During normal developmental processes, oligodendrocytes (OLs), the myelinating cells of the central nervous system (CNS), differentiate from their precursor cells (OPCs). During development, there are three waves of OPC differentiation, allowing for rapid population of the developing CNS, although not all OPCs will survive until adulthood of the organism1. Since their discovery2, many terms have been utilized to describe OPCs including “NG2” cells, “polydendrocytes” and “O2A” cells1. OPCs remain resident in the CNS once the brain completes development1,3,4, acting as a reservoir to replace lost OLs in case of injury4 and to serve a number of homeostatic functions5,6. Thus, OPCs serve as an important therapeutic target in demyelinating diseases and regenerative medicine.

Early in the glial lineage, tri-potent glial-restricted progenitors (GRPs) give rise to either type-1 astrocytes directly, astrocyte-restricted progenitors (ARP), or bipotent O2A cells. These O2A cells can further differentiate into type-2 astrocytes or OL-fated OPCs2,7,8,9. The progression from GRPs to specific differentiated cell types requires tightly regulated developmental steps to ensure a balanced production of astroglial and oligodendroglial populations in the CNS. Early studies on GRP and O2A-cells have suggested that certain in vitro conditions can bias their differentiation towards either astrocytic or oligodendroglial lineages2,10.

The growing interest in OL lineage cells and their roles in disease pathology underscores the need for reliable and reproducible research models. While murine models have yielded valuable insights into OL biology, significant species differences in OL development, myelination and response to injury present challenges in translating findings to humans11,12. These limitations are further emphasized by the absence of myelin pathology in many rodent models carrying disease-causing mutations in genes associated with white matter disorders13. Studying development in a human in vitro model allows us to improve our understanding of unique occurrences that may not be estimated in other organisms.

Access to primary human OL-lineage cells is typically limited. Thus, human induced pluripotent stem cell (iPSC) models are of increasing importance towards advancing the basic understanding of OPC biology and myelination, and for improving the translatability of future therapeutics targeting these cells and their remyelination potential. However, current efforts are limited by the difficulties in obtaining pure OL cultures from human iPSCs. Astrocytic cells, which spontaneously arise from shared OL/astrocyte progenitors, frequently dominate these cultures, impacting the final yield of OLs. To address this, our group recently published a method to generate OL lineage cells from iPSCs more efficiently14. Despite the increased yield of mature OLs, early progenitor heterogeneity remained an issue.

In the present study, we analyzed the pool of progenitors produced at discrete steps of our OL generation protocol. We identified 3 populations (GRPs, ARPs and O2A) and isolated them using a previously tested in-house fluorescence- activated cell sorting (FACS) pipeline and assessed the differentiation potential of each population to confirm the contribution of each subpopulation to the final proportion of cell types. With single-cell RNA sequencing (scRNAseq), we identified three distinct developmental lineages and the distinct molecular mechanisms contributing to each. This study aims to shed light on human developmental dynamics in an in vitro model. This information will serve to set a benchmark for the study of complex glial cell dynamics in healthy and disease contexts thereby allowing for the fine-tuning of further works exploring the development and remyelination capacity of these cells.

With this in mind, we generated OL-lineage cells using our previously published protocol14 (Fig. 1A). To study the contribution of various glial precursor populations, we sequenced our cultures at different timepoints throughout the protocol (Day 75, Day 85 and Day 95) (Fig. 1A, C). These timepoints comprise the majority of glial progenitors (Day 75), committed OPCs (Day 85) and pre-OLs/OLs (Day 95) (Fig. 1A). Following unsupervised clustering, we identified populations of GRPs, O2A and ARPs (Fig. 1B, C) as well as various sub-populations of OPCs and astrocytes using canonical markers (Fig. 1B, C). We further generated cluster-specific markers for identification, provided in supplemental files (Supplemental Fig. 1, Supplemental Data 1). Next, we asked which glial clusters responded to our differentiation media, namely at our selected timepoints (Day 75, Day 85 and Day 95) where media composition is changed (see materials and methods). Interestingly, at each of the selected timepoints, we observed stable proportions of both proliferative and resting OPC clusters, as well as the O2A and GRP populations (Fig. 1D). Clusters with the largest changes in response to the media included “Astro_3”, “Astro_1”, “ARP”, “pre OL” and “late OPC” (Fig. 1D). Alongside generation of pre OLs, the media also influences astrocytic differentiation with sequential changes of “Astro_2”, “Astro_1” and “Astro_3” populations. These observations further solidify our group’s previous findings14, where OL differentiation was observed to coincide with the generation of astrocytes and retention of glial progenitors.

