A gene regulatory network explains RET-EDNRB epistasis in Hirschsprung disease

Abstract

Disruptions in gene regulatory networks (GRNs), driven by multiple deleterious variants, potentially underlie complex traits and diseases. Hirschsprung disease (HSCR), a multifactorial disorder of enteric nervous system (ENS) development, is associated with at least 24 genes and seven chromosomal loci, with RET and EDNRB as its major genes. We previously demonstrated that RET transcription in the ENS is controlled by an extensive GRN involving the transcription factors (TFs) RARB, GATA2 and SOX10 and other HSCR genes. We now demonstrate, using human and mouse cellular and animal models, that EDNRB is transcriptionally regulated in the ENS by GATA2, SOX10 and NKX2.5 TFs. Significantly, RET and EDNRB expression is regulated by their shared use of GATA2 and SOX10, and in turn, these TFs are controlled by EDNRB and RET in a dose-dependent manner. This study expands the ENS development GRN to include both RET and EDNRB, uncovers the mechanistic basis for RET-EDNRB epistasis and emphasizes how functionally different genes associated with a complex disorder can be united through a common GRN.

Figure 1

Identifying putative enhancers at the EDNRB locus. (A) A 350 kb genomic segment comprising the core TAD in eight human cell lines containing 11 putative CREs defined using ENCODE DHS and H3K4me1 or H3K27ac marks from a 108-day-old human fetal intestine. The locations of all ENCODE TF ChIP-seq sites and the EDNRB gene are indicated. (B) In vitro luciferase assays in SK-N-SH cells show significant enhancer activity in 5 of the 11 putative CREs (marked in green) compared to a promoter-only control. Error bars represent SEs of the mean (^*P < 0.01, ^**P < 0.001) for three independent biological replicates for each luciferase assay.

Figure 2

Identification of the cognate TFs bound to EDNRB enhancers. (A) Genomic map of the EDNRB locus with location for element E10 and ChIP-qPCR assays using a SOX10 antibody show enrichment of binding as compared to the background in SK-N-SH cells. The specificity of this binding is shown by siRNA knockdown of SOX10 with a concomitant reduction in the ChIP-qPCR signal. (B) Analogous assays for two putative GATA2 enhancers (E1 and E9) and ChIP-qPCR assays using a GATA2 antibody show enrichment of binding to E9 but not E1. (C) ChIP-qPCR assays on enhancer region E7 demonstrate specific binding of NKX2.5 but not NF-κB. Error bars represent SEs of the mean (^*P < 0.01, ^**P < 0.001) for three independent biological replicates for each ChIP.

Figure 3

TF-mediated in vitro control of gene expression. (A) siRNA-mediated knockdown of SOX10, GATA2 and NKX2.5 but not RARB in SK-N-SH cells downregulates EDNRB transcription. (B) siRNA-mediated knockdown of GATA2 affects activity of enhancer E9 but not E1, further confirming GATA2 binding to E9 only. Similar experiments, including using control siRNAs, show the specificity of binding of NKX2.5 and SOX10 to E7 and E10, respectively. (C) siRNA-mediated knockdown of GATA2, SOX10 and NKX2.5 demonstrates that loss of SOX10 also has an effect on activity of GATA2 regulated enhancer E9, highlighting feedback between the two TFs. The other enhancers are only affected by loss of their cognate TF. Pairwise comparisons are against a vector with basal promoter-only (black) for measuring enhancer activity and between untransfected and siRNA transfected cells for measuring TF specificity for both (B) and (C). Error bars represent SEs of the mean (^*P < 0.01, ^**P < 0.001) for five independent biological replicates in all experiments.

Figure 4

RET–EDNRB GRN in the developing mouse ENS. Gene expression of Ednrb, Gata2, Nkx2.5 and Ret in the developing mouse gut at embryonic stage E11.5 and E12.5 in wild-type and Sox10 heterozygote embryos shows that Ednrb, Gata2 and Ret are transcriptionally affected at both developmental stages by loss of Sox10 expression. The effect on Nkx2.5 is only observed at E12.5. All pairwise comparisons are between wild-type and heterozygous embryonic guts within each developmental stage. Error bars represent SEs of the mean (^*P < 0.01, ^**P < 0.001) for three independent embryos for each developmental stage and genotype.

Figure 5

Transcriptional feedback between RET, EDNRB and their TFs. (A) EDNRB gene expression in human SK-N-SH cells with increasing doses of EDNRB siRNA (12–40 μM). The transcriptional effect is seen in RET, SOX10 and GATA2 when EDNRB levels are significantly reduced to below 50% of wild-type levels. (B) Experiments as in (A) for RET loss of expression (12–25 μM) in SK-N-SH cells show a transcriptional feedback on EDNRB, SOX10 and GATA2 but only when RET expression falls below 50% of wild-type levels. NKX2.5 and RARB levels remain unchanged in both experiments. All pairwise comparisons are with transfections using control siRNAs. Error bars represent SEs of the mean (^*P < 0.01, ^**P < 0.001) for five independent biological replicates.

Figure 6

The RET–EDNRB GRN. RET and EDNRB are the two major genes for ENS development and harbor multiple mutations leading to HSCR. These genes are coregulated within a larger GRN controlled by at least two common TFs with feedback and feed forward loops as significant features. The grey colored components of the GRN (GDNF, GFRA1 and CBL) were deciphered in our previous study (7).