构建M1型巨噬细胞相关lncRNA签名预测肿瘤免疫微环境.pdf

Construction of an M1 macrophage-related lncRNA signaturefor predicting the tumor immune microenvironmentQiWu,YimingLiu,QingsongHu,andHuihuiWuSchool of Basic Medical Sciences,Division of Life Sciences and Medicine,University of Science and Technology of China,Hefei 230026,ChinaCorrespondence:HuihuiWu,E-mail:2023TheAuthor(s).ThisisanopenaccessarticleundertheCCBY-NC-ND4.0license(http:/creativecommons.org/licenses/by-nc-nd/4.0/).Cite This:JUSTC,2023,53(9):0903(12pp)ReadOnlineAbstract:LongnoncodingRNAs(lncRNAs)areconsideredcrucialmoleculesassociatedwiththetumormicroenviron-ment(TME)andtumorimmunemicroenvironment(TIM).Macrophagesareimportantmembersoftheimmunesystem,andM1macrophagefunction-associatedlncRNAsstillneedtobefurtherinvestigated.Inthisstudy,alncRNAsignaturewas constructed based on transcriptome differences between high and low M1 macrophage infiltration cohorts.ThislncRNAsignatureincludedsevenlncRNAs:LINC01494,ZDHHC20-IT1,LINC01450,LINC00871,EVX1-AS,KIF25-ASandAADACL2-AS1,andallofthemwereupregulatedinpatientslackingM1macrophages,indicatingtheirrolesininhibitingmacrophageinfiltrationandpolarizingtotheM1subtype,leadingtoanimmuneexclusionTME,whichhasbeendemonstratedtobecloselycorrelatedwithpoorprognosis.ThislncRNAsignaturenotonlypredictedundesirableclinicaloutcomesbutwasalsoassociatedwiththeimmunosuppressiveenvironmentofthetumorregion,whichismediatedbyhinderingantigenpresentationandprocessingprogress.Inaddition,thepredictivevalueofthislncRNAsignatureforim-mune checkpoint inhibition(ICI)therapy was also evaluated,which further enriched and strengthened the power oflncRNAsinpredictingtheimmunotherapyresponserate.Keywords:coloncancer;M1macrophage;lncRNA;tumorimmunemicroenvironment;immunecheckpointtherapyCLC number:R735.3Document code:A1 IntroductionAtotalof1.9millioncoloncancerpatientswerenewlydia-gnosedin2020,causingover900,000deaths,rankingsecondinallcancer-relateddeaths1.Theincidenceratecontinuestorisearoundtheworld.Studieshaveidentifiedthatbadlife-stylesandhabits,suchassmoking2,3,high-fatdiet4andalco-holism5,increasetheincidenceriskofcoloncancer.Apartfromexternalenvironmentalfactors,approximately2%to5%ofcoloncancerpatientsarederivedfromintestinalfamilialgeneticdiseasessuchasLynchsyndromeandfamilialadeno-matouspolyposis(FAP)6,7.GenemutationsinhMSH2andhMLH1,whichareresponsibleformismatchrepair(MMR),were detected in most Lynch syndrome patients,whichgreatlyincreasestheriskofcoloncanceroncogenesisifleftuntreated8.APClossormutationisahazardfactorthatcon-tributes to Wnt-catenin signaling pathway dysregulation,furtherleadingtoFAPorevencoloncancer8,9.LongnoncodingRNAs(lncRNAs)aredefinedas200ntnoncoding RNAs that cannot be translated into functionalproteins.Recently,lncRNAs,asregulatorymolecules,havedisplayedattractiveprospectsinthefunctionalregulationofcells,especially in the field of cancer research.AbnormallncRNAtranscriptionhasbeenfoundinvarioustypesofcan-cer.Forinstance,PCGEM1andPRNCR1canbindtothean-drogenreceptorandpromoteprostatecancerprogression10.In pace with the development and rise of immunotherapy,light has been shed on the relationships between lncRNAsandthetumorimmunemicroenvironment1114.Macrophagesplayacrucialroleininnateandadaptiveim-muneprocesses,whichworkforpathogenorsenescentcellclearanceandhomeostasismaintenance.Aclassicclassifica-tionofmacrophagesisbasedonspecificfunctions,categoriz-ingmacrophages into M1 and M2 subtypes.M1 macro-phagesaregenerallyresponsibleforpro-inflammatoryfunc-tions and kill cancerous cells15,16,while M2 macrophagesmainlyexertrepairingoranti-inflammatoryactivitiesinin-juredtissue15,17,18andpro-tumorfunctioninthebackgroundof cancer19,20.Tumor-associated macrophages(TAMs)arenowattracting increasing attention.Studies have