistry reveals intensely stained OLIG2⁺ cells in the white matter of control donors. OLIG2^low cells are detectable in the normal-appearing white matter (NAWM) and at the active lesion rim (insets b). NAWM is defined as white matter that shows no alteration in anti-PLP signal compared with controls and has the greatest distance to the lesion on the respective section. c Immunofluorescent double-labeling for ASPA and OLIG2 reveals ASPA⁺OLIG2⁺ oligodendrocytes in the white matter of control donors. ASPA⁺OLIG2^low cells are present in the NAWM (arrows in d; h), while ASPA⁺OLIG2^⁻ cells can be found in the active lesion rim and core (arrows in e; i). g The overall density of OLIG2⁺ cells is significantly reduced in the active lesion edge and core compared to the white matter of control donors. g–i one-way ANOVA followed by Tukey’s multiple comparisons test. MS donors n = 7; control donors n = 5. Scale bars: a PLP and LN3 300 µm, a OLIG2 100 µm; b 50 µm; c 40 µm, insets c 25 µm

Next, we examined OLIG2 expression at the single-cell level using OLIG2 and ASPA immunofluorescence double-labeling, quantified it in defined MS lesion regions, and compared the findings with oligodendrocytes from control white matter. We first quantified OLIG2^lowASPA⁺ oligodendrocytes using predefined workflows. The proportion of OLIG2^lowASPA⁺ cells showed an upward trend in NAWM and was significantly higher at the lesion rim and in the lesion core (Fig. 6h: Ctrl WM 4.56% ± 1.06; NAWM 12.93% ± 2.26; Lesion rim 17.54% ± 2.85; Lesion core 26.31% ± 7.11). When we screened for OLIG2⁻ASPA⁺ cells, we again found a significant proportional increase in the lesion rim and core compared to the white matter of control donors (Fig. 6i: Ctrl WM 1.1% ± 0.21; NAWM 7.21% ± 2.41; lesion rim 7.74% ± 1.96; lesion core 15.83% ± 3.93).

Siponimod treatment associates with preserved OLIG2 protein expression during metabolic stress

We next examined whether pharmacological modulation of sphingosine-1-phosphate (S1P) signaling alters OLIG2 expression in the cuprizone model. We used siponimod, a selective S1PR1/S1PR5 modulator previously shown to limit oligodendrocyte degeneration and enhance cell survival in this model [5]. We first assessed the impact of siponimod treatment on OLIG2⁺ cell density after 1 week of cuprizone intoxication.

The density of OLIG2⁺ cells was reduced after 1 week of cuprizone intoxication in both the vehicle and siponimod groups compared to controls. Consistent with previous reports on the protective effects of siponimod in the cuprizone model [5], siponimod treatment resulted in a significantly higher OLIG2⁺ cell density (Fig. 7b: Ctrl 1766.0 ± 40.0 cells/mm²; 1 week Cup-Veh 936.0 ± 101.4 cells/mm²; 1 week Cup-Sipo 1219.0 ± 79.2 cells/mm²) and a downward trend in ATF3⁺ oligodendroglia (i.e., OLIG2⁺ATF3⁺ cell density) compared to the vehicle group (Fig. 7b: Ctrl 20.9 ± 8.4 cells/mm²; 1 week Cup-Veh 723.4 ± 87.2 cells/mm²; 1 week Cup-Sipo 509.2 ± 111.0 cells/mm²). In line with our previous findings (relate to Fig. 2), the proportion of OLIG2^low cells was significantly higher among OLIG2⁺ATF3⁺ oligodendrocytes than among ATF3⁻ oligodendrocytes in vehicle-treated mice, whereas this difference was absent in siponimod-treated animals (Fig. 7c: 1 week Cup-Veh ATF3⁻ 44.36% ± 5.03; 1 week Cup-Veh ATF3⁺ 84.58% ± 1.49; 1 week Cup-Sipo ATF3⁻ 52.50% ± 6.30; 1 week Cup-Sipo ATF3⁺ 48.97% ± 8.56). Nuclear OLIG2 fluorescence intensity was significantly reduced in ATF3⁺ oligodendrocytes after 1 week of cuprizone intoxication (Fig. 7d: 1 week Cup-Veh ATF3⁻ 2.99 AU ± 0.26; 1 week Cup-Veh ATF3⁺ 2.11 AU ± 0.08); however, this reduction in OLIG2 fluorescence intensity was not seen when cuprizone intoxication was parallel with siponimod treatment (Fig. 7d: 1 week Cup-Sipo ATF3⁻ 2.65 AU ± 0.21; 1 week Cup-Sipo ATF3⁺ 2.70 AU ± 0.29). These findings indicate that siponimod treatment preserves higher nuclear OLIG2 protein levels in stressed oligodendrocytes during early cuprizone intoxication.

