Circulating miRNA Biomarkers for Schizophrenia?

The Psychiatric Genomics Consortium recently derived a significant ensemble of common variants through its landmark mega genome-wide association study of schizophrenia (1). While this provides the foundations for variant profiling and the emergence of trait biomarkers, the predictive power of polygenic risk scoring defined by receiver-operating characteristic curve analysis is reported to be below the threshold for clinical utility (2). As an alternative, other nucleic acids, proteins, and various metabolites could support predictive function or serve as “state” biomarkers in the personalized management of schizophrenia (3).

In this issue, Wei et al. (4) explore the potential of circulating serum microRNA (miRNA) levels to inform schizophrenia diagnosis and antipsychotic drug-associated remission. The authors extracted RNA from plasma collected from a large group of subjects with psychiatric illness (N=726)—mostly confirmed cases of schizophrenia—and control subjects (N=400) matched for age, gender, smoking history, and education level. The analysis of these samples was divided into discovery, validation, and follow-up cohorts (Figure 1). In the discovery phase, plasma RNA from 150 case subjects and 150 control subjects were each pooled and subjected to both small RNA sequencing (sRNA-seq) and miRNA TaqMan Low Density Array (TLDA) analysis. The authors then selected a subset of differentially expressed miRNAs (schizophrenia/control pooled plasma) with more than a twofold difference by sRNA-seq and a 10-fold difference by TLDA that were consistent by direction on each platform. These criteria yielded eight miRNAs—miR-122, miR-130a, miR-130b, miR-193a-3p, miR-193b, miR-502-3p, miR-652, and miR-886-5p—which were then analyzed separately in the 300 individual samples of the discovery cohort by targeted quantitative polymerase chain reaction (qPCR) analysis.

Under these conditions, allowing the full heterogeneity of the discovery set to be accounted for, only two molecules, miR-130b and miR-193a-3p, remained significantly altered in the schizophrenia group compared with the control group. Fortunately, the disease-associated up-regulation of these molecules was robust and was supported in a large validation sample consisting of another 400 case subjects and 213 control subjects. The elevated levels of these two miRNAs were shown to be relatively specific to schizophrenia, as no significant difference was observed between 400 control subjects and 162 subjects with nonschizophrenia psychiatric disorders, including major depression, anxiety, bipolar disorder, and other atypical psychiatric diseases. The 400 participants in the schizophrenia group were followed up after 12 months of regular treatment with the atypical antipsychotics risperidone and aripiprazole. Interestingly, the plasma levels of miR-130b and miR-193a-3p appeared to normalize with respect to the control group in patients who were in remission (>6 months, N=79/107) according to the amelioration of their symptom severity on the Positive and Negative Syndrome Scale as defined by the Remission in Schizophrenia Working Group (5). In nonresponsive individuals (N=28) who exhibited acute-phase symptoms, plasma levels of these molecules remained relatively high, suggesting that they may be indicative of psychotic illness.

While this is not the first investigation of plasma miRNA levels in schizophrenia, the scale of this project is impressive, and the results provide some promise that circulating small RNA levels could be informative about the disorder and possibly other psychiatric syndromes. Ideally, it would have been desirable to have complete genome-wide screens of each individual sample rather than pooled samples. This would have been expensive and perhaps difficult if only small amounts of plasma were available; however, it may have revealed more molecules with the potential to establish complex molecular signatures with greater capacity for discrimination in heterogeneous populations. The authors concede that this is a limitation of the study and potentially responsible for some of the discrepancy between the two analytical approaches used in the primary screening of pooled sample groups. It may also be responsible for the high attrition rate between molecules selected from pooled analysis and the qPCR. To further validate the application of miR-130b and miR-193a-3p, follow-up studies should also endeavor to determine the sensitivity and specificity of this pattern of altered miRNA expression through receiver-operating characteristic curve analysis.