Epigenomic profiling of retinal progenitors reveals LHX2 is required for developmental regulation of open chromatin

Abstract

Retinal neurogenesis occurs through partially overlapping temporal windows, driven by concerted actions of transcription factors which, in turn, may contribute to the establishment of divergent genetic programs in the developing retina by coordinating variations in chromatin landscapes. Here we comprehensively profile murine retinal progenitors by integrating next generation sequencing methods and interrogate changes in chromatin accessibility at embryonic and post-natal stages. An unbiased search for motifs in open chromatin regions identifies putative factors involved in the developmental progression of the epigenome in retinal progenitor cells. Among these factors, the transcription factor LHX2 exhibits a developmentally regulated cis-regulatory repertoire and stage-dependent motif instances. Using loss-of-function assays, we determine LHX2 coordinates variations in chromatin accessibility, by competition for nucleosome occupancy and secondary regulation of candidate pioneer factors.

Keywords: Developmental neurogenesis; Epigenomics.

Fig. 1

Pairwise comparison of ATAC-Seq data identifies open chromatin regions from early and late RPCs, with a broad overrepresentation of LHX2 related motifs. a Workflow for epigenomic profiling of RPC. b Venn-diagram represents open chromatin regions identified in early- and late-stage VSX2 (CHX10)—GFP-positive RPCs. c Hierarchical clustering of known motifs in the vertebrate genome (left, Jaspar 2016 non-redundant vertebrates core). Inset represents the major cluster of probabilistically assigned position weight matrices (PWMs) identified in open chromatin regions from early- and late-stage RPCs (scale = 1 node/pixel) (linkage = average; similarity threshold cor = 0.6, ncor = 0.4, w = 5) (Lhx2 instances as blue pixel strokes). d Relative RNA-Seq expression for homeobox transcription factors. e Representative LHX2 logos (MA0700.1, k-mer sig = 300, e-value = 1e-300), with positional variations of the same motif instance. f Known motifs enrichment in open chromatin regions from early and late RPCs by binomial scoring of PWMs. g LHX2 footprints from open chromatin regions identified in early- and late-stage RPCs. h Custom tracks of RPC-derived ATAC-Seq profiles and aged-matched LHX2 ChIP-Seq feature the Vsx2 locus (mm9)

Fig. 2

LHX2 regulates cell cycle genes and the Notch signaling pathway by targeting promoters and non-coding elements in nucleosome free regions associated with active enhancers. a LHX2 motif density around the centers of ChIP-Seq peaks at E14 and P2. b Enrichment of LHX2 ChIP-Seq peaks distributed across different genomic regions (log₂ fold enrichment). c–f Percentage of LHX2 target genes in retinal term assigned to at least one age-matched LHX2 ChIP-Seq peak. Enrichment was computed between target genes in term and total number of known genes in term (Supplementary Table 4 and Supplementary Data 1) for peaks shared at E14 and P2 (c), stage-specific peaks (d), and all peaks detected at E14 (e) and P2 (f). RNA-Seq from age-matched Lhx2 cKO retinas identifies Lhx2-dependent genes sets. (*p-value < 0.05, ***p-value < 0.001). Asterisks in parenthesis refer to p-values before Bonferroni–Hochberg correction (target genes populating the enriched ontologies are in Supplementary Data 2). g, h Heatmaps of raw reads from LHX2 ChIP-Seq peaks are plotted across nucleosome centered regions from age-matched ATAC-Seq samples. Each row represents a 3 kb window centered at maximum read pile-up. LHX2 motif occurrence in open chromatin regions is displayed and reported in Supplementary Tables 6. LHX2 motifs co-localize with RNA-Seq raw reads. Meta-profiles of the class II enhancer-associated H3K27ac marks were compiled at bidirectionally transcribed regions from the E14 (g) and P2 RPCs fractions (h) (background in gray, opposite strands replicates in hue). i, j Functional enrichment of Lhx2-dependent genes sets encoded within open chromatin regions (Supplementary Tables 6) by binomial distribution

Fig. 3

Binding sites for transcription factors with predicted pioneer function co-occur with LHX2 peaks. a, b Hierarchical clustering of LHX2 ChIP-Seq regulatory motifs and assigned representative logos are represented at E14 (a) and P2 (b) (linkage = average; similarity threshold cor = 0.6, ncor = 0.4, w = 5). The most enriched cluster comprises LHX2 and multiple variations of the same motif. c, d Known transcription factor motifs preferentially enriched in either E14 or P2 LHX2 ChIP-Seq peaks and identified by pairwise comparison are shown. Percentage of motif occurrences are reported for input and background datasets. The relative expression level of the corresponding transcription factor mRNA at E14/P2 is indicated by the blue/red color gradient, respectively

Fig. 4

LHX2 affects local and global chromatin accessibility. a, b ATAC-Seq read distribution was plotted relative to the center of LHX2 footprints from controls and age-matched Lhx2 cKO. c, d Read distribution profiles across nucleosome free regions (nfr) are derived from ATAC-Seq control and Lhx2 cKO, plotted around age-matched LHX2 ChIP-Seq peaks centers and across all the identified age-matched open chromatin regions in a 3 kb window. The densitometric difference in read coverage between control and Lhx2 cKO is reported on the right. e, f Read distribution profiles flanking nucleosomes centered regions (ncr) are derived from ATAC-Seq control and Lhx2 cKO and plotted across all the identified age-matched open chromatin regions in a 3 kb window

Fig. 5

LHX2 affects local and global chromatin accessibility by epistatic and steric regulation of transcription factors with high pioneer potential. a, b Correlation between LHX2 occupancy and local accessibility at LHX2 target sites in age-matched open chromatin regions in E14 and P2 control conditions and in Lhx2 cKO. Two-tailed t-test statistics is reported. For loss of accessibility statistics at LHX2 targets, refer to Supplementary Table 7. c, d Paired variations in gene expression and local chromatin accessibility are displayed for loci encoding LHX2 targeted transcription factors in embryonic Lhx2 cKO (Supplementary Data 3) (c) and developmentally regulated post natal targets (d). Representative transcription factors are highlighted with arrows (green/red for up/downregulation by RNA-Seq). e, f Correlation of open chromatin profiles obtained from ATAC-Seq flow-sorted fractions from control and Lhx2 cKO, normalized by the library size. Two-tailed t-test statistics is reported. For global loss of accessibility refer to Supplementary Table 8. g Cumulative distribution of differentially expressed transcription factors across intervals of pioneer potential. Transcription factors with predicted pioneer potential based on P2 RPCs accessibility profiles compared to E14 were identified and filtered based on RNA-Seq. Kolmogorov–Smirnov test was applied to putative pioneer TF with differential (blue) or comparable (red) expression between time points. Differentially expressed genes are distributed across higher intervals of pioneer potential than those shared between time points. h Footprint counts for individual transcription factors with predicted pioneer activity matching LHX2 co-occurrent motifs are shown in control and Lhx2 cKO from E14 and P2 retina. For predicted pioneer potential, refer to Supplementary Table 12

Fig. 6

LHX2 coordinates variations in chromatin accessibility at the Sox2 locus. Custom tracks of E14 and P2 LHX2 ChIP-Seq reads and age-matched ATAC-Seq from purified RPCs and post mitotic precursors were configured on the mm10 UCSC murine genome assembly. Regions of interest targeted by LHX2, where variations by RNA-Seq and/or ATAC-Seq coverage are observed in Lhx2 cKO retinas are highlighted. Association with H3K27ac-labeled enhancer elements is also indicated. The LHX2-coordinated regulatory module is defined as follows: regulatory regions that display variations in ATAC-Seq coverage without a corresponding variation in the nearest transcript by E14 Lhx2 cKO RNA-Seq are putatively assigned (arrows) to the nearest gene exhibiting variation in transcript levels

Fig. 7

Footprinting analysis and competition for nucleosome occupancy by predicted pioneer factors reflect their developmentally regulated reliance on LHX2. a–f ATAC-Seq average cut profiles showing footprints for LHX2, SOX2, NF-I (NFIA/B/X), KLF4/9, ASCL1, and HES5 from E14 and P2 control and Lhx2 cKO samples. g–l Nucleosome occupancy at motif centers is reported for LHX2, SOX2, NF-I, FIA/B/X), KLF4/9, ASCL1, and HES5 in E14 and P2 control and Lhx2 cKO samples