Next, we investigated whether siponimod could also prevent the loss of OLIG2 expression in mature CNPase⁺ oligodendrocytes after 3 weeks of cuprizone intoxication (relate to Fig. 3). After 3 weeks, only a marginal reduction in the density of CNPase⁺ cells was observed compared to controls in both groups (Fig. 8b: Ctrl 173.5 ± 9.5 cells/mm²; 3 weeks Cup-Veh 146.3 ± 11.0 cells/mm²; 3 weeks Cup-Sipo 140.2 ± 9.8 cells/mm²). While CNPase⁺OLIG2⁻ cells were virtually absent in controls, approximately one-third of CNPase⁺ cells lacked OLIG2 expression after 3 weeks of cuprizone intoxication and vehicle treatment. In contrast, this increase in CNPase⁺OLIG2⁻ mature oligodendrocytes was not observed in siponimod-treated mice (Fig. 8c: Ctrl 2.98 ± 1.34 cells/mm²; 3 weeks Cup-Veh 41.63 ± 11.0 cells/mm²; 3 weeks Cup-Sipo 5.85 ± 1.77 cells/mm² and Fig. 8d: Ctrl 1.75% ± 0.71; 3 weeks Cup-Veh 29.33% ± 7.69; 3 weeks Cup-Sipo 4.73% ± 1.43).

Fig. 8.

Siponimod prevents the emergence of OLIG2⁻CNPase⁺ oligodendrocytes in the cortex during 3 weeks of cuprizone intoxication. a Representative double-labeling for CNPase and OLIG2 in the mouse cortex after 3 weeks of cuprizone intoxication with or without siponimod treatment. CNPase⁺ cells with intense membranous/perinuclear staining that lack nuclear OLIG2 expression emerge after 3 weeks of cuprizone exposure (arrowheads). b The overall density of CNPase⁺ cells shows a slight tendency to decrease following cuprizone intoxication, yet not statistically significant. c, d In contrast, the number of mature oligodendrocytes completely negative for OLIG2 increases substantially and significantly after 3 weeks of cuprizone intoxication, both in absolute and relative terms. This OLIG2⁻ phenotype is almost completely abolished by siponimod administration. b–d Kruskal–Wallis test followed by Dunn’s multiple comparisons test. All groups n=4. Scale bar: 200 µm, scale bar inset: 15 µm

To strengthen the mechanistic interpretation of the protective siponimod effect, we analyzed markers associated with metabolic stress (i.e., Serpina3n) and apoptosis (i.e., cleaved caspase-3) in the corpus callosum of vehicle- versus siponimod-treated mice (Supplementary Figure S7a, c). At the early 1-week time point, densities of Serpina3n⁺ and cleaved caspase-3⁺ cells were comparable between vehicle and siponimod groups (Supplementary Figure S7b, d). However, by week three, siponimod treatment markedly attenuated the accumulation of Serpina3n⁺ stressed cells and significantly reduced the density of cleaved caspase-3⁺ cells compared to the vehicle group (Supplementary Figure S7b, d).

