Epigenomic Profiling of Retinal Progenitors Reveals LHX2 Is

ARTICLE

https://doi.org/10.1038/s42003-019-0375-9 OPEN Epigenomic proﬁling of retinal progenitors reveals LHX2 is required for developmental regulation of open chromatin

Cristina Zibetti 1, Sheng Liu2,3, Jun Wan2,3, Jiang Qian2,4 & Seth Blackshaw 1,2,5,6,7 1234567890():,;

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.

1 Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. 2 Department of Ophthalmology, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. 3 Department of Medical and Molecular Genetics, Indiana University School of Medicine, Indianapolis, IN 46202, USA. 4 The Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. 5 Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. 6 Center for Human Systems Biology, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. 7 Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA. Correspondence and requests for materials should be addressed to C.Z. (email: [email protected])

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 1 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

etinal neurogenesis occurs through partially overlapping transgene12 from embryonic day (E)14 and post natal day (P)2 temporal windows giving rise to individual cell types, as retina, representing early and late stages of retinal progenitors R 1,2 1 result of temporal patterning and context-dependent competence, respectively . The brightest fraction of GFP-positive regulatory functions governed by transcription factors (TFs)3–6.A cells expresses the retinal progenitor-specific markers VSX2 subset of transcription factors, commonly referred to as pioneer (Chx10), mKi67 and CCND1 and was separated from the dim factors, occupy compacted chromatin regions to initiate or and GFP-negative (Supplementary Fig. 1A–I). Proliferating ret- mediate chromatin opening, enabling nucleosomal accessibility to inal progenitor cells will be hereafter referred to as RPCs or GFP the transcriptional apparatus and instructing recruitment of + and the post mitotic pool as GFP−. Flow-sorted cells were accessory factors to reprogram cellular identity. profiled by RNA-Seq and ATAC-Seq, relying on direct in vitro Even though the expression patterns have been extensively transposition of sequencing adapters into native chromatin13. characterized for most of the retinal transcription factors, the The preliminary screening of the ATAC-Seq-derived open mechanisms by which these coordinate epigenetic changes across chromatin regions identified LHX2 as top candidate transcription retinal development remain elusive, as they are limited to select factor. ChIP-Seq was then performed on LHX2 and the related cases and often perturbed by experimental biases intrinsic to cell binding sites were integrated and compared with the age-matched cultures systems and specific genetic background. RNA-Seq and ATAC-Seq profiles from control and knock-out The LIM homeodomain transcription factor LHX2 has been retinas (Fig. 1a and Supplementary Fig. 1J, K). characterized in a broad range of developmental contexts and it is required for the eye field specification and the morphogenesis of Open chromatin unbiased screening reveals top candidate TF. the optic cup through cell autonomous regulation of gene The ATAC-Seq-derived open chromatin regions were identified expression7. While Lhx2 mutants are anophthalmic, display by pairwise comparison of normalized read counts from VSX2 hypoplasia of the neocortex and severe anemia, neuroretina (Chx10)-positive cell fractions at E14 and P2 and assessed for specific loss-of-function of Lhx2 causes severe microphthalmia, reproducibility (Supplementary Fig. 2A–D). The open chromatin loss of expression of a subset of retinal progenitor cells (RPC)- regions shared between time points represent 2.4% of the genome specific genes and ectopic expression of hypothalamic genes8. and those specific to either E14 or P2 RPCs represent 0.72 and Ocular expression of Lhx2 has been observed as early as 0.80% of the genome, respectively (Fig. 1b). embryonic day e8.5, it can be detected in the retinal neuro- We next scanned open chromatin regions common to E14 and epithelium and retinal pigmented epithelium at e10 and ciliary P2 RPCs for enriched oligonucleotide motifs to identify candidate margin by e14, becoming progressively restricted to the inner transcription factor binding sites. Hierarchical clustering of nuclear layer. In the post natal retina, LHX2 is expressed in mitotic probabilistically assigned position weight matrices revealed that retinal progenitor cells that co-express VSX2 and mKi67. By post homeobox domain-containing transcription factor motifs con- natal day P7, LHX2 expression becomes restricted to differentiating stitute the prevalent cluster, representing 284 out of 373 putative Müller glia and is preserved in adult Müller glia, a subset of bipolar matches (Fig. 1c), followed by C2H2- type, GC-rich and KLF zinc and starburst amacrine cells9. Early embryonic deletion of Lhx2 finger proteins, then POU-homeodomain, MADS box and results in proliferative defects of glial committed precursors while CCAAT-Binding factors; individual motif instances were found early post natal ablation, as well as overexpression9 results in the associated to NFY factors and the insulator protein CTCF, loss of glial markers and dysmorphic apical structures. involved in chromatin looping and conformation. Among the While temporally controlled conditional knockouts indicate putatively assigned homeobox transcription factors for which a that LHX2 induces and stabilizes the retinal glial fate during first motif instance could be found, just few are expressed in RPCs post natal week of development, by sustaining the expression of (Fig. 1d) and many represented positional variations of the LHX2 the Hes5-mediated Notch signaling effectors9,10, its function does consensus (Fig. 1e and Supplementary Fig. 2E). LHX2 enrichment not seem to be confined to the differentiation of glial precursors: was inferred by binomial scoring of known position weight early embryonic ablation of Lhx2 at e12.5 results in the depletion matrices as the most represented transcription factor motif in of multipotent retinal progenitors and overproduction of retinal open chromatin regions common to early and late RPCs (Fig. 1f ganglion cells (RGCs), while overproduction of rods is observed and Supplementary Table 1), and variably represented in stage- three days later. This suggests LHX2 may contribute to the specific accessible regions (Supplementary Tables 2 and 3). LHX2 maintenance of embryonic progenitors in a proliferative and occupancy resulted into detectable footprints at both stages multipotent state, by restricting their exit from the cell cycle and (Fig. 1g). ChIP-Seq was performed for the top candidate allowing for the generation of later cell types within discrete transcription factor LHX2 in E14 and P2 mouse retina and temporal windows11. integrated with age-matched RNA-Seq and ATAC-Seq profiles To investigate the mechanistic basis by which LHX2 orches- from flow-sorted RPCs (Figs. 1h and 2a and Supplementary trates diverse developmental programs, we perform flow cyto- Fig. 3A–G). metry sorted analysis of dissociated retinal cells obtained from the neuro-retinal reporter Chx10-Cre:eGFP transgenic line12 that allows preferential retention of proliferating retinal progenitors at Dynamic allocation of LHX2-binding sites across development. early and late stages of retinal development. We query the LHX2 ChIP-Seq-binding sites are predominantly located in deriving epigenomic and transcriptional profiles obtained from intergenic and intronic regions of the genome (Supplementary natively extracted retinal progenitor cells for LHX2 representa- Fig. 4A) and enriched in promoter regions, 5′UTR, introns and tion and function over the course of retinal neurogenesis and we ncRNA at both time points (Fig. 2b). The associated genes are show LHX2 controls chromatin accessibility by modulating enriched for functions related to the control of retinal develop- expression and co-opting action of developmentally regulated ment (Supplementary Fig. 4B, C) and the binding sites exhibit transcription factors with high pioneer potential. developmental differences in chromatin accessibility (Supple- mentary Fig. 4D). The LHX2 cis-regulatory repertoire (ChIP-Seq peaks) (Supple- Results mentary Fig. 3G) was annotated based on proximity to the closest Fluorescence-activated cell sorting was adopted to isolate retinal TSS profiled by RNA-Seq from age-matched RPCs (Supplemen- progenitor cells from mice expressing the Chx10-Cre:eGFP tary Table 4) and annotated after the manually curated lists of

2 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE

a c E14 P2 Homeobox

ATAC-Seq Ctrl; Lhx2 cKO 10 nodes RNA-Seq

ChIP-Seq Homeobox 00.40.8 Distance (node/pixel)

b 10 nodes Vertebrate PWMs

h Scale 500 kb mm9 25145 68823 41419 chr12: 85,500,000 86,000,000 86,500,000 1761 E14 ATAC-Seq GFP+ r1

5416 E14 ATAC-Seq GFP+ r2 E14 Chx10-GFP+ P2 Chx10-GFP+

254 d e E14 Lhx2 ChIP-Seq r1 Vsx2 0.851 597 Lhx2 E14 Lhx2 ChIP-Seq r2

Prrx2 0.875 254 Msx2 E14 isotype control Assigned

Barx2 926 0.785 E14 GFP+ P2 ATAC-Seq GFP+ r1 Pdx1 P2 GFP+ 1486 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 P2 ATAC-Seq GFP+ r2 0 100 200 300 ′UTR UTR 5 ′ (FPKM) Correlation to Lhx2 3 936 P2 Lhx2 ChIP-Seq r1 f g Lhx2 Lhx2 1 Sox2 183 Sox4 P2 Lhx2 ChIP-Seq r2 Neurod1 0.8 Pax6 Tcf12 Sox9 43 Otx2 0.6 P2 Lhx2 ChIP-Seq r3 Klf9 Ascl1 E2f1 Known Olig2 0.4 81 Maz Klf4 P2 isotype control Tcf4 Gfy 0.2 E14