demon-stratedthatacomplexTMEcanreprogrammacrophagesintodifferentsubtypes2123.Atthesametime,thequestionaboutwhereTAMsoriginatealsoarose,andmanystudieshavede-scribed that TAMs are recruited by chemotactic factorssecretedfromtumorcells24,suchasCCL225,CCL526,andCXCL1227.Inthis study,we mainly focused on TME M1 macro-phages,andbioinformaticapproacheswereusedtofinddif-ferencesintranscriptomelevelsbetweendifferentCOADpa-tientpopulations,whichweredividedbyM1proportions.M1macrophage-related lncRNAs were identified and furtherscreenedbysurvivalanalysis.TheremaininglncRNAswerecollectedintoasignaturetotestitsprognosticvalueandper-formanceintheTCGA-COADcohortandotherindependentArticlehttp:/Received:December 30,2022;Accepted:April 07,202309031DOI:10.52396/JUSTC-2022-0185JUSTC,2023,53(9):0903validationdatasets.2 Materials and methods2.1 Data collectionCount and fragments per kilobase of per million(FPKM)RNA-Seqdataofcoloncancer(TCGA-COAD,n=453)andcorrespondingclinicalphenotypeinformationwereobtainedfromUCSCXENA(http:/xena.ucsc.edu/).Independentval-idationmicroarraydatasetswithclinicalinformationandsur-vivaldataweredownloadedfromtheGEOdatabase(https:/www.ncbi.nlm.nih.gov/geo/)underaccessionnumbersGSE14333(n=226,https:/www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14333),GSE17537(n=55,https:/www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE17537),GSE17536(n=177,https:/www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE17536),GSE17538(n=244,https:/www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE17538),and GSE39582(n=585,https:/www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39582).2.2 Tumor microenvironment infiltrated immune cellproportionTheCIBERSORTalgorithm28wasusedtodeconvoluteim-munecellproportionsfrombulkRNA-SeqdatabasedontheLM22immunecellsignature,whichcontainsBcells(naiveandmemory),plasmacells,CD8Tcells,CD4Tcells,follicu-larhelperTcells,regulatoryTcells(Tregs),gammadeltaTcells,NK cells(resting and activated),monocytes,macro-phages(M0,M1,M2),dendriticcells(restingandactivated),mastcells(restingandactivated),neutrophilsandeosinophils.M1 macrophages were extracted for further evaluation andanalysis.2.3 Differentially expressed lncRNA detection andlncRNA signature construction 0.05DESeq229 was used to find differentially expressed genesbetweenthehigh(M1-Mhigh)andlow(M1-Mlow)M1macro-phageinfiltrationgroups,whichwerestratifiedbythesurviv-alpackageandsurvminerpackageinR.|logFC|1andad-justedpvalueweresettofilterupregulatedanddown-regulatedgenes.LncRNAinformationwasdownloadedfromtheLNCipediadatabase(https:/lncipedia.org/)30.Theupreg-ulatedgenesintheM1-MlowgroupwereintersectedwiththelncRNAlisttoobtainupregulatedlncRNAsintheM1-Mlowgroup.These lncRNAs were further narrowed by survivalanalysis to filter survival-correlated lncRNAs for lncRNAsignatureconstruction.2.4 Gene set enrichment analysis and functional enrich-ment analysisThescoreofthelncRNAsignaturewascalculatedbyGSVA,and differential biological processes(BPs)between thelncRNAscorehighandlncRNAscorelowgroupswereanalyzedthroughgenesetenrichmentanalysis(GSEA)softwarede-veloped by the Broad Institute31,32.Gene sets were down-loadedfromMolecularSignaturesDatabasev7.5.1,GSEAof-ficialwebsite(http:/www.gsea-msigdb.org/gsea/msigdb/index.jsp),C5ontologygenesetsandHhallmarkgenesetswerequeriedfordownstreamanalysis.Inaddition,mRNAsnegativelycorrelated(cor 0.3,pvalue 0.05)withthelncRNA signature were selected for further gene ontology(GO)and functional enrichment analysis,which were per-formed by the online enrichment tool g:Profiler(https:/biit.cs.ut.ee/gprofiler/gost)33.2.5 Genomic mutation profileMutation information of TCGA-COAD was downloadedfromtheTCGAofficialwebsite,andthepackagemaftools34was used to visualize the mutation profile.The top 