Monocarboxylate transporter 2 regulates maintenance of myelin and axonal integrity by oligodendrocytes

Abstract

Myelin alterations, tightly linked to axonal degeneration, are common in neurodegenerative diseases, including multiple sclerosis (MS). However, the metabolic mechanisms that sustain white matter integrity remain elusive. Monocarboxylates are important energy fuels, but their role in myelinating oligodendrocyte function remains unclear. Here, we show that myelinating oligodendrocytes express high affinity monocarboxylate transporter 2 (MCT2), which is downregulated in progressive MS. While deletion of MCT2 in the mouse spinal white matter using oligodendrotropic AAV injection does not affect oligodendrocyte survival, it downregulates lipid synthesis-associated enzymes and increases inflammation, leading to a failure of myelin maintenance. These changes, not evidenced in AAV-control mice that only show mild inflammation, are accompanied by axonal upregulation of lactate dehydrogenase A and injury, effects alleviated by ketogenic diet. Therefore, our findings show that oligodendroglial MCT2 regulates myelin maintenance and axonal support under mild inflammation. This appears disrupted in progressive MS but might be compensated for by specific metabolic therapies to preserve white matter integrity.

Metabolic determinants of white matter integrity remain elusive. Here, the authors show downregulation of MCT2 in oligodendrocytes in progressive MS, and then demonstrate its role in myelin maintenance in mice, where the effects of MCT2 loss can be alleviated by a ketogenic diet.

Introduction

Metabolic alterations are a well-recognized feature of many, if not all, neurodegenerative disease¹. These include multiple sclerosis (MS)^2–5, an inflammatory demyelinating disease of the central nervous system (CNS) that represents the leading cause of acquired non-traumatic disability in young adults⁶.

Energy deficit observed in the CNS of patients suffering from neurodegenerative diseases is thought to significantly contribute to brain dysfunction, as the brain is a major consumer of body’s oxygen and glucose (around 20%) while it represents only a 2% of the body’s weight⁷. Synthesis of myelin sheaths by oligodendrocytes imposes a high metabolic cost⁸. Myelinating oligodendrocytes require metabolic fuels to sustain lipid and protein synthesis to ensure myelin maintenance, but also to provide metabolic support to the axons⁹. Because of high metabolic demands of oligodendroglia, the energy deficit described in MS and other neurodegenerative diseases may significantly compromise the function of these cells, including myelination and the support of axonal integrity, as oligodendrocytes and myelin are highly vulnerable to energy deprivation¹⁰. Thus, elucidating metabolic pathways that underlie oligodendroglial function and how these might be affected in MS is crucial to develop therapies to prevent myelin alterations/loss and neurodegeneration, thus disease progression.

Monocarboxylates (MCs), a family of molecules that comprises lactate, pyruvate, acetate, and ketone bodies among others, are increasingly recognized as important metabolic fuels for the CNS^9,11–14. Monocarboxylate trafficking into/out of cells is mediated by monocarboxylate transporters (MCTs). MCTs 1–4 execute the proton-linked transport of MCs across the plasma membrane. MCT1,2, and 4 are expressed in the brain, and out of these, MCT2 has the highest affinity¹⁵. This transporter plays an important role in synaptic function, the most energy-consuming process in the brain, as it allows neurons to import astrocyte-derived lactate to be used as energy fuel, according to the astrocyte-neuron lactate shuttle (ANLS) hypothesis¹⁶. Gene expression studies on CNS cells of both mice and humans have shown that the expression of Slc16a7, the gene that encodes MCT2, is not restricted to neurons as it is also detected in oligodendroglial cells, including mature oligodendrocytes, and in microglia^17–21.

Because of its high metabolic cost, similarly to synaptic function, oligodendroglial myelination might require not only glucose, but also a substantial input of additional metabolic fuels⁹. Literature suggests MCs as important carbon sources of myelin lipids: lactate has been shown as a precursor for oligodendroglial lipid synthesis in vitro²², and ketone bodies are preferentially incorporated into myelin lipids over glucose in young rats²³. Yet, the mechanisms that control MC import into oligodendrocytes are not clear. Oligodendrocytes are considered to express predominantly MCT1, the intermediate affinity MCT. The expression of this transporter has been detected mainly on myelin both ex vivo²⁴ and in vivo²⁵. This transporter has been proposed as the primary mediator of the rescue effect of lactate on myelination under low glucose conditions ex-vivo²⁴. However, oligodendroglia-specific knockout of MCT1 had no effects on myelination in vivo until old age²⁶ suggesting that either MCs are not required for myelination in young and adult mice, or that other MCTs might be involved.

Based on the findings that Slc16a7 is expressed in oligodendrocytes, we hypothesized that this high affinity monocarboxylate transporter might contribute to the metabolic fitness of these cells, enabling them to maintain myelin sheaths and sustain axonal integrity. Here, we demonstrated protein expression of MCT2 on myelinating oligodendrocytes in mice and humans. Next, we investigated potential changes in both gene and protein expression of this transporter in MS and showed that it is decreased even in the normally appearing white matter (NAWM) of patients as compared to control subjects. We then used Cre recombinase-encoding oligodendrotropic AAV injection in the spinal white matter of MCT2^lox/lox mice to induce loss-of-function in vivo and observed that loss of MCT2 in oligodendrocytes in the setting of AAV-induced mild inflammation leads to demyelination in absence of oligodendrocyte death, associated with a decreased expression of lipid synthesis-associated enzymes fatty acid synthase (FASN) and acyl-coenzyme A synthetase short-chain family member 2 (ACSS2). Furthermore, axonal damage was observed, coincident with increased axonal expression of lactate dehydrogenase A (LDHA). Finally, pathological consequences of MCT2 deletion were attenuated by provision of ketogenic diet. We conclude that MCT2 enables oligodendrocytes to efficiently import monocarboxylates to support both the high cost of lipid synthesis for myelin maintenance and provision of axons with metabolic fuels, at least in the presence of mild neuroinflammation, which appears relevant for NAWM in MS. We also suggest that MCT2 loss can be compensated by other, lower-affinity, transporters by increasing extracellular ketone body concentrations via ketogenic diet. Thus, our results provide an insight into oligodendrocyte metabolism that could be exploited to design metabolic therapies to enhance oligodendroglial function in diseases such as MS.

Results

MCT2 is expressed by oligodendrocytes in mice and humans

As mentioned above, the expression of Slc16a7 has been reported in human and mouse oligodendrocytes. Thus, we first aimed to investigate whether this mRNA expression is also translated into MCT2 protein expression by performing immunohistochemical analyses on both mouse and human CNS tissue. To investigate the expression by mouse oligodendrocytes we performed co-labeling studies using oligodendrocyte transcription factor 2 (Olig2) as oligodendroglial marker, adenomatous polyposis coli protein (APC)/CC1 as a marker of mature oligodendrocytes, and 2 commercial antibodies for MCT2. One of these antibodies, suitable for western blot, was confirmed to recognize the band corresponding to the molecular weight of MCT2, on the extract of brain tissue. Moreover, some labelings were also performed by a previously validated home-made antibody²⁷. In all cases, MCT2 positive cells were observed in gray and white matter, both in the spinal cord (Fig. 1A–E) and the brain (Fig. S1A–B). In the gray matter (gm), MCT2+ cells were identified as APC/CC1+ satellite oligodendrocytes (50 ± 13.32% of MCT2+ cells; Fig. 1C), while the rest were NeuN+ neurons (Fig 1B). In the white matter (wm), MCT2 expression was detected on axons, as expected, but also on cells that were identified as oligodendrocytes by APC/CC1 expression (Fig. 1D,E). 83.2 ± 5.11% of the spinal WM oligodendrocytes were positive for MCT2 (Fig. 1F), which constituted 86.47 ± 6.04 % of MCT2 expressing cells in the white matter (Fig. 1G). The remaining MCT2+ cells in the wm were identified as Olig2+APC/CC1- cells (Fig. S1A), indicating that they were oligodendrocyte progenitors (OPCs). We also investigated MCT2 expression on post-mortem sections of human cerebellum. To label human MCT2 we used a previously validated antibody²⁸. As the highest Slc16a7 transcript expression in human CNS is detected on neurons²⁹, we first verified whether our antibody labels neuronal cells. Indeed, and similarly to the results obtained in rodent cerebellum³⁰, we observed robust MCT2 positivity, specifically on neurons in the Purkinje cell layer (PCL)(Fig. S2A), and the glomeruli in the granule cell layer (GCL) (Fig. S2B). Strong MCT2 positivity was detected on the rosettes, structures that consist of post-synaptic granule cell dendrites and pre-synaptic terminals of mossy fibers (Fig. S2B1), consistent with previous observations that neuronal MCT2 in mice is enriched at synapses³⁰. After having confirmed the validity of the antibody in labeling neurons, we investigated MCT2 expression in the white matter in both control subjects and patients with MS. In both, we also observed MCT2 positive cells in the cerebellar white matter. Co-labeling with human oligodendroglia marker SOX10, revealed that these MCT2+ cells were oligodendrocytes (Fig. 1H, Fig. S2C). In addition, these cells were frequently aligned in triplets, typical of interfascicular myelinating oligodendrocytes (Fig. 1H). SOX10+ cells in the GCL also expressed MCT2 (Fig. S2C). Thus, MCT2 is expressed by myelinating oligodendrocytes throughout the mouse CNS, and in the human cerebellar white matter.

Fig. 1. Oligodendrocytes express MCT2 in mouse and human CNS.

A MCT2 labeling on coronal sections of the mouse spinal cord reveals MCT2 positive cells both in the gray matter (GM) and white matter (WM). B Co-labeling for NeuN (red) and MCT2 (green) in the mouse spinal cord gray matter shows MCT2 expression on neurons (NeuN+ MCT2 + , white arrows) and MCT2+ NeuN- cells (purple arrows). C Co-labeling for MCT2 (green) and APC (red) in the gray matter of the spinal cord. MCT2 is observed on neurons (MCT2 + APC-, white arrows) as well as on MCT2 + APC+ oligodendrocytes, including satellite oligodendrocytes, indicated by purple arrows. D Co-labeling for MCT2 (green), DAPI (blue), neurofilament (NF, in red) and MBP (gray) in the mouse spinal cord white matter show that MCT2 is expressed on cells indicated by purple arrows (MCT2 + DAPI + NF-) and on axons, characterized by MCT2 + NF+ staining and indicated by white arrows. E Co-labeling for MCT2 (green) and APC (red) in the mouse spinal cord white matter reveals MCT2 + APC+ oligodendrocytes, indicated by purple arrows. Scale bars for A-D = 20 µm, insets = 10 µm. F Quantification of the percentage of APC+ oligodendrocytes that express MCT2 in the spinal cord white and gray matter of 4–5-month-old mice. G Quantification of the percentage of MCT2+ cells that express mature oligodendrocyte marker APC/CC1 in the spinal cord. n = 3 mice. Data presented as mean ± SEM. H–I Images of human cerebellar white matter. Labeling for MCT2 (green) and Sox10 (purple), marker of oligodendroglia. Co-labeling for Sox10 and MCT2. Scale bar=20 µm, inset scale bar=10 µm. Colabelling was observed in all 8 human subjects analyzed. MCT2 monocarboxylate transporter 2, NeuN Neuronal Nuclei, NF neurofilament, APC Anti-Adenomatous Polyposis Coli, DAPI 4’,6-diamidino-2-phenylindole, MBP myelin basic protein, Sox10 SRY-box transcription factor 10, WM white matter, GM gray matter.

MCT2 gene and protein expression changes in MS

We then investigated whether MCT2 gene and protein expression by oligodendrocyte were altered in MS.

