Fig. 5: Exploratory: substances of abuse may be associated with similar dysregulation of neurotransmission-related gene expression in BA10 as SCHIZ and BP.

Within a set of exploratory analyses, we estimated the differential expression (Log2FC) in BA10 in our qPCR dataset associated with a variety of substances of abuse (tobacco, cannabinoids, stimulants, opioids) while controlling for diagnosis. Substance exposure was defined by indication of usage within the subjects’ clinical records, family interviews, toxicology reports, and coroners reports. To reduce false discovery due to multiple comparisons, this exploratory analysis was limited to the 20 genes with the most reliable diagnosis effects in BA10 (listed in Fig. 3). The pattern revealed by our exploratory analysis was consistent across categories of substances of abuse (Supplementary Fig. S18D), but the largest effects were observed with opioid exposure (n = 8 subjects). A Replication: opioid exposure in our BA10 qPCR dataset was associated with similar differential expression (Log2FC) to what had been observed previously in opioid use disorder in the DLPFC using RNA-Seq by Seney et al. [41] (n = 15 diagnosis-related genes, BA10 Pritzker qPCR vs. DLPFC Seney RNA-Seq: R = 0.79, P = 0.000499). B Relationship with diagnosis: In general, the effects of diagnosis on gene expression (Log2FC) in BA10 measured in our qPCR study correlated positively with the effects of a variety of substances of abuse (Log2FC; tobacco, cannabinoids, stimulants, opioids) as estimated while controlling for diagnosis in our dataset. The top scatterplot uses opioid exposure as an example to illustrate the similarity between the effects associated with substances of abuse and diagnosis in our BA10 qPCR dataset (n = 20 diagnosis-related genes, x axis: Blue = BP Log2FC, Green = SCHIZ Log2FC, opioid Log2FC vs.: BP: R = 0.90, P = 5.04e-08, SCHIZ: R = 0.91, P = 2.21e-08). The bottom scatterplot shows that the similarity between the effects of opioid use and diagnosis replicates when using the effects of opioid use disorder measured in the DLPFC Seney et al. [41]. RNA-Seq dataset (n = 15 diagnosis-related genes, opioid use disorder Log2FC vs.: BP: R = 0.89, P = 6.43e-06, SCHIZ: R = 0.76, P = 0.000678). C Opioid use appears to amplify diagnosis-related dysregulation in our BA10 qPCR dataset: Example boxplots illustrate the effect of opioid exposure within our diagnosis groups for top diagnosis-related genes (DRD4: effect of diagnosis (P = 0.0190), effect of opioids (P = 0.000208); HTR2B: effect of diagnosis (P = 9.94e-05), effect of opioids (P = 9.42e-09); SST: effect of diagnosis (P = 0.0111), effect of opioids (P = 0.000146)). Plotting conventions follow Fig. 1, with “yes” and “no” indicating exposure to opioids. Differential expression statistics are derived from the multilevel model used previously, but with opioid use added as a predictor. Differential expression results for other genes and substances of abuse can be found in Supplementary Fig. S19. Full statistical reporting for correlations can be found in Supplementary Table S9.