molecular subtype-speciﬁc immunocompetent models of …007914; kppt model). in order to induce cre...

Tumor Biology and Immunology

Molecular Subtype-Specific ImmunocompetentModels of High-Grade Urothelial CarcinomaReveal Differential Neoantigen Expression andResponse to ImmunotherapyRyoichi Saito1, Christof C. Smith1,2, Takanobu Utsumi1, Lisa M. Bixby1,Jordan Kardos1,3, Sara E.Wobker4, Kyle G. Stewart1, Shengjie Chai1,5,Ujjawal Manocha1, Kevin M. Byrd4, Jeffrey S. Damrauer1, Scott E.Williams1,4,Benjamin G. Vincent1,2,6, and William Y. Kim1,3,6,7,8

Abstract

High-grade urothelial cancer contains intrinsic molecular sub-types that exhibit differences in underlying tumor biology and canbe divided into luminal-like and basal-like subtypes. We describehere the first subtype-specific murine models of bladder cancerand show that Upk3a-CreERT2; Trp53L/L; PtenL/L; Rosa26LSL-Luc

(UPPL, luminal-like) and BBN (basal-like) tumors are morefaithful to human bladder cancer than the widely used MB49cells. Following engraftment into immunocompetent C57BL/6mice, BBN tumors were more responsive to PD-1 inhibition than

UPPL tumors. Responding tumors within the BBNmodel showeddifferences in immune microenvironment composition, includ-ing increased ratios of CD8þ:CD4þ and memory:regulatory Tcells. Finally, we predicted and confirmed immunogenicity oftumor neoantigens in each model. These UPPL and BBN modelswill be a valuable resource for future studies examining bladdercancer biology and immunotherapy.

Significance: This work establishes human-relevant mousemodels of bladder cancer. Cancer Res; 78(14); 3954–68.�2018 AACR.

IntroductionIn the United States, bladder cancer is the fifth most common

malignancy with approximately 79,000 new cases and nearly17,000 deaths expected in 2017 (1). Bladder cancer is comprisedof both low-grade and high-grade tumors. Although low-gradetumors are almost uniformly noninvasive (Ta), high-gradetumors can become muscle-invasive and metastatic.

Multiple studies have now identified distinct RNA expressionsubtypes within both low- and high-grade bladder cancer (2–10).

Building upon the work of Hoglund and colleagues (5), we alongwith others have recently described distinct subtypes of high-grade muscle-invasive urothelial carcinoma, which we havetermed luminal-like and basal-like, that have gene expressionpatterns that appear to be consistent with differentiation states ofnormal urothelium and reflect gene expression patterns andbiology between breast and bladder cancer (2–4, 11).

Cisplatin-based chemotherapy has been the only FDA-approved therapy to treat advanced bladder cancer for over twodecades until the recent approval of immune checkpoint anti-bodies targeting the PD-1/PD-L1 axis. PD-1 axis blockade inducesa response in approximately 20% to 30% of patients withadvanced urothelial carcinoma, with the premise that activationof immune checkpoint pathways result in active immunosup-pression (12–17). Response to PD-1 axis inhibition in urothelialbladder cancer has been associated with a number of intrinsictumor features such as tumor mutational burden and tumormolecular subtype, as well as tumor microenvironment featuressuch as the presence of PD-L1–expressing tumor-infiltratingimmune cells, CD8þ cytotoxic T cells in the tumor, and expressionof effector T-cell genes by gene expression profiling (13).

Multiple immunocompetent mouse models of bladdercancer currently exist including the carcinogen-inducedmodels: MB49 (DMBA-derived cell line) and BBN [N-butyl-N-(4-hydroxybutyl)nitrosamine] (18, 19) as well as numerousautochthonous, genetically engineered murine (GEM) models(20), some of which progress to muscle-invasive bladdercancer and metastasis (21–24).

We report here the generation of a novel GEM model of high-grade, muscle-invasive bladder cancer that faithfully recapitulates

1Lineberger Comprehensive Cancer Center, University of North Carolina atChapel Hill, Chapel Hill, North Carolina. 2Department of Microbiology/Immu-nology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina.3Department of Genetics, University of North Carolina at Chapel Hill, Chapel Hill,North Carolina. 4Department of Pathology and Laboratory Medicine, Universityof North Carolina at Chapel Hill, Chapel Hill, North Carolina. 5Curriculum inBioinformatics and Computational Biology, University of North Carolina atChapel Hill, Chapel Hill, North Carolina. 6Department of Medicine, Division ofHematology/Oncology, University of North Carolina at Chapel Hill, Chapel Hill,North Carolina. 7Department of Urology, University of North Carolina at ChapelHill, Chapel Hill, North Carolina. 8Department of Pharmacology, University ofNorth Carolina at Chapel Hill, Chapel Hill, North Carolina.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

R. Saito and C.C. Smith contributed equally to this article.

Corresponding Authors: William Y. Kim, University of North Carolina, 450 WestDr., Chapel Hill, NC 27599. Phone: 919-966-4765; Fax: 919-966-8212; E-mail:[email protected]; and Benjamin G. Vincent, [email protected]

doi: 10.1158/0008-5472.CAN-18-0173

�2018 American Association for Cancer Research.

CancerResearch

Cancer Res; 78(14) July 15, 20183954

Cancer Research. by guest on September 2, 2020. Copyright 2018 American Association forhttps://bloodcancerdiscov.aacrjournals.orgDownloaded from

http://crossmark.crossref.org/dialog/?doi=10.1158/0008-5472.CAN-18-0173&domain=pdf&date_stamp=2018-6-28

https://bloodcancerdiscov.aacrjournals.org

the luminal molecular subtype of bladder cancer: Upk3a-CreERT2;Trp53L/L; PtenL/L; Rosa26LSL-Luc (UPPL) mice. This model is char-acterized by papillary histology and decreased levels of immuneinfiltration relative to basal tumors derived from BBN-treatedanimals, a pattern that is similar to human disease (3, 5, 11). Wehave generated cell line adoptive transfer models for luminal-likeUPPL tumors aswell as for basal tumors derived fromBBN-treatedanimals. Cell line–derived tumors from the UPPL model main-tain luminal-like characteristics, such as high expression of Ppargand Gata3 gene signatures. Moreover, gene expression profilesfrom BBN andUPPLmodelsmore closelymap to human bladdercancer and to normal murine urothelial cells than the commonlyused MB49 model, which appears to more closely resemblefibroblasts. As models of bladder cancer biology in immunocom-petent mice, these models can be used to interrogate subtype-specific responses to immune checkpoint inhibition and otherimmunotherapy strategies in vivo.

Materials and MethodsMouse models and establishment of mouse bladder cancer celllines

All animal studies were reviewed and approved by TheUniversity of North Carolina at Chapel Hill Institutional AnimalCare and Use Committee. For the BBN carcinogen-inducedmouse bladder cancer model, C57BL/6 mice (Charles RiverLaboratories) were continuously exposed to 0.05% N-butyl-N-(4-hydroxybutyl) nitrosamine (BBN) in drinking water. Trp53and Pten conditional knockout mice were obtained from TheJackson Laboratory (STOCK: 008462) and Terry Van Dyke(National Cancer Institute, Bethesda, MD; ref. 25), respectively,and crossed with Upk3a-CreERT2 allele (The Jackson LaboratorySTOCK: 015855) and the Rosa26LSL-Luciferase allele (The JacksonLaboratory, STOCK: 005125; UPPL model) or crossed with Krt5-CreERT2 allele (a gift from Brigid Hogan, Duke University, Dur-ham,NC) andRosa26LSL-tdTomato (The Jackson Laboratory, STOCK:007914; KPPT model). In order to induce Cre recombination inthe bladder of UPPL or KPPT mice, 5 mg of tamoxifen was givenorally by gavage in both the UPPL and KPPT model. In the KPPTmodel, transurethral injection of 4-hydroxy-tamoxifen was alsoperformed. Tumor development was regularly monitored bybladder ultrasonography.

Mice were sacrificed for the humane endpoints as follows. Forthe autochthonousmousemodels,micewere sacrificed forweightlossmore than 10%of the initial weight or tumor size diameter of>7mmas evaluatedbybladder ultrasound. Inour studies, allmicewere sacrificed because of tumor size. The endpoint for allograftmodels was tumor volume >500 mm3, skin ulcer formation, orweight loss greater than 20% body weight.

Generation of UPPL1541 and BBN963 cell linesOnce the bladder tumors became>7mmindiameter, theywere

harvested for pathologic evaluation, in vitro analysis, and forestablishing cell lines. Tumors were dissociated and digested withcollagenase anddispase (Roche). The dissociated tumor cells wereresuspended in growth media and plated to a plastic plate asdescribed previously (26). Cell lines were passaged more than 10times before use. Mycoplasma testing was performed monthlywhile cells were in culture.

MB49 cell lines were obtained from Molly Ingersol (InstitutPasteur, Paris, France). Mycoplasma testing was performedmonthly while cells were in culture.

