Gene expression elucidates functional impact of polygenic risk for schizophrenia

Only five genes in the MHC locus were ranked highly by Sherlock and passed concordance evaluation: C4A, HCG17, VARS2, HLA-DMB, and BRD2.

“Nevertheless, only five genes in this locus were ranked highly by Sherlock and passed evaluation for concordance of associations (Supplementary data file 2): C4A, HCG17, VARS2, HLA-DMB, and BRD2.”

When CMC eQTLs at FDR ≤ 5% were intersected with HBCC eQTLs at FDR ≤ 20%, 93% showed the same direction of allelic effect.

“Moreover, concordance of the direction of allelic effect was high, with 93% of eQTL showing the same direction of effect when intersecting CMC eQTL (FDR ≤ 5%) with even a liberally defined set of HBCC eQTL (FDR ≤ 20%).”

Schizophrenia cases carried more common variant alleles previously associated with schizophrenia risk than controls, with P = 1.6 × 10−8.

“As expected3, SCZ cases inherited an increased number of common variant alleles previously associated with SCZ risk (P = 1.6 × 10−8, Supplementary Fig. 1D).”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

Of the 108 previously reported schizophrenia GWAS loci, 73 contain cis-eQTL SNPs for one or more genes at FDR ≤ 5%.

“Of the 108 SCZ GWAS loci previously reported3, 73 harbor cis-eQTL SNPs for one or more genes (FDR ≤ 5%).”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

Overexpression of TSNARE1 in zebrafish embryos significantly decreased head size by 9.5%, P < 0.001.

“Relative to control embryos, overexpression of TSNARE1 or CNTN4 resulted in a significant decrease in head size, 9.5% (P < 0.001) and 3.5% (P = 0.018), respectively, while SNAP91 or CLCN3 showed no statistically significant effect (Fig. 3A, B).”

The observed number of 73 loci with cis-eQTL SNPs is consistent with chance expectation.

“To determine if 73 out of 108 loci were larger than that expected by chance, we conducted an experiment that randomly chose such loci in the genome; it showed that 73 loci with cis-eQTL SNPs is consistent with chance expectation (data not shown).”

Overexpression of TSNARE1 increased proliferating PH3-positive cells by 9.5%, P = 0.018.

“Overexpression of CNTN4 and suppression of furin_a led to a 9.8% (P = 0.003) and a 29.8% (P < 0.001) decrease, respectively, in proliferating cells marked by phospho-histone3 (PH3), and overexpression of TSNARE1 led to a 9.5% increase (P = 0.018) in proliferating cells (N = 20 per experiment; Fig. 3C, D).”

Overexpression of CNTN4 in zebrafish embryos significantly decreased head size by 3.5%, P = 0.018.

“Relative to control embryos, overexpression of TSNARE1 or CNTN4 resulted in a significant decrease in head size, 9.5% (P < 0.001) and 3.5% (P = 0.018), respectively, while SNAP91 or CLCN3 showed no statistically significant effect (Fig. 3A, B).”

After removing genes with inconsistent evidence, 33 genes remained highlighted in 18 of the 108 GWAS loci.

“After removing genes where additional evaluation indicated lack of consistency (Supplementary Fig. 7B), there were 33 genes highlighted in 18 of the 108 GWAS loci (Supplementary data file 2).”

Sherlock analysis suggested co-localization of GWAS risk and eQTL association signals for 84 genes in 30 of the 108 schizophrenia GWAS loci at adjusted P < 0.05.

“The results suggested that GWAS risk and eQTL association signals co-localized for 84 genes in 30 of these loci (adjusted P < 0.05; Supplementary Fig. 6A, Supplementary data file 2).”

Body length and somitic structures were similar across embryos, suggesting the head-size observations were unlikely due to gross developmental delay.

“Body length and somitic structures were similar across all embryos, suggesting that our observations were unlikely due to gross developmental delay.”

Modulation of FURIN, TSNARE1, and CNTN4 led to a significant increase in apoptotic cells in the head region, P < 0.001.

“we performed TUNEL staining on injected embryos, and determined that modulation of all three target genes led to a significant increase in apoptotic cells in the head region corresponding to our head size measurements (N = 20 per experiment; P < 0.001; Fig. 3E, F).”

There was a strong correlation between schizophrenia risk alleles and up-regulation of C4A expression, Spearman ρ = 0.66 and P < 10−16.

“we found a strong correlation between the risk alleles for SCZ and up-regulation of expression of C4A (complement component 4A; Spearman’s ρ = 0.66, P < 10−16).”

IGF2 was the most strongly down-regulated gene, with a 1.33-fold change.

“Insulin-like growth factor 2 (IGF2), the most strongly down-regulated gene (1.33-fold, Fig. 5D), can rescue neurogenesis and cognitive deficits in certain mouse models of schizophrenia44.”

• cites [44] Ouchi, Y. et al. Reduced adult hippocampal neurogenesis and working memory deficits in the Dgcr8-deficient mouse model of 22q11.2 deletion-associated schizophrenia can be rescued by IGF2. J. Neurosci. 33, 9408–9419 (2013). sourceNot checked yet

  Not checked in the free run.

The study found no evidence of differential expression between schizophrenia cases and controls among the genes implicated by GWAS risk loci.

“In contrast, we did not find evidence for case-control differential expression among the implicated GWAS risk genes.”

FURIN knockdown in hiPSC neural progenitor cells significantly decreased total radial migration for all three individuals, with decreases of 1.16-fold, 1.23-fold, and 1.22-fold.

“FURIN knockdown in the hiPSC NPCs resulted in significantly decreased total radial migration for all three individuals (C1: 1.16-fold decrease, P < 0.0017; C2: 1.23-fold decrease, P < 3 × 10−6; C3: 1.22-fold decrease, P < 2 × 10−6) (Fig. 4).”

Four co-expression modules contained an excess of differentially expressed genes.

“Four modules stand out in harboring an excess of differentially expressed genes (Fig. 6A, Supplementary data file 6).”

A model incorporating allele frequency differences and predicted allele effects on expression produced a distribution of expected differential expression across genes that closely matched the observed distribution in the CMC dataset.

“Modeling the differential in allele frequencies and the predicted effect of alleles on gene expression demonstrates that the distribution of expected differential expression, across genes, is quite similar to the observed distribution from the CMC data (Fig. 7A).”

• Also stated in: § DISCUSSION · boosted

  “Our model demonstrates that the distribution of expected differential expression, across genes, is quite similar to the observed distribution from the CMC data (Fig. 7).”

The M2c module was enriched for gene sets from prior genetic studies, including the ARC protein complex, FMRP targets, neuronal markers, PSD proteins, and NMDA receptors.

“Gene sets identified in prior genetic studies that highlighted certain neurobiological functions are also enriched in the M2c module, including the activity-regulated cytoskeleton-associated (ARC) protein complex, targets of fragile X mental retardation protein (FMRP), neuronal markers, post-synaptic density (PSD) proteins, and NMDA receptors (Fig. 6A).”

TACR3 was the most highly up-regulated protein-coding gene, with a 1.24-fold change.

“The most highly up-regulated protein-coding gene is tachykinin receptor 3 (TACR3, NK3 receptor, 1.24-fold, Fig. 5D).”

