2018-肠道微生物组调节黑素瘤患者对抗PD-1免疫疗法的反应.pdf

CANCER IMMUNOTHERAPYGut microbiome modulates responseto antiPD-1 immunotherapy inmelanoma patientsV.Gopalakrishnan,1,2*C.N.Spencer,2,3*L.Nezi,3*A.Reuben,1M.C.Andrews,1T.V.Karpinets,3P.A.Prieto,1 D.Vicente,1K.Hoffman,4S.C.Wei,5A.P.Cogdill,1,5L.Zhao,3C.W.Hudgens,6D.S.Hutchinson,7T.Manzo,3M.Petaccia de Macedo,6T.Cotechini,8T.Kumar,3W.S.Chen,9S.M.Reddy,10R.Szczepaniak Sloane,1J.Galloway-Pena,11H.Jiang,1P.L.Chen,9 E.J.Shpall,12K.Rezvani,12A.M.Alousi,12R.F.Chemaly,11S.Shelburne,3,11L.M.Vence,5P.C.Okhuysen,11V.B.Jensen,13A.G.Swennes,7F.McAllister,14E.Marcelo Riquelme Sanchez,14Y.Zhang,14E.Le Chatelier,15L.Zitvogel,16N.Pons,15J.L.Austin-Breneman,1|L.E.Haydu,1E.M.Burton,1J.M.Gardner,1E.Sirmans,17J.Hu,18A.J.Lazar,6,9T.Tsujikawa,8A.Diab,17H.Tawbi,17I.C.Glitza,17W.J.Hwu,17S.P.Patel,17S.E.Woodman,17R.N.Amaria,17M.A.Davies,17J.E.Gershenwald,1P.Hwu,17J.E.Lee,1J.Zhang,3L.M.Coussens,8Z.A.Cooper,1,3 P.A.Futreal,3C.R.Daniel,4,2N.J.Ajami,7J.F.Petrosino,7M.T.Tetzlaff,6,9P.Sharma,5,19J.P.Allison,5R.R.Jenq,3#J.A.Wargo1,3#*Preclinical mouse models suggest that the gut microbiome modulates tumor responseto checkpoint blockade immunotherapy;however,this has not been well-characterizedin human cancer patients.Here we examined the oral and gut microbiome of melanomapatients undergoing antiprogrammed cell death 1 protein(PD-1)immunotherapy(n=112).Significant differences were observed in the diversity and composition ofthe patient gut microbiome of responders versus nonresponders.Analysis of patientfecal microbiome samples(n=43,30 responders,13 nonresponders)showedsignificantly higher alpha diversity(P 0.01)and relative abundance of bacteria ofthe Ruminococcaceae family(P 0.01)in responding patients.Metagenomic studiesrevealed functional differences in gut bacteria in responders,including enrichmentof anabolic pathways.Immune profiling suggested enhanced systemic and antitumorimmunity in responding patients with a favorable gut microbiome as well as ingerm-free mice receiving fecal transplants from responding patients.Together,thesedata have important implications for the treatment of melanoma patients withimmune checkpoint inhibitors.Tremendousadvanceshavebeenmadeinthetreatment of melanoma and other cancersby using immune checkpoint inhibitors tar-getingthecytotoxicTlymphocyteassociatedantigen 4(CTLA-4)and programmed celldeathprotein1(PD-1);however,responsestothesetherapiesareoftenheterogeneousandnotdurable(13).Ithasrecentlyemergedthatfactorsbeyondtumor genomics influence cancer developmentand therapeutic responses(47),including hostfactors such as the gastrointestinal(gut)micro-biome(810).A number of studies have shownthat the gut microbiome may influence anti-tumor immune responses by means of innateand adaptive immunity(11,12)and that thera-peutic responses may be improved through itsmodulation(13,14);however,this has not beenextensively studied in cancer patients.Tobetterunderstandtheroleofthemicrobiomein response to immune checkpoint blockade,weprospectivelycollectedmicrobiomesamplesfrompatientswithmetastaticmelanomastartingtreat-ment with antiPD-1 therapy(n=112 patients)(fig.S1 and table S1).Oral(buccal)andgut(fecal)microbiome samples were collected attreatmentinitiation,andtumorbiopsiesandbloodsampleswerecollectedatmatchedpretreatmenttimepointswhen possible,to assess for genomic alterationsas well as the density and phenotype of tumor-infiltrating and circulating immune cell subsets(Fig.1A and fig.S2).Taxonomic profiling using16SribosomalRNA(rRNA)genesequencingwasperformed on all available oral and gut samples,withmetagenomicwhole-genomeshotgun(WGS)sequencingperformedona subset(n=25).Eligi-blepatients(n=89)wereclassifiedasresponders(R,n=54)or