Toward a biaxial model of “bipolar” affective disorders: Further exploration of genetic, molecular and cellular substrates

Abstract

Current epidemiologic and genetic evidence strongly supports the heritability of bipolar disease. Inconsistencies across linkage and association analyses have been primarily interpreted as suggesting polygenic, nonMendelian and variably-penetrant inheritance (i.e., in terms of interacting disease models). An equally-likely explanation for this genetic complexity is that trait, locus and allelic heterogeneities (i.e., a heterogeneous disease model) are primarily responsible for observed variability at the population level. The two models of genetic complexity are not mutually-exclusive, and are in fact likely to co-exist both in trait determination and disease expression. However, the current model proposes that, while both types of complex genetics are likely central to observable affective trait spectra, inheritance patterns, gross phenotypic categories and treatment-responsiveness in affective disease (as well as the widespread inconsistencies across such studies) may be primarily explained in terms of a heterogeneous disease model. Gene–gene, gene–protein and protein–protein interactions, then, are most likely to serve as trait determinants and ‘phenotypic modifiers’ rather than as primary pathogenic determinants. Moreover, while locus heterogeneity indicates the presence of multiple susceptibility genes at the population level, it does not necessitate polygenic inheritance at the individual or pedigree level. Rather, it is compatible with the possibility of mono- or bigenic determination of disease susceptibility within individuals/pedigrees. More specifically, the biaxial model proposes that integration of specific findings from genetic linkage and association studies, ion channels research as well as pharmacologic mechanism, phenotypic specificity and effectiveness studies suggests that each gene of potential etiologic significance in primary affective illness might be categorized into one of two classes, according to their primary role in neuronal functioning—neuroelectrical and neurochemical. The class(es) of primary genetic alteration (i.e., neuroelectrical, neurochemical or both) determines the type, while the locus and specific allelic variant determines the direction, of pathologic trait alteration(s). In addition to the class, locus and allelic variant of the primary genetic alteration, the cellular- and system-level expressions—including functional trait interaction therein—determine the nature and degree of clinical expression of each trait. Finally, the type, direction and presence of functional interaction between pathologic alterations would indicate the most appropriate pharmacology.

Introduction

While affective diseases are genetically complex in that there is no single, population-wide monogenic cause, decades of familial segregation, twin and adoption studies support the strong heritability of affective–and especially ‘bipolar’–disorders (Matthews and Reus, 2003, Kelsoe, 2003). The majority of studies have implicated complex inheritance, and supported polygenic and/or oligogenic causation at the population level, though some segregation analyses have implicated simpler inheritance patterns in pedigrees for these disorders including autosomal dominant, recessive and X-linked (see reviews by Matthews and Reus, 2003, Homer et al., 1997, Kelsoe, 2003). Probandwise twin concordance estimates have ranged from 0.33 to 0.90 for monozygotic (MZ) twins and from zero to 0.30 for dyzygotic (DZ) twins (Matthews and Reus, 2003, Kieseppa et al., 2004, Kelsoe, 2003). Heritability estimates as high as 0.93 for bipolar (Kieseppa et al., 2004) and 0.50 for unipolar (Oswald et al., 2004) disorders have also been reported. Variability in these estimates may derive from a host of conceptual and methodologic factors, including presumed pathophysiology and genetic disease models, sampling frames, case ascertainment methods, technologic precision in genome analyses, statistical analytic methods employed and, especially, phenotype definitions. Despite significant advances in knowledge and technology and the evolution of our pathophysiologic understanding of affective disease, inconsistencies remain the rule across study outcomes.

In the post-Genomic era, the search for genetic etiologies has been dominated by large-scale linkage and association analyses, while earlier studies relied more upon candidate gene approaches, in which single genes were selected, based upon extant pathophysiologic hypotheses, for genotyping and follow-up pedigree linkage analysis. To date, the great majority of candidate gene studies–pre- and post-Genome–have examined genes related to monoaminergic pathways including receptors and transporters and metabolic enzymes (e.g., dopamine and serotonin receptors and transporters, TH, MAO, COMT) (Craddock and Jones, 2001, Ikeda et al., 2002, Matthews and Reus, 2003, Oswald et al., 2004, Ryu et al., 2004, Hattori et al., 2005). Yet, even as hypothetical and empirical orientation in affective disorders has moved downstream from neurotransmitter (NT) production and availability, toward neurotransmitter–receptor interactions, receptor complex organization, second messengers, and intracellular signaling, no study has unequivocally supported a definitive role for any single neurochemical locus across affective pathologies (Manji and Lenox, 2000, Craddock and Jones, 2001, Neves-Pereira et al., 2002, Sklar et al., 2002, Matthews and Reus, 2003, Oswald et al., 2004, Petronis, 2003, Hasler et al., 2004, Hattori et al., 2005). Although some studies have examined the potential role of calcium channels in affective disorders (Levy and Janicak, 2000, Yoon et al., 2001), presumably because of their direct involvement in neurotransmitter release, very few researchers in the post-Genomic era have returned upstream to reconsider other neuroelectrical proteins, which enjoyed earlier intuitive appeal.