RNA/DNA extraction, library prep, and RNA sequencing orwhole-exome sequencing

RNA was extracted from the primary tumors and the estab-lished cell lines using an RNeasy Kit (QIAGEN), and DNA wasextracted from primary tumors, established cell lines, andtail clippings using a DNeasy Kit (Qiagen). Whole-exome andtranscriptome library preparation was performed using AgilentSureSelect XT All Exon and Illumina TruSeq Stranded mRNALibrary Preparation Kits, respectively. Libraries were sequencedvia 2 � 100 runs on an Illumina HiSeq 2500 at the UNC HighThroughput Sequencing Facility.

RNA sequencing analysisSequence reads were aligned to the murine genome (mm9),

and gene expression was generated as reads per kilobase of exonmodel permillionmapped reads per genebyusingMapSplice andupper quartile normalized via RSEM (University of KentuckyBioinformatics Labs, Lexington, KY; ref. 27).

RNA sequencing (RNA-seq) data were normalized for varia-tions in read counts, log2 transformed, and median centeredbefore analysis. When combining datasets, we adjusted forbatch effects using the surrogate variable analysis R package(version 3.12.0; R Foundation). Subtype calls were made usingthe BASE47 classification algorithm based on the median-centered expression of Mus musculus homologs of genes foundin the classifier (3). Clustering was done using average linkageclustering with a centered correlation similarity metric.Immune gene signature scores were derived as described pre-viously (11).

Gene data were grouped into immune gene signatures, whichweremurine orthologs of signatures previously identified throughunsupervised clustering and gene expression profiling of sortedimmune cells (11, 28, 29). Gene data werematched to predefinedimmune gene signature clusters via Entrez IDs. Each gene signa-turewas calculated as the average value of all genes included in thesignature. Differential expression for each gene signature wasanalyzed between tumor models and treatment groups viaANOVA (one-way ANOVA), adjusted for multiple testing usingan FDR of 0.05. To determine the prognostic value of eachimmune gene signature, linear univariate correlation modelingwas used with signature/clinical variable as a continuous variablecompared with tumor size. Heat map of the log10 transformed Pvalue of gene signature correlations was displayed with colorgradient calculated via:

�log10 pValueð Þ � log10 0:05ð ÞÞ � sign of coefficient

Gata3 and Pparg gene signaturesThe PPARy gene signature was derived by determining

the genes that are significantly upregulated (samr packageFDR < 0.05) in UMUC9 cells treated with rosiglitazone, aPPARy agonist in the GSE47993 dataset (11). The GATA3 genesignature was pulled from the BIOCARTA curated gene signa-ture set in MSigDB. Gene expression data have been depositedGSE112973.

PvClustThe significance of clustering nodes was determined using the

pvclust R package (version 2.0-0, R Foundation; ref. 30). Signif-icance of all nodes was calculated with a correlation distancemetric and average linkage clustering.

UPPL Bladder Cancer GEM Model

www.aacrjournals.org Cancer Res; 78(14) July 15, 2018 3955



qPCR normalization for TCR/BCR repertoire profilingTumor RNA concentrations were determined using a Qubit

RNA BR Assay Kit, 1:200 in dilution buffer. Using a QiagenQuantitect Reverse Transcription Kit, cDNAwas synthesized from50 ng to 1 mg starting total RNA. RNA derived from column-purified T/B cells was included as a positive control, and DI H2Owas included as a negative control. The reaction was carried outaccording to the manufacturer's instructions, using a Veriti ther-mocycler (Applied Biosystems).

Quantitative PCR was performed in triplicate with 0.5 mmol/Lof each forward and reverse primers, 0.1 mmol/L TaqMan probe[T-cell receptor (TCR): FAM reporter with TAMRA quencher;B-cell receptor (BCR): VIC reporter with TAMRA quencher],cDNA (2.5 mL), and Bio-Rad SsoAdvanced Universal ProbeSupermix (2�) and DI H2O for a final volume of 10 mL perwell. Cycling conditions for TCR and BCR were both set for 45cycles of recommended TaqMan conditions for the QuantStu-dio 6 Flex system.

Purified T/B-cell cDNA was used for positive control andcalibration curve, and the template-free cDNA synthesis reactionwas used for negative control. The calibration curve was deter-mined using Ct values from purified T/B-cell cDNA, 10-foldserially diluted in nuclease-free water ranging from 1:0 (cDNA:H2O, v/v) to 1:1� 1012. For both T- and B-cell calibration curves,Ct values were detectable as dilute as 1:1� 105, with a coefficientof determination of >0.99 for the linear fit of log10(dilution)versus Ct. Each sample's Ct value was read out as the ratio of T- orB-cell cDNA to total cDNA.

50 RACE amplification of TCR/BCR sequencesBased on qPCR results, all tumor samples were normalized by

T- or B-cell RNA starting template. Using a Clontech SMARTerRACE 50/30 Kit, cDNA was generated using the manufacturer'sprotocol. cDNA was diluted with tricene/EDTA buffer, and 50

RACE was carried out using the manufacturer's protocol with0.5 mmol/L custom barcoded gene-specific reverse primer, using aVeriti thermocycler (Applied Biosystems Veriti 96-well) with thefollowing cycling conditions:

30 cycles:* 94�C, 30 seconds* 68�C, 30 seconds* 72�C, 3 minutes

50 RACE products were pooled, and clean-up/concentrationwere performed using a Zymogen Genomic DNA Clean & Con-centrator. Pools of samples were eluted in 32 mL of nuclease-freewater heated to 70�C. DNA concentration was measured using aQubit DNA HS Assay Kit. Purity (A260/280nm and A260/230nm

ratios) was determined using a ND-1000 spectrophotometer.Pooled DNA (1–5 mL) was visualized in a 1.5% agarose gelto confirm the presence of proper band sizes (TCR and BCR:400–500 bp).

TCR/BCR repertoire profilingFor TCR/BCR repertoire studies, pooled TCR or BCR amplicons

were size selected using a Sage Science Pippin Prep 1.5% agarosecassette (HTC1510). Bands were size selected at 450 to 650 bp.After size selection, samples were analyzed using either Agilent2100 Bioanalyzer or Tapestation to ensure purity. IlluminaMiSeqlibrary preparation was performed using a KAPABiosystemsDNA

Preparation Kit. Libraries were run at 6 pmol/L on an IlluminaMiSeq using a 600-cycle kit (2� 300 paired-end), with 15% PhiXspike-in.

Mouse allograft model and treatment by anti–PD-1 antibodyBBN963 and UPPL1541 cell lines were injected subcutane-

ously in C57BL/6J mice at 1 � 107 and 1 � 106 cells, respec-tively. Once tumors reached 200 mm3 in tumor volume,treatment either by anti–PD-1 antibody (clone RMP1-14,Millipore) or isotype control IgG (Sigma) by intraperitonealinjection was started. The treatment was administered once aweek at a dose of 10 mg/kg. The tumor size was measured bycaliper weekly or twice weekly.

Flow cytometryFor tissue dissociation, tissues were homogenized in cold

media using the GentleMACs Dissociator, and the samples werepassed through a70-mmcell strainer using a 5-mL syringe plunger.The samples were centrifuged for 7 minutes at 290 RCF, 4�C,decanting the supernatant. The remaining pellet was resuspendedinto 1 mL of ACK lysis buffer (150 mmol/L NH4Cl, 10 mmol/L,KHCO3, 0.1 nmol/L Na2EDTA in DPBS, pH 7.3) for 2 minutes atroom temperature before quenching with 10 mL of cold media.The samples were centrifuged for 7 minutes at 290 RCF, 4�C,resuspended in 10 mL of cold media, and passed through a40-mm cell strainer. Cell counting was performed by running adiluted aliquot of sample on a MACSQuant flow cytometer,counting lymphocytes as gated by forward scatter area versusside scatter area.

Samples were washed and resuspended in cold DPBS, nor-malized by count, and transferred onto a 96-well V-bottomplate at 2.5 � 106 lymphocytes per well. Cells were resus-pended in FVS700 viability stain (BD, 1:1,000 dilution in100 mL DPBS) for 40 minutes on ice. Wells not receivingviability staining were resuspended in DPBS. Cells werewashed twice in staining buffer (0.02% NaN3, 2% BSA inDPBS), resuspended in 50 mL Fc block (1:50 dilution instaining buffer), and incubated on ice for 15 minutes. Anti-body master mix was added to samples at 50 mL per samplewith final antibody concentrations as indicated in Supplemen-tary Table S1 (all mAbs from BD Biosciences). Please seeSupplementary Table S1 for list of antibodies.

Cells were incubated on ice in the dark for 45 minutes andwashed twice with staining buffer. Cells were fixed in 2% para-formaldehyde overnight. The following morning, a minimum of100,000 events were collected for each sample on a BD LSRFor-tessa flow cytometer. FlowJo flow software Version 10 (Treestar)was used for analyses. FluorescenceMinusOne controlswere usedto guide gating strategies.