Tcf3 Average ATAC-Seq signal Gli3 Shared P2 RefSeq genes Foxp1 0 0 2 × 103 4 × 103 –100 –50 0 50 100 Vsx2 103 3 × 103 Distance to motif center –ln Adj (p-val) Mammal cons (FDR < 0.05)

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) retinal cell-type specific/enriched genes (Supplementary Data 1). Fig. 5E, F). Transcriptional dependence of a given retinal gene on Gene Ontology enrichment was computed based on the number LHX2, first inferred by ChIP-Seq peaks proximity to the closest of LHX2 target genes in term, for which at least one nearest peak TSS, was confirmed by comparison of LHX2 annotated Peaks with could be found, populating a given ontology (Fig. 2c–f). RNA-Seq from age-matched Lhx2 cKO (Supplementary Data 2.1– In order to evaluate whether LHX2 directly impacts the retinal 2.3 and Supplementary Fig. 5G) and gene ontology enrichment transcriptome, RNA-Seq transcriptional profiles were also obtained analysis was repeated (Fig. 2c–f). LHX2 ChIP-Seq peaks shared from E14 (Chx10Cre;Lhx2lox/lox retinas) (Supplementary Fig. 5A±D) between time points were preferentially associated with late-stage and P2 (Lhx2 lox/lox retinas electroporated with pCAG-Cre-GFP at RPCs genes and, to a lesser extent, with Müller glia and P0andisolatedbyFACS)Lhx2 cKO conditions (Supplementary anterodorsal hypothalamus (Fig. 2c), while peaks found only at

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 3 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

a b Promoter 0.008 5UTR Intron ncRNA 0.006 Intergenic TTS 0.004 E14 3UTR P2 Exon E14 rRNA P2 Lhx2 motifs 0.002 Lhx2 ChiP-Seq snoRNA per bp Peak pseudo miRNA 0.000 –2000–10000 1000 2000 –2 –1 0 1 2

Distance from center log2 (FC) cd * Late RPCs ****** Early RPCs *** Early RPCs Late RPCs (*) Amacrine Amacrines * RGCs RGCs E14 RNA-Seq Lhx2 cKO Muller Glia (*) (*) Bipolars (*) Bipolars Muller Glia E14 specific peaks Rods E14,P2 RNA-Seq Lhx2 cKO Cones (*) P2 RNA-Seq Lhx2 cKO Cones E14,P2 shared peaks Horizontals P2 specific peaks by shared peaks Genes IDs bound Hypothal. (*) Known retinal genes Hypothal. (*) (*) Known retinal genes

(*) Gene IDs bound by Horizontals stage-specific peaks Rods 020406080 0 20406080 Percentage Percentage

ef Early RPCs *** Late RPCs * Late RPCs Early RPCs Amacrines Amacrines Muller glia (*) Muller glia (*) RGCs RGCs Bipolars Bipolars Cones Cones P2 RNA-Seq Lhx2 cKO Hypothal. E14 RNA-Seq Lhx2 cKO Rods P2 Lhx2 peaks

E14 Lhx2 peaks Rods Horizontals Known retinal genes Gene IDs bound by Gene IDs bound by Horizontals Known retinal genes Hypothal. 0204060 0204060 Percentage Percentage

ghP2 Lhx2 E14 Lhx2 Lhx2 motif ChIP-Seq Lhx2 motif ChIP-Seq –1500 0 1500 –1500 0 1500 –1500 0 1500 –1500 0 1500 0.0014 P2 0.002 E14 ATAC-Seq ATAC-Seq E14 H3K27-Ac GFP+ P2 H3K27-Ac GFP+ 0.0004 0.0002 –1500 0 1500 –1500 0 1500 20 10 P2 RNA-Seq E14 RNA-Seq 6 10 GFP+ P2 RNA-Seq GFP+ E14 RNA-Seq 2 2 –5000 0 5000 –1500 0 1500 –5000 0 5000 –1500 0 1500 P2 0.05 E14 0.05 ATAC-Seq ATAC-Seq GFP+ 0.01 GFP+ 0.01 Distance from center Distance from center ij E14 ATAC-Seq, RNA-Seq GFP+ modules P2 ATAC-Seq, RNA-Seq GFP+ modules E14 ATAC-Seq, RNA-Seq GFP+,RNA-Seq Lhx2 cKO DEG P2 ATAC-Seq, RNA-Seq GFP+, RNA-Seq Lhx2 cKO DEG G0 and early G1 RNA polII transcription pre-Init. Notch signaling pathway DNA replication Anaphase promoting complex Regulatory RNA pathways DNA replication Oligodendrocyte differentiation RNA PolI promoter opening Lens epithelium apoptosis G2/M checkpoints Oculomotor apraxia G2/M DNA damage checkpoint Retina hypoplasia Cell cycle Axonal dystrophy 0 5 10 15 0246810

–log10 (p-val) –log10 (p-val)

