29 July 2010. A clue to the possible role of the ZNF804A gene in psychotic disorders comes from a new study in the July Archives of General Psychiatry. Gary Donohoe of Trinity College in Dublin, Ireland, and colleagues found evidence that the risk variant found in genomewide association studies of psychosis, rather surprisingly, is associated with better performance on certain working and episodic memory tasks in schizophrenia, and thus could define a group of people whose cognition remains relatively intact despite their other symptoms. This group, the authors speculate, may comprise a genetically unique type of schizophrenia that arises through a distinct pathway—a notion that, if backed by future studies, may affect psychiatric research, diagnosis, and treatment.
The single-nucleotide polymorphism (SNP) rs1344706, located in an intron of the zinc finger protein 804A (ZNF804A) gene, came within a hair of being the first to cross the statistical significance finish line in the genomewide association sweepstakes, and a combined psychosis sample that included bipolar cases nudged it into genomewide significance (see SRF related news story). Further genomewide association and other studies have supported the link between rs1344706 and schizophrenia and bipolar disorder in independent samples (see SRF related news story; Zhang et al., 2010; and SZGene entry).
The gene and its protein product remain rather mysterious. Beyond the observation that the eponymous zinc finger domain suggests the protein binds DNA and thus might regulate genes, there are data to show that a putative mouse homolog is expressed in brain. Interestingly, a new study indicates that this mouse homolog is regulated by the Hoxc8 protein (Chung et al., 2010). Hox genes are generally known for their dominant roles in embryonic patterning early in development, but there is also evidence for their role in the adult brain, and even in behavioral disturbances in a possible animal model of obsessive-compulsive behavior (see SRF related news story).
The wider lens of brain activation
The current finding supports the notion that variation in ZNF804A may affect schizophrenia through patterns of brain activation. In a previous magnetic resonance imaging study of healthy subjects performing a cognitive task, Meyer-Lindenberg's group, at the University of Heidelberg in Germany, tied the risk variant to altered connectivity within and between the dorsolateral prefrontal cortex and the hippocampus, areas thought to act up in schizophrenia (see SRF related news story). While that study found no association between the SNP and task performance, a later one from the same group tied the variant not only to changes in brain activation, but also to impaired social cognition in healthy subjects (Walter et al., 2010).
Such findings caused the international team, led by first author James Walters, Cardiff University, Cardiff, Wales, to think that rs1344706 might be associated with cognitive performance in schizophrenia. They considered cognition a possible schizophrenia endophenotype, in part because a variety of cognitive deficits play a key role in the disorder (Bowie and Harvey, 2005). Performance on neuropsychological tests may relate more closely than diagnosis to pathogenesis and, hence, may offer a better window into how the illness develops.
Walters and colleagues examined the possible relationship between the genotype and cognition in patients with schizophrenia and healthy subjects in two stages. The discovery stage involved genotyping an Irish sample for rs1344706 and asking subjects to complete standard tests of cognitive function to obtain measures of IQ, working memory, episodic memory, and attention. The researchers then used analysis of variance to determine whether the SNP was associated with cognitive performance. The second stage entailed using the same methods to try to replicate, in an independent sample of German subjects, any significant associations found in stage one.
The Irish cohort consisted of 297 patients who met DSM-IV criteria for schizophrenia and 165 healthy controls. The German replication sample included 251 patients with schizophrenia, all of whom had experienced symptoms for at least six months, and 1,472 healthy controls who had been randomly selected from the community. Subjects ranged from 18 to 65 years old.
A surprise finding
Not surprisingly, each cognitive indicator—IQ, episodic memory, working memory, and attention—showed that patients had worse cognitive functioning than controls. Less expected were the findings regarding the association between the zinc finger genotype and cognitive deficits. Since the work by Meyer-Lindenberg's group had tied the A (risk) allele at rs1344706 to abnormal brain activation in areas involved in memory and to psychosis, which often comes with cognitive deficits, it stood to reason that carriers of the risk genotype should perform worse than non-carriers on cognitive tests. However, in a counterintuitive twist, carriers of the risk genotype actually performed better than non-carriers.