Gene (Slc16a7) expression changes

The analyses of Slc16a7 expression changes in MS was performed in silico using single nucleus RNA-seq data published by Trobisch et al. 2022²⁰. Among the identified cell types, these authors defined four clusters of oligodendrocytes. Two of these were primarily composed of control oligodendrocytes and defined as homeostatic clusters SLC5A11 and LINC01608, while the other two, containing high proportions of reactive oligodendrocytes and defined as clusters HSPA1A and SGCZ, were amplified in MS (Fig. 2A). We first investigated Slc16a7 expression across the 4 clusters, after which we compared levels of expression among the clusters. We observed that the two homeostatic clusters present higher expression than the two reactive ones, with significant differences detected between SLC5A11 and the rest of the clusters (Fig. 2B; adjusted p values SLC5A11 vs LINC01608: 0.002, SLC5A11 vs HSPA1A: 0.0162 and SLC5A11 vs SGCZ: 4.06e-8). Of note, SLC5A11 was validated in the original publication as the cluster predominantly associated with control white matter. We then investigated potential changes in Slc16a7 expression in patients with MS versus controls. In the SLC5A11, the homeostatic cluster enriched in the white matter, Slc16A7 expression significantly decreased in patients with MS (Fig. 2C; p value: 0.034), whereas the comparison of general oligodendrocyte Slc16a7 expression levels (all clusters combined) showed only a tendency for a decrease in the MS group. As there was a large age gap within both control and MS groups (35-88 years), the subjects were then stratified by age, with the cut-off at 65 years (that allowed us to have sufficient subjects in both “young” and “old” groups, for both controls and patients with MS). In subjects younger than 65 years, Slc16a7 mRNA levels were similar between controls and MS patients (adjusted p value: 0.509). However, with age, MS patients older than 65 years showed a significantly decreased Slc16a7 expression levels compared to controls older than 65 (Fig. 2D; adjusted p value: 0.0406). Moreover, because our analyses included oligodendrocytes from different regions (brain vs spinal cord), and regional differences in gene expression among different human oligodendrocyte populations have been documented³¹, we then re-analysed Slc16a7 expression by oligodendrocytes from the cortex and the spinal cord, including age as a control variable in the statistical models, in order to investigate whether these regions were differentially affected. While in the cortex we observed only a non-significant tendency for a decrease, in the spinal cord, in which oligodendrocytes maintain longer internodes³² thus have larger synthetic needs, a larger decrease in Slc16a7 expression was detected (Fig. 2E–F; p value cortex: 0.66, p value spinal cord: 0.08).

Fig. 2. Slc16A7 expression changes in progressive MS.

A Number of oligodendrocytes (Y-axis) analysed by cluster (X-axis) and condition (bar color); n (cells/subjects) = SLC5A11 (control=1621/10, MS = 1230/10), LINC01608 (control=2163/10, MS = 1593/9), HSPA1A (control=722/10, MS = 2629/10), SGCZ (control=226/8, MS = 877/10). B–F Normalized and logarithmically scaled Slc16A7 gene expression values across different comparisons Statistical analyses were performed using Seurat and DESeq2. Differential expression (performed on raw pseudobulk counts) was assessed using a two-sided negative binomial generalized linear model (as implemented in DESeq2) with Wald test to determine statistical significance. In the case of multiple comparisons, Benjamini–Hochberg (false discovery rate <0.05) test was used to adjust p-values. Values were Log10-transformed for visualization purposes, and these are shown in figures. B oligodendrocyte clusters (adjusted p values SLC5A11 vs LINC01608: 0.002, SLC5A11 vs HSPA1A: 0.0162 and SLC5A11 vs SGCZ: 4.06e-8, Benjamini–Hochberg test; n cells per cluster: OL-SLC5A11 (n = 2851 from 20 subjects), OL-LINC01608 (n = 3756 from 19 subjects), OL-HSPA1A (n = 3,351 from 20 subjects), and OL-SGCZ (n = 1103 from 18 subjects). C SLC5A11 (white matter-enriched) oligodendrocyte cluster (p value: 0.034, Benjamini-Hochberg test; control (n = 1621 cells), MS (n = 1230 cells) from 10 subjects each), D stratification by age groups (adjusted p value: 0.0406, Benjamini-Hochberg test; n cells per group: control over 65 (n = 2262 from 4 subjects), control under 65 (n = 2470 from 6 subjects), MS over 65 (n = 893 from 3 subjects), MS under 65 (n = 5463 from 7 subjects), E cerebral cortex (n = 3370 cells from 4 subjects for control and n = 5257 cells from 4 subjects for MS); F spinal cord (n = 1237 cells from 3 subjects for control and n = 893 cells from 3 subjects for MS). Boxplots represent the median (horizontal line) and interquartile range (box, 25^th and 75^th percentiles). Whiskers extend to the smallest and largest values within 1.5 times the interquartile range. Outliers are shown as large black dots, while small gray dots represent individual samples. Statistical significance is indicated by * for p-value (single comparison) or adjusted p-value (multiple comparisons) <0.05. MS multiple sclerosis, OL oligodendrocytes.

Thus, these data suggest a decrease in Slc16a7 expression by oligodendrocytes in patients with progressive MS that affects preferentially the spinal cord and appears attributable to both the appearance of reactive oligodendrocyte clusters, as well as the decrease in the expression within the homeostatic cluster SLC5A11.

MCT1 is another MCT that imports monocarboxylates, although with a significantly lower affinity than MCT2¹¹. Thus, we also investigated variations in Slc16a1, the gene encoding MCT1, using the same analyses as those performed for Slc16a7. First, we observed that the expression levels of Slc16a1 in oligodendrocytes are lower than those of Slc16a7. In addition, unlike Slc16a7 that was enriched in the white matter cluster SLC5A11, Slc16a1 expression was significantly lower in this cluster as compared to the homeostatic cluster LINC01608 and reactive clusters HSPA1A and SGCZ (adjusted p values = 0.00029, 0.016, and 4.06e-8, respectively) (Fig. S3A). No significant differences in the Slc16a1 expression within the white matter cluster were detected between controls and MS (Fig. S3B). The analyses with the age cutoff at 65 years showed a non-significant decreasing tendency (p = 0.0954) between controls vs patients older than 65 years (Fig. S3C). In addition, while a non-significant increase in Slc16A7 expression had been observed in the controls with age (Fig. 2D), changes in control Slc16a1 expression with age were not observed (Fig. S3C). Although age-adjusted regional analyses of Slc16a1 expression changes in oligodendrocytes in MS showed decreasing tendencies, these differences were far from reaching statistical significance both in the cortex (p = 0.42; Fig. S3D) and the spinal cord (p = 0.28; Fig. S3E). Thus, the pattern of Slc16A1 expression among human oligodendrocyte clusters is different from that of Slc16a7, and the extent of change in progressive MS is much lower.

MCT2 protein expression changes in MS

We then examined MCT2 protein expression in post-mortem sections of the human cerebellum in control subjects versus patients with MS. Clinical features of the subjects analysed are presented in Table S1. Images of sections cut from the blocks were first stained with Luxol Fast Blue (myelin staining), cresyl violet (neuronal cell bodies), and major histocompatibility complex II (MHC II; marker for inflammatory cells) to characterize the tissue. We used Sox10 antibody as a marker for oligodendroglia. SOX10+ and MCT2+ cells were quantified at the rims and lesion cores of chronic and chronic active lesions, and within the normal appearing white matter (NAWM) of MS and non-neurological controls (Fig. 3D).

Fig. 3. Changes in MCT2 expression by oligodendroglia in patients with progressive MS.

A SOX10+cell numbers in MS NAWM vs controls. Mean ± SEM. B Lines represent different patients; in 2/4 patients, SOX10+numbers increase perilesionally, 1/4 shows no change, and in 2/4 SOX10+ numbers decrease. Friedman test P = 0.0417. A significant downregulation of SOX10 was observed in the lesion core compared to the rim (perilesion), *p = 0.04, Dunn’s multiple comparison. All patients show SOX10+ cell depletion in the lesion core. C Co-immunohistochemistry for SOX10 (red) and MCT2 (green) in the white matter of a control subject. D Sections of of cerebellar tissue of the patient with MS harboring a chronic active lesion. Low power image shows Luxol fast blue (light blue) staining for myelin, Cresyl Violet (dark blue) staining for neuronal cell bodies, and MHC II (black) staining for inflammatory cells. Squares indicate areas analyzed on adjacent sections by immunofluorescence. Images of co-immunolabelling for SOX10 (red) and MOG (gray) show Sox10+ cells are present in the NAWM and perilesion but are largely depleted from the lesion core. Co-immunohistochemistry for SOX10 (red) and MCT2 (green) in the NAWM (E)and perilesion (F). G, H Quantification of MCT2/SOX10 colocalization. G The percentage of SOX10+ cells that express MCT2 in control subject white matter is significantly higher than in the MS NAWM, two-sided Welch’s unpaired t-test, Mean ± SEM. *p = 0.0246, and H Comparison of SOX10+ cells that express MCT2 between MS NAWM, perilesion and lesion core, Friedman test (P = 0.0093). Dunn’s multiple comparisons test detects significant differences between NAWM and the lesion core, *p = 0.0240. Controls N = 4 and MS patients N = 4. C, E, F Purple arrows: MCT2+cells; light blue arrows =MCT2- cells. Scale bars: C, E, F 20 µm, D 1mm, D insets 50 μm. NAWM normal appearing white matter. PL perilesion. L lesion core.

Although numbers of SOX10+ cells per area were generally increased in the NAWM of MS cases versus controls (669.2 ± 55.55 cells/mm² in controls and 1325 ± 406.2 cells/mm² in MS), the variability among MS cases was important, thus the differences between the two experimental groups were not significant (Fig. 3A). We then investigated the numbers of SOX10+ cells in the lesion rim or perilesion and the lesion core of different MS cases and compared them to those in the corresponding NAWM. The number of SOX10+ cells per area in the perilesion was 1436 ± 575.0, and in the lesion core decreased to 43.52 ± 41.43. Statistical differences were found between the perilesion (rim) and the lesion core (Fig. 3B).

Then, we analysed the co-localization between MCT2 and SOX10, in controls versus in MS NAWM and changes in MS tissue according to the lesion proximity (Fig. 3C–H). The percentage of SOX10+ cells that express MCT2 decreased from 49.92 ± 8.093% in controls to 16.84 ± 7.566% in NAWM MS (p = 0.0246) (Fig. 3G). Moreover, this percentage in patients with MS further decreased to 3.133 ± 1.145% in the rim (perilesion) and to 0% in the lesion core. A statistically significant difference was detected between the NAWM and lesion core.

Thus, MS patients show a decrease in MCT2 expression by oligodendroglial cells in the NAWM compared to the controls, and this expression becomes negligible in the lesion proximity (perilesion), although oligodendroglial cells are present in significant numbers. Importantly, chronic lesion core contains very low numbers of SOX10+cells that do not express MCT2.

Altogether, we conclude that MCT2 is expressed by myelinating oligodendrocytes throughout the mouse central nervous system and in the white matter of the human cerebellum, and that this expression decreases in progressive MS.

MCT2 deletion in myelinating oligodendrocytes in the mouse spinal cord leads to demyelination

MCT2 is a high affinity transporter for lactate and ketone bodies. These molecules have been associated with lipid synthesis in oligodendrocytes^22,23,33. Thus, we hypothesized that MCT2 may be important for lipid synthesis and myelin maintenance. To investigate the role of MCT2 in myelin maintenance, we aimed to delete the Slc16a7, the gene coding for MCT2, in myelinating oligodendrocytes. For that, we used MCT2^lox/lox mice³⁴ and the oligodendrotropic AAV, Olig001, with capsid tropism for myelinating oligodendrocytes^35–37. As spinal cord oligodendrocytes maintain long internodes³², and we detected a preferential Slc16a7 decrease in the spinal cord in MS (Fig. 2F), we tested the effect of Slc16a7 deletion in the spinal cord oligodendrocytes. Thus, we injected the Olig001 construct carrying the Cre recombinase tagged with GFP (Olig001-Cre-GFP) or the control Olig001-GFP AAV into wildtype (wt) or MCT2^lox/lox mouse dorsal funiculus (Fig. 4A). Mice were sacrificed at 3 weeks post injection, to allow adequate AAV expression and recombination. GFP expression in the white matter was observed in both groups, locally around the injection site (Fig. S4A). Transduced cells were Olig2+ (62.19 ± 14.86%)(Fig. S4D) and APC/CC1+ (76.42 ± 16.38%) (Fig. 4B), demonstrating their identity as oligodendroglia and post-mitotic oligodendrocytes, respectively. GFP label never colocalized with IBA1, a microglia/macrophage marker (Fig. S4E). Transduced cells were also negative for astrocyte marker GFAP (Fig. S4F), except for occasional cells (less than 1%) in some animals. We also verified the expression of Cre- recombinase using Cre-specific antibodies and observed co-expression with GFP in Olig001-Cre-GFP injected, but not Olig001-GFP injected animals (Fig. 4C). Thus, using this experimental paradigm, we successfully expressed Cre in mature oligodendrocytes in the white matter that we could follow by monitoring GFP expression.

