Transcriptional dynamics of the oligodendrocyte lineage and its regulation by the brain erythropoietin system

Abstract

Oligodendrocytes differentiate from oligodendrocyte progenitor cells (OPC) in early postnatal development, but some oligodendrogenesis is maintained throughout adulthood, where oligodendrocyte lineage dynamics may contribute to neuroplasticity, adaptive myelination, and myelin repair. Here, we studied the effect of erythropoietin (EPO) and its receptor (EPOR) on oligodendrocyte lineage dynamics employing murine hippocampus and its myelinated fibers as model region. Using multiple stage-specific markers and single-nuclei-RNA-seq data, we find that EPO stimulates all oligodendroglial lineage cells directly, driving differentiation/maturation. Differential gene expression analysis reveals multiple EPO-regulated mRNAs, including downregulated transcripts for GABA-A receptors, fitting the known inhibition of oligodendrocyte maturation by GABA. Importantly, analogous oligodendrocyte responses are seen when endogenous EPO expression in brain is stimulated by hypoxia. Mice lacking EPOR from mature oligodendrocytes show subtle deficiencies of adult myelination in hippocampal fimbria and mild working memory deficits. These gain- and loss-of-function experiments may further suggest EPO as clinically safe treatment for remyelination therapies.

Fig. 1. Effects of rhEPO/placebo on OPC numbers, proliferation, and apoptosis.

a Schematic representation depicting OL differentiation and maturation, along with the respective stage markers. OPC, oligodendrocyte progenitor cells; PreOL, premyelinating oligodendrocytes; COP, committed oligodendrocyte progenitors; Myelinating OL, myelinating oligodendrocytes; NFOL, newly formed oligodendrocytes; MOL, mature myelinating oligodendrocytes. b Experimental design: Starting on P28, wildtype C57Bl/6 mice received rhEPO/placebo intraperitoneally, either once (24 h experiment) or every other day during 1 week or 3 weeks. Mice were sacrificed 24 h after the last injection, in the week 4 experiment after 1 week of break. c Representative images and d quantification of OPC (NG2⁺OLIG2⁺ cells) in CA1 and CA3 following 24 h and 1 week of rhEPO/placebo, respectively. n = 7 mice (CA1 placebo), 6 mice (CA1 rhEPO), 7 mice (CA3 placebo), 6 mice (CA3 rhEPO) for the 24 h timepoint; n = 6 mice for all groups at w1. e Sketch summarizing the course of OPC numbers at 24 h, 72 h, 1 week, and 4 weeks after initiation of rhEPO/placebo treatment. Information on the 72 h and 4 week time point derived from Hassouna et al. 2016. f Representative images and g percentage of proliferative OPC in CA1 at 24 h. n = 6 mice (placebo), 5 mice (rhEPO). h Representative images and i % apoptotic OPC in CA1 at 24 h. n = 6 mice (placebo), 5 mice (rhEPO). Scale bars in c, f, h: 50 µm; mean ± SEM presented; mouse n numbers given in bars; unpaired two-tailed Student’s t-test (d, g, i); a, b, e Created in BioRender. ye, l. (2025) https://BioRender.com/s3nhxid. Source data are provided as a Source Data file.

Fig. 2. EPO induces OPC proliferation and migration in vitro.

a Experimental design delineating the set-up of OPC cultures and subsequent in vitro procedures. b Representative culture images and c quantification of proliferating OPCs (BrdU⁺PDGFRα⁺ / PDGFRα⁺) after 24 h rhEPO treatment. Data were collected from three independent OPC cultures (biological replicates), each from pooled brains of P7 mice. For each condition, 8–9 fixed-area ROIs ( ≈ 2.5 mm²) were analyzed using FIJI. Each dot represents one ROI. d Primary OL were prepared via MACS sorting from C57/BL6N WT mice (P7) and imaged after 1 day of culture in vitro (DIV1) for 8 h. Cells were treated with 0.3, 1 IU/ml rhEPO or an equivalent volume of the vehicle (EPREX). Representative images of the treated cells are shown. e Quantification of the migration distance of OPC given as “track displacement” for average values of all evaluated cells per ROI and underneath of the fastest 20% OPC per ROI. Data were collected from three independent OPC cultures (biological replicates). For each condition, 4 fixed-area ROIs (40x objective) were imaged and analyzed using Imaris. Scale bars in b, d: 50 µm; mean ± SEM presented; one-way ANOVA with Dunnett’s test (c), unpaired two-tailed Student’s t-test (e); a: Created in BioRender. ye, l. (2025) https://BioRender.com/s87m1ln. Source data are provided as a Source Data file.

