Research ArticleDISEASES AND DISORDERS

Deconvolution of transcriptional networks identifies TCF4 as a master regulator in schizophrenia

See allHide authors and affiliations

Science Advances  11 Sep 2019:
Vol. 5, no. 9, eaau4139
DOI: 10.1126/sciadv.aau4139
  • Fig. 1 An overview of the current study.

    (A and B) Obtaining gene expression data generated from DLPFC (CMC data) and later CNON data as an independent validation. (C) Transcriptional network reconstruction based on the RNA-seq data using algorithm for reconstruction of accurate cellular networks (ARACNe) and identification of TCF4 as a disease-relevant MR. (D) Knockdown of TCF4 in hiPSCs derived from a patient with SCZ and two unaffected individuals. (E) Differentiation of hiPSC into neural progenitor cells (NPCs) and glutamatergic neurons (Glut_Ns). (F) RNA-seq of the generated neuronal cells with and without TCF4 knockdown. (G) Downstream inferences and analysis of the conducted experiments.

  • Fig. 2 Summary of TCF4 regulons in the CMC and CNON data.

    (A) TCF4 motif from the JASPAR database used in the analysis. (B) TCF4 binding site enrichment scores (based on TCF4 motifs from the JASPAR database) among the TCF4 regulons compared to that of a set of random genes. (C) Enrichment P values of the TCF4 regulons, compared with several gene sets, including the predicted TCF4 sites from ATAC-seq, the differentially expressed genes from our TCF4 knockdown experiments, and the predicted TCF4 binding sites in neuroblastoma cells by ChIP-seq (P value obtained from Fisher’s exact test). (D) A schematic of the network created from CMC data as well as the TCF4 targets. (E) TCF4 targets from CMC and CNON data combined with some of the associated pathways. (F) List of overlapping predicted TCF4 targets with an ATAC-seq on hiPSC–Glut_Ns and two ChIP-seq datasets.

  • Fig. 3 TCF4 knockdown in hiPSC-derived neural progenitor cells (NPCs) and Glut_Ns (from a SCZ patient line).

    (A) Representative immunofluorescence (IF)–staining images of hiPSCs. hiPSCs are stained positive for pluripotency markers TRA-1-60 (green) and NANOG (red). (B) Representative IF-staining images of hiPSCs. hiPSCs are stained positive for OCT4 (green) and SSEA-4 (red). (C) Representative IF-staining images of NPCs (3 days after plating cells): NPCs are stained positive for PAX6 (green) and NESTIN (red). (D) Representative IF-staining images of NPCs (3 days after plating cells): NPCs are stained positive for NESTIN (red) and SOX2 (green). (E) Representative image of day 14 neurons stained positive for MAP2 (green) and vGLUT1 (red). (F) TCF4 knockdown efficiency measured by qPCR in NPCs (3 days after plating). (G) TCF4 knockdown efficiency measured by qPCR in Glut_Ns (14 days after neuronal induction). (H) Volcano plot of DE genes in NPCs. (I) Volcano plot of DE genes in Glut_Ns. (J) TCF4 expression levels at days 3 and 14 in RNA-seq data (K) Overlap of the DE genes upon TCF4 knockdown with a list of GWAS-implicated SCZ risk genes (up- and down-regulated genes are shown in red and blue, respectively; white cells indicate that the gene is not DE). (L) Overlap of the DE genes upon TCF4 knockdown with a list of credible GWAS risk loci (up- and down-regulated genes are shown in red and blue, respectively; white cells indicate that the gene is not DE). Scale bars, 50 μm in all images. Cell nuclei are stained with 4′,6-diamidino-2-phenylindole (DAPI) (blue) (A to E), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is used as the endogenous control to normalize the TCF4 expression for qPCR. Error bars, mean ± SD (n = 4). ***P < 0.001, Student’s t test, two-tailed, heteroscedastic.

  • Fig. 4 TCF4 knockdown in hiPSC-derived Glut_Ns in two independent cell lines CD07 and CD09 from unaffected control subjects.

    (A and B) IF staining of hiPSC-derived Glut_Ns from the CD07 control line. (C and D) IF staining of hiPSC-derived Glut_Ns from the CD09 control line; MAP2, green; vGlut1, red; DAPI, blue. Scale bar, 50 μm. (E and F) Real-time qPCR (qRT-PCR) result of TCF4 expression level in two different lines, demonstrating knockdown efficiency; GAPDH is used as the endogenous control to normalize the TCF4 expression for qPCR; error bars, mean ± SD (n = 6 to 8); ***P < 0.001, Student’s t test, two-tailed, heteroscedastic. (G) Percentage of vGlut1-positive cells derived from two different hiPSC lines; cell counts were from five images in each line. (H) TCF4 expression levels in knockdown (K1, K2, and K3) and control samples (C1, C2, and C3) in the generated RNA-seq data on CD07 and CD09. (I) Pathway enrichment analysis results on the identified DE genes in CD07 (color legend indicates the number of overlapped DE genes with the corresponding pathway). (J) Pathway enrichment analysis results on the identified DE genes in CD09 (color legend indicates the number of overlapped DE genes with the corresponding pathway). (K) Correlation plot of the log fold changes (logFC) of the Glut_Ns from the SCZ cell line (“old”) and CD07. (L) Correlation plot of the log fold changes of the Glut_Ns from the SCZ cell line (“old”) and CD09. ns, not significant.

