MyoD induced enhancer RNA interacts with hnRNPL to activate target gene transcription during myogenic differentiation

Abstract

Emerging evidence supports roles of enhancer RNAs (eRNAs) in regulating target gene. Here, we study eRNA regulation and function during skeletal myoblast differentiation. We provide a panoramic view of enhancer transcription and categorization of eRNAs. Master transcription factor MyoD is crucial in activating eRNA production. Super enhancer (se) generated seRNA-1 and -2 promote myogenic differentiation in vitro and in vivo. seRNA-1 regulates expression levels of two nearby genes, myoglobin (Mb) and apolipoprotein L6 (Apol6), by binding to heterogeneous nuclear ribonucleoprotein L (hnRNPL). A CAAA tract on seRNA-1 is essential in mediating seRNA-1/hnRNPL binding and function. Disruption of seRNA-1-hnRNPL interaction attenuates Pol II and H3K36me3 deposition at the Mb locus, in coincidence with the reduction of its transcription. Furthermore, analyses of hnRNPL binding transcriptome-wide reveal its association with eRNAs is a general phenomenon in multiple cells. Collectively, we propose that eRNA-hnRNPL interaction represents a mechanism contributing to target mRNA activation.

Fig. 1. Elucidation of enhancer transcription in muscle cells.

a GRO-seq detected eRNAs that were up-, down-regulated or unchanged in myotube (MT) vs myoblast (MB) cells, and the changes were correlated with H3K27ac remodeling. b Expression of neighboring genes associated with up- or down-regulated eRNAs in MT vs MB. c The eRNAs were categorized into ‘stable’ (captured by GRO-seq and total RNA-seq) and ‘unstable’ transcripts (captured only by GRO-seq); the divergent enhancer transcription was further categorized into three types of pairs: Bi-stable (both directions generate stable transcripts), Uni-stable (only one direction generates stable transcript) and Unstable (both directions generate unstable transcripts). d Distribution of the above types of divergent eRNAs. e Box plot showing the read density (RPM) of GRO-seq signals or Pol II binding on the above three types of eRNAs in MT. f Analyzing total RNA-seq or GRO-seq revealed that a higher percentage of SEs in MB or MT gave rise to eRNAs (eRNA+) compared to TEs. g seRNAs displayed higher level of GRO-seq signals compared to teRNAs. h TF hotspot regions showed markedly higher levels of GRO-seq signals compared to non-hotspot enhancer regions. i Genomic snapshots of representative eRNAs identified from MB- (left) or MT-expressed SE (right), showing H3K4me1, H3K4me2, H3K27ac, H3K4me3, Pol II and H3K36me3 ChIP-seq profiles, and GRO-seq in MB and MT cells. The red bar highlights the SE region. The “transcript” track indicates transcript units identified through GRO-seq. GRO-seq signals are displayed in “+” (red) and “−” (light green) strands separately. j qRT-PCR measurement of expression dynamics of several MT seRNAs during 120 h differentiation course of C2C12 myoblast. k seRNA expressions were measured in the muscles after cardiotoxin (CTX) injection induced regeneration. n = 3 per group. Data in j and k represent the average of three independent experiments ± s.d. Data in b, e, and h are presented in boxplots. Center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range. Statistical analyses in b, e, and h were done by Mann–Whitney non-parametric test; ***P < 0.001. Source data are provided as a Source Data file.

Fig. 2. MyoD plays a crucial role in inducing MT eRNAs.

a Unsupervised clustering of TF binding at eRNA TSSs in MB or MT. The color code indicates the Pearson correlation coefficient (PCC) between two TFs at their binding sites. b Comparison between GRO-seq, total and PolyA⁺ RNA-seq signals in the MyoD binding proximal regions (±1 kb of the center of MyoD-binding sites) at SEs and TEs. c MyoD ChIP-seq signals within ±2.5 kb flanking TSSs of up-regulated (left) and down-regulated (right) eRNAs. d Distribution of averaged GRO-seq signals from SEs or TEs in WT or MyoD knockout (MyoD^−/−) cells. e Illustration of eRNA down-regulation in MyoD^−/− vs WT cells by GRO-seq tag counts on seRNA-1, -2, and -5. The bar graph shows the quantification of GRO-seq signals in RPKM. f qRT-PCR measurement of seRNAs in the differentiating MyoD^−/− vs WT cells. g qRT-PCR detection of seRNAs from 48-h-differentiated C2C12 cells transfected with either control or MyoD siRNA. h Top: 10T1/2 cells were transfected with either control or MyoD expressing plasmid; the cells were collected in growth medium (GM) or differentiate medium for 48 h (DM). The relative expression of seRNAs, MyoD and Myogenin were measured by semi-quantitative RT-PCR. Bottom: Western blot confirmed MyoD overexpression. i MyoD ChIP-PCR at the TSS of seRNA-1 or seRNA-2 in MT cells. j Luciferase reporter activity of seRNA promoter was detected in 48-hr-differentiated C2C12 cells transfected with either control or MyoD siRNA. k Luciferase reporter activity of the above seRNA promoter in 10T1/2 cells overexpressing MyoD. l Distribution of MyoD ChIP-seq signals on the TSSs of seRNA-1, -2, and -5 in 10T1/2 cells overexpressing MyoD. Data in f, g, i represent the average of three independent experiments ± s.d. Data in b are presented in boxplot. Center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range. Statistical analysis was performed by two-way ANOVA with Sidak’s post-hoc test (f) or two-tailed unpaired Student’s t-test (g, i, j, k), n.s., not significant, *P < 0.05, **P < 0.01 and ***P < 0.001. Source data are provided as a Source Data file.

