General description of prefrontal cortical miRNA expression
From the 265 distinct, human miRNAs included on our array, 244 were detected (1.5-fold over background) in the PFC tissue of ≥60% of the study subjects. These included robust detection of miRNAs previously known to be expressed in the brain (for example, let-7a to let-7i) as well as brain-specific miRNAs (for example, miR-124a and miR-125b) (Additional data file 1) [11
miRNA expression in schizophrenia versus unaffected comparison subjects
Assuming a false discovery rate (FDR) of 5%, 16 miRNAs were differentially expressed in PFC of schizophrenia subjects (n = 13) or schizoaffective disorder (n = 2) versus PFC of 21 psychiatrically unaffected individuals (Table ). Of the 16 distinguished miRNAs, 15 were expressed at lower (fold change 0.63 to 0.89) and one at higher (fold change 1.77) levels than in the psychiatrically unaffected comparison subjects. A heat map based on cluster analysis illustrates the differentiated expression levels of these probes (Figure ). Controlling on brain pH, postmortem interval (PMI), and hemisphere, and excluding the two subjects with schizoaffective disorder from the analyses did not substantially affect these results (Additional data file 2).
Differentially expressed miRNAs from the prefrontal cortex of subjects with schizophrenia compared to psychiatrically healthy subjects
An miRNA expression map shows differentiated genes as determined by SAM analysis. Yellow indicates low expression and blue indicates high expression, relative to the median.
Quantitative RT-PCR verification of microarray results
The expression levels of 12 selected miRNAs were also determined by quantitative RT-PCR (qRT-PCR) in our lab (Additional data file 3). For the eight miRNAs distinguished by being expressed at lower microarray levels in schizophrenia samples versus comparison samples, seven were also expressed at lower levels with qRT-PCR (Figure ). For four of the seven, the difference in expression was significant with p < 0.05, consistent with microarray findings for the same miRNAs. The eighth miRNA, miR-7, was found by qRT-PCR to have higher levels in schizophrenia than comparison subjects, but the difference in expression was not significantly different between groups (p = 0.23); we have not determined a cause for this one discrepancy of PCR versus microarray results. We also compared expression of four miRNAs that were not differentially expressed in the microarray results, and found none to be differentially expressed by qRT-PCR (p > 0.05).
Figure 2 miRNA microarray fold changes can be compared with delta-delta C(t) functions of qRT-PCR data (see Materials and methods). The comparisons are over four samples from schizophrenia patients and four samples from psychiatrically unaffected comparison subjects. (more ...)
Effect of haloperidol exposure on miRNA expression
Since all of the schizophrenia subjects were treated or had previously been treated with antipsychotics and none of the psychiatrically unaffected subjects were reported to have such a treatment history, we endeavored to evaluate the effect of antipsychotic treatment on miRNA expression. We compared expression of 179 rat miRNAs in haloperidol-treated and -untreated rats. With a FDR of 5% we found that three miRNAs were expressed at higher levels in the haloperidol-treated rats: miR-199a, miR-128a, and miR-128b. None of these miRNAs was differentially expressed in the PFC of schizophrenia patients (Additional data file 4).