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

4 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE

E14 were selectively associated with genes expressed in early-stage Examples of such LHX2 regulated genes include the locus RPCs (Fig. 2d). Conversely, late RPCs were overrepresented in the encoding Ndnf, upregulated in P2 Lhx2 cKO cells and the late- P2 ChIP-Seq dataset (Fig. 2d). stage RPC and Müller glia -enriched Car2 locus, which is More than 70% of the amacrine and retinal ganglion cell- downregulated in P2 Lhx2 cKO cells. enriched genes were induced in the E14 Lhx2 cKO retinas, The majority of LHX2 target sites are transcriptionally followed by horizontals, bipolars and late RPCs related genes, accessible: 76.66% of the E14 LHX2 cis-regulatory sites fall in while Müller glia, early RPCs and photoreceptors related genes GFP-positive open chromatin regions and 61.72% of the P2 were prevalently downregulated. Conversely, photoreceptor- ChIP-Seq peaks co-occur with P2 GFP-positive open chromatin enriched genes were prevalently induced in the P2 Lhx2 cKO regions (Supplementary Table 7). retinas while horizontals related genes were lost (Supplementary We next scanned LHX2 ChIP-Seq peaks from early and late Fig. 5H and Supplementary Data 2.4). As a fraction of LHX2 retinal stages to infer transcription factors co-occurrences. ChIP-Seq-binding sites fall in chromatin regions that are not Multiple transcription factor motif instances were found within transcriptionally active in RPCs at P2 (Supplementary Fig. 4D), a broad range of similarity to the known LHX2 consensus and LHX2 may potentially repress retinal type specific targets within exhibited differential occupancy, possibly underlying differences critical windows to prevent their ectopic expression, as suggested in affinity and/or combinatorial interaction with other co-factors by previous knock-out studies11. In total, >30% of the LHX2 (Supplementary Fig. 6A–I). peaks shared between time points and 20% of the stage-specific Novel LHX2 motifs were overrepresented within each ChIP- peaks could be assigned by proximity to differentially expressed Seq dataset (Fig. 3a, b), followed by putative co-factors; some are genes detected from the comparison between E14 and P2 RPCs, Yamanaka factors, known to regulate the developmental signaling potentially accounting for 10–14% of the differentially expressed network in mouse embryonic stem cells14 and to exert a pioneer transcriptome. Most of the differentially expressed genes, as well function in non-retinal developmental contexts15,16. The closest as genes constitutively enriched in early and late RPCs compared assignments were confirmed at each developmental stage by to the corresponding post mitotic fraction, were associated with known motifs (Supplementary Fig. 7A) with preferential stage-specific LHX2 peaks (Supplementary Table 4). Importantly, representation of homeobox/bHLH transcription factors at E14 47% of all differentially expressed RPCs associated genes show (Fig. 3c) and CTF/NF-I and forkhead factors at P2 (Fig. 3d), dependence on Lhx2 expression in at least one developmental although a wider variety was seen in the repertoire of P2 time point exhibiting variations in expression upon Lhx2 loss-of- preferential motifs (Supplementary Fig. 7B). Specifically, among function (Supplementary Fig. 5G, Supplementary Table 5, and LHX2 co-occurrent motifs, the SoxB1 related family members Supplementary Data 2). This indicates a central role of LHX2 in (SOX2)16 and MADS transcription factors were found at E14 the regulation of RPCs transcriptome. (Fig. 3a) while Kruppel like factors (KLF9/13)14, NF-I family members (NFIA/NFIB/NFIC/NFIX)17, SoxE (SOX8/9)18, and ASCL119 were found at P2 (Fig. 3b). LHX2 binds within coordinately accessible chromatin modules. In total, 4.25% and 3.48% of open chromatin regions overlapped with age-matched LHX2 ChIP-Seq peaks identified at E14 and LHX2 regulates local and global chromatin accessibility.To P2, respectively, and a higher fraction of closely co-localizing assess whether LHX2 regulates chromatin accessibility, we com- peaks was observed at E14 compared to P2 (Supplementary pared ATAC-Seq profiles from purified RPCs at E14 and P2 to Table 6). those obtained from age-matched conditional knock-outs Notably, LHX2 ChIP-Seq peaks were preferentially co-localized (Fig. 4a–f and Supplementary Fig. 8A–E). In E14 Chx10Cre; with positioned nucleosomes at E14 (Fig. 2g) but bimodal at P2 Lhx2lox/lox retinas, the read coverage and footprint depth at LHX2 (Fig. 2h), suggesting LHX2 may compete for nucleosomes center were fourfold and eightfold reduced, respectively (Fig. 4a), binding in early RPCs. Overall, 9.87% of the nucleosomes while in P2 Lhx2lox/lox retinas, a detectable, yet less pronounced interspersed open chromatin regions identified in E14 RPCs, and reduction in the read coverage and footprint depth could be 9.25% of those identified at P2 are coupled with bidirectionally observed (Fig. 4b), possibly as a result of LHX2 perdurance in P0 transcribed sites to which LHX2 binding is associated. These electroporated cells, following CRE-mediated deletion. represent 19.27% of the high-confidence LHX2 ChIP-Seq peaks at Upon Lhx2 loss-of-function variations in local accessibility E14 (0.76% co-localizing) and 17.37% at P2 (0.73% co-localizing). occur at LHX2 target sites (Fig. 4c, d upper panel, Fig. 5a–d, and The subset of LHX2 targets that flanked positioned nucleosomes Supplementary Table 7) and a correlation between local within open chromatin regions were enriched at 5′UTR and accessibility and LHX2 occupancy was observed at both time = fi promoter regions at E14 (Fig. 2g, log2FE 6.27, 5.45, respec- points (Fig. 5a, b). Speci cally, 8% of the LHX2 targets lose = tively) and P2 (Fig. 2h log2FE 6.33, 5.2). Furthermore, LHX2- accessibility in Lhx2 cKO at P2, while 40% of the LHX2 targets binding sites were coupled with bimodally distributed H3K27ac are affected at embryonic stages (Supplementary Table 7, and profiles at P2, a characteristic of active enhancers. Exons and Supplementary Figs. 9 and 10), indicating LHX2 contributes to ncRNA were also found among the enriched categories (E14 the maintenance of chromatin accessibility of its binding sites. = = log2FE 2.05, 2; P2 log2FE 2.28, 2.16, respectively), where co- The effect of Lhx2 loss-of-function on chromatin accessibility localization of transcripts with H3K27ac meta-profiles may was not merely focal, but extended genome-wide. While ATAC- potentially underlie the biogenesis of enhancer-templated eRNAs. Seq profiles from E14 whole retina (Fig. 5e) and P2 electroporated The LHX2 targets identified in the E14 RPCs open chromatin control (Fig. 5f) showed high correlation with those from each regions were involved in regulatory RNA pathways, promoter corresponding flow-sorted RPCs fraction, overall, 50% of the opening and cell cycle checkpoints (Fig. 2i) and those in P2 RPCs globally profiled open chromatin regions are lost in E14 population were enriched for functional categories related cell Lhx2 cKO retinas and 22% are lost at P2 (Fig. 4c, d lower panel, cycle progression, Notch signaling and DNA replication, as well as Fig. 4e, f, and Supplementary Table 8), suggesting Lhx2- phenotypes such as retina hypoplasia and axonal dystrophy dependent mechanisms may concur to the global loss of (Fig. 2j). Enrichment for the related ontologies was confirmed by chromatin accessibility. This may result from the regulation of comparison with Lhx2-dependent gene sets, identified by RNA- chromatin remodeling factors and pioneer transcription factors Seq from the age-matched Lhx2 cKO (Supplementary Table 6.2). that are expressed in the developing RPCs. Indeed, multiple

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 5 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

0 0.4 0.8 Distance Homeobox Homeobox

p-val 1e–1802 p-val 1e–1528 0.97 -Lhx2 0.97 -Lhx2 (GSE48068) (GSE48068)

p-val 1e–866 p-val 1e–244 0.55 -Nfic Sox2 MA0161.1 (GSE11431) (jaspar)

p-val 1e–215 p-val 1e–749 0.63-MF0005.1 0.712 -Klf7 forkhead class 0.66 -Klf4 (jaspar) (jaspar)

p-val 1e–73 p-val 1e–643 Smad2 0.631 -Tcf7 (GSE29422) MA0769.1 (jaspar)

cdE14 stage-enriched PWMs P2 stage-enriched PWMs E14 P2(GFP+) (FPKM) E14 P2 (GFP+) (FPKM)

Lhx2 (homeobox) 82.06% / 66.15% Nfi-halfsite (CTF) 49.68% / 33.39% Isl1 (homeobox) 82.86% / 69.76% Foxo1 (forkhead) 57.38% / 41.55% Pax6 (paired,homeobox) 6.94% / 3.79% Foxm1 (forkhead) 41.85% / 27.11% Meis1 (homeobox) 49.34% / 43.54% Smad3 (MADS) 68.66% / 54.81% Tgif1 (homeobox) 75.19% / 71.01% Etv1 (ETS) 38.25% / 25.37% Tgif2 (homeobox) 76.08% / 72.11% Ptf1a (bHLH) 58.07% / 44% Pbx1 (homeobox) 3.42% / 2.33% Ascl1 (bHLH) 41.67%/ 28.58% Sox3 (HMG) 58.84% / 55.41% Tbx5 (T-box) 69.42% / 55.96% Pknox1 (homeobox) 8.46% / 7.04% Crx (homeobox) 68.36% / 55.44% Pbx3 (homeobox) 9.29% / 7.96% Smad4 (MADS) 43.72% / 31.23% Sox2 (HMG) 38.11% / 36.27% Zfx (Zf) 30.24% / 19.50% Sox9 (HMG) 32.01% / 30.66% Maz (Zf) 30.80% / 20.49% known motifs Maff (bZIP) 11.16% / 10.69% Foxp1 (forkhead) 20.39% / 11.80% Rfx1 (HTH) 6.09% / 5.78% Max (bHLH) 17.15% / 9.48% Mef2a (MADS) 13.89% / 13.53% Zbtb12 (Zf) 11.32% / 5.25% Usf2 (bHLH) 6.32% / 6.27% Olig2 (bHLH) 55.38% / 43.88% Rfx5 (HTH) 8.19% / 8.13% Smad2 (MADS) 42.70% / 31.67% Sox4 (HMG) 33.23% /33.23% Klf10 (Zf) 13.76% / 7.16% Zbtb33 (Zf) 0.13% / 0.17% Gabpa (ETS) 25.36% / 16.44% Tead2 (TEA) 13.02% / 13.19% Neurod1 (bHLH) 27.75%/18.67%

0 100 200 300 400 0 100 200 300 400 –ln Adj.(p-val ) (FDR < 10–4) –ln Adj.(p-val ) (FDR < 10–4)

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

LHX2 cis-regulatory elements that exhibit variation in local Supplementary Tables 10 and 11), as result of transcriptional de- accessibility upon Lhx2 loss-of-function and, that are assigned by regulation and/or loss of steric interaction. proximity to Lhx2-dependent genes sets, encode for chromatin Some retinal progenitor-expressed transcription factors, whose remodeling factors (Supplementary Table 6.2) and transcription motifs are predicted to co-occur with LHX2 peaks, (Fig. 3c, d and factors with predicted pioneer function (Fig. 5c, d and Supplementary Fig. 7A, B) have been previously characterized for Supplementary Table 9). their pioneer function in non-retinal contexts, such as KLF (KLFf9)13, SoxB1 family members15, and NF-I (NFIA/NFIB/ Lhx2-dependent TFs pioneer chromatin opening. We next NFIC/NFIX)16. Sox2 is expressed in early and late-stage RPCs determined whether binding by known and putative pioneer and is also a transcriptional target of LHX2 (Fig. 5c) for which factors expressed in RPCs occurs (Fig. 5g) and tested whether it coordinated variations in local accessibility are found within a was altered by loss of LHX2 (Supplementary Fig. 5H and range of 500 Kb from the TSS (Fig. 6).

