As new studies continue to elucidate the link between structural variants and disease, these variants are increasingly associated with multiple distinct diseases, complicating their diagnostic utility. For example, a recurrent deletion on chromosome 15q13.3 has been associated with intellectual disability, schizophrenia, autism, and idiopathic generalized epilepsy, while another deletion on chromosome 16p11.2 has been associated with intellectual disability, obesity, schizophrenia, and sporadic cases of autism.
In an environment where high-density genotyping arrays and high-throughput sequencing allow researchers to associate rare variants with numerous genomic disorders, how can geneticists use this information for clinical diagnosis or counseling?
That is the question that a team of researchers from the University of Washington and Signature Genomics tried to tackle in a recent study. Described in a paper in the New England Journal of Medicine last month, the researchers settled upon a "multiple-site model" to aid in identifying rare variants that may, in combination with additional variants associated with particular conditions, be contributing to patients' disorders.
Lead author Santhosh Girirajan, now an associate professor of biochemistry and molecular biology and anthropology at Penn State, said that the study evolved out of research he conducted in Evan Eichler's lab at UW in Seattle. In particular, Girirajan and colleagues found that a deletion on 16p12.1 was enriched in cases with neurodevelopmental disorders, but was inherited from a parent 95 percent of the time, not de novo.
"It was confusing for us," Girirajan told BioArray News. "If the parent is apparently normal, and the child is affected, there should be something going on," he said. Girirajan and colleagues decided to look at other site variants in the genome, and found a "sizable number" that were larger than 500 kilobases.
"They were rare, they were large, and they are also seen in individuals with the 16p12.1 deletion," said Girirajan. "When we went back to the parents, we found out that many of them had phenotypes that weren't so obvious that they would go to a clinic," he said. "But in addition to that deletion, having this extra CNV pushed the child toward having a more severe condition."
After refining these findings into the multiple-site model, which explains how multiple copy number variants can interact to cause more severe disease, Girirajan and partners at UW and Signature wanted to see how the hypothesis generalized to other variants.
"The next question was, there were so many CNVs associated with different phenotypes, could we see that there was a correlation between multiple CNVs and the variability of these phenotypes in terms of severity or manifestation?" said Girirajan. "Can you say that you need one hit to cause something, and you need another hit that leads you to autism or a different hit that leads you to schizophrenia or different outcomes?"
To answer those questions, the team analyzed the genomic context of 72 large, rare copy number variants known to be associated with a genomic disorder or potentially associated with disease. They also examined the relationship between phenotypic severity and the total size and number of copy number variants.
The study drew upon Signature's data on 32,587 samples from children who had been tested from 2008 through 2010 and diagnosed with developmental delay with or without congenital malformations. The data was produced using Signature's SignatureChipWG, a whole-genome bacterial-artificial-chromosome microarray, and SignatureOS, an oligonucleotide-based microarray manufactured by Agilent Technologies and Roche NimbleGen.
To assess CNVs for a specific phenotype, they examined children with sporadic autism as part of the Simons Simplex Collection, a repository of genetic samples from 2,700 families, each of which has one child affected with an autism spectrum disorder, and unaffected parents and siblings. In that part of the analysis, they generated calls for 841 probands, 1,651 parents, and 793 siblings using Illumina 1M and 1M-Duo arrays.
According to the paper, from the original set of 32,587 samples, the researchers focused on 2,312 children who were determined to have one of the 72 primary-site copy-number variants. These CNVs mapped to 39 distinct genomic regions and were associated with clinical features involving various organ systems, including developmental delay or intellectual disability, autism, cardiac abnormalities, speech deficits, craniofacial features, and other previously defined congenital malformations.
The authors observed "considerable variation" in the phenotypes associated with several recurrent copy-number variants. The finding was complicated by the identification of apparently normal or mildly affected carrier parents with certain CNVs, "suggesting that these variants are critical but not sole determinants of phenotype," they wrote in the paper.
Additional analysis led the team to propose that a combination of rare and disruptive variants can contribute to different phenotypic outcomes, including intellectual disability, epilepsy, autism, and schizophrenia. They distinguished between primary mutations that make persons susceptible to a particular disease from secondary mutational events that modify the outcome and severity of that disease.
Another factor highlighted by the research team was the "mode of inheritance" for both kinds of mutations. "Even after excluding known pathogenic variants, we found that children with two or more rare and large variants of unknown significance were eight times as likely to be classified as having developmental delay as were population controls," the authors wrote. They also found a bias toward maternal transmission of second-site variants, and proposed that males, by having one X chromosome, were more vulnerable than females to the effects of large variants.
"An average male will require fewer mutational events to cross the threshold to disease," they wrote.
The authors concluded that the "overall burden of genes that are affected by large variants may eventually be of prognostic usefulness, allowing clinicians to better anticipate long-term outcomes when the variants are discovered in affected persons."
Still, Jay Ellison, medical director at Signature Genomics, cautioned that it was too early for the model to be applied clinically.
"This model provides a partial explanation for the variable expressivity seen with certain genomic disorders," Ellison told BioArray News. "From a clinical perspective, it cannot yet offer much from a predictive standpoint," he said, as "much more work needs to be done to identify which secondary CNVs influence the phenotype, before using this type of information for individual cases."
Girirajan agreed that more work needs to be done.
"This is only paving the way toward more understanding of how people should get away from thinking every disorder behaves like Mendelian disorders and think about it in a whole-genome sense," Girirajan said. "When you look at complex orders, like autism, it is hard to say that one gene or genic region is causing autism," he said. "There are some syndromic genes, but there is always going to be some genetic background that is modifying the phenotypic outcome."