Fig. 3. snRNA-seq analyses of OL upon rhEPO/placebo treatment.

a Two-dimensional UMAP showing the 5 clusters obtained using unbiased transcriptome analysis of 12043 single nuclei, derived from hippocampi of mice treated with rhEPO/placebo. Each dot represents a single nucleus. Clusters were manually annotated by cross-referencing established markers (subfigure b). MOL1/2 were separated based on their calculated pseudotime (MOL2 pseudotime>MOL1 pseudotime, Wilcoxon rank sum test, p < 0.001). b DotPlot illustrating the expression pattern of established OL markers. Average expression denotes the RNA expression of each marker normalized within respective clusters. These markers are used to confirm the identity of the clusters. MOL1 and MOL2 exhibit a higher level of Apc expression separating them from NFOL lineage. c Heatmap representing the differentially expressed genes (DEG, N = 637) upon rhEPO/placebo for each cluster. The table beneath the plot shows the numbers of DEG for each cluster. DEG are represented by their average log2 fold change (minimum average log2FC ± 0.25; adjusted p value ≤ 0.05, ‘bimod’ test, multiple correction done using Bonferroni). Red represents a higher expression value on rhEPO treated samples when compared to placebo samples, while blue represents a lower expression value for the same comparison. The full list of DEG is provided in Supplementary Table 1. d Multiple violin plots showing the expression level and distribution of DEG from the OPC cluster, which belong to the Gene Ontology (GO) term gamma-aminobutyric acid signaling pathway (GO:0007214). Note that all DEG from this GO term were downregulated under rhEPO. P-values calculated using the ‘bimod’ test and adjusted using Bonferroni. Gray represents the expression values from placebo samples and red the ones from samples treated with rhEPO. e Multiple violin plots showing the expression level and distribution of DEG from the OPC cluster, which belong to the GO term glutamate receptor signaling pathway (GO:0007215). Again, all DEG from this GO term were downregulated under rhEPO. P-values calculated using the ‘bimod’ test and adjusted using Bonferroni. Gray represents the expression values from Placebo samples and red the ones from samples treated with rhEPO. f Selected GO terms enriched significantly (over-representation analysis, adjusted p value < 0.05, Benjamini-Hochberg procedure, calculated using clusterProfile) by the analysis of those genes that constructed the rhEPO exclusive regulons (see SCENIC analysis in Methods section). The full list of regulons and GO terms is provided in Supplementary Tables 3 and 4. Source data are provided as a Source Data file.

Fig. 4. Effects of rhEPO on OL differentiation and maturation.

a Overview of the hippocampus showing OPC/COP/NFOL marker BCAS1 (red), OPC marker NG2 (yellow), and OL marker OLIG2 (green) staining. Left: White frames denote areas a1-a4, magnified on the right. a1 illustrates early OPC (BCAS1^-NG2⁺OLIG2⁺), characterized by small soma with few ramifications; a2 illustrates late OPC (BCAS1⁺NG2⁺OLIG2⁺) characterized by large soma with more ramifications; a3 illustrates COP (BCAS1⁺NG2^-OLIG2⁺) characterized by large soma with numerous ramifications; a4 show NFOL (BCAS1⁺NG2^-OLIG2⁺) characterized by large soma with parallel arranged ramifications; white solid arrows indicate early OPC, late OPC, COP and NFOL, respectively. b–d Quantification of late OPC/COP/NFOL (BCAS1⁺OLIG2⁺), late OPC (BCAS1⁺NG2⁺OLIG2⁺), and COP/NFOL (BCAS1⁺NG2^-OLIG2⁺) in CA1 and CA3 after 24 h of rhEPO/placebo. e Representative images displaying COP marker BMP4, OL marker OLIG2, and DAPI staining in CA1 after 1 week of rhEPO/placebo. White arrow heads indicate BMP4⁺NG2^- OL. f Quantification of COP (BMP4⁺NG2^-OLIG2⁺) in CA1 and CA3 after 1 week of rhEPO/placebo. g Representative images displaying late OPC/COP/NFOL marker BCAS1, MOL marker CC1, OL marker OLIG2, and DAPI staining. White solid arrow indicates NFOL (BCAS1⁺CC1⁺OLIG2⁺); white hollow arrows indicate MOL1/2 (BCAS1^-CC1⁺OLIG2⁺). h–j Quantification of late OPC/COP/NFOL, late OPC/COP, and NFOL in CA1 and CA3 after 3 weeks of rhEPO/placebo. k–m Quantification in CA1 and CA3 of late OPC/COP/NFOL, late OPC/COP (BCAS1⁺CC1^-OLIG2⁺), and NFOL (BCAS1⁺CC1⁺OLIG2⁺) at 4 weeks after initiation of rhEPO/placebo. n Quantification of MOL (CC1⁺ OLIG2⁺) at 1 week, 3 weeks and 4 weeks after initiation of rhEPO/placebo. o Sketch summarizing the course of MOL number changes at 1 week, 3 weeks and 4 weeks after initiation of rhEPO/placebo. Scale bars in a,e,g: 50 µm; mean ± SEM presented, n (number of mice) given inside the bars; unpaired two-tailed Student’s t-test, except for b (CA1), m (CA1), n (MOL W3, CA3), where two-tailed Mann-Whitney U test was performed. o Created in BioRender. ye, l. (2025) https://BioRender.com/s87m1ln. Source data are provided as a Source Data file.