Analysis and statisticsAll flow cytometry, TCR/BCR sharing, and Shannon entropy

statistics were calculated with Mann–Whitney U test.For TCR/BCR amplicon sequencing analyses, raw .fastq files

were demultiplexed by barcode sequences of the gene-specificprimers. Sorted R1 and R2 files were respectively merged.Sequencing quality was confirmed through the FastQC qualitycontrol tool. TCR and BCR amplicon data were analyzed viaIMGT/HighV-QUEST. Data were converted into standard in-labformat, and downstream analysis was performed with customscripts as well as the tcR R package.

Saito et al.

Cancer Res; 78(14) July 15, 2018 Cancer Research3956



Neoantigen predictionC57BL/6mice were given a single subcutaneous flank injection

of BBN963, UPPL1541, or MB49 cells. Tumor growth was mon-itored until tumors reached 100 mm3, at which point mice werehumanely sacrificed with CO2 asphyxiation followed by cervicaldislocation. Tumors were dissected for downstream DNA/RNAextraction as described above. Matched normal DNA wasextracted from tail-clippings or liver from the mouse in whichthe cell lines were respectively derived. Library prep and sequenc-ingwere performed as described above. Bioinformatics predictionof neoantigens was performed as described previously (11).Predicted neoantigens were filtered on expression in all replicateswith >5� read support.

Vaccine/ELISPOT assay for neoantigen immunogenicityPredicted neoantigen peptides were synthesized by New Eng-

land Peptide, using custom peptide array technology. C57BL/6micewere vaccinatedwithpredicted neoantigenpeptides, given asa subcutaneous injection of a pool of 8 equimolar peptides(5 nmol total peptide) and 50 mg poly(I:C) in PBS. A secondidentical injectionwas repeated 6 to 7days after primary injection.Mice were humanely sacrificed with CO2 asphyxiation followedby cervical dislocation 5 to 6 days after second injection. Spleenswere harvested andprepared into a red blood cell lysed, single-cellsuspension. Splenocytes were plated in triplicate at 5 � 105 cellsper 100 mLmedia onto an IFNg capture antibody-coated ELISPOTplate (BD Biosciences) for 48 hours, along with 1 nmol of a singlepeptide against which the respective mouse was vaccinated. IFNgexpressionwas comparedwith splenocytes incubatedwith vehiclecontrol.

Neoantigen-enriched T-cell cocultureC57BL/6 mice were vaccinated with either a pool of the top 8

predicted BBN963 neoantigens or irrelevant peptide (SIINFEKL)control, with a second identical booster given 7 days after primaryvaccine. One week after secondary vaccination, spleens wereharvested and prepared into a red blood cell lysed, single-cellsuspension. T cells were isolated usingMiltenyimurine Pan TCellIsolation Kit II. Using previously described methods (31), T cellswere expanded in the presence of bonemarrow–derived dendriticcells pulsed with a single peptide against which the derivativemouse was vaccinated against. Seven days following ex vivoexpansion, 1 � 105 T cells were cocultured 10:1 with BBN963cells onto an IFNg capture ELISPOT plate for 72 hours. Controlsincluded T cell only, BBN963 only, and media only negativecontrols, as well as antigen-enriched T cells cocultured withrespective peptide-pulsed UPPL1541 cells as positive control.Signal intensity was read out using an ELISPOT plate reader.T-cell/BBN963 coculture spot counts were subtracted from theirrespective T-cell only control, and then taken as a percentage ofthe counts from their respective peptide-pulsed target positivecontrols.

ResultsInactivation of Pten and Trp53 in Uroplakin3a-expressing cellsresults in muscle-invasive, high-grade urothelial carcinoma

Our previously published studies describing luminal-like andbasal-like molecular subtypes of bladder cancer demonstratedthat these subtypes reflect the gene expression patterns of thedifferentiation states of the normal urothelium (3). Basal-like

bladder tumors harbor gene expression patterns most similar tobasal and intermediate cell layers of the bladder, whereas luminaltumors harbor gene expression patterns most similar to umbrellacells (2, 3). To determine whether different cells of origin accountfor the differential gene expression patterns between basal-likeand luminal-like high-grade, muscle-invasive bladder cancer,we conditionally inactivated Pten and Trp53 in Keratin5 (K5)or Uroplakin3a (Upk3a)-expressing basal/intermediate andumbrella/intermediate cell layers, respectively, using previouslyreported K5-CreERT2 (32) and Upk3a-CreERT2 (The JacksonLaboratory) transgenic mice. Dual inactivation of Pten and Trp53by surgical injection of adenoviral cre into the bladder has beenpreviously shown to induce bladder cancer in mice (24).Using standard animal husbandry, we generated cohorts ofUpk3a-CreERT2; Trp53L/L; Pte/L/L; Rosa26LSL-Luc mice (hereaftertermed "UPPL") as well as K5-CreERT2; Trp53L/L; PtenL/L;Rosa26LSL-tdTomato mice (hereafter called "KPPT") that were back-crossed 10 times to a C57BL/6 background. BothUPPL and KPPTmice were gavaged with tamoxifen every other day for 3 dosesstarting at 6 to 8 weeks of age to induce CreERT2 activity. Serial invivo luminescence (forUPPLmice) and ultrasound of the bladderwere used to monitor for tumor development and growth. UPPLmice demonstrated gradually increasing luminescence signal overtime in the region of the bladder (Fig. 1A and B). In addition, byultrasound, papillary-appearing tumors began to be apparent at amedian of 58 weeks (Fig. 1C and D).

In contrast, KPPTmice administered tamoxifen by gavage diedrapidly of epithelial hyperplasia of the snout, paws, and papillaryskin lesions (Supplementary Fig. S1A–S1C). This likely representsthe inactivation of Pten and Trp53 (and pursuant epithelialovergrowth) in K5-expressing basal cells in multiple organsincluding the epidermis, trachea, and gastrointestinal tract. In anattempt to activate K5-CreERT2 solely in the K5-expressing basalcells of the bladder, we administered 4-hydroxytamoxifen(4-OHT) intravesically at various concentrations (2,000 and200 nmol/L). Mice injected with intravesical 4-OHT at2,000 nmol/L exhibited a similar but attenuated phenotype toKPPT mice that had been gavaged with tamoxifen (Supplemen-tary Fig. S1B and S1C) and had a shortened survival. In contrast,KPPTmice injectedwith intravesical 4-OHT at 200 nmol/L had anextended survival but did not develop bladder tumors despiteCre-mediated recombination as evidenced by increasedtdTomato signal over the region of the bladder by IVIS imaging(Supplementary Fig. S1D). Moreover, histologic examination ofthe bladders of mice injected with intravesical 4-OHT at 2,000 or200 nmol/L showed no significant histologic changes of theurothelium (Supplementary Fig. S1E).

Approximately 95%ofUPPLmice developed tumors within 77weeks (Fig. 1D). Grossly, bladder tumors in UPPLmice appearedto be papillary in nature (Fig. 1E), which is a feature documentedto be enriched in the luminal-like molecular subtype (2, 4).Histologically, the UPPL tumors were characterized as high gradeby an expert genitourinary pathologist (S.E. Wobker) and werefound to be of varying tumor stage (Fig. 1F) as well as rarelymetastatic (Fig. 2A). UPPL tumors also hadmicroscopic papillaryfeatures (Fig. 2B–D) and some had prominent squamous differ-entiation (Fig. 2D and E). In addition, UPPL tumorswere noted tohave different depths of invasion into the bladder wall includingboth lamina propria invasion (Figs. 1F and 2F) and muscularispropria invasion (Fig. 2G). In keepingwith the known field defectof urothelial tumors in human disease, tumors were also noted to





form in the renal pelvis and ureters of about a third of mice (Figs.1E and 2A, H, and I). Finally, rare macroscopic metastases wereseen (Figs. 2A, J, and K).

BBN and UPPL models are basal- and luminal-like models,respectively, of human bladder cancer

Bladder tumors induced by the carcinogen N-Butyl-N-(4-hydroxybutyl) (BBN) have been previously documented to har-bor a number of histologic features (e.g., squamous differentia-tion, Supplementary Fig. S2A–S2D) and gene expression patternsknown to be found in basal-like bladder tumors (7).We thereforeestablished 11 independent BBN-induced bladder tumors bycontinuously administering 0.05% BBN in drinking water asdescribed previously (19). Given the papillary nature of UPPLtumors, we hypothesized that they correspond to a luminal-likemolecular subtype. We therefore performed global transcriptomeprofiling of 9 UPPL and 11 BBN mouse tumors using RNA-seq.We first performed molecular subtype classification using ourpreviously published BASE47 (bladder cancer analysis of sub-types by gene expression; ref. 3) subtype classifier and found that

8 of the 9 UPPL tumors had high correlation to the luminalcentroid of gene expression (Fig. 3A). To further validate ourobservation, we coclustered the UPPL and BBN murine tumorswith human tumors from the The Cancer Genome Atlas (TCGA;n ¼ 408) using genes with corresponding homologs across thespecies and found that the majority of UPPL and BBN tumorscoclustered with human luminal-like tumors and basal-liketumors, respectively (Fig. 3B).