Together, these data place the siponimod effect on OLIG2 protein levels in the context of reduced stress-associated and apoptotic readouts at the intermediate stage (i.e., 3 weeks) of cuprizone intoxication.

OLIG2 protein loss occurs despite sustained Olig2 mRNA expression

In the next step, we investigated whether the loss of OLIG2 protein expression results from reduced Olig2 mRNA levels or a post-transcriptional mechanism. To address this, we reanalyzed previously published gene array data obtained after 1 week of cuprizone intoxication [32] and RNA sequencing data collected after 3 weeks of cuprizone intoxication with paralleled siponimod treatment [5]. Transcriptional profiling revealed a dynamic induction of stress-associated genes. We observed a significant upregulation of the metabolic stress marker Serpina3n after 1 week of cuprizone exposure (Fig. 9a). Notably, in animals intoxicated with cuprizone for 3 weeks paralleled by Siponimod treatment, Serpina3n mRNA levels were no longer significantly elevated compared to controls (Fig. 9b), in line with the observed suppression of Serpina3n protein accumulation in these animals (refer to Supplementary Figure S7). Regarding the ISR pathway, Atf3 showed a trend toward upregulation at 1 week in the array data (Fig. 9a) and reached significance in the 3 week Cuprizone + Siponimod group (Fig. 9b). However, qPCR analysis confirmed a significant increase in Atf3 mRNA levels already after 1 week (Fig. 9d, h), validating the early transcriptional activation of the ISR.

Fig. 9.

Transcriptome changes and Olig2 mRNA levels after cuprizone intoxication. Volcano plots of differentially expressed genes of 1 week cuprizone samples (a) and 3 weeks Cuprizone + Siponimod samples (b) compared to controls. Note that key oligodendrocyte transcription factors (i.e., Olig2, Olig1, Sox10) remain stably expressed. mRNA expression of Plp1, Atf3, Cnp and Olig2 after 1 week of cuprizone intoxication in the cortex (c–f) and corpus callosum (g–j), respectively. k–i A combined OLIG2 immunohistochemistry and Olig2 RNA in situ hybridization showing OLIG2⁺Olig2RNA⁺ cells in the corpus callosum of control mice (arrows) and OLIG2⁻Olig2RNA⁺ cells after 1 week of cuprizone (arrowheads). Note the apoptotic cell lacking both OLIG2 protein and Olig2 RNA (asterisk). c–j Kruskal–Wallis test followed by Dunn’s multiple comparisons test. l one-way ANOVA followed by Tukey’s multiple comparisons test. a, b n = 4; c–j Ctrl n = 6, Cup-Veh and Cup-Sipo n = 4; l both groups n = 4. Scale bar: 50 µm, insets: 20 µm, high-power insets: 10 µm

Interestingly, both transcriptomic datasets show that mRNA levels of key oligodendrocyte transcription factors, including Olig1, Olig2, and Sox10, remain stable and are neither significantly up- nor downregulated (Fig. 9a, b). This finding contrasts sharply with our protein data, which reveal a marked reduction in OLIG2 protein expression. In contrast, transcripts of oligodendrocyte-specific and myelin-related genes—including Plp1, Cnp, and Aspa—are significantly downregulated, especially after early cuprizone intoxication (i.e., 1 week; Fig. 9a, b).

Atf3 expression was robustly upregulated following cuprizone intoxication in vehicle-treated animals. Cuprizone intoxication in the presence of siponimod treatment also resulted in elevated Atf3 levels; however, this increase did not reach statistical significance (Fig. 9d, h—Cortex: Ctrl 100% ± 31.66; 1 week Cup-Veh 571.6% ± 95.89; 1 week Cup-Sipo 469.3% ± 169.4; Corpus callosum: Ctrl 100% ± 27.6; 1 week Cup-Veh 1,196% ± 220.7; 1 week Cup-Sipo 826.3% ± 134.7).

