Tumor immune microenvironment-based classifications of bladder cancer for enhancing the response rate of immunotherapy

Immunotherapy is a potential way to save the lives of patients with bladder cancer, but it only benefits approximately 20% them. A total 4,028 cancer were collected for this study. Unsupervised non-negative matrix factorization and nearest template prediction algorithms employed classification. We identified immune non-immune classes from The Cancer Genome Atlas Bladder Urothelial Carcinoma (TCGA-BLCA) training cohort. 150 most differentially expressed genes between these two extracted, classification reappeared in 20 validation cohorts. For activated exhausted subgroups, stromal activation signature was assessed by NTP algorithm. Patients class showed highly enriched signatures immunocytes, while subgroup also exhibited transforming growth factor (TGF)-?1, cancer-associated extracellular signatures. immune-activated lower genetic alteration better overall survival. Anti-PD-1/PD-L1 immunotherapy more beneficial subgroup, checkpoint blockade therapy plus TGF-? inhibitor or an EP300 might achieve greater efficacy immune-exhausted subgroup. Novel molecular classifier innovative cancer. IntroductionBladder 10th frequent tumor globally exhibits high rate recurrence.1Sanchez A. Wszolek M.F. Niemierko Clayman R.H. Drumm M. Rodríguez D. Feldman A.S. Dahl D.M. Heney N.M. Shipley W.U. et al.Incidence, Clinicopathological Risk Factors, Management Outcomes Nonmuscle Invasive Recurrence after Complete Response Trimodality Therapy Muscle Cancer.J. Urol. 2018; 199: 407-415Crossref PubMed Scopus (26) Google Scholar major challenge clinical care short-term recurrence non-muscle-invasive (NMIBC), as well shortened survival muscle-invasive (MIBC) patients, especially those distant metastases, 5-year whom less than 10%.2Hautmann R.E. Gschwend J.E. de Petriconi R.C. Kron Volkmer B.G. Cystectomy transitional cell carcinoma bladder: results surgery series neobladder era.J. 2006; 176 (discussion 491–492): 486-492Crossref (318) Scholar,3Chamie K. Litwin M.S. Bassett J.C. Daskivich T.J. Lai J. Hanley J.M. Konety B.R. Saigal C.S. Urologic Diseases America ProjectRecurrence high-risk cancer: population-based analysis.Cancer. 2013; 119: 3219-3227Crossref (192) In mass, normal cells, blood vessels, cytokines that surround support vitality cells compose microenvironment (TME). Crosstalk exists TME. Tumor can alter TME, TME promote spread tumors.From side, composed multitude heterogenetic characteristics, including gene mutations, copy number alterations, neoantigens, infiltration immunocytes. Several teams have established classifications among Mo al.4Mo Q. Nikolos F. Chen Tramel Z. Lee Y.C. Hayashi Xiao Shen Chan K.S. Prognostic Power Differentiation Gene Signature Carcinomas.J. Natl. Inst. 110: 448-459Crossref (63) generated 18-gene MIBC reflect urothelial differentiation predict outcomes; basal differentiated groups are highest lowest risk scores, respectively. Damrauer al.5Damrauer J.S. Hoadley K.A. Chism D.D. Fan C. Tiganelli C.J. Wobker S.E. Yeh J.J. Milowsky M.I. Iyer G. Parker Kim W.Y. Intrinsic subtypes high-grade hallmarks breast biology.Proc. Acad. Sci. USA. 2014; 111: 3110-3115Crossref (575) developed BASE47, 47 gene-based classifier, separate luminal-like basal-like tumors. Robertson al.6Robertson A.G. Al-Ahmadie H. Bellmunt Guo Cherniack A.D. Hinoue T. Laird P.W. Akbani R. al.TCGA Research NetworkComprehensive Molecular Characterization Muscle-Invasive Cancer.Cell. 2017; 171: 540-556.e25Abstract Full Text PDF (1078) further consensus hierarchical clustering luminal-papillary, luminal-infiltrated, luminal, basal/squamous, neuronal subtypes. However, classifiers focused on outcomes, not microenvironment. Therefore, our goals provide comprehensive insight into response diverse inner features generate screen suited immunotherapy.The (NMF) algorithm multiplicative updates algorithm; decompose V matrices, W H.7Devarajan Nonnegative factorization: analytical interpretive tool computational biology.PLoS Comput. Biol. 2008; 4: e1000029Crossref (265) Similar principal component analysis (PCA) independent (ICA), NMF use limited components original observed data, which contain huge volumes.8Gaujoux Seoighe flexible R package nonnegative factorization.BMC Bioinformatics. 