Studies before the introduction of operational diagnoses showed pooled probandwise concordances of 46 % for MZ and 14 % for DZ pairs (Gottesman & Shields, 1982). Subsequent studies employing operational diagnoses have shown probandwise concordances of 41–79 % for MZ and 0–17 % for DZ pairs (Cardno & Gottesman, 2001). Thus the results of studies from both periods are consistent with a genetic contribution to the aetiology of schizophrenia. Further issues related to the twin method and studies of heritability are discussed below.

The best study of bipolar disorder using non-operational criteria (Bertelsen, Harvald, & Hauge, 1977) showed probandwise concordances of 62 % for MZ and 8 % for DZ pairs.

Subsequently, a pilot study in Sweden based on self-report questionnaires found probandwise concordances for DSM-III-R bipolar disorder of 38 % in MZ and 4 % in DZ pairs (Kendler, Pedersen, Johnson, Neale, & Mathe, 1993).

Studies based on the Maudsley twin register in London have found probandwise concordances for DSM-IV bipolar disorder (types I and II combined) of 40 % in MZ and 5.4 % in DZ pairs (McGuffin et al., 2003) and, in the same sample, probandwise concordances for non-hierarchical Research Diagnostic Criteria (RDC) mania of 36.4 % in MZ and 7.4 % in DZ pairs (Cardno et al., 1999).

A Finnish national population register study (Kieseppä, Partonen, Haukka, Kaprio, & Lannqvist, 2004) found very similar probandwise concordances for DSM-IV bipolar disorder type I of 43 % in MZ and 6 % in DZ pairs.

Thus, the results of twin studies are consistent with a genetic contribution to the aetiology of bipolar disorder.

Adoptee studies (Heston, 1966; Rosenthal, Wender, Kety, Welner, & Schulsinger, 1971), adoptee’s family studies (Kety et al., 1994), and cross-fostering studies (Wender, Rosenthal, Kety, Schulsinger, & Welner, 1974) all show higher rates of schizophrenia and related disorders in the biological relatives of probands than controls. The data of Kety et al. (1994) from Denmark have been re-analysed applying DSM-III criteria, with consistent results (Kendler, Gruenberg, & Kinney, 1994): 3/38 (7.9 %) of first-degree biological relatives of proband adoptees had DSM-III schizophrenia, compared with 1/107 (0.9 %) of first-degree relatives of control adoptees. A further Finnish study (Tienari et al., 2003) found a higher morbid risk of DSM-III-R schizophrenia and related disorders in the adopted-away offspring of mothers with these disorders, compared with the adopted-away offspring of mothers who did not have these disorders (22.5 % vs. 4.4 %).

More recently, further evidence has come from the large Swedish population register study of Lichtenstein et al. (2009) (Table 6.1). The relative risk of schizophrenia in adopted offspring of a parent with schizophrenia was 13.7, which provides further confirmation of the elevated risk of schizophrenia in adopted-away biological relatives of those with schizophrenia.

An adoptee’s family study of bipolar disorder (Mendlewicz & Rainer, 1977) defined by Feighner criteria found a trend towards a higher rate of bipolar disorder in biological (14 %) than adoptive parents (3 %) of 29 adoptees with bipolar disorder, which became significant when a range of affective disorders in parents were included. None of the biological or adoptive parents of 22 unaffected control adoptees had bipolar disorder. In a further adoptee’s family study (Wender, Kety, & Rosenthal, 1986) using clinical diagnoses, the rates of affective disorders were non-significantly higher in biological relatives of probands with bipolar disorder (4.2 %), than in biological relatives of control adoptees (2.3 %).

The evidence from adoption studies of bipolar disorder has been notably enhanced by the Swedish study of Lichtenstein et al. (2009) (Table 6.1), which found a significantly elevated relative risk of 4.3 for bipolar disorder in adopted-away offspring of a parent with bipolar disorder.

The heritabilities of DSM-III-R and ICD-10 schizophrenia have been estimated as 88 % (95 % CI 83–92 %) and 83 % (95 % CI 74–90 %), respectively, from pooled results of twin studies (Cardno & Gottesman, 2001). In a later meta-analysis of 12 twin studies using any diagnostic classification, Sullivan, Kendler and Neale (2003) estimated the heritability of schizophrenia to be 81 % (95 % CI 73–90 %), with a small but significant common familial environmental effect of 11 % (95 % CI 3–19 %), and the remaining 8 % of variance in liability due to individual-specific environmental effects.

