The importance of endophenotypes in schizophrenia research

Abstract

Endophenotypes provide a powerful neurobiological platform from which we can understand the genomic and neural substrates of schizophrenia and other common complex neuropsychiatric disorders. The Consortium on the Genetics of Schizophrenia (COGS) has conducted multisite studies on carefully selected key neurocognitive and neurophysiological endophenotypes in 300 families (COGS-1) and then in a follow up multisite case–control study of 2471 subjects (COGS-2).

Endophenotypes are neurobiologically informed quantitative measures that show deficits in probands and their first degree relatives. They are more amenable to statistical analysis than are “fuzzy” qualitative clinical traits or confoundingly heterogeneous diagnostic categories. Endophenotypes are also viewed as uniquely informative in traditional diagnosis-based as well as emerging NIMH Research Domain (RDoC) contexts, offering a bridge between the two approaches to psychopathology classification and research. Endo- or intermediate phenotypes are heritable, and in the COGS-1 cohort their level of heritability is in the same range as is the heritability of schizophrenia itself, using the same statistical methods and subjects to assess both. Because we can demonstrate endophenotypes link to both gene networks and neural circuits on the one hand and also to real-life function, endophenotypes provide a critically important bridge for “connecting the dots” between genes, cells, circuits, information processing, neurocognition and functional impairment and personalized treatment selection in schizophrenia patients.

By connecting schizophrenia risk genes with neurobiologically informed endophenotypes, and via the use of association, linkage, sequencing, stem cell and other strategies, we can provide our field with new neurobiologically informed information in our efforts to understand and treat schizophrenia. Evolving views, data and new analytic strategies about schizophrenia risk, pathology and treatment are described in this Viewpoint and in the accompanying Special Issue reports.

Introduction

Endophenotypes have played a crucial role in advancing our understanding of the gene to phene knowledge gap of schizophrenia (e.g., Gottesman and Gould, 2003, Braff et al., 2007). The use of endophenotypes as neurobiologically informed quantitative measures is now rapidly increasing in schizophrenia research (cf. Fig. 1). Endophenotypes are quantitative laboratory based measures, hidden from the view of the “naked eye” of clinical observation. Endophenotype deficits are observed in groups of schizophrenia patients (SZ) relative to Healthy Control Subjects (HCS). First degree relatives of SZ probands show intermediate values. Gottesman and Shields' (1973) transformative concept has yielded a plethora of highly informative endophenotype studies of schizophrenia, an accelerating harvest that still continues (cf. Gottesman and Gould, 2003, Braff et al., 2007, Tan et al., 2008, Glahn et al., 2014) as illustrated by Fig. 1.

As quantitative measures, endophenotypes offer a significant statistical analytic advantage over the DSM-based qualitative and somewhat fuzzy clinical phenotypes. It is also crucial to point out that endophenotypes are viewed as crucial, strategically important measures in the NIMH Research Domain Criteria (RDoC) literature as indicated by the title of the commentary “Endophenotypes: Bridging genomic complexity and disorder heterogeneity” (Insel and Cuthbert, 2009). But are endophenotypes really simpler and more proximal to genes than diagnostic categories? I would posit that anyone who has interviewed several hundred schizophrenia patients and then administered the LNS working memory test (see Lee et al., 2015--in this issue) to the same patients would doubtlessly say that yes endophenotypes are simpler than the disorder. This admittedly face validity view is reinforced by multiple analytic strategies as discussed below (e.g., Gottesman and Gould, 2003, Braff et al., 2007). Are endophenotypes also closer to genomic substrates than fuzzy clinical diagnoses? Yes, endophenotypes do fill the gene to phene gap in our knowledge. Please see discussion below and related references (cf. Gottesman and Gould, 2003, Braff et al., 2007, Tan et al., 2008, Glahn et al., 2014). Beginning in 2003, the Consortium on the Genetics of Schizophrenia (COGS-1), characterized 300 families of schizophrenia probands with at least one unaffected sibling and both parents available for testing (cf. Calkins et al., 2007, Swerdlow et al., 2015). COGS-2 is a follow-up study expanded to include almost 2500 case–control participants, using the 12 primary and some additional COGS-1 endophenotypes, embedded in a carefully quality assured demographic, clinical and functional outcome database (e.g., cf. Calkins et al., 2007). Additional forthcoming genomic characterization and statistical genetic analyses partly using the Psychiatric Genomics Consortium (PGC) 550K PsychChip and other genomic platforms will follow.

An interesting challenge about endophenotype deficits arises in family studies. Clinically unaffected first degree relatives of schizophrenia patients have at least some level of endophenotype deficits but these relatives are not affected by schizophrenia itself. Why are these endophenotype (and putatively gene variation) carriers not affected by the clinical phenotype of schizophrenia? Perhaps clinically unaffected relatives have a subthreshold summation of genetic and non-genetic risk factors and a below threshold genomic and endophenotype burden (e.g. Glahn et al., 2014). Alternatively, risk genes may interact with protective genes and/or protective environmental factors, reflecting the dynamic interplay of multiple risk and protective genetic (G) and environmental (E) observations (Gottesman and Gould, 2003, Braff et al., 2007, Braff, 2012, Glahn et al., 2014). This issue of gene burden thresholds and opposing protective factors is perhaps one of the most crucial yet relatively unexplored areas of schizophrenia research.

The risk algorithm for schizophrenia is multifactorial and complex, and there have been long standing efforts to identify which “high risk” children with endophenotypic and genetic risk factors ultimately cross the threshold from risk and “convert” to developing a psychotic illness often after a normal or near normal appearing childhood. After a period of initial Jacobean revolutionary zeal (Braff and Braff, 2013) algorithms designed to identify which “high-risk” children do convert to having schizophrenia have undergone many iterations. In this context, some risk-associated endophenotypes such as mismatch negativity (MMN) (see Light et al., 2015, in this issue) can actually be measured in utero. Since many reports indicate that the onset of neural circuit disrupting events occurs before birth, the optimal time for effective early identification and intervention might ultimately be in the prenatal period (Swerdlow, 2011). But, prenatal pharmacological intervention based on a risk algorithm is a strategy fraught with profound ethical and practical long term medical, scientific, legal and social challenges. It is more likely that benign (low side effect profile) cognitive and sensory training interventions will be utilized early in life in a high risk population of endophenotype deficit burdened children. In this broad context, what is our present and likely future state of knowledge and future strategies of endophenotype research in schizophrenia?