3 August 2012. The rare genetic glitches with large contributions to schizophrenia risk may lie on the rarest end of the spectrum, according to a study published online August 2 in the American Journal of Human Genetics. From David Goldstein’s lab at Duke University in Durham, North Carolina, the study tries to validate 5,155 variants identified by sequencing 166 schizophrenia cases in a follow-up cohort of 2,617 cases and 1,800 controls, but comes up empty-handed: though some variants occurred exclusively in cases, none significantly associated with the disorder. This suggests that these variants are very rare, indeed, and determining whether they constitute true genetic risk factors requires larger sample sizes, along with gene-centered methods that combine variants hitting the same gene.
Schizophrenia appears to have its roots in a mixed economy of genetic factors, ranging from common variants that only slightly increase risk to rare ones with large effects (see SRF related news story). The rare-but-nasty category has been dominated by copy number variants (CNVs) in which segments of DNA containing many genes are deleted or duplicated. Next-generation sequencing promises to identify similarly rare single base changes, termed single nucleotide variants (SNVs), that could implicate a single gene. But sequencing the genomes or protein-encoding exomes of people with schizophrenia reveals thousands of variants (see SRF related news story). This has left researchers tossing about for principled ways of discriminating the causal from the inconsequential.
The new study tries one approach, in which the investigators cull a set of variants identified through exome sequencing, then look for additional supportive evidence in a separate, follow-up cohort large enough to validate any "moderately rare" variants. Defined as variants with a minor allele frequency of 1-5 percent, these occur more frequently than what is typically deemed "rare" (minor allele frequency <1 percent), but less frequently than the common variants (minor allele frequency >5 percent) explored in genomewide association studies (GWAS).
Follow the variants
First author Anna Need and colleagues sequenced exomes or genomes of 166 people with schizophrenia. This discovery sample largely consisted of treatment-resistant cases, and came from Finland or the United States. Of the 337,312 coding variants that turned up in sequencing, the researchers focused on the rarer ones, with a minor allele frequency of 5 percent or less, or those occurring exclusively in cases when compared to a control group of 307 sequenced samples. Of these, they selected those likely to change protein function, with the variant introducing a new stop codon or destroying a stop codon (nonsense); landing in a splice site; occurring within CNV regions already linked to schizophrenia or other neurodevelopment disorders; and resulting in an amino-acid change (missense) that was ranked as "probably damaging" by a tool that predicts protein function. This procedure winnowed the variants down to 5,788 for further exploration, 5,155 of which were genotyped successfully in a follow-up cohort.
Within the discovery cohort, none of the 5,788 variants was overrepresented in cases compared to controls with the study-wide level of significance corrected for multiple testing (p <1.5 x 10-7), but 428 did meet a lower criterion of p <0.05. Because true schizophrenia-related variants might lie among the false positives in this group, the researchers included these 428 in the set of 5,155 variants genotyped in the follow-up sample of 2,617 cases (which was not dominated by treatment-resistant cases) and 1,800 controls. No variant significantly associated with schizophrenia in the follow-up cohort, or when combining the follow-up sample with the discovery sample.
An exclusive club, for now
Because true risk variants that are scarcer than "moderately rare" would not be expected to reach statistical significance with this sample size, the researchers also tracked the seemingly case-specific variants from the discovery cohort in the follow-up cohort. These variants consisted of "non-private" ones found twice or more in discovery cases but not in discovery controls, and found once in discovery cases but not in discovery controls. This undid the seemingly case-exclusive nature of most of these. For example, of the non-private variants found exclusively in cases in the discovery cohort, 60 percent were found also in controls in the follow-up group. Of the private ones, 39 percent were found in controls.
But a handful of variants remained, with some found in additional cases. None of the genes containing these SNVs will ring many bells for schizophrenia researchers, however. The top hit was a variant found in five cases, and not in the 2,120 follow-up controls or in 5,379 control samples from the Exome Variant Server database. This variant was a missense mutation in KL, a gene more widely known for roles in the renal and cardiovascular systems. Intriguingly, this gene is also linked to vitamin D metabolism, which recalls the epidemiological evidence for vitamin D deficiency as a risk factor for schizophrenia (see SRF related news story).
Twenty-three other variants were found in three or more cases, and not in any of the follow-up or Exome Variant Server controls. Other genes included ZNF804B, a relation of the schizophrenia risk factor ZNF804A; PCLO, which encodes a component of synaptic machinery; and EPB41L1, which encodes a protein that associates with the AMPA subtype of glutamate receptors.
The researchers conclude that their study does not offer much support for a contribution of moderately rare variants to schizophrenia risk. Still, they point out that, because they dealt only with exome SNVs, this leaves unexamined the regulatory regions highlighted by miR-137, a non-coding microRNA that regulates expression of other genes and ranks as a top hit in the largest-yet GWAS of schizophrenia (see SRF related news story). The researchers also note ethnic and clinical differences in the makeup of their discovery and follow-up cohorts, which may have muddled validation.
But if their study does begin to sketch the truth, how best to go about detecting the rare variants? Collecting larger sample sizes could certainly help, but the researchers also propose shifting from variant-centered to gene-centered analyses (Dering et al., 2011). By considering variants that affect the same gene together, these gene-based collapsing methods not only ease the difficulties of statistics on rare events, but also move closer to a description of the biological pathways disrupted in schizophrenia.—Michele Solis.
Need AC, McEvoy JP, Gennarelli M, Heinzen EL, Ge D, Maia JM, Shianna KV, He M, Cirulli ET, Gumbs CE, Zhao Q, Campbell CR, Hong L, Rosenquist P, Putkonen A, Hallikainen T, Repo-Tiihonen E, Tiihonen J, Levy DL, Meltzer HY, Goldstein DB. Exome Sequencing Followed by Large-Scale Genotyping Suggests a Limited Role for Moderately Rare Risk Factors of Strong Effect in Schizophrenia. Am J Hum Genet 2012 August 10; 91: 1-10.