In contrast to our immunohistochemical results, Olig2 mRNA expression was not significantly altered in either the cortex or the corpus callosum of cuprizone-intoxicated mice. Siponimod treatment led to a minor but statistically significant increase in Olig2 expression in the corpus callosum (Fig. 9f, j—Cortex: Ctrl 100.0% ± 3.8; 1 week Cup-Veh 95.7% ± 7.4; 1 week Cup-Sipo 96.5% ± 9.1, Corpus callosum: Ctrl 100.0% ± 5.2; 1 week Cup-Veh 99.34% ± 5.9; 1 week Cup-Sipo 125.6% ± 12.0). Since bulk RNA data can be confounded by shifts in the relative abundance of different cell types within the analyzed tissue, we next examined whether the observed uncoupling between Olig2 mRNA and OLIG2 protein expression also occurs at the single-cell level. To this end, we performed combined OLIG2 immunolabeling and Olig2 RNA in situ hybridization (RNA-ISH) on the same tissue section. As expected, in control animals nearly all cells positive for Olig2 mRNA also showed anti-OLIG2 immunoreactivity. In contrast, after 1 week of cuprizone intoxication, we observed a significant increase in Olig2 mRNA⁺ cells lacking detectable OLIG2 protein (Fig. 9k). Notably, the density of OLIG2⁺ cells detected in this experiment was lower than that observed in classical immunohistochemistry (compare Fig. 1). We attribute this to several methodological factors: (i) the lower sensitivity of alkaline phosphatase-based IHC compared to DAB-based labeling and (ii) the absence of heat-induced epitope retrieval in the sequential IHC-RNA-ISH protocol, which may reduce detection sensitivity. Consequently, OLIG2⁺ and OLIG2^low oligodendrocytes are likely underrepresented in this experimental setup.

To assess whether stress-induced protein loss extends to other lineage-defining factors, we examined SOX10 and OLIG1—both well-established oligodendrocyte markers. Similar to Olig2, neither Sox10 (Fig. 9a, b and Supplementary Figure S8c) nor Olig1 (Fig. 9a, b) is significantly regulated at the transcriptional level under cuprizone-induced metabolic stress. However, using chromogenic immunohistochemistry, both markers showed a significant reduction in overall cell density during the early phase of cuprizone intoxication (Supplementary Figure S8 and S9). To determine whether this reduction reflects cell loss or specific protein loss within surviving oligodendrocytes, we performed double immunofluorescence with CNPase. OLIG1 immunolabeling revealed a rapid and transient reduction in protein detectability: after 1 week of cuprizone intoxication, the majority (> 80%) of CNPase⁺ oligodendrocytes were OLIG1-negative (Supplementary Figure S9), followed by a return to near-baseline levels at week three. SOX10, in contrast, appeared more stable within the CNPase⁺ population during the early phase of cuprizone intoxication. Although the nuclear signal intensity appeared reduced in many oligodendrocytes at week one (refer to Supplementary Figure S8d, arrows), the majority of CNPase⁺ cells remained detectable as SOX10⁺. Consequently, a significant proportion of CNPase⁺SOX10⁻ oligodendrocytes was only observed at the 5-week time point (Supplementary Figure S8e: 5 weeks Cup 48.15% ± 7.23). These findings indicate that, similar to OLIG2, the oligodendroglial transcription factors SOX10 and OLIG1 are subject to a reduction in protein expression levels in response to metabolic stress, yet with different temporal dynamics.