The M2c module was enriched for axon guidance, postsynaptic membrane, transmission across chemical synapses, and voltage-gated potassium channel complexes.

“In particular, the M2c module is enriched for multiple categories, including axon guidance, postsynaptic membrane, transmission across chemical synapses, and voltage-gated potassium channel complexes (Fig. 6C).”

No obvious unifying biological theme emerged from the differentially expressed genes and isoforms on the basis of pathway enrichment analysis.

“No obvious unifying biological theme emerges from this set of genes and isoforms on the basis of pathway enrichment analysis (Supplementary data file 4).”

Analyses of cell composition do not support global differences in cellular composition in DLPFC tissue between schizophrenia cases and controls.

“Analyses of cell composition, however, do not support global differences in the cellular composition in DLPFC tissue from SCZ versus control subjects.”

Gene co-expression was organized into modules of distinct cellular and functional categories, consistent with prior brain transcriptome studies.

“Consistent with prior studies of the brain transcriptome15,47–50, we find gene co-expression to be organized into modules of distinct cellular and functional categories (Supplementary data file 7).”

• cites [15] Zhang, B. et al. Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer's disease. Cell 153, 707–720 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [47] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011). sourceNot checked yet

  Not checked in the free run.

• cites [48] Torkamani, A., Dean, B., Schork, N.J. & Thomas, E.A. Coexpression network analysis of neural tissue reveals perturbations in developmental processes in schizophrenia. Genome Res. 20, 403–412 (2010). sourceNot checked yet

  Not checked in the free run.

• cites [49] Oldham, M.C. et al. Functional organization of the transcriptome in human brain. Nat. Neurosci. 11, 1271–1282 (2008). sourceNot checked yet

  Not checked in the free run.

• cites [50] Roussos, P., Katsel, P., Davis, K.L., Siever, L.J. & Haroutunian, V. A system-level transcriptomic analysis of schizophrenia using postmortem brain tissue samples. Arch. Gen. Psychiatry 69, 1205–1213 (2012). sourceNot checked yet

  Not checked in the free run.

There were 239 differentially expressed isoforms between schizophrenia cases and controls: 94 up-regulated and 145 down-regulated.

“We identified 239 isoforms differentially expressed between SCZ cases and controls: 94 up-regulated and 145 down-regulated.”

To achieve 80% power for detecting a significant case-control difference in FURIN expression, 11,784 cases and 11,784 controls would be required.

“For example, 11,784 cases and 11,784 controls would be needed to have 80% power to detect a significant case-control difference in FURIN expression.”

The pattern of large observed differential expression is not observed when the sample size is increased to 250.

“Notably, this pattern not seen when the N is raised to 250.”

Only a handful of disease-genotype interaction associations were significant after FDR control.

“Nevertheless, to potentially identify eQTL that differ by disease state, a disease-genotype interaction term was also explicitly tested, but only a handful of such associations were found to be significant after controlling for FDR.”

Across the 20 samplings, the median number of FDR < 5% differentially expressed genes was 216 (mean = 315, sd = 261, 25th percentile = 92, 75th percentile = 562).

“Across the 20 samplings, the median number of FDR < 5% differentially expressed genes was 216 (mean = 315, sd = 261, 25th percentile = 92, 75th percentile = 562).”

1,584 genes had no cis-eQTL at FDR ≤ 5% but had at least one significant isoQTL.

“There were, however, 1,584 genes having no cis-eQTL (FDR ≤ 5%) that nevertheless had at least one significant isoQTL.”

If 5% of the 693 genes were false, the threshold of 34.7 genes was exceeded in 9 of 1000 experiments, slightly less than 1%.

“If 5% of the 693 were false, the threshold established of 34.7 genes is exceed in 9 out of 1000 experiments, slightly less than 1%.”

In permutation experiments, the average number of significant genes at FDR ≤ 0.05 was 4.3, well below the 693 found in the sample.

“Of these experiments, the average number of significant genes at FDR ≤ 0.05 was 4.3, well below 693 found in our sample.”

When the 20-60 age group (172 samples, mean age 45.5) and the complementary 60-or-older group (362 cases and controls, mean age 77.8) were processed independently, their differential expression t-statistics correlated at r = 0.18, P < 2 × 10−16.

“we explicitly compared the differential expression between the aged 20–60 individuals (172 samples, mean age 45.5) to an analysis of the complementary age 60 or older cohort (362 cases and controls, mean age 77.8) by independently processing the data for each of those sub-cohorts. The differential t-statistics between these independent sub-cohorts were correlated (r = 0.18, P < 2 × 10−16)”

For discovery differentially expressed genes, the median Pearson correlation of t-statistics with the replication cohort was 0.79 (mean = 0.75, sd = 0.16, 25th percentile = 0.67, 75th percentile = 0.88).

“For these FDR < 5% “discovery” differentially expressed genes, the median Pearson correlation of t-statistics with the “replication” cohort was 0.79 (mean = 0.75, sd = 0.16, 25th percentile = 0.67, 75th percentile = 0.88).”

The stricter filter increased replication in HBCC to 36% at FDR ≤ 5% in both cohorts.

“This stricter filter increases the replication rate in HBCC to 36% at FDR ≤ 5% in both cohorts.”

The stricter trans-eQTL filter yielded 13,114 trans-eQTL (about 7% of all trans-eQTL SNP-gene pairs at P < 10−5) across 661 genes.

“yielded a final reduced subset of 13,114 trans-eQTL (~7% of all trans-eQTL SNP-gene pairs at P < 10−5) across 661 genes.”

In the M2c module, genes with de novo mutations have higher intramodular connectivity than genes not genetically associated with schizophrenia (T test t=5.1, P=2.9×10−6).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

In the M2c module, genes with CNVs do not have higher intramodular connectivity than genes not genetically associated with schizophrenia (T test t=0.88, P=0.4).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

For schizophrenia cases, the cross-validation of module reproducibility yielded a median connectivity correlation between discovery and replication cohorts of 0.77 (mean 0.76, SD 0.06, IQR 0.73-0.78).

“The median Pearson correlation of connectivity values among the “discovery” and “replication” cohorts was 0.77 (mean = 0.76, sd = 0.06, 25th percentile = 0.73, 75th percentile = 0.78) and 0.80 (mean = 0.78, sd = 0.07, 25th percentile = 0.70, 75th percentile = 0.84) for cases with SCZ and controls, respectively.”

The overlap between drug signature datasets and M2c was significant after multiple-testing correction.

“While the overlap was significant after correcting for multiple testing”

Three out of 18 drug signature datasets overlapped with the M2c module.

“identified an overlap for 3 out of 18 drug signature datasets with M2c.”

There was a small positive correlation (Pearson r = 0.095, p < 10−16) between the independent t-statistics for PRS effects on expression in controls and the case-control expression differences in the full cohort.

“there was a significantly positive, but small, correlation (Pearson r = 0.095, p < 10−16) between the independent t-statistics for the effect of PRS on expression in controls and those we found for case-control expression differences in the full cohort”

In the M2c module, genes with PGC SCZ2 GWAS loci have higher intramodular connectivity than genes not genetically associated with schizophrenia (T test t=2.6, P=0.013).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

At nominal P < 0.05, module eigengenes M0c, M16c, M6c, M26c, M28c, M32c, M7s and M12s were associated with Institution, RIN or Library batch.

