Background Microsatellite instability (MSI) is an important biomarker for immunotherapy in solid tumors, resulting from deficient mismatch repair (dMMR) and leading to elevated mutational burden. Standard methods for detecting MSI include (1) immunohistochemistry to detect loss of Lynch Syndrome genes, (2) polymerase chain reaction of conserved microsatellite regions, and (3) next-generation sequencing-based methods, including MSIsensor and MANTIS.^{1 2} Relatedly, a tumor mutational burden (TMB) of at least 10 mutations per megabase is also a biomarker for immunotherapy in solid tumors, irrespective of MSI status.³ Here we demonstrate a deep learning-based digital pathology approach to identify dMMR, irrespective of MSI status, providing a quantitative method to identify additional patients that may respond to immunotherapy.

Methods Convolutional neural network models trained using hematoxylin and eosin-stained whole slide images were used to extract human interpretable features (HIFs) quantifying the tumor microenvironment (TME) in TCGA colorectal cancer (CRC; n=464) and endometrial cancer (EC; n=517) cohorts. MSI status was determined based on previous analyses2, and TMB was quantified by whole-exome sequencing. Mutational signatures indicative of dMMR were identified using the deconstructSigs R package.⁴ The cohort was split into 4 groups: MSI, MSS/dMMR, MSS/TMB-high, and MSS. Mann-Whitney tests identified pairwise associations between HIFs and molecular subpopulations. Multivariable logistic regression models were developed to predict MSI and dMMR status. Repeated 5-fold cross-validation for 1000 iterations was used to quantify performance metrics.

Results Molecular subtypes are summarized in table 1. In CRC, MSI was associated with general immune cell infiltration into the tumor, while both MSI and MSS/dMMR were associated with increased immune cell infiltration, particularly macrophages, into cancer gland lumens. Multivariable TME HIF models successfully differentiated MSI from non-MSI (median AUROC: 0.86) and MSS/dMMR from MSS (median AUROC: 0.65), including MSS/TMB-high. In EC, pairwise analysis revealed 61, 34, and 11 TME-related HIFs associated with MSI, MSS/dMMR, and MSS/TMB-high when compared to MSS tumors (FDR p < 0.05), respectively. However, when comparing MSI, MSS/dMMR, and MSS/TMB-high to each other, no HIFs passed FDR correction.

Conclusions Understanding drivers of immunotherapy response is key to improving patient outcomes. Here, we show that TME composition (e.g., elevated immune cell infiltration) is associated with MSI status in CRC and EC. Importantly, these features are also present in some MSS patients, with and without dMMR. These results suggest that digital pathology models may enable identification of previously unidentified patients likely to respond to immunotherapy.

References

1. Chen ML, Chen JY, Hu J, Chen Q, Yu LX, Liu BR, Qian XP, Yang M. Comparison of microsatellite status detection methods in colorectal carcinoma. Int J Clin Exp Pathol. 2018;11(3):1431–1438.

2. Bonneville R, Krook MA, Kautto EA, Miya J, Wing MR, Chen HZ, Reeser JW, Yu L, Roychowdhury S. Landscape of Microsatellite Instability Across 39 Cancer Types. JCO Precis Oncol. 2017;2017:PO.17.00073.

3. Sha D, Jin Z, Budczies J, Kluck K, Stenzinger A, Sinicrope FA. Tumor Mutational Burden as a Predictive Biomarker in Solid Tumors. Cancer Discov. 2020;10(12):1808–1825.

4. Rosenthal R, McGranahan N, Herrero J, Taylor BS, Swanton C. DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution. Genome Biol. 2016;17:31.