Multi-omic analyses identify molecular targets of Chd7 that mediate CHARGE syndrome model phenotypes

Abstract

CHARGE syndrome is a developmental disorder that affects 1 in 10,000 births, and patients exhibit both physical and behavioral characteristics. De novo mutations in CHD7 (chromodomain helicase DNA binding protein 7) cause 67% of CHARGE syndrome cases. CHD7 is a DNA-binding chromatin remodeler with thousands of predicted binding sites in the genome, making it challenging to define molecular pathways linking loss of CHD7 to CHARGE phenotypes. To address this problem, here we used a previously characterized zebrafish CHARGE model to generate transcriptomic and proteomic datasets from larval zebrafish head tissue at two developmental time points. By integrating these datasets with differential expression, pathway, and upstream regulator analyses, we identified multiple consistently dysregulated pathways and defined a set of candidate genes that link loss of chd7 with disease-related phenotypes. Finally, to functionally validate the roles of these genes, CRISPR/Cas9-mediated knockdown of capgb, nefla, or rdh5 phenocopies behavioral defects seen in chd7 mutants. Our data provide a resource for further investigation of molecular mediators of CHD7 and a template to reveal functionally relevant therapeutic targets to alleviate specific aspects of CHARGE syndrome.

Figure 1.

Loss of chd7 causes dysregulation of genes associated with neurodevelopmental and transcriptional regulation functions. Principal component analysis (PCA) of A. quantified RNAs for each sample and replicate, and B. quantified proteins for each sample and replicate. Heatmap of scaled Log₂ Fold Change (Log₂FC) with hierarchical clustering by similar fold change patterns of C. Differentially Expressed Genes (DEGs) with a p-value < 0.05 in at least one sample (n = 6114) and D. Differentially Expressed Proteins (DEPs) with a p-value < 0.05 in at least one sample (n = 553). Heatmap of scaled Log₂FC with hierarchical clustering by similar fold change patterns of E. DEGs that overlap with a list of neurodevelopmental genes generated from GO Accession terms and Ensembl (Neuro-GO) (n = 553) and F. DEPs that overlap with the Neuro-GO list (n=89). C-F. Slices (left, numbered in black and white) by hierarchical clustering of similar fold change patterns.

Figure 2.

Neural-related transcripts with protein counterparts change developmentally from 3 dpf to 5 dpf due to loss of chd7. Volcano plot of Differentially Expressed Genes (DEGs) A. 5 dpf Wild Type (WT) compared to 3 dpf WT samples, B. 5 dpf Heterozygous (HT) compared to 3 dpf HT samples, and C. 5 dpf Homozygous Mutant (MUT) compared to 3 dpf MUT samples. Ingenuity Pathway Analysis (IPA) pathway enrichment by patterns of DEGs from D. 5 dpf WT compared to 3 dpf WT samples, E. 5 dpf HT compared to 3 dpf HT samples, and F. 5 dpf MUT compared to 3 dpf MUT samples. Overlap of DEGs in WT 5 dpf v 3 dpf, HT 5 dpf v 3 dpf, and MUT 5 dpf v 3 dpf G. DEGs with p-value < 0.05, H. DEGs up-regulated with Fold Change (FC) > 1 and p-value < 0.05, I. DEGs down-regulated with FC < 1 and p-value < 0.05. J. Log₂ FC for DEGs selected for most differing FC from S1. and all DEGs from overlap with Neuro-GO list S2. underlined in red and S3. underlined in blue.

Figure 3.

Early neurodevelopmental transcripts with protein counterparts and neural signaling pathways emerge due to loss of chd7 at 3 dpf. Volcano plot of Differentially Expressed Genes (DEGs) comparing A. 3 dpf Heterozygous (HT) compared to 3 dpf Wild Type (WT) samples and B. 3 dpf Homozygous Mutant (MUT) compared to 3 dpf WT samples. Volcano plot of Differentially Expressed Proteins (DEPs) comparing C. 3 dpf HT compared to 3 dpf WT samples and D. 3 dpf MUT compared to 3 dpf WT samples. Ingenuity Pathway Analysis (IPA) pathway enrichment by patterns of DEGs from E. 3 dpf HT compared to WT samples and F. 3 dpf MUT compared to WT samples. IPA pathway enrichment by patterns of DEPs from G. 3 dpf HT compared to WT samples and H. 3 dpf MUT compared to WT samples. I. Overlap of DEGs with p-value < 0.05 from 3 dpf HT v WT and 3 dpf MUT v WT. J. Overlap of DEPs with p-value < 0.05 from 3 dpf HT v WT and 3 dpf MUT v WT. K. Overlap of DEGs with p-value < 0.05 from 3 dpf HT v WT and MUT v WT (purple) and DEPs with p-value < 0.05 from 3 dpf HT v WT and MUT v WT (green). L. Overlap of 3 dpf DEGs from I. and DEPs from J. and Neuro-GO list. M. Log₂ Fold Change for all DEGs and DEPs from MUT v WT comparison from overlap in K. and selected from overlap in L. are underlined.