Fig. 4. MCT2 deletion in myelinating oligodendrocytes using AAV-mediated Cre-Lox approach.

A Schematic presentation of the experimental strategy employed to delete Slc16a7 in mature oligodendrocytes. Wildtype (wt) and MCT2^lox/lox mice were injected with oligodendrotropic Olig001-AAV carrying either GFP or Cre-GFP construct in the spinal cord dorsal white matter. Mice were sacrificed at 3 and 6 weeks post injection. B Co-immunolabelling for GFP in green and mature oligodendrocyte marker APC/CC1 in red. Arrowheads indicate double-labelled cells. C Co-labeling for GFP (green) and Cre-recombinase (magenta) in MCT2^lox/lox mice injected with Olig001-GFP or Olig001-Cre-GFP.Arrowheads indicate double-labelled cells, only apparent in Olig001-Cre-GFP injected MCT2^lox/lox mice. D Co-immunolabelling for GFP in green and MCT2 in magenta. White arrows indicate GFP+ cells. Yellow arrows indicate GFP- cells. MCT2^lox/lox mice injected with Olig001-GFP show comparable intensity of MCT2 labeling between GFP+ and GFP- cells. WT mice injected with Olig001-Cre-GFP also showcomparable intensity of MCT2 labeling between GFP+ and GFP- cells. In MCT2^lox/lox mice injected with Olig001-Cre-GFP (AAV-MCT2KO mice), while GFP- cells show MCT2 staining comparable to the AAV-controls, only residual staining is observed on GFP+ cells. E The proportion of GFP+ cells that are MCT2+ significantly decreases in AAV-MCT2KO compared to AAV-controls. Two-sided Unpaired student t-test, Mean ± SEM (*) p = 0.0007. n = 4 mice per group. All scale bars=10 µm. Mouse drawing in A is from Pixabay https://pixabay.com/vectors/search/lab%20mouse/.

We also observed variable degrees of transduction in gray matter (Fig. S4B,C), likely due to AAV spreading. In the gray matter, transduced cells were identified as APC/CC1+ oligodendrocytes, but also as APC/CC1- cells, of neuronal morphology.

We then investigated whether MCT2 protein expression was diminished in recombined cells. We performed co-immunolabellings for MCT2 and GFP. MCT2 expression was observed on GFP+ cells, indistinguishably from GFP- cells in the white matter of MCT2^lox/lox mice injected with Olig001-GFP AAV, as well as in wt mice injected with Olig001-Cre-GFP (Fig. 4D). Yet very weak or absent MCT2 labeling was observed on many GFP+ cells in MCT2^lox/lox mice injected with Olig001-Cre-GFP, while MCT2 expression was present on GFP- cells. The proportion of cells positive for MCT2 among the GFP+ population decreased from 83.60 ± 3.52% in MCT2^lox/lox mice injected with Olig001-GFP (from herein referred to as AAV-controls) to 33.61 ± 9.94 in MCT2^lox/lox mice injected with Olig001-Cre-GFP (from herein referred to as AAV-MCT2KO) (Fig. 4E). Thus, Olig001-Cre-GFP transduction in MCT2^lox/lox mice successfully decreased levels of MCT2 in transduced cells.

We then investigated the evolution of axonal and myelin protein expression following MCT2 deletion in white matter oligodendrocytes. For this, we generated an additional cohort of AAV-control vs AAV-MCT2KO mice that we sacrificed 6 weeks after Olig001-AAV injection. Co-immunolabellings for GFP, SMI31 (phosphorylated neurofilament) and myelin basic protein (MBP) revealed transduced oligodendrocytes surrounded by myelinated axons in the AAV-controls at both 3 and 6 weeks. Yet, in the AAV-MCT2KO, GFP+ cells were frequently observed within the areas of myelin loss, already evident at 3 weeks, and still present at 6 weeks (Fig. 5A). Quantification of MBP+ area fraction revealed a significant decrease in AAV-MCT2KO injected white matter at both time points, decreasing those values from 63.98 ± 7.58% to 38.85 ± 2.98% at 3 weeks and from 61.84 ± 1.78% to 32.90 ± 1.89% at 6 weeks (Fig. 5B).

Fig. 5. Demyelination, axonal injury and microgliosis in AAV-MCT2 KO mice.

A Co-labeling for GFP in green, SMI31 in red, and MBP in gray in AAV-control and AAV-MCT2KO mice at 6 weeks post Olig001-AAV injection. B The percentage of MBP- labelled area within the total WM area significantly decreases in AAV-MCT2KO at 3 weeks (Student unpaired t-test (**) p = 0.0074, n = 6 AAV-controls vs 7 AAV-MCT2KO, Mean ± SEM) as well as at 6 weeks (Student unpaired t-test (***) p < 0.0001, n = 3 AAV-controls vs 5 AAV-MCT2KO, Mean ± SEM). C No significant changes in the percentage of SMI31-labelled area between the groups (n = 3 for all cases except for 6 weeks AAV-MCT2KO, where n = 5). D Co-labeling for GFP in green and SMI32 in magenta in AAV-control and AAV-MCT2KO mice at 6 weeks post Olig001-AAV injection. E White matter area fraction labelled with SMI32 is significantly increased in AAV-MCT2KO at 3 weeks [(*) p = 0.0272 (two-tailed Welch’s t test, Mean ± SEM)] and 6 weeks after Olig001-AAV injection [(*) p = 0.0357 (two-tailed Mann–Whitney test. Mean ± SEM.)]. F Co-labeling for GFP in green and IBA1 in red in AAV-control and AAV-MCT2KO mice at 3 weeks post Olig001-AAV injection. G The percentage of Iba1 labelled area within the total white matter area is significantly increased at 3 weeks post Olig001-AAV injection in AAV-MCT2KO mice (two-tailed Student unpaired t-test, (**) p = 0.0018, Mean ± SEM), but not at 6 weeks (p = 0.1244). n = 3 for AAV-controls at both time points, n = 4 for AAV-MCT2KO at 3 weeks and n = 5 for AAV-MCT2KO at 6 weeks. All scale bars=10 µm.

Axonal injury in AAV-MCT2KO mice

We then investigated axonal integrity by performing immunolabellings for SMI31 (phosphorylated neurofilament) and SMI32 (non-phosphorylated neurofilament). Although we observed slight disorganization of SMI31 staining in AAV-MCT2KO mice (Fig. 5A), quantification of percentage area labelled with SMI31 revealed no significant differences between the groups neither at 3 nor at 6 weeks (Fig. 5C). We then assessed axonal damage by performing immunohistochemistry for SMI32, associated with axonal pathological states³⁸. The percentage of the area labelled with anti-SMI32 significantly increased in AAV-MCT2KO mice both at 3 and 6 weeks (Fig. 5E; AAV-controls 4.87 ± 0.79%, AAV-MCT2KO 13.05 ± 1.74%, *p = 0.027 at 3 weeks; AAV-controls 0.70 ± 0.16% vs AAV-MCT2KO 4.62 ± 0.96%, *p = 0.0357, 6 weeks), and swollen axons positive for SMI32 were evident in AAV-MCT2KO but not AAV-control mice (Fig. 5D–E).

Next, we investigated inflammation in AAV-MCT2KO mice. We performed labelings for Iba1+ cells to investigate presence of myeloid cells, in AAV-control vs AAV-MCT2KO mice both at 3 and 6 weeks post Olig001-AAV injection. Iba1+ (myeloid)(Fig. 5F) and CD45+ (hematopoietic/activated microglia) cells (Fig. S7D) were present in the white matter areas of both AAV-control and AAV-MCT2KO mice. The percentage of white matter area occupied by Iba1+ staining significantly increased in AAV-MCT2KO mice at 3 weeks (8.423 ± 1.538% in AAV-controls, 22.46 ± 1.651% in AAV-MCT2KO, **p = 0.0018, Fig. 5F–G). The increase persisted but was no longer significant at 6 weeks (8.65 ± 4.53% in AAV-controls and 18.87 ± 3.50% in AAV-MCT2KOs, Fig. 5G). In addition, quantification of CD3+ cells (T cells) at 6 weeks revealed a trend towards an increase in the AAV-MCT2KO (56.37 ± 45.01 CD3+ cells/mm² for AAV-control and 138.80 ± 52.21 CD3+ cells/mm² for AAV-MCT2 KO). Thus, Olig001-AAV transduction itself in the white matter leads to a low degree of inflammation, which is upregulated in AAV-MCT2KO, particularly at early time points.

To gain a better insight into white matter changes in AAV-MCT2KO, we performed ultrastructural analyses using transmission electron microscopy (TEM). To ensure that the analyses were performed in the transduced white matter area, we combined TEM with immunogold labelings for GFP in AAV-control vs AAV-MCT2KO mice sacrificed at 6 weeks post AAV injection. These analyses (Fig. 6A) revealed that in AAV-controls, oligodendrocytes, identified by their immunogold labeling for GFP and Olig2, exhibited a typical rounded morphology, dark spherical nucleus, with some small chromatin clumps, slightly dense cytoplasm, with abundant ribosomes and short cisternae of rough endoplasmic reticulum (ER) and were surrounded by thick myelin-enveloped axons. Except for the small areas surrounding the needle tract, in which evidence of complete remyelination (thin myelin) was evident, the myelin status in these animals appeared largely normal, and myelin-debris-filled macrophages were absent. Conversely, in AAV-MCT2KOs, oligodendrocytes displayed an altered/reactive morphology, with lighter nuclei and enlarged cytoplasm (Fig. 6A), but persistence of short cisternae of the ER (Fig. 6A red square inset). Of note, it was confirmed that these cells were indeed oligodendrocytes as they were positive for Olig2 (Fig. S5). Demyelinated axons were visible in the vicinity of these cells, as well as microglial cells, some of which had many lipid droplets within their cytoplasm indicating they had ingested myelin debris (arrows). Microglial cells were also observed in contact with myelin sheaths that appeared separated from the axon by an empty space (Fig. 6C, green asterisk), indicating myelin disintegration. Besides demyelinated axons, we also observed axons with very thin myelin sheaths (Fig. 6C, “t”), demyelinated axons showing accumulation of electron-dense bodies indicative of axonal damage (Fig. 6C, red asterisk), and occasional evidence of axonal degeneration reflected by the presence of myelin sheaths with axonal debris inside (Fig. 6C, blue asterisk).

Fig. 6. Immunogold labeling for GFP combined with transmission electron microscopy shows ultrastructural alterations of oligodendroglia, myelin, and axonal integrity in AAV-MCT2KO mice.