The study tested two main kinds of memory: working memory and episodic memory. The working memory tasks assessed both verbal and spatial working memory. In the Irish cohort, the ZNF804A genotype interacted with case-versus-control status to predict variance in both kinds of working memory. Analyses limited to control subjects, in which the genotype was only weakly related to performance, suggested that the interaction resulted mainly from the genotype’s effect in patients.
In patients, the genotype explained 2.8 percent of the variance in verbal working memory and 4.4 percent in spatial working memory. Post-hoc tests limited to patients revealed better working memory in AA homozygotes than in CC homozygotes.
The study examined two kinds of episodic memory: immediate and delayed logical memory. For immediate logical memory, genotype interacted with subject group, again reflecting an association in patients that explained almost 3 percent of the variance in performance. Post-hoc tests showed that homozygous and heterozygous risk allele carriers performed better than homozygous subjects who lacked the allele. Analyses also found a significant, but weaker, interaction for delayed episodic memory.
Turning to the German cohort, the researchers were able to replicate the significant associations found in the Irish sample. Again, genotype and case-versus-control status interacted to explain variance in scores for all of the memory tests, a finding driven by the gene’s effects in patients only. For each kind of memory, patients with the AA genotype performed best. In the German patients, genotype explained about 3 percent of the variance in spatial working memory, verbal working memory, and immediate and delayed episodic memory.
While research results seldom replicate as cleanly as they did in this study, finding less impaired cognition in carriers of an allele associated with psychosis defied expectations. In search of an explanation, Walters and colleagues hypothesized that the zinc finger gene variant might be defining a subgroup of patients with schizophrenia whose cognition had been relatively unscathed. If so, they thought, the association should be greatest in analyses restricted to subjects with higher cognitive ability. Sure enough, as IQ rose, the association between the genotype and schizophrenia grew stronger. Similar results emerged when the researchers used memory performance instead of IQ to select the highest-ability groups.
In trying to make sense of the findings, Walters and colleagues considered the possibility that they might result from spurious influences. Statistically, they were able to rule out potential confounding variables, including sociodemographic characteristics, medication, and severity of clinical symptoms. They further reasoned that the gene must not encode for cognition in general, because otherwise it would have affected performance in healthy subjects and, besides, the greatest association with schizophrenia appeared in subjects with relatively high cognitive ability. Rather, the researchers interpreted their findings as evidence that ZNF804A may contribute to a distinct kind of psychosis that develops through a different pathway than other kinds of schizophrenia that bring on greater cognitive impairment. If true, this finding could affect the usefulness of cognition as a general endophenotype for the heterogeneous disorder called schizophrenia.
Taking the results one step further, the researchers speculate that the increased connectivity between the hippocampus and the dorsolateral prefrontal cortex found in previous research “could represent a neural mechanism that spares episodic and working memory in patients by allowing processing of memory information in both structures.” Healthy subjects would not need this compensatory mechanism; therefore, only patients would benefit from it.
Finally, these findings could have implications for the classification of psychotic disorders in general and schizophrenia in particular. Taking a cue from other complex genetic diseases—such as breast cancer, in which the presence of the HER2 gene shapes research and treatment—Walters and colleagues state that their findings support subtyping patients with schizophrenia to clarify underlying molecular and biological pathways. They hint that the psychosis subtype defined by the zinc finger variant may cross diagnostic boundaries to include bipolar disorder. They write, “If confirmed, defining the molecular etiology involved in this group may have important diagnostic, prognostic, and therapeutic implications.”—Victoria L. Wilcox and Hakon Heimer.
Walters JT, Corvin A, Owen MJ, Williams H, Dragovic M, Quinn EM, Judge R, Smith DJ, Norton N, Giegling I, Hartmann AM, Möller HJ, Muglia P, Moskvina V, Dwyer S, O'Donoghue T, Morar B, Cooper M, Chandler D, Jablensky A, Gill M, Kaladjieva L, Morris DW, O'Donovan MC, Rujescu D, Donohoe G. Psychosis susceptibility gene ZNF804A and cognitive performance in schizophrenia. Arch Gen Psychiatry. 2010 Jul;67(7):692-700. Abstract