The Chd4 chromatin remodeller controls temporal programs during retinal neurogenesis

During retinal development, neurons and glia are produced in a highly stereotyped and evolutionarily conserved birth order. Landmark studies have demonstrated that the sequence of cell production is temporally encoded. Transcription factors and heterochromatic determinants have been elsewhere implicated in the temporal regulation of progenitor competence states, but the underlying mechanisms are not well understood. Since chromatin remodelling complexes interact with both transcription factors and heterochromatic complexes such as polycomb, we focused on the nucleosome remodeller Chd4. We generated conditional knockouts (cKOs) of Chd4 – a key chromatin remodelling enzyme in neural progenitors. Chd4 cKOs exhibited a marked expansion in early-born retinal ganglion cells. Postnatally, rod photoreceptors were drastically underproduced. As a result, progenitors failed to be exhausted during late phases of development, leading to a striking increase in the production of Müller glia. Clonal retroviral lineaging and histological marker analyses suggest that these effects were independent of alterations in cell death or proliferation. Next, we examined the effect of Chd4 on the genome and transcriptome, focusing on retinal progenitors at P0/P1. Multi-seq single-cell transcriptomics demonstrated that deletion of Chd4 created divergent gene expression profiles and developmental trajectories. ATAC-seq experiments performed on sorted P1 retinal progenitors revealed that chromatin accessibility was significantly increased at ~10 000 enhancer elements and ~4000 genes in the Chd4 cKO, demonstrating that Chd4 is required to repress gene expression and genome accessibility. Finally, to determine whether Chd4 regulates retinal cell-type production by altering competence windows, we performed EdU birthdating. These experiments revealed that cell fates were altered without affecting competence windows. Thus, despite a very strong shift in the production of early-born and late-born cell types, our data suggest that chromatin remodelling does not directly regulate competence windows. Overall design: Cut&run-seq was performed using the Cutana cut&run kit (14-0500; Epicypher), following the manufacturer's instructions. Briefly, P0 retinas were dissected. Cells were dissociated with mechanical trituration in order to retain the glycoproteins necessary for adhering to the concavanalin beads. 500 000 cells were used in each reaction. Cut&run seq was performed using antibodies against Chd4 (Abcam ab72418) or Mbd3 (Fortis A302-529A). Non-specific IgG was used as a negative control. Libraries were prepared using the NEBNext® UltraTM II Library Prep Kit (New England Biolabs). Paired-end 150 sequencing was performed using the NextSeq 500 platform (Illumina) to a read-depth of ~20-35 million reads per sample. For ATAC-seq, we followed Buenrostro et al. [54]. Briefly, 75,000 cells were sorted from Chx10-Cre-YFP animals using YFP to purify RPCs. Cells were lysed in cold lysis buffer (10 mM Tris-Cl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, and 0.1% IGEPAL CA-630). Lysed nuclei were tagmented using 6.5 µl of TDE1 transposase from the Nextera DNA Flex Library kit (Illumina). Samples were purified using Zymo-Spin IC columns (Zymo), and libraries constructed according to the Nextera workflow. Libraries were cleaned up using the AMPure XP kit (Beckman Coulter). PE 150 sequencing was performed using the NextSeq 500 platform. Bioinformatic analysis was performed using Galaxy. FASTQ files were processed via Fastq Groomer and Trimmomatic, and then mapped to the mm9 genome using Bowtie2. Peak calling was performed with Macs2.