NEW YORK (GenomeWeb) – A University of North Carolina-led team has demonstrated the potential prognostic importance of tumor and stroma subtypes in pancreatic cancer.
As they reported yesterday in Nature Genetics, the researchers combined microarrays with a mathematical method known as non-negative matrix factorization (NMF) to test hundreds of tumor, stroma, and normal pancreas samples. The approach helped discern between tumor, stroma, and normal cells, with tumor expression patterns distinguishing classical tumor subtypes from basal-like tumors heralding poorer outcomes.
The analysis also pointed to the potential of gleaning prognostic clues from gene expression patterns in tumor-neighboring stroma cells. For instance, stroma belonging to a 'normal' subtype seemed to help rein in tumor spread, while an 'activated' stromal subtype was associated with survival times.
"We believe these results will set the groundwork for future clinical trials, allow treatments to be assigned based on the subtypes, and guide the development of new therapies," senior author Jen Jen Yeh, a researcher affiliated with the University of North Carolina at Chapel Hill's Lineberger Comprehensive Cancer Center and the UNC School of Medicine, said in a statement.
Yeh added that the work "helps make sense of researchers' conflicting findings about stroma — that it can either promote or be a barrier to tumor spread."
Past genomic studies of PDAC tumors have been tricky, the researchers explained, in part because stromal cells tend to intermix with the tumor and with nearby normal cells.
With that in mind, Yeh and her colleagues took a crack at using NMF to virtually microdissect these intimately mixed cells from one another based on subtle gene expression differences.
With Agilent array-based expression profiles for 145 primary PDAC tumors, 61 metastatic PDAC tumors, along with 46 normal pancreatic samples, 88 distant site adjacent normal samples, and 17 related cell lines, the team was able to splice and dice up samples using clues from gene expression signatures associated with each cell type.
After verifying these results through RNA sequencing on dozens of primary tumors, PDAC-derived xenografts, cell lines, or cancer-associated fibroblast lines, the researchers looked at the information that could be gleaned from these gene expression patterns.
On the tumor side, they explained, the approach distinguished classical PDAC tumors from basal-like PDAC tumors, which resemble basal-like breast and bladder cancer subtypes.
As in those cancer types, the team found that basal-like PDACs were associated with reduced survival times — 11 months as a median — and a one-year survival rate of 44 percent. For individuals with the tumors from the classical subtype, on the other hand, the median survival time stretched out to 19 months, with 70 percent of individuals surviving at least a year.
The study's authors noted that further research is needed to work out whether basal-like PDAC tumors show treatment responses that resemble those in basal-like subtypes from other cancers.
Likewise, they noted that it may eventually be helpful to take treatment cues from gene expression profiles present in the stroma, which made up almost half of the cells in the primary tumor samples considered, on average.
In particular, the team found that individuals with activated stroma — marked by elevated expression of some immune response- and signaling-related genes — had a median survival time of just 15 months, with 60 percent of patients surviving up to one year. In contrast, the normal stroma subtype was linked to a two-year median survival time and a one-year survival rate of 82 percent.
The approach provided other hints about PDAC, too, including information about the mutations present in primary and metastatic tumor samples from the same individuals.