Subsequently, Lichtenstein et al. (2009). Estimated heritability based on their large national register study of family and adoption data. They found a somewhat lower, but still substantial, heritability of 64 % (95 % CI 62–68 %) with common environmental effects of 45 % (95 % CI 4–7 %). In a further study, based on data from the Danish national register study of parent–offspring risks (Gottesman et al., 2010), a similar heritability estimate for schizophrenia was found (67 % (95 % CI 64–71 %)) (Wray & Gottesman, 2012). (Technically, this last result was for the familiality of schizophrenia, as genetic and common environmental effects could not be separated.)

The reason for the difference from the twin study results is not clear, but taking all these results together, it can still be concluded that the heritability of schizophrenia is substantial and between 60 and 80 %, with a modest contribution from common environmental effects.

The pilot twin study in Sweden based on self-report questionnaires (Kendler, Pedersen, Neale, & Mathe, 1995) estimated the heritability of DSM-III-R bipolar disorder to be 79 %, with no common environmental effects (AE model best fitting on grounds of parsimony).

In the Maudsley twin series, the heritability of DSM-IV bipolar disorder (types I and II combined) was estimated to be 85 % (95 % CI 73–93 %) with no common environmental effects (AE model best fitting on grounds of parsimony and zero point estimate for common environmental effects with ACE model) (McGuffin et al., 2003). In the same sample, the heritability of non-hierarchical RDC mania was estimated to be 84 % (95 % CI 69–93 %) (AE model best fitting on grounds of parsimony—point estimate for common environmental effects of 10 % with ACE model) (Cardno et al., 1999).

In the Finnish register sample, the heritability of DSM-IV bipolar disorder type I was estimated to be 93 % (95 % CI 69–100 %) (AE model best fitting on grounds of parsimony—with ACE model), heritability 67 % (95 % CI 0–99 %), and common environmental effects 25 % (95 % CI 0–82 %) (Kieseppä et al., 2004). The sample on which the model fitting was based (but not the concordances given above) included two pairs where the proband had schizoaffective disorder, bipolar type, in addition to 25 pairs where the proband had bipolar disorder. The parameter estimates in this study are less precise than in the Maudsley twin series due to the smaller sample size (27 vs. 67 probandwise pairs).

Again the large family and adoption study of Lichtenstein et al. (2009) found a somewhat lower heritability estimate for bipolar disorder. For non-hierarchical diagnoses, heritability was 59 % (95 % CI 56–62 %) with common environmental effects of 3 % (95 % CI 2–6 %); and for hierarchical diagnoses, heritability was 55 % (95 % CI 51–61 %) with common environmental effects of 3 % (95 % CI 2–7 %).

Thus, as for schizophrenia, the heritability of bipolar disorder is substantial, with somewhat lower estimates in family/adoption than twin studies and with the possibility of a modest common environmental effect.

In the Maudsley twin series, heritability for the broad phenotype of any psychotic disorder has been estimated to be 90 % (95 % CI 68–94 %), with a point estimate of zero for common environmental effects (Rijsdijk, Gottesman, McGuffin, & Cardno, 2011).

All of the heritability estimates in this section were based on liability-threshold models (Falconer, 1965; Neale & Cardon, 1992).

A large number of genome-wide linkage studies of schizophrenia and bipolar disorder have been carried out, based on families with multiple affected members in different generations and on affected sibling pairs. Many candidate regions have been identified, including on chromosomes 1q, 6p, 8p, 13q, 10p, 10q, and 22q for schizophrenia and 6q, 9p, 10q, 12q, 13q, 14q, 18, and 22q for bipolar disorder (reviewed by Craddock et al., 2005) including some regions in common to both disorders, but results have been difficult to replicate consistently across studies.

A subsequent meta-analysis of 32 linkage studies of schizophrenia (Ng et al., 2009) found genome-wide significant evidence of linkage on chromosome 2q and suggestive evidence on 5q, and a secondary analysis of 22 studies of European ancestry found suggestive evidence on 8p. A meta-analysis of 11 linkage studies of bipolar disorder (McQueen et al., 2005) found genome-wide significant evidence of linkage on chromosomes 6q and 8q.