6 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE

ab0.15 0.3 E14 Ctrl r1 P2 Ctrl r1 0.10 E14 Ctrl r2 0.2 P2 Ctrl r2 E14 Lhx2 KO r1 P2 Lhx2 KO r1 E14 Lhx2 KO r2 P2 Lhx2 KO r2 0.05 0.1 Coverage

0.00 0.0 100 300 500 700 900 100 300 500 700 900 –900 –700 –500 –300 –100 1500 1100 1300 –900 –700 –500 –300 –100 1500 1100 1300 –1500 –1300 –1100 –1500 –1300 –1100 Distance to Lhx2 footprint Distance to Lhx2 footprint

cdE14 E14 P2 P2 ATAC-Seq ATAC-Seq ATAC-Seq ATAC-Seq ctrl Lhx2 KO ctrl Lhx2 KO

–1500 0 1500–1500 0 1500 –1500 0 1500–1500 0 1500 E14 Lhx2 peaks P2 Lhx2 peaks

E14 P2 ATAC-Seq ATAC-Seq Ctrl Ctrl

NFR10–1 NFR 10–1 ×105 ×105

efE14 ATAC-Seq E14 ATAC-Seq P2 ATAC-Seq P2 ATAC-Seq ctrl Lhx2 KO ctrl Lhx2 KO –15000 1500 –15000 1500 –15000 1500 –15000 1500 E14 ATAC-Seq P2 ATAC-Seq Ctrl Ctrl

Lost Lost –15000 1500 –15000 1500

NCR NCR

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

First, we computed the probability of occupancy for candidate SOX2 and NF-I exhibited a 93 and 84% reduction in the transcription factor binding sites in the genome to which position footprints count (Supplementary Table 10), together with KLF4, weight matrices can be assigned and where focal depletion of the ASCL1, and HES5; SOX2 and NF-I had altered mean scores ATAC-Seq reads is observed (footprints). We then estimated (Supplementary Table 11) and meta-profiles around motif centers transcription factor-specific chromatin opening indexes from (Fig. 7b, c). Variations in mean scores and footprinting meta- ATAC-Seq developmental time points15 at P2 compared to E14. profiles were also observed for the predicted pioneer KLF4/9, as Finally, we iterated the analysis in the Lhx2 cKO conditions. well as ASCL1 and HES5 (Fig. 7d–f), known neurogenic and Hence, we predicted recruitment of putative pioneer factors, by gliogenic factors respectively19,20, expressed in late-stage taking into account transcription factors expression patterns and RPCs21,22. Other transcription factors that exhibit a high pioneer querying ATAC-Seq profiles for detectable footprints (Fig. 7a–f), potential in the retina were transcriptionally targeted by LHX2 chromatin accessibility profiles and derivative nucleosome (Fig. 5c, d) and/or predicted to interact with LHX2, such as SOX9, occupancy (Fig. 7g–l) in RPCs at E14 and P2. SMAD3, ZFX, and ETV5, because variations in recruitment at In E14 Chx10Cre;Lhx2lox/lox retinas, the average cut frequency active sites (Supplementary Table 11) and in the average footprint at LHX2 motif center was reduced and in P2 Lhx2lox/lox retinas a count (Fig. 5h and Supplementary Table 12) were observed upon minor reduction in the average cut frequency (Fig. 7a) was also Lhx2 loss-of-function. observed. Overall, 96% of LHX2 footprints from E14 RPCs were Nucleosome occupancy at the aforementioned motifs was lost in E14 Lhx2 cKO retinas and 39% were lost at P2 altered in Lhx2 cKO samples (Fig. 7g–l), suggesting that LHX2 (Supplementary Tables 10 and 11). Footprints associated with may facilitate or stabilize recruitment of cognate factors on SOX2, NF-I, SOX8/9 and ASCL116–19 were statistically reduced chromatin regions, prior to or upon chromatin opening. in E14 Lhx2 cKO (Fig. 5h), indicating that the recruitment of LHX2 seemed to compete for nucleosome occupancy (Fig. 7g), these transcription factors was altered following loss of LHX2. as its meta-profiles were reminiscent of those previously

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 7 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

abE14 ATAC-Seq ctrl P2 ATAC-Seq ctrl E14 ATAC-Seq Lhx2 cKO P2 ATAC-Seq Lhx2 cKO 10 215 R 2 0.72 2 R 2 0.37 210 25 5 2 2 2 R 0.71

R 0.23 P2 ctrl E14 ctrl 0 2 20 l Stat t I = T a/2 l Stat t I = T a/2 p-val < 0.0001 ATAC-Seq coverage ATAC-Seq p-val < 0.0001 coverage ATAC-Seq –5 2–5 2 20 25 210 215 20 25 210 215 220 E14 Lhx2 ChIP-Seq occupancy P2 Lhx2 ChIP-Seq occupancy (fraction reads/10 Mio) (fraction reads/10 Mio)

c –4 d log2(FC) P2/E14 GFP+ RNA-Seq Tox2 Casz1 Klf4 –3 Foxn4 Nfib Neurod1 Hes5 Nfia Sox2 Lmo4 Meis1 –2 Pbx1 –2 (FC) Neurod4 (FC) 2 2 Rax log Notch1 log –1 Hey2 Foxp4 Ndnf Car2 P2 ATAC-Seq Lhx2 KO P2 ATAC-Seq

E14 ATAC-Seq Lhx2 KO E14 ATAC-Seq Foxa1 20 40 –5 5 –40 –20 –15 –10 10 15 log (FC) Dlx2 2 1 Nr2e3 E14 RNA-Seq Lhx2 KO Atoh8 Lhx2 dependent, 2 P2/E14 GFP+ DEG P2/E14 GFP+ DEG Lhx2 dependent, 3 non DEG

efE14 ATAC-Seq ctrl P2 ATAC-Seq ctrl E14 ATAC-Seq Lhx2 cKO P2 ATAC-Seq Lhx2 cKO 210 215 R 2 = 0.76 R 2 = 0.88 210 25 25 2 2 0 R = 0.37 R = 0.87 2 20 l Stat t I = T a/2 l Stat t I = T a/2 E14 whole retina

ATAC-Seq coverage ATAC-Seq p-val < 0.0001

ATAC-Seq coverage ATAC-Seq p-val < 0.0001 2–5 P2 ctrl electroporated 2–5 2–5 20 25 210 215 2–5 20 25 210 215 ATAC-Seq coverage ATAC-Seq coverage E14 GFP+ sorted P2 GFP+ sorted (fraction reads/10 Mio) (fraction reads/10 Mio) g h Sox8/9 *** 1 Klf4 Zfx * 0.8 Smad3 * K-S Gabpa * 0.6 p-val 0.024 Sox2 *** NF-I ***

ECDF E14 ctrl 0.4 Etv5 * E14 Lhx2 KO Putative pioneer Putative Lhx2 *** 0.2 *** P2 ctrl Significant expr.change Hes5 No significant expr.change P2 Lhx2 KO 0 Ascl1 * –20 020 406080 16 64 256 1024 4096 Pioneer values 16384 65536 ATAC-Seq footprints

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

8 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE

Lhx2 PioneerPioneer TF/TF/ Lhx2 Lhx2 cofactorcofactor ChIP-Seq ChIP-Seq ChIP-Seq identified PPioneerioneer TF/TF/ unknown function cofactorcofactor

Ctrl Lhx2 cKO RNA-Seq ATAC-Seq

Scale 100 kb mm10 100 kb mm10 200 kb mm10 chr3: 34,450,000 34,550,000 34,650,000 34,750,000 34,800,000 34,950,000 35,150,000 34,500,000 34,600,000 34,700,000 34,800,000 35,050,000

485.33 513.33 E14 ATAC-Seq GFP+ r1,r2

E14 ATAC-Seq ctrl1, ctrl2 176.68 176.68

E14 ATAC-Seq Lhx2 cKO r1, r2 89.42 148.29

49.49 102.41 E14 Lhx2 ChIP-Seq r1,r2 E14 isotype ctrl 61.46 33.42 E14 H3K27 Ac r1,r2 E14 isotype ctrl P2 ATAC-Seq GFP+ r1,r2 281.29 319.74