“At nominal P value < 0.05 we found an association of M0c, M16c, M6c, M26c, M28c, M32c, M7s and M12s MEs with Institution, RIN or Library batch.”

The WGCNA module label permutation approach showed a strong dependency on module size in this cohort (fitting index r2 > 0.95 based on a quadratic model).

“The permutation approach implemented in the WGCNA package (module label permutation in the test network) shows a strong dependency on module size in our cohort (fitting index r2 > 0.95 based on a quadratic model).”

For rodent drug signatures, the direction of effect was significantly less concordant than expected by chance: 1 of 8 genes (P = 0.004), 10 of 55, and 122 of 388 for sets appearing in ≥4, ≥3, and ≥2 studies, respectively.

“for each of these gene signatures, the direction of effect was significantly less concordant than expected by chance (1 of 8 [P = 0.004], 10 of 55, and 122 of 388 for those in ≥4 studies, ≥3 studies, and ≥2 studies, respectively).”

There was no significant effect of Drug versus NonDrug signatures within M2c on loss or gain of connectivity in controls compared to schizophrenia cases (Kolmogorov-Smirnov test P=0.054).

“We did not observe any significant effect (Kolmogorov–Smirnov test: P = 0.054).”

There was no significant association between clinical/technical covariates and module eigengene values from control and schizophrenia networks at FDR < 20% (range of Pearson's r: −0.16 to 0.21).

“There was no significant association at FDR < 20% (range of Pearson’s r: −0.16 to 0.21).”

For control subjects, the same cross-validation yielded a median connectivity correlation between discovery and replication cohorts of 0.80 (mean 0.78, SD 0.07, IQR 0.70-0.84).

“The median Pearson correlation of connectivity values among the “discovery” and “replication” cohorts was 0.77 (mean = 0.76, sd = 0.06, 25th percentile = 0.73, 75th percentile = 0.78) and 0.80 (mean = 0.78, sd = 0.07, 25th percentile = 0.70, 75th percentile = 0.84) for cases with SCZ and controls, respectively.”

The blue (M2c) module showed no association with any clinical/technical covariate at P < 0.1.

“We found no association of the blue (M2c) module with any covariate at P < 0.1”

The distribution of differential expression significance (−log10 P) did not differ significantly between M2c genes with and without a drug signature (Kolmogorov-Smirnov test P=0.54).

“We did not find a significant difference in the distribution of −log10 P values for genes that have or do not have drug signature (drug versus non-drug: Kolmogorov–Smirnov test: P = 0.54).”

In the macaque antipsychotic treatment analysis, no genes were differentially expressed after multiple test correction.

“While no genes were considered differentially expressed after multiple test correction”

No convincing patterns were found for the primary or exploratory gene set enrichment hypotheses after correction for multiple set testing, adjustment for confounders such as gene size, and combining results over multiple enrichment tests.

“we found no convincing patterns to the primary or exploratory hypotheses, after correction for multiple set testing, confounders such as gene size, and after combining results over multiple enrichment tests.”

Knockdown of FURIN in human neural progenitor cells yields abnormal migration.

“knockdown of FURIN in human neural progenitor cells yields abnormal migration”

• Also stated in: § Functional dissection of genes highlighted · plain

  “FURIN knockdown in the hiPSC NPCs resulted in significantly decreased total radial migration for all three individuals”

About 20% of schizophrenia risk loci contain variants that could contribute to altered gene expression and liability.

“Using this resource, ~20% of schizophrenia loci have variants that could contribute to altered gene expression and liability.”

• Also stated in: § eQTL signatures at SCZ risk loci point to specific genes · hedged

  “Combining the gene and isoform data, 20 of 108 GWAS loci (19%) had evidence suggesting that mis-regulated gene expression could, in part, explain the genetic association with schizophrenia”

The CommonMind Consortium sequenced RNA from dorsolateral prefrontal cortex of 258 schizophrenia cases and 279 controls, producing a resource of gene expression and its genetic regulation.

“The CommonMind Consortium sequenced RNA from dorsolateral prefrontal cortex of schizophrenia cases (N = 258) and control subjects (N = 279), creating a resource of gene expression and its genetic regulation.”

In five schizophrenia risk loci, only a single gene was implicated: FURIN, TSNARE1, CNTN4, CLCN3, or SNAP91.

“In five loci, only a single gene was involved: FURIN, TSNARE1, CNTN4, CLCN3, or SNAP91.”

• Also stated in: § eQTL signatures at SCZ risk loci point to specific genes · plain

  “For functional follow-up, we focused on the five single-gene loci encoding known proteins implicated at the gene level.”

An independent cohort yields similar differential expression results.

“and an independent cohort yields similar results”

• Also stated in: § Gene expression is subtly disrupted in schizophrenia · plain

  “there was high correlation of test statistics for differential expression in CMC compared to HBCC for the differentially expressed genes also present in the HBCC data (480 of 693), Pearson correlation r = 0.58 (P < 10−16)”

Of 693 genes showing significant case-control differential expression, their fold changes are at most 1.33.

“Of 693 genes showing significant case/control differential expression, their fold changes are ≤ 1.33”

• Also stated in: § Gene expression is subtly disrupted in schizophrenia · plain

  “All had modest fold changes (Fig. 5B), with a mean of 1.09 and range 1.03–1.33”

Of 108 schizophrenia GWAS loci, 73 harbor cis-eQTL SNPs for one or more genes at FDR ≤ 5%.

“Of the 108 SCZ GWAS loci previously reported3, 73 harbor cis-eQTL SNPs for one or more genes (FDR ≤ 5%).”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

The eQTL model identified 2,154,331 significant cis-eQTLs at FDR ≤ 5% for 13,137 (80%) of 16,423 genes.

“The model identified 2,154,331 significant cis-eQTLs, (i.e., within 1 Mb of a gene) at a false discovery rate (FDR) ≤ 5%, for 13,137 (80%) of 16,423 genes.”

Combining gene and isoform data, 20 of 108 GWAS loci (19%) had evidence that mis-regulated gene expression could partly explain the genetic association with schizophrenia.

“Combining the gene and isoform data, 20 of 108 GWAS loci (19%) had evidence suggesting that mis-regulated gene expression could, in part, explain the genetic association with schizophrenia”

32.8% of eQTL SNPs predict expression of more than one gene.

“32.8% of eQTL SNPs (“eSNPs”) predict expression of more than one gene.”

Sherlock identified co-localization of GWAS risk and eQTL association signals for 84 genes in 30 of the 108 loci (adjusted P < 0.05).

“The results suggested that GWAS risk and eQTL association signals co-localized for 84 genes in 30 of these loci (adjusted P < 0.05; Supplementary Fig. 6A, Supplementary data file 2).”

• cites [25] He, X. et al. Sherlock: detecting gene-disease associations by matching patterns of expression QTL and GWAS. Am. J. Hum. Genet. 92, 667–680 (2013). sourceNot checked yet

  Not checked in the free run.

The CommonMind Consortium cohort is an order of magnitude larger than prior RNA sequencing studies.

“We sequenced a cohort of SCZ and control subjects that is an order of magnitude larger than prior RNA sequencing studies.”

