Endophenotypes: Bridging Genomic Complexity and Disorder Heterogeneity

s will be familiar to most readers of this journal, endophenotypes are relatively well-specified physiological or behavioral measures that are considered to occupy the terrain etween disease symptoms and risk genotypes. The endophenoype concept has been an enduring feature in psychiatry for nearly 0 years, since its introduction in the classic monograph by Gottesan and Shields (1). However, endophenotypes have enjoyed ccelerating popularity in recent years due to their potential assoiations with genomic data to provide a way forward for parsing omplex, heterogeneous disease phenotypes such as schizophreia. Several excellent articles on theory and data for schizophrenia ndophenotypes have appeared recently (e.g., [2,3]), but detailed onsideration of these articles is beyond the scope of this brief ommentary. Rather, we wish to note some salient issues with espect to endophenotype research for schizophrenia and other sychiatric disorders. The two articles on candidate endophenoypes relevant to schizophrenia that appear in this issue, each xcellent in their own way, represent topical illustrations of these ssues. A major appeal of endophenotypes stems from the presumpion that they are proximal and thus more tightly linked to enotypes. However, the nature of endophenotypes invoked in arious studies varies considerably. In some articles, endopheotypes are defined as specific measures in particular response ystems (e.g., prepulse inhibition [PPI], the P300 event-related otential, a cognitive task such as the n-back); in other reports, ndophenotypes are themselves broad constructs, such as neuocognitive deficits or personality traits, that could well be econstructed further into their own subendophenotypes. Tan t al. (4) reflect this varying usage in adopting the term “interediate phenotype” on the grounds that many variables typically alled endophenotypes are not hidden or occult as per the classic efinition. Such ambiguity raises considerable potential for conusion and problems with replication and could contribute to ecent negative reports about genetic effect sizes for endopheotypes (e.g., [5]). The two articles in this issue represent these differing senses of ndophenotypes. The Wessman et al. (6) study reports statistically efined broad symptom clusters in probands with schizophrenia hat relate in distinctive ways to disrupted-in-schizophrenia-1 DISC1) and neuregulin 1 (NRG1). In contrast, the Roussos et al. (7) rticle employs specific response measures—startle amplitude and PI—to study a particular catechol-O-methyltransferase (COMT) olymorphism in healthy subjects and its interaction with the COMT nhibitor tolcapone. Both articles offer significant contributions to he genotype-behavior literature, yet the endophenotype umbrella eems rather stretched to encompass both designs. We suggest that potentially useful distinction might be to retain “endophenotype” hen a specific response measure is employed (such as PPI or a ell-defined performance task) and adopt the Tan et al. (4) term