Genetic and epigenetic determinants of neurogenesis and myogenesis

Abstract

The regulatory networks of differentiation programs have been partly characterized; however, the molecular mechanisms of lineage-specific gene regulation by highly similar transcription factors remain largely unknown. Here we compare the genome-wide binding and transcription profiles of NEUROD2-mediated neurogenesis with MYOD-mediated myogenesis. We demonstrate that NEUROD2 and MYOD bind a shared CAGCTG E box motif and E box motifs specific for each factor: CAGGTG for MYOD and CAGATG for NEUROD2. Binding at factor-specific motifs is associated with gene transcription, whereas binding at shared sites is associated with regional epigenetic modifications but is not as strongly associated with gene transcription. Binding is largely constrained to E boxes preset in an accessible chromatin context that determines the set of target genes activated in each cell type. These findings demonstrate that the differentiation program is genetically determined by E box sequence, whereas cell lineage epigenetically determines the availability of E boxes for each differentiation program.

Figure 1. NEUROD2 binds genome-wide in P19 cells differentiated into neurons

(A) Differentiation of P19 cells by expression of NEUROD2. Immunostaining of P19 cells before (top panels) and 72 hours after (bottom panels) transduction with NEUROD2 lentivirus (green: Tuj1 antibody; red: NEUROD2 antibody; blue: DAPI). (B) NEUROD2 antibody is specific for NEUROD2. Immunoprecipitation of ³⁵S-labeled in vitro translated bHLH proteins relative to 10% input (lane 1) with either non-specific IgG (lane 2) or 2 different NEUROD2 antibodies (lanes 3 and 4). (C) NEUROD2 binding is associated with, but does not reliably predict, gene up-regulation. NEUROD2 ChIP-Seq and microarrays were performed in P19 cells before and 72 hours after transduction with NEUROD2 lentivirus. NEUROD2 peak height (Y-axis, square root transformation) of binding sites located within the CTCF domain of gene TSSs is plotted against the log-2 fold change in mRNA expression (X-axis) in smooth scatter plot (left) and boxplot (right) binned by level of activation. The blue trend line in the scatter plot was computed using the loess local regression method; in the boxplot, the vertical bounds represent the 25^th and 75^th percentile, the width represents the size of the dataset, the dot is the median value, and the whisker extends to the extreme value (minimum or maximum), bounded by 1.5 times IQR (25^th and 75^th interquartile range) from the box See also Figure S1.

Figure 2. NEUROD2 and MYOD possess shared and private binding sites

(A) E-box motif enrichment at NEUROD2 and MYOD peaks (a and d) demonstrates central dinucleotide and flanking sequence preferences that consist of a factor-specific motif (b and e) and a shared motif (c and f). (B) Motifs enriched (see Experimental Procedures) under NEUROD2 peaks in P19 cells compared to background. All motifs posses z-values > 5 based on a logistic regression model, with an approximate p-value of < 10⁻⁷ (ratio: enriched/depleted ratio of motifs; fg.frac, bg.frac: fraction of foreground/background sequences that contain at least one motif occurrence). (C) NEUROD2 and MYOD bind with higher affinity to their private E-box sequences. Top: EMSA using translated NEUROD2 or MYOD and E12 mixed with probes containing identical flanking sequences and either a MYOD-preferred (probe A), or NEUROD2-preferred (probe B) E-box and competed with cold A or B probe as shown above each lane. * indicates E12 homodimer. Bottom: EMSA using probes containing either a MYOD-preferred (probe C) or NEUROD2-preferred (probe D) E-box with flanking sequence from a natural site. (D) Comparison of the top 30,000 peaks (30K) bound by NEUROD2 in P19 cells (X-axis) and MYOD in MEFs (Y-axis) demonstrates ~20% overlap of binding sites. From the origin, bins represent the top 3K peaks, then the top 6K peaks, etc, as determined by peak height rank. Colors represent the proportion of sites bound by both NEUROD2 and MYOD (see Experimental Procedures). (E) NEUROD2 and MYOD shared E-box sequence correlates with shared binding sites. Comparison of binding site overlap between NEUROD2 in P19 cells and MYOD in MEFs restricted to the top 30K peaks centered on a GC E-box demonstrates ~40% overlap. See also Figure S2.