Schizophrenia cases inherited an increased number of common variant alleles previously associated with schizophrenia risk (P = 1.6 × 10−8).

“As expected3, SCZ cases inherited an increased number of common variant alleles previously associated with SCZ risk (P = 1.6 × 10−8, Supplementary Fig. 1D).”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

A random experiment showed that 73 loci with cis-eQTL SNPs is consistent with chance expectation.

“To determine if 73 out of 108 loci were larger than that expected by chance, we conducted an experiment that randomly chose such loci in the genome; it showed that 73 loci with cis-eQTL SNPs is consistent with chance expectation (data not shown).”

After removing genes where additional evaluation indicated lack of consistency, 33 genes in 18 of the 108 GWAS loci were highlighted.

“After removing genes where additional evaluation indicated lack of consistency (Supplementary Fig. 7B), there were 33 genes highlighted in 18 of the 108 GWAS loci (Supplementary data file 2).”

The CommonMind Consortium sample finds a substantial number of genes with previously undetected eQTLs.

“In addition to containing the vast majority of eQTL found in the literature, the CMC sample finds a substantial number of genes with previously undetected eQTL (Table 1).”

93% of eQTLs showed the same direction of allelic effect when intersecting CommonMind Consortium eQTLs with HBCC eQTLs.

“Moreover, concordance of the direction of allelic effect was high, with 93% of eQTL showing the same direction of effect when intersecting CMC eQTL (FDR ≤ 5%) with even a liberally defined set of HBCC eQTL (FDR ≤ 20%).”

The pattern of large observed differential expression in small studies is not seen when the sample size is raised to 250.

“Notably, this pattern not seen when the N is raised to 250.”

No obvious unifying biological theme emerges from pathway enrichment analysis of the differentially expressed genes and isoforms.

“No obvious unifying biological theme emerges from this set of genes and isoforms on the basis of pathway enrichment analysis (Supplementary data file 4).”

Modeling allele frequency differences and predicted allele effects on expression shows that the expected distribution of differential expression across genes closely matches the observed distribution in the CMC data.

“Modeling the differential in allele frequencies and the predicted effect of alleles on gene expression demonstrates that the distribution of expected differential expression, across genes, is quite similar to the observed distribution from the CMC data (Fig. 7A).”

• Also stated in: § DISCUSSION · boosted

  “Our model demonstrates that the distribution of expected differential expression, across genes, is quite similar to the observed distribution from the CMC data (Fig. 7).”

• AssumptionRests on an assumption: Case-control differences in allele frequencies for schizophrenia risk variants are small.

The analyses show that smaller studies would have notably larger variability in differential expression estimates.

“Our analyses show that these studies would have notably larger variability”

Using PGC schizophrenia allele frequencies, 11,784 cases and 11,784 controls would be needed for 80% power to detect a significant case-control difference in FURIN expression.

“For example, 11,784 cases and 11,784 controls would be needed to have 80% power to detect a significant case-control difference in FURIN expression.”

Hierarchical clustering of differentially expressed genes showed case-control distinctions independent of institution, sex, age at death, ethnicity, and RIN.

“As expected, hierarchical clustering of the differentially expressed genes showed case-control distinctions but were independent of institution, sex, age at death, ethnicity, and RIN (Fig. 5A).”

Studies with only 25 cases and 25 controls can yield estimates of differential expression exceeding twofold even when no gene is differentially expressed.

“even when no gene is differentially expressed, studies with only 25 cases and 25 controls can lead to estimates of differential expression exceeding twofold.”

Only five genes in the MHC locus were ranked highly by Sherlock and passed evaluation for concordance: C4A, HCG17, VARS2, HLA-DMB, and BRD2.

“Nevertheless, only five genes in this locus were ranked highly by Sherlock and passed evaluation for concordance of associations (Supplementary data file 2): C4A, HCG17, VARS2, HLA-DMB, and BRD2.”

Modulation of FURIN, TSNARE1, and CNTN4 led to a significant increase in apoptotic cells in the head region corresponding to head size measurements.

“we performed TUNEL staining on injected embryos, and determined that modulation of all three target genes led to a significant increase in apoptotic cells in the head region corresponding to our head size measurements (N = 20 per experiment; P < 0.001; Fig. 3E, F).”

The study found no evidence of case-control differential expression among genes implicated by GWAS risk for schizophrenia.

“In contrast, we did not find evidence for case-control differential expression among the implicated GWAS risk genes.”

There is a strong correlation between schizophrenia risk alleles and up-regulation of C4A expression (Spearman's ρ = 0.66, P < 10−16).

“we found a strong correlation between the risk alleles for SCZ and up-regulation of expression of C4A (complement component 4A; Spearman’s ρ = 0.66, P < 10−16).”

• cites [32] Sekar, A. et al. Schizophrenia risk from complex variation of complement component 4. Nature 530, 177–183 (2016). sourceNot checked yet

  Not checked in the free run.

Gene co-expression is organized into modules of distinct cellular and functional categories.

“Consistent with prior studies of the brain transcriptome15,47–50, we find gene co-expression to be organized into modules of distinct cellular and functional categories (Supplementary data file 7).”

• cites [15] Zhang, B. et al. Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer's disease. Cell 153, 707–720 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [47] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011). sourceNot checked yet

  Not checked in the free run.

• cites [48] Torkamani, A., Dean, B., Schork, N.J. & Thomas, E.A. Coexpression network analysis of neural tissue reveals perturbations in developmental processes in schizophrenia. Genome Res. 20, 403–412 (2010). sourceNot checked yet

  Not checked in the free run.

• cites [49] Oldham, M.C. et al. Functional organization of the transcriptome in human brain. Nat. Neurosci. 11, 1271–1282 (2008). sourceNot checked yet

  Not checked in the free run.

• cites [50] Roussos, P., Katsel, P., Davis, K.L., Siever, L.J. & Haroutunian, V. A system-level transcriptomic analysis of schizophrenia using postmortem brain tissue samples. Arch. Gen. Psychiatry 69, 1205–1213 (2012). sourceNot checked yet

  Not checked in the free run.

Gene sets from prior genetic studies are enriched in the M2c module, including the ARC protein complex, FMRP targets, neuronal markers, PSD proteins, and NMDA receptors.

“Gene sets identified in prior genetic studies that highlighted certain neurobiological functions are also enriched in the M2c module, including the activity-regulated cytoskeleton-associated (ARC) protein complex, targets of fragile X mental retardation protein (FMRP), neuronal markers, post-synaptic density (PSD) proteins, and NMDA receptors (Fig. 6A).”

TACR3 is the most highly up-regulated protein-coding gene (1.24-fold).

“The most highly up-regulated protein-coding gene is tachykinin receptor 3 (TACR3, NK3 receptor, 1.24-fold, Fig. 5D).”

IGF2 is the most strongly down-regulated gene (1.33-fold).

“Insulin-like growth factor 2 (IGF2), the most strongly down-regulated gene (1.33-fold, Fig. 5D)”

FURIN knockdown in hiPSC neural progenitor cells significantly decreased total radial migration for all three individuals (C1: 1.16-fold, C2: 1.23-fold, C3: 1.22-fold).