P2 ATAC-Seq GFP- r1,r2 366.07 159.97

P2 ATAC-Seq ctrl 1,ctrl2 253.33 184.17

P2 ATAC-Seq Lhx2 cKO r1,r2 231.33 124

P2 Lhx2 ChIP-Seq r1,r2,r3 100 108 P2 isotype ctrl 167.24 25.99 P2 H3K27 Ac P2 isotype ctrl Mir637 Sox2ot UCSC genes,RefSeq 3.296_ Sox2ot 0- Gm3143 Placental Mammal Conservation by Phylo Mir189 Sox2 Gm38509 log (FC) log (FC) 2 2 E14 ATAC-Seq Lhx2 cKO E14 ATAC-Seq Lhx2 cKO –2.13 –1.70 log2 (FC) Mir6378 E14 ATAC-Seq Lhx2 cKO log2 (FC) miRNA –1.76 E14 ATAC-Seq Lhx2 cKO –1.57 log (FC) log (FC) 2 2 E14 RNA-Seq Lhx2 cKO E14 ATAC-Seq Lhx2 cKO –1.72 Sox2ot q-val > 0.05 log (FC) Dnajc19 2 log (FC) E14 RNA-Seq Lhx2 cKO 2 E14 ATAC-Seq Lhx2 cKO –1.81 q-val > 0.05 log2 (FC) Gm38509 Sox2 Gm3143 E14 RNA-Seq Lhx2 cKO lincRNA q-val > 0.05 ncRNA log2 (FC) E14 RNA-Seq Lhx2 cKO q-val > 0.05 Sox2 -TSS log2 (FC) E14 RNA-Seq Lhx2 cKO -1.41 q-val < 0.001

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 described for the non-canonical pioneer factor NFIB17. In E14 rescued levels of nucleosome occupancy (Fig. 7k). Moreover, we and P2 Lhx2 cKO retinas, an overall increase in nucleosomes noticed a global reduction in nucleosome occupancy 500 bp coverage occurs around LHX2 motif center, consistent with a around NF-I, KLF4/9, and HES5 motif centers in E14 Lhx2 cKO reduction in chromatin accessibility observed upon Lhx2 loss-of- retinas and observed the opposite effect at post natal stages. No function. Similarly, nucleosome depletion was observed at ASCL1 difference in competition for nucleosome binding could be motif center in control conditions. Conversely, chromatin noticed directly on NF-1 and KLF4/9 motif centers following compaction occurs following Lhx2 loss-of-function, as seen from Lhx2 loss-of-function (Fig. 7i, j, l). Chromatin compaction also

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 9 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

g a Lhx2 E14 Lhx2 P2 Lhx2 E14 Lhx2 P2 0.966 1 1 0.975 0.8 0.8 0.96 0.97 0.6 0.6 0.965 0.4 0.4

signal 0.955 Ctrl Ctrl occupancy 0.96 Nucleosome

ATAC–Seq 0.2 0.2 Ctrl Lhx2 cKO Ctrl Lhx2 cKO Lhx2 cKO Lhx2 cKO 0 0 0.95 0.955 –100 –50 0 50 100 –100 –50 0 50 100 –300–200–1000 100 200 300 –300–200 –1000200300 100 Distance to motif center Distance to motif center Distance to motif center Distance to motif center

bhSox2 E14 Sox2 P2 Sox2 E14 Sox2 P2

1 1 0.965 0.975 0.8 0.97 0.6 0.96 0.5 0.965 0.4 signal 0.955 Ctrl Ctrl Ctrl 0.96 0.2 occupancy Ctrl ATAC-Seq Lhx2 cKO Lhx2 cKO Nucleosome Lhx2 cKO 0 0 0.95 0.955 Lhx2 cKO –100 –50 0 50 100 –100 –50 0 50 100 –400 –200 0 200 400 –400 –200 0 200 400 Distance to motif center Distance to motif center Distance to motif center Distance to motif center

ciNF-I E14 NF-I P2 NF-I E14 NF-I P2 1 1 0.934 0.96 0.8 Ctrl 0.8 0.932 0.955 0.6 Lhx2 cKO 0.6 0.93 0.4 0.4 0.95

signal 0.928 Ctrl

occupancy Ctrl ATAC-Seq 0.2 0.2 Ctrl 0.945

Nucleosome 0.926 Lhx2 cKO Lhx2 cKO Lhx2 cKO 0 0 0.924 0.94 –100 –50 0 50 100 –100 –50 0 50 100 –400 –200 0 200 400 –400 –200 0 200 400 Distance to motif center Distance to motif center Distance to motif center Distance to motif center j d Klf4/9 E14 Klf4/9 P2 Klf4/9 E14 Klf4/9 P2 1 1 Ctrl Ctrl 0.8 0.915 0.95 Lhx2 cKO Lhx2 cKO 0.6 0.91 0.945 0.5 0.94 0.4 0.905

signal 0.935 0.9 Ctrl Ctrl 0.2 occupancy ATAC-Seq 0.93 Nucleosome Lhx2 cKO Lhx2 cKO 0 0 0.895 0.925 –100 –50 0 50 100 –100 –50 0 50 100 –400 –200 0 200 400 –300–200–1000200300 100 Distance to motif center Distance to motif center Distance to motif center Distance to motif center

ekAscl1 E14 Ascl1 P2 Ascl1 E14 Ascl1 P2 1 1 0.915 0.955 Ctrl 0.8 Ctrl 0.95 Lhx2 cKO Lhx2 cKO 0.91 0.6 0.945 0.5 0.4 0.905 0.94 signal 0.2 occupancy Ctrl 0.935 Ctrl Nucleosome ATAC-Seq 0 0 0.9 Lhx2 cKO 0.93 Lhx2 cKO –100 –50 0 50 100 –100 –50 0 50 100 –400 –200 0 200 400 –400 –200 0 200 400 Distance to motif center Distance to motif center Distance to motif center Distance to motif center

flHes5 E14 Hes5 P2 Hes5 E14 Hes5 P2 1 1 Ctrl Ctrl 0.908 0.8 Lhx2 cKO 0.8 Lhx2 cKO 0.95 0.906 0.6 0.6 0.904 0.945 0.94 0.4 0.4 0.902 signal 0.9 0.935

ATAC-Seq 0.2 0.2 Ctrl occupancy 0.898 0.93 Ctrl Nucleosome Lhx2 cKO 0 0 0.896 0.925 Lhx2 cKO –100 –50 0 50 100 –100 –50 0 50 100 –400 –200 0 200 400 –400 –200 0 200 400 Distance to motif center Distance to motif center Distance to motif center Distance to motif center

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 occurs around SOX2 motif center in the Lhx2 cKO (Fig. 7h), with Discussion dips in nucleosomes distribution flanking SOX2 motif centers, In this study, we report a central role for LHX2 in the global suggesting SOX2 may preferentially be recruited on nucleosome control of chromatin accessibility in retinal progenitor cells, we free regions at E14, potentially acting as a settler factor whose identify possible mechanisms by which LHX2 may mediate ret- binding occurs in proximity to pre-existing open chromatin inal transcriptional networks across development and we provide states, as established by pioneer factors15. what is, to our knowledge, the first systematic query of candidate