Currently, very few cell lines exist for modeling bladder cancerin immunocompetent mice; therefore, we set out to generateadditional cell lines that could be utilized in future studies. Inparticular, MB49 cells have long been the workhorse of syngeneicbladder cancer cell lines (18) for studies requiring an immuno-competent host. Given the long latency of tumor formation in theUPPLmodel, we established tumor cell lines frombothUPPL andBBN tumors using the conditional reprogramming of cells (CRC)method described previously (26). Specifically, transplantablecell lines were established from BBN (BBN963) and UPPL(UPPL1541) tumors (Fig. 3C) and have been confirmed to growinC57BL/6mice. In parallel, using theCRCmethod,we generated

4 wks

8 wks

36 wks

48 wks

56 wks

A UPPL B C

D

54 wks

68 wks

72 wks

76 wks

UPPL

Mean luminescence UPPL

0 20 40 60

106

107

108

109

Time post tamoxifen (wks) Lu

min

esce

nce

(p/c

m2 /

sec/

sr)

0 20 40 60 800

50

100

Time post tamoxifen (wks)

Tum

or-fr

ee s

urvi

val (

%)

E

R KidneyL Kidney

Bladder F

TaT1

37%

50%

13%

T2 or greater

Figure 1.

Inactivation of Pten and Trp53 in Upk3a-expressing cells results in high-grade muscle-invasive bladder tumors. A, Bioluminescent images of UPPL mice atindicated time points. B,Quantification of luminescence over the region of the bladder. C, Ultrasound images of bladder tumor formation. D, Tumor-free survival asdetected by ultrasound. E, Gross images of the kidneys and bladder from a tumor-bearing UPPL mouse. F, Tumor stage assessed histologically based onhuman TNM staging.

Saito et al.




three primary cell lines derived fromnormalmouseurotheliumoftamoxifen-treated K5-CreERT2; Rosa26LSL-tdTomato mice, hereaftercalled KT mice (KT1044, KT1975, and KT1970) as a normalreference for comparison. Interestingly, the vast majority of epi-thelial cells that grew in vitro from CRC culture of KT mouse

bladders expressed tdTomato, suggesting they at some point hadexpressed K5 (Supplementary Fig. S3).

To assess the similarities between our newly generatedmodels and MB49 cells, we performed whole-transcriptomeprofiling on the BBN963 and UPPL1541 cell lines, MB49 cells,

Bladder

tumor

Upper tra

ct tumor

Distan

t meta

stasis

0

50

100

PositiveNegative

% o

f Cas

es

100 μmol/L 200 μmol/L

B C

D E F

G H I

A

J K

100 μmol/L 200 μmol/L 200 μmol/L

200 μmol/L200 μmol/L 100 μmol/L

100 μmol/L 200 μmol/L

Figure 2.

Representative histology of UPPL tumors. A, Bar graph indicating the percentage of UPPL mice that developed bladder tumors, upper tract tumors, anddistantmetastases at the timeof sacrifice.B, Low-power viewof papillary-appearingUPPL tumor.C,High-power viewof papillaryUPPL tumor.D, Low-power viewofpapillary tumor with squamous histology. E, High-power view of squamous histology. F, UPPL tumor showing lamina propria invasion. G, UPPL tumor withmuscularis propria invasion. H, Upper tract tumor demonstrating invasion into the renal parenchyma. I, High-power view of urothelial tumor invading renalparenchyma. J, Cervical lymph node metastases. K, High-power view of cervical lymph node metastases.





BASE47 Centroid correlationC

orre

latio

n to

bas

al c

entr

oid

−1.0

−0.5

0.0

0.5

1.0

BBN primary tumor UPPL primary tumor

Luminal (human)Basal (human)

UPPL (mouse)

BBN (mouse)

BA

SE

47 G

enes

Basal LuminalBBN 10 1

UPPL 1 8M

B49

BB

N96

3

UP

PL1

541

Tumor fragment Feeder cells

Epithelial islands

Epithelial islands

Feeder cells

Immortalizedcell lines

3T3 MB49 BBN963 UPPPL1541

A B C

D E

I

H

-3.00-2.00 -1.00 0.00 1.00 2.00 3.00

MB

49B

BN

963

UP

PL1

541

KT1

044

KT1

975

KT1

970

−2 −1 0 1 2

−2 0 1 2-1

GMB49 -vs. BBN963

MB49 -vs- UPPL1541

F Gata3 Activity

BBN963 UPPL1541

0.3

0.2

0.1

0.0

-0.1

Gat

a3 s

igna

ture

Z-s

core

Ppa

rg s

igna

ture

Z-s

core

BBN963 UPPL1541

Pparg Activity

0.2

-0.2

0.0

-0.4 P = 0.002 P = 0.016

ILK Signaling

Glucocorticoid receptor signaling

Role of tissue factor in cancer

PKCÎ¸ Signaling in T lymphocytes

p53 Signaling

Semaphorin signaling in neurons

Hepatic fibrosis/hepatic stellate cell activation

GM−CSF Signaling

Regulation of the epithelial−mesenchymal transition pathway

B Cell activating factor signaling

0 1 2 3 4 5 6 7

Axonal guidance signaling

Molecular mechanisms of cancer

Role of macrophages, fibroblasts and endothelial cells in rheumatoid arthritis

Glioblastoma multiforme signaling

p53 Signaling

Regulation of the epithelial−mesenchymal transition pathway

Role of osteoblasts, osteoclasts and chondrocytes in rheumatoid arthritis

Glioma signaling

Breast cancer regulation by stathmin1

Wnt/Î²−catenin Signaling

−Log (P value)

−Log (P value)

0 2 4 6 8 10 12

EpCAM

Cou

nt

Actin

K5

K14

E-Cadherin

Vimentin

3T3J

2M

B49

MBT2

BBN963

Positive z-score

Negative z-scoreNo activity pattern available

z-score = 0

Ratio

Figure 3.

BBN and UPPL tumors recapitulate the human basal and luminal molecular subtypes of bladder cancer. A, Waterfall plot of the correlation to the basal centroidfor BBN and UPPL primary tumors, with an accompanying confusion matrix indicating the subtype calls. B, Unsupervised clustering heatmap of BBN andUPPL primary tumor samples with the TCGA BLCA dataset across genes in the BASE47 classifier. C, Representative flowchart of the workflow to transition aprimary tumor extracted fromamousemodel to a cell line anda cell line–derived tumor.D,UnsupervisedclusteringheatmapofMB49, BBN963, UPPL1541, andnormalurothelium (KT) cell lines across the top 10% most differentially expressed genes across the samples. E, Unsupervised clustering heatmap of MB49, BBN963,and UPPL1541 cell line–derived tumors across the top 10% most differentially expressed genes across the samples. F, Box plots of Pparg and Gata3 genesignature scores from RNA-seq data of BBN963 and UPPL1541 cell line–derived tumors. G, IPA analysis plots showing activated pathways in MB49 cell line–derivedtumors relative to BBN963 and UPPL1541 cell line–derived tumors. H, Flow cytometry plot for EpCAM expression in 3T3, MB49, UPPL1541, and BBN963cell lines. I, Western blot of whole-cell lysates from 3T3, MB49, MBT2, and BBN963 cell lines blotted for the indicated antibodies.

Saito et al.




3T3 cells, and three primary mouse urothelial cell lines(KT1044, KT1975, and KT1970). Unsupervised hierarchicalclustering of the cell lines on differentially regulated genesacross samples (Supplementary Table S2) demonstrated thatMB49 cells had transcriptome profiles that differed significant-ly from the other cell lines (Fig. 3D) when tested by multiscalebootstrap resampling (P ¼ 0.0, Supplementary Fig. S4A),whereas BBN963 and UPPL1541 cells had transcriptome pro-files that more closely resembled normal urothelial (KT) cells.To ensure that we had not tainted our MB49 cells, we obtainedMB49 cells from an independent source (Phil Abbosh, FoxChase Cancer Center, Philadelphia, PA) and performed tran-scriptome profiling. We found that the transcript level (acrossall genes) is highly correlated when comparing "UNC MB49"with "FCCC MB49" (R ¼ 0.94 respectively, Supplementary Fig.S4B), suggesting our MB49 cells were genuine. Intriguingly,hierarchical clustering of MB49 cells with 3T3 cells, our threeprimary mouse urothelial cell lines, BBN963 cells, andUPPL1541 cells demonstrated that the MB49 cells coclusteredwith 3T3 cells (Supplementary Fig. S4C) significantly byPVClust (Supplementary Fig. S4D). To examine the RNAexpression profiles of these cell lines in the context of thetumor microenvironment, we generated RNA expression dataon cell line–derived tumors from MB49, BBN963, andUPPL1541 cells. Clustering of these cell line–derived tumorsusing differentially regulated genes (Supplementary Table S3)again demonstrated that MB49 tumors have significantlydivergent transcriptome profiles when tested by multiscalebootstrap resampling (P ¼ 0.0, Supplementary Fig. S5A) anddemonstrate that this finding is not merely an artifact of cellculture (Fig. 3E). Finally, reassuringly, Pparg and Gata3 genesignatures were upregulated in UPPPL1541 tumors relative toBBN963 tumors (Fig. 3F), demonstrating that cell line–derivedUPPL1541 tumors maintain molecular features of a luminal-like molecular subtype.