2010; 11: 367Crossref (654) has been applied reveal biomarkers, classify subtypes, prognosis tumors recent works.9Zeng Vo A.H. Mao Clare Khan S.A. Luo Y. pathway discovery using factorization.J. Biomed. Inform. 2019; 96: 103247Crossref (16) Scholar, 10Esposito Boccarelli Del Buono N. An NMF-Based Methodology Selecting Biomarkers Landscape Genes Heterogeneous Cancer-Associated Fibroblast Populations.Bioinform. Insights. 2020; 14 (1177932220906827)Crossref (9) 11Meng Zhou Lu X. Bian Zhang L. Hao Liang Immune drives outcomes prostate implications immunotherapy.Mol. Oncol. (Published online December 18, 2020)https://doi.org/10.1002/1878-0261.12887Crossref (29) 12Zhou Y.J. Zhu G.Q. X.F. Zheng K.I. Wang Q.W. J.N. Yan F.R. Li X.B. Identification tumour microenvironment-based subgroups gastric immunotherapeutic implications.Cancer Immunol. Immunother. 69: 1057-1069Crossref (23) ScholarWe enrolled (NTP) distinguish different immunophenotypes cohort novel definition could illumination cancer.ResultsIdentification module derivation cancer4,028 involved, along matched clinicopathological information, expression profiles (Figure 1). performed virtual microdissection TCGA-BLCA To obtain robust module, we preset respective numbers five 10. When nine, first strongly enrichment defined 2A). top weighted exemplar reflected characteristics (Table S1). According ontological biological processes, associated T helper 1 (Th1)/Th2 differentiation, receptor signaling, B signaling (all p < 0.05; Table S2). Subsequently, redefined 408 immune-enriched non-immune-enriched via 2B). Furthermore, multidimensional scaling (MDS) random forest (RF) define precise 2C). Figure 2D, distributions modules, weight, clustering, final classes, score shown.Figure 2Recognition algorithmShow full caption(A) Nine modules algorithm, gathered recognized module. (B) Heatmap showing clusters, divided clustering. (C) modified clusters classes. (D) score.View Large Image ViewerDownload Hi-res image Download (PPT)Several immune-associated S3) help confirm each patient determined single-sample set (ssGSEA). increased immunocytes compare class, (as 13 signature, CD8+ NK. Metagene), B.P. metagene), macrophages, tertiary lymphoid structure (TLS), cytolytic activity (CYT), interferon (IFN) 3A). analyzed KEGG pathways GSEA, revealing (including cell-, natural killer leukocyte-associated pathways), chemokine pathways, antigen processing presentation, adhesion molecules, complement coagulation cascades), proinflammatory FC-Epsilon-RI-, NOD-like receptor-, FC gamma R-mediated phagocytosis pathways) all From Figures 2, 3A (top panel), S1 Tables S1, S2, S3, microdissected cohort, class.Figure 3The heterogeneity phenotypes subgroupShow Division characterization three immunophenotypes. CYT, score; TITR, tumor-infiltrating Tregs; MDSC, myeloid-derived suppressor cell; TLS, structure; C-ECM, matrix. Difference lymphocyte abundance. PD-L1 mRNA level. amplification deletion, arm levels focal levels. (E) mutation burden. (F) neoantigens. (G) Specific mutant (H) WT, wild-type; IM-Act, subgroup; IM-Exh, subgroup.View (PPT)Tumor distinguished cellsFibroblasts, mesenchymal (MSCs), (ECM) key stroma connect cells.13Valkenburg K.C. Groot A.E. Pienta K.J. Targeting improve therapy.Nat. Rev. Clin. 15: 366-381Crossref (473) Especially during late stages tumors, epigenetic alterations driven components.14Hanahan Coussens L.M. Accessories crime: functions recruited microenvironment.Cancer Cell. 2012; 21: 309-322Abstract (2789) MSCs act inherent regulators secrete inhibiting soluble factors surface markers suppress regulate impact proliferation induction regulatory (Tregs).15Sivanathan K.N. Gronthos S. Rojas-Canales Thierry B. Coates P.T. Interferon-gamma modification stem cells: autologous allogeneic allotransplantation.Stem Cell Rep. 10: 351-375Crossref (108) Scholar,16van Megen K.M. van ’t Wout E.T. Lages Motta Dekker Nikolic Roep B.O. Activated Mesenchymal Stromal Cells Process Present Antigens Regulating Adaptive Immunity.Front. 