A Images of white matter in AAV-control and AAV-MCT2KO mice. Upper left: GFP+ cell in the AAV-control mouse surrounded by myelinated axons. Lower left: Low power images reveal oligodendrocytes and myelinated axons in the AAV-control mouse. Upper right: GFP+ cell in AAV-MCT2KO mouse next to demyelinated axons. Inset shows short, dilated cisternae of rough ER, typical of an oligodendrocyte. Red arrows in A indicate gold particles visualized as dense dots. Lower right: Low power images of AAV-MCT2KO mouse show demyelinated axons (D), numerous microglial cells (mi) and myelin debris filled microglia/macrophages (arrows). B Quantification of myelinated vs demyelinated axons in AAV-controls and AAV-MCT2KOs. The percentage of myelinated axons is significantly decreased in AAV-MCT2KOs (*) p = 0.03, while that of demyelinated axons is significantly upregulated in AAV-MCT2KO (*) p = 0.0308. Two-sided Unpaired t-test, Mean ± SEM. For AAV-controls n = 4 and for AAV-MCT2KO n = 4. C TEM images of AAV-MCT2KO mouse dorsal funiculus. An axon (green *) on which myelin sheath is separated by an empty space, potentially indicating detachment, in close contact with microglial cell. A demyelinated axon with accumulation of electron dense bodies, indicative of axonal injury (red *). An axon with thin myelin is also observed (t). Axonal degeneration reflected by a myelin sheath that contains debris, but not an axon (blue *). Scale bars A = 1,5 µm, C = 2 µm.

We then quantified myelinated and demyelinated axons throughout the dorsal funiculus in which we detected GFP labeling in AAV-control vs AAV-MCT2KO mice. The percentage of myelinated axons significantly decreased from 98.11 ± 0.54% in AAV-controls to 78.13 ± 5.23% in the AAV-MCT2KOs. Conversely, the percentage of demyelinated axons significantly increased from 1.89 ± 0.54% in AAV-controls to 21.87 ± 5.23% in AAV-MCT2KOs (Fig. 6B).

Thus, MCT2 deletion in white matter oligodendrocytes leads to demyelination, axonal damage, and altered oligodendroglial morphology.

Deletion of MCT2 in spinal cord neurons does not lead to demyelination

Although the transduction with Olig001-AAV in the white matter was restricted to mature oligodendrocytes, in the neighboring gray matter we also observed transduced neurons. While these neurons do not project their axons in the dorsal funiculus (thus excluding axonal effects on myelin integrity in that area), it could be that MCT2 deletion in these neurons could alter extracellular environment and induce demyelination in the neighboring white matter. To investigate whether MCT2 deletion in spinal cord neurons leads to demyelination of the neighboring white matter, we deleted MCT2 specifically in these neurons and not oligodendrocytes. We injected the AAV carrying Cre-GFP or the GFP (control) under neuron-specific hSyn promoter, that was previously used to study the effect of MCT2 deletion on hippocampal neurons³⁴, in the spinal cord of MCT2^lox/lox mice using the same injection settings as those used for Olig001-AAV injection (Fig. S6A). As expected, GFP expression was observed only on NeuN+ cells in gray matter, and no GFP expression was observed in white matter (Fig. S6B,D), confirming previously reported neuronal specificity. Colabellings for GFP and MCT2 showed MCT2 + GFP+ cells in AAV-hSyn- GFP injected MCT2^lox/lox mice, but faint or no expression on GFP+ cells in AAV-hSyn-Cre-GFP injected MCT2^lox/lox mice (Fig. S6C, white arrows). In both groups, intense MCT2 positivity was detected on GFP- cells (Fig. S6C, blue arrows). We then analyzed the expression of SMI31 (axonal marker), MBP (myelin marker) (Fig. S6D) and Iba1 (microglia/macrophage marker; Fig. S6E) in the neighboring white matter. Quantification of SMI31+ staining (Fig. S6F), MBP+ labelled area (Fig. S6G), and Iba1+ labelled area (Fig. S6H) did not reveal differences between the AAV-hSyn-GFP- and AAV-hSyn-Cre-GFP injected mice, indicating that myelination as well as axonal and microglia numbers status were similar between the two groups.

Thus, demyelination seen in MCT2^lox/lox mice injected with oligodendrotropic Olig001-Cre-GFP-AAV is due to MCT2 loss in oligodendrocytes and not neighboring neurons.

Demyelination in AAV-MCT2KO mice is not due to oligodendrocyte death

In principle, demyelination can occur either because myelinating cells die or because they fail to maintain myelin. We first investigated whether MCT2 deletion leads to oligodendrocyte apoptosis. We performed co-immunolabellings for cleaved caspase 3 (CASP3), a marker of apoptotic cells, GFP (Olig001-AAV-transduced cells), and Olig2, an oligodendroglial marker. We observed CASP3+ cells both in AAV-control and AAV-MCT2KO mice, mostly at 3 weeks after injection (Fig. S7A). Their numbers did not differ between the groups and markedly decreased at 6 weeks (at 3 weeks, 257 ± 118.8 CASP3+ cells AAV-controls vs 216.4 ± 119.2 CASP3+ cells in the AAV-MCT2KO; 6 weeks 57.22 ± 23.26 controls group vs 35.52 ± 20.85 MCT2KO; Fig. S7B). Yet, no colocalization was seen between CASP3 and GFP (Fig. S7A) suggesting that transduced cells were not the ones undergoing apoptosis. In addition, GFP+ cell numbers were similar between AAV-control and AAV-MCT2KO mice both at 3 and 6 weeks (200.4 ± 25.74 in controls versus 206.0 ± 29.64 in MCT2KOs at 3 weeks and 191.20 ± 19.54 in AAV-controls versus 192.0 ± 23.51 in AAV-MCT2KOs at 6 weeks; Fig. S7C), showing that transduced cells persist in the AAV-MCT2KO thus excluding the possibility that MCT2-deficient oligodendrocytes may be undergoing non-apoptotic cell death. We then verified whether non-transduced oligodendrocytes were CASP3+ but did not observe CASP3 expression on Olig2+ cells. Moreover, the total numbers of mature oligodendrocytes (APC/CC1 + ) were not statistically different between the groups (APC+ cells/mm² for AAV-controls 440.0 ± 58.79, and for AAV-MCT2KO 359.8 ± 63.37). Thorough characterization of CASP3+ cells identity revealed these were CD45+ haematopoietic cells in both groups (Fig. S7D), which is consistent with previous reports showing that infiltrating inflammatory cells undergo apoptosis in the CNS^39,40.

Thus, MCT2 deletion in oligodendrocytes does not lead to oligodendroglial death which suggests that MCT2KO oligodendrocytes survive but fail to properly maintain myelin.

Knockdown of Slc16a7 (mct2) in rat oligodendroglial cultures does not affect lineage progression but diminishes the extent of MBP+ sheaths

We then aimed to explore the effect of MCT2 deletion in cultured oligodendrocytes on the expression of markers of oligodendroglial maturity and extension of pseudo-myelin sheaths. Thus, we performed shRNA-mediated knock-down of Slc16a7 in primary rat oligodendroglia and investigated its effects at 5 days in vitro.

To perform loss of Slc16a7 function, we used a combination of four shRNA encoding lentiviruses targeting three distinct exons of the rat Slc16a7 gene and co-expressing a GFP reporter (referred to Slc16a7-KD) or with a scrambled RNA-encoding lentiviral vector co-expressing a GFP reporter (referred to as Control). At 5 days in vitro, transduced cells (GFP + ) were visible in both control and Slc16a7-KD cells (Fig. S8A) and no significant differences were observed in the numbers of GFP+ cells. RT-qPCR for Slc16a7 gene showed a statistically significant 2.4-fold decrease between control vs Slc16a7-KD cultures (Fig. S8B) thus validating the Slc16a7 knock-down. RT-qPCR for Mbp, Plp1, and Myrf (Fig. S8C–E), markers of oligodendroglial maturity, showed no significant differences between the two groups. We also performed immunocytochemical analyses for MBP (pseudomyelinating oligodendroglia marker) and Olig2 (marker of oligodendroglia). The proportion of Olig2+GFP+ cells positive for MBP was similar between control and Slc16a7-KD cultures (Fig. S8F). We then quantified the surface area occupied by MBP+ oligodendroglial processes in transduced cells (GFP + ), and we observed a significant decrease in the area occupied by MBP+ processes in Slc16a7-KD cultures (Fig. S8A,G). Thus, while Slc16a7-KD in vitro does not affect oligodendroglial numbers nor their maturation state, it significantly decreases the amount of MBP+ pseudomyelin sheaths maintained by these cells.

AAV-MCT2KO oligodendrocytes show reduced expression of enzymes involved in fatty acid synthesis

As our results showed demyelination in absence of oligodendroglial death in AAV-MCT2KO mice, we hypothesized that AAV-MCT2KO oligodendrocytes survive but fail to maintain myelin. As myelin is mostly lipids, we focused on the expression of enzymes involved in fatty acid (FA) synthesis.

We investigated the expression of fatty acid synthase (FASN), an enzymatic complex dimer that catalyzes the first step of FA synthesis, namely the conversion of acetyl-co-A to palmitate (Fig. 7A). We observed extensive FASN expression in white matter (Fig. 7B). In AAV-control mice, the absolute majority of GFP+ cells were also FASN+ (Fig. 7B, yellow arrows). However, in the AAV-MCT2KO mice, numerous GFP + FASN- cells were observed (Fig. 7B, purple arrows). Quantification of GFP + FASN+ cells showed a significant reduction in the proportion of FASN-expressing cells within the GFP population (89.98 ± 2.32% in AAV-control vs 47.57 ± 6.08% in AAV-MCT2KO mice at 3 weeks and 86.20 ± 2.03% in AAV-control vs 55.65 ± 4.95% in AAV-MCT2KO at 6 weeks, Fig. 7C).

Fig. 7. Reduced expression of 2 major enzymes that regulate fatty acid synthesis in AAV-MCT2KO oligodendrocytes.

A Schematic representation of fatty acid synthase (FASN) and acyl-coenzyme A synthetase short-chain family member 2 (ACSS2) involvement in fatty acid synthesis. ACSS2 regulates lipid synthesis via two distinct, not mutually exclusive mechanism. First, ACSS2 catalyzes the conversion of acetate (that can be imported via MCT2) to acetyl co-A. Acetyl Co-A then can be used as a substrate for, on one hand, palmitate synthesis by FASN (1^st step of fatty acid synthesis), or on the other hand, for histone acetylation which enhances the expression of the genes that regulate lipid synthesis, including the gene encoding FASN. B Co-immunolabelling for GFP in green and FASN in gray. Yellow arrows indicate GFP + FASN+ cells, purple arrows indicate GFP + FASN- cells. C Quantification of the percentage of GFP+ cell expressing FASN at 3 weeks and 6 weeks. At 3 weeks (***) p = 0.0002, two-tailed Student’s unpaired t-test. Mean ± SEM. n = 5 for AAV-controls and n = 7 for AAV-MCT2KO. At 6 weeks, (**), p = 0.0040, two-tailed Student’s unpaired t-test. Mean ± SEM. n = 3 for AAV-controls and n = 5 for AAV-MCT2KO. D Co-immunolabelling for GFP in green and ACSS2 in red. White arrows indicate GFP + ACSS2+ cells and are more numerous in AAV-controls. E Quantification of the percentage of GFP+ cells expressing ACSS2 shows a significant decrease in AAV-MCT2KO at 3 weeks, (*) p = 0.0440, two-tailed Student unpaired t-test and at 6 weeks (**) p = 0.0070, two-tailed Student unpaired t-test. Mean ± SEM. n = 5 and 3 in AAV-controls at 3 and 6 weeks respectively and n = 4 and N = 5 in AAV-MCT2KO at 3 and 6 weeks, respectively. All scale bars = 20 µm.

We then investigated the expression of another enzyme involved in lipid synthesis, acyl-coenzyme A synthetase short-chain family member 2 (ACSS2), which catalyzes the production of acetyl-CoA from acetate (Fig. 7D). Besides lipid synthesis, ACSS2 is crucial for energy generation and histone acetylation. We observed expression of ACSS2 oligodendrocytes in the Olig001-AAV transduced white matter, some of which were GFP+ cells (Fig. 7D, white arrows). In AAV-control mice, 64.52 ± 8.23% of GFP+ cells were ACSS2 + . However, in the AAV-MCT2KO mice, the proportion of ACSS2 expressing cells within the GFP population decreased to 26.73 ± 13.96% at 3 weeks. At 6 weeks, 53.33 ± 11.59% of AAV-control GFP+ cells in the WM were ACSS2 + , while in the AAV-MCT2KO these values decreased to 14.93 ± 3.32% (Fig. 7E). Therefore, our results show that demyelination after MCT2 deletion in oligodendrocytes is not due to oligodendroglial death but appears correlated to significantly reduced levels of two important enzymes that regulate the process of lipogenesis (FASN and ACSS2), suggesting a failure in myelin maintenance.

