3.1. MHC Class II Expression in Cingulate Cortex by Exon Array in Bipolar Disorder Compared to Controls (Experiment 1)
The data analysis for the Affymetrix Human Exon 1.0 ST array used an ANOVA with diagnosis and probesets as between subject factors. There were 11,807 transcript identifiers analyzed composed of 230,659 probesets. Affymetrix array transcript identifiers that passed this filter for diagnosis by probeset interaction are listed (Supplementary Tables 2–4), and there were 30 genes (29 known genes as ENSG00000185790 did not map to a known RefSeq gene) that passed Bonferroni multiple comparisons correction for diagnosis by probeset. The analysis was restricted to male control and bipolar patients to minimize heterogeneity of expression due to sex differences, as well as there being unequal and fewer females in the study. There were no genes that passed Bonferroni corrections for diagnosis factor.
Two MHC Class II genes (HLA-DPA1 and CD74) passed Bonferroni correction (Supplementary Tables 2–4). Post hoc analysis of HLA-DPA1 showed downregulation of HLA-DPA1 mRNA in the ACC of bipolar disorder patients relative to healthy control subjects using the Affymetrix Human Exon 1.0 ST array data (GEO Accession Number: GSE78246) (Table 4) along the entire length of the gene (Figure 2A). There was a statistically significant decrease in BD in ACC on HLA-DPA1 gene expression (p = 0.013); however, the interaction effect between diagnosis and probesets was highly significant (p = 1.6 × 10−6). The individual probesets in the exon array that were significantly decreased in bipolar subjects in ACC compared to controls were the following: Affymetrix probesets: 2950331, 2950332, 2950333, 2950336, 2950341, 2950342, 2950343, 2950346, 2950348 (Figure 2A). CD74 was significantly decreased in expression in BD compared to controls by exon array (p = 0.038, Figure 2B), and showed a highly significant interaction effect between diagnosis and probesets (p = 3.3 × 10−5). For the other MHC Class II transcript, HLA-DRB1, the transcript level was not altered in BD compared to controls (p = 0.27), nor was there a significant interaction effect.
The HLA-DRB1 diagnosis by probeset was not significant (p = 0.93). A variable expression of HLA-DPA1 between subjects by exon array probesets (Figure 3) suggested that exons 2, 3, and 4 might be sites of alternative splicing, which were investigated by PCR sequencing (see below). For validation of the exon array results, we chose additional qPCR candidates based upon microarray results shown in Supplementary Table 1 of significant evidence of diagnosis by probeset. Our validations Figure 4 were consistent for microarray and exon array fold changes (r = 0.80).
The Affymetrix exon array fold changes were consistent with the result shown for exon array of anterior cingulate cortex.