Fig. 3. seRNA-1 and seRNA-2 regulate target gene expression.

a Genomic snapshots of mouse seRNA-1 and seRNA-2 genes showing Pol II and histone marks ChIP-seq profiles, Poly A⁺ RNA-seq and GRO-seq in MB and MT cells. The red bar highlights the SE region and the red box indicates the seRNA locus. GRO-seq signals are displayed in “+” (red) and “−” (light green) strands separately. b Top: schematic illustration of genomic structure of mouse seRNA-1 relative to the neighboring genes Mb and Apol6. Bottom: Left: Product of RACE cloning (5′ and 3′) of seRNA-1; Right: Detection of seRNA-1 molecules (red) in MT by single molecule RNA FISH. Scale bar, 5 μm. c Top: schematic illustration of genomic structure of mouse seRNA-2 relative to the neighboring genes Atp1a1 and Igsf3. Bottom: Left: Product of RACE (5′ and 3′) cloning of seRNA-2; Right: FISH detection of seRNA-2 in MT. Scale bar, 5 μm. Quantification of FISH signals in b, c corresponding to seRNA transcripts in MB and MT cells. Cells with at least one spot in the nucleus were regarded as “transcribing”. DNA (blue) was stained with DAPI. A representative image was shown. d qRT-PCR analysis of RNAs purified from nuclear and cytosolic fractions of C2C12 cells. e qRT-PCR detection of seRNA-1 and seRNA-2 in the differentiating SCs isolated from muscles of Tg: Pax7-nGFP mice. f Left: qRT-PCR detection of seRNA-1 and neighboring genes from 48-h-differentiated C2C12 cells transfected with either control or seRNA-1 siRNA (si-se1#1 or si-se1#2). Right: qRT-PCR measurement of expression kinetics of seRNA-1 and the Mb and Apol6 during C2C12 differentiation. g The above experiments were performed for seRNA-2 and its neighboring Atp1a1 and Igsf3 genes. Data represent the average of three independent experiments ± s.d. Statistical analysis was performed by one-way ANOVA with Tukey’s post-hoc test (e), or two-tailed unpaired Student’s t-test (f, g) n.s., not significant, *P < 0.05, **P < 0.01 and ***P < 0.001. Source data are provided as a Source Data file.

Fig. 4. seRNA-1 and seRNA-2 interact with hnRNPL.

a RNA pull-down assay was performed using biotinylated seRNA or antisense control RNAs in nuclear lysates of MT cells and the purified proteins were run on SDS-PAGE. The highlighted bands were extracted and subjected to mass spectrometry (MS) analysis. b Western blot (WB) analysis confirmed the specific association of seRNA-1 or -2 with hnRNPK and hnRNPL. No association with MED1, RAD21, RBBP5, YY1, and MyoD proteins was detected. c RNA immunoprecipitation (RIP) was performed with antibodies against hnRNPK or hnRNPL in non-crosslinked differentiating cells and followed by qRT-PCR analysis of retrieved RNAs. Enrichment was determined as RNAs associated to hnRNPK or hnRNPL IP relative to IgG control. d The indicated deletion fragments of seRNA-1 or -2 were in vitro generated and used for RNA pull-down assay to map the interacting region with hnRNPL. e The indicated Myc-tagged hnRNPL variants (D1, D2, D3, and D4) were transfected into 293 T cells and the whole cell lysates were used for RNA pull-down assay with biotinylated seRNA to map the hnRNPL domain interacting with seRNA-1 or -2. Left: Western blot showing the overexpression of the above variants. D2 and D4 interacted with seRNA-1 or seRNA-2; Right: schematic of structures of the above variants. RRM, RNA recognition motif. f C2C12 cells were transfected with siRNAs targeting hnRNPL or scrambled negative control (si-NC). At 24 h post transfection, the cells were switched to DM for 48 h. The expression of seRNA target genes was measured by qRT-PCR. g Nuclear run-on assay was performed to measure nascent transcription of Mb or Atp1a1 in the above cells transfected with si-hnRNPL#1. Data represent the average of three independent experiments ± s.d. Statistical analysis was done by two-tailed unpaired Student’s t-test (f, g), n.s., not significant, *P < 0.05, **P < 0.01 and ***P < 0.001. Source data are provided as a Source Data file.