Figure 4.

At 5 dpf down-regulated transcripts with protein counterparts and inhibition of pathways emerge. Volcano plot of Differentially Expressed Genes (DEGs) comparing A. 5 dpf Heterozygous (HT) compared to 5 dpf Wild Type (WT) samples and B. 5 dpf Homozygous Mutant (MUT) compared to 5 dpf WT samples. Volcano plot of Differentially Abundant Proteins (DEPs) comparing C. 5 dpf HT compared to 5 dpf WT samples and D. 5 dpf MUT compared to 5 dpf WT samples. Ingenuity Pathway Analysis (IPA) pathway enrichment by patterns of DEGs from E. 5 dpf HT compared to WT samples and F. 5 dpf MUT compared to WT samples. Ingenuity Pathway Analysis (IPA) pathway enrichment by patterns of DEPs from G. 5 dpf HT compared to WT samples and H. 5 dpf MUT compared to WT samples. I. Overlap of DEGs with p-value < 0.05 from 5 dpf HT v WT and DEGs with p-value < 0.05 from 5 dpf MUT v WT. J. Overlap of DEPs with p-value < 0.05 from 5 dpf HT v WT and 5 dpf MUT v WT. K. Overlap of DEGs with p-value < 0.05 from 5 dpf HT v WT and MUT v WT (purple) and DEPs with p-value < 0.05 from 5 dpf HT v WT and MUT v WT (green). L. Overlap of 5 dpf DEGs and DEPs from K. and Neuro-GO list. M. Log₂ FC for selected DEGs and DEPs from MUT v WT comparison from overlap in K. and overlap in L. underlined.

Figure 5.

Transcript and protein expression patterns inform downstream pathway, regulator, and functional analysis. Ingenuity Pathway Analysis (IPA) canonical pathway enrichment by patterns of A. DEGs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons and B. DEPs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons. IPA upstream regulator enrichment by patterns of C. DEGs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons and D. DEPs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons. IPA diseases and functions enrichment by patterns of E. DEGs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons and F. DEPs with p-value < 0.05 from 3 dpf and 5 dpf HT v WT and MUT v WT comparisons. G. Overlap of DEGs with p-value < 0.05 3 dpf and 5 dpf HT v WT and MUT v WT comparisons and DEPs with p-value < 0.05 from HT v WT and MUT v WT comparisons. H. Overlap of DEGs and DEPs from G. with public ChIP-seq data from Schnetz 2010. I. Overlap of DEGs and DEPs from G. with public ChIP-seq data from Reddy 2021. J. Overlap of DEGs and DEPs from G. with Neuro-GO list. K. Log2 FC of selected candidate DEGs and DEPs from MUT v WT comparison plotted at 3 dpf and 5 dpf, and table below noting presence in ChIP-seq or Neuro-GO list.

Figure 6.

Knockdown of candidate genes causes phenocopy of CHARGE model phenotypes A. Short-Latency C-bend (SLC) frequency plotted against acoustic stimulus intensity for each group of larvae fit with a nonlinear regression sigmoidal dose response curve. B. Area under the curve (AUC) plotted for each individual larva’s SLC response frequency. C. Long-Latency C-bend (LLC) frequency plotted against acoustic stimulus intensity for each group of larvae fit with a nonlinear regression second order polynomial (quadratic) curve. D. AUC plotted for each individual larva’s LLC response frequency. E. Short-term Habituation (STH) SLC frequency plotted against time for each group of larvae fit with nonlinear regression one-phase decay curve. F. STH AUC for each individual larva’s SLC response frequency. G. STH LLC frequency plotted against time for each group of larvae fit with a nonlinear regression second order polynomial (quadratic) curve. H. STH AUC for each individual larva’s LLC response frequency. I. Pre-pulse Inhibition percent. J. Dark flash response frequency. K. Light flash response frequency. L. Frequency of morphological defects observed for each group of larvae. (Bar graphs: Mean ± Standard Deviation, Line graphs: Mean ± SEM, Ordinary one-way ANOVA with Dunnetts’s multiple comparisons, * p-value < 0.05, ** p-value < 0.01, *** p-value < 0.001, **** p-value < 0.0001).