Discussion

Here we identify a robust, stress-responsive loss of OLIG2 protein in oligodendrocytes that emerges rapidly in vivo and precedes overt oligodendroglial cell degeneration and demyelination. Across a toxin-induced model (i.e., cuprizone), a reversible metabolic paradigm (i.e., chronic starvation), and post-mortem MS lesions, we observe OLIG2^low expressing or OLIG2-negative oligodendrocytes enriched among cells expressing ISR readouts (e.g., ATF3) and by the broader metabolic stress marker Serpina3n. Notably, OLIG2 protein loss occurs without detectable reduction in Olig2 mRNA by qPCR, bulk transcriptomics, or in situ hybridization, indicating a post-transcriptional mechanism that is invisible to RNA-level profiling. Pharmacological modulation of sphingosine-1-phosphate (S1P) signaling with siponimod attenuates OLIG2 protein loss during cuprizone exposure, suggesting that this stress-induced state is pharmacologically pliable. Together, these data delineate an early, potentially reversible phase of oligodendrocyte stress characterized by loss of lineage-defining transcription factors.

The metabolic demands of myelin biogenesis and maintenance render oligodendrocytes highly susceptible to mitochondrial dysfunction and lipid peroxidation—central pathological features of cuprizone toxicity and active MS lesions [12, 36]. Prior studies emphasized loss of myelin proteins and oligodendrocyte death as terminal read-outs of injury [20, 64]. Our findings add an earlier layer: a transient reduction or even loss of OLIG2 protein that accompanies stress-engaged oligodendrocytes, including cells with ISR activation (ATF3) and a broader metabolic stress signature (Serpina3n). A recent study by Nguyen et al. [39] noted a loss of OLIG2 immunoreactivity in mature oligodendrocytes during the chronic phase of cuprizone-induced demyelination (4 weeks of cuprizone). Here, we demonstrate that the OLIG2^low phenotype appears already within the first week of cuprizone-induced metabolic stress. Crucially, this early reduction in OLIG2 protein occurs without detectable changes in Olig2 mRNA, consistent with post-transcriptional regulation.

The observed post-transcriptional OLIG2 loss explains why RNA‑based approaches, including bulk RNA‑seq and single-cell/single nucleus RNA sequencing, may underestimate stressed oligodendrocytes when relying on Olig2 transcript abundance or when inferring cell identity from transcriptomic signatures alone. Importantly, the reversibility of OLIG2 protein loss in our starvation paradigm implies that relief from stress can restore OLIG2 protein levels, arguing against a fixed reprogramming event and supporting a dynamic mechanism. While our data demonstrate recovery from an OLIG2^low state under metabolic stress, it remains unclear whether a complete loss of OLIG2 protein marks oligodendrocytes destined for cell death or represents a potentially reversible condition. Interestingly, Nguyen et al. further reported that OLIG2 immunoreactivity re-emerges during remyelination. They also provided evidence that a subset of OLIG2^low or negative mature oligodendrocytes can remain engaged in myelin production [39]. Together, these observations argue for dynamic post-transcriptional control of OLIG2, and we outline potential mechanisms below.

Loss of a lineage-defining transcription factor such as OLIG2 raises concerns about destabilization of the oligodendrocyte gene‑regulatory network. We consider two (not mutually exclusive) interpretations. First, OLIG2 protein loss may represent an adaptive pause that temporarily reduces energy-demanding transcriptional programs while the cell endures acute metabolic crisis, analogous to synaptic downscaling during neuronal stress [44, 48]. In this scenario, OLIG2^low cells remain primed for recovery once the stress subsides, consistent with the reversibility observed after nutrient restoration. Second, if OLIG2 expression remains persistently depressed or undergoes repeated suppression, the network may cross a threshold toward maladaptive identity loss, reduced myelin–gene expression, and impaired remyelination capacity. This interpretation is supported by the temporal coincidence of peak oligodendrocyte DNA damage in the cortex with the highest absolute density of OLIG2-negative mature oligodendrocytes after 3 weeks of cuprizone intoxication (Figs. 3 and 4).