“FURIN knockdown in the hiPSC NPCs resulted in significantly decreased total radial migration for all three individuals (C1: 1.16-fold decrease, P < 0.0017; C2: 1.23-fold decrease, P < 3 × 10−6; C3: 1.22-fold decrease, P < 2 × 10−6) (Fig. 4).”

The M2c module is enriched for axon guidance, postsynaptic membrane, transmission across chemical synapses, and voltage-gated potassium channel complexes.

“In particular, the M2c module is enriched for multiple categories, including axon guidance, postsynaptic membrane, transmission across chemical synapses, and voltage-gated potassium channel complexes (Fig. 6C).”

Four modules harbor an excess of differentially expressed genes.

“Four modules stand out in harboring an excess of differentially expressed genes (Fig. 6A, Supplementary data file 6).”

A set of 1,584 genes lacked any cis-eQTL at FDR ≤ 5% yet had at least one significant isoQTL.

“There were, however, 1,584 genes having no cis-eQTL (FDR ≤ 5%) that nevertheless had at least one significant isoQTL.”

Applying the stricter trans-eQTL filter raised the HBCC replication rate to 36% at FDR ≤ 5% in both cohorts.

“This stricter filter increases the replication rate in HBCC to 36% at FDR ≤ 5% in both cohorts.”

The P value distribution for PRS association with expression showed inflation consistent with a non-uniform distribution (π1 = 0.22).

“there was inflation of the P value distribution101 consistent with a non-uniform distribution (π1 = 0.22).”

• cites [101] Storey, J.D. & Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. USA 100, 9440–9445 (2003). sourceNot checked yet

  Not checked in the free run.

• AssumptionRests on an assumption: The CMC cases carry an aggregate common polygenic schizophrenia risk burden.

In independent processing of the age 20-60 sub-cohort (172 samples, mean age 45.5) and the age 60+ sub-cohort (362 cases and controls, mean age 77.8), the differential expression t-statistics correlated at r = 0.18 (P < 2 × 10−16).

“The differential t-statistics between these independent sub-cohorts were correlated (r = 0.18, P < 2 × 10−16)”

After removing trans-eQTL in which the eSNP lay within 10 Mb of the associated gene, the final subset had 13,114 trans-eQTL (about 7% of all trans-eQTL SNP-gene pairs at P < 10^-5) across 661 genes.

“Additionally excluding trans-eQTL where the eSNP was within 10 Mb of the associated gene (since such scenarios are perhaps instances of cis-eQTL for regions with larger LD blocks), yielded a final reduced subset of 13,114 trans-eQTL (~7% of all trans-eQTL SNP-gene pairs at P < 10−5) across 661 genes.”

For each rodent signature, directional concordance was significantly below chance (1 of 8 [P = 0.004], 10 of 55, and 122 of 388 for ≥4, ≥3, and ≥2 studies).

“However, for each of these gene signatures, the direction of effect was significantly less concordant than expected by chance (1 of 8 [P = 0.004], 10 of 55, and 122 of 388 for those in ≥4 studies, ≥3 studies, and ≥2 studies, respectively).”

• AssumptionThe authors state this as an assumption: The directional-concordance hypergeometric test assumes that drug-responsive up- and down-regulated genes were drawn at random from genes up- or down-regulated in the CMC SCZ differential expression tests.

In 20 random splits, the discovery cohort had a median of 216 genes with FDR < 5% differential expression (mean = 315, sd = 261, 25th percentile = 92, 75th percentile = 562).

“Across the 20 samplings, the median number of FDR < 5% differentially expressed genes was 216 (mean = 315, sd = 261, 25th percentile = 92, 75th percentile = 562).”

The clozapine signature showed significant directional concordance (22 of 31 genes concordant, one-sided P = 0.013).

“while the clozapine signature showed significant concordance, despite the lack of enrichment in the previous test (22 of 31 genes concordant, one-sided P = 0.013).”

• AssumptionThe authors state this as an assumption: The directional-concordance hypergeometric test assumes that drug-responsive up- and down-regulated genes were drawn at random from genes up- or down-regulated in the CMC SCZ differential expression tests.

In the monkey drug analysis, multiple-testing correction left no gene as differentially expressed.

“While no genes were considered differentially expressed after multiple test correction”

The correlation between control PRS-expression t-statistics and full-cohort case-control expression t-statistics was positive but small (Pearson r = 0.095, p < 10−16).

“there was a significantly positive, but small, correlation (Pearson r = 0.095, p < 10−16) between the independent t-statistics for the effect of PRS on expression in controls and those we found for case-control expression differences in the full cohort”

• AssumptionRests on an assumption: The CMC cases carry an aggregate common polygenic schizophrenia risk burden.

For GWAS loci, intersection with DLPFC-expressed genes left 87 loci covering at least one gene and 489 genes in total.

“After intersecting with DLPFC-expressed genes from this paper, there were 87 loci spanning one or more genes, encompassing a total of 489 genes.”

In the controls-only PRS expression analysis, no gene reached significance after FDR correction at 5%.

“While no single gene was found to be significantly associated with PRS after correction for multiple testing using an FDR approach (cutoff of 5%)”

• AssumptionRests on an assumption: The CMC cases carry an aggregate common polygenic schizophrenia risk burden.

The haloperidol signature was significantly less concordant in direction than expected by chance (35 of 237 genes concordant, one-sided P < 1 × 10−4).

“We found the haloperidol signature significantly less concordant than expected by chance (35 out of 237 genes concordant, one-sided P < 1 × 10−4)”

• AssumptionThe authors state this as an assumption: The directional-concordance hypergeometric test assumes that drug-responsive up- and down-regulated genes were drawn at random from genes up- or down-regulated in the CMC SCZ differential expression tests.

Two rodent signatures showed significantly lower directional concordance than chance after multiple-testing correction (Pbonferroni = 0.01 and 0.01).

“though two signatures were significantly less concordant than expected by chance after multiple test correction (Pbonferroni = 0.01 and 0.01, Supplementary Table 2B).”

• AssumptionThe authors state this as an assumption: The directional-concordance hypergeometric test assumes that drug-responsive up- and down-regulated genes were drawn at random from genes up- or down-regulated in the CMC SCZ differential expression tests.

For CNV regions, conditioning on DLPFC expression left 12 regions spanning 127 genes.

“After conditioning on DLPFC expression, the 12 regions spanned 127 genes.”

After Bonferroni correction, no rodent gene signature reached strict significance (Pbonferroni > 0.05).

“none of the gene signatures were strictly significant after Bonferroni correction (Pbonferroni > 0.05)”

For genes with FDR < 5% in discovery, the median correlation of t-statistics between discovery and replication cohorts was 0.79 (mean = 0.75, sd = 0.16, 25th percentile = 0.67, 75th percentile = 0.88).

“For these FDR < 5% “discovery” differentially expressed genes, the median Pearson correlation of t-statistics with the “replication” cohort was 0.79 (mean = 0.75, sd = 0.16, 25th percentile = 0.67, 75th percentile = 0.88).”

The number of samples with mRIN P values below 0.05 and 0.01 was smaller than expected by chance.

“There were 17 and 3 samples with P value < 0.05 and < 0.01, respectively. One would expect a total of 29 and 5 samples to have P values of this magnitude by chance alone.”

