lncRNA Functional Networks in Oligodendrocytes Reveal Stage-Specific Myelination Control by an lncOL1/Suz12 Complex in the CNS

Abstract

Long noncoding RNAs (lncRNAs) are emerging as important regulators of cellular functions, but their roles in oligodendrocyte myelination remain undefined. Through de novo transcriptome reconstruction, we establish dynamic expression profiles of lncRNAs at different stages of oligodendrocyte development and uncover a cohort of stage-specific oligodendrocyte-restricted lncRNAs, including a conserved chromatin-associated lncOL1. Co-expression network analyses further define the association of distinct oligodendrocyte-expressing lncRNA clusters with protein-coding genes and predict lncRNA functions in oligodendrocyte myelination. Overexpression of lncOL1 promotes precocious oligodendrocyte differentiation in the developing brain, whereas genetic inactivation of lncOL1 causes defects in CNS myelination and remyelination following injury. Functional analyses illustrate that lncOL1 interacts with Suz12, a component of polycomb repressive complex 2, to promote oligodendrocyte maturation, in part, through Suz12-mediated repression of a differentiation inhibitory network that maintains the precursor state. Together, our findings reveal a key lncRNA epigenetic circuitry through interaction with chromatin-modifying complexes in control of CNS myelination and myelin repair.

Keywords: EZH2; OPC; chromatin remodeling; functional genomics; long noncoding RNA; myelination; neural precursors; oligodendrocyte differentiation; polycomb repressive complex; remyelination.

Figure 1. De Novo Mapping of lncRNA Transcriptome in OL Lineage Cells

(A) Schematic procedure for discovering lncRNAs in primary mouse OPCs and OLs. (B) Cumulative frequency plot of coding potential CPAT scores for mRNA and lncRNA genes. (C) Cumulative frequency plot of PhastCons conservation scores for mRNA and lncRNA genes. (D) Violin plot of log₂ maximum expression values (FPKM, fragments per kilobase of transcript per million mapped reads) for mRNA and lncRNA genes. Student's t test with Welch's corrections. (E) Box plot of tissue specificity index (Ts score) for mRNA and lncRNA genes. Ts scores are calculated by analysis of fractional expression in OLs across all the tissues examined. Whiskers show the minimum and maximum, and boxes extend from the first to the third quartiles, with cross lines at the medians; Student's t test with Welch's corrections. (F) Heat map representation of ChIP-seq signal density for H3K4me3, H3K27ac, and Sox10 for 5 kb centered on predicted lncRNA TSSs in rat OPCs and OLs. (G) Histogram depicting ChIP-seq signal density on all predicted lncRNA promoters for H3K27ac, H3K4me3, and Sox10 in primary rat OLs. (H) Heat maps showing top differentially expressed protein coding (left) and lncRNA (right) genes during mouse OPC maturation. (I) Cytoscape representation of the 19 largest MCL clusters in the lncOL and protein-coding gene co-expression network from all tissues and cells examined. (J) GO enrichment for the 19 largest MCL clusters. The most significant GO categories are displayed. (K) Heat map displaying Pearson correlation coefficients of lncOL and protein-coding genes in MCL cluster I. Columns and rows represent protein-coding genes and lncRNAs, respectively. Myelin-associated genes and lncOLs are highly correlated.