miRNA and Affymetrix U133A probe relationships
We considered whether the observed pattern of lower expression of some miRNAs in schizophrenia subjects was related to lower pri-miRNA transcription. We adopted the previously published method of Thomson and colleagues [29
], where the pri-miRNA expression was determined from existing archived mRNA microarray results from the PFC of the same study subjects. A total of 52 of the miRNAs included in this study could be mapped to a primary transcript that was present among the mRNA transcripts accessible with the Affymetrix U133A array. All but three of the miRNAs with corresponding U133A probes were from the introns of protein-coding genes (host genes). The mean expression of only two of the Affymetrix U133A probes was significantly different between groups (ELM2
hosting miR-330 with p
= 0.03; MYH6
hosting miR-208 with p
= 0.03). However, these differences were not significant after correction for multiple comparisons (p
We then focused on the five miRNAs expressed at significantly lower levels in schizophrenia that also had a U133A probe that included the pri-miRNA transcript (miR-26b, miR-9-3p (alias miR-9*), miR-24, miR-7, and miR-30e). The ratio of mature miRNA to primary miRNA transcripts was lower for schizophrenia versus controls for all 5 miRNAs, and the difference in ratios reached statistical significance for 3 of the 5 (miR-26b, p = 0.009; miR-9-3p, p = 0.002; and miR-24, p = 0.037). For the one miRNA that was expressed at a significantly higher level in schizophrenia subjects, miR-106b, the ratio was also significantly higher (p = 0.003 and p = 0.006 for the two associated Affymetrix pri-miRNA probes). In the remaining 46 miRNAs with a corresponding Affymetrix U133A probe for their pri-miRNA transcripts, the ratio of miRNA to host mRNA was significantly lower for two pri-miRNAs (primary transcripts for miR-218, p = 0.021, and miR-9, p = 0.006) and significantly higher for five (miR-482, p = 0.015; miR-190, p = 0.018; miR-105, p = 0.02; miR-148b, p = 0.027; miR-218, p = 0.02). Thus, we found that the miRNA:U133A probe ratios of the schizophrenia group were significantly different from those of the comparison group for 4 of the 6 differentiated miRNAs but only 7 of the 46 nondifferentiated miRNAs (p = 0.013, Fisher's exact test) (Additional data file 5).
Common motifs near the pre-miRNA:pri-miRNA junction
We hypothesized that the system regulating processing of the pri-miRNA to pre-miRNA might involve a motif within the pri-miRNA and upstream of the single-stranded RNA (ssRNA)-double-stranded RNA (dsRNA) junction that would lend selectivity to this process. Specifically, we hypothesized that an upstream motif of some kind is shared by the 15 miRNAs that were found to be downregulated in our tests of schizophrenia PFC samples. To seek bioinformatic indications, we focused on source pre-miRNAs that were isolated (no other pre-miRNAs within 1,000 bases), yielding 11 distinguished, isolated pre-miRNAs: miR-7-1, miR-7-2, miR-7-3, miR-9-1, miR-9-2, miR-9-3, miR-26b, miR-30a, miR-30b, miR-30d, miR-30e. Of these, miR-9-1 and miR-30a can yield two mature miRNAs; the others yield one. Furthermore, miR-7-2, miR-9-2, miR-9-3, miR-30b, and miR-30d are intergenic, and the others are intronic in coding genes.
Using a combination of approaches, we found the motif UGAGNCUU upstream of pre-miRNA sequences for miR-26b, miR-30a, miR-30b, and miR-7-1. We also found GUCNCUUC upstream of pre-miRNAs miR-9-1, miR-9-2, miR-9-3, miR-7-3, and miR-30e. Thus, both 8 nt motifs are found upstream of 9 of the 11 isolated, distinguished pre-miRNAs. Lastly, instances of UGUUNNAAGAUG were found upstream of pre-miRNAs for miR-30d and miR-7-2 at the same distance, 108 bases, and not within 500 bases upstream of any other human, isolated pre-miRNAs. For displays of the motifs and the bases between motifs and junctions, see Additional data file 6; clustering of the number of bases in each such interval is displayed in Figure . Bioinformatic searches by us have found neither shared motifs that are positioned at similarly clustered distances from the junctions nor strong general homology among the 11 upstream regions.
Figure 3 Regarding the 11 isolated miRNAs distinguished in schizophrenia, this figure shows the distances (numbers of bases) from shared 5' motifs we discovered (two 8 nt and two 12 nt motif sequences in the pri-miRNA) to the ssRNA-dsRNA junctions at starts of (more ...)
Importantly, the same 8 nt motifs UGAGNCUU and GUCNCUUC are absent from the 500-base 5' regions of most undistinguished pre-miRNAs. That is, the same motifs are also upstream of only 13 isolated, undistinguished pre-miRNAs among a total of 192 isolated, human pre-miRNAs, and some of the 13 are sequentially similar as mature miRNAs to the 11 distinguished ones. However, carefully designed and executed in vivo experiments would be needed to determine whether the above or any other motifs are actually functional; the above motifs are intriguing, but their bioinformatic properties are certainly not a proof of common regulation of coordinated pre-miRNA excision.