30mutatedgenesinlncRNAscorehighandlncRNAscorelowpa-tientswereextractedforfurtheranalyses.2.6 Statistical analysisKaplanMeiersurvivalplotsandlogrankpvalueswerecar-riedouttoidentifydifferentsurvivalprobabilitiesbetweentheM1high and M1low groups,which were determined by the“surv_cutpoint”functioninthe“survminer”Rpackage.Mul-tivariate Cox regression analyses were performed by the“coxph”functioninthe“survival”Rpackage.TheSpearmancorrelationcoefficientandcorrespondingpvalueswerecal-culatedbythe“cor.test”function.MannWhitneytestswereperformed to distinguish differences between two differentgroups.Statistical analyses were performed using R v4.1.3withthenecessaryRpackagesandGraphPadPrism7.0.3 Results3.1 Identification of differentially expressed lncRNAsbetween M1-Mhigh and M1-M1lowTodiscovertherelationshipbetweeninfiltrationofM1mac-rophagesinthetumormicroenvironmentandpatientprognos-is,theM1macrophage(M1-M)proportionofeachTCGA-COADpatientcalculatedbyCibersortwasgroupedintotwocohorts,the M1-Mlow(n=47)group and the M1-Mhigh(n=369)group,which were divided by the“surv_cutpoint”functionfromthesurvivalpackageinR(cutpoint=0.007690675).Patients with lower M1 macro-phage infiltration had significantly shorter overall survivaltimesthanpatientswithhigherM1macrophageinfiltration(Fig.1a,logrankp=0.0082).Thesameresultwasobtainedinanotherindependentvalidationdataset(Fig.1b,GSE14333,log rank p=0.0068,cutpoint=0.0453332;Fig.1c,integrateddataset of GSE17536 and GSE17538,log rank p=0.0472;Fig.1d,GSE39582,cutpoint=0.003866829,logrankp=0.0088,cutpoint=0.005733291).Toinvestigatethediffer-encesbetweentheM1-MlowandM1-Mhighgroups,differen-tiallyexpressedgeneswerecalculatedbytheDESeq2pack-age,and a total of 1634 DEGswere detected,which con-tained194upregulatedgenesand1440downregulatedgenesin the M1-Mlow group.Then,differentially expressedlncRNAs(DElncRNAs)wereextractedbyintersectingDEGswith the lncRNA list from the LNCipedia database,and atotalof71DElncRNAswereobtained(Fig.1e),whichcon-tained 13 upregulated lncRNAs and 58 downregulatedlncRNAs(Fig.1f,Wilcoxon rank sum test).Here,upregu-latedlncRNAsintheM1-Mlowgroupwereselectedforfur-theranalysis,andtheoverexpressionoftheselncRNAsmayM1macrophagelncRNAfeaturepredictsimmunecheckpointtherapyoutcomeWuetal.09032DOI:10.52396/JUSTC-2022-0185JUSTC,2023,53(9):0903mmmdFig.1.M1-MpredictstheclinicalprognosisofcoloncancerpatientsandDElncRNAdetection.(ad)KaplanMeiercurveofTCGA-COADdataset,GSE14333,integrateddatasetofGSE17536andGSE17538,GSE39582,respectively,stratifiedbyM1macrophageproportion.(e)VenndiagramshowsDElncRNAsthroughintersectingDEGs(n=1634)withthelncRNAlist(n=13076),71DElncRNAscontaining58downregulatedlncRNAsand13upreg-ulatedlncRNAs.(f)VolcanoplotofDElncRNAs,withupregulatedlncRNAsshowninred(n=13),downregulatedlncRNAsshowninblue(n=58,Wil-coxonranksumtest)andnonsenselncRNAsshowningray(n=1180,Wilcoxonranksumtest).Wuetal.09033DOI:10.52396/JUSTC-2022-0185JUSTC,2023,53(9):0903beresponsibleforlowerM1macrophageinfiltrationintheTME.3.2 lncRNA signature construction and evaluationSurvivalanalysisof13upregulatedlncRNAswascarriedout.7 lncRNAs stood out(Fig.2ag):LINC01494(log rankp=0.0011,cutpoint=0.04112853),ZDHHC20-IT1(p=0.0043,cutpoint=0.3504403),LINC01450(p=0.012,cutpoint=0.01601357),LINC00871(p=0.0013,cutpoint=0.081546),EVX1-AS(p=0.034,cutpoint=1.016255),KIF25-AS1(p=0.0013,cutpoint=0.07708173),AADACL2-AS1(p=0.0037,cutpoint=0.1166067),astheiroverexpressionsig-nificantly correlated with shorter survival time.TheselncRNAswerecollectedintoalncRNAsignatureforfurtherevaluationsandstudies.ToassessthelncRNAsignatureperformanceonprognosisprediction,the lncRNA signature score of each