10 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE pioneer factors that may contribute to the establishment and Secondary regulation of candidate pioneer factors may be maintenance of open chromatin states in the native context of the achieved transcriptionally, by Lhx2-coordinated variations in local developing murine retina. accessibility, paired with epistatic variations in gene expression, as First, an unbiased search for transcription factor motifs enri- is the case of Sox2, and/or by steric interaction of predicted co- ched in accessible chromatin regions from early and late retinal factors, both of which would presumably result in altered tran- progenitor cells identified LHX2 as the most represented cano- scription factors recruitment and nucleosome remodeling. nical transcription factor. LHX2 cis-regulatory repertoire changes Transcriptional dependence of candidate pioneer factors on over time, where the selective occupancy of retinal targets is LHX2 can be exemplified with the following cases, wherein var- reflected into gene variations upon Lhx2 loss-of-function. At iations in local accessibility at their closest cis-regulatory sites embryonic stages, LHX2 preferentially targets known early retinal bound by LHX2 co-occur with variations in expression levels progenitor-related genes while, at post natal stages, LHX2 targets upon Lhx2 loss-of-function and/or across development. These known late progenitor-related genes. Divergence of LHX2 motif genes include Hes5, Neurod4, Meis1, Rax, Dlx2, and Neurod1. instances from the consensus results in developmentally regulated Additional transcription factors, whose encoding genes are occupancy and may possibly underlie preferential targeting of transcriptionally targeted by LHX2, exhibit high predicted pio- retinal type specific targets in RPCs. neer potential such as ATF1, ATOH7, CREB1, DEAF1, EGR1, Individual overexpression of the Sry-related high-mobility ETS2, ETV5, HEY1, HEYL, MYCN, NR2F2, PROX1,REST, SP4, group (HMG) transcription factors SOX2, SOX8, and SOX9 in TEAD1, VAX2, VSX2, ZBTB33, and ZFX. the post natal retina has been shown to promote amacrine cells The E14 retinal transcriptome shows greater Lhx2-dependence formation in Lhx2 cKO mice, but not in wild-type animals. than would be expected based on peaks association by proximity Likewise, electroporation of the Notch intracellular domain to the closest promoter region. This may partially reflect the N1ICD in Lhx2 cKO mice inhibits RPCs maintenance while preponderant distribution of LHX2-binding sites at intergenic promoting the expression of gliogenic markers, yet the phenotype regions, exerting distal regulation of target genes. On the other was not observed electroporating the wild type retina10, sug- hand, the extent to which Lhx2 loss-of-function impacts retinal gesting LHX2 may deflect default developmental pathways by gene expression in early RPCs, may depend on collateral tran- combinatorial interaction with co-factors, such as SOX9, scriptional networks evoked by LHX2, whereby the observed restraining their ability to give rise to cellular diversity. variations in gene expression following E14 Lhx2 loss-of-function While LHX2 has a prominent role in the regulation of the would be amplified by Lhx2-dependent transcription factors. transcriptome of early progenitors, a more moderate effect can be Among the many transcription factors predicted to co-occur seen at post natal stages. The lower extent to which Lhx2 loss-of- with LHX2 ChIP-Seq peaks, some had predicted pioneer potential function affects gene expression at post natal stages may be, in for which we observed variations in footprints counts upon Lhx2 part, result of LHX2 residual expression in neonatal electro- loss-of-function at E14. Affinity might be affected at both stages porated cells, where the perdurance of the GFP reporter in pro- for many of the predicted interactors, as shown in their meta- genitors exiting the cell cycle after transfection may contribute to profiles distribution and mean scores from motif centers. More- the profiles analyzed. However, differential motif enrichment over, nucleosomes depletion was observed at motif center of NF- analysis indicates the post natal retinal transcriptome relies on a I, KLF4/9, ASCL1, and HES5, while variations in nucleosomes broader set of transcriptional networks, which may compensate coverage were detected upon Lhx2 loss-of-function, suggesting for Lhx2 ablation, limiting the magnitude of the transcriptional LHX2 may be necessary to stabilize candidate pioneer co-factors variations. This would be particularly relevant in the post natal prior to or during recruitment to target sites. retina as opposed to the embryonic knock-out, as the endogenous While several of the predicted pioneer transcription factors expression of LHX2 that precedes its experimental ablation at P0, affected by LHX2, such as HES5, SOX8/9, and NFIA, play hence a phasic drop in protein expression, may enable a feed- important roles in the control of retinal gliogenesis17–20, a process forward mechanism. that is dependent on LHX210, Ascl1 plays an essential role in Notably, immunostaining for P27 Kip1 and GLUL in Müller conferring neurogenic competence to late-stage RPCs21,22. Glia is reduced both by overexpression23 and downregulation of Among additional LHX2 targets, Pax6 and Sox2 instead show Lhx2 in the neonatal retina9, suggesting a balanced amount of broad temporal expression profiles25,26 and exert distinct func- LHX2 protein may be necessary to mediate the gliogenic process tions across development27–31. Some of the known and candidate and a graded transcriptional response of target genes. Tran- pioneer factors that exhibit developmentally regulated scriptional de-regulation of P27 Kip1 and Glul may ultimately nucleosome-binding signatures, like NF-I, SOX2, KLF4, may result from threshold responses that override the feed-forward preserve chromatin in a highly compacted state at embryonic regulatory loop controlling Lhx2 expression. states, while mediating chromatin unfolding later in development. LHX2 regulatory role likely relies on remodeling of nucleosomes units, variably coupled with active enhancers marks, flanked by Methods active promoter and non-coding regions within modules of coor- Experimental design. The experimental pipeline involves the generation and dinately accessible chromatin. As a result, Lhx2 loss-of-function integration of high-throughput sequencing libraries from purified fractions of affects genes involved in the cell cycle checkpoints, DNA replication, murine retinal progenitor cells (RPCs). Fluorescence-activated cell sorting (FACS) was adopted to isolate RPCs from mice expressing the RPC-specific Chx10-Cre: retinal disease phenotypes and the Notch signaling pathway, con- eGFP transgene12 from embryonic day (E)14 and post natal day (P)2 retina, 9,10 sistent with previous findings . LHX2 has been previously shown representing early and late stages of RPC competence, respectively. Flow-sorted to act as a regulator of genes that control cortical neuronal subtype cells were profiled by RNA-Seq and ATAC-Seq, relying on direct in vitro trans- identity by interaction with members of the nucleosome and histone position of sequencing adapters into native chromatin. ChIP-Seq was then performed on a select transcription factor candidate, identified in the preliminary deacetylase remodeling complex, although the mechanistic basis screening of the ATAC-Seq-derived open chromatin regions. ChIP-Seq-binding 24 behind its regulatory role had not been investigated . sites for the candidate factor were ultimately integrated and compared with the age- LHX2 affects both local and global chromatin accessibility by matched RNA-Seq and ATAC-Seq profiles from control and knock-out retinas. For competition for nucleosome occupancy, whereby secondary reg- each experimental condition involving a point-source factor and broad regions (ChIP-Seq and ATAC-Seq, respectively) a minimum of 20 million uniquely ulation of chromatin remodeling factors and pioneer factors is mappable reads or ≥10 million uniquely mappable reads for each biological likely to concur to the widespread decrease in accessibility, replicate were collected, according to the ENCODE’s guidelines. For point-source observed upon Lhx2 loss-of-function. datasets, non-redundant mapped reads were retained for downstream analysis.

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 11 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9

