4 May 2015
May 5, 2015. Although the categories of mental illness ensconced in the Diagnostic and Statistical Manual (DSM) still serve the field, many feel that there are more biologically relevant ways of parsing these disorders. Looking into brain-related phenotypes, such as a thorough detailing of cognitive function or measures of electrical activity in the brain, may find commonalities between different diagnoses, or extract subtypes within a disorder such as schizophrenia. Two efforts to find such phenotypes, called intermediate phenotypes or endophenotypes, were on display at the International Congress on Schizophrenia Research: results from the Bipolar-Schizophrenia Network for Intermediate Phenotypes (B-SNIP) consortium, which has been looking across three disorders that share psychosis as a symptom (schizophrenia, schizoaffective disorder, and bipolar disorder), and the Consortium on Genomics of Endophenotypes in Schizophrenia (COGS).
In his plenary talk on Tuesday afternoon, March 31, Brett Clementz of the University of Georgia, Athens, presented a new categorization of psychosis based on biomarkers that index brain function collected by B-SNIP. These included measures of cognitive function, eye movement control, and responses to auditory stimulation collected from nearly 1,000 people with psychosis. Analyzing these biomarkers in combination broke psychosis into three different "biotypes" that did not map onto the diagnostic categories. Though Biotype 1 contained more schizophrenia cases and Biotype 3 more bipolar, there was good representation of all diagnoses within each biotype, Clementz said, adding, "We didn't reinvent the wheel."
The biotypes differed from each other in measures of social function, cognitive impairment, and gray matter volume. For example, Biotype 1 had the most profound cognitive and social impairments, whereas Biotype 2 showed enhanced sensorimotor reactivity, perhaps indicating a hyperaware state. Another difference came from genes: the polygenic risk score that sums a person's genetic risk factors for schizophrenia was much higher for Biotype 1 than it was for a schizophrenia diagnosis.
One audience member asked whether these biomarkers could identify subgroups within schizophrenia, but Clementz said that so far they didn't have the statistical power to find them within the smaller schizophrenia group. Another audience member wondered whether these biotypes would be stable over time, and another encouraged them to take into consideration a person's stage of illness, as some biomarkers may be state-based (reflecting consequences of being ill) and others trait-based (stemming from risk for psychosis). Preliminary analyses did not find that biotypes predicted treatment response, though Clementz said it is somewhat difficult to tell, given the participants' varied medication histories. Overall, the effort illustrates the utility of measures that don't cleanly distinguish between diagnostic categories. "We have to consider heterogeneity as our friend rather than enemy," Clementz said. "It can be leveraged to understand pathophysiology."
On Monday, March 30, an afternoon session featured the efforts of COGS to dissect endophenotypes within schizophrenia. The consortium has measured an array of phenotypes in 296 people with schizophrenia and their families (COGS-1) and 2,500 people with schizophrenia or unrelated controls (COGS-2). This kind of deep phenotyping on a moderately large scale has taken 10 years, but its proponents argue COGS is positioned to disentangle the heterogeneity of schizophrenia and extract clear insights into the nature of the disorder.
With COGS-2 just finished and a new issue of Schizophrenia Research highlighting some of its results, Larry Seidman of Harvard Medical School, Boston, Massachusetts, presented the cognition results, which focused on non-social forms of cognition, including measures of attention, vigilance, memory, and goal maintenance. People with schizophrenia in both COGS cohorts were impaired in verbal memory, working memory, and attention. In general, smokers performed worse than non-smokers did. The two COGS cohorts were similarly impaired, staving off earlier worries that the need to recruit intact families for COGS-1 might pick up milder forms of schizophrenia.
Gregory Light of the University of California, San Diego, described the addition of two electrophysiological measures to the COGS battery while the study was in progress. Adopting a simplified, 2-channel electroencephalogram (EEG) system, Light reported successfully measuring mismatch negativity (MMN) and P300 auditory event-related potentials in the COGS-2 cohort; known to be abnormal, on average, in schizophrenia, these measures showed similar impairments in the COGS-2 group. But Light noted that demographic factors such as ethnicity and smoking substantially influenced these signals and so will have to be taken into account when exploring the use of these measures.
If endophenotypes lie closer to the root causes of illness, then using them as the key phenotype may lead to the more obvious genetic signals than a schizophrenia diagnosis would. Twelve phenotypes have already been used in linkage and candidate gene studies (Greenwood et al., 2011, and Greenwood et al., 2013), but apparently the COGS-1 dataset can be mined for more phenotypes to unearth more genes. Tiffany Greenwood, also of UCSD, presented results based on another nine endophenotypes that discriminated between schizophrenia and controls, and that showed some evidence of being heritable. Candidate gene analysis, in which single nucleotide polymorphisms (SNPs) near genes already suspected in schizophrenia are assayed, highlighted 11 genes, several of which were associated with glutamate signaling. A linkage analysis of these nine phenotypes also picked up signals in regions containing candidate genes such as CSMD1 and DISC1, and glutamate-related genes such as DLGAP2 and GRIN3A. Greenwood said that a genomewide association study (GWAS) of these endophenotypes is in the works.
The symposium ended with a pep talk from David Braff of the University of California, San Diego, who said that they were on the cusp of realizing the importance of COGS. In the next three to five years, they expect to have more comprehensive genetic data in hand—including GWAS, sequencing, and epigenetic data—against which to compare their endophenotypes. He contrasted the COGS approach of deeply phenotyping 5,000 people with that of the current drive by the Psychiatric Genomics Consortium (PGC) to acquire tens of thousands of people with a minimal phenotype of a schizophrenia diagnosis, adding that COGS will help "close the gap between knowledge and wisdom." Beyond helping to pinpoint genetic contributors to schizophrenia, the COGS data could also help in understanding symptoms, vocational and social outcomes, and treatment response. The challenge, Braff said, will be figuring out how to parse this massive dataset into usable units to help patients.—Michele Solis.