Therefore, the combined mRNA expression of key signal transduction pathway components and regulators of the extracellular tumor microenvironment in combination with certain direct regulators of immune activity (ICOS, DPP4) appears to predict priming for responsiveness to CTLA-4 blockade

Therefore, the combined mRNA expression of key signal transduction pathway components and regulators of the extracellular tumor microenvironment in combination with certain direct regulators of immune activity (ICOS, DPP4) appears to predict priming for responsiveness to CTLA-4 blockade. We found that relative to healthy normal controls expression of the response predictive genes was preserved in responders but significantly down regulated in non-responders. U test results were determined for all individual predictor genes. Candidate synergistic 2-gene pairs included either predictors or a predicator plus an enhancer variable. We used Statistical Innovations CORExpress 1.1 commercial software, which employs correlated component regression analysis T863 and handles multicolinearity due to correlated predictors effectively even with high-dimensional data. The software was run in two-component mode for synergistic gene pair analysis. The final list of synergistic gene pairs, or 2-gene core models, was trained to predict response in the discovery dataset, tested next in the validation dataset based on objective response and then tested using one-year survival as the criterion. Over 260 pre-treatment response-predictive 2-gene core models were validated for objective response and survival. Larger, optimal pre-treatment classifier models were constructed by combining validated 2-gene core models using CORExpress correlated component regression package. The software Rabbit Polyclonal to NPHP4 was run in three-component, step-down mode starting with validated pre-treatment core models to eliminate weaker genes. At each step, resulting classifier models were validated for response and survival on the validation dataset. The discovery datasets AUC was checked with publically-available MedCalc version 17 ROC analysis and which may exceed the sample size value?=?0.005) in the phase III but not the phase II where the (value?=?0.346.) Table ?Table33 shows eight examples of phase III pre-treatment response-predictors that are not statistically significant in the phase II study. Table 3 Examples of genes predictive for response in the discovery but not validation datasets thead th rowspan=”1″ colspan=”1″ Gene /th th rowspan=”1″ colspan=”1″ em N /em ?=?210 Discovery br / Pre-treatment br / ANOVA t-test /th th rowspan=”1″ colspan=”1″ em N /em ?=?150 Validation br / Pre-treatment br / ANOVA t-test /th /thead CD280.0260.158CD800.0120.368FAIM30.0080.638FYN0.0060.962IL18BP0.0200.958IL320.0210.686IL7R0.0090.590INPP4B0.0060.740 Open in a separate window The T863 data highly suggest that the expression of the 15-genes in the signature represent expression levels of particular genes needed for robust immune responses against cancer. Expression of these genes may identify patients whose immune systems are already primed to have an anti-cancer immune response. Therefore the level at which the 15 genes were expressed in discovery dataset responders was compared to expression in a set of 50 blood bank healthy normal volunteers. Unexpectedly only 6 genes demonstrated differential expression being either up or down regulated when responders were compared to healthy normal controls of which 5 of these 6 genes were enhancer genes. Eight of the genes demonstrated equivalent expression between responders and healthy normal controls (Table ?(Table4)4) of which 7 of these 8 genes were predictor genes. T863 Table 4 Relative gene expression of the 15 genes comprising the pre-treatment signature comparing responders in the discovery dataset to healthy volunteers and to non-responders thead th rowspan=”3″ colspan=”1″ 15-Gene Pre-Treatment Model /th th rowspan=”3″ colspan=”1″ Predictor or Enhancer Variable /th th rowspan=”1″ colspan=”1″ Blood Bank /th th rowspan=”3″ colspan=”1″ Difference Normals versus Responders /th th colspan=”2″ rowspan=”1″ Phase 3 Discovery Dataset /th th rowspan=”1″ colspan=”1″ /th th rowspan=”1″ colspan=”1″ em N /em ?=?50 /th th rowspan=”1″ colspan=”1″ em N /em ?=?28 /th th rowspan=”1″ colspan=”1″ em N /em ?=?182 /th th rowspan=”1″ colspan=”1″ Difference /th th rowspan=”1″ colspan=”1″ Healthy Normals /th th rowspan=”1″ colspan=”1″ T863 Responders /th th rowspan=”1″ colspan=”1″ Non-Responders /th th rowspan=”1″ colspan=”1″ Responders vs Non-responders /th /thead Responders Equivalent to Normals?ITGA4Predictor14.20.0214.2214.610.39?LARGEPredictor22.00.0922.0922.970.88?CDK2Predictor19.60.0919.6919.910.22?TIMP1Enhancer15.00.1015.114.95?0.15?DPP4Predictor18.50.1218.6218.950.33?NRASPredictor17.10.1317.2317.440.21?ERBB2Predictor23.0?0.1822.8223.230.41?NAB2Predictor20.0?0.2919.7120.040.33Responders T863 Upregulated Compared to Normals?ADAM17Enhancer18.50.3218.1818.360.18?RHOCEnhancer16.90.3916.5116.630.12?TGFB1Enhancer13.40.4512.9513.050.10?CDKN2APredictor21.40.6320.7721.160.39Responders Downregulated Compared to Normals?HLADRAEnhancer12.10.4812.5812.640.06?MYCEnhancer17.70.8218.5318.670.14Measurement of Gene Expression Not Available?ICOSPredictorN/AN/A22.3222.780.46 Open in a separate window As discussed above, the 15-gene signature contains 6 enhancer variable genes which do not independently predict response but rather enhance the predictive ability of the 9 predictor genes. While gene expression of most of the enhancers differed between responders and healthy normals only 1 1 (potentially 2) of the 9 predictive genes demonstrated differential expression (ICOS is a predictive gene but its mRNA expression was not available for measurement in the healthy normals). Given that almost all of the predictor genes showed equivalent expression in responders and healthy normals we hypothesized that the predictor genes were differentially expressed in the non-responders. As shown in Table ?Table44 all eight evaluable predictive genes were in fact down-regulated in the non-responders relative to responders. However all six enhancers showed no significant change in gene expression between responders and non-responders. Discussion An ideal biomarker should be obtained easily with minimal risk to the patient. Biomarkers based on mRNA transcript gene expression profiling obtained from whole blood have enormous.