Combined biomarker for prediction of response to an immune checkpoint inhibitor in metastatic gastric cancer

Abstract

Purpose: Immunecheckpointblockades(ICB)havebeensuccessfulingastriccancer (GC).However,themajorityofunselectedpatientswithGCfailtorespondtoICB.Itis crucialtoidentifyprecisebiomarkerstopredictresponsetoICB. Methods: Geneexpressionprofilingofformalin-fixedandparaffin-embeddedGCtissues from25patientstreatedwithICB(pembrolizumab)targetingprogrammedcelldeathpro-tein1(PD-1)wasperformedusingNanoString(NanoStringTechnologies).Fordevelop-mentofagenesignaturetopredictresponsetoICB,differentialgeneexpressionanalysis withlinearregressionmodelingwasperformedwithareaunderthecurvepackagesinR. Results: Fromtheanalysis,10genesweredifferentiallyexpressedbetweenpatientswith responseandnoresponsetoICB(P< 0.01).Toidentifyabiomarkerpredictingresponseto ICB,fourgeneswereselectedbasedon|log 2 (foldchange)|≥ 1.AftercalculatingtheIMmu-notherapyAgainstGastrIcCancer(IMAGiC)score,patientsweredividedintotwogroups:to beresponderandtobenon-responder,accordingtoResponseEvaluationCriteriainSolid Tumors(RECIST)guidelines.TheIMAGiCscorewassignificantlyassociatedwithRECIST groups(P = 0.0057),Epstein-Barrvirus(P = 0.048),andtumormutationalload(P = 0.023); however,wasnotsignificantlycorrelatedwithmicrosatelliteinstabilitystatus(P = 0.14) andprogrammeddeathligand1(PD-L1)expression(P= 0.095).ToreproduceIMAGiCwith differenttechnology,weretestedtheresultswithaquantitativereal-timepolymerase chainreaction(qRT-PCR)method,andtheprecisionofreproductionof87.5%.Invalida-tioncohortwith17samplesfromtheongoingtrialwithnivolumab,theprecisionofIMAG-iCqRT-PCRwas100%. Conclusion: OuridentifiedgenesignaturesandproposedIMAGiCmodelforpredicting responsetopembrolizumabinpatientswithGCshowedvalidity.