We next performed Ingenuity Pathway Analysis comparingMB49, UPPL1541, and BBN963 cell line–derived tumors. Gen-eral pathways related to cancer were enriched in BBN963tumors relative to UPPL1541 tumors (Supplementary Fig.S5B). In contrast, pathways related to fibrosis and epithelial-to-mesenchymal transition (EMT) appeared to be highly upre-gulated in MB49 cell line–derived tumors compared with eitherthe BBN963 or UPPL1541 tumors (Fig. 3G). Based on theseobservations, we examined the expression of a set of epithelialmarkers in the mouse bladder cell lines. Assessment of EpCAMby flow cytometry demonstrated that a significant proportionof BBN963 and UPPL1541 cells expressed cell surface EpCAMwhile MB49 cells had little to no EpCAM expression, similar tothe mouse fibroblast line 3T3 (Fig. 3H). In keeping with thisfinding, we also noted that MB49 cells did not express K5 orK14 in immunoblots of whole-cell lysates (Fig. 3I), implyingthat MB49 cells have lost characteristic urothelial cytokeratinexpression patterns potentially from undergoing EMT. Further-more, we noted that MB49 cell line–derived tumors had rel-atively high and low expression of vimentin and Cdh1(E-cadherin), respectively, consistent with MB49 cells beingmore mesenchymal than BBN963 and UPPL1541 cell line–derived tumors (Fig. 3I; Supplementary Fig. S6). In aggregate,these findings suggest that MB49 cells and tumors more closelyresemble fibroblasts than urothelial cells and highlight thepotential benefit of our models.

BBN963 tumors demonstrate evidence of an antigen-drivenT-cell response

Human basal-like and luminal-like bladder cancers demon-strate different patterns of immune infiltration and are alsocorrelated with differential response to checkpoint inhibitortherapy (11, 13), suggesting subtype-specific differences in thetumor-immune microenvironment. Immune gene signatureexpression derived from previously published studies were com-pared among 11 BBN and 9 UPPL models (11, 28, 29, 33).Consistent with immune gene signature patterns observed inhuman tumors, BBN (basal-like) tumors demonstrated greateroverall expression of immune gene signatures (Fig. 4A; Supple-mentary Fig. S7A) than did UPPL (luminal-like) tumors, includ-ing those for T cells, B cells, dendritic cells, other innate immunecells, and immunosuppression (11).

To further explain the observed immunologic differencesbetween BBN and UPPL tumors, we performed flow cytometricanalysis in cell line–derived BBN963 and UPPL1541 tumors.Comparing the frequency of tumor infiltrating lymphocytes (TIL)by flow cytometry from BBN963 and UPPL1541 tumors, weobserved significantly greater frequencies of CD3þ T cells (Fig.4B), as well as increased ratio of CD8þ cytotoxic T cells to CD4þ

helper T cells (Fig. 4C) in BBN963. Somewhat surprisingly, theproportion of B cells was higher in UPPL tumors; however, theoverall proportion of B cells in the lymphocytic infiltrate was low(Fig. 4D). To further characterize the phenotype of the tumor-infiltrating T cells, expressions of CD44 and CD62L were used toidentify na€�ve (CD44�, CD62Lþ), central memory (CM; CD44þ,CD62Lþ), and effector memory (EM; CD44þ, CD62L�) popula-tions. Among CD4þ T cells, UPPL1541 tumors were enriched forna€�ve T cells, whereas the frequency of the total memory pool(CD44þ) was significantly greater in BBN963 tumors (Fig. 4E).Moreover, EM and CM frequencies both trended higher inBBN963. AmongCD8þT cells, theCM frequencywas significantlygreater in BBN963. In addition, CD4þ FoxP3þ regulatory T cellstrended toward higher frequency in BBN963. Thus, memorysubpopulations of both CD8þ and CD4þ T cells had increasedfrequencies in the BBN tumors, suggesting the presence of anantigen driven T-cell response in BBN963.

Analyzing BBN963 and UPPL1541 cell line–derived tumors byTCR repertoire profiling, BBN963 tumors demonstrated a higherdegree of clonotype sharing between animals (Fig. 4F and G),suggesting that there may be greater convergent repertoire selec-tion in BBN tumor-infiltrating T cells in the context of an antigen-driven response. To examine whether the increased immuneinfiltration and TCR repertoire sharing seen in BBN tumors wereassociated with the number of targetable tumor antigens, weperformed neoantigen prediction on BBN963 and UPPL1541cells and not unexpectedly observed significantly higher neoanti-gen burden in BBN963 compared with UPPL1541 (Fig. 4H;Supplementary Tables S4 and S5). This, in combination with theincreased TCR repertoire sharing, further supports the hypothesisthat the immune infiltration seen in BBN tumors is driven by anantigen-specific immune response.

To investigate the functional significance of immune infiltrat-ing T-cell subpopulations,weperformedunivariable linear regres-sion with frequency of T-cell phenotypic subpopulations as acontinuous predictor variable and tumor mass as the responsevariable in untreated mice. In BBN963, the frequency of total andna€�veCD8þT cellswas positively associatedwith tumormass, andthe frequency of CD4þmemory, CD4þCM, total CD8þmemory,





andCD8þEMTcellswere all inversely correlatedwith tumormass(Supplementary Fig. S7B). In UPPL1541, no features were pos-itively associated with tumor mass, whereas CD8þ total memoryand specifically CD8þ CM T cells were both weakly inversely

correlated with tumor mass. These associations are suggestive oftumor-infiltratingmemory T cells being functional and capable ofantitumor activity in both BBN andUPPL, with greater functionalsignificance in BBN963.

0

5

10

15Percentage CD3+

0.0

0.5

1.0

1.5

2.0Percentage B220+

0.0

0.5

1.0

1.5

FrequencyCD8+:CD4+ T cell

CD8+ EM

CD8+ CM

CD8+ Total memory

CD8+ Naïve

Total CD8+

CD4+ FoxP3+

CD4+ EM

CD4+ CM

CD4+ Total memory

CD4+ Naïve

Total CD4+

BBN963

UPPL1541

-0.0002

0.0000

0.0002

0.0004

0.0006

0.0008

TCR Sharing

0e+00

2e−04

4e−04

6e−04Shared clonotypes

TCR sharing

Immunosuppression

T Cell

NK Cell

Macrophage

Granulocyte

Dendritic cell

B Cell

Color key

BBNUPPL

BBN96

3UPP

L154

1

BBN96

3UPP

L154

1

BBN96

3UPP

L154

1

PredictedNeoAg burden

−3 −1 1 2 3-2 0

A B C D

E F

G H

T1T2T3T4T5T6T1T2T3T4T5T6

T1

T2

T3

T4

T5

T6

T1

T2

T3

T4

T5

T6

BBN963

BB

N96

3

UP

PL1

541

UPPL1541

0 20 40 60 80 100

T-cell phenotypes

Percentage of parent

BBN963UPPL1541

***

*

**

*

MB49

BBN963

UPPL1541

0

100

200

300

400

Pred

icte

d ne

oant

igen

cou

nt

Figure 4.

Immune characterization of UPPL1541 and BBN963 subtype-specific bladder models. A, Immune gene signature expression across 9 UPPL and 11 BBN primary tumors.B, Flow cytometric characterization of tumor-infiltrating T cells (CD3þ) in cell line–derived (allograft) BBN963 and UPPL1541 tumors. Each datapoint represents anindependentmouse.C,FrequencyofCD8þ toCD4þTcells in cell line–derivedBBN963andUPPL1541 tumors.D,Flowcytometric characterizationof tumor-infiltratingBcells(B220þ) in cell line–derived BBN963 and UPPL1541 tumors. E, T-cell phenotypic subpopulations in cell line–derived BBN963 and UPPL1541 tumors, with significantlyincreased phenotypes highlightedwith respective colors (Mann–WhitneyU test, � , P < 0.05; ��, P < 0.001). F andG,Heatmap (F) and respective quantification (G) of T-cellreceptor clonotype sharingwithin cell line–derivedBBN963andUPPL1541 tumors, derived fromwhole-tumorRNA-basedT-cell receptor amplicon sequencing.H,Predictedneoantigen burden (class I and II, >500 nmol/L predicted binding affinity) in MB49, BBN963, and UPPL1541 cell line–derived tumors.