After multiple-set correction, adjustment for confounders like gene size, and combining enrichment tests, neither primary nor exploratory gene sets showed convincing patterns.

“we found no convincing patterns to the primary or exploratory hypotheses, after correction for multiple set testing, confounders such as gene size, and after combining results over multiple enrichment tests.”

In the CMC cohort, the differential expression t-statistics for the 172 youngest cases and controls (mean age 45.5, range 20-60) correlated with those from the full cohort at Pearson r = 0.62.

“for the 172 cases and controls whose age of death was youngest (mean age 45.5, range 20–60), the t-statistics for differential expression were highly correlated with those from the full cohort (Pearson r = 0.62)”

If 5% of the 693 differentially expressed genes were false positives, the threshold of 34.7 genes would be exceeded in 9 out of 1,000 permutation experiments, slightly less than 1%.

“If 5% of the 693 were false, the threshold established of 34.7 genes is exceed in 9 out of 1000 experiments, slightly less than 1%.”

Testing a disease-by-genotype interaction yielded only a few significant associations after FDR correction.

“Nevertheless, to potentially identify eQTL that differ by disease state, a disease-genotype interaction term was also explicitly tested, but only a handful of such associations were found to be significant after controlling for FDR.”

The correlation for the youngest subset was not significantly lower than that for 100 random subsets of the same size (mean r = 0.69, SD = 0.13; P = 0.28).

“yet somewhat lower (though not significantly, P = 0.28) than for 100 random subsets of the same size (mean r = 0.69, standard deviation = 0.13)”

At nominal P < 0.05, the M0c, M16c, M6c, M26c, M28c, M32c, M7s, and M12s module eigengenes were associated with Institution, RIN, or Library batch.

“At nominal P value < 0.05 we found an association of M0c, M16c, M6c, M26c, M28c, M32c, M7s and M12s MEs with Institution, RIN or Library batch.”

In this cohort, the WGCNA module-label permutation method depended strongly on module size (quadratic-model fitting index r2 > 0.95).

“The permutation approach implemented in the WGCNA package (module label permutation in the test network) shows a strong dependency on module size in our cohort (fitting index r2 > 0.95 based on a quadratic model).”

There was no significant effect of Drug versus NonDrug signatures within the M2c module on loss or gain of connectivity in controls compared with schizophrenia cases (Kolmogorov-Smirnov P=0.054).

“We also explored whether “Drug” versus “NonDrug” signatures within the M2c module show a different effect for loss or gain of connectivity in controls compared to SCZ. We did not observe any significant effect (Kolmogorov–Smirnov test: P = 0.054).”

Permuted sets also showed large module-size differences in network statistics.

“In the permuted sets, we again observed large differences of the network statistics with module size.”

No significant association was found between clinical or technical covariates and module eigengene values at FDR < 20% (range of Pearson r: −0.16 to 0.21).

“There was no significant association at FDR < 20% (range of Pearson’s r: −0.16 to 0.21).”

There was no significant difference in the distribution of −log10 P values for differentially expressed genes with versus without a drug signature (Kolmogorov-Smirnov P=0.54).

“We did not find a significant difference in the distribution of −log10 P values for genes that have or do not have drug signature (drug versus non-drug: Kolmogorov–Smirnov test: P = 0.54).”

In the M2c module, genes with de novo mutations had higher intramodular connectivity than genes not associated with schizophrenia (t=5.1, P=2.9×10−6).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

In cross-validation, the median Pearson correlation of connectivity between discovery and replication cohorts was 0.77 for schizophrenia cases and 0.80 for controls.

“The median Pearson correlation of connectivity values among the “discovery” and “replication” cohorts was 0.77 (mean = 0.76, sd = 0.06, 25th percentile = 0.73, 75th percentile = 0.78) and 0.80 (mean = 0.78, sd = 0.07, 25th percentile = 0.70, 75th percentile = 0.84) for cases with SCZ and controls, respectively.”

In the M2c module, genes associated with PGC SCZ2 GWAS loci had higher intramodular connectivity than genes not associated with schizophrenia (t=2.6, P=0.013).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

The blue (M2c) module showed no association with any covariate at P < 0.1.

“We found no association of the blue (M2c) module with any covariate at P < 0.1, indicating that our differential co-regulated results are not biased from clinical or technical covariates.”

In the M2c module, there was no significant difference in intramodular connectivity between genes with CNVs and genes not associated with schizophrenia (t=0.88, P=0.4).

“We found a significant effect for PGC SCZ2 GWAS loci (T test: t = 2.6; P = 0.013) and de novo mutations (T test: t = 5.1; P = 2.9 × 10−6) but no CNVs (T test: t = 0.88; P = 0.4), where genes associated with SCZ have higher intramodular connectivity.”

Schizophrenia risk is highly polygenic, arising from many genetic variants each with small effects.

“SCZ is truly complex and appears to result from a myriad of genetic variants exerting small effects on disease risk4,5, conforming closely to a classical polygenic model.”

• cites [4] Purcell, S.M. et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 460, 748–752 (2009). sourceNot checked yet

  Not checked in the free run.

• cites [5] Purcell, S.M. et al. A polygenic burden of rare disruptive mutations in schizophrenia. Nature 506, 185–190 (2014). sourceNot checked yet

  Not checked in the free run.

Post-mortem expression studies in schizophrenia suggest subtle abnormalities across several brain regions and cell types.

“Post-mortem gene expression studies of SCZ cases suggest subtle abnormalities in multiple brain regions including the prefrontal and temporal cortices, hippocampus, and several specific cell types8.”

• cites [8] Horváth, S., Janka, Z. & Mirnics, K. Analyzing schizophrenia by DNA microarrays. Biol. Psychiatry 69, 157–162 (2011). sourceNot checked yet

  Not checked in the free run.

Existing expression studies are probably underpowered to detect the subtle effects expected from a complex disease in tightly regulated brain tissue.

“These studies are probably underpowered to detect subtle effects that might be expected to arise as a result of this complex disease and within tightly regulated brain tissue9, among other limitations of existing microarray-based gene expression studies10.”

• cites [9] Mistry, M., Gillis, J. & Pavlidis, P. Meta-analysis of gene coexpression networks in the post-mortem prefrontal cortex of patients with schizophrenia and unaffected controls. BMC Neurosci. 14, 105 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [10] Hitzemann, R. et al. Introduction to sequencing the brain transcriptome. Int. Rev. Neurobiol. 116, 1–19 (2014). sourceNot checked yet

  Not checked in the free run.

PCR validation showed high correlation, r > 0.5, with normalized RNA-seq expression for most genes tested.

“Validation using PCR showed high correlation (r > 0.5) with normalized expression from RNA-seq for the majority of genes assessed (Supplementary Fig. 3).”

Schizophrenia GWAS SNPs occur more often than chance in cis-regulatory functional genomic elements such as enhancers or eQTL SNPs.

“Indeed, GWAS SNPs associated with SCZ risk occur more often than expected by chance in cis-regulatory functional genomic elements, such as enhancers or eQTL SNPs3,21–24.”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

• cites [21] Roussos, P. et al. A role for noncoding variation in schizophrenia. Cell Rep. 9, 1417–1429 (2014). sourceNot checked yet

  Not checked in the free run.