694Crossref (37) processes decreasing factors, IFN-?, necrosis (TNF)-?, interleukin (IL)-1?, promoting type 2 IL-10 IL13.17Soboslay Lüder C.G. Riesch Geiger S.M. Banla Batchassi E. Stadler Schulz-Key Regulatory effects Th1-type (IFN-gamma, IL-12) Th2-type (IL-10, IL-13) parasite-specific cellular responsiveness Onchocerca volvulus-infected humans exposed endemic controls.Immunology. 1999; 97: 219-225Crossref (36) 18Aggarwal Pittenger Human modulate responses.Blood. 2005; 105: 1815-1822Crossref (3599) 19Selleri Dieng M.M. Nicoletti Louis I. Beausejour Le Deist Haddad Cord-blood-derived downmodulate CD4+ T-cell inducing IL-10-producing Th1 cells.Stem Dev. 22: 1063-1075Crossref reason, previously used separation immunocyte immunophenotypes, status. 11.0% (45/408) remaining 110 (27.0%, 110/408) belonged phenotype 2; Cancer-associated ECM (C-EC) regulated fibroblasts recruit immunosuppressive (TGF)-? accepted immunosuppressor microenvironment, Tregs MDSCs status TME.20Batlle Massagué Transforming Growth Factor-? Signaling Immunity Cancer.Immunity. 50: 924-940Abstract (781) 21Furukawa Wisel Tang Impact Immune-Modulatory Drugs Cell.Transplantation. 2016; 100: 2288-2300Crossref (73) 22Groth Hu Weber Fleming V. Altevogt P. Utikal Umansky Immunosuppression mediated (MDSCs) progression.Br. Cancer. 120: 16-25Crossref (282) 23Berraondo Sanmamed Ochoa M.C. Etxeberria Aznar M.A. Pérez-Gracia J.L. Rodríguez-Ruiz M.E. Ponz-Sarvise Castañón Melero Cytokines immunotherapy.Br. 6-15Crossref (373) These evaluated ssGSEA, revealed (TITRs), WNT/TGF-?, TGF-?1-activated, C-ECM higher 3A; TIM-3 LAG3 reported be exhaustion status,24Dong Bao PD-1 LAG-3 loss function chronic hepatitis B.BMC 20: 27Crossref (54) Scholar,25Liu J.F. Wu Yang L.L. Deng W.W. W.F. Sun Z.J. Blockade TIM3 relieves immunosuppression through reducing head neck cancer.J. Exp. Res. 37: 44Crossref (59) found similar subgroups; (p = 0.008) 0.218) Based (bottom panel) separated subgroups. score, (MDSC), WNT/TGF-? validated LAG3.Table 2Summary parameters TCGA-BLCA, GSE32894, E-MTAB-1803 cohortsTCGA-BLCA (n 408)GSE32894 308)E-MTAB-1803 70)Age?7023014342>7017816528SexMale30122859Female1078011StageaSix samples lacked stage data TCGA database, GEO: GSE32894.Ta–116–T11197–T21918524T3157728T443118GradebThree grade date GSE32894.G1/low2148–G2–1034G3/high38415466SmokingcTwo alive 84 GSE32894.No109––Yes286––StatusAlive22919928Dead1772542a Six GSE32894.b Three GSE32894.c Two GSE32894. Open table new tab Heterogeneity classesTo genes, compared (TIL) abundance anteriorly estimated hematoxylin eosin (H&E) staining,26Saltz Gupta Hou Kurc Singh Nguyen Samaras Shroyer K.R. Zhao Batiste al.Cancer NetworkSpatial Organization Correlation Tumor-Infiltrating Lymphocytes Using Deep Learning Pathology Images.Cell 23: 181-193.e7Abstract (400) TIL 0.001; 3B), consistent groups. 3C). (CNA), burden (TMB), neoantigens exhibit crosstalk activation. deletion at both (pArm-del 0.001, pFocal-del 0.007) CNA (pArm-Amp 0.733, pFocal-Amp 0.065) 3D), positive association deletion. With TIMER, twice confirmed deletion; deep arm-level PD-1, CTLA-4; checkpoints linked decreased infiltration, neutrophils, dendritic S3).The TMB 0.01; 3E), neoantigen level no difference 0.109; 3F). specific mutations S4A). Mutations TP53 (53.5% versus 43.1%, 0.051), TTN (52.9% 39.5%, 0.011), PIK3CA (28.0% 17.0%, 0.007), RB1 (26.0% 13.0%, 0.001) appeared frequently S4B). ERBB2 0.035), KMT2A 0.013), PKHD1 MDN1 0.015) noted 3G), 0.020), HMCN1 0.014), AKAP9 0.003), MACF1 0.016) 3H). lead S4C). 