nonresponders(NR,n=35)on thebasisofradiographicassessmentusingtheresponseevaluation criteria in solid tumors(RECIST 1.1)(15)at 6 months after treatment initiation.Pa-tients were classified as R if they achieved anobjectiveresponse(completeorpartialresponseor stable disease lasting at least 6 months)or NR(progressive disease or stable disease lasting lessthan 6 months).This classification accounts forthesubsetofpatientswhomayderivelong-termdisease benefit despite not achieving a bona fideRECIST response and has been used in numer-ous published studies of patients on checkpointblockade(1619).Of note,patients in R and NRgroups were similar with respectto age,gender,primary type,prior therapy,concurrent systemictherapy,and serum lactate dehydrogenase(tableS2).Prior genomic analyses have demonstratedthatpatientswithtumorsthathaveahighermuta-tional load are more likely to respond to antiCTLA-4(16,20,21)or antiPD-1 therapy(2124);however,a high mutational load alone appearsneither sufficient nor essential for response.Inthis cohort,the total number of mutations andspecific melanoma driver mutations were withincomparableparametersbetweenRandNRafterantiPD-1 therapy(fig.S3),though the numberoftumorsavailableforsequencing(n=10,R=7,NR=3)was limited and may have reduced ourabilitytodetectasignificantassociationbetweenmutational burden and response.We first assessed the landscape of the oralandgutmicrobiomeinallavailablesamplesinpatients(n=112)with metastatic melanoma using 16SRESEARCHGopalakrishnan et al.,Science 359,97103(2018)5 January 20181 of 71Department of Surgical Oncology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.2Department of Epidemiology,Human Genetics and Environmental Sciences,University of Texas School of Public Health,Houston,TX 77030,USA.3Department of Genomic Medicine,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.4Department of Epidemiology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.5Department of Immunology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.6Department of Translational Molecular Pathology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.7Department of Molecular Virologyand Microbiology,Baylor College of Medicine,Houston,TX 77030,USA.8Department of Cell,Developmental and Cell Biology,Oregon Health and Sciences University,Portland,OR 97239,USA.9Department of Pathology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.10Department of Breast Medical Oncology,The University of Texas MD AndersonCancer Center,Houston,TX 77030,USA.11Department of Infectious Diseases,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.12Department of Stem CellTransplantation,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.13Department of Veterinary Medicine and Surgery,The University of Texas MD Anderson CancerCenter,Houston,TX 77030,USA.14Department of Clinical Cancer Prevention,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.15Centre de Recherche de Jouy-en-Josas,Institut National de la Recherche Agronomique,78352 Jouy-en-Josas,France.16Centre dInvestigation Clinique Biothrapie,Institut Gustave-Roussy,94805 Villejuif Cedex,France.17Department of Melanoma Medical Oncology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.18Department of Biostatistics,The University of Texas MD AndersonCancer Center,Houston,TX 77030,USA.19Department of Genitourinary Medical Oncology,The University of Texas MD Anderson Cancer Center,Houston,TX 77030,USA.*These authors contributed equally to this work.Present address:University of Rochester James P.Wilmot Cancer Center,Rochester,NY 14642,USA.Present address:A.C.Camargo Cancer Center,So Paolo,Brazil.Present address:Moffitt Cancer Center,Tampa,FL 33612,USA.