Saito et al.




BBN963 tumors stratify by response to PD-1 axis inhibitionThe relative overexpression of immune gene signatures and

evidence of an antigen driven T-cell response in BBN963 cell line–derived tumors are suggestive of possible greater responsivenessto immune checkpoint inhibitor therapy inBBN963. Accordingly,we observed dramatic decreases in mean tumor volume inBBN963 following anti–PD-1 therapy, while UPPL1541 tumorsdemonstrated only modest control of tumor growth (Fig. 5A andB). Despite the mean tumor size being substantially controlled inBBN963 following anti–PD-1 therapy, the growth pattern ofindividual tumors demonstrated a mixed-response pattern (Fig.5C). Therewas heterogeneity of anti–PD-1 response inUPPL1541tumors as well (Fig. 5D). The difference in responsivenessbetween BBN963 and UPPL1541 tumors did not appear to besecondary to differential expressionof PD-L1 as bothBBN963 andUPPL1541 cell lines upregulated PD-L1 expressionwhen exposedto IFNg (Fig. 5E).

In order to elucidate the immune correlates of these twophenotypes in response to anti–PD-1 therapy, we repeatedanti–PD-1 antibody treatments (Fig. 6A) and analyzed the TILpopulations among anti–PD-1 responder and nonresponderBBN963 tumors once response class could be determined on dayþ14 following tumor inoculation. Surprisingly, no significantchanges were observed in the overall T- and B-cell infiltrationfrequencies by flow cytometry in responders versus nonrespon-ders (Fig. 6B). In addition, phenotyping of tumor-infiltrating Tcells demonstrated only significantly greater frequencies of totalCD4þ among nonresponders, with subtle, nonsignificant varia-tions among other T-cell phenotypic subpopulations (Fig. 6C).Despite these minimal differences, comparison of subpopulationratios demonstrated an overall significant increase in the frequen-cy of total memory-to-regulatory T cells as well as significantlyhigher ratios of CD8þ to CD4þ T cells in responders (Fig. 6D).To further examine the role of TILs among responder and

BA

0

100

200

300

400

Days posttreatment start Days posttreatment start

% T

umor

vol

ume

(mea

n)0

500

1,000

1,500

% T

umor

vol

ume

(mea

n) IgG (n = 8)PD1Ab (n = 8)

BBN963 UPPL1541

P = 0.0003 (ANOVA) P = 0.9127 (ANOVA)

IgG (n = 6)

PD1Ab (n = 6)

0 7 14 21 280 4 7 11 14

BBN963 UPPL1541

IFNγ + Anti–PD-L1

Anti–PD-L1

IFNγ + Control IgG

C D

E

BBN963 Anti–PD-1treated only

UPPL1541 Anti–PD-1treated only

400

300

200

100

00 3 6 10 14

% T

umor

vol

ume

% T

umor

vol

ume

Days posttreatment start Days posttreatment start

2,000

1,500

1,000

500

0

PD1Ab (n = 6)PD1Ab (n = 8)

0 7 14 21 28

Nor

mal

ized

to m

ode

Nor

mal

ized

to m

ode

100

80

60

40

20

0

100

80

60

40

20

0-103 0 103 104 105 -103 0 103 104 105

Figure 5.

Anti–PD-1 treatment of BBN963 andUPPL1541 cell line–derived tumors. A,Mean tumor volume of C57BL/6 micebearing BBN963 tumors treated witheither control IgG or anti–PD-1antibody. Anti–PD-1 treatment wasbegun when tumors reached200 mm3. B, Mean tumor volume ofC57BL/6 mice bearing UPPL1541tumors treated with either control IgGor anti–PD-1 antibody. Anti–PD-1treatment was begun when tumorsreached 200mm3.C, Tumor volumeofindividual mice from A. D, Tumorvolume of individual mice from B. E,BBN963 and UPPL1541 cells weretreated with IFNg , and flow cytometrywas used to detect cell surface PD-L1expression.





0 10 20 300

200

400

600

Days posttreatment start

% T

umor

vol

ume

BBN963PD (progressive disease)

SD/PR (stable disease/partial response)

CR (complete response)

A B

C D

Untreate

d contro

l

Responders

Nonresponders

Nonresponders

0

2

4

6

8

Percentage CD3+

Untreate

d

Responders

-2

0

2

4

6

8

Percentage B220+

0

2

4

6

8

FrequencyCD8+:CD4+ T cells

0

2,000

4,000

6,000

FrequencyTmem:Tregs

E F

2

3

4

5

IgH Shannonentropy index

Sha

nnon

ent

ropy

inde

x

Responder

Nonresponder

Responder

Nonresponder

Responder

Nonresponder

BCR SharingT1

T2

T3

T4

T5

T6

T1

T2

T3

T4

T5

T1

T2

T3

T4

T5

T6

T1

T2

T3

T4

T5

Responder

Nonresponder

Res

pond

er

Non

resp

onde

r

Significance

fdr_nr_siggenesfdr_re_siggenes

TgdTFHTemTcmpDC

NK_CellsNK_CD56dim_Cells

NK_CD56bright_CellsNeutrophilsMast_Cells

iDCEosinophils

DCaDCTIC

MacTh1_ClusterMacrophages

Mac_CSF1LCK

CD68_ClusterIL_8

Cytotoxic_CellsTReg

Th17_CellsTh2_CellsTh1_Cells

T_Helper_CellsGO_TCR_Signaling

T_CellsT_Cell

T_Cell_ClusterCD8_T_CellsCD8_Cluster

CD8IGG_Cluster

GO_BCR_SignalingBCell_60gene

B_CellsB_Cell

B_Cell_ClusterImmunosuppression

NRR

Pre

dict

ors

Univariate correlationBBN963 IGS vs. tumor size

0 20 40 60 80 100

CD8+ EM

CD8+ CM

CD8+ Total memory

CD8+ Naïve

Total CD8+

CD4+ FoxP3+

CD4+ EM

CD4+ CM

CD4+ Total memory

CD4+ Naïve

Total CD4+

T-cell phenotypes

Percentage of parent

RespondersNonresponders

* *

**

*

G

0.5

0.0

-0.5

Figure 6.

Description and immune characterization of BBN963 mixed-response phenotype. A, Tumor growth curves of anti–PD-1 treated, cell line–derived BBN963tumors, showing responders (black) and nonresponders (red) to therapy. B, Flow characterization of tumor-infiltrating T cells (CD3þ) and B cells (B220þ) inuntreated (black), responder (red), and nonresponder (blue) BBN963 tumors.C, T-cell phenotypic subpopulations in cell line–derivedBBN963 andUPPL1541 tumors,with significantly increased phenotypes highlighted with respective colors (Mann–Whitney U test, � , P < 0.05). D, Frequency of memory to regulatory T cells(CD3þCD44þ: CD3þCD4þFoxP3þ) and CD8þ to CD4þ T cells in responder versus nonresponder BBN963 tumors. E, Univariable correlation of tumor sizeto immune gene signature expression in responder (left) and nonresponder (right) BBN963 tumors. Shannon entropy index (F) and receptor clonotype sharing (G)of tumor-infiltrating B-cell receptor heavy chain expression in responder and nonresponder BBN963 tumors.

Saito et al.




nonresponder tumors, we calculated immune gene signaturesderived from total tumor RNA-seq and independently correlatedthese signatures to tumor mass. Although nonresponders onlydemonstrated modest inverse correlation between a single CD8þ

T-cell immune signature and tumor mass, responder tumor masswas inversely correlated with multiple immune cell signatures,most significantlywith cytotoxic T-cell andCD8þ T-cell signatures(Fig. 6E). These data in aggregate demonstrate the potentialimportance of the balance of effector to suppressor T-cell sub-populations, rather than just absolute numbers, in mediating anantitumor response in responding BBN963 tumors.

To address the role of clonality of the tumor-infiltratinglymphocytes in controlling tumor growth in anti–PD-1responsive BBN963 tumors, we performed TCR and BCRrepertoire profiling of responder and nonresponder wholetumor RNA. We observed a modest, nonsignificant increasein TCR clonotype sharing (Supplementary Fig. S8A and S8B)and no differences in Shannon entropy (Supplementary Fig.S8C) between responder and nonresponder tumors. In con-trast, BCR profiling demonstrated significantly lower Shannonentropy indices (Fig. 6F) and significantly greater clonotypesharing among responders (Fig. 6G), suggesting a potentiallyimportant role for B-cell clonal shift in mediating response inthe BBN963 model.