Fig. 1: Generation of OL-lineage cells by growth factors results in the retention of glial precursor populations.
figure 1

A Schematic of the OL-lineage generation protocol from human iPSCs. Created in BioRender. Durcan, T. (2025) https://BioRender.com/uydn3pb. B Dimplot following unsupervised clustering of two iPSC lines at Day 75, Day 85 and Day 95 of the protocol following scRNAseq, all shown. Number of cells that passed QC and were included in the analysis are as follows: 971, 1218, 1153 for line 3450 at Day 75, Day 85 and Day 95 respectively. For line 81280, cell numbers included in the analysis are 1146, 2974, and 2412 at Day 75, Day 85 and Day 95 respectively. C Dot plot representation the average expression of key identifying genes used to label clusters. D Bar plot representing cluster proportions at selected stages of the differentiation protocol.

Full size image

We further assessed the lineage potential of identified glial progenitor populations by employing FACS, isolating them using a combination of antibodies previously described by our lab14. We utilized combinations of the following: A2B5 + PDGFRα+ for OPCs, A2B5 + PDGFRα- for GRP and CD44 + PDGFRα- for ARP, confirmed in our scRNAseq dataset (Fig. 1C). We next cultured the populations in OL maturation media for 21 days followed by immunofluorescence staining to confirm the identity of the glial subpopulations. Specifically, most GRPs expressed astrocytic markers S100β and GFAP (Fig. 2A, B) with a small subpopulation displaying low levels of OPC/OL markers PDGFRα and O4 (Fig. 2C, D). ARPs gave rise almost exclusively to S100β and GFAP-positive cells and displayed high levels of the nuclear factor 1 A (NF1A, Fig. 2C), associated with astrocytic lineage fate15. In turn, O2A/OPC cells predominantly differentiated into OLIG2, PDGFRα and O4-positive cells (Fig. 2C, D) with some sparse cells expressing the myelin marker MBP, which was completely absent in the other populations (Fig. 2D). Thus, while both GRPs and O2A cells could differentiate to OPCs, only OPCs arising from O2A cells had the capacity to differentiate into mature OLs.

Fig. 2: Isolated glial precursor population varied response to differentiation factors.
figure 2

Immunofluoresence staining to characterize sorted glial precursor sub-populations after differentiation. GRP glial-restricted glial progenitors, ARP astrocyte-restricted progenitors, O2A oligodendrocyte-type-2 astrocyte progenitors. Nuclei are stained in blue with Hoechst 33342, scale bar:50 um. A SOX10 in red, GFAP in green. B S100β in red, OLIG2 in green. C PDGFRα in red, NF1A in green. D O4 in red, MBP in green.

Full size image

To explore the cell fate decisions of our glial precursor subpopulations, we analyzed their developmental trajectories in silico16. Following guided analysis starting at the GRP cluster, we identified three separate lineages, lineage 1 (terminating at astro_3 in our dataset), lineage 2 (rest OPC) and lineage 3 (pre OL) (Fig. 3A). Lineages 2 and 3 have been identified as essential OPC subpopulations in the rodent brain17, which are retained throughout postnatal life. To understand the molecular drivers underlying the divergence of lineages 2 and 3, we evaluated the top differentially expressed genes along these two lineages (Fig. 3B, C). Top genes (fold change, p > 0.05) include SGCD, RNF144A, DNM3, and PKP4 contributing to lineage 3 (Fig. 3C), and complete gene lists are included in Supplemental Data 3. The significantly upregulated genes in lineage 3 corresponded to GO terms related to oligodendrocyte differentiation and myelination (Supplemental Fig. 2, Supplemental Data 4, 5). Our data shows that although OPCs are generated with our differentiation protocol, only a proportion of these are destined to become mature OLs while the other are retained as an OPC reservoir containing both resting and proliferative cells.

Fig. 3: Only one precursor population is destined to give rise to mature OLs.
figure 3

A Dim plots representing pseudotime lineages as generated by Slingshot originating from the “GRP” cluster. Lineages end at the “Astro_3”, “rest OPC”, and “pre OL” clusters. The Feature Plots represent the topmost significant genes (ordered by Log2FC, p > 0.05) which contribute to lineage 2 (B) and lineage 3 (C) following subset for OL-lineage cells and patternTest.

Full size image

The differentiation of OL-lineage cells from glial precursor populations is a dynamic and multifaceted process. The sequencing of cultures at various timepoints (Day 75, 85, and 95) revealed stable populations of OPCs, O2A precursors, and GRPs, suggesting that these progenitors maintain their identities even under varying conditions. The biological meaning of the persistence of these precursors remains to be established and would require further study. Following pseudotime analysis, we observed the retention of a cluster of OPCs capable of generating mature OLs as well as a reservoir of resting OPCs. Resting, or homeostatic OPCs17 have a broad range of functions which are important for CNS development and maintenance. Their roles include angiogenesis5 and mediation of inflammation within the CNS6. Additionally, the differential gene expression analysis at the bifurcation point of the two identified lineages offered insight into the molecular mechanisms that drive lineage commitment (Fig. 3, Supplemental Fig. 2, Supplemental Data 2–5). The presence of a resting OPCs reservoir can provide a novel target for further works addressing remyelination and human developmental processes.