|Present address:Harvard University,Cambridge,MA 02138,USA.Present address:MedImmune,Gaithersburg,MD 20878,USA.#Theseauthors contributed equally to this work.*Corresponding author.Email:jwargomdanderson.orgon December 28,2018 http:/science.sciencemag.org/Downloaded from sequencing,noting that both communities wererelativelydiversewithahighabundanceofbacte-ria of the order Lactobacillales in the oral micro-biomeandBacteroidalesinthefecalmicrobiome(Fig.1B).Bipartite network analysis(25)demon-stratedaclearseparationofcommunitystructurebetween the oral and fecal microbiomes in termsof both matched and aggregate samples(fig.S4),suggesting that these communities are distinctin terms of their compositional structure.Loss ofmicrobialdiversity(dysbiosis)isassociatedwithchronichealthconditions(2628)andcancer(810)andisalsoassociatedwithpooroutcomesofcertainformsofcancertherapy,includingallogeneicstemcell transplant(29).Based on these data,we ex-amined the diversity of the oral and gut micro-biomes in eligible patients on antiPD-1 therapyand found that alpha diversity,or within-samplediversity,ofthegutmicrobiomewassignificantlyhigher in R(n=30)compared to NR(n=13)usingseveralindices(P0.1%abun-dance)at the order level in oral(n=109,top)and fecal(n=53,bottom)samples by 16S rRNA sequencing.(C)Inverse Simpson diversity scores of thegut microbiome in R(n=30)and NR(n=13)to antiPD-1 immunotherapyby Mann-Whitney U rank sum(MW)test.Error bars represent the distributionof diversity scores.(D)Phylogenetic composition of fecal samples(n=39)at the family level(0.1%abundance)at baseline.High blue,11.63(inverseSimpson score),n=13,intermediate(gold,7.46 to 11.63,n=13),and low(red,7.46,n=13)diversity groups were determined using tertiles of inverseSimpson scores.(E)Kaplan-Meier(KM)plot of PFS by fecal diversity:high(median PFS undefined),intermediate(median PFS=232 days),and low(median PFS=188 days).High versus intermediate diversity(HR 3.60,95%CI 1.02 to 12.74)and high versus low(HR 3.57,95%CI 1.02 to 12.52)byunivariate Cox model.(F)Principal coordinate analysis of fecal samples(n=43)by response using weighted UniFrac distances.*P 0.05;*P 3.(E)Differentially abundantgut bacteria in R(blue)versus NR(red)by MW test false-discovery rate(FDR)adjusted within all taxonomic levels.(F)Pairwise comparisons by MWtest of abundances of metagenomic species identified by metagenomic WGSsequencing in fecal samples(n=25)for R(n=14,blue)and NR(n=11,red).*P 0.05;*P 0.01.Colors reflect gene abundances visualized as“barcodes”with the following order of intensity:white(0)light blue blue green yellow orange red for increasing abundance,where each colorchange corresponds to a fourfold abundance change.In these barcodes,metagenomic species appear as vertical lines(coabundant genes in asample)colored according to the gene abundance.RESEARCH|REPORTon December 28,2018 http:/science.sciencemag.org/Downloaded from response to therapy(12,14,15,23),we sought todetermine if differences existed inthe oral orgutmicrobiomes of R and NR to antiPD-1 therapy.To test this,we first compared an enrichment ofoperational taxonomic units(OTUs)in R versusNR,demonstrating that distinct sets of rare lowabundance OTUs were associated with responseto antiPD-1 therapy,with enrichment of ordersClostridiales in R and Bacteroidales in NR in thegut microbiome(P 0.01;Fig.2,A and B,andfig.S9,A and C).No significant differences inenrichment were noted in the oral microbiomeof R versus NR(fig.S9,B and D,and fig.S10).To further explore these findings,we performedhigh-dimensional class comparisons using lineardiscriminant analysis of effect size(LEfSe)(31),whichagaindemonstrateddifferentiallyabundantbacteria in the fecal microbiome of