Upregulation of LDHA in AAV-MCT2KO mice

We hypothesized that absence of MCT2 would reduce the ability of oligodendrocytes to import monocarboxylates and use them as metabolic fuels, forcing these cells to increase glucose metabolism. Thus, we then investigated the expression of lactate dehydrogenase A (LDHA), enzyme that converts pyruvate into lactate, the expression of which increases when glycolysis is upregulated⁴¹. Gene expression data¹⁹ and a recent study²⁵ show that oligodendrocytes under control conditions express very low levels of ldha. We performed co-labelings for LDHA and GFP in AAV-control vs AAV-MCT2KO mice at 6 weeks post Olig001-AAV injection. Both in AAV-controls and AAV-MCT2KO LDHA expression was observed on blood vessels and cells of macrophage/microglial morphology (Fig. 8A). However, in AAV-MCT2KO but not in AAV-controls, LDHA expression was also observed on a minor subset of GFP+ oligodendrocytes (Fig. 8B). This expression was strong in a low proportion of cells (arrow in Fig. 8B), and weaker in others (Fig. 8B arrowheads). Quantification of the percentage of GFP+ cells that were positive for LDHA showed a significant difference between the AAV-control group and the AAV-MCT2KO group (*p = 0.0333), although this percentage remained low (Fig. 8D). An important difference in LDHA expression pattern between the AAV-controls and AAV-MCT2KO was that, in the latter, strong LDHA expression was detected on axons positive for SMI32 + , particularly the swollen ones (Fig. 8C). Quantification revealed a significant increase in LDHA+ axons in AAV-MCT2KO mice (Fig. 8E; *p = 0.0199).

Fig. 8. Increase in LDHA expression in AAV-MCT2KO.

A Immunolabelling for LDHA in the dorsal funiculus of AAV-control and AAV-MCT2KO mice. Dashed lines delimit the white matter. B Co-labeling for LDHA (gray), GFP (green) and APC/CC1 (red) reveals LDHA + , AAV-transduced oligodendrocytes in AAV-MCT2KO. Arrow indicates an AAV-MCT2KO oligodendrocyte labelled by LDHA antibody, arrowheads indicate faintly labelled ones. C Co-labeling for LDHA (gray) and SMI32 (magenta) reveals strong LDHA expression by SMI32+ (damaged) axons in AAV-MCT2KO. Arrows indicate swollen SMI32+ axons that express LDHA. D Quantification of GFP+ cells positive for LDHA reveals absence of colocalization in AAV-controls and an increase in AAV-MCT2KO mice (*p = 0.0333; two-tailed unpaired t test with Welch’s correction). E Significant increase in LDHA+ axons in AAV-MCT2KO mice (*p = 0.0199, two-tailed unpaired t test with Welch’s correction). All data presented as Mean ± SEM, n = 3 for AAV-controls and n = 5 for AAV-MCT2KO. Scale bars A, B = 10 µm, C = 20 µm.

We wondered whether LDHA upregulation in axons occurs because of demyelination, or because of metabolic alterations induced by oligodendroglial MCT2 deficiency. To investigate whether axonal LDHA is simply a consequence of demyelination, we investigated LDHA expression in LPC model of demyelination in the same area (dorsal funiculus). Within the demyelinated area at 14 days post lesion (dpl), we observed LDHA upregulation, with predominantly cellular staining pattern compatible with macrophage localization and morphology in these lesions (Fig. S9A). These observations are in agreement with previous work showing LDHA expression by leukocytes in demyelinating lesions⁴². We did not observe axonal LDHA staining in the lesions. However, axonal pattern of staining was observed in the non-demyelinated tissue neighbouring the lesion (Fig. S9A,B). These observations suggest that axonal LDHA upregulation is not a simple consequence of demyelination.

Thus, our data indicate upregulation of LDHA in AAV-MCT2KO oligodendrocytes and in neighbouring axons, which is associated with axonal damage but does not appear to be a mere consequence of demyelination.

Ketogenic diet attenuates pathological changes in AAV-MCT2KO white matter

We observed that MCT2 deletion in oligodendrocytes results in axonal damage associated with increased LDHA expression, the latter suggesting increased axonal glycolysis and lactate production. Because increased neuronal glycolysis leads to neuronal dysfunction⁴³, we aimed to investigate whether exposure of mice to ketogenic diet, a low glucose diet, could alleviate axonal damage upon AAV-MCT2KO in oligodendrocytes, by reducing glucose availability and shifting neuronal metabolism towards ketone consumption, which was shown as neuroprotective (reviewed in ref. ¹³). Moreover, we also aimed to investigate whether AAV-MCT2KO oligodendrocytes can survive under low glucose/high monocarboxylate conditions, when cells rely on alternative energy fuels such as monocarboxylates.

We first quantified GFP cells to investigate the persistence of AAV-MCT2KO oligodendrocytes under ketogenic diet. We did not observe significant differences between the groups (GFP+ cells/mm²: AAV-control 116.3 ± 42.36, AAV-MCT2KO 143 ± 24.29, AAV-control keto 127.6 ± 32, AAV-MCT2KO keto 139.6 ± 26.46; Fig. 9A,B). Quantification of microglia/macrophages marker IBA1 revealed significant differences between the groups (one way ANOVA P = 0.0371), as well as significant increase in AAV-MCT2KO as compared with both AAV-control and AAV-control keto (P = 0.012 and 0.0163, respectively, Benjamini Hochberg post-test, Fig. S10A). AAV-MCT2KO keto group showed variable amounts of inflammation, and no significant differences were detected between this group and any of the other groups (P = 0.253 (AAV-control vs AAV-MCT2KO keto), 0.334 (AAV-control keto vs AAV-MCT2KO keto), and 0.097 (AAV-MCT2KO vs AAV-MCT2KO keto), Benjamini Hochberg post-test, Fig. S10A). Quantification of the area occupied by MOG staining in 4 groups revealed significant differences (one-way ANOVA 0.0023). Surprisingly, multiple comparisons revealed significant differences not only between AAV-control vs AAV-MCT2KO and AAV-control keto vs AAV-MCT2KO, but also AAV-MCT2KO and AAV-MCT2KO keto (P = 0.0005, 0.0014, and 0.0094, respectively, Benjamini-Hochberg post-test), whereas no statistical differences were observed when comparing AAV-control vs AAV-MCT2KO keto and AAV-control keto vs AAV-MCT2KO keto (Fig. 9A,C), suggesting that myelin loss is attenuated in AAV-MCT2KO keto. LDHA labeling was scarce in both AAV-control and AAV-MCT2KO mice treated with ketogenic diet, detected only on a few cells with mononuclear morphology (Fig. S10B). Importantly, quantification of SMI32 labeling revealed significant differences between AAV-MCT2KO group treated with standard diet with AAV-control and AAV-control keto groups as well as between AAV-MCT2KO group treated with standard diet and AAV-MCT2KO treated with ketogenic diet, (% SMI32 labelled area: 4.87 ± 0.79 AAV-control, 13.05 ± 1.74 AAV-MCT2KO, 5.9 ± 1.04 AAV-control keto, 5.53 ± 2.81 AAV-MCT2KO keto, Fig. 9D). Thus, ketogenic diet diminishes myelin and axonal alterations observed upon MCT2 deletion in oligodendrocytes.

Fig. 9. Ketogenic diet alleviates demyelination and axonal damage in AAV-MCT2KO mice.

A Co-immunohistochemistry for GFP in green and myelin marker MOG in gray, and labeling for SMI32 (non-phosphorylated neurofilament) in magenta, a marker of axonal damage, in AAV-control, AAV-MCT2KO, AAV-control keto, and AAV-MCT2KO keto mice. B Quantification of GFP+ cells per area shows no significant differences between the groups. Mean ± SEM. C Quantification of MOG labelled area shows significant differences between the groups (One-way ANOVA F^3,8 = 12.34b, P = 0.0023**; Benjamini–Hochberg post-hoc test: AAV-control vs AAV-MCT2KO p = 0.0005***, AAV-MCT2KO vs AAV-control keto p = 0.0014**, AAV-MCT2KO vs AAV-MCT2KO keto p = 0.0094**) Mean ± SEM. D Quantification of area labelled with SMI32 shows significant differences between the groups (One-way ANOVA, F^3,8 = 4.65, p = 0.0365*; Benjamini–Hochberg post-hoc test: AAV-control vs AAV-MCT2KO p = 0.012*, AAV-MCT2KO vs AAV-control keto p = 0.022*, AAV-MCT2KO vs AAV-MCT2KO keto p = 0.017*). Mean ± SEM. In all cases n = 3 mice per group. All scale bars 20 µm.

Discussion

Although oligodendroglial expression of MCT2-encoding gene Slc16a7 has been detected in mice^17–19 and humans^20,21, potential function of MCT2 in myelinating oligodendrocytes has not been previously addressed. The expression of this high affinity monocarboxylate transporter by oligodendrocytes could empower these cells to efficiently import monocarboxylates under physiological conditions to sustain their extensive metabolic needs. While oligodendrocytes rely on astrocyte-derived lipids to an extent⁴⁴, myelin generation requires oligodendroglial fatty acid synthesis⁴⁵. Moreover, interfering with lipid synthesis in adult myelinating oligodendrocytes disrupts myelin structure⁴⁶. Although literature highlights that myelin lipid synthesis is significantly supported by ketone bodies during the development²³ and by lactate in vitro²², whether monocarboxylate family contributes to lipid synthesis during myelin maintenance in vivo by mature oligodendrocytes is not clear. Our present data suggest this may be the case and highlight the role for MCT2 in this process.

Myelinating oligodendrocytes have been considered for a long time as lactate producers rather than consumers, and as cells that predominantly rely on glucose consumption, the rate of which depends on axonal activity^47,48. Thus, oligodendroglial glycolysis⁴⁹ and transfer of lactate via MCT1 to the axons⁵⁰ have been evoked as the basis of myelin maintenance and support of axonal integrity in the CNS. However, more recent reports suggest that oligodendroglial MCT1 is dispensable for these processes in non-aged mice²⁶. According to our present work, lack of myelination deficits in non-aged MCT1 oligodendroglial mutants could be explained by MCT2-mediated import of monocarboxylate family of metabolites including lactate, pyruvate, ketone bodies, and acetate, all of which have been indicated as important metabolic supporters of the brain^12,14,51,52. It could be that oligodendrocytes efficiently capture lactate via MCT2 from the extracellular space and shuttle it directly to the axons (as oligodendrocytes do not express LDH thus cannot metabolize lactate²⁵), which also may be the case for ketone bodies. Moreover, because oligodendrocytes express ketolytic enzymes^17,53 and, as we show here, acetate-metabolizing enzyme ACSS2, MCT2-mediated import of ketone bodies and/or acetate might represent an important mechanism underlying the generation of large pools of acetyl Co-A in oligodendrocytes, leading to a state of “nutritional abundance” that will activate lipid synthesis in these cells (Fig. 10A)^54,55. Importantly, here we show that both ACSS2 and FASN (that catalyzes the first step of fatty acid synthesis) are downregulated after MCT2 deletion, suggesting a decrease in lipid synthesis. It has been shown that short chain fatty acids such as acetate^56,57, as well as ketone bodies⁵⁸ induce ACSS2 expression. Thus, reduced capacity of oligodendrocytes to import acetate and/or ketone bodies upon MCT2 deletion could potentially explain the decrease in the expression of ACSS2, which could then decrease the expression of FASN by epigenetic mechanisms⁵⁴. Another mechanism that might explain a decrease in FASN expression is that the reduction in intracellular levels of acetyl Co-A (the substrate of FASN) after MCT2 deletion can be interpreted by the cell as a state of “nutritional deprivation” that triggers mechanisms leading to both decreased expression and increased degradation of FASN^59,60. Such a state of “nutritional deprivation” likely reduces the provision of metabolites to the axons by AAV-MCT2KO oligodendrocytes which, in combination with myelin defects observed, will lead to compensatory upregulation of LDHA by the axons and axonal damage. Thus, our data directly support a link between MCT2, ACSS2 expression and fatty acid synthesis, although other mechanisms of MCT2 action cannot be excluded.

