NEW YORK – An international team of researchers has used transcriptomic data from ulcerative colitis patients to develop a predicted polygenic transcriptional risk score, or PPTRS, that can identify UC-affected individuals at fivefold elevated risk of progressing to surgical resection of the large bowel.
In a paper published on Thursday in the American Journal of Human Genetics, the Georgia Institute of Technology-led team noted that 5 percent to 10 percent of people with UC require bowel resection, or colectomy, within five years of diagnosis, but that polygenic risk scores based on genome-wide association studies generally don't provide meaningful prediction of progression to surgery. However, studies of Crohn's disease have shown that gene expression profiling of GWAS-significant genes provides some stratification of risk of progression to complicated disease through transcriptional risk scoring, or TRS.
In their paper, the researchers demonstrated that a measured TRS based on bulk rectal gene expression in a cohort of UC patients had a positive predictive value approaching 50 percent for colectomy. Single-cell profiling demonstrated that the disease-associated genes were active in multiple diverse cell types from both the epithelial and immune compartments, and expression quantitative trait locus analysis identified genes with differential effects at baseline and the one-year follow-up, the researchers said. But for the most part, they found that differential expression associated with colectomy risk was independent of local genetic regulation.
Overall, their data suggested that prediction of gene expression from relatively small transcriptome datasets can be used in conjunction with transcriptome-wide association studies for stratification of risk of disease complications.
The researchers began by performing differential expression analysis between baseline rectal RNA-seq biopsies of individuals in the PROTECT multicenter pediatric inception cohort study of response to standardized colitis therapy. Analyses were done on 21 affected individuals who progressed to colectomy and 310 who did not. They identified downregulation of 783 transcripts in the individuals who underwent colectomy and upregulation of 1,405 transcripts overall.
They also obtained rectal biopsy RNA-seq data for 92 affected individuals at week 52 and observed a marked shift in gene expression at follow-up, prompting them to ask whether local regulation of the gene expression might contribute to this effect. They found that there were 72 SNPs that were significantly regulating 308 genes at both time points.
Further examination of the expression of colectomy-associated genes in a single-cell RNA-seq dataset obtained from rectal biopsies provided strong evidence that both epithelial and immune cells contributed to the risk of disease progression, the researchers said.
The researchers then performed a TWAS to capture the effects of all polymorphisms within 1 Mb of each transcript expressed in the PROTECT rectal biopsies and then used the weights to predict gene expression in a validation cohort from the UK Biobank. They tested for differential predicted gene expression in 70 percent of the validation samples and discovered about 800 genes either upregulated or downregulated in UC-affected individuals relative to non-IBD control individuals. They then derived a PPTRS for UC based on the effect sizes of the minor alleles and applied it to the remaining 30 percent of the validation samples, as well as to the PROTECT genotypes, and found that the PPTRS efficiently discriminated UC-affected individuals from non-IBD control individuals.
Significantly, it also discriminated the individuals who underwent colectomy versus those who didn't in both the UK Biobank and PROTECT.
"More extensive single-cell profiling, combined with cell-type-specific genetic analysis of gene expression, is likely to lead to the development of even better transcriptional risk signatures," the authors concluded. "It is also likely that such focused and personalized analysis may highlight specific pathological mechanisms active in particular affected individuals."
They did note, however, that these results were limited by the relatively small sample size of colectomies in the PROTECT study, and that validation of cross-ancestry assessments — and the evaluation of the consistency of gene expression prediction across populations — should be a high priority.
In an email, corresponding author and GIT researcher Greg Gibson noted that while the study's multiple layers of replication show that transcriptional profiling of the rectum greatly enhances risk stratification for risk of colectomy, this was not a clinical trial, so the approach is not yet approved for evaluation of patients.
"We hope that it will progress to implementation in the near future," he added. "The prediction from genotypes alone is less likely to have clinical utility since the precision is still quite low, so that aspect is more research oriented."
He further noted that the approach he and his colleagues used could also be applied to a wide range of diseases, and that they are pursuing that research.