BBN963 and UPPL1541 models express targetable,immunogenic neoantigens

With the recent interest in neoantigens as biomarkers of immu-notherapy response and therapeutic targets for personalizedimmunotherapy, we sought to identify and validate neoantigentargets in our subtype-specific bladder models. Using previouslydescribed methods (11), neoantigens were predicted in BBN963,MB49, and UPPL1541 cell line–derived tumors (Fig. 7A), select-ing for predicted class I and II binders by NetMHCpan andNetMHCIIpan (affinity < 500 nmol/L). Using 5x RNA-seq cov-erage and expression in all replicates as cutoffs, we observedBBN963 to have the greatest number of class I (48) and class II(18) predicted neoantigens, followed by MB49 (I: 29; II: 8), andlastly by UPPL1541 (I: 2; II: 5). To validate the immunogenicpotential of these predicted neoantigens, we synthesized 96 of110 predicted neoantigens and performed vaccination/ELISPOTanalyses to identify the ability of each peptide to induce an IFNgresponse in T cells stimulated by neoantigen peptide-pulseddendritic cells. Mice were vaccinated with a pool of eight random,equimolar peptides, and splenocytes derived from vaccinatedmice were subsequently pulsed with one of the eight peptideson an IFNg capture ELISPOT plate (Fig. 7B and C). Based on thenumber of spots induced by each peptide, we observed MB49 tocontain the most highly immunogenic class I and II peptides,holding eight of the top 10 neoantigens by IFNg response. This isfollowed by BBN963, and lastly by UPPL1541, which containedonly one peptide within the top 15. Class I and II neoantigenswere equally represented among the top binders, with four and sixof 10 top neoantigens predicted as class II and class I, respectively.Finally, to test the potential for these peptides to generate neoanti-gen-enriched T-cell populations, we performed two rounds ofvaccination in wild-type C57BL/6 mice using the top eightBBN963 neoantigens, followed by ex vivo stimulation using oneof the respective peptides. Coculture of these neoantigen-enrichedT cells with BBN963 tumor cells demonstrated an IFNg responseover that of irrelevant peptide control in five of eight peptides,

emphasizing the potential of these peptides for use in neoantigen-based BBN963 treatment models (Fig. 7D).

DiscussionWe describe here the first molecular subtype-specific murine

model of luminal-like bladder cancer. The BBN model wasderived from carcinogen-treated mice that develop spontaneousbladder tumors and exhibit a basal phenotype as previouslydescribed by others (7). We extend these findings by demonstrat-ing that the BBN model also has an immune infiltration patternthat is consistentwith that of humanbasal tumors (11). TheUPPLmodel was derived from mice with directed knockout of Trp53and Pten in the urothelial umbrella cells under the control of theUroplakin-3a promoter. UPPL tumors have papillary histology,decreased immune infiltration, and decreased response to PD-1inhibition relative to BBN tumors. Characterized by gene expres-sion profiling, these models reflect human bladder cancer andnormal urothelium more closely than does the commonly usedMB49model, which appears tomore closely resemble fibroblasts.These results imply the BBN and UPPL models will prove to bevaluable resources in studying bladder cancer in immunocom-petent animals, and they will be a unique resource with which tostudy molecular subtype-specific biology and treatment effects.

Although dual inactivation of Trp53 and Pten in Upk3a-expres-sing cells led to robust bladder tumor formation, inactivation ofthese same genes in K5-expressing cells did not result in anyapparent neoplasia or preneoplastic changes. Althoughwehad setout with the goal of answering whether inactivation of Trp53 andPten in K5-expressing basal and intermediate urothelial cells waspermissive for tumorigenesis, we are hesitant to conclude toomuch from our negative findings given the significant technicaldifferences between how Trp53 and Pten were inactivated inUpk3a-expressing cells (systemic tamoxifen by gavage) and K5-expressing cells (local 4-OHT administration into the bladder).Nonetheless, at face value, our results suggest that dual inactiva-tion of Trp53 and Pten are sufficient to initiate bladder tumors inUpk3a-expressing but not K5-expressing urothelium.

The immune microenvironments of the mouse models alsoreflect patterns seen in human disease, with increased overallimmune infiltration seen in the BBN basal-like model (11). Inparallel with the active immune response, BBN tumors showedincreased expression of genes associated with immunosuppres-sion, which is presumably an adaptive response to suppressand/or evade antitumor immunity. In the BBN (basal-like) mod-el, we noted increased memory polarization, which is compli-mented by significantly higher expression of EM T-cell immunegene signatures in human basal versus luminal tumors amongTCGA BLCA samples (Supplementary Fig. S9). These patternsmirror those observed in our prior study of human bladder cancer(11) and imply that tumor immunobiology and mechanisms ofresistance to immunotherapy may differ by tumor molecularsubtype. Importantly, there were a number of aspects noted inthedescribedmousemodels that havenotbeen as yet examined inhuman tumors. In the BBN (basal-like) model, we notedincreased T-cell clonotype sharing, suggesting the presence of anactive antigen-driven response in these tumors. In contrast, UPPL(luminal-like) tumors showed decreased overall immune infil-tration along with an increased frequency of na€�ve T cells, con-sistent with immune exclusion and lack of antigen experience,respectively.





Our study also highlights several limitations of the widely usedMB49 murine bladder cancer model. We demonstrate that MB49tumors show lack of characteristic urothelial cytokeratins, lack ofEpCAM expression, and a profound skewing toward havingundergone epithelial-to-mesechymal cell transition. Comparedwith BBN and UPPL tumors, MB49 had a wide transcriptomicdistance from both normal murine urothelial cells and moreclosely resemble immortalized fibroblasts. These results suggestthat althoughMB49may be adequate (or in some cases preferred)for studying someaspects of bladder cancer biology (e.g., the post-EMT state), themodels reported in our study should gainwide usein the translational bladder cancer research community, both fortheir subtype specificity and increased fidelity to human bladdercancer gene expression profiles.

Although recent work highlights the mutational faithfulnessof the BBN model to human bladder cancer (34) one limitationof the UPPL1541 and BBN963 models is the lack of drivermutations that are also known drivers in human basal-like andluminal-like bladder tumors. Major bladder cancer driver muta-tions such as MLL, FGFR3, and ARID1A are notmutated in eithermodel, and suspected driver mutations in our BBN963 line suchas TCF4, FGFR2, and ITK are not seen at high frequencies inhuman disease (Supplementary Fig. S10). This limitation is notunique to our models, as it is also a feature of MB49. As RNAtranscription is downstream of genetic events such as mutationsand gene fusions and upstream of protein translation, we feel thetranscriptomic fidelity of our models to human bladder cancer isevidence that these models can be used to faithfully study

A B

C D

Tumor normalDNA/RNA-seq

Variantcalling

NetMHCPanepitope prediction

Filter byRNA expression

Screen for immnogenicityvaccine/ELISPOT

Spo

ts (n

egat

ive

cont

rol s

ubtra

cted

) 800

600

400

200

0

Predicted neoantigen peptide

Neoantigen-specific T-cell

mBLCA NeoantigenELISPOT screen MBA9

BBN963UPPL1541

G1 F4

H7

C5

C4

G9

+ Control Mediacontrol

BBN963 coculture ELISPOT

% P

ositi

ve c

ontr

ol

60

50

40

30

20

10

0

Figure 7.

Neoantigen prediction and validation in BBN963, UPPL1541, and MB49. A, Schema for neoantigen prediction workflow, using tumor DNA, tumor RNA, andmatched normal DNA to call mutations via UNSeqr, epitope prediction to identify predicted class I and II binders, and vaccine/ELISPOT validation. B, Summary ofvaccine/ELISPOT results in MB49, BBN963, and UPPL1541, with background subtracted counts ranked by number of spots and representative figures of highlyimmunogenicwells (G1, C4, G9), weakly immunogenic wells (F4, H7, C5), and controls.C, Summary of top eight predicted neoantigens inMB49 (red), BBN963 (blue),and top four predicted neoantigens in UPPL1541 (green), including sequence, predicted MHC class, and rank among all screened peptides within all threemodels. D, IFNg ELISPOT results of BBN963 neoantigen-enriched T cells cocultured with BBN963 tumors, as a percentage peptide-pulsed target positivecontrol. Blue dashed line marks IFNg intensity of irrelevant peptide (SIINFEKL) enriched T cells cocultured with BBN963.

Saito et al.




bladder cancer biology in general and subtype-specific biology inparticular. As both BBN and UPPL tumors exist as transplantablecell lines syngeneic with the C57BL/6 background and are able togrow both subcutaneously and orthotopically in the bladder,genetic manipulation strategies such as CRISPR/Cas9 may beused to manipulate these tumors should researchers desire tostudy effects of specific mutations in the context of subtype-specific tumors. A second limitation of the UPPL model, espe-cially for tumor immunology studies, is accounting for thepotential effect of PTEN inactivation on the immune microen-vironment. PTEN has been found to promote nuclear import ofthe transcription factor Interferon Regulatory Factor 3 (IRF3),thereby positively regulating type I IFN induction (35). PTENloss has also been associated with impaired T-cell tumor ingressand T-cell mediated cytotoxicity in a preclinical model of mel-anoma where human tumor cell lines engineered with PTENsilenced and to express the murine MHC class I molecule H-2Db

were injected into immunocompromised mice followed bytransfer of murine antigen-specific T cells and antigen-presentingcells 7 days later (36). Thus, it is possible that the UPPL modelmay be skewed toward an immune evaded or suppressed phe-notype due to PTEN loss.

