Repression of developmental transcription factor networks triggers aging-associated gene expression in human glial progenitor cells

Abstract

Human glial progenitor cells (hGPCs) exhibit diminished expansion competence with age, as well as after recurrent demyelination. Using RNA-sequencing to compare the gene expression of fetal and adult hGPCs, we identify age-related changes in transcription consistent with the repression of genes enabling mitotic expansion, concurrent with the onset of aging-associated transcriptional programs. Adult hGPCs develop a repressive transcription factor network centered on MYC, and regulated by ZNF274, MAX, IKZF3, and E2F6. Individual over-expression of these factors in iPSC-derived hGPCs lead to a loss of proliferative gene expression and an induction of mitotic senescence, replicating the transcriptional changes incurred during glial aging. miRNA profiling identifies the appearance of an adult-selective miRNA signature, imposing further constraints on the expansion competence of aged GPCs. hGPC aging is thus associated with acquisition of a MYC-repressive environment, suggesting that suppression of these repressors of glial expansion may permit the rejuvenation of aged hGPCs.

Fig. 1. Transcriptomic characterization of FACS-isolated human fetal GPCs.

A Workflow of bulk and scRNA-Sequencing of CD140a⁺, CD140a^-, and A2B5⁺/PSA-NCAM^--selected 2nd trimester human fetal brain isolates. B Principal component analysis of all samples across two batches. Numbers indicate pairing of samples (fetal CD140⁺ n = 8; CD140a^-, n = 5; A2B5⁺/PSA-NCAM^-, n = 3; biologically independent samples). C Venn diagram of CD140a⁺ vs CD140a^- and CD140⁺ vs A2B5⁺/PSA-NCAM^- differentially-expressed gene sets (FDR < 0.01 and absolute log₂-fold change > 1, calculated with DESeq2). D Log₂-fold changes of significant curated genes for each geneset. Missing bars were not significant. E UMAP plot of the primary cell types identified during scRNA-Seq analysis of all FACS-isolated hGPCs derived from 20 week gestational age human fetal VZ/SVZ. F-G UMAPs of PSA-NCAM^-/A2B5⁺ (F) vs. CD140a⁺ (G) human fetal cells. H Violin plots of cell type-selective marker genes. I Select feature plots of transcription factor regulons (calculated in SCENIC), predicted to be significantly activated in fetal hGPCs (FDR < 0.01, Wilcoxon rank-sum test). Source data are provided as a Source Data file. Error bars ± SEM.

Fig. 2. Adult human GPCs are transcriptionally and functionally distinct from fetal hGPCs.

A Workflow of bulk RNA-Seq analysis of human adult and fetal GPCs. B Principal component analysis of all samples across three batches (fetal CD140a⁺, n = 12; fetal A2B5⁺/PSA-NCAM^-, n = 3; adult A2B5⁺, n = 3; biologically independent samples). C Venn Diagram of Adult vs Fetal differentially expressed gene sets. (FDR < 0.01, Log2FC > 1, calculated with DESeq2). D IPA network of curated significant terms and genes (FDR < 0.001). Node size is proportionate to node degree. Color corresponds to enrichment in either adult (red) or fetal (blue) populations. E Bar plots of significant IPA terms by module. Z-Scores indicate predicted activation in fetal (blue) or adult (red) hGPCs. F Log₂-fold changes and heatmap of network gene TPMs. Source data are provided as a Source Data file. Error bars ± SEM.

Fig. 3. Inference of transcription factor activity implicates a set of transcriptional repressors in the establishment of adult hGPC identity.

A Normalized enrichment score plots of significantly enriched transcription factors predicted to be active in fetal and adult GPCs. Each dot is a motif whose size indicates how many genes in which that motif is predicted to be active; color represents the window around the promoter at which that motif was enriched. B Heatmap of enriched TF TPMs, and (C) log-fold changes for both fetal hGPC isolates vs adult GPCs (fetal CD140a⁺, n = 12; fetal A2B5⁺/PSA-NCAM^-, n = 3; adult A2B5⁺, n = 3; biologically independent samples). D–G Predicted direct transcription factor activity of curated genes split into (D) fetal activators; (E) fetal repressors; (F) adult activators; and (G) adult repressors. Color indicates differential expression in adult (red) or fetal (blue) hGPCs; shape dictates type of node (octagon, repressor; rectangle, activator; oval, other target gene). Boxed and circled genes indicate functionally-related genes contributing to either glial progenitor/oligodendrocyte identity, senescence/proliferation targets, or upstream or downstream TFs that were also deemed activated. Source data are provided as a Source Data file. Error bars ± SEM.