TCGA-COADpatientwascalculatedbytheGSVApackageinR.Allscoredpatientswerestratifiedinto2groupsbasedontheme-dianlncRNAsignaturescore(medianscore=0.04486603),andtheKaplanMeiercurveshowedthatpatientswithhigherlncRNAsignaturescoreshadsignificantlyshorteroverallsur-vivaltimes(Fig.2h,left,logrankp=0.00024).Additionally,inthelncRNAscorehighgroup,patientshadlowerinfiltrationof M1 macrophages,suggesting that the lncRNA signaturescoreisnegativelycorrelatedwithM1macrophagesintheTME(Fig.2i,p0.0001,MannWhitneytest).Inthevalida-tioncohortofGSE17537,thelncRNAsignaturescorealsocorrelatedwithashorteroverallsurvivaltime(Fig.2h,right,logrankp=0.011,cutpoint=0.4367792).MultivariateCoxregressionanalysisofthelncRNAsigna-tureshowedthatitservedasanindependentriskfactorthatsignificantly correlated with patient prognosis(Fig.2j).Alower lncRNA score acted as a protective factor,which isconsistentwithearliersurvivalanalysisresults.Inaddition,withtheprogressionofcoloncancerstage,thelncRNAscoreincreased(Fig.2k).The analysis above indicated that thelncRNAsignatureisasignificantriskfactorcorrelatedwithpatientprognosis.3.3 lncRNA signature correlates with immune inertnessGSEAwasperformedtodetectdifferentiallyactivatedpath-ways or biological processes between the lncRNA scorehighand lncRNA scorelow groups.Notably,in 50 hallmark genesets,IL2-STAT5signaling(NES=2.12,FDRqval=0.001),interferongammaresponse(NES=2.11,FDRqval=0.000),IL6-JSK-STAT3signaling(NES=2.08,FDRqval=0.001),interferon alpha response(NES=1.90,FDR q val=0.008)andcomplement(NES=2.09,FDRqval=0.001)weresigni-ficantlyenrichedinthelncRNAscorelowgroup(Fig.3a),in-dicatingthatthelncRNAscorehighcohortwasweakinantigenpresentation-associatedpathways.ThetumorimmunemicroenvironmentwasassessedviatheESTIMATE package in R.The tumor immune score ofTCGA-COAD patients was calculated,and its correlationwiththelncRNAsignaturescorewasalsotested.TheresultssuggestedthatahigherlncRNAsignaturescorewasnegat-ivelycorrelatedwiththeimmunescore(Fig.3b,Rs=0.302,p=3.9351010),indicatingalackofimmunecellswithintheTMEofcoloncancer.Commonly,lncRNAsexerttheirfunc-tions by manipulating mRNAs to regulate correspondingdownstreampathways.Hence,we calculated the relation-ships between protein-coding gene expression and thelncRNAsignaturescore.mRNAswithaSpearmancorrela-tioncoefficient 0.3andp 0.05(n=141)wereremovedforGOenrichmentandfunctionalenrichmentanalyses.Not-ably,theresultsshowedthatthesemRNAswerelargelyen-riched in immune-associated reactome pathways(Fig.3c,Fishersexacttest)andbiologicalprocesses(Fig.3d,Fishersexacttest),indicatingthatmembersinthelncRNAsignaturemayimpede immune-associated gene expression and func-tion.3.4 lncRNA signature negatively correlates with antigenpresentation and processing and immune cellchemotaxisImmune cell recruitment to the TME depends on tumor-relatedantigensthatcanberecognizedbyantigen-presentingcellssuchasDCsandmacrophages.Inaddition,chemokinessecretedbytumorcellsorstromalcellsintheTMEmediateextratumor boundary immune cell chemotaxis.Antigenprocessing-andpresentation-relatedgenesetswereincludedtoevaluatetheirrelationshipwiththelncRNAsignature.TheresultsshowedthatthehigherthelncRNAsetscorewas,thelower the score for antigen presentation and processing(Fig.4a,Rs=0.307,p=2.0541010).Additionally,thechemo-taxis score of each sample in the lncRNA scorehigh andlncRNAscorelowgroupswasassessed,andtheresultsshowedthatthelncRNAscorehighgrouphadalowerchemotaxisscore,which indicated an immune exclusion microenvironment(Fig.4b,MannWhitneytest).Theseresultsprovideaddition-alevidencethatpatientswithahighlncRNAscoremayalsohaveanimmune-coldTME.Inaddition,wetestedtherela-tionshipbetweenthelncRNAscoreandgenesthatpositivelyregulate antigen processing and presentation and positivelyregulatechemotaxis.Wefoundthattheseimmune-positive-regulatinggenesweremostlynegativelycorrelatedwiththelncRNAscore(Fig.4c,d),whichfurtherconfirmedtheinert-