Figure 2. OL-Enriched lncRNAs Regulate OL Differentiation

(A) UCSC genome browser tracks of lncOL1–4 loci depict RNA-seq signals for OPCs and OLs, along with ChIP-seq signals for H3K4me3 and H3K27ac. Bottom tracks depict de novo transcript models created from Cufflinks and RefSeq annotations. (B) qRT-PCR analyses of lncOL1–lncOL4 levels in different murine tissues at P14. (C) qRT-PCR analyses of Mbp, Myrf, and lncOL1–lncOL4 transcript levels in mouse spinal cords at indicated stages. Data are presented as mean ± SEM, n = 3 animals at each time point. (D) qPCR analyses for lncOL1–lncOL4 in primary mouse OPCs and OLs. Data are presented as mean ± SEM, n = 3 independent experiments; ***p < 0.001; paired Student's t test. (E) qPCR validation of efficiency of siRNA knockdown of lncRNAs. Data are presented as mean ± SEM; n = 4 independent experiments, **p < 0.01, ***p < 0.001;paired Student's t test. (F) qRT-PCR analyses of OL-differentiation-associated genes following treatments with scrambled or lncRNA-targeted siRNAs. Data are presented as mean ±SEM; n = 4 independent experiments, *p < 0.05, **p < 0.01, ***p < 0.001; one-way ANOVA with post hoc Tukey's test. (G) qRT-PCR analyses of OL-differentiation associated genes in rat OLs transduced with control or lncRNA-overexpressing vectors. Data are presented as mean ± SEM; n = 4 independent experiments, *p < 0.05, **p < 0.01, ***p < 0.001; paired Student's t test.

Figure 3. lncOL1 Is a Conserved OL-Specific lncRNA that Regulates OPC Differentiation

(A) Genome browser tracks of the lncOL1 locus in mouse OLs (top) and its corresponding genomic locus in rat OLs (bottom). Top tracks depict RNA-seq signals for mouse OPCs and OLs and ChIP-seq signals for H3K4me3 and H3K27ac (Lu et al., 2016). Bottom tracks represent RNA-seq signals for rat OPCs, OLs, astrocytes (Astro), and neural precursor cells (NPCs) and ChIP-seq signals for H3K4me3, H3K27ac, Olig2, and Sox10 in rat OPCs and OLs. (B) qRT-PCR analyses of lncOL1, Myrf, and Mbp transcripts in primary mouse OPCs, iOLs, and mOLs. Data are presented as mean ± SEM, n = 3 at each timepoint. (C and D) In situ hybridization showing expression of lncOL1 in the cerebral white matter (C) and spinal cord (D) at the indicated time points. Arrows point to lncOL1-expressing cells. CC, corpus callosum. Scale bar, 100 μm. (E) In situ hybridization for lncOL1 (blue) and immunohistochemistry for Olig2, Sip1/Zeb2, CC1, or PDGFRα (brown) at P14. Arrows indicate lncOL1⁺ cells. Arrowheads indicate PDGFRα⁺ OPCs without lncOL1 expression. Scale bar, 10 μm. (F) FISH for lncOL1 in primary OPCs and OLs differentiated for 3 days. Nuclei are counterstained with DAPI. Scale bar, 5 μm. (G) Immunostaining for MBP and Olig2 after 48 or 72 hr of differentiation after PDGFAA withdrawal in control or lncOL1-shRNA transduced mouse OLs. Scale bar, 50 μm. (H) Histogram depicting the percentage of MBP⁺ cells among GFP⁺ transduced cells. Data are presented as mean ± SEM; n = 3 independent experiments; *p < 0.05, **p < 0.01; one-way ANOVA with post hoc Tukey's test. (I) Schematic presentation of control and lncOL1-overexpressing vectors used for in utero electroporation. lncOL1 transcription is driven by bidirectional promoters in pBI. (J and K) Cortical sections at E17.5, 3 days after electroporation with control (pBI-zsgreen) or lncOL1 overexpressing (pBI-lncOL1/Zsgreen) vectors. (K) shows high magnification of transduced cells co-labeled with GFP, MBP, and Olig2. Scale bars in (J), 50 mm; scale bars in (K), 25 μm.

Figure 4. lncOL1-Deficient Mice Exhibit Defects in the Onset of OL Differentiation and Myelination