R versus NRGopalakrishnan et al.,Science 359,97103(2018)5 January 20184 of 7Faecalibacterium BacteroidalesHigh abundanceLow abundanceHigh abundanceLow abundance*100500Days elapsedDays elapsed%Progression-free0 200 400 600100500%Progression-free0 200 400 600*BiosynthesisDegradation-2.29-1.53-0.7600.761.532.29Type 205101520Number of patients0.000.250.500.75Type 1Type 21.00OtherVeillonellalesAcidaminococcalesLactobacillalesErysipelotrichalesDesulfovibrionalesEnterobacterialesBurkholderialesClostridialesBacteroidalesPatientscrOTU abundanceResponderNon-responderLowHighcrOTU community typeType 1*ResponderNon-responderSecondary metabolite biosynthesisCarbohydrate biosynthesisNucleoside and nucleotide biosynthesisAmino acid biosynthesisCell structure biosynthesisFatty acid and lipid biosynthesisAromatic compund biosynthesisAmines and polyamines biosynthesisMetabolic regulator biosynthesisAminoacyl-tRNA chargingAldehyde degradationDegradation/Utilization/Assimilation-OtherPolymeric compound degradationBile acid degradationC1 compound utilization and assimilationFatty acid and lipid degradationAlcohol degradationNucleoside and nucleotide degradationAmino acid degradationAmines and polyamines degradationSecondary metabolite degradationBiosynthesisDegradation/Utilization/AssimilationCofactors,prosthetic groups biosynthesisGeneration of precursor metabolites Carboxylates degradationAromatic compounds degradationCarbohydrate degradationInorganic nutrient metabolismFig.3.Abundance of crOTUs within the gut micro-biome is predictive of response to antiPD-1immunotherapy.(A)Top:Unsupervised hierarchicalclustering by complete linkage of Euclidean distancesby crOTU abundances in 43 fecal samples.Bottom:Stacked bar plot of relative abundances at the orderlevel by crOTU community type.(B)Association ofcrOTU community types with response to antiPD-1 byFishers exact test:crOTU community type 1(black,n=11;R=11,NR=0)and crOTU community type 2(orange,n=32;R=19,NR=13).R,blue bars;NR,redbars.(C)Comparison KM plot PFS curves by log-ranktest in patients with high abundance(dark blue,n=19,median PFS=undefined)or low abundance(light blue,n=20,median PFS=242 days)of Faecalibacterium(topPFS curve)or with high abundance(dark red,n=20,median PFS=188 days)or low abundance(light red,n=19,median PFS=393 days)of Bacteroidales(bottomPFS curve).(D)Unsupervised hierarchical clustering ofpathway class enrichment calculated as the number ofMetaCyc pathways predicted in the metagenomes offecal samples from 25 patients(R=14,NR=11).Columns represent patient samples(R,blue;NR,red),and rows represent enrichment of predicted MetaCycpathways(blue,low enrichment;black,mediumenrichment;yellow,high enrichment).Black text,biosynthetic pathways;blue text,degradative pathways.*P 0.05.RESEARCH|REPORTon December 28,2018 http:/science.sciencemag.org/Downloaded from inresponsetoantiPD-1therapy,withClostridialesorder and Ruminococcaceae family enriched in RandBacteroidales order enriched in NR(Fig.2,C and D).No major differences were observed intheoralmicrobiomebetweenRandNR,withtheexceptionofhigherBacteroidalesinNRinresponsetoantiPD-1therapy(fig.S11).Pairwisecomparisonswerethenperformedforbacterialtaxaatalllevelsbyresponse.Inadditiontoconfirmingtheprevioustaxonomic differences,these analyses identifiedthe Faecalibacterium genus as significantly en-riched in R(Fig.2E and table S3).MetagenomicWGSsequencingfurtherconfirmedenrichmentofFaecalibacteriumspeciesinadditiontoothersinR,whereas Bacteroides thetaiotaomicron,Escherichiacoli,andAnaerotruncuscolihominiswereenrichedin NR(Fig.2F and table S4).Importantly,the gutmicrobiome was shown to be relatively stable overtime in a limited number of longitudinal samplestest