Systems analysis of de novo mutations in congenital heart diseases identified a protein network in the hypoplastic left heart syndrome

Abstract

Despite a strong genetic component, only a few genes have been identified in congenital heart diseases (CHDs). We introduced systems analyses to uncover the hidden organization on biological networks of mutations in CHDs and leveraged network analysis to integrate the protein interactome, patient exomes, and single-cell transcriptomes of the developing heart. We identified a CHD network regulating heart development and observed that a sub-network also regulates fetal brain development, thereby providing mechanistic insights into the clinical comorbidities between CHDs and neurodevelopmental conditions. At a small scale, we experimentally verified uncharacterized cardiac functions of several proteins. At a global scale, our study revealed developmental dynamics of the network and observed its association with the hypoplastic left heart syndrome (HLHS), which was further supported by the dysregulation of the network in HLHS endothelial cells. Overall, our work identified previously uncharacterized CHD factors and provided a generalizable framework applicable to studying many other complex diseases. A record of this paper's Transparent Peer Review process is included in the supplemental information.

Keywords: biological networks; congenital heart diseases; hypoplastic heart; personal genomes; single cell; systems biology.

Figure 1.. RUN Identified a highly connected network in CHD.

A. Overview of the study design. B. PCGC candidate genes tended to occupy central positions on the protein interaction network. PCGC candidate genes (red) were identified by dosage-sensitive genes affected by de novo loss-of-function (LoF) mutations from the PCGC CHD cohorts. The same procedure also identified affected genes (blue) from the matched unaffected sibling cohort. Network connectivity indicates the number of interacting partners for each protein on the network. As an independent control experiment, the same comparison was also performed on genes affected by de novo synonymous mutations, whose functions were presumably neutral. P-values were derived from Wilcoxon rank-sum test. C. PCGC candidate genes were more likely to maintain mutual interactions on the network. The fractions of interacting proteins were computed among the candidate and control genes identified from proband and sibling cohorts, respectively. Genes affected by de novo synonymous mutations were used as an independent control experiment. P-values were derived from the Fisher’s exact test. D. The identified network component has substantially increased reachability to PCGC CHD candidate proteins relative to all other proteins on the network. P-values were derived from Wilcoxon rank-sum test.

Figure 2.. Functional characterization of the CHD network.

A. An overview of the identified network seeded with PCGC candidate proteins (grouped by CHD subtypes that were color coded), and the orange nodes were novel proteins identified by the RUN algorithm. The subtype annotations were derived from the original publication(Jin et al., 2017), where CTD, HTX and LVO stand for conotruncal defects, heterotaxy and left ventricular outflow tract obstruction, respectively. Other indicates more than one subtype was associated with the corresponding proteins. B. Temporal expression of the network genes across heart developmental stages. Human genes were mapped onto mouse orthologs, and hierarchical clustering revealed two expression components of the network, where Group-I (G-I) genes displayed preferential expression from embryonic stem cells (ESC) to E7.3, whereas Group-II (G-II) genes exhibited substantial expression enrichment from E7.3 to postnatal and adult stages. Close examination of Group-II genes further revealed two subcluster structure, where Group-II-A (G-II-A) was preferentially expressed across fetal developmental stages and Group-II-B (G-II-B) was more specific in postnatal stages, particularly strong in the adult heart. The log2-based gene expression was normalized into z-scores across time.

Figure 3.. Expression analysis of the identified CHD network.

A. The expression of Group-I (G-I) genes was peaked at Carnegie Stage 12–13 (gestational week 4) and was then quickly reduced to a level even below the transcriptome background at Carnegie Stage 16–23 (gestational week 5–8). In contrast, Group-II (G-II) genes did not display significant expression increase from the transcriptome background. P-values were derived from Wilcoxon rank-sum test. B. Group-II-A (G-II-A) genes showed significantly increased expression in the fetal heart from postconceptional day 96 to day 147, whereas Group-I (G-I) genes and Group-II-B (G-II-B) genes did not display statistical significance (p > 0.05, Wilcoxon rank-sum test). C. Group-II-A (G-II-A) genes showed significantly increased expression in the fetal heart in the gestational weeks 19 and 28 based on RNA-seq data. The statistical significance was not observed from other gene groups. P-values were derived from Wilcoxon rank-sum test.