• cites [22] Richards, A.L. et al. Schizophrenia susceptibility alleles are enriched for alleles that affect gene expression in adult human brain. Mol. Psychiatry 17, 193–201 (2012). sourceNot checked yet

  Not checked in the free run.

• cites [23] Trynka, G. et al. Chromatin marks identify critical cell types for fine mapping complex trait variants. Nat. Genet. 45, 124–130 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [24] Bharadwaj, R. et al. Conserved higher-order chromatin regulates NMDA receptor gene expression and cognition. Neuron 84, 997–1008 (2014). sourceNot checked yet

  Not checked in the free run.

In the CMC cohort, ethnicity was similar between schizophrenia cases and controls.

“Ethnicity was similar between cases and controls (Caucasian 80.7%, African-American 14.7%, Hispanic 7.7%, East Asian 0.6%, Supplementary Figs. 1B, C).”

Combining gene and isoform data, 20 of 108 schizophrenia GWAS loci, about 19%, had evidence that mis-regulated gene expression could partly explain the genetic association with schizophrenia.

“Combining the gene and isoform data, 20 of 108 GWAS loci (19%) had evidence suggesting that mis-regulated gene expression could, in part, explain the genetic association with schizophrenia: 18 cis-QTL loci (cis-eQTL for 33 genes + 2 genes with cis-isoQTL), one locus implicated only by cis-isoQTL (SNX19), and one trans-eQTL association for IMMP1L at a GWAS locus on chr7.”

Human FURIN mRNA rescued the phenotype induced by either morpholino, providing evidence for specificity.

“Importantly, expression of human FURIN mRNA could rescue the phenotype induced by either morpholino, providing evidence for specificity (Supplementary Fig. 8).”

The possibility of cell-type-specific effects does not compromise the majority of the study's results.

“We do not expect, however, that the scenario will compromise the bulk of our results, all of which complement the genomic studies of this disease.”

The larger differential expression in earlier studies is consistent with Winner's Curse or false positives because those studies had much smaller sample sizes.

“because earlier studies tend to be far smaller in sample size, their larger differential expression is consistent with either the well-known “Winner’s Curse”41 or false positives that may occur in smaller samples.”

• cites [41] Xiao, R. & Boehnke, M. Quantifying and correcting for the winner's curse in genetic association studies. Genet. Epidemiol. 33, 453–462 (2009). sourceNot checked yet

  Not checked in the free run.

The dysregulation of M2c in schizophrenia was not due to medication effect or clinical and technical confounds.

“The dysregulation of M2c in SCZ is not due to medication effect or clinical and technical confounds.”

Hierarchical clustering of differentially expressed genes showed case-control distinctions that were independent of institution, sex, age at death, ethnicity, and RIN.

“As expected, hierarchical clustering of the differentially expressed genes showed case-control distinctions but were independent of institution, sex, age at death, ethnicity, and RIN (Fig. 5A).”

Changes in FURIN, TSNARE1, and CNTN4 expression levels induce subtle neuroanatomical variation in multiple brain regions.

“Taken together, the data support the hypothesis that changes in FURIN, TSNARE1, and CNTN4 expression levels induce subtle neuroanatomical variation in multiple brain regions.”

Schizophrenia is polygenic.

“Our findings show schizophrenia is polygenic and highlight the utility of this resource for mechanistic interpretations of genetic liability for brain diseases.”

Thousands of samples are needed to ensure 80% statistical power to observe differential expression between cases and controls for genes implicated at schizophrenia-associated eQTLs.

“Finally, our results imply a need for thousands of samples to ensure 80% statistical power to observe differential expression between cases and controls for the genes implicated at SCZ-associated eQTL, e.g., the five genes of interest above.”

The results from smaller studies that report large differential expression are called into question.

“This calls into question results from smaller studies that report large differential expression.”

The cross-validation results strongly support the robustness of the differential expression results.

“This strongly supports the robustness of the differential expression results described herein.”

Genome-wide RNA-seq-based quantification gives good estimates of true gene expression in DLPFC tissue.

“We thus conclude that the genome-wide RNA-seq-based quantification provides good estimates of true gene expression in DLPFC tissue.”

Age at death has only a modest impact on case-control expression differences among adult cohorts.

“suggesting that age at death may have only a modest impact among adult cohorts.”

Sherlock's underlying principle is that genetic-driven expression changes in risk genes should also appear as genetic association of the same SNPs with disease.

“Its underlying principle is that genetic-driven changes in expression levels of risk genes (discovered as eSNPs) should ultimately also manifest as genetic association of those same SNPs with disease (GWAS SNPs).”

• cites [25] He, X. et al. Sherlock: detecting gene-disease associations by matching patterns of expression QTL and GWAS. Am. J. Hum. Genet. 92, 667–680 (2013). sourceNot checked yet

  Not checked in the free run.

Analysis using the mRIN software concluded that the RNA quality of the samples was adequate.

“We therefore conclude the RNA quality of the samples is adequate.”

In the M2c module, genes associated with schizophrenia have higher intramodular connectivity than genes not associated.

“where genes associated with SCZ have higher intramodular connectivity.”

The replication of connectivity in the cross-validation used a completely independent sub-cohort comprising 20% of the brain samples.

“since this replication process occurs in a completely independent sub-cohort of 20% of the brain samples.”

The high connectivity correlations between discovery and replication cohorts strongly support the robustness of the gene-gene correlation structure.

“This strongly supports the robustness of the gene-gene correlation structure, since this replication process occurs in a completely independent sub-cohort of 20% of the brain samples.”

The analysis provides additional evidence that density loss in schizophrenia is not driven by medication effects.

“This analysis provides additional evidence that the density loss in SCZ is not driven by medication effects.”

At least some of the schizophrenia case-control expression differences in CMC may be driven by underlying genetic differences between cases and controls.

“consistent with at least some of the SCZ case-control differences in CMC perhaps being driven by underlying genetic differences between the SCZ cases and controls.”

Case-control expression differences show some consistency across the lifespan.

“arguing for some consistency of case-control differences across the lifespan.”

The differential co-regulated results are not biased by clinical or technical covariates.

“indicating that our differential co-regulated results are not biased from clinical or technical covariates.”

The results do not support the hypothesis that drugs drive loss of density in schizophrenia modules through altered transcript abundance of target genes.

“Therefore, our results do not support the hypothesis that drugs drive the loss of density through alteration in the transcript abundance of target genes.”

The CommonMind Consortium resource is useful for mechanistic interpretations of genetic liability for brain diseases.

“highlight the utility of this resource for mechanistic interpretations of genetic liability for brain diseases”

Gene co-expression implicates a network relevant for schizophrenia.

“Gene co-expression implicates a network relevant for schizophrenia.”

The findings show that schizophrenia is polygenic.

“Our findings show schizophrenia is polygenic”

More than one hundred genetic loci contain variants associated with schizophrenia.

“Over 100 genetic loci harbor schizophrenia associated variants”

• Also stated in: § INTRODUCTION · plain

  “>100 common variants3”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

Schizophrenia results from many genetic variants with small effects, fitting a polygenic model.

“SCZ is truly complex and appears to result from a myriad of genetic variants exerting small effects on disease risk4,5, conforming closely to a classical polygenic model.”