Our observation that siponimod treatment attenuates the loss of OLIG2 protein expression in cuprizone-intoxicated animals suggests that preservation of oligodendrocyte integrity under metabolic stress is associated with maintenance of lineage-defining transcription factor expression. Rather than indicating a direct stabilization of OLIG2 protein by siponimod, these findings are more consistent with an indirect effect whereby reduced cellular stress permits sustained OLIG2 protein expression. In this context, maintaining a minimal level of lineage-defining transcription factors during metabolic challenge may support oligodendrocyte resilience and functional stability. Pharmacological interventions that alleviate oligodendrocyte stress and thereby prevent the collapse of transcriptional identity could be particularly beneficial in vulnerable microenvironments, facilitating recovery from sublethal injury.

The presence of OLIG2^low expressing or OLIG2⁻ oligodendrocytes in MS lesions suggests that human oligodendrocytes also undergo a transient identity‑dampened state in vivo. Because OLIG2 is widely used to identify oligodendroglia in human tissue [57], its stress‑sensitive loss carries two important implications. Biologically, it may mark cells at risk that remain potentially recoverable if stress is relieved and lineage-defining transcription factors are restored. Methodologically, oligodendrocyte counts that rely solely on OLIG2 immunoreactivity may underestimate the number of oligodendroglial cells in active lesions, thereby biasing lesion staging and mechanistic interpretation.

Although OLIG2 is considered one of the most reliable markers for oligodendroglia in histology, we observed that up to one-third to one-half of mature cortical oligodendrocytes became negative for OLIG2 after cuprizone intoxication (Figs. 3 and 8), and approximately 15% of ASPA⁺ oligodendrocytes lacked OLIG2 in active MS lesions (Fig. 6). Notably, at 3 weeks of cuprizone intoxication, ATF3⁺ cells rarely co-labeled with OLIG2 in vehicle-treated animals, whereas siponimod increased the fraction of ATF3⁺ cells that still displayed detectable OLIG2 signal to roughly one-third (Supplementary Figure S10). This shift is consistent with siponimod delaying the transition from a stressed OLIG2⁺ state to an OLIG2^– state, for example by prolonging the persistence of oligodendrocytes under ongoing metabolic challenge. Alternatively, siponimod may influence OLIG2 protein abundance or detectability in stressed cells (e.g., by mitigating stress-related downregulation), thereby increasing the probability of capturing ATF3⁺ oligodendrocytes within the OLIG2⁺ compartment.

These findings highlight a major challenge in quantifying and phenotyping stressed oligodendrocytes. While incorporating additional lineage markers such as ASPA, CA2, or CNPase can help reduce ‘false negatives’ associated with OLIG2 loss, a multi-marker approach remains essential. Indeed, we demonstrate that markers like CNPase or ASPA can reliably identify mature oligodendrocytes that have become undetectable by anti-OLIG2 immunolabeling. Nonetheless, each marker poses its own limitations—particularly CNPase, where dense myelin can obscure single-cell resolution. Our transcriptomic data (Fig. 9) reveal that unlike Olig2, the mRNA levels of lineage markers, such as Aspa and Cnp, are significantly downregulated under stress. The fact that these proteins remain detectable by immunohistochemistry despite transcriptional repression likely reflects their longer half-lives as structural or metabolic proteins, compared to the rapid turnover of transcription factors like OLIG2 [35, 49]. Thus, although structural lineage markers may persist longer, their eventual loss likely reflects a combination of transcriptional repression and protein degradation, whereas the loss of OLIG2 protein most likely represents a specific and early post-transcriptional event. Consequently, detection of stressed oligodendrocytes may benefit from a combinatorial approach integrating multiple protein markers together with complementary features such as nuclear morphology or mRNA in situ hybridization.