Fig. 10. Hypothetical model integrating the role of oligodendroglial MCT2 in the context of previous findings on white matter metabolism.

A In control white matter, glucose (Glu; light blue circles) and monocarboxylates (ketone bodies (KBs), lactate (Lact), acetate (Ac), pyruvate (Pyr); brown circles) enter parenchyma via transporters expressed by endothelial cells lining the blood vessels (BV). Glucose is imported by astrocytes, lactate is generated and shuttled outside via MCT4¹⁶. Astrocytes also metabolize fatty acids to KBs and export these via MCT4⁹³. Lactate⁹⁴ and KBs can also be transferred to oligodendrocytes via connexin channels. MCT2 expressed by oligodendrocytes, given its high affinity, is predicted to efficiently import monocarboxylates from the extracellular space. Among these, KB, pyruvate and acetate can be metabolized to ATP and/or used as precursors for lipid synthesis. Lactate is likely not metabolized by oligodendrocytes because of low LDH expression²⁵ but may be shuttled via myelinic channels and MCT1 on myelin to axons that transport it via MCT2. Oligodendrocytes also import glucose via Glut1, connexin hemichannels, or gap junctions from astrocytes⁹⁴. Glucose can be directly shuttled to the axons or metabolized to pyruvate, which may then be shuttled to the axons²⁵. B MCT2 deletion in oligodendrocytes is expected to significantly diminish monocarboxylate import under standard conditions, thus decreasing both the availability of carbon molecules and ATP for lipid synthesis. This leads to metabolic deficit that impairs myelin maintenance and likely compromises axonal metabolic support, resulting in axonal LDHA upregulation and damage. C Ketogenic diet increases KB entry to the CNS via endothelial MCT1. Moreover, it induces metabolic reprogramming of astrocytes allowing these to import KB via increased expression of MCT1⁶³. KB could then be delivered to oligodendrocytes via connexin channels. High extracellular KB concentrations might also allow lower affinity transporters such as MCT1, to import KB into oligodendrocytes and sustain lipid synthesis and axonal support. Moreover, ketones can be directly imported and metabolized by neurons/axons⁶³.

We have observed upregulation of LDHA by a small subset of AAV-MCT2KO oligodendrocytes that likely reflects increased glucose utilization through glycolytic pathway (to compensate for decreased monocarboxylate availability) that exceeds oxygen availability in some cells, thus resulting in lactate production. On one hand, this might compromise myelin maintenance because upregulation of glycolysis will likely divert glucose from pentose phosphate pathway (PPP), needed to produce NADPH that is required for lipid synthesis. Reduced rates of PPP might also interfere with antioxidant production, thus decreasing the capacity of oligodendrocytes to deal with oxidative stress and possibly leading to lipid peroxidation, which might interfere with myelination. On the other hand, increased expression of LDHA might simply reflect attempts of oligodendrocytes to compensate for monocarboxylate absence that are, however, insufficient to satisfy oligodendrocyte homeostasis, myelin synthesis and axonal support, at least in our setting of mild neuroinflammation associated with AAV injection.

Demyelination observed after AAV-mediated MCT2 deletion in myelinating oligodendrocytes was not associated with oligodendrocyte death. These results resemble previous work showing that altered nutritional supply in cats and monkeys leads to myelin loss despite oligodendrocyte preservation^61,62. Restoration of normal metabolite supply within a limited time frame in these models resulted in remyelination by pre-existing oligodendrocytes^61,62. While we do not know whether re-expression of MCT2 by oligodendrocytes in our model would restore their myelinating capacity, we did observe that failure of myelin maintenance is attenuated if AAV-MCT2KO mice are subjected to a ketogenic diet. In this scenario, high ketone concentrations might allow for compensation of defective MCT2 by other MCTs or other transporters (e.g., connexin hemichannels). These transporters have lower affinity for monocarboxylates than MCT2 and thus likely cannot compensate for MCT2 deficiency under physiological conditions (low extracellular MC concentrations) (Fig. 10B) but might be able to do so when extracellular ketone concentrations increase, which is expected under ketogenic diet. Moreover, under ketogenic diet, astrocytes undergo metabolic reprogramming and import ketones⁶³, which could then be shuttled through connexin channels to oligodendrocytes (Fig. 10C).

Besides myelination, another crucial function of oligodendrocytes in the CNS is the support of axonal integrity. This support depends not only on the presence of myelin but also on its composition^64–66, presumably because molecular alterations in myelin can affect metabolite and organelle provision to the axons by oligodendrocytes via myelinic channels^67–69. We have observed that upon MCT2 deletion in oligodendrocytes, axons upregulate LDHA. Although basal levels of neuronal LDHA expression appear required for fast axonal vesicular transport⁷⁰, LDHA upregulation by neurons has been previously correlated to increased angiogenesis and neurodegeneration⁷¹, altered cognition in aged mice⁷², as well as to elevated metabolic demand from chronic neuronal activation⁷³. Thus, LDHA upregulation likely reflects compensatory changes in axonal metabolism such as increased axonal glycolysis and lactate production. These adaptations possibly take place to mitigate altered myelination and axonal support by oligodendrocytes that lack MCT2 and, therefore likely enter a “metabolite-sparing” state that prioritizes their own survival over myelin lipid synthesis and metabolic support to the axons. The resulting axonal energy stress might then explain increased expression of axonal damage marker SMI32, which labels the non-phosphorylated form of the neurofilament (NF), given that NF phosphorylation is an energetically demanding process⁷⁴ which might become compromised if axonal metabolic support by oligodendrocytes and/or myelin decreases.

To summarize, while at present time we cannot exclude a direct contribution of oligodendroglial MCT2 to metabolite shuttling from oligodendrocytes to the axons, we believe that the axonal pathology observed after AAV-MCT2KO in oligodendrocytes is the consequence of oligodendroglial metabolic deficit that diminishes both myelin lipid synthesis (which in itself can damage the axons as myelin lipids have been suggested as axonal energy fuels^75–77) and the amount of metabolites to share with the axons. These findings therefore support the link between oligodendroglial and axonal “metabolic wellbeing”.

We observed that LPC-induced demyelination (associated with oligodendrocyte death) does not lead to LDHA upregulation in demyelinated axons, suggesting that myelin loss itself does not upregulate axonal lactate production. Lack of axonal LDHA upregulation in the context of LPC demyelination might be explained by the fact that metabolically stressed oligodendrocytes are absent in the lesions (as, in this model, oligodendrocytes die), and demyelination is complete, which means that demyelinated axons likely import metabolites directly through specific transporters on their surface. In the perilesional white matter, oligodendrocytes and myelin persist, but the proximity of a hypermetabolic area (demyelinating lesion with hypercellularity and ongoing myelin repair) might induce metabolic disturbances reflected by perilesional axonal LDHA expression. This scenario is compatible with previous reports showing that myelin is a risk factor for the axons within inflamed, thus metabolically compromised, environment⁷⁸.

We observed that axonal LDHA expression and damage after AAV-MCT2KO in oligodendrocytes were alleviated by ketogenic diet. Direct neuronal import of ketone bodies during ketogenic diet likely enables neurons to increase ketone utilization, alleviating energy deficit and consequent axonal damage. In line with our data showing axonal protection by ketogenic diet, exposure of mice with experimental autoimmune encephalomyelitis (EAE) to ketogenic diet was found to increase ketolytic metabolism in neurons and astrocytes and was neuroprotective⁶³.

It is important to highlight that monocarboxylates imported by MCT2 may fulfill functions other than that of ATP and/or lipid precursors. Lactate and ketone bodies have been shown as substrates for protein lactylation⁷⁹ and β-hydroxybutyrylation⁸⁰, respectively. Such modifications on histones can affect the expression of genes that regulate metabolism⁸¹, which suggests that even if oligodendrocytes do not express enzymes that allow them to convert lactate into pyruvate (that can be further metabolized in different metabolic pathways), lactate imported via MCT2 could still play an important role in regulating oligodendrocyte metabolism by histone lactylation. Such roles of lactate in oligodendrocytes should be addressed by future studies. Moreover, acetate and ketone body deprivation in MCT2-deficient oligodendrocytes could lead to a decrease in acetyl CoA, thus deficient histone (and other protein) acetylation, further affecting oligodendroglial function.

Our data show downregulation of Slc16a7/MCT2 expression by oligodendrocytes in progressive MS. We first performed in silico analyses of Slc16a7 in different oligodendrocyte clusters described by Trobisch and colleagues²⁰, in control vs MS subjects. These data show predominant Slc16a7 expression in the homeostatic cluster associated with white matter (SLC5A11 cluster), which decreases in MS. Cluster-specific enrichment for Slc16a7 expression in human oligodendrocytes has also been reported previously. Seeker et al.³¹ showed Slc16a7 expression in OligoC cluster, one of the 2 clusters enriched in the spinal cord (https://seeker-science.shinyapps.io/shiny_app_multi/). Importantly. Sadick, O’Dea et al.²¹ showed a significantly enriched Slc16a7 expression in oligodendrocyte cluster 2, the one associated with increased cholesterol (lipid) synthesis (https://liddelowlab.shinyapps.io/Sadick/). Moreover, this expression was unchanged in Alzheimer’s disease²¹. Thus, it might be that Slc16a7 changes in oligodendrocytes are disease specific. Importantly, our analyses of MCT2 protein expression in cerebellar tissue of control subjects vs patients with progressive MS showed a decrease in the percentage of oligodendroglial cells that express MCT2 in the NAWM in MS compared to control subject white matter. In addition, oligodendroglial MCT2 expression further decreased at lesion borders, suggesting that the proximity of lesions leads to MCT2 downregulation in oligodendroglia in progressive MS. These changes will likely impair the capacity of oligodendroglia to import monocarboxylates, which will interfere with their capacity to synthesize lipids and sustain axonal metabolism, as suggested by our loss-of -function data in the mouse model, in which the setting of mild inflammation resembles lesion-proximal NAWM. Thus, our data support the notion that oligodendroglial dysfunction in progressive MS is present beyond lesion areas leading to diffuse myelin/axonal damage in the NAWM that underlies “silent” disease progression⁸².

While our loss-of-function studies suggest that MCT2 is an important regulator of myelination and axonal support by oligodendrocytes, the animal model we used has limitations that must be considered when interpreting the results. We expressed Cre-recombinase using AAVs, which was associated with a basal inflammation state even in the control mice. Therefore, it is possible that this inflammation, albeit at low levels, might increase oligodendroglial need for MCT2, resulting in myelin damage when MCT2 is absent, but that MCT2 deletion in a different, non-inflammatory setting would have less consequences. This should be addressed by future studies using inducible oligodendroglial MCT2 mutants. However, we believe that our findings that oligodendroglial MCT2KO in the context of low-grade inflammation leads to myelin and axonal pathology, alleviated by ketogenic diet, are highly relevant for MS, particularly its progressive forms in which low-grade inflammation is present in the NAWM.