These results suggest that while multiple glial progenitor populations are present, only a specific subset ultimately gives rise to mature OLs, with the remaining cells retained or generated astrocytes leading to heterogeneity in the dish. Here we provide a workflow for the isolation and interrogation of early OL-lineage cells. Our flow cytometry panel to subdivide glial progenitors provides an avenue for the generation of a purer OL culture by the isolation of O2A cells (Fig. 2). Further, by elucidating which transcripts are prominent for fate switching in the OL-lineage (Fig. 3), we can then find compounds to aid in generation of late-stage OLs. This work provides a deeper understanding of how glial precursor populations contribute to OL development in human cells, and a novel model for studying myelination, remyelination, and glial cell plasticity in neurological diseases.

Methods

Cell lines

The lines used in this study include 345018 and 81280 (obtained from Genome Quebec). The complete profiles of the iPSCs have been published previously14. The use of iPSCs and stem cells in this research was approved by the McGill University Health Centre Research Ethics Board (DURCAN_IPSC/2019–5374).

OL-lineage generation from human iPSCs

Glial progenitors were generated from human iPSCs according to our previous publication14. At day 75 of the protocol, we have a mixed-glial culture consisting of committed glial progenitors, astrocytes, and OPCs (Fig. 1B). Passaging our mixed-progenitor culture allows for expansion and use for differentiation as needed. For differentiation, a commercially available basal medium is supplemented with growth serum (Astrocyte Medium with Astrocyte Growth Supplements, ScienCell) for 10 days (Day 85), followed by 10 ng/mL IGF-1 (Peprotech) and 60 ng/mL T3 (Sigma Aldrich) for the following 11 days to reach a majority of OL-lineage cells at a total of 21 days (Day 95).

Single cell sequencing

Library preparation, sequencing, alignment

Cells generated from both 3450 and 81280 (one cell prep per condition), were collected using TrypLE (Gibco), resuspended in PBS (Gibco) with 1% BSA Fraction V (Gibco). Cells were counted and then fixed using the Parse fixation workflow that allows samples to be stored at -80°C until the barcoding and library steps are carried out. Between 100,000 and four million cells may enter the fixation. On the day of library preparation, 2770 cells from each sample are combinatorically barcoded, with each cell receiving three unique barcodes over three consecutive splitting and pooling steps. Barcoded cells are then split into sublibraries, lysed, reverse transcribed, and barcoded once more in a sub library-specific manner. Sublibraries were sequenced at a depth of 50k reads/cell through the McGill Genome Centre Platform using an Illumina NovaSeq 6000. Generated Fastq files are read into the Parse Bioscience analysis pipeline (version 1.0.5p) on the high-performance Digital Research Alliance of Canada computing cluster Béluga to assign samples to barcodes, combine sublibraries, and undergo initial quality control. Analysis (Seurat, Slingshot, tradeSeq) The Parse pipeline output was read into R v4.3.3 using Seurat v5.1.0 to further quality control and process the data. Filters were set to nFeature_RNA < 12000 & nFeature_RNA > 300, nCount_RNA < 50000, and percent.mt < 25. Slingshot v2.10.016 was used for the pseudotime analysis, and tradeSeq v1.16.019 for evaluation of DEGs associated with lineage divergence. GO Enrichment Analysis was performed with clusterProfiler v4.10.120.

Fluorescence activated cell sorting

Live sorting was performed using the aria fusion cell sorter (BD Biosciences). Cells were collected and passed through 40 µM mesh to ensure single cell suspension. Cells were stained with CD44 (Biolegend), PDGFRα (BD Optibuild) and A2B5 (Miltenyi biotec) as previously described14. Cells were placed into culture post-FACS for differentiation and analysis by immunofluorescence imaging.

Immunofluorescence staining and imaging

Cells were stained as previously described14. Briefly, cells were stained with Anti-O4 (R&D), SOX10 (R&D), PDGFRα (Cell Signaling), OLIG2 (Millipore), NF1A (Sigma), GFAP (Dako), and S100β (Sigma) to characterize cells resulting from differentiation post-FACS. Images were acquired with a Zeiss Axio Observer Z1 Inverted Microscope using 20x magnification objective (N.A 0.8) and a ZEISS Axiocam 506 mono camera.