Colorectal Cancer: Microsatellite Instability/Mismatch Repair Testing in the Era of Digital Pathology

See “Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning,” by Echle A, Grabsch HI, Quirke P, et al, on page 1406. See “Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning,” by Echle A, Grabsch HI, Quirke P, et al, on page 1406. It has been >15 years since universal microsatellite instability (MSI) or mismatch repair (MMR) testing has been recommend for all cases of colorectal carcinoma.1Hampel H. Frankel W.L. Martin E. et al.Screening for the Lynch syndrome (hereditary nonpolyposis colorectal cancer).N Engl J Med. 2005; 352: 1851-1860Crossref PubMed Scopus (1048) Google Scholar MSI/MMR testing is a proven prognostic marker and can help identify patients with possible hereditary cancers (Lynch syndrome).2Phipps A.I. Alwers E. Harrison T. et al.Association between molecular subtypes of colorectal tumors and patient survival, based on pooled analysis of 7 international studies.Gastroenterology. 2020; 158: 2158-2168 e4Abstract Full Text Full Text PDF PubMed Scopus (12) Google Scholar,3Boland C.R. Goel A. Microsatellite instability in colorectal cancer.Gastroenterology. 2010; 138: 2073-2087 e3Abstract Full Text Full Text PDF PubMed Scopus (1137) Google Scholar More recently, checkpoint inhibition therapy has been found to be effective in tumors that demonstrate MSI, making MSI/MMR testing a pan-cancer therapeutic biomarker.4Ruiz-Banobre J. Goel A. DNA mismatch repair deficiency and immune checkpoint inhibitors in gastrointestinal cancers.Gastroenterology. 2019; 156: 890-903Abstract Full Text Full Text PDF PubMed Scopus (24) Google Scholar Given its importance, multiple organizations, including the American Society of Clinical Oncology, Association for Molecular Pathology, College of American Pathologists, European Society for Medical Oncology, and National Institute for Health and Care Excellence, have advocated for reflective MSI or MMR testing on all colorectal adenocarcinomas.5Stjepanovic N. Moreira L. Carneiro F. et al.Hereditary gastrointestinal cancers: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up†.Ann Oncol. 2019; 30: 1558-1571Abstract Full Text Full Text PDF PubMed Scopus (52) Google Scholar,6Sepulveda A.R. Hamilton S.R. Allegra C.J. et al.Molecular biomarkers for the evaluation of colorectal cancer: guideline from the American Society for Clinical Pathology, College of American Pathologists, Association for Molecular Pathology, and American Society of Clinical Oncology.Arch Pathol Lab Med. 2017; 141: 625-657Crossref PubMed Scopus (38) Google Scholar Despite the call for universal testing, several barriers exist for the implementation of universal MSI testing. In this issue of Gastroenterology, Echle et al7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar demonstrate how the adoption of digital pathology can help to overcome some of the barriers that exist for universal MSI testing.7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar The underlying concept is that through the application of deep learning algorithms, image analysis on routine hematoxylin and eosin–stained slides can identify tumors with MSI or abnormal MMR expression. At its core, MSI testing is done to identify tumors in which DNA MMR is not functioning, resulting in the accumulation of errors through cell replication. These errors are most notable in areas with DNA repeats. Inactivation of MMR can occur in association with hereditary syndromes with an underlying alteration in the MMR genes (Lynch syndrome), or, more commonly, through somatic inactivation or mutation in the MMR genes. The identification of these tumors can be done either by looking for the underlying molecular changes in the genome (MSI testing) or through immunohistochemical (IHC) testing evaluating the expression of the most common gene products associated with MMR (MLH1, MSH2, MSH6, and PMS2). Normal expression of the MMR gene products is referred to as proficient MMR expression, whereas the loss of expression is typically referred to as deficient MMR expression (dMMR). As Echle et al7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar describe in their article, although these molecular and IHC assays can identify most patients with MSI/dMMR, there is no gold standard test or evaluation method. The most common test used to identify patients with MSI is IHC. IHC is cheaper and more commonly available than molecular assays. In colorectal cancer (CRC), the sensitivity and specificity for MMR–IHC range from 80.8% to 100% and 80.5% to 91.9%, respectively.5Stjepanovic N. Moreira L. Carneiro F. et al.Hereditary gastrointestinal cancers: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up†.Ann Oncol. 2019; 30: 1558-1571Abstract Full Text Full Text PDF PubMed Scopus (52) Google Scholar,8Snowsill T. Coelho H. Huxley N. et al.Molecular testing for Lynch syndrome in people with colorectal cancer: systematic reviews and economic evaluation.Health Technol Assess. 2017; 21: 1-238Crossref Scopus (43) Google Scholar At this level, MMR-IHC is considered a reliable tool for the routine identification of MSI in CRC and other tumors. Echle et al7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar have demonstrated that image analysis using a deep learning algorithm can identify CRC with MMR or dMMR with a sensitivity of 95% and specificity of 67%, levels similar to those seen with current clinical testing. It is expected that image analysis could identify CRC with underlying MSI/dMMR. CRC which are MSI have known morphologic changes including tumor-infiltrating lymphocytes, poorly differentiated tumor areas, and peritumoral inflammation. Several assessment algorithms have been previously developed that use morphologic and clinical features to predict the likelihood of MSI in CRC.9Greenson J.K. Huang S.C. Herron C. et al.Pathologic predictors of microsatellite instability in colorectal cancer.Am J Surg Pathol. 2009; 33: 126-133Crossref PubMed Scopus (150) Google Scholar Unfortunately, other organ systems, such as endometrial cancer, in which routine MSI/MMR testing has also been advocated, do not have such associated morphologic findings.10Carlson J. McCluggage W.G. Reclassifying endometrial carcinomas with a combined morphological and molecular approach.Curr Opin Oncol. 2019; 31: 411-419Crossref PubMed Scopus (15) Google Scholar The lack of morphologic features in other systems raises the question of using deep learning image analysis for MSI/dMMR testing in other organ sites. Because sites will still need to use IHC or molecular methods to test for MSI in other organ systems there may be limited interest in the adoption of image analysis for MSI/dMMR testing related to a single organ system. The number of required biomarkers to determine the optimal treatment for patients with cancer is expanding; as a result, molecular analyses of carcinomas is becoming the standard of care for the treatment of metastatic carcinoma for multiple disease sites. The molecular profiling of tumors is also a fundamental component for many clinical trials looking at tumor-agnostic and tumor-specific therapies. RAS and BRAF testing is recommended for all metastatic CRC in which chemotherapy is being considered.6Sepulveda A.R. Hamilton S.R. Allegra C.J. et al.Molecular biomarkers for the evaluation of colorectal cancer: guideline from the American Society for Clinical Pathology, College of American Pathologists, Association for Molecular Pathology, and American Society of Clinical Oncology.Arch Pathol Lab Med. 2017; 141: 625-657Crossref PubMed Scopus (38) Google Scholar,11Van Cutsem E. Cervantes A. Adam R. et al.ESMO consensus guidelines for the management of patients with metastatic colorectal cancer.Ann Oncol. 2016; 27: 1386-1422Abstract Full Text Full Text PDF PubMed Scopus (1608) Google Scholar Because multiple genes (KRAS, NRAS, BRAF) are being analyzed, many molecular testing kits have moved to advanced molecular testing platforms such as next generation (or massive parallel) sequencing. Recent studies have shown that MSI can be determined using next-generation sequencing.12Gan C. Love C. Beshay V. et al.Applicability of next generation sequencing technology in microsatellite instability testing.Genes (Basel). 2015; 6: 46-59Crossref PubMed Scopus (29) Google Scholar,13Zhu L. Huang Y. Fang X. et al.A novel and reliable method to detect microsatellite instability in colorectal cancer by next-generation sequencing.J Mol Diagn. 2018; 20: 225-231Abstract Full Text Full Text PDF PubMed Scopus (38) Google Scholar The best method, or methods, for determining the underlying molecular phenotype when IHC, molecular analysis and digital pathology are available has yet to be determined. When discussing the limitations of their analysis, Echle et al7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar state that the ground truth used to train the deep learning system are imperfect. This lack of a gold standard and correlation with clinical outcomes applies to all other studies evaluating methods to determine CRC MSI or dMMR status. Studies evaluating clinical outcomes, tied to the specific indication for the selected test, are required to determine the optimal testing strategy to determine the best therapy and management for patients. Digital pathology is expanding internationally. This includes the use of digital pathology for clinical use. This article shows how, once digital pathology is adopted into diagnostic use, image analysis can be used to further classify tumors, including the identification of tumors with a specific molecular phenotype. Other reports have demonstrated how image analysis can identify molecular phenotypes or biomarker findings based on the hematoxylin and eosin stain alone in pancreatic, lung, prostate, and gastric cancers, as well as CRC.14S NK Wilson G.W. Grant R.C. et al.Morphological classification of pancreatic ductal adenocarcinoma that predicts molecular subtypes and correlates with clinical outcome.Gut. 2020; 69: 317-328Crossref PubMed Scopus (24) Google Scholar,15Acs B. Rantalainen M. Hartman J. Artificial intelligence as the next step towards precision pathology.J Intern Med. 2020; 288: 62-81Crossref PubMed Scopus (50) Google Scholar These sites have been limited to single centers and lack the ability to be extrapolated to other sites. Echle et al7Echle A. Grabsch H.I. Quirke P. et al.Clinical-grade detection of microsatellite instability in colorectal tumors by deep learning.Gastroenterology. 2020; 159: 1406-1416Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar have highlighted the importance of ensuring image analysis is focused on the nature of the specimen of interest and how international, large-scale collaboration to strengthen deep learning algorithms and allow for tools to be shared between multiple, diverse institutions. It is unlikely that a single site will have the >5000 annotated specimens needed to optimize a deep learning system to the point where it can be clinically validated and considered ready for diagnostic testing. The authors of this study have done impressive work, demonstrating how digital pathology can add value beyond the hematoxylin and eosin slide. The authors have demonstrated the value and necessity for multisite collaboration to optimize algorithms and develop systems which can be used in different practice settings. Incorporating deep learning and image analysis algorithms into a digital pathology practice allows for classification of tumors, including predicting molecular biomarkers without the added expense of IHC or molecular testing. The work done by Echle et al can serve as a template to demonstrate how to refine and evaluate digital “molecular” biomarkers. Clinical-Grade Detection of Microsatellite Instability in Colorectal Tumors by Deep LearningGastroenterologyVol. 159Issue 4PreviewMicrosatellite instability (MSI) and mismatch-repair deficiency (dMMR) in colorectal tumors are used to select treatment for patients. Deep learning can detect MSI and dMMR in tumor samples on routine histology slides faster and less expensively than molecular assays. However, clinical application of this technology requires high performance and multisite validation, which have not yet been performed. Full-Text PDF Open Access