The dissociation between preserved Olig2 transcripts and reduced OLIG2 protein points to post-transcriptional control. In principle, OLIG2 protein abundance is determined by the interplay of its translation, nuclear import and export, post-translational modifications, and proteostatic regulation. Two non-exclusive mechanisms could plausibly account for the rapid depletion of OLIG2 protein under metabolic stress. First, activation of the ISR—which we capture by ATF3/DDIT3—promotes eIF2α phosphorylation and attenuates global cap-dependent translation [1, 2, 23, 55]. Given the relatively short half-life reported for OLIG2 [35], even transient translational repression could rapidly lower nuclear OLIG2 protein levels without measurable changes at the mRNA level. Second, stress-activated signaling may impact OLIG2 protein stability and subcellular distribution. For example, Akt-dependent phosphorylation has been linked to nuclear export of OLIG2 [43] and this phosphorylation-dependent regulation has been discussed in the context of OLIG2 destabilization in neural stem cells [37]. While our data do not distinguish reduced synthesis from enhanced turnover, these pathways provide plausible routes for rapid protein-level changes at stable transcript levels. Interestingly, a similar dissociation between preserved transcripts and reduced protein levels has been reported in cultured OPC exposed to oxidative stress or hypoxia [9, 16], suggesting that stress-induced uncoupling of Olig2 mRNA and protein may be a conserved response across the oligodendrocyte lineage. While our data is suggestive for post-transcriptional alterations, definitive mechanistic insight will require direct measurement of OLIG2 synthesis and degradation rates in vivo.

Several limitations warrant consideration. First, immunohistochemical detection may be influenced by epitope masking or fixation-related artifacts; consistent findings using multiple antibodies partially mitigate this concern but do not eliminate it. Second, siponimod acts on multiple cell types and we did not disentangle S1PR1- vs. S1PR5-dependent effects or differentiate cell-autonomous from non-cell-autonomous mechanisms. Third, we did not determine whether the preservation of nuclear OLIG2 levels represents a functionally meaningful alteration at the cellular level or merely a bystander effect. Fourth, our analysis focused on early time points; it remains to be established whether preserving OLIG2 protein expression confers long-term protection against oligodendrocyte damage or facilitates remyelination. Finally, our human data derive from cross-sectional post-mortem tissue, which cannot resolve temporal dynamics or treatment responsiveness. Additionally, analyses of human post-mortem tissue warrant caution due to technical variables, such as post-mortem delay and fixation duration, which can affect epitope preservation. While our inclusion of appropriate controls and the detection of OLIG2 in NAWM mitigates some technical concerns, the heterogeneity of lesion pathology means that OLIG2 loss should be interpreted as a feature of active stress within a lesion, rather than a uniform marker of all MS pathology.

Mechanistically, it will be important to clarify whether OLIG2 loss arises from translational repression, enhanced proteasomal or autophagic degradation, altered nuclear transport, or specific post-translational modifications under stress conditions. Longitudinal in vivo imaging or fate mapping in OLIG2 reporter lines could determine to what extent OLIG2^low and OLIG2-negative cells re-express OLIG2 and whether they make a meaningful contribution to remyelination, as suggested in previous work [39]. Furthermore, it would be interesting whether siponimod promotes that trajectory. Genetically, receptor-specific manipulations—such as oligodendrocyte-restricted S1pr5 deletion and microglial/astrocytic S1pr1 perturbation—will help delineate the cellular basis of siponimod’s protective effects and its mechanistic impact on OLIG2 expression. Translationally, it will be valuable to assess whether OLIG2 dynamics can be monitored in accessible biospecimens (e.g., CSF-derived extracellular vesicles) or via advanced imaging readouts, and whether early pharmacologic stabilization of OLIG2 correlates with improved myelin stability and remyelination in chronic demyelinating diseases.