(A) CRISPR/Cas9-mediated knockout at the lncOL1 locus. Two sgRNA recognition sites for lncOL1 deletion are demarcated in red. Primer pairs 1–5 (P1–P5) were used for qRT-PCR analysis of knockout in different lncOL1 exons. (B) Validation of lncOL1-KO mouse lines by PCR genotyping. Wild-type allele: 319 bp; mutant allele: 654 bp. (C) qPCR analyses of lncOL1 transcript levels with primer pairs P1–P5 in brains from lncOL1-KO and control littermates at P12. Data are presented as mean ±SEM; n = 3 animals; **p < 0.01, ***p < 0.001; Student's t test. (D) The spinal white matter at P1 from control and lncOL1 null animals was co-immunostained for CC1 and Olig2. Scale bar, 50 μm. (E) Quantification of CC1⁺OLs as a percentage of total Olig2⁺ cells in control and lncOL1-KO spinal cords at P1. Data are presented as mean ± SEM;n =3animals;*p < 0.05; Student's t test. (F) Immunofluorescence labeling for Olig2 and MBP in control and lncOL1-KO corpus callosum at P6. Scale bars, 50 μm. (G) Quantification of MBP⁺ OLs as a percentage of all Olig2⁺ cells within the corpus callosum of control and lncOL1-KO mice at P6. Data are presented as mean ± SEM; n = 4 animals; *p < 0.05; Student's t test. (H) Quantification of PDGFRα⁺ OPC density in control and lncOL1-KO brains (Br) and spinal cords (SC). Data are presented as mean ± SEM; n = 3 to 4 animals. (I and J) Representative electron micrographs of transverse optic nerve sections from control and lncOL1 -KO animals at P9 (I) and P17 (J). Scale bar, 2 μm. (K) Quantification of the percentage of axons myelinated at P9 and P17. Data are presented as mean ± SEM; each data point is an average for a single animal; n = 4 animals; *p < 0.05, ***p < 0.001; Student's t test. (L) Scatter plots of g-ratios of individual fibers from lncOL1 -KO and littermate controls at P17. Between 300 and 400 axons from three mice of each genotype were analyzed; ***p < 0.001. (M) Immunostaining for Ki67, Sox10, and PDGFRα on control and lncOL1 null OPCs cultured with PDGFAA. Scale bar, 50 μm. (N) Percentage of Ki67⁺ proliferating and Ki67^– non-proliferating OPCs among Sox10⁺ PDGFRα⁺OPCs from control or lncOL1 null mice. Data are presented as mean ± SEM; n = 3 independent experiments. (O) Immunostaining for MBP, CNP, and Olig2 on mouse OPCs from control and lncOL1 null mice. Cells differentiated with T3 for 24 and 72 hr are shown. Scale bar, 50 μm. (P) The proportions of CNP⁺ (left) or MBP⁺ (right) OLs in mouse OL cultures from control and lncOL1 null mice. Data are presented as mean ± SEM; n = 2–4 independent experiments; *p < 0.005, **p < 0.01; one-way ANOVA with post hoc Tukey's test.

Figure 5. lncOL1 Is Critical for Proper Myelin Repair after Demyelination

(A) qRT-PCR analyses for lncOL1 and Tcf7l2 levels in spinal cord lesions at various time points post LPC-induced demyelination. Data are presented as mean ± SEM; n = 4 animals at each time point. (B) In situ hybridization for lncOL1 in representative spinal cord lesions from 8-week-old wild-type control mice at 7 and 14 dpl. Scale bars, 100 μm. (C) Quantification of lncOL1-expressing cells in unlesioned and lesioned regions of spinal cords of wild-type control mice at 7 and 14 dpl. Data are presented as mean ± SEM; n = 3; Student's t test. (D) In situ hybridization for lncOL1, Plp, Cnp, and Mag in representative spinal cord lesions from control and lncOL1-KO mice at 14 dpl. Scale bars, 100 μm. (E) Quantification of mature OLs in 14 dpl demyelinated spinal cord of lncOL1 null and control mice. Data are presented as mean ± SEM; n = 9, ***p < 0.001; Student's t test. (F) Immunostaining for CC1 and Olig2 in representative spinal cord lesions from control and lncOL1-KO mice at 14 dpl. Scale bar, 100 μm. (G) Quantification of CC1⁺ OLs in lesions at 14 dpl. Data are presented as mean ± SEM; n = 9, **p < 0.01; Student's t test. (H) Immunostaining for PDGFRα and Olig2 in representative LPC-induced lesions from control and lncOL1-KO mice at 14 dpl. Scale bar, 100 μm. (I) Quantification of PDGFRα⁺ OPCs in lesions at 14 dpl. Data are presented as mean ± SEM; n = 9. (J) Representative electron microscopy images of LPC-induced lesions from control and lncOL1-KO mice at 16 dpl. Scale bar, 2 μm. (K) The percentage of remyelinated axons in LPC-induced lesions of control and lncOL1-KO mice at 16 dpl. Data are presented as mean ± SEM; n = 3, *p < 0.05; Student's t test. (L) The myelin g-ratio in LPC-induced lesions of control and lncOL1-KO mice at 16 dpl. n = 3 animals/genotype, p < 0.001; Student's t test.