Building upon earlier findings related to the potentially greater overall number of CAG repeats among bipolar patients (Lindblad et al., 1995, O'Donovan et al., 1995, O'Donovan et al., 1996, Saleem et al., 2001), Wittekindt et al. (1998) identified a specific association between a highly polymorphic CAG trinucleotide repeat (TNR) in a neuroelectrical gene–i.e., the potassium channel gene, KCNN3–among schizophrenic, but not bipolar I subjects. Between 1998 and 2001, several follow-up studies across different ethnic populations (Guy et al., 1999, Putzhammer et al., 2005, Hawi et al., 1999, Saleem et al., 2001, Bowen et al., 2000, Jin et al., 2001, Meira-Lima et al., 2001, Ujike et al., 2001), failed to replicate Wittekindt's specific CAG repeat KCNN3 association in bipolar subjects. To the best of this author's knowledge, no further studies have explored the possibility of affective disorder association with either TNR or with far more common single nucleotide polymorphism (SNP) variants in any of the 168 known Na⁺, K⁺ and Cl⁻ channel isoforms— ie., those channel proteins most directly involved in the regulation of neuronal activation.

Similarly, while physiologic investigations identified alterations in cellular transmembrane potentials and/or ATPase activity in bipolar subjects (Hokin-Neaverson and Jefferson, 1989a, Hokin-Neaverson and Jefferson, 1989b, Looney and el-Mallakh, 1997, El-Mallakh and Wyatt, 1995, Ponizovsky et al., 2003), very few genetic studies (see review, Craddock and Jones, 2001) have explored the potential etiologic significance of overlap between affective susceptibility loci and Na⁺–K⁺-ATPase isoforms. Mynett-Johnson et al., in a 1998 case-control study in an Irish population, demonstrated allelic association between bipolar I disorder and a Na⁺–K⁺-ATPase subunit gene (ATP1A3). One attempt to replicate these findings, examining candidate ATP1A3 and ATP1B3 polymorphisms in an older order Amish population failed to do so (Philibert et al., 2001). Finally, a study (Li et al., 2000) examining the association of excess CAG repeats in the ATPase beta-1 subunit gene (ATP1B), and two (Jacobsen et al., 2001, Jones et al., 2002) examining a specific (i.e., Darier's Disease) mutation at the ATPase alpha-2 subunit (ATP2A2), also did not find significant association with affective illness. No further examinations of these or any of the dozens of other ATPase genes were identified in the published literature.

Interestingly, in a very recent study, Meyer et al. (2005) conducted mutation analysis of a potassium–chloride co-transporter and follow-up case-control study in a large sample of bipolar and schizophrenic subjects. The findings demonstrated that three variants in this gene (SLC12A6) were co-inherited with schizophrenia and significantly associated with bipolar disorder.

Considering the widespread inconsistencies across candidate gene analyses alongside the weak and inconsistent associations found across large-scale, genomewide investigations (Bailer et al., 2002, Badenhop et al., 2002, Cichon et al., 2001, Segurado et al., 2003, MacIntyre et al., 2003, Ewald et al., 2003, Fallin et al., 2004, Faraone et al., 2004, Camp et al., 2005, Lambert et al., 2005, Shink et al., 2005b, Venken et al., 2005), and summarized in psychiatric genetics reviews (Craddock et al., 2001, Berrettini, 2002, Sklar, 2002, Maier et al., 2003, Anguelova et al., 2003, Matthews and Reus, 2003, Oswald et al., 2004, Castren, 2005, Hattori et al., 2005, Kendler, 2005) there is general agreement that a population-wide monogenic cause will not be identified. The failure of linkage and association studies to support Mendelian inheritance at the population level, has led many investigators to conclude that affective illness in the individual are necessarily poly- or oligogenic, quantitative, and variably penetrant in their inheritance and profoundly influenced by stochastic mechanisms and environmental factors (e.g., Berrettini, 1998, Glazier et al., 2002, Petronis, 2003, Hasler et al., 2004). This may be the consequence of premature conclusions that a disease manifesting complex inheritance at the population level (i.e., as demonstrated through association analyses in either population-based or clinical samples) necessarily derives from complex genetic etiology at the individual level. An alternative view would allow the possibility that genetic etiology in the individual may be simple (i.e., mono- or bigenic), but manifest complexly in the population, via locus, allelic and trait heterogeneity, that results in a diverse spectrum of weak association findings in population-based studies. Recognition of such individual-population distinctions in causative interpretations, as advocated by some cytogenetic researchers and computational geneticists (Pickard et al., 2005a, Pickard et al., 2005b, MacIntyre et al., 2003, Thornton-Wells et al., 2004, Thornton-Wells, 2005), may allow for more efficient and systematic pathophysiologic investigation than previously suggested, and illuminate potential trajectories for clinical approaches and drug development.

As described by Thornton-Wells et al. (2004), allelic, locus and trait heterogeneities are population-level characteristics of genetic disease that collectively comprise heterogeneous or competing disease models, as distinguished from multifactorial or interacting disease models, in which another set of characteristics, including gene–gene, gene–protein, protein–protein, and gene–environment interactions in the individual is operative (Fig. 1). Despite an apparent trend in contemporary neuropsychiatric research to assume a primary role for the latter models in neuropsychiatric disease, there have been few specific and testable hypotheses–of either set of models–presented in the literature. Moreover, exploration of the potential contribution of either set of factors requires careful prospective design and analytic specification.