Fig. 5. OL dynamics in dentate gyrus, induction of endogenous EPO, and consequence of EPOR deletion in MOL.

a–d Differential OL dynamics in the dentate gyrus upon rhEPO/placebo. a Area of analysis. b Quantification of late OPC/COP/NFOL at 24 h, 3 and 4 weeks following rhEPO/placebo initiation. c Quantification of NFOL at 3 and 4 weeks after start of rhEPO/placebo. d Quantification of MOL following 1 week, 3 weeks and 4 weeks after initiation of rhEPO/placebo. e, f Induction of endogenous EPO by motor cognitive challenge and inspiratory hypoxia. e Experimental design: Starting on P28, mice are exposed to normoxia (21% O₂; N), functional hypoxia induced by complex running wheel (CRW) performance, inspiratory hypoxia (12% O₂; H), or CRW upon inspiratory hypoxia (H + CRW), for 3 weeks. f Quantification of late OPC/COP/NFOL in CA1 and CA3. g–k Effect of rhEPO in mice with EPOR deletion in MOL. g Experimental design: CnpCre::EPOR mice received rhEPO once (24 h) or for 3 weeks and brains were examined at week 4. h Quantification of OPC and i late OPC/COP/NFOL in CA1 and CA3 at 24 h. j Quantification of late OPC/COP/NFOL and k MOL in CA1 and CA3 at 4 weeks; mean ± SEM; n (number of mice) given inside the bars; unpaired two-tailed Student’s t-test except for i (CA3), where two-tailed Mann-Whitney U test was performed; in f, Tukey’s one-way ANOVA. a, e, g Created in BioRender. ye, l. (2025) https://BioRender.com/s87m1ln. Source data are provided as a Source Data file.

Fig. 6. Deletion of EPOR in MOL has network-relevant influence on myelination.

a Experimental design using CnpCre::EPOR mice and controls. b Representative images and c quantification of myelin (MBP⁺ area) in CA1 and CA3. d Representative images with microglia marker IBA1 and phagocytic (activation) marker CD68 and e,f quantification of microglia and activated microglia. g, h Quantification of OPC and MOL in CA1 and CA3 of CnpCre^+/-::EPOR^+/+ and CnpCre^+/-::EPOR^fl/fl mice. i Representative EM images from fimbria of CnpCre::EPOR mice, showing no appreciable difference between genotypes. j Highly significant right-shift of myelinated axon calibers in fimbria, but no difference in myelin thickness (g-ratio) and myelinated axons/µm. k Representative EM images from corpus callosum of CnpCre::EPOR mice, revealing no systematic alteration of myelinated axon calibers in fimbria, and no difference in myelin thickness (g-ratio) and myelinated axons/µm. m Representative EM image from CA1 of CnpCre::Epor mice with n unchanged myelinated axons/µm. o Comparison of Jacobian determinants in MRI of CnpCre^+/-::EPOR^fl/fl versus CnpCre^+/-::EPOR^+/+ mice: red/yellow color indicates local tissue volume expansions, blue/cyan local reductions. p Volumetric MRI comparison of fimbria and striatum; scale bars: b, d 50 µm, I, k, m 1 µm; mean ± SEM presented, with n (number of mice) given inside the bars; two-tailed Mann-Whitney U test was performed for n, asymptotic two-sample Fisher-Pitman permutation tests for j and l, unpaired two-tailed Student’s t-tests for the others. a Created in BioRender. ye, l. (2025) https://BioRender.com/s87m1ln. Source data are provided as a Source Data file.

Fig. 7. Mild cognitive impairment in mice with EPOR deletion in MOL.

a Experimental outline of the behavioral test battery of CnpCre::EPOR mice. Behavioral characterization was started at the age of 9 weeks and completed at the age of 50 weeks. b Higher cognition was assessed in IntelliCage and cognition index per day in all challenges plotted. Below the graphs, a schematic outline of the IntelliCage paradigms is depicted; HAB: Habituation; PL: Place Learning; RL: Reversal Learning; MRL: Multiple Reversal Learning; EXT: Extinction; ELM: Episodic-like memory; CW: Clockwise; CCW: Counterclockwise; N numbers in bars; mean ± SEM; two-tailed two-sample t test for normal and two-tailed Mann-Whitney U test for non-normal distributions. Only statistically significant p-values were shown, the non-significant values were 0.0993 (PL), 0.4795 (MRL), 0.2786 (ELM acquisition), 0.6007 (ELM retrieval), 0.2053 (ELM rev acquisition), 0.7789 (ELM rev retrieval). A more detailed overview of the results can be found in supplementary table 5. Source data are provided as a Source Data file.