The first wave of association studies focused on genetic markers in functional candidate genes, that is, genes with a known function that had a plausible role in the aetiology of schizophrenia or bipolar disorder. A weakness of this approach was that, due to our limited knowledge about the pathophysiology of these disorders, it was difficult to find strong candidate genes. Significant associations from meta-analysis, or with some independent replication, have been found for markers in a range of candidate genes, including 5HT2a DRD2, DRD3, and DRD4 for schizophrenia and 5HTT, MAOA, COMT, and BDNF for bipolar disorder (reviewed by Craddock, O’Donovan, & Owen, 2005; Farmer, Elkin, & McGuffin, 2007; Barnett & Smoller, 2009; see also http://​www.​szgene.​org). However, none have yet been established as fully confirmed susceptibility genes.

The second wave of association studies focused on the investigation of positional candidate genes in regions of interest from linkage studies. This approach led to the identification of a number of positional candidate genes for schizophrenia, including DTNBP1 (dysbindin, 6p) (Straub et al., 2002), NRG1 (neuregulin 1, 8p) (Stefansson et al., 2002) and DAOA (D-amino acid oxidase activator, 13q) (Chumakov et al., 2002). Markers in DAOA have also been associated with bipolar disorder (Hattori et al., 2003), and in NRG1 with manic episodes accompanied by mood-incongruent psychotic symptoms (Green et al., 2005). As with the candidate gene association studies, the positional candidate association findings have been replicated in some independent samples, although the associated alleles and haplotypes have been only partly consistent across samples. Again, none have been fully confirmed as susceptibility genes, but they remain under investigation.

The third wave involves genome-wide association studies (GWAS), which have become technically feasible in recent years. The first promising GWAS result in schizophrenia was for an SNP in the zinc finger binding protein 804A gene (ZNF804A) (2q32) (O’Donovan et al., 2008). Evidence for the association strengthened when a sample of patients with bipolar disorder were included, suggesting that one or more alleles in the vicinity of ZNF804A influence risk to a broader phenotype including both disorders. These findings were further substantiated in a larger meta-analysis (Williams et al., 2011).

Subsequent increases in sample size, which resulted from increasing collaboration between research groups, have led to the discovery of genome-wide significant associations at markers in the major histocompatibility complex (MHC) region on chromosome 6 (Shi et al., 2009), and additional significant SNPs in or near neurogranin (NRGN) (11q24) and transcription factor 4 (TCF4) (18q21) (Stefansson et al., 2009).

Following the inclusion of bipolar disorder in the landmark Wellcome Trust Case Control Consortium GWAS (Consortium, 2007), larger collaborative studies of bipolar disorder identified genome-wide significant associations with markers in the alpha 1C subunit of the L-type voltage-gated calcium channel (CACNA1C) (12p13) and ankyrin 3 (ANK3) (10q21) (Ferreira et al., 2008), making these leading candidate genes for bipolar disorder.

The largest GWAS ‘mega-analysis’ of schizophrenia to date, from the Psychiatric GWAS Consortium (PGC) (Ripke et al., 2011), involved over 50,000 participants. It found genome-wide significant associations at seven loci, five of which are new (1p21.3, 2q32.3, 8p23.2, 8q21.3, and 10q24.32–q24.33) and two of which have been previously implicated (6p21.32–p22.1 and 18q21.2). The strongest of the new findings was for a marker in microRNA 137 (MIR137), which has a role in the regulation of neuronal development. The study also provided further evidence for loci that increase risk of both schizophrenia and bipolar disorder (CACNA1C, ANK3, and ITIH3–ITIH4), where the level of statistical significance increased when samples including both disorders were included in the combined analysis. In keeping with the focus of GWAS, the identified risk variants are likely to each have a small effect on risk (odds ratios around 1.1). This study did not include ZNF804A among its top findings, but this locus remains a strong candidate due to the evidence from other studies and emerging evidence of associations with endophenotypes (see below).

Concurrently, the PGC published the largest GWAS of bipolar disorder to date, involving over 60,000 participants (Sklar et al., 2011). This strengthened the evidence for CACNA1C and showed a new genome-wide significant association with a marker in ODZ4 (11q14). Association findings at ANK3 were genome-wide significant in the primary analysis, but fell short of this in the combined analysis. However, ANK3 remains a leading candidate gene for bipolar disorder.

