An immune responsive tumor microenvironment imprints into PBMCs and predicts outcome in advanced pancreatic cancer: lessons from the PREDICT trial.
Prognosis in advanced pancreatic ductal adenocarcinoma (aPDAC) is particularly poor, only few patients benefit from treatment, and there are few biomarkers. The PREDICT trial examined whether first-line time-to-treatment failure (TTF1) predicts second-line treatment failure (TTF2) in aPDAC patients but found no association. We hypothesized that the tumor immune microenvironment (TiME) could correlate with the outcome in this trial and assessed whether tissue features were reflected in peripheral blood.
PREDICT patients received 5-FU/LV plus nanoliposomal irinotecan as second-line treatment. We stratified patients by shortest vs. longest TTF2 and analyzed 20 treatment-naïve tumor tissues samples via transcriptomics and immunohistochemistry. Peripheral blood mononuclear cells (PBMCs) from 82 patients collected prior to second-line therapy underwent flow cytometry and gene expression profiling. A machine learning pipeline integrated PBMC and clinical data to predict second-line outcome including external validation in 30 patients.
Long-TTF2 tumors exhibited an immune-active ("hot") TiME with cytotoxic CXCR3+CD8+-T-cell infiltration. PBMC analysis showed that these immune features were reflected in peripheral blood after one line of treatment. A novel 7-feature PBMC-based model ("TTF2Pred") accurately predicted TTF2 and overall survival, outperforming clinical or CA19-9 models and was confirmed in an external validation cohort. Long-TTF2 patients exhibited more circulating CXCR3⁺-T-cells and plasmacytoid dendritic cells. Short-TTF2 patients had more platelet-leukocyte aggregates.
An immune-active, treatment-naïve TiME predicts a better second-line outcome, and these characteristics imprinted into PBMCs obtained after one line of chemotherapy. We here first describe a minimally invasive, PBMC-based predictor of second-line outcome as a powerful prognostic tool for triaging patients.
ClinicalTrials.gov NCT03468335 (registered March 15, 2018).
PREDICT patients received 5-FU/LV plus nanoliposomal irinotecan as second-line treatment. We stratified patients by shortest vs. longest TTF2 and analyzed 20 treatment-naïve tumor tissues samples via transcriptomics and immunohistochemistry. Peripheral blood mononuclear cells (PBMCs) from 82 patients collected prior to second-line therapy underwent flow cytometry and gene expression profiling. A machine learning pipeline integrated PBMC and clinical data to predict second-line outcome including external validation in 30 patients.
Long-TTF2 tumors exhibited an immune-active ("hot") TiME with cytotoxic CXCR3+CD8+-T-cell infiltration. PBMC analysis showed that these immune features were reflected in peripheral blood after one line of treatment. A novel 7-feature PBMC-based model ("TTF2Pred") accurately predicted TTF2 and overall survival, outperforming clinical or CA19-9 models and was confirmed in an external validation cohort. Long-TTF2 patients exhibited more circulating CXCR3⁺-T-cells and plasmacytoid dendritic cells. Short-TTF2 patients had more platelet-leukocyte aggregates.
An immune-active, treatment-naïve TiME predicts a better second-line outcome, and these characteristics imprinted into PBMCs obtained after one line of chemotherapy. We here first describe a minimally invasive, PBMC-based predictor of second-line outcome as a powerful prognostic tool for triaging patients.
ClinicalTrials.gov NCT03468335 (registered March 15, 2018).
Authors
Lahusen Lahusen, Lutz Lutz, Fang Fang, Kirchner Kirchner, Albus Albus, Kluck Kluck, Karthaus Karthaus, Schwarzer Schwarzer, Siegler Siegler, Kleger Kleger, Ettrich Ettrich, Becher Becher, Höfling Höfling, Siveke Siveke, Budczies Budczies, Tannapfel Tannapfel, Stenzinger Stenzinger, Cheung Cheung, Eiseler Eiseler, Seufferlein Seufferlein
View on Pubmed