Figure 6. lncOL1 Interacts with Suz12/PRC2 Complex to Regulate OL Differentiation

(A) HEK293T cells overexpressing HA-tagged Suz12 and pBI-lncOL1 were lysed and precipitated with anti-Suz12 and control IgG, and analyzed by qRT-PCR for lncOL1 and Gapdh. Data are presented as mean ± SEM; n = 3 independent experiments. *p < 0.05; Student's t test. (B) Immunoblot for HA-Suz12 after RNA pull-down with biotin-labeled lncOL1 RNA or biotin-labeled control RNA (Bmp4). Pull-down by streptavidin magnetic beads with no RNA probe was used as negative control. (C) qRT-PCR quantification of lncOL1 and Gapdh in Suz12 immunoprecipitates from mouse OLs. Data are presented as mean ± SEM; n = 3 independent experiments. ***p < 0.001; Student's t test. (D) Immunoblot for endogenous Suz12 in differentiating Oli-Neu cells after RNA pull-down with biotin-labeled lncOL1 or biotin-labeled control RNA (Bmp4). Pulldown by streptavidin beads without RNA probe as a negative control. The larger band appeared as a non-specific protein associated with beads. (E) qRT-PCR quantification of lncOL1 and Gapdh in Suz12, Eed, and Ezh2 immunoprecipitates using RNA immunoprecipitation (RIP) assay from differentiating oligodendroglial cells (Oli-Neu). Data are presented as mean ± SEM; n = 3 to 4 independent experiments. *p < 0.05, **p < 0.01; Student's t test. (F) P14 mouse spinal cord transverse sections immunostained for Suz12, CC1, and PDGFRα. Scale bar, 100 μm. Boxed area is shown at higher magnification on right; scale bar, 20 μm. (G) Immunoblotting for indicated proteins in rat OPCs at 0, 1, 3, and 5 days of culture under differentiation conditions. (H) qRT-PCR analyses of Suz12 and Eed in Oli-Neu cells transduced with retroviral vectors for expression of shRNAs targeting Renilla luciferase (Rluc) as a control (Ctrl), Suz12, or Eed. Data are presented as mean ± SEM; n = 3, *p < 0.05, **p < 0.01; Student's t test. (I) Immunostaining for MBP and Olig2 after 72 hr of differentiation induced by T3 in GFP⁺ mouse OPCs transduced with retroviral vectors for expression of shRNAs targeting Renilla luciferase, Suz12, or Eed. White arrows indicate control transfected MBP⁺ OLs; arrowheads indicate shSuz12- or shEed-transduced cells, respectively. Scale bar, 50 μm. (J) The percentage of MBP⁺ among GFP⁺ transduced cells. Data are presented as mean ± SEM; n = 3 independent experiments; *p < 0.05, **p < 0.01; one-way ANOVA with post hoc Tukey's test. (K and L) Immunofluorescence labeling for Olig2 and MBP (K) or CC1 (L) in the control and Ezh2cKO corpus callosum at P7. Scale bars, 50 μm. (M) Percentage of CC1⁺ OLs and PDGFRα⁺ OPCs among Olig2⁺ cells in P7 control and Ezh2cKO corpus callosum. Data are presented as mean ± SEM; n = 2–4 animals; *p < 0.05, **p < 0.01; Student's t test. (N) Immunostaining for MBP and Olig2 after 72-hr differentiation after PDGFAA withdrawal. DMSO or 1mM Unc1999 were added to culture medium as indicated. Scale bar, 25 μm. (O) Histogram depicting the percentage of MBP⁺ cells among GFP⁺ cells. Data are presented as mean ± SEM; n = 4 independent experiments; *p < 0.05, ***p < 0.001; one-way ANOVA with post hoc Tukey's test. (P) qRT-PCR analyses of OL-differentiation associated genes in rat OLs transduced with control or lncOL1-overexpressing vectors in the absence or presence of 1μM Unc1999. Data are presented as mean ± SEM; n = 4 independent experiments; *p < 0.05, **p < 0.01; one-way ANOVA with post hoc Tukey's test. (Q) qRT-PCR of myelin-associated gene expression in OLs transduced with lncOL1-overexpressing vectors with or without siRNAs against Suz12. Data are presented as mean ± SEM; n = 4 independent experiments. *p < 0.05; ***p < 0.001; one-way ANOVA with post hoc Tukey's test.

