Study Finds Immune Enrichment Improves Outcomes in Pancreatic Cancer With Activated but Not Normal Stroma

Get Permission

ALTHOUGH IMMUNOTHERAPIES have had limited success in pancreatic cancer to date, findings from a molecular analysis of the tumor microenvironment suggest that certain subtypes may be more responsive to treatment. According to data presented at the 2019 Society of Surgical Oncology Annual Cancer Symposium, immune infiltration appears to improve outcomes in patients with “activated stroma.”1 What’s more, these differences do not appear to be due to tumor intrinsic factors such as neoantigen load or differences in immune populations, the authors of the study reported.

Robert J. Torphy, MD

Robert J. Torphy, MD

“We think that the distribution of the immune infiltration in the tumor microenvironment, or potentially the proximity of this infiltration to the tumor epithelial cell, may be contributing to these improved outcomes in patients with activated stroma and high immune infiltration,” said lead author of the study, Robert J. Torphy, MD, a general surgery resident at the University of Colorado School of Medicine, Aurora.

Study Rationale

AS DR. TORPHY explained, the tumor microenvironment in pancreatic ductal adenocarcinoma consists of abundant stroma, which may interfere with the effective treatment of pancreatic cancer by impairing drug delivery and suppressing the immune response. Better understanding of the tumor microenvironment will help to identify new treatment strategies or more personalized treatment approaches, he added, but the heterogeneity of the tumor microenvironment poses a challenge.

Dr. Torphy and colleagues have utilized an approach called virtual deconvolution to identify and study tumor and stroma molecular subtypes of pancreatic cancer. Although the tumor epithelial cell component clusters into “classical” and “basal-like” subtypes, said Dr. Torphy, the stroma clusters into “normal” and “activated” subtypes, which are enriched for the expression of genes associated with quiescent fibroblasts and activated pancreatic stellate cells, respectively.

“A great deal of work has gone into evaluating the role of the tumor subtypes of pancreatic cancer, but a lot less work has gone into understanding the role of stromal subtypes and the immune compartments of pancreatic cancer,” stated Dr. Torphy.

“We think that the distribution of the immune infiltration in the tumor microenvironment … may be contributing to these improved outcomes in patients with activated stroma and high immune infiltration.”
— Robert J. Torphy, MD

Tweet this quote

Study Details

FOR THIS STUDY, Dr. Torphy and colleagues developed an algorithm called DECODER to perform an unbiased deconvolution of bulk tumor RNA expression from patients in The Cancer Genome Atlas (TCGA) data set. DECODER was used to score immune enrichment and to classify tumors as immune high versus low. Initial validation using hematoxylin and eosin–stained diagnostic slides from the TCGA showed that DECODER immune scores were significantly higher in tumors that had infiltration present on the slides vs tumors with no evidence of immune infiltration. The investigators then evaluated the association between immune enrichment as scored by DECODER and overall survival across stroma subtypes.

Dr. Torphy and co-investigators included 146 patients with no evidence of metastatic disease who all underwent resection of their primary tumors. Clinical and pathologic factors were similar between patients with normal vs activated stroma.

Stromal Subtype Has Prognostic Value

AS DR. TORPHY reported, the investigators observed no difference in overall survival between tumors with normal vs activated stroma. However, DECODER-high immune tumors had improved outcomes compared with tumors found to have DECODER-low immune infiltration (P = .0126).

“We were surprised to see that in the normal stroma subtype of tumors, immune enrichment had no prognostic effect, but in the activated stroma subtype, tumors that had high immune infiltration had significantly improved overall survival compared with tumors with low immune infiltration, with a medium survival of approximately 23.0 vs 15.8 months,” said Dr. Torphy. He noted that this association was confirmed in a multivariable analysis controlling for patient age, race, and tumor characteristics (hazard ratio [HR] = 0.30, P = .001).


  • De novo compartment deconvolution and weight estimation of tumor samples
  • Integrated computational framework for performing de novo deconvolution
  • Used to break down a new tumor sample into its components
  • Estimates cellular compartment weights (tumor, stroma, immune, etc) within a sample
  • Can be applied to any data platform with nonnegative value

The researchers then attempted to identify factors allowing for immune infiltration to improve survival in tumors with activated but not normal stroma. The analysis of mutations per megabase and neoantigen load from single nucleotide variations or insertion and deletion mutations showed no differences in mutations or predicted neoantigen load between the stromal subtypes. More granular analysis of immune infiltration, including CD8 T cells, regulatory T cells, and macrophages, found similar immune population enrichment across tumors with normal and activated stromal subtypes.

Furthermore, DECODER immune scores were also correlated with the cytolytic index (based on the expression of granzyme A and perforin 1) in both stromal subtypes, but there was no difference between tumors with normal or activated stroma. Finally, markers of T-cell activation and exhaustion, including PD-1, LAG3, TIM3, and TIGIT, were shown to be correlated with DECODER immune scores but did not differ between stromal subtypes.

“Interestingly, in our cohort of patients, we saw that tumors with activated stroma and high immune infiltration had significantly more immune clusters than tumors with normal stroma or activated stromal tumors with low immune infiltration,” Dr. Torphy observed.

According to the authors, future work is needed to determine whether specific cell populations are differentially enriched and, if so, whether this has biologic significance. “We hope to expand this analysis using a larger data set and perform immunohistochemical staining of tumors with activated stroma and normal stroma to look more specifically at the spatial distribution of the infiltrating lymphocytes in the tumor microenvironment,” Dr. Torphy concluded.

“For now, the work underscores how far we remain from leveraging such data to a successful immunotherapy strategy for patients with pancreatic cancer.”
— Susan Bates, MD

Tweet this quote

Predicting Response: The ‘Holy Grail’ for Researchers

SUSAN BATES, MD, a medical oncologist at NewYork-Presbyterian and the Herbert Irving Comprehensive Cancer Center at Columbia University, commented that understanding why certain cancers fail to respond to immunotherapy and how to change that has become the holy grail for cancer researchers. “Pancreatic cancer has proven particularly difficult,” said Dr. Bates. “Beyond the few tumors that bear known inducers of high tumor mutation burden, such as mismatch-repair deficiency [and subsequent microsatellite instability], immunotherapy has failed.” 

Although multiple investigators have found the immune microenvironment in most pancreatic cancer to be “cold,” lacking the requisite activated T cells for an immune response, continued Dr. Bates, a variety of pathways and approaches might improve the infiltration of activated T cells in the pancreatic microenvironment.

“Rather than a traditional histologic look at the mix of immune cells infiltrating the cancer, Dr. Torphy has taken a molecular approach, ratifying prior observations showing that RNA expression data derived from bulk tumor RNA can be used to make inferences about immune-cell activation in the tumor microenvironment,” said Dr. Bates. She noted that although a high immune score was independently associated with prolonged survival, whether this can be exploited therapeutically remains to be determined.”

“These data build on recent understandings of the importance of the immune microenvironment and raise hope that RNA signatures could eventually allow a precision medicine approach to understand which tumors in which patients need ‘immune enrichment’ and activation—perhaps directing a particular type of immune activation,” Dr. Bates concluded. “But, for now, the work underscores how far we remain from leveraging such data to a successful immunotherapy strategy for patients with pancreatic cancer.”

DISCLOSURE: Drs. Torphy and Bates reported no conflicts of interest.


1. Torphy RJ, Peng XL, Moffitt RA, et al: Immune enrichment improves outcomes only in activated stroma subtype of pancreatic cancer. 2019 Society of Surgical Oncology Annual Cancer Symposium. Abstract 2. Presented March 29, 2019.