Classical twin studies aim to separate the contributions of genetic predisposition (‘nature’) and environmental exposures (‘nurture’) to the expression of human complex traits. These studies compare trait similarity (i.e., concordance rate or correlation) of monozygotic (MZ) twins (genetically identical) with that of dizygotic (DZ) twins (who share 50% of their genes): a higher similarity in MZ compared with DZ twins suggests genetic effects on the trait, assuming that environmental factors are shared within twin pairs to the same extent regardless of zygosity (‘Equal Environments Assumption’) (Boomsma et al. Reference Boomsma, Busjahn and Peltonen2002). Results of classical twin studies proved to be useful in encouraging subsequent efforts to identify the specific gene variants and environmental determinants aetiologically involved in an extraordinary variety of phenotypes.

Aetiological research often uses the matched case-control design, where subjects with (‘cases’) and without (‘controls’) a given condition are made similar for a number of characteristics (‘confounders’), and are compared with respect to the prevalence of an exposure factor. A problem with this design is that cases and controls can only be matched on measured variables (e.g., age, gender, education, etc.), while there may be unmeasured variables (e.g., genetic background, intrauterine or early postnatal environmental exposures, etc.) which may contribute to the observed differences between the two groups. Apart from the randomised experiment which is not always feasible when studying humans, one of the best observational approaches to handle this problem is the ‘discordant-twin design’. This method uses, as case-control dyads, twin pairs where one twin carries the targeted condition and the co-twin does not. In this way, matching is obtained for age, a key factor in several traits, and can be easily achieved for gender, by considering MZ pairs or same-gender DZ pairs. More importantly, cases and controls are automatically matched also for genotype (totally for MZ pairs, partially for DZ pairs), as well as for in-utero exposures and family background. Within such a scenario, the main challenges to causal inference in observational research are neutralised (McGue et al. Reference McGue, Osler and Christensen2010).

When the condition of interest is a disease, the discordant-twin design compares affected twins with healthy co-twins, and allows the identification of specific aetiological factors. Furthermore, the comparison with an external control group of healthy twins makes it possible to assess if a given trait (e.g., a biological marker) is related to the genetic vulnerability to develop the disease (‘endophenotype’). Over the years, the design has been applied to various psychiatric diseases, mainly using MZ twins. In this editorial, we briefly revise some discordant-twin studies that focused on bipolar disorder (BD), coming across different domains, from neuropsychology, to neuroendocrinology, to neuroimaging, up to epigenetics and proteomics.

BD is a severe mental illness characterised by recurrent episodes of mania and depression, which affects approximately 1% of the population (Merikangas et al. Reference Merikangas, Jin, He, Kessler, Lee, Sampson, Viana, Andrade, Hu, Karam, Ladea, Medina-Mora, Ono, Posada-Villa, Sagar, Wells and Zarkov2011). Classical twin studies indicated a substantial genetic influence on BD liability, with environmental factors also playing a non-negligible role (Edvardsen et al. Reference Edvardsen, Torgersen, Røysamb, Lygren, Skre, Onstad and Oien2008). Accordingly, both genetic and non-genetic factors were targeted by aetiological research, including investigations on BD discordant twins.

In a small study on cognitive performance, Gourovitch et al. (Reference Gourovitch, Torrey, Gold, Randolph, Weinberger and Goldberg1999) found that MZ twins with BD were impaired on measures of visuospatial functioning and verbal memory, as compared with the unaffected co-twins. This suggested that some neuropsychological deficits may be consistent features of BD, related to disease itself.

A major body of research on BD regards the evaluation of candidate endophenotypes, which increase the power of genetic analysis, and also help in reshaping the classical nosological systems and diagnostic categories. Vonk et al. (Reference Vonk, van der Schot, Kahn, Nolen and Drexhage2007) investigated autoimmune thyroiditis with raised levels of thyroperoxidase antibodies (TPO-Abs) as a biological marker for the transmission of the bipolar genotype, by considering bipolar and healthy control twin pairs. Results showed significantly increased mean TPO-Abs levels in discordant versus healthy twin pairs, but no difference between bipolar patients and their non-bipolar co-twins, suggesting that autoimmune thyroiditis, with TPO-Abs as marker, is a possible BD endophenotype.

Although callosal volume reduction has been observed in BD patients, it is still unclear whether this deficit reflects genetic vulnerability to the illness. To shed light on this issue, Bearden et al. (Reference Bearden, van Erp, Dutton, Boyle, Madsen, Luders, Kieseppa, Tuulio-Henriksson, Huttunen, Partonen, Kaprio, Lönnqvist, Thompson and Cannon2011) mapped corpus callosum morphology with MRI techniques on proband BD twins, their non-BD co-twins and demographically matched control twins. The observation that probands, but not their co-twins, had significant callosal thinning and area reduction (most pronounced in the genu and splenium) relative to healthy twins was consistent with a disease-related deficit, rather than with a genetic reflection of BD vulnerability.

In recent years, research started to focus on the potential role of epigenetic factors in the aetiology of psychiatric diseases, and BD is no exception. Epigenetic mechanisms affect gene expression mainly through DNA methylation, without involving the genomic DNA sequence. In this context, disease-discordant MZ twins represent the ideal design because they allow one to explore epigenomic variation independently of genomic sequence variation. Dempster et al. (Reference Dempster, Pidsley, Schalkwyk, Owens, Georgiades, Kane, Kalidindi, Picchioni, Kravariti, Toulopoulou, Murray and Mill2011) were the first to perform a systematic genome-wide analysis of methylation differences between MZ twins discordant for BD or schizophrenia. They found considerable differences in DNA methylation between affected and unaffected twins, many of them located in the vicinity of genes previously implicated in psychosis, such as GPR24 and CTNNA2 in BD. These data support epigenetic alterations as relevant in BD and schizophrenia aetiology.

The latest advances in proteomics, combined with the MZ discordant-twin design, gave additional input to biomarkers identification for BD. The first such study is that of Kazuno et al. (Reference Kazuno, Ohtawa, Otsuki, Usui, Sugawara, Okazaki and Kato2013), who performed proteomic analysis of lymphoblastoid cells derived from one single pair of BD discordant twins. Their results indicate phosphoglycerate mutase 1 (PGAM1), involved in glycolysis, as a new candidate BD biomarker.

In conclusion, the discordant-twin design is a valuable tool for aetiological research. A future greater use of this design in the psychiatric field is desirable, to gain further relevant insights in the aetiology of major psychoses, particularly BD. The increasing potential of twin registries worldwide (Hur & Craig, 2012) will facilitate the identification of discordant twin pairs, encouraging the application of such a unique approach.

Table 1. Published discordant-twin studies on bipolar disorder

BD, bipolar disorder; MZ, monozygotic; DZ, dizygotic.