In the ImVigor 210 study, Rosenberg and colleagues showedthat a subset of luminal tumorsweremore likely to respond toPD-1 axis inhibition using an anti–PD-L1 antibody (13). We reporthere that response to monotherapy using an mAb against Pd-1was effective in the BBNbut not theUPPLmodel. This represents apotential discrepancy between our murine models and theirhomonymous human subtypes; however, that conclusion istempered by two considerations: (i) Our group's luminal versusbasal predictor is different from that used by the ImVigor inves-tigators, and unfortunately despite publication the ImVigor RNA-seq data have yet to bemade public, and (ii) the subset of luminaltumors more likely to respond was also more heavily immuneinfiltrated and skewed toward effector T-cell expression, whichmay have marked them as basal by our classifier and/or representa subset of luminal tumors not modeled by UPPL.

The BBN model exhibited a mixed response to anti–PD-1therapy, which allowed us to evaluate differences in the tumorimmune microenvironment between responders and nonre-sponders. Responsive tumors showed higher degrees of CD8þ-to-CD4þ T-cell infiltration and memory-to-regulatory polari-zation of the tumor-infiltrating T cells. The former finding isconsistent with prior results in human bladder cancer (13),whereas the latter is to our knowledge the first report ofmemory T-cell polarization associating with response toimmune checkpoint inhibition in bladder cancer. This isconsistent with the hypothesis that tumor clearance is aug-mented by generation of T-cell memory. Although multipleT-cell immune gene signatures were strongly correlated withtumor mass in responders, we observed an overall lack of TCRclonotype sharing increase in tumor-infiltrating T cells inresponders compared with nonresponders. In addition, signif-icant changes to the B-cell clonotype sharing and diversity were

observed, suggesting B cells may play an important role in theresponse to checkpoint inhibitor therapy. Presumably, thispattern of TCR and BCR expression could be explained byseveral hypotheses: (i) that effector T-cell clones capable ofpromoting antitumor immunity are only present in respondersbut their frequencies are too low to result in discriminatingdifferences in global T-cell diversity changes, or (ii) thateffector T-cell clones capable of promoting antitumor immu-nity are present in both responders and nonresponders, but thepresence of specific B-cell clones is necessary to mediate theirfunction. Ongoing studies are evaluating pretreatment tumorfeatures and on-treatment features measurable from theperipheral blood that associate with eventual response totherapy, as well as to test novel combinations of immuno-therapy agents. Thus, this model provides a novel mixed-response platform to study efficacy and mechanisms of immu-notherapy in bladder cancer.

Disclosure of Potential Conflicts of InterestB.G. Vincent is a consultant/advisory boardmember for NanoString, Inc. No

potential conflicts of interest were disclosed by the other authors.

Authors' ContributionsConception and design: R. Saito, C.C. Smith, J. Kardos, J.S. Damrauer,B.G. Vincent, W.Y. KimDevelopment of methodology: R. Saito, C.C. Smith, T. Utsumi, L.M. Bixby,J. Kardos, J.S. Damrauer, B.G. Vincent, W.Y. KimAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): R. Saito, C.C. Smith, T. Utsumi, L.M. Bixby,K.G. Stewart, U. Manocha, K.M. Byrd, J.S. Damrauer, S.E. WilliamsAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): R. Saito, C.C. Smith, L.M. Bixby, J. Kardos,S.E. Wobker, S. Chai, J.S. Damrauer, B.G. Vincent, W.Y. KimWriting, review, and/or revision of the manuscript: R. Saito, C.C. Smith,L.M. Bixby, J. Kardos, S.E. Wobker, J.S. Damrauer, B.G. Vincent, W.Y. KimAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): C.C. Smith, J. KardosStudy supervision: S.E. Williams, B.G. Vincent

AcknowledgmentsWeacknowledge themembers of theKim, Vincent, and Serody labs for useful

discussions as well as Lineberger Bioinformatics Core, the UNCMouse Phase 1Unit (MP1U) for technical support. We thank Phil Abbosh for sharing MB49cells. This work was supported by the Bladder Cancer Advocacy Network(BCAN) Innovation Award (to W.Y. Kim) and Young Investigator Award(to R. Saito), University Cancer Research Fund (UCRF; to W.Y. Kim andB.G. Vincent), NIH K08 Award K08DE026537 (to K.M. Byrd), and UNCOncology Clinical Translational Research Training Program, 5K12CA120780(to B.G. Vincent).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received January 17, 2018; revised February 22, 2018; accepted May 3, 2018;published first May 21, 2018.

References1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CA Cancer J Clin

2017;67:7–30.2. Choi W, Porten S, Kim S, Willis D, Plimack ER, Hoffman-Censits J, et al.

Identification of distinct basal and luminal subtypes of muscle-invasive

bladder cancer with different sensitivities to frontline chemotherapy.Cancer Cell 2014;25:152–65.

3. Damrauer JS, Hoadley KA, Chism DD, Fan C, Tiganelli CJ, Wobker SE,et al. Intrinsic subtypes of high-grade bladder cancer reflect the





hallmarks of breast cancer biology. Proc Natl Acad Sci U S A 2014;111:3110–5.

4. Cancer Genome Atlas Research Network. Comprehensive molecular char-acterization of urothelial bladder carcinoma. Nature 2014;507:315–22.

5. Sj€odahl G, Lauss M, L€ovgren K, Chebil G, Gudjonsson S, Veerla S, et al. Amolecular taxonomy for urothelial carcinoma. Clin Cancer Res 2012;18:3377–86.

6. Volkmer J-P, Sahoo D, Chin RK, Ho PL, Tang C, Kurtova AV, et al. Threedifferentiation states risk-stratify bladder cancer into distinct subtypes. ProcNatl Acad Sci U S A 2012;109:2078–83.

7. Rebouissou S, Bernard-Pierrot I, de Reyni�es A, Lepage M-L, Krucker C,Chapeaublanc E, et al. EGFR as a potential therapeutic target for a subset ofmuscle-invasive bladder cancers presenting a basal-like phenotype. SciTransl Med 2014;6:244ra91–1.

8. Dyrskjøt L, Thykjaer T, Kruhøffer M, Jensen JL, Marcussen N, Hamilton-Dutoit S, et al. Identifying distinct classes of bladder carcinoma usingmicroarrays. Nat Genet 2003;33:90–6.

9. Hedegaard J, Lamy P, Nordentoft I, Algaba F, Høyer S, Ulhøi BP, et al.Comprehensive transcriptional analysis of early-stage urothelial carcino-ma. Cancer Cell 2016;30:27–42.

10. Robertson AG, Kim J, Al-Ahmadie H, Bellmunt J, Guo G, Cherniack AD,et al. Comprehensive molecular characterization of muscle-invasive blad-der cancer. Cell 2017;171:540–556.

11. Kardos J, Chai S, Mose LE, Selitsky SR, Krishnan B, Saito R, et al. Claudin-low bladder tumors are immune infiltrated and actively immune sup-pressed. JCI Insight 2016;1:e85902.

12. Powles T, Eder JP, Fine GD, Braiteh FS, Loriot Y, Cruz C, et al. MPDL3280A(anti-PD-L1) treatment leads to clinical activity in metastatic bladdercancer. Nature 2014;515:558–62.

13. Rosenberg JE, Hoffman-Censits J, Powles T, van der Heijden MS, Balar AV,Necchi A, et al. Atezolizumab in patients with locally advanced andmetastatic urothelial carcinoma who have progressed following treatmentwith platinum-based chemotherapy: a single-arm, multicentre, phase 2trial. Lancet 2016;387:1909–20.

14. Sharma P, Callahan MK, Bono P, Kim J, Spiliopoulou P, Calvo E, et al.Nivolumab monotherapy in recurrent metastatic urothelial carcinoma(CheckMate 032): a multicentre, open-label, two-stage, multi-arm, phase1/2 trial. Lancet Oncol 2016;17:1590–8.

15. Massard C, GordonMS, Sharma S, Rafii S,Wainberg ZA, Luke J, et al. Safetyand efficacy of durvalumab (MEDI4736), an anti-programmed cell deathligand-1 immune checkpoint inhibitor, in patients with advanced urothe-lial bladder cancer. J Clin Oncol 2016;34:3119–25.

16. Bellmunt J, de Wit R, Vaughn DJ, Fradet Y, Lee J-L, Fong L, et al. Pem-brolizumab as second-line therapy for advanced urothelial carcinoma. NEngl J Med 2017;376:1015–26.

17. Apolo AB, Infante JR, Balmanoukian A, Patel MR, Wang D, Kelly K, et al.Avelumab, an anti-programmed death-ligand 1 antibody, in patients withrefractory metastatic urothelial carcinoma: results from a multicenter,phase Ib study. J Clin Oncