Simple Algorithm for the Prediction of Low-Level Microsatellite Instability (MSI-L) in Colorectal Cancer
Inti Zlobec, Michel P Bihl, Alessandro Lugli. Institute of Pathology, University of Bern, Bern, Switzerland; Institute of Pathology, University Hospital of Basel, Basel, Switzerland
Background: Colorectal cancers with low-level microsatellite instability (MSI-L) are thought to represent a subgroup of highly aggressive tumors leading to unfavorable outcome. Whereas MSI-high (MSI-H) cancers can be distinguished from microsatellite stable (MSS) cases using immunohistochemistry for mismatch repair markers MLH1, MSH2, MSH6 and PMS2, the same is not true for MSI-L since both MSI-L and MSS will result in a positive immunohistochemical pattern. The aim of this study was to establish a discriminating protein profile for MSI-L colorectal cancers.
Design: MSI was assessed by analysing BAT25, BAT26, D2S123, D5S346, and D17S250. MSS and MSI-L were defined as instability in 0 and 1 marker, respectively. A tissue microarray was stained and analysed for 41 different proteins selected from a larger pool of markers to represent major signalling pathways (WNT, RAS/MAPK, TGF-beta, and AKT), stem cell markers, cellular processes such as cell cycle, proliferation, apoptosis, angiogenesis, metastasis and markers of differentiation and mucin expression. Simple and multiple logistic regression and receiver operating characteristic (ROC) curve analysis were performed and area under the curve (AUC) was evaluated.
Results: From 172 cases included in this study, 98 (57%) were MSS and 20 (11%) were MSI-L. Only loss of nuclear beta-catenin (OR (95%CI): 0.21 (0.1-0.9); p=0.027), Cdx2 (OR (95%CI): 0.23 (0.1-0.63); p=0.005), CK20 (OR (95%CI): 0.16 (0.03-0.7); p=0.018) and membranous E-cadherin (OR (95C%): 0.2 (0.1-0.55; p=0.002) could significantly predict MSI-L from MSS. More importantly, in multivariate analysis, beta-catenin, CK20 and Cdx2 were independent predictors of MSI-L and together strongly discriminating (AUC=0.82). These features could be combined into a single score and visualized as an algorithm for the prediction of MSI-L. The parameter estimates from logistic regression analysis were used as weights for each marker then a total score and probability of MSI-L were obtained.
Conclusions: These results suggest that in contrast to MSS, MSI-L cancers may not frequently arise through deregulation of WNT pathway signaling. In addition, a simple combined score of beta-catenin, CK20 and Cdx2 may be useful as a predictive algorithm for the MSI-L status.
Tuesday, March 20, 2012 1:00 PM
Poster Session IV # 66, Tuesday Afternoon