Fig. 5. CAAA tract is indispensable for seRNA-1 function.

a Schematic illustration of CAAA deletion medicated by CRISPR-Cas9 editing in C2C12 cells. Top: black box depicts the deletion region. Primer set 1 (1F and 1R) was used for cDNA genotyping. Primer set 2 (2F and 2R) was used for qRT-PCR analysis. Middle: CAAA tract sequence is highlighted in red. sgRNAs were designed to delete the underlined sequence encompassing the CAAA tract. Bottom: Result from Sanger sequencing confirmed the CAAA deletion. b Biotinylated seRNA-1 transcripts of full length (FL) or 5′ end fragment with or without CAAA deletion (ΔCAAA) were used in RNA pull-down assay to reveal seRNA-1 binds hnRNPL through the CAAA tract embedded in its 5′ region. c hnRNPL RIP was performed in WT or the generated CAAA deletion (KO) cells to show hnRNPL/seRNA-1 binding was abolished in the KO cells as compared to WT control cells. Enrichment was determined as RNA associated to hnRNPL IP relative to IgG control. d Expression of the associated target genes, Mb and Apol6, was decreased but seRNA-1 was increased in the KO cells as measured by qRT-PCR. e Schematic illustration of the CRISPR-Cas9 mediated in vivo deletion of CAAA tract. Pax7^Cas9 mouse at postnatal 10 (P10) age was intramuscularly (IM) injected with 5 × 10¹¹ vg AAV9-sgRNAs viruses and the infected muscles were recovered for analysis at 4 weeks later (P38). n = 3 per group. f Detection of CAAA excision by genomic PCR in muscle tissues injected with AAV9-sg-seRNA-1, compared to AAV-sg-Control. Un-edited product, 304 bp; deletion product (red asterisk), 220 bp. g qRT-PCR was performed to measure the levels of seRNA-1, Mb and Apol6 in the above injected muscles. n = 3 per group. Data represent the average of three independent experiments ± s.d. Statistical analysis was done by two-tailed unpaired Student’s t-test (c, d, g), *P < 0.05, **P < 0.01, and ***P < 0.001. Source data are provided as a Source Data file.

Fig. 6. seRNA-1 modulates hnRNPL, RNA Pol II and H3K36me3 at Mb locus.

a Western blot analysis of hnRNPL cellular distribution in differentiating C2C12 untreated or treated with RNaseA. Whole cell extracts (WCE), nuclei (N). Relative levels of hnRNPL were normalized with H3K36me3 and measured by Image J. b qPCR quantification of RNA (left) and DNA (right) recovered after lacZ ChIRP or seRNA-1 ChIRP with two different biotinylated probe sets (even and odd) in C2C12 MT. P1, P3 and P4 are three different primer pairs for detecting seRNA-1 RNA. Mb, seRNA-1 and Atp1a1 indicate the primers corresponding to the promoter regions. c ChIP-PCR of hnRNPL, CCNT1, CDK9, Pol II, H3K36me3, and KMT3a at regions (1, 2, 3, and 4) across Mb locus in MT vs MB. d Co-IP assay was performed using antibodies against hnRNPL or CCNT1 in C2C12 MT and the interaction between endogenous hnRNPL and CCNT1 or CDK9 was detected. * IgG light chain. e ChIP-PCR of hnRNPL, Pol II, and H3K36me3 at the above Mb loci in control or hnRNPL knockdown cells (#1 or #2). f ChIP-PCR of hnRNPL, CDK9, CCNT1 and H3K36me3 at the above Mb loci in control or seRNA-1 knockdown cells (#1 or #2). g ChIP-PCR of hnRNPL at the above Mb loci in the two CAAA KO cell lines compared to WT control. h Overexpression of hnRNPL in C2C12 cells decreased the expression of seRNA-1 and Mb but not Apol6. i Schematic illustration of dCas9 mediated tethering of seRNA-1 wild type or the CAAA KO mutant (ΔCAAA) to the seRNA-1 TSS or Mb promoter. j qRT-PCR detection of seRNA-1 and Mb in the above cells. k ChIP-PCR of hnRNPL binding at the indicated seRNA-1 or Mb promoter following the above tethering. Data represent the average of three independent experiments ± s.d. Statistical analysis was done by two-tailed unpaired Student’s t-test (b, c, e–h, j, k). n.s., not significant, *P < 0.05, **P < 0.01, and ***P < 0.001. Source data are provided as a Source Data file.