Anomalies in CNS metabolism are well recognized in MS and thought to play an important role in disease progression, which is why ensuring adequate metabolic support to stimulate myelin maintenance and repair might be crucial^2,9. However, provision of adequate energy fuels will be effective only if corresponding transporters are expressed. Our work suggests that MCT2 regulates myelin maintenance and axonal support, but is downregulated on oligodendrocytes in MS, which suggests that these cells have an impaired capacity to perform a high affinity import of monocarboxylates, at least under physiological conditions. These findings might therefore question the usefulness of potential monocarboxylate provision strategies (e.g., ketogenic diet, direct monocarboxylate supplementation etc.) as a supporting therapy in MS. However, in our mouse model, demyelination and axonal damage due to AAV-mediated MCT2 deletion in oligodendrocytes were attenuated by ketogenic diet. This might be due to direct axonal effects but also the fact that in situations of increased monocarboxylate (ketone) extracellular concentrations high affinity transporters are dispensable because other transporters of lower affinity can capture and import these molecules. This might be particularly the case for MCT1, known to be upregulated in the CNS during ketogenic diet⁶³. Our in silico analyses of Slc16a1, the gene encoding MCT1, a lower affinity MCT, showed no significant changes in the white matter oligodendrocyte cluster in MS. Moreover, we observed increased expression of this gene in reactive oligodendrocyte clusters as compared to the homeostatic white matter cluster (the opposite of what was seen with Slc16a7). This suggests that, if MCT1 protein expression corresponds to that of mRNA, increased monocarboxylate supply may be effective in stimulating myelin preservation in MS as it would enable oligodendrocytes to import these molecules via MCT1. Thus, future studies should imperatively investigate potential changes in oligodendroglial MCT1 protein expression in MS NAWM to reveal whether metabolic therapies to increase monocarboxylate provision may be of interest to stimulate myelin maintenance in MS.

In conclusion, our findings identify oligodendroglial MCT2 as a metabolic regulator of lipid synthesis and myelin maintenance, at least under conditions of low-grade inflammation. In such context, loss of MCT2 leads to impaired myelin integrity and axonal metabolic stress. Increasing ketone body availability compensates for this deficit, likely by engaging alternative transport pathways and restoring substrate supply to oligodendrocytes and axons. These results link monocarboxylate transport to white matter stability upon low grade to moderate inflammatory challenges and suggest that targeted metabolic interventions may help preserve axonal integrity in progressive neurodegenerative diseases.

Methods

Mice

Experiments with mice were approved by the Animal Ethic Committee of the University of the Basque Country (UPV/EHU) and followed the European Communities Council Directive 2010/63/EU, approval number M20_2019_244. Both males and females were used. Mice were housed in groups not larger than 5 mice per cage, in an environment with controlled temperature (24 °C), humidity 45-60%, and 12-hour light/dark cycles, with food and water ad libitum.

A previously described MCT2 floxed mouse line³⁴ (MGI:6473153, background strain C57BL6J) and control C57BL6J (Janvier) 4-6 months old mice were used for the experiments.

AAV generation

The Cre-GFP construct⁸³ was purchased from Addgene (https://www.addgene.org/68544) and cloned into the oligodendrotropic AAV Olig001³⁷ generously provided by Sara Powell and Thomas McCown (University of North Carolina) at the University of North Carolina Vector Core.

AAV injection

Spinal cord injections were performed as described previously^84,85. Prior to the surgery mice were anaesthetized by an intraperitoneal injection of ketamine (90 mg/kg) and xylazine (20 mg/kg) cocktail in NaCl. Then, two longitudinal incisions into longissimus dorsi at each side of the vertebral column were performed to remove the tissue covering the column was removed. Animals were placed in a stereotaxic frame, the 13th thoracic vertebra was fixed in between the bars designed for manipulations on mouse spinal cord (Stoelting, Wood Dale, IL), and intervertebral space was exposed by removing the connective tissue. An incision into dura mater was performed using a 30-gauge needle, and 0.5 µL of AAV solution was injected using a glass micropipette attached via a connector to a Hamilton’s syringe and mounted on stereotaxic micromanipulator. The injection site was marked with sterile charcoal. The muscle sheaths were sutured with 3/0 Monocryl, and the skin incision was closed with sterile surgical clips.

MCT2^lox/lox mice injected with Olig001-Cre-GFP AAV are referred to as AAV-MCT2KO. Wildtype (wt) mice injected with Olig001-Cre-GFP and MCT2^lox/lox mice injected with Olig001-GFP were both used as controls. After confirming that Olig001-Cre-GFP injection in wt mice did not lead to pathological changes observed in AAV-MCT2KO, MCT2^lox/lox mice injected with Olig001- GFP were used as specific AAV-controls for quantification experiments, as those mice were from the same litters as AAV-MCT2KO and housed in the same ages, for we considered these as the best possible controls.

Ketogenic diet experiments

Teklad custom ketogenic and control for KETO diets were purchased from Envigo. Ten days after AAV injection (minimum time required for AAV expression in the tissue), mice injected with Olig-001 AAV were switched to ketogenic diet (10% protein, <1% carbohydrate and 89% fat). Ketone levels in the blood were monitored using the GlucoMen Day METER 2K Blood Glucose/β-Ketone Monitoring System and corresponding ketone strips. Mice were sacrificed at 3 weeks post AAV injection (11 days after the initiation of the ketogenic diet).

Perfusion and tissue processing

Mice were euthanized with an overdose of pentobarbital and perfused with a 2% paraformaldehyde (PFA; Electron Microscopy Sciences, Hatfield, PA) solution in phosphate-buffered saline (PBS, pH 7.4) for immunohistochemistry (IHC) experiments or a 2% PFA and 0.5% glutaraldehyde solution in 0.1 M phosphate buffer (PB) for transmission electron microscopy (EM) experiments.

For IHC analyses, the brains and spinal cords were extracted and post-fixed for 20 min in 2% PFA. After rinsing 3 times with NaCl, the tissue was equilibrated in 15% sucrose, embedded in a 7% gelatine (type A porcine skin, Sigma)/15% sucrose solution in PBS, snap-frozen using isopentane (Sigma-Aldrich) at −55 °C, and subsequently stored at −80 °C prior to cryostat sectioning. Spinal cord cryosections, of 12 µm in thickness, were cut coronally on Super Frost Plus slides (Thermofisher) using the Leica CM1950 Cryostat. Slides were then air-dried for a minimum of 30 min and stored at −20 °C.

For EM analyses, spinal cords were post-fixed overnight in 2% PFA and 0.5% glutaraldehyde, washed 3 times in 0.1 M PB and stored in 0.1 M PB + 0.05% Sodium azide prior to resin processing and embedding.

Human material

Post-mortem human MS and control brain samples were obtained from the UK MS Tissue Bank (Dr R. Reynolds, London). Informed consent had been obtained from each patient, and this study was approved by the London Multicenter Research Ethics Committee (MREC 02/2/39).

Snap-frozen sections from postmortem cerebellum samples were analyzed to study the expression pattern of MCT2. Four controls and 4 MS patients were analyzed. The clinical features of these human subjects and lesions analyzed are detailed in Table S1.

Immunohistochemistry (IHC)

Slides with spinal cord sections were air-dried at room temperature for 45min-1h and then rehydrated in TBS 1x for 30 min. Antigen retrieval was performed by heating the sections in a citrate-based retrieval buffer (Vector Laboratories/Dako) for 45 s in a microwave. After serial washes, blocking buffer (0.1% Triton X-100, 10% goat serum in 1x TBS) was applied for 30 min, followed by incubation with primary antibodies (Table S2) diluted in the blocking buffer, overnight at 4 °C. The next day, samples were washed and incubated with the secondary antibodies (Table S3) for an hour. Finally, slides were counterstained with DAPI, washed and mounted with Fluoromount-G.

For human tissue, histological assessment of the lesions was performed using Luxol Fast blue/Cresyl violet and Oil-red-O staining. Lesions were classified according to their inflammatory activity (KP1 and MHC-II immunolabelling) and on the basis of previously described histological criteria^86,87.

Cryostat sections (12 µm thickness) of snap-frozen control and MS cerebellum were rehydrated in PBS. Thereafter, slides were post-fixed for 20 min with PFA 2% and washed, to posteriorly microwave them in low-pH unmasking solution (Vector Laboratories). Sections were then preincubated in blocking buffer (10% normal goat serum, 0.1% Triton-X 100 in PBS) for 1 h and incubated overnight with primary antibodies (Table S2) at 4 °C. After overnight incubation, slides were extensively washed in PBS 0.1% Triton X-100 and incubated with appropriate secondary antibodies (Table S3) for an hour. Nuclei were counterstained with DAPI and slides mounted with Fluoromount-G.

Mouse tissue imaging and quantification

Imaging of mouse tissue sections was performed using a Leica TCS STED CW SP8X confocal microscope and the Zeiss LSM 880 confocal microscope equipped with a 40x oil-immersion objective.

The area containing Olig-001 AAV-transduced cells was localized based on the expression of GFP reporter. Images were acquired at 40x and/or 63x magnification to cover the entire dorsal funiculus. At least two spinal cord levels separated by 144 µm minimum were imaged per mouse. Analyses were carried out using the ImageJ/Fiji software (version 1.54, NIH). Cells positive for GFP, for the marker of interest, and cells co-expressing GFP and the marker of interest were counted. The results are presented either as cells/mm² or as the percentage of cells within the GFP+ population.

In the case of GFAP, MOG, MBP, and CD45 staining, a positivity threshold was set in the corresponding channel, and quantifications were performed as the percentage of area labelled within the ROI of interest. For colocalization studies, cells labeled on both channels were counted and the results were expressed as numbers of co-labelled cells per area or the percentage of the total.

In the case of neurofilaments, positivity thresholds were applied to sections labelled with anti-SMI31 (phosphorylated NF) or -SMI32 (non-phosphorylated NF) antibodies. The area labelled for these antigens was then quantified using the “area fraction” setting in Fiji. Numbers of labelled axons were quantified using “Analyze Particles” function in Fiji with the circularity set as 0.10-1.00, and lower particle area threshold of 1.5µm².

The analyses of axonal LDHA expression were performed on cross-sectioned axons using specific Fiji macros⁸⁸, except where specified. First, LDHA and SMI32 channels were extracted from multichannel z-stacks by selecting the plane with the highest SMI32 mean gray value (MGV) plus/minus 2 z-steps. ROIs were manually drawn in order to exclude areas where axons are longitudinally organized (rare, as the absolute majority of the axons in dorsal funiculus are coronally cut in transverse spinal cord sections), thus selecting the regions with circular orthogonally cut axons. In order to avoid manual bias while applying axonal masks, we applied the “Moments method threshold” colocalization quantification via BIOP-Jacop plugin⁸⁹, and calculated Manders coefficient, that was considered as quantification of “axonal LDHA expression” in Fig. 9F.

To illustrate myelin-axon status in the injected white matter, a 3D reconstruction of the colocalization between SMI31-MBP and GFP was generated using the Software Leica Application Suite X (LAS X).

Electron microscopy

For pre-embedding immunostaining for GFP and Olig2 spinal cords were cut into 50 µm thick slices using a Leica VT1000S vibratome (Leica Biosystems, Wetzlar, Germany). For aldehyde inactivation, sections were incubated in 1% sodium borohydride in PB and washed. Then, they were permeabilized by cryoprotection in 25% sucrose in PB followed by freeze-thaw cycles in methylbutanol. For the immunogold labeling, sections were incubated in blocking solution I (0.3% BSAc (Aurion) in PB) for an hour and then, anti-GFP (1:200; Aves Labs, Table S2) primary antibody diluted in blocking solution was added from 36 up to 60 h at 4 °C and in agitation. Samples were washed after and incubated in blocking solution II (0.5% BSAc and 0.1% fish gelatin (Aurion) in PB) for an hour before incubation with the secondary antibody conjugated to colloidal gold (Aurion, 1:50, Table S3) for 24 h. Afterwards, for the silver enhancement, sections were washed and then immersed in equal volumes of reagents A and B (Aurion) in a dark chamber, to amplify the signal. Lastly, gold toning was performed by gold incubation, 0.05% gold chloride for 10 min, and posterior incubation in 0.3% sodium thiosulfate for two times.

For the embedding process, sections were washed with 0.1 M PB and post-fixed with osmium tetroxide. Samples underwent a sequence of increasing ethanol concentrations washes for dehydration before uranyl acetate incubation. Propylene oxide was used after as an intermediate between alcohol washes and the final embedding in araldite. Semithin (1.5 um) sections were cut for identification of the area of interest, and ultrathin (70–80 nm) sections were obtained and stained with lead citrate for analysis.