3B–3H, S4, conclude abundance, TMB, diverse.Reappearance external cohortsExternal cohorts recapitulate respect 1; 2). (DEGs) S4) chosen seed regenerate subclasses GenePattern NMFConsensus, then method.Table 1Summary detailed information cohortsDatasetData arrayPatientsReferenceTCGA-BLCARNA sequencing408https://xenabrowser.net/datapages/?cohort=GDC%20TCGA%20Prostate%20Cancer%20(PRAD)&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443E-MTAB-4321RNA sequencing476https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-4321/IMvigor210Illumina HiSeq 2500348http://research-pub.gene.com/IMvigor210CoreBiologies/GSE32894Illumina HumanHT-12 V3.0 beadchip308https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE32894GSE83586Affymetrix 1.0 ST Array307https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83586GSE87304Affymetrix Exon Array305https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE87304GSE128702Affymetrix Array256https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE128702GSE13507Illumina human-6 v2.0 beadchip164https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE13507GSE129871Illumina 2000 (Homo sapiens)158https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE129871GSE120736Illumina V4.0 beadchip145https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE120736GSE39016Affymetrix Array141https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39016GSE128701Affymetrix Array136https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE128701GSE124035Affymetrix Array133https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE124305GSE86411Illumina WG-DASL R2 beadchip132https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE86411GSE48276Illumina beadchip116https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE48276GSE128192Illumina beadchip112https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE128192GSE31684Affymetrix U133 Plus 2.0 Array93https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE31684GSE134292Illumina 4000 sapiens)80https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE134292GSE93527Affymetrix Transcriptome Array 2.079https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE93527E-MTAB-1803Affymetrix GeneChip 2.070https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-1803/GSE69795Illumina beadchip61https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE69795 GSE32894 60.7% (187/308) signatures; 121 enrichment, 42 79 High scores WNT/TGF?, TGF?-1-activated, 3; S5).Table distribution newly cohortsDatasetNo. patientsImmunophenotype distribution, n (%)Immune activatedImmune exhaustedNon-immuneTCGA-BLCA40845 (11.03)110 (26.96)253 (62.01)E-MTAB-432147674 (15.55)111 (23.32)291 (61.13)IMvigor21034885 (24.43)142 (40.8)121 (34.77)GSE3289430842 (13.64)79 (25.65)187 (60.71)GSE8358630763 (20.52)93 (30.29)151 (49.19)GSE8730430559 (19.34)85 (27.87)161 (52.79)GSE12870225672 (28.13)88 (34.38)96 (37.50)GSE1350716423 (14.02)36 (21.95)105 (64.02)GSE12987115826 (16.46)27 (17.09)105 (66.46)GSE12073614521 (14.48)35 (24.14)89 (61.38)GSE3901614116 (11.35)31 (21.99)94 (66.67)GSE12870113642 (30.88)34 (25.00)60 (44.12)GSE12403513332 (24.06)54 (40.6)47 (35.34)GSE8641113222 (16.67)36 (27.27)74 (56.06)GSE4827611624 (20.69)29 (25.00)63 (54.31)GSE12819211226 (23.21)36 (32.14)50 (44.64)GSE316849314 (15.05)34 (36.56)45 (48.39)GSE1342928013 (16.25)16 (20.00)51 (63.75)GSE935277913 (16.46)15 (18.99)51 (64.56)E-MTAB-18037013 (18.57)19 (27.14)38 (54.29)GSE69795619 (14.75)19 (31.15)33 (54.10) other cohorts, replicated displayed 3 S5 S6. ranged 11.3% 30.9%, proportion 17.1% 40.8%. 18 As expected, Tregs, combined 1, 3, S4 S6, suggest stably precisely divide immune-activated, immune-exhausted, phenotypes. reappear cohorts.Favorable anti-PD-L1 subgroupTo evaluate profile 348 IMvigor210 large phase II