We propose that OLIG2 protein loss is a rapid, stress-coupled, and reversible hallmark of oligodendrocyte vulnerability that precedes irreversible degeneration. Because transcript-level profiling cannot detect OLIG2 protein loss, RNA-based datasets may not capture the corresponding protein-level pathology, underscoring the limitations of using OLIG2 as a single lineage marker in pathological contexts. The ability of siponimod to attenuate OLIG2 loss positions S1P receptor signaling as a modulator of oligodendrocyte identity during metabolic stress and motivates therapeutic strategies that maintain lineage-defining transcription factor pools to preserve myelin integrity and potentially enable effective remyelination.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ASPA

Aspartoacylase

ATF3

Activating transcription factor 3

ATP

Adenosine triphosphate

AU

Arbitrary units

CA2

Carbonic anhydrase II

CC

Corpus callosum

CMCS

1% Carboxymethylcellulose sodium solution

CNPase

2′,3′-Cyclic-nucleotide 3′-phosphodiesterase

CNS

Central nervous system

CSF

Cerebrospinal fluid

Ctrl

Control

Cup

Cuprizone

CX

Cortex

DAPI

4′,6-Diamidino-2-phenylindole

DDIT3

DNA damage-inducible transcript 3

DNA

Deoxyribonucleic acid

e.g.

exempli gratia

GFAP

Glial fibrillary acidic protein

γH2AX

Histone H2AX phospho Ser139

H₂O₂

Hydrogen peroxide

HLA-DR

Human leukocyte antigen-DR isotype

i.e.

id est

IBA1

Ionized calcium-binding adapter molecule 1

IHC

Immunohistochemistry

ISH

In situ hybridization

ISR

Integrated stress response

ISRIB

Integrated stress response inhibitor

mRNA

Messenger ribonucleic acid

MS

Multiple sclerosis

NADPH

Nicotinamide adenine dinucleotide phosphate

NAWM

Normal appearing white matter

NG2

Neural/glial antigen 2

NGS

Next-generation sequencing

OLIG1/2

Oligodendrocyte transcription factor 1/2

OPC

Oligodendrocyte progenitor cells

PBS

Phosphate-buffered saline

PERK

Protein kinase R (PKR)-like endoplasmic reticulum kinase

PFA

Paraformaldehyde solution

PLP

Myelin proteolipid protein

PMD

Post-mortem Delay

qPCR

Quantitative real-time polymerase chain reaction

RefSeq

Reference sequencing

ROI

Region(s) of interest

S1PR

Sphingosine-1-phosphate receptor

Serpina3n

Serine (or cysteine) peptidase inhibitor, clade A, member 3N

SOX10

SRY-box transcription factor 10

Tris/EDTA

Tris(hydroxymethyl)aminomethane/ethylenediaminetetraacetic acid

Author contributions

Conceptualization: H.K., L.H. Investigation/methodology and formal analysis: H.K., L.H., E.P., L.W., S.M., V.H., L.B. Resources: H.K., N.B., A.Z., L.F., S.A., M.K. Visualization/figure preparation: H.K., L.H. Writing—original draft: H.K., L.H. Writing—review and editing: H.K., L.H., E.P., N.B., A.Z., L.F., M.K. Supervision: H.K., L.H., M.K. All authors read and approved the final manuscript.

Funding

Open Access funding enabled and organized by Projekt DEAL. Hannes Kaddatz was supported by the Rostock Academy of Science (RAS) and Centre for Transdisciplinary Neurosciences (CTNR), Rostock University Medical Center and received intramural funding (FORUN program #889113, Rostock University Medical Center). Emil Pril has received a scholarship from the Hertie foundation (P1250011, DrPrg SoSe 25 Kipp/Pril). Newshan Behrangi received intramural funding (FORUN program #889103, Rostock University Medical Center). Siponimod was kindly provided by Novartis Pharma AG, Basel, Switzerland, MTA: #PLSMTA19AUG177.

Data availability

All custom analysis scripts are provided in the Supplementary Information. Raw and processed image data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

H.K. received travel support and a speaker honorarium from Novartis. Novartis had no role in study design, data collection/analysis, decision to publish, or preparation of the manuscript. All other authors declare no competing interests.