The samples were examined and images were acquired at 80 kV on a FEI Tecnai G2 Spirit microscope (FEI Company, Hillsboro, OR) equipped with a Xarosa digital camera (20 megapixel resolution) using Radius image acquisition software (EMSIS GmbH, Münster, Germany).

TEM images were analyzed using Image J, and myelinated and demyelinated axons were counted.

Quantification of MCT2 and SOX10 in human samples

All tissues obtained from multiple sclerosis patients consisted of parts of the cerebellum, where different types of lesions were found. For the quantification of MCT2 and SOX10, images were acquired in the active zones of 3 active lesions, 2 chronically active lesions, and 1 shadow plaque lesion.

Images were acquired in the core region of each lesion, in the perilesion, and in the normal-appearing white matter. These areas were identified based on Luxol fast Blue, Cresyl violet, and MHCII staining. Images were acquired in these areas using a camera (AxioCam; Zeiss, Jena, Germany) attached to a Zeiss Cell Observer with Apotome module (Zeiss).

SOX10+ and MCT2+ cells, and co-labeled cells within the lesion core, were quantified as the number of positive cells per unit area or as percentage of the total SOX10+ cells.

In vitro studies

Rat primary OPC cultures were prepared according to the Paris Brain Institute ethical guidelines (agreement n° P390R). Both male and female pups were used. Primary OPC cultures were obtained from cerebral cortices of Wistar rats (Janvier Labs) at P0-P1, and after 11-12 days in vitro (DIV), OPC retrieval was performed by differential shaking as previously described⁹⁰, except that the differential adhesion process was repeated three times to allow the remaining microglia and astrocytes to adhere, while the OPC were kept in suspension. Isolated OPC plated and kept in “proliferation medium” containing DMEM and Nutrient Mixture F-12 Ham (1:1)(Dutscher) supplemented with 2% B27 (GIBCO), 10 µg/mL of biotin, 1% of N1 mix, 10 ng/mL of FGF and 10 ng/mL of PDGF (all from Sigma-Aldrich), or in “differentiation medium” (same medium, but without FGF and PDGF). The N1 mix is composed of DMEM, 1 µg/µL of selenite, 80.5 mg of putrescine and 25 mg of transferrin (all from Sigma Aldrich). OPC were then transduced at a multiplicity of infection (MOI) of 4 with either lentiviral shRNA scramble control particles (OriGene #TR30021V, referred to as control, Scrambled sequence 5’ GCACTACCAGAGCTAACTCAGATAGTACT 3’) or Slc16a7 rat shRNA lentiviral particles (OriGene #TL710093V, referred to as Slc16a7-KD; The kit contained 4 shRNAs, they were combined at a MOI of 1 each. TL710093VA: 29735 ACGAAGAGACTCAGTAAGGTATCAACAGC Rt 741NM_017302, TL710093VB: 29735 ATTCTTGGCTCCGTATGCTAAGGACAAAG Rt 878 NM_017302, TL710093VC: 29735 AAGCCGTCACGGTATTCTTCAATGACATC Rt199 NM_017302 and TL710093VD: 29735 AATCAGCCTCGCAGGCATCCAAAGAAATG Rt 1414 NM_017302) under the control of CMV promoter and kept in proliferation medium for 12 h at +37 °C, 5% CO2. Proliferation medium was replaced with differentiation medium. Both lentiviral vectors also encoded for green fluorescent protein (GFP), which allowed to visualize transduced cells. For immunocytochemistry, at 6 DIV cells were fixed with 2% PFA for 10 min at room temperature and then washed twice with 0.1 M PBS. Cells were incubated with blocking buffer (0.M PBS 0.1% Triton-X100 and 10% normal goat serum (NGS; Eurobio) for 10 min at room temperature. Cells were then incubated with primary antibodies overnight (Table S2) in 0.025% Triton X-100, 2% NGS, 0.1 M PBS at 4 °C, washed, and incubated with secondary antibodies (Table S3) in 0.025% Triton X-100, 2% NGS, 0.1 M PBS for an hour at room temperature. Nuclei were counterstained with the Hoechst dye (SIGMA, dilution 1/1000). Coverslips were mounted using Fluoromount G (SouthernBiotech).

RNA extraction and RT-qPCR

Cells in 12-well plates were washed once with 0.1 M PBS and total RNA was extracted using the RNeasy Mini Kit, following the manufacturer’s instructions (QIAGEN #74104). cDNA was then synthesized using the High-Capacity cDNA reverse transcription kit with RNAse inhibitor (Applied Biosystem). Finally, qPCR reaction was performed using TaqMan Fast Advanced Master Mix (Applied Biosystem) and specific probes. Ppia (TaqMan Gene Expression Assays Rn00690933_m1; Applied Biosystems) housekeeping control gene was used. Threshold cycle (Ct) numbers were calculated for specific genes: slc16a7 (TaqMan Gene Expression Assays Rn00571479_m1; Applied Biosystems), mbp (TaqMan Gene Expression Assays Mm01266402_m1; Applied Biosystems), plp1 (TaqMan Gene Expression Assays Mm01297210_m1; Applied biosystems) and myrf (TaqMan Gene Expression Assays Rn01454573_m1; Applied biosystems) genes. The exact TaqMan assays sequences are proprietary of Thermofisher and are not provided by the supplier. All probes are linked with a 6-carboxyfluorescein fluorochrome (FAM). The ΔΔCt method was used to calculate mRNA abundancy and mRNA fold changes.

Cell imaging and analyses

Cells were imaged using an Axio Scan.Z1 slide scanner with a 20X objective. Transduction rate was defined by the ratio of the number of Olig2+GFP+ cells over the total number of Olig2+ cells, while percentage of differentiated cells was defined as the number of Olig2+MBP + GFP+ cells over the total number of Olig2+GFP+ cells. Olig2+ cell density was defined by the ratio of the number of Olig2+ cells over the total number of DAPI+ cells. At least three individual experiments and 300 cells were analyzed in each experimental condition and time point.

MBP+ membrane surface area (in µm²) of individual OL was measured by delineating the MBP+ edge of Olig2+GFP+ pseudo-myelin membranes. Five individual experiments and 19-33 cells per experimental condition and time point were analyzed. Quantifications were performed on Zen 10.0 software (Zeiss).

Statistical analyses

GraphPad Prism 8 software was used to perform statistical analyses (GraphPad Software, CA, USA). Data were presented as the mean ± standard error of the mean (SEM). All data sets were tested for normality and homoscedasticity. For the in vivo (mouse manipulations) and post-mortem studies (control vs MS tissue): the comparison of two groups, if data passed normality test, was performed using parametric two-tailed unpaired t-test; if data did not pass the normality test, two-tailed Mann–Whitney’s test was applied; comparisons between more than two groups were performed according to the normality test results either using One-way Analysis of Variance (ANOVA), or in the case of negative normality test, using Kruskal Wallis test. For the analysis of slc16a7 knockdown in vitro: in the case of qPCR analysis, data were presented as fold change with respect to Ppia gene, log2 normalized, and two-tailed paired t test was performed for data that passed the normality test. If data did not pass the normality test, comparisons were performed using the two-tailed Wilcoxon matched-pairs signed rank test. For immunocytochemistry analyses, data on the number of Olig2 + , MBP + , and GFP+ cells were analyzed using two-tailed unpaired student t test, or Mann–Whitney test, depending on the results of the normality test. For the quantification of MBP surface area, data were analyzed and, according to the negative normality test, two-tailed Wilcoxon Signed Rank test was used, with theoretical/predicted median value of fold change set to 1. Data were analyzed as technical replicates belonging to 5 independent experiments. Statistical significance was represented as p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***).

In silico analyses of Slc16a7 and Slc16a1 expression

Single nucleus transcriptomic count matrix and metadata information were retrieved from https://ms-cross-regional.cells.ucsc.edu (accession date: 20 November 2023; downloaded object: integrated seurat.rds), corresponding to the publication of Trobisch et al. 2022²⁰. Bioinformatic analyses were conducted with the R programming language (version 4.1.2) (https://www.R-project.org). First, processing was performed following the standard Seurat R pipeline (package version 5.0.1.)⁹¹. We selected nuclei annotated as oligodendrocytes that met the quality control criteria established by the original authors. We retained nuclei that expressed at least 250 different genes, had a minimum of 400 counts, and contained less than 5% mitochondrial genes. Normalization was performed with the NormalizeData function, and gene expression levels were visualized using the VlnPlot function. The differential expression analysis for Slc16a7 and Slc16a1 genes across groups was performed with DESeq2⁹². The pseudobulk raw counts were generated by Seurat’s AggregateExpression function, on which we applied the negative binomial generalized linear model, adjusting for age when indicated in the results section. For multiple comparison testing, p-values were adjusted by the Benjamini and Hochberg procedure (Benjamini and Hochberg, 1995), and significantly differential expression defined as false discovery rate <0.05. The code used for the in silico analyses is available at: 10.5281/zenodo.19466297.

Third party templates used in Figures

Mouse drawing used in Figs. 4 and S6 is from Pixabay https://pixabay.com/vectors/search/lab%20mouse/.

Templates used in Fig. 10: connexins-adapted from NIAID Visual & Medical Arts. 10/7/2024. Perforin Pore. NIAID BioArt Source. bioart.niaid.nih.gov/bioart/402. For MCT channels: Image adapted from Servier Medical Art (https://smart.servier.com/), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization: V.T.; Methodology: F.G.G., J.M.G.V., B.N.O., V.T.; Investigation: L.I.U., L.M.H., I.S.S., R.M.G., M.A.M., M.J.U.N., J.C.C., S.C., C.D., L.M.C., M.A.B., A.C.G., M.D., J.L.Z., F.G.G., J.M.G.V., B.N.O., V.T.; Formal analysis: L.I.U., L.M.H., I.S.S., S.C., V.T.; Writing—Original Draft: L.I.U., L.M.H., V.T.; Writing—Review & Editing: L.I.U., L.M.H., I.S.S., R.M.G., M.A.M., L.M.C., C.N., J.L.Z., L.P., F.G.G., J.M.G.V., C.M., B.N.O., V.T.; Visualization: L.I.U., L.M.H., I.S.S., V.T.; Resources: C.N., M.D., J.L.Z., L.P., F.G.G., J.M.G.V., C.M., B.N.O., V.T.; Supervision: F.G.G., J.M.G.V., B.N.O. and V.T.; Project administration: V.T.; Funding Acquisition: B.N.O. and V.T.

Peer review

Peer review information

Nature Communications thanks Julia Edgar and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by: Ministerio de Ciencia, Innovación y Universidades-España (SAF2015-74332-JIN and PID2023-152688OB-I00 to V.T., and PID2022-143020OB-I00 to C.M.); ARSEP Foundation and France Sclèrose en Plaques Foundation to V.T. and B.N.O. (ARSEP1317, France SEP-10, and France SEP-51), Consellería de Educación, Universidades y Empleo-Generalitat Valenciana (CIDEXG/2023/23) to V.T., BIOEF-EITB-Maratoia (BIO23/EM/008) to V.T., NeurATRIS (partial funding by ‘NeurATRIS ANR-11-INBS-0011’, of the French Investissements d’Avenir Program run by the Agence Nationale pour la Recherche) to V.T. and BNO, CIBERNED (CB06/05/0076 to C.M. and CB06/05/1131 to J.M.G.V.), Walk On Project grant to V.T., and Gobierno Vasco (IT1203-19 and IT-1551-22 to C.M.). L.I.U. was a recipient of the Basque Government PhD studentship. I.S.S. was supported by a predoctoral grant FPU20/03544 funded by the Spanish Ministry of Universities. F.G.G. is supported by PID2021-124430OA-I00 funded by MCIN/AEI/10.13039/501100011033 and by “ERDF A way of making Europe”, and by CIAICO/2023/149 funded by the Consellería de Educación, Cultura, Universidades y Empleo de la Generalitat Valenciana.

Data availability

The data that support the findings of this study are available from the corresponding author upon request. Source data are provided with this paper.

Code availability

The code used for the in silico analyses (Fig. 2 and S3) is available at 10.5281/zenodo.19466297.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74488-w.