Spatial and temporal mapping of de novo mutations in schizophrenia to a fetal prefrontal cortical network

Abstract

Genes disrupted in schizophrenia may be revealed by de novo mutations in affected persons from otherwise healthy families. Furthermore, during normal brain development, genes are expressed in patterns specific to developmental stage and neuroanatomical structure. We identified de novo mutations in persons with schizophrenia and then mapped the responsible genes onto transcriptome profiles of normal human brain tissues from age 13 weeks gestation to adulthood. In the dorsolateral and ventrolateral prefrontal cortex during fetal development, genes harboring damaging de novo mutations in schizophrenia formed a network significantly enriched for transcriptional coexpression and protein interaction. The 50 genes in the network function in neuronal migration, synaptic transmission, signaling, transcriptional regulation, and transport. These results suggest that disruptions of fetal prefrontal cortical neurogenesis are critical to the pathophysiology of schizophrenia. These results also support the feasibility of integrating genomic and transcriptome analyses to map critical neurodevelopmental processes in time and space in the brain.

Figure 1. De novo mutations in persons with schizophrenia and unaffected siblings

(A) All de novo point mutations in 189 persons. Observed numbers of events fit a Poisson distribution with m = 0.87 events per person. R² = 0.97 for goodness of fit of observed to expected values under Poisson assumption. (B) De novo point mutations in 189 offspring by terciles of paternal age. Presence of at least one de novo point mutation is associated with older paternal age (X² = 13.96, p = 0.0009). (C). Presence of at least one de novo putatively damaging mutation in schizophrenia probands and unaffected siblings. 45% (47/105) of the probands and 30% (25/84) of the unaffected siblings carried at least one damaging de novo mutation (X²=4.45, P=0.035). (D) Genes harboring de novo damaging mutations in persons with schizophrenia. Colors represent types of mutations: blue are missense, red are nonsense, black are frameshift, green are splice, and orange are CNVs.

Figure 2. Interconnectedness of protein-protein interaction (PPI) networks based on genes harboring de novo mutations

(A–B) The 54 genes harboring de novo damaging mutations in schizophrenia cases yield a PPI network of 18 genes (A) and 23 interactions (B), indicated by the vertical dotted lines. For comparison, 10,000 sets of 54 genes were selected at random from 264 genes harboring de novo damaging mutations in unaffected sibs (controls). Each of these 10,000 gene sets also yields a PPI network. The numbers of genes and connecting links for each of these 10,000 control networks was plotted in panels A and B, respectively. P-values for enrichment of genes and connections in the case network were estimated by the proportion of control networks with more genes or interactions than the case network. The network based on cases is enriched both for number of genes (P=0.002) and for number of connections between genes (P=0.0005). (C–D) Analogous networks based on genes harboring de novo benign events in cases and in controls. The networks based on genes with de novo benign events in cases do not differ from networks based on genes with de novo benign mutations in controls.

Figure 3. Interconnectedness of transcriptional co-expression networks, at various developmental stages and in different brain regions, based on genes harboring de novo damaging mutations

Co-expression of genes harboring de novo damaging mutations in cases and in controls was evaluated using RNASeq data from the BrainSpan Atlas. Gene pairs were defined as co-expressed if |R| > 0.8 for their RNASeq expression levels across all tissues from a given brain region and a given developmental stage. Networks were created for co-expressed gene pairs as described for Figure 2B. Dotted lines indicate numbers of connections (edges) in networks created using genes with de novo damaging mutations in cases. Histograms represent distributions of the numbers of edges in 10,000 simulated networks using genes with de novo damaging mutations in controls. The most significant enrichment for co-expression of genes mutant in schizophrenia was observed in frontal cortex during fetal development (P=0.00006; Table S6). There was no enrichment for co-expression of genes with de novo benign mutations in schizophrenia compared to controls (Figure S2).

Figure 4. Interconnectedness of transcriptional co-expression networks from four subregions of the fetal frontal cortex

Tissues from the fetal frontal cortex with RNASeq data in the BrainSpan Atlas were classified by subregion. Networks were then created, as described for Figure 3, for genes with de novo damaging mutations in cases and in controls (left panels) and for de novo benign mutations in cases and in controls (right panels). The greatest enrichments for co-expression of genes with de novo damaging events were observed in the dorsolateral and ventrolateral prefrontal cortex (Table S7). There were no regions enriched for co-expression of genes harboring de novo benign mutations in cases.

Figure 5. Network of genes harboring damaging de novo mutations in schizophrenia that are co-expressed in fetal dorsolateral (DFC) and ventrolateral (VFC) prefrontal cortex

Merging data from networks of protein-protein interaction (PPI) network and of co-expression in fetal DFC and fetal VFC yields a network of 50 genes (nodes) with 126 connections (edges). Red lines represent co-expressed genes; blue lines represent physical connections between protein products of the genes.

Figure 6. Expression patterns in frontal cortex of genes with de novo damaging mutations in schizophrenia

Gene expression across all periods of development (13 post-conception weeks to 23 years) for each gene harboring a de novo damaging mutation in schizophrenia is depicted for four brain regions: dorsolateral prefrontal cortex (DFC), anterior (rostral) cingulate (medial prefrontal) cortex (MFC), orbital frontal cortex (OFC), and ventrolateral prefrontal cortex (VFC). Expression levels were obtained from BrainSpan Atlas of Developing Human Brain (2013). RPKM values were max-min normalized and Loess smoothed (Cleveland et al., 1992) across time points.