Figure 4.. Validating novel functions of the identified genes in regulating fetal heart development.

A-C. Network clustering identified 33 local clustering structures on the identified network, where cluster #4 (A), #2 (B), #3 (C) were presented as representative pathways regulating heart development. D. Gene ontology enrichment of the differentially expressed genes in iPSC-CMs upon siRNA knockdown of TEAD2, TLK1 and RBBP5, respectively. These differentially expressed genes consistently displayed strong functional enrichment for heart development and cardiac muscle contraction. The color intensities of the circles represent false discovery rates (FDRs). Sizes of the circles represent the enrichment scores. E,F,L,M. RNA-seq identified differentially expressed genes in iPSC-CMs upon siRNA knockdown of TEAD2 (E), TLK1 (F), RBBP5 (L) and ASH2L (M), respectively. X-axis is the mean expression of each gene in iPSC-CMs, and Y-axis indicates their respective fold changes upon siRNA knockdown (siRNA treatment vs. siRNA control). Genes with false discovery rates (FDRs) less than 0.05 were highlighted in red. G-K. Cellular contractility assay in the iPSC-CMs. siRNA knockdown against TEAD2 in iPSC-CMs resulted in a marked reduction of the cardiomyocyte beating rate (G), increased contraction velocity (H), contraction deformation distance(I), relaxation velocity J. and relaxation deformation distance (K) relative to the siRNA control. P-values were derived from t-test (*, p < 0.05; **, p<0.01; ***, p < 0.001; ****, p < 0.0001). Error bars represent standard error of the mean. N. Cellular contractility assay in the iPSC-CMs. RBBP5 knockdown in iPSC-CMs displayed an increased beating rate. P-values were derived from t-test (*, p < 0.05; **, p<0.01; ***, p < 0.001; ****, p < 0.0001). Error bars represent standard error of the mean. O-R. Cellular contractility assay in the iPSC-CMs. ASH2L knockdown in iPSC-CMs showed increased contraction velocity (O), contraction deformation distance (P), relaxation velocity (Q) and relaxation deformation distance (R). P-values were derived from t-test (*, p < 0.05; **, p<0.01; ***, p < 0.001; ****, p < 0.0001). Error bars represent standard error of the mean. S-T. ASH2L^+/− knockout lines clone 1 (S) and clone 2 (T) significantly reduced the differentiation efficiencies into cardiomyocytes (TNNT2 positive cells) from iPSCs. P-values were derived from paired t-test (*, p < 0.05; **, p<0.01; ***, p < 0.001; ****, p < 0.0001). Error bars represent standard error of the mean.

Figure 5.. Group-II genes on the network were enriched for pathogenic mutations in probands with HLHS.

A-B. Rare non-synonymous variants displayed a significant increase in mutational deleteriousness (CADD (A), VARITY (B)) in HLHS probands, but not in individuals with ASD, VSD, TGA or TOF, relative to unaffected siblings. HLHS, ASD, VSD, TGA and TOF stand for hypoplastic left heart syndrome, atrial septal defects, ventricular septal defects, transposition of the great arteries, and tetralogy of fallot. P-values were derived from Wilcoxon rank-sum test. C. Significant differences in mutational deleteriousness (CADD) of rare synonymous mutations were not observed in any CHD subtypes relative to unaffected siblings. P-values were derived from Wilcoxon rank-sum test. D. Excluding individuals comorbid with ASD further boosted statistical significance (Group-II genes). P-values were derived from Wilcoxon rank-sum test.

Figure 6.. Single-cell analysis of the network in the HLHS heart.

A. Across all the cell types in the HLHS left ventricle, Group-II (G-II) genes showed the strongest expression reduction in the endothelium cells. P-values were derived from Wilcoxon rank-sum test, B. Examining two subgroups of Group-II (G-II) genes revealed significant downregulation of both subgroups (G-II-A and G-II-B) in the endothelium cells. P-values were derived from Wilcoxon rank-sum test. C. Only Group-II-A (G-II-A) genes displayed significant expression reduction in the conduction system in the HLHS left ventricle. P-values were derived from Wilcoxon rank-sum test.