Researchers from Brigham and Women's Hospital have identified a mutation in the glucocorticoid-induced transcript 1 gene that is associated with poor response to commonly prescribed inhaled glucocorticoid asthma therapies.
The variant only accounts for about six and a half percent of the overall variability in response to glucocorticoid treatment, however, so the researchers are now expanding their search to find additional SNPs that could be combined into a predictive panel.
Using a genome-wide association and family-based screening method, the group profiled subjects randomly assigned to glucocorticoids in the Childhood Asthma Management Program clinical trial and identified the functional variant rs37973 in GLCCI1 as having a "significant" association with treatment response. They then replicated this pharmacogenetic association in 935 subjects from four independent populations, the group reported in a paper published last month in the New England Journal of Medicine.
The single SNP may serve as a foundation for a panel of markers that could accurately predict which patients are more or less likely to respond well to treatment, Kelan Tantisira, one of the paper's lead authors, told PGx Reporter.
Since the mutation only explains around six and a half percent of the overall variability in response, the researchers are now using Bayesian statistical approaches to try to reveal additional SNPs that may have a "multiplicative effect" and could be built into a predictive signature, said Tantisira.
"If you looked at the figures on the relative response for all of the four independent replication populations, the wild type was certainly much, much higher in response, and the mutant allele really hovered around zero response," he said. "But standard error bars suggest that not all the variability was accounted for."
Tantisira said that from an ideal pharmacogenomic prediction perspective, a test should account for at least 50 percent of response variability. "Having said that, though," he added, "[6.5 percent] is really huge compared to other complex trait findings out there [from genome-wide association studies]."
"It's not big enough in and of itself to do true prediction, but it is substantial."
Finding a reliable predictor for response to asthma medication is a pressing goal for the Brigham and Women's Pharmacogenetics of Asthma Treatment research group, Tantisira said, because "between 20 and 30 percent of patients really don't respond to corticosteroids at all."
Tantisira and his colleague Scott Weiss received $9.8 million in NIH funding for the asthma PGx project last year as part of the Pharmacogenomics Research Network (PGx Reporter 7/8/2010).
While some pharmacogenetics projects focus on the genetic underpinnings of adverse drug reactions, the Brigham and Women's team has been focusing on drug efficacy, "primarily because it is lack of response to drugs that actually leads to morbidity in asthma," Tantisira said.
"If you're not taking the right medicine, those are [the patients] who repeatedly end up in the ER or worse. So that really is the motivation," he said.
The Brigham & Women's study follows another asthma PGx study published in NEJM in August. That paper discussed the results from a Phase II study that identified a protein biomarker predictive of response to Roche's investigational drug lebrikizumab, an interleukin-13 inhibitor intended for patients who don't response to inhaled corticosteroids (PGx Reporter 8/10/2011).
A Family-Based Approach
Tantisira and his colleagues initially set out with the hypothesis that a genome-wide association study would identify novel variants associated with glucocorticoid response. But because of the limited subject numbers available in the treatment arms of asthma clinical trials, they also adopted a family-based screening algorithm approach, described by Harvard researcher Kristel Van Steen in 2005 in Nature Genetics, which takes advantage of parental data to increase the statistical power to detect associations, Tantisira explained.
The researchers performed genotyping and family-based analysis on 118 white, non-Hispanic parent-child trios in the CAMP trial who were assigned to treatment with the inhaled glucocorticoid budesonide, the group reported in their paper. They evaluated the 100 highest powered SNPs identified by the family-based statistics for associations with treatment response, and chose 13 for replication, of which 12 were successfully genotyped in four additional patient cohorts from four separate asthma studies.
The group found only one SNP, rs37972 on the GLCCI1 gene, associated with decreased treatment response in three of the four replication populations. Treatment response was measured by the subjects' forced expiratory volume in one second.
To validate that they had found a functional mutation, the team resequenced GLCCI1, finding that another variant, rs37973, was in complete linkage disequilibrium with rs37972, and was a down-regulator of GLCCI1 expression.
Rs37973 was associated with significantly less expiratory volume, they found, with patients homozygous for the wild-type allele having improved response to glucocorticoids as compared to those homozygous for the mutant allele, which was "consistent with the initial rs37972 association," the authors wrote.
After adjusting for age, sex, height, and population, the researchers arrived at an odds ratio of 1.52 for a poor response with rs37973, meaning subjects who were homozygous mutants were about two and a half times as likely to have a response in the lowest quartile as those who were wild-type homozygous, the group reported.
Overall, the group characterized the variant as representing about 6.6 percent of the overall variability of treatment response.
According to Tantisira, while this is too low to serve as a predictive marker on its own, the researchers are already making some progress in finding additional mutations that could have an additive or even multiplicative effect on that percentage.
"We are currently working on trying to identify other factors using different types of genome-wide approaches," said Tantisira. "We only looked at the top one hundred variants, and there are certainly others potentially of interest that we could also look at in our replication populations."
The researchers are also looking "at how genes interact with each other," he said. "Can we find a few variants that together, instead of predicting only six or two percent, perhaps that six percent interacting with the two might actually explain twelve percent of the variability?"
Tantisira said the team has been taking a Bayesian statistical approach to map the genes most closely associated with that outcome "to try and figure out can we generate a predictive network."
The researchers have already found a promising connection between GLCCI1 and a variant previously reported in the CRHR1 gene. "We have noted a multiplicative effect just between GLCCI1 and that older variant," Tantisira said.
"[This] suggests even just limiting to those two SNPs, we can do much better in terms of predictive value than either alone."
In an accompanying editorial on the group's study, Jeffrey Drazen, editor in chief of NEJM, wrote that the clinical effect of the SNP is "appreciable but not overwhelming" and noted that around 16 percent of the population would be homozygous for the genotype responsible for diminished response.
He advocated that the next step — for the Brigham and Women's team, as well as the Roche team working on lebrikizumab — should be to mount clinical trials where patients are stratified based on the response-associated mutation.
Tantisira said that a clinical trial using the GLCCI1 SNP alone might be premature at this point. "It would all depend on your outcome parameters and how you specifically define them," he said.
The biggest challenge, he noted, is "the huge group in the middle" that is difficult to correlate with a response genotype.
"If you are limiting things to the phenotypic as well as genotypic extremes, I think you have a much better chance of actually being able to significantly say, 'Yeah, there could be something here doing a prospective trial,'" he said.
However, "if you include phenotypic heterogeneity and say, 'We'll take all comers,' it might be more difficult to justify doing a prospective trial at this time."
Have topics you'd like to see covered in Pharmacogenomics Reporter? Contact the editor at mashford [at] genomeweb [.] com.