These GWAS results provide the strongest evidence to date for statistical associations of genetic markers with schizophrenia and bipolar disorder, and many more associations are likely to be identified as sample sizes increase further.

In addition to investigations of individual genetic markers, GWAS have jointly analysed many markers to provide molecular evidence that the genetic contribution to the aetiologies of schizophrenia and bipolar disorder includes a polygenic component (i.e. due to the cumulative influence of many genetic variants of small effect) and that there is partial overlap in the polygenic contribution to schizophrenia and bipolar disorder (Purcell et al., 2009).

In addition to the long-known 22q11 microdeletion, many more chromosomal deletions or duplications between one kilobase and several megabases in size have been found in the human genome (Sebat et al., 2004; Stankiewicz & Lupski, 2010).

Two main types of study have investigated the role of such CNVs in schizophrenia and, more recently, in bipolar disorder. The first of these looks at the total genomic load of CNVs and whether CNVs are more common across the genome in cases than in controls. They have shown that large rare deletions and duplications are significantly more common in schizophrenia cases than controls (Stone et al., 2008; Walsh et al., 2008).

Secondly, studies have looked for an excess of CNVs at specific chromosomal locations. In large cohorts of patients with schizophrenia, some genomic hotspots have been found to contain structural variants associated with the disorder. CNVs at a number of different loci including 1q21, 2p16 (NRXN1), 3p26, 3q29, 5p13, 7q11, 7q22, 7q36 (VIPR2), 15q11, 15q13, 16q11, 16p13, 17p12, 17q12, and 22q11 (reviewed by Doherty, O’Donovan, & Owen, 2012; Gejman, Sanders, & Kendler, 2011) have been found to be significantly over-represented in patients with schizophrenia compared to controls. Most loci span multiple genes, but those at 2q16 (NRXN1) and 7q36 (VIPR2) only involve single genes.

Although the CNVs have relatively large effects on risk (odds ratios 2 to 26: Doherty et al., 2012), they are generally not thought to be sufficient to cause schizophrenia on their own, as most are also found occasionally in unaffected individuals.

There is evidence that the associated CNVs at specific chromosomal locations are likely to be relatively recent de novo mutations, which are selected out of the population in just a few generations. They continue to be found because of relatively high mutation rates at these loci (Rees, Moskvina, Owen, O’Donovan, & Kirov, 2011).

Studies of CNVs have also provided evidence of partial genetic overlap between schizophrenia, autism, learning disability/mental retardation, ADHD, and idiopathic generalised epilepsy because these disorders are also associated with CNVs in similar chromosomal regions (Burbach & van der Zwaag, 2009; Doherty et al., 2012; Williams et al., 2010).

Some initial studies of bipolar disorder suggested that, in contrast to schizophrenia, there was no increased burden of CNVs (Grozeva et al., 2010). However, there is now some evidence of an excess of CNVs in bipolar disorder (Malhotra et al., 2011; Zhang et al., 2009), for example, in early onset cases, and more studies are underway.

A common approach to reducing the considerable clinical heterogeneity in schizophrenia is to perform factor analysis of psychotic symptoms. This frequently results in three main quantitative psychotic symptom dimensions (positive, negative, and disorganised) (Andreasen et al., 1995; Liddle, 1987). Evidence is currently strongest for the disorganised dimension having a substantial genetic component. It has consistently shown significant familial aggregation in pairs of affected siblings (Burke, Murphy, Bray, Walsh, & Kendler, 1996; Cardno et al., 1999; Hamshere et al., 2011; Loftus, DeLisi, & Crow, 1998; Rietkerk et al., 2008) or other relatives (McGrath et al., 2009; Wickham et al., 2001) or other relatives and has an estimated heritability in twins with psychotic disorders of 84 % (95 % CI 18–93 %) (Rijsdijk et al., 2011). The genetic influences appear to be partly due to modifying effects independent of susceptibility to psychotic disorders. Studies of familial/genetic influences on other symptom dimensions have been less consistent in their results, but there is evidence of familiality in some studies (Burke et al., 1996; Cardno et al., 1999; Hamshere et al., 2011; Kendler et al., 1997; Loftus, Delisi, & Crow, 2000; Vassos et al., 2008).