  • Fig. 5 Enrichment of SCZ de novo SNVs in TCF4-associated gene expression changes from NPCs and Glut_Ns.

    (A) Analysis framework. (B) Enrichment P values (−log10-transformed) of all DE genes at day 3 (NPCs). The P value on the y axis obtained from a hypergeometric test was used to test the statistical significance of each overlap (number of shared genes between lists) and using all protein coding genes as a background set. (C) Enrichment P values (−log10-transformed) of all DE genes at day 14 (Glut_Ns). (D) Enrichment of SCZ dnSNVs in TCF4-associated gene expression changes.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/9/eaau4139/DC1

    Fig. S1. Representation of the top active MRs identified in the CMC data.

    Fig. S2. GOSlim summary of the up-regulated genes on day 3.

    Fig. S3. GOSlim summary of the down-regulated genes on day 3.

    Fig. S4. GOSlim summary of the up-regulated genes on day 14.

    Fig. S5. GOSlim summary of the down-regulated genes on day 14.

    Fig. S6. Biological pathways altered upon TCF4 knockdown in hiPSC-derived neuronal cells.

    Fig. S7. MetaCore analysis of top 2000 dysregulated genes upon TCF4 knockdown in hiPSC-derived neurons.

    Fig. S8. MetaCore analysis of all gene sets dysregulated upon TCF4 knockdown in hiPSC-derived neurons.

    Fig. S9. The identified TCF4 interactome and their contributions to expression of TCF4.

    Fig. S10. Batch correction of CD07 cell line data.

    Fig. S11. Correlations between the expression profile (summarized as median gene expression of each gene across samples) of the CMC, CNON, and NPC data with the 53 tissues from the GTEx consortium.

    Table S1. Targets of the identified hub genes in the CMC network.

    Table S2. Network structure created from the CMC data.

    Table S3. Network structure created from the CNON data.

    Table S4. Targets of the identified hub genes in the CNON network.

    Table S5. Targets of five MRs in the CMC and CNON data.

    Table S6. List of the altered pathway upon TCF4 knockdown in NPCs and Glut_Ns in the SCZ cell line.

    Table S7. List of differentially expressed genes in NPCs and Glut_Ns in the SCZ cell line.

    Table S8. MetaCore enrichment analysis on NPCs in the SCZ cell line.

    Table S9. MetaCore enrichment analysis on Glut_Ns in the SCZ cell line.

    Table S10. List of TCF4 targets in other studies.

    Table S11. VIPER enrichment scores of the identified MRs in the CMC and CNON data.

  • Supplementary Materials

    The PDF file includes:

    • Fig. S1. Representation of the top active MRs identified in the CMC data.
    • Fig. S2. GOSlim summary of the up-regulated genes on day 3.
    • Fig. S3. GOSlim summary of the down-regulated genes on day 3.
    • Fig. S4. GOSlim summary of the up-regulated genes on day 14.
    • Fig. S5. GOSlim summary of the down-regulated genes on day 14.
    • Fig. S6. Biological pathways altered upon TCF4 knockdown in hiPSC-derived neuronal cells.
    • Fig. S7. MetaCore analysis of top 2000 dysregulated genes upon TCF4 knockdown in hiPSC-derived neurons.
    • Fig. S8. MetaCore analysis of all gene sets dysregulated upon TCF4 knockdown in hiPSC-derived neurons.
    • Fig. S9. The identified TCF4 interactome and their contributions to expression of TCF4.
    • Fig. S10. Batch correction of CD07 cell line data.
    • Fig. S11. Correlations between the expression profile (summarized as median gene expression of each gene across samples) of the CMC, CNON, and NPC data with the 53 tissues from the GTEx consortium.

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • Table S1 (Microsoft Excel format). Targets of the identified hub genes in the CMC network.
    • Table S2 (Microsoft Excel format). Network structure created from the CMC data.
    • Table S3 (Microsoft Excel format). Network structure created from the CNON data.
    • Table S4 (Microsoft Excel format). Targets of the identified hub genes in the CNON network.
    • Table S5 (Microsoft Excel format). Targets of five MRs in the CMC and CNON data.
    • Table S6 (Microsoft Excel format). List of the altered pathway upon TCF4 knockdown in NPCs and Glut_Ns in the SCZ cell line.
    • Table S7 (Microsoft Excel format). List of differentially expressed genes in NPCs and Glut_Ns in the SCZ cell line.
    • Table S8 (Microsoft Excel format). MetaCore enrichment analysis on NPCs in the SCZ cell line.
    • Table S9 (Microsoft Excel format). MetaCore enrichment analysis on Glut_Ns in the SCZ cell line.
    • Table S10 (Microsoft Excel format). List of TCF4 targets in other studies.
    • Table S11 (Microsoft Excel format). VIPER enrichment scores of the identified MRs in the CMC and CNON data.

    Files in this Data Supplement:

Navigate This Article