Figure 7. lncOL1 Acts as a Modulatory Factor for Suz12 to Regulate Transcriptome Dynamics during OL Differentiation

(A) A heat map showing genes differentially expressed in control and lncOL1 null OLs cultured under differentiation conditions for 24 hr. Modules I and II represent upregulated and downregulated genes compared to wild-type, respectively. Data are from two separate cultures. (B) Box plots for mRNA levels of module I and II genes in OPCs and newly formed OLs (NFOs). (C) Box plots comparing the levels of active histone marks H3K27ac and H3K4me3 in the promoters of the genes downregulated (module I) and upregulated (module II) in lncOL1 null OL lineage cells. (D and E) GO biological process terms enriched among mRNA genes that show significantly decreased (D) or increased (E) expression upon lncOL1 deletion relative to control. (F) qRT-PCR analyses of genes involved in different regulatory arms of OL differentiation. Data are presented as mean ± SEM; n = 3; *p < 0.05, ***p < 0.001; Student's t test. (G) ChIP-seq density heat maps for Suz12 and H3K27ac within 2.5 kb on either side of the Suz12 and H3K27ac peak centers in OPCs. The sites are ranked in descending order of Suz12 intensity. (H) Box plots for log₂ fold change in expression levels of the genes targeted with or without Suz12 in NFOs and myelinating OLs (MOs) (Zhang et al., 2014) compared to OPCs. (I) Gene signature enrichment analysis plot comparing the Suz12-targeted genes (top 500 genes with Suz12-binding peak within ± 50 kb around TSS) in wild-type or lncOL1-KO mouse OLs. NES, normalized enrichment score. FDR, false discovery rate. (J) GO biological process terms enriched among mRNA genes that are targeted by Suz12, with significantly higher expression levels in lncOL1 null OLs. (K) Genome browser tracks over select gene loci with ChIP-seq density mapping of Suz12, H3K27ac, and H3K4me3 from rat OLs. (L) qRT-PCR analyses of Cyp1b1, H19, and Igf2 in control and lncOL1-KO OLs. Data are presented as mean± SEM; n =3independent experiments, ***p < 0.001; Student's t test. (M) qRT-PCR analyses of Suz12, Cyp1b1, H19, and Igf2 in primary mouse OLs treated with scrambled or Suz12-targeted siRNAs. Data are presented as mean ± SEM; n = 5 independent experiments. **p < 0.01; Student's t test. (N) ChIP-qPCR analyses for Suz12 at the promoters of Igf2, H19, and Cyp1b1 loci or control genomic regions (NC) in primary mouse OLs from control and lncOL1 knockout animals. Data are presented as mean ± SEM; n = 3 independent experiments. ***p < 0.001; Student's t test. (O) Schematic diagram for the lncOL1-mediated control of OL differentiation. When lncOL1 expression is turned on, it associates with Suz12-mediated PRC2 complex and silences the OPC gene regulatory program that antagonizes their differentiation process, and thereby allows the activation of the OL differentiation program. When lncOL1 expression is off, the OPC-promoting gene program is de-repressed and activated, which inhibits OL differentiation processes. Other lncRNAs may cooperate with lncOL1 to regulate the OL differentiation process. In box plots (B), (C), and (H), whiskers show the minimum and maximum, and boxes extend from the first to the third quartiles, with cross lines at the medians; Mann-Whitney-Wilcoxon test.