• cites [4] Purcell, S.M. et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 460, 748–752 (2009). sourceNot checked yet

  Not checked in the free run.

• cites [5] Purcell, S.M. et al. A polygenic burden of rare disruptive mutations in schizophrenia. Nature 506, 185–190 (2014). sourceNot checked yet

  Not checked in the free run.

The CommonMind Consortium resource can serve as a catalogue of regulatory variants underlying the molecular basis of schizophrenia and other brain disorders.

“This resource can serve as a useful catalogue of regulatory variants underlying the molecular basis of SCZ and other brain disorders.”

PCR validation showed high correlation (r > 0.5) with normalized RNA-seq expression for most genes assessed.

“Validation using PCR showed high correlation (r > 0.5) with normalized expression from RNA-seq for the majority of genes assessed (Supplementary Fig. 3).”

Prior gene expression studies are probably underpowered to detect subtle effects expected from schizophrenia and tightly regulated brain tissue.

“These studies are probably underpowered to detect subtle effects that might be expected to arise as a result of this complex disease and within tightly regulated brain tissue9, among other limitations of existing microarray-based gene expression studies10.”

• cites [9] Mistry, M., Gillis, J. & Pavlidis, P. Meta-analysis of gene coexpression networks in the post-mortem prefrontal cortex of patients with schizophrenia and unaffected controls. BMC Neurosci. 14, 105 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [10] Hitzemann, R. et al. Introduction to sequencing the brain transcriptome. Int. Rev. Neurobiol. 116, 1–19 (2014). sourceNot checked yet

  Not checked in the free run.

GWAS SNPs associated with schizophrenia risk occur more often than expected by chance in cis-regulatory functional genomic elements, such as enhancers or eQTL SNPs.

“Indeed, GWAS SNPs associated with SCZ risk occur more often than expected by chance in cis-regulatory functional genomic elements, such as enhancers or eQTL SNPs3,21–24.”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

• cites [21] Roussos, P. et al. A role for noncoding variation in schizophrenia. Cell Rep. 9, 1417–1429 (2014). sourceNot checked yet

  Not checked in the free run.

• cites [22] Richards, A.L. et al. Schizophrenia susceptibility alleles are enriched for alleles that affect gene expression in adult human brain. Mol. Psychiatry 17, 193–201 (2012). sourceNot checked yet

  Not checked in the free run.

• cites [23] Trynka, G. et al. Chromatin marks identify critical cell types for fine mapping complex trait variants. Nat. Genet. 45, 124–130 (2013). sourceNot checked yet

  Not checked in the free run.

• cites [24] Bharadwaj, R. et al. Conserved higher-order chromatin regulates NMDA receptor gene expression and cognition. Neuron 84, 997–1008 (2014). sourceNot checked yet

  Not checked in the free run.

Ethnicity was similar between schizophrenia cases and controls.

“Ethnicity was similar between cases and controls (Caucasian 80.7%, African-American 14.7%, Hispanic 7.7%, East Asian 0.6%, Supplementary Figs. 1B, C).”

Post-mortem gene expression studies suggest subtle abnormalities in multiple brain regions and cell types in schizophrenia.

“Post-mortem gene expression studies of SCZ cases suggest subtle abnormalities in multiple brain regions including the prefrontal and temporal cortices, hippocampus, and several specific cell types8.”

• cites [8] Horváth, S., Janka, Z. & Mirnics, K. Analyzing schizophrenia by DNA microarrays. Biol. Psychiatry 69, 157–162 (2011). sourceNot checked yet

  Not checked in the free run.

Prior analyses using brain eQTL datasets derived from older technologies pointed to fewer than a handful of associations between schizophrenia GWAS risk loci and gene expression.

“Prior analyses using brain eQTL datasets derived from older technologies have pointed to less than a handful of such associations3.”

• cites [3] Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511, 421–427 (2014). sourceNot checked yet

  Not checked in the free run.

The data support that changes in FURIN, TSNARE1, and CNTN4 expression levels induce subtle neuroanatomical variation in multiple brain regions.

“Taken together, the data support the hypothesis that changes in FURIN, TSNARE1, and CNTN4 expression levels induce subtle neuroanatomical variation in multiple brain regions.”

The study's modeling results call into question findings from smaller studies that report large differential expression.

“This calls into question results from smaller studies that report large differential expression.”

An assessment of age at death or cell type proportions suggests these variables do not explain significant differential expression.

“An assessment of the impact of age at death or cell type proportions suggests that these variables do not explain significant differential expression (Supplementary Fig. 12).”

Because genome-wide surveys test many genes, variability in differential expression estimates can produce large observed differential expression even when no gene is differentially expressed.

“and because genome-wide surveys test a large number of genes, that variability can translate into large observed differential expression”

The results imply a need for thousands of samples to ensure 80% statistical power to observe differential expression for genes implicated at schizophrenia-associated eQTLs.

“Finally, our results imply a need for thousands of samples to ensure 80% statistical power to observe differential expression between cases and controls for the genes implicated at SCZ-associated eQTL, e.g., the five genes of interest above.”

The RNA quality of the samples is adequate.

“We therefore conclude the RNA quality of the samples is adequate.”

Expression of human FURIN mRNA rescued the phenotype induced by either morpholino, providing evidence for specificity.

“Importantly, expression of human FURIN mRNA could rescue the phenotype induced by either morpholino, providing evidence for specificity (Supplementary Fig. 8).”

Genes identified by differential expression analysis display various degrees of cell-type specificity.

“The data suggest that genes identified by DE analysis display various degrees of cell-type specificity.”

RNA-seq quantification across the genome yields accurate estimates of true gene expression in dorsolateral prefrontal cortex tissue.

“We thus conclude that the genome-wide RNA-seq-based quantification provides good estimates of true gene expression in DLPFC tissue.”

The Pitt and non-Pitt groups differ in ways other than age.

“other factors besides age also differ between the Pitt and non-Pitt samples”

Among adult cohorts, age at death appears to have at most a small influence on differential expression.

“suggesting that age at death may have only a modest impact among adult cohorts”

Case-control differences in gene expression show some consistency across the lifespan.

“arguing for some consistency of case-control differences across the lifespan”

Some CMC case-control expression differences may reflect genetic differences between cases and controls.

“consistent with at least some of the SCZ case-control differences in CMC perhaps being driven by underlying genetic differences between the SCZ cases and controls.”

• AssumptionRests on an assumption: The CMC cases carry an aggregate common polygenic schizophrenia risk burden.

Cross-validation provides strong evidence that the differential expression findings are robust.

“This strongly supports the robustness of the differential expression results described herein.”

The results do not support the hypothesis that drugs drive loss of density through altered transcript abundance of target genes.

“Therefore, our results do not support the hypothesis that drugs drive the loss of density through alteration in the transcript abundance of target genes.”

This analysis provides additional evidence that density loss in schizophrenia is not driven by medication effects.

“This analysis provides additional evidence that the density loss in SCZ is not driven by medication effects.”

The high connectivity correlations support the robustness of the gene-gene correlation structure.

“This strongly supports the robustness of the gene-gene correlation structure, since this replication process occurs in a completely independent sub-cohort of 20% of the brain samples.”