Figure 3. The neurogenic and myogenic programs correlate with NEUROD2 and MYOD binding to private sites

(A) Scatter plot of private and shared NEUROD2 and MYOD peaks within the promoter regions (+/− 2kb from the TSS) of genes up-regulated by NEUROD2 in P19 cells (ND.P19), MYOD in MEFs (MD.MM), or up-regulated by both (shared). The number of reads is represented in square root transformation. Sites are further characterized by PWM score (see Experimental Procedures): green, NEUROD2 private site; red, MYOD private site; black, shared site. Genes with multiple TSS were excluded. (B) Sites occurring within 2 kb of a TSS plotted against the fold change in gene activation for (a) NEUROD2 and (b) MYOD. (Y-axis: square root transformation of peak height; X-axis: log-2 fold change in gene expression). (c) box plot of fold change in gene activation (log 2) comparing private and shared sites for NEUROD2 and MYOD. There is greater gene activation associated with private sites based on a Wilcoxon Rank Sum test for NEUROD2 (p < 10⁻⁶) and for MYOD (p = 0.027). Using a threshold of 2-fold change in expression, 21.1% of genes associated with NEUROD2 private sites have fold change >/= 2, compared to 5.4% of genes associated with shared sites (p = 8.2e-9 per Fisher’s exact test). For MYOD, 25.1% of genes associated with private peaks have fold change >/= 2, and 23.5% for shared peaks (p = 0.67), whereas 11.8% of genes associated with private sites have a fold-change >/= 8 compared to 6.3% of genes associated with a shared site (p = 0.03). (C) Reporter constructs containing paired E-boxes with the indicated central nucleotides were transfected into P19 cells with NEUROD2 (top) or MEFs with MYOD (bottom). *p-value < 0.05 by t-test compared to vector without E-box insertion (pGL3); error bars represent 1 standard deviation. (D) Scatter plot of peak height derived from native ChIP-Seq for acetyl-histone 4 in P19 cells prior to (X-axis) and after (Y-axis) transduction with NEUROD2. Shared, NEUROD2-induced change in acetylation at sites shared sites; ND.private, acetylation at NEUROD2 private sites; MD.private, acetylation at sites not bound by NEUROD2 in P19 cells. Number of reads are shown in square root transformation. (E) Y-axis represents the number of raw reads from native ChIP-Seq for acetyl-histone 4, divided by strand (blue: + strand, red: - strand). X-axis represents nucleotide position centered on the E-box closest to the summit of either the private (left half) or shared (right half) NEUROD2 peaks. There is little histone acetylation in P19 cells at baseline (bottom panels), and a significant increase in histone acetylation after differentiation with NEUROD2 (top panels). See also Table S3.

Figure 4. Chromatin accessibility is necessary but not sufficient for NEUROD2 and MYOD binding

(A) Scatter plots comparing (a) NEUROD2 binding sites in P19 cells and (b) MYOD binding sites in MEFs with PvuII nuclease accessibility at these sites. NEUROD2 and MYOD peak height (Y-axis) and the normalized accessibility of PvuII sites (X-axis, see Experimental Procedures for calculation) are represented in square root transformation. Only PvuII sites in the context of good MYOD and NEUROD2 motif matches with PWM scores >/= 14 are included. Blue line is the fitted loess curve. (B) Motif enrichment analysis comparing bound and unbound sites within PvuII accessible areas for (a) NEUROD2 in P19 cells and (b) MYOD in MEFs (ratio: enriched/depleted ratio of motifs; fg.frac, bg.frac: fraction of foreground/background sequences that contain at least one motif occurrence). (C) Plot of E-box PWM (Y-axis) for (a) NEUROD2 and (b) MYOD bound and unbound sites within PvuII accessible regions demonstrates a higher average PWM at bound regions. (D) Plot of the number of E-boxes at PvuII accessible regions either bound or unbound by (a) NEUROD2 in P19 cells or (b) MYOD in MEFs. Colors represent the number of E-boxes located within the 200bp window of a PvuII accessible site. X-axis is the frequency of sites containing the depicted number of E-boxes. See also Figure S3.

Figure 5. NEUROD2 and MYOD binding between cell types is strongly determined by chromatin accessibility

(A) Comparison of the top 30,000 peaks (30K) bound by NEUROD2 and MYOD in the same cell type (left: P19 cells, right: MEFs) demonstrates a ~30% overlap of peaks. From the origin, bins represent the top 3K peaks, then the top 6K peaks, etc, as determined by peak height rank. Color scale represents the percentage of peaks bound by both NEUROD2 and MYOD (see Experimental Procedures). Bins are presented with their corresponding p-values for peak height. (B) Restriction of the comparison in (A) to the top 7–10,000 peaks (7–10K) containing a RRCAGCTGG E-box. (C) Further restriction of the comparison in (B) to the top 1,500–3,000 peaks (1.5–3K) containing a RRCAGCTGG E-box with a high nuclease accessibility score (normalized value > 2). (D) Comparison of the top 30,000 peaks (30K) bound by NEUROD2 in both P19 and MEFs (left) or MYOD in both P19 and MEFs (right). (E) Restriction of the comparison in (D) to the top 7–10,000 peaks (7–10K) containing a RRCAGCTGG E-box. (F) Further restriction of the comparison in (E) to the top 1,500–3,000 peaks (1.5–3K), and containing a RRCAGCTGG E-box with a high nuclease accessibility score.