Fig. 4. Induction of features of aging via adult hGPC-enriched repressors.

A Schematic outlining the structure of four distinct doxycycline (Dox)-inducible EGFP lentiviral expression vectors, each encoding one of the transcriptional repressors: E2F6, IKZF3, MAX, or ZNF274. B Induced pluripotent stem cell (iPSC)-derived hGPC cultures (line C27^,) were transduced with a single lentivirus or vehicle for 1 day, and then treated with Dox for the remainder of the experiment. At 3, 7 and 10 days following initiation of Dox-induced transgene expression, hGPCs were isolated via FACS for qPCR. C qPCRs of Dox-treated cells showing expression of each transcription factor, vs matched timepoint controls (n = 3–5 independent experiments/condition) D qPCR fold-change heatmap of select aging related genes. Within timepoint comparisons to controls were calculated via post hoc estimated marginal means tests of linear models following regression of a cell batch effect. E Immunocytochemistry for p16 (left) and p21 (right) in EGFP⁺ cells at 7 days post infection (n = 3 biologically independent samples). Post hoc pairwise comparisons of each overexpression condition to control in (C–E) were calculated via estimated marginal means tests of linear models, following regression of a cell batch effect. FDR adjusted p-values: *<0.05, ** <0.01, *** <0.001. Source data are provided as a Source Data file; exact p values are listed there. Error bars ± SEM.

Fig. 5. E2F6 and ZNF274 overexpression each induce an aged transcriptome in hGPCs.

scRNA-sequencing of E2F6 and ZNF274 overexpression (OE) with a GFP lentivirus used as control, 7 days following doxycycline (Dox) treatment. A UMAPS of isolated GPC populations showing PDGFRA expression in addition to E2F6 and ZNF274 for demonstration of overexpression (EGFP CTR: 921 GPCs; E2F6 OE: 1085 GPCs, ZNF274 OE—906 GPCs). B Venn diagram of both differentially expressed genesets (FDR < 0.05, Log₂FC > 0.25). C, D Upset plots indicating repressed genes shared between fetal vs adult hGPCs and E2F6 I or ZNF274 (D) OE cells, and their respective targets, as determined via rCisTarget from these scRNA-seq OE experiments and the fetal vs adult bulk RNA-seq experiments. E Scoring via AUCell of the rCisTarget determined regulons of either E2F6 or ZNF274. Black lines indicate median expression. F, G Heatmaps indicating Log₂ fold changes and significances of adult vs. fetal GPCs (F) and E2F6 or ZNF274 vs EGFP CTR (G). H Significant direct targets of E2F6 or ZNF274 OE. I qPCR validation of select direct targets (indicated in blue in F; n = 4 biologically independent samples). Central lines indicate the median, boxes show the interquartile range, and the whiskers indicate 1.5 times the interquartile range. Outliers are plotted beyond the whiskers. J Dot plot of curated IPA terms (FDR < 0.05). K KI67 staining of EGFP⁺ cells (Dox Ctr is of DAPI⁺, n = 3 biologically independent samples). L. β-Galactosidase⁺ staining of DAPI⁺ cells (n = 3 biologically independent samples). Statistics for I, K, and L were calculated via estimated marginal means tests of linear models, following regression of cell batch effect. FDR adjusted p-values: *<0.05, **<0.01, ***<0.001, ****<0.0001. Source data are provided as a Source Data file; exact p values are listed there. Error bars ± SEM.

Fig. 6. MicroRNAs drive hGPC transcriptional maturation in tandem with transcription factor activity.

A. Principal component analysis of miRNA microarray samples from human A2B5⁺ adult and CD140a⁺ fetal GPCs (n = 4 biologically independent samples). B. Log2 fold change bar plots and heatmap of all differentially expressed miRNAs (FDR < 0.01, calculated in limma). C Characterization bubble plot of enrichment of miRNAs, versus the average log2 FC of its predicted gene targets. D, E Curated signaling networks of (D) fetal (top) and (E) adult (bottom) enriched miRNAs and their predicted targets. Source data are provided as a Source Data file. Error bars ± SEM.