Animals. CD-1 mice of either sex were euthanized at embryonic day 14 (E14) and per position = 1, normalization to input, or isotype control) and subject to post natal day 2 (P2) according to Johns Hopkins IACUC-approved protocols. comparison with ATAC-Seq peaks (hypergeometric test, ln p-val) where co- Timed pregnant CD-1 mice were obtained from Charles River Laboratories. occurrence was defined by physical overlap, allowing a max distance of 20 bp from Chx10-Cre:GFP mice12 were purchased from the Jackson Laboratories. Retinas peak summits over 3000 bp for confocality. For direct comparison of ChIP-Seq and were freshly dissected, incubated in a suspension of papain and DNAse for 30 min ATAC-Seq data, high-confidence peaks with the highest differential in accessibility at 37 °C, inactivated with bovine serum albumin, resuspended in equilibrated between nucleosomal units and flanking nucleosomes-free, transposon-accessible Earle’s balanced salt solution and subject to fluorescence-activated cell sorting regions, and a minimum distance of 300 bp were identified from two ATAC-Seq (98–99% purity) where viability was assessed by propidium iodide exclusion. Cell replicates. fractions were collected on poly-D-lysine coated slides, fixed in 4% paraformalde- Genes that showed enriched or specific expression in retinal progenitors were hyde for 10 min, permeabilized in TritonX-100 and stained for CHX10 (Cat.# identified based on the RNA-Seq data generated in this study. Genes that showed X1179P, Exalpha), GFP (Cat.# 600-101-215, Rockland), mKi67 (Cat.# RM-9106- cell-type enriched or specific expression in adult retina were extracted from other S1,Thermo Scientific), or CCND1 (Cat.# SC-450, Santa Cruz). The brightest published studies using RNA-Seq, microarray-based analysis of isolated cells, fraction, differing of fourfold mean intensity for GFP relative to the dim fraction, in situ hybridization or immunohistochemistry of intact retina. Association of was always retained for subsequent processing and hereafter referred to as GFP- ChIP-Seq peaks with genes expressed in specific retinal cell types was evaluated by positive, RPC-enriched fraction. Lhx2 conditional embryonic knockouts were Fisher’s exact test and corrected for multiple hypothesis testing obtained by crossing Chx10-Cre:GFP with Lhx2lox/lox mice, and harvesting at E1432. (Bonferroni–Hochberg). Post natal Lhx2 knockouts were generated by electroporation of pCAG-Cre-GFP construct into P0.5 wild-type CD-1 animals or Lhx2lox/lox animals. Retinas were Motif enrichment analysis. Hierarchical clustering of probabilistically assigned harvested at P2, dissociated, and GFP-positive electroporated cells were isolated by motifs (Jaspar 2016 non-redundant core vertebrates)41 was done with the following FACS. Overall electroporation efficiency was 2–3%. parameters (linkage = average; similarity threshold cor = 0.6, ncor = 0.4, rl = 5, where ncor is Pearson’s correlation (cor) by relative alignment length (rl) divided ATAC-Seq, RNA-Seq, and ChIP-Seq analysis. Chromatin derived from flow- by the overall alignment length. (e-val = p-val × enriched oligos)42,43. sorted Chx10-Cre-GFP + ve and GFP-ve retinal fractions was processed as pre- For ChIP-Seq analysis, binomial probability analysis of regulatory transcription viously described13. Briefly, chromatin was extracted and processed for Tn5- factor motifs was calculated in the overall set of high-confidence peaks (minimum mediated tagmentation and adapter incorporation, according to the Manufacturer’s of two normalized replicates). The regions size was empirically determined and protocol (Nextera DNA sample preparation kit, Illumina®) at 37 °C for 30 min. motifs were found by cumulative binomial distribution of known position weight Reduced-cycle amplification was carried out in presence of compatibly matrices assuming a random representation of decamers. Motif finding was indexed sequencing adapters. The quality of the libraries was assessed by fluoro- performed on the repeats masked mm9 murine genome, optimized for the top metric DNA incorporation-based assay (Thermo Fisher ScientificTM) and auto- enriched 20 putative motifs, randomized and repeated twice to estimate FDR44. mated capillary electrophoresis (Agilent Technologies, Inc.) and up to four samples per lane were pooled and run as 50 bp paired ends on a HiSeq2500 Illumina Footprint analysis. FIMO45 was used to scan the genome for candidate binding sequencer. sites of different transcription factors. We then used BPAC to identify the actual RNA was processed with Qiagen RNAeasy Mini kit, subject to DNAse binding sites among these candidate sites40. The number of active binding sites was digestion, and samples with a minimum RNA integrity number (RIN) of seven analyzed at E14 and P2. Genome-wide changes in footprint counts and nucleosome were further processed for sequencing. Libraries were prepared using Illumina occupancy for individual transcription factor motifs were estimated for all candi- ’ TruSeq RNA Sample kit (Illumina, San Diego, CA) following the manufacturer s date binding sites at E14 and P2. Footprints scores were calculated for LHX2, fl recommended procedure. Brie y, total RNA was denatured at 65 °C for 5 min, KLF9/13, NF-I (NFIA/B/X), and SOX2. T-test on mean footprints scores dis- fi cooled on ice, puri ed and incubated at 80 °C for 2 min. The eluted mRNA was tributed 200 bp from motif centers were calculated for the paired control and Lhx2 fragmented at 94 °C for 8 min and converted to double stranded cDNA, end cKO conditions. repaired, A-tailed, and ligated with indexed adaptors and run on a MiSeq Illumina sequencer. The quality of the libraries was assessed by fluorimetric RNA incorporation-based assay (Thermo Fisher ScientificTM) and automated capillary Statistical analysis. Co-occurrence statistics for point-source and broad regions of electrophoresis (Agilent Technologies, Inc). interest was computed by hypergeometric test with a default minimum overlap of fi ChIP was performed as described previously33. Whole dissected retinas were 1 bp, unless otherwise speci ed. Coverage was adopted as reproducibility metrics dissociated in a collagenase I suspension, cross-linked in 1% formaldehyde for 15 for ChIP-Seq and ATAC-Seq experimental replicates (fraction of reads/10 Million min, and quenched in 125 mM glycine. The extracted nuclei were sheared to non-redundant uniquely mappable reads) and FPKM for RNA-Seq, where corre- ’ ’ fi produce 100–500 bp fragments by means of probe sonication. Chromatin was lation was reported by Pearson s or Spearman s coef cient. Binomial probability immunoprecipitated with anti-Lhx2 (Cat.# SC-19344, Santa Cruz Biotechnology), analysis of regulatory transcription factor motifs was applied genome-wide to rabbit anti-H3K27Ac (Cat.# ab4729, Abcam), or the corresponding isotype identify enriched position weight matrices (PWMs) and clustered by average ’ controls (Abcam); retained on agarose beads (Invitrogen), washed and purified by linkage. Fisher s exact test was adopted to compute gene ontology enrichment and – organic extraction. Amplicons corresponding to the targeted transcription factors corrected for multiple hypothesis (Bonferroni Hochberg) and RNA-Seq-derived fl and other candidate cis-regulatory regions that exhibit LHX2-binding sites and gene sets from ow-sorted retinal cell fractions and control vs. experimental syntenic unbound regions had been verified by SYBR qRT-PCR (Agilent conditions (Lhx2 cKO) with q-val (FDR) < 0.05 were retained for downstream fi Technologies) as we previously reported10. Amplicons exhibiting enrichment of analysis, unless otherwise speci ed. Two-tailed t-test was adopted to compare the anti-LHX2 fractions over isotype control include Six6, Fgf15, Rax, Vsx2, Dct, Car2, average footprints counts for candidate pioneer factors in control and Lhx2 knock- NeuroD1, Neurod4, and Neurog2. Libraries were processed according to the out conditions. manufacturer’s protocol (TruSeq Nano DNA Library Prep Kit). The quality of the libraries was assessed by fluorometric DNA incorporation-based assay (Thermo Reporting Summary. Further information on experimental design is available in Fisher) and automated capillary electrophoresis (Agilent Technologies) and the Nature Research Reporting Summary linked to this article. libraries (100–150 bp single read, paired ends) were run on a HiSeq2500 Illumina sequencer. Data availability Auxiliary data, files, and tables have been deposited in the GEO repository, Accession Peak calling and retinal gene ontology analysis. RNA-Seq reads were aligned to Number GSE99818. Metadata descriptions for the GEO repository are available in 34 the mouse transcriptome (mm9 UCSC build) using Tophat2 , and differentially Supplementary Data 4. All data needed to evaluate the conclusions in the paper are fi 35 = expressed genes (DEGs) were identi ed by Cuffdiffs , imposing a cutoff q-val present in the paper and/or the Supplementary Materials. Additional data are available 0.05 for pairwise comparison. from the authors upon request. Bowtie2 was used for ChIP-Seq and ATAC-Seq read alignment on the mouse genome (mm9)36. Uniquely mappable reads from ChIP-Seq were retained for peak calling by MACSs37 (band width = 300, mfold = 5, 50, d = 200, max tags per Received: 1 September 2018 Accepted: 11 March 2019 position = 1, min false discovery rate (FDR) q-val cutoff = 1E-02, λ = 1000–10,000 bp unless indicated otherwise). Open chromatin regions were identified in ATAC-Seq data using MACS237. Correlations between open chromatin states identified by pairwise comparison of normalized read counts in P2 and E14 GFP-positive flow-sorted retinal progenitor cells was computed with Jaccard38. The Jaccard index was estimated as the number References of peaks that overlap between two peak files, divided by the union of the two files. 1. Livesey, F. J. & Cepko, C. L. Vertebrate neural cell-fate determination: lessons Footprints and nucleosomes were identified as described previously15,39,40. from the retina. Nat. Rev. Neurosci. 2, 109–118 (2001). Annotation was performed by proximity to the closest transcription start site. 2. Kohwi, M. & Doe, C. Q. Temporal fate specification and neural progenitor High-confidence ChIP-Seq peaks were identified from at least two experimental competence during development. Nat. Rev. Neurosci. 14, 823–838 (2013). replicates (Poisson p-val threshold = 0.0001, min FE = 4, FDR = 0.001, max tags

12 COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0375-9 ARTICLE

3. Elliott, J., Jolicoeur, C., Ramamurthy, V. & Cayouette, M. Ikaros confers early 33. Shang, Y., Hu, X., DiRenzo, J., Lazar, M. A. & Brown, M. Cofactor dynamics temporal competence to mouse retinal progenitor cells. Neuron 60,26–39 (2008). and sufficiency in estrogen receptor-regulated transcription. Cell 103, 843–852 4. Mattar, P., Ericson, J., Blackshaw, S. & Cayouette, M. A conserved regulatory (2000). logic controls temporal identity in mouse neural progenitors. Neuron 85, 34. Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence 497–504 (2015). of insertions, deletions and gene fusions. Genome Biol. 14, R36 (2013). 5. Yang,Z.,Ding,K.,Pan,L.,Deng,M.&Gan,L.Math5determinesthecompetence 35. Trapnell, C. et al. Differential gene and transcript expression analysis of RNA- state of retinal ganglion cell progenitors. Dev. Biol. 264,240–254 (2003). seq experiments with TopHat and Cufflinks. Nat. Protoc. 7, 562–578 6. Li, S. et al. Foxn4 controls the genesis of amacrine and horizontal cells by (2012). retinal progenitors. Neuron 43, 795–807 (2004). 36. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. 7. Porter, F. D. et al. Lhx2, a LIM homeobox gene, is required for eye, forebrain, Methods 9, 357–359 (2012). and definitive erythrocyte development. Development 124, 2935–2944 (1997). 37. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, 8. Roy, A. et al. LHX2 is necessary for the maintenance of optic identity and for R137 (2008). the progression of optic morphogenesis. J. Neurosci. 33, 6877–6884 (2013). 38. Rau, A., Gallopin, M., Celeux, G. & Jaffrezic, F. Data-based filtering for 9. De Melo, J. et al. Lhx2 is an essential factor for retinal gliogenesis and Notch replicated high-throughput transcriptome sequencing experiments. signaling. J. Neurosci. 36, 2391–2405 (2016). Bioinformatics 29, 2146–2152 (2013). 10. De Melo, J., Clark, B. S. & Blackshaw, S. Multiple intrinsic factors act in 39. Schep, A. N. et al. Structured nucleosome fingerprints enable high-resolution concert with Lhx2 to direct retinal gliogenesis. Sci. Rep. 6, 32757 (2016). mapping of chromatin architecture within regulatory regions. Genome Res. 25, 11. Gordon, P. J. et al. Lhx2 balances progenitor maintenance with neurogenic 1757–1770 (2015). output and promotes competence state progression in the developing retina. J. 40. Liu, S. et al. BPAC: A universal model for prediction of tran-scription factor Neurosci. 33, 12197–12207 (2013). binding sites based on chromatin accessibility. BMC Bioinform. 18, 355 12. Rowan, S. & Cepko, C. L. Genetic analysis of the homeodomain transcription (2017). factor Chx10 in the retina using a novel multifunctional BAC transgenic 41. Mathelier, A. et al. JASPAR 2016: a major expansion and update of the open- mouse reporter. Dev. Biol. 271, 388–402 (2004). access database of transcription factor binding profiles. Nucl. Acid. Res 44, 13. Buenrostro, J. D., Giresi, P. G., Zaba, L. C., Chang, H. Y. & Greenleaf, W. J. D110–D115 (2016). Transposition of native chromatin for fast and sensitive epigenomic profiling 42. Van Helden, J. Regulatory sequence analysis tools. Nucl. Acids Res 31, of open chromatin, DNA-binding proteins and nucleosome position. Nat. 3593–3596 (2003). Methods 10, 1213–1218 (2013). 43. Castro-Mondragon, J. A., Jaeger, S., Thieffry, D., Thomas-Chollier, M. & Van 14. Liu, X. et al. Yamanaka factors critically regulate the developmental signaling Helden, J. RSAT matrix-clustering: dynamic exploration and redundancy network in mouse embryonic stem cells. Cell Res. 18, 1177–1189 (2008). reduction of transcription factor binding motif collections. Nucl. Acid. Res. 45, 15. Sherwood, R. I. et al. Discovery of directional and nondirectional pioneer 13 e119 (2017). transcription factors by modeling DNase profile magnitude and shape. Nat. 44. Heinz, S. et al. Simple combinations of lineage-determining transcription Biotechnol. 32, 171–178 (2014). factors prime cis-regulatory elements required for macrophage and B cell 16. Lodato, M. A. et al. SOX2 co-occupies distal enhancer elements with distinct identities. Mol. Cell 38, 576–589 (2010). POU factors in ESCs and NPCs to specify cell state. PLoS Genet. 9, e1003288 45. Grant, C. E., Bailey, T. L. & Noble, W. S. FIMO: scanning for occurrences of a (2013). given motif. Bioinformatics 27, 1017–1018 (2011). 17. Denny, S. K. et al. Nfib promotes metastasis through a widespread increase in chromatin accessibility. Cell 166, 328–342 (2016). 18. Muto, A., Iida, A., Satoh, S. & Watanabe, S. The group E Sox genes Sox8 and Acknowledgements Sox9 are regulated by Notch signaling and are required for Muller glial cell We thank Hongkai Ji for help with statistical analysis, Anand Venkataraman for com- development in mouse retina. Exp. Eye Res. 89, 549–558 (2009). ments on the manuscript, Hong Wang for help with dissections, the JHMI Deep 19. Brzezinski, J. A. T., Kim, E. J., Johnson, J. E. & Reh, T. A. Ascl1 expression sequencing and microarray core, the JHU GRCF high-throughput sequencing center, the defines a subpopulation of lineage-restricted progenitors in the mammalian JHSPH Flow Cytometry and Cell Sorting Core Facility. This work was supported by the retina. Development 138, 3519–3531 (2011). National Institutes of Health (R01EY020560 to Seth Blackshaw, R01EY024580 and 20. Hojo, M. et al. Glial cell fate specification modulated by the bHLH gene Hes5 R01EY023188 to Jiang Qian) in mouse retina. Development 127, 2515–2522 (2000). 21. Tomita, K., Nakanishi, S., Guillemot, F. & Kageyama, R. Mash1 promotes Author contributions neuronal differentiation in the retina. Genes Cells 1, 765–774 (1996). C.Z. carried out experiments and data analysis. S.L, J.W., and J.Q carried out analysis. 22. Hufnagel, R. B., Le, T. T., Riesenberg, A. L. & Brown, N. L. Neurog2 controls C.Z., S.B., and J.Q. conceived the study and wrote the paper. the leading edge of neurogenesis in the mammalian retina. Dev. Biol. 340, 490–503 (2010). 23. De Melo, J. et al. Ldb1- and Rnf12-dependent regulation of Lhx2 controls the Additional information relative balance between neurogenesis and gliogenesis in the retina. Supplementary information accompanies this paper at https://doi.org/10.1038/s42003- Development 145,1–10 (2018). 019-0375-9. 24. Muralidharan, B. et al. LHX2 interacts with the NuRD complex and regulates cortical neuron subtype determinants Fezf2 and Sox11. J. Neurosci. 37, Competing interests: The authors declare no competing interests. 194–203 (2017). 25. Molyneaux, B. J., Arlotta, P., Menezes, J. R. & Macklis, J. D. Neuronal subtype Reprints and permission information is available online at http://npg.nature.com/ specification in the cerebral cortex. Nat. Rev. Neurosci. 8, 427–437 (2007). reprintsandpermissions/ 26. Bassett, E. A. & Wallace, V. A. Cell fate determination in the vertebrate retina. Trends Neurosci. 35, 567–573 (2012). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 27. Surzenko, A., Crowl, T., Bachleda, A., Langer, L. & Pevny, L. SOX2 maintains published maps and institutional affiliations. the quiescent progenitor cell state of postnatal retinal Muller glia. Development 140, 1445–1456 (2013). 28. Wang, Z., Yasugi, S. & Ishii, Y. Chx10 functions as a regulator of molecular Open Access This article is licensed under a Creative Commons pathways controlling the regional identity in the primordial retina. Dev. Biol. Attribution 4.0 International License, which permits use, sharing, 413, 104–111 (2016). adaptation, distribution and reproduction in any medium or format, as long as you give 29. Sakurai, K. & Osumi, N. The neurogenesis-controlling factor, Pax6, inhibits appropriate credit to the original author(s) and the source, provide a link to the Creative proliferation and promotes maturation in murine astrocytes. J. Neurosci. 28, Commons license, and indicate if changes were made. The images or other third party 4604–4612 (2008). material in this article are included in the article’s Creative Commons license, unless 30. Warren, N. et al. The transcription factor, Pax6, is required for cell indicated otherwise in a credit line to the material. If material is not included in the proliferation and differentiation in the developing cerebral cortex. Cereb. article’s Creative Commons license and your intended use is not permitted by statutory – Cortex 9, 627 635 (1999). regulation or exceeds the permitted use, you will need to obtain permission directly from 31. Tetreault, N., Champagne, M. P. & Bernier, G. The LIM homeodomain the copyright holder. To view a copy of this license, visit http://creativecommons.org/ fi transcription factor Lhx2 is required to specify the retina eld and synergistically licenses/by/4.0/. cooperates with Pax6 for Six6 transactivation. Dev. Biol. 327,541–550 (2009). 32. Mangale, V. S. et al. Lhx2 selector activity specifies cortical identity and suppresses hippocampal organizer fate. Science 319, 304–309 (2008). © The Author(s) 2019

COMMUNICATIONS BIOLOGY | (2019) 2:142 | https://doi.org/10.1038/s42003-019-0375-9 | www.nature.com/commsbio 13