Pancreatic cancer is often treated as a homogeneous disease, but new research reveals distinct subtypes of tumor and stromal gene expression
Michael Schubert |
Exocrine tumors make up about three-quarters of pancreatic cancer, and of those, the vast majority are ductal adenocarcinomas (PDACs). Unfortunately, these are also among the most lethal pancreatic cancers, with only one in 25 patients surviving for five years. But despite this low rate of treatment success, pancreatic cancers are still treated as a single entity – although the disease has been exhaustively sequenced, very few new genetic mutations have come to light, which suggests that the tumors are very alike. So why do clinicians see such different outcomes in different patients? What guides the tumors to spread to different locations, or allows some patients to survive for many years while others progress quickly?
A hallmark of PDAC is the extensive involvement of the stroma – the tissue surrounding the tumor. This tissue is poorly understood, has very few genetic mutations, and makes it difficult to capture precise molecular information from the tumor itself. Previous researchers have attempted to surmount this hurdle by either isolating their studies to looking at tumors with high tumor cell content (1) or microdissecting for the pure tumor cell population (2). But although both of these are valid strategies for studying the tumor compartment, they fail to gather data on PDAC’s defining characteristic: its low ratio of tumor to stromal content.
A research group at the University of North Carolina’s Lineberger Comprehensive Cancer Center wanted to better understand the nature of PDACs, which meant collecting information both on the tumors themselves and on the surrounding stroma. “In order to differentiate between stroma- and tumor-specific signals, we used a computational strategy called ‘non-negative matrix factorization’ (NMF) to separate out gene expression components (3),” said study leader Jen Jen Yeh, UNC Lineberger member, associate professor and the vice chair for research in the UNC School of Medicine Department of Surgery. “This is an unbiased strategy that forced us to understand at the back end what the gene expression components mean. Once we understood what the gene expression of each compartment represented, we realized that NMF was able to separate out all the different tissue compartments – essentially performing a virtual microdissection.” For instance, if they analyzed a metastatic liver sample, they were able to distinguish a normal liver compartment, a tumor compartment that was similar to the primary PDAC tumor, and some PDAC stroma.
The study involved using NMF to analyze 357 samples – 145 primary and 61 metastatic tumors, 17 cell lines, and 46 pancreas and 88 distant site adjacent normal samples. “After confirming that we were able to distinguish between different tissue compartments,” said Yeh, “we were able to identify two different tumor-specific subtypes, the ‘classical’ and the ‘basal-like,’ and two different stroma-specific subtypes, the ‘normal’ and the ‘activated’.”
Why is this so significant? When the patient populations were divided by tumor subtype, the data revealed that those with basal-like tumors had a significantly worse median survival time (11 vs. 19 months) and one-year survival rate (44 vs. 70 percent) than those with classical tumors. The news isn’t all bad for basal-like tumors, though; the subtype also showed better response to adjuvant therapy, with a hazard ratio of 0.38 compared with 0.76 in the classical subtype. Interestingly, the researchers found that all of the tested cell lines were classified as basal-like, meaning that researchers who use pancreatic cancer cell lines may not be getting a complete picture of the disease’s diversity. Stromal subtype has a similar effect; when the patient populations were divided by stromal subtype, those with activated stroma had a worse median survival time (15 vs. 24 months) and one-year survival rate (60 vs. 82 percent) than those with normal stroma.
Differential gene expression allowed the research group to classify PDAC into four distinct subtypes – classical tumors with normal stroma, classical tumors with activated stroma, basal-like tumors with normal stroma, and basal-like tumors with activated stroma. Although even these basic categories have prognostic relevance, Yeh says that there are still two major limitations in the current understanding of PDAC. One is our knowledge of the stroma; although many groups have made a great deal of progress with that in the last few years, there’s still much more to do. The second is our lack of understanding of metastatic disease. “It’s rare that we have patient samples from metastatic sites,” says Yeh, “and to truly make headway into the treatment of pancreatic cancer, we as clinicians should make a concerted effort to enroll patients in clinical studies and obtain permission to store specimens as frozen tissue or create patient-derived xenografts, so that the field will have have a library of metastatic samples for investigating biology and treatment options.”
Thus far, Yeh and her colleagues are excited to have confirmed their subtypes in large external cohorts – and they now believe that, in the short-term, these subtypes can be used as diagnostic and prognostic tools. In the longer term, though, they plan to examine whether or not the subtypes can be used in decision-making to help tailor therapy types and timing for patients. This will become even more important as the repertoire of therapies for pancreatic cancer increases. Once it’s possible to biologically explain the behavior of different pancreatic cancer subtypes, clinicians can be encouraged to stop thinking of it as one homogeneous disease. Pathologists will then be able to use that knowledge in conjunction with existing evaluations to inform patients’ individual therapy plans.
Mission: Impossible by Michael Schubert
Tapping a Rich Vein by Christopher Chapman
New Ab-ilities in Imaging by Michael Schubert
Friend or Foe? by Anne Marie Lennon
(Chemo)resistance Is Futile by Michael Schubert
Breaching Cancer’s Defenses by Ingunn Stromnes
An Epigenetic Epiphany by Michael Schubert
Building a Better Mousetrap by Michael Schubert
- AV Biankin et al., “Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes”, Nature, 491, 399–405 (2012). PMID: 23103869.
- AK Witkiewicz et al., “Whole-exome sequencing of pancreatic cancer defines genetic diversity and therapeutic targets”, Nat Commun, 6, 6744 (2015). PMID: 25855536.
- RA Moffitt et al., “Virtual microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal adenocarcinoma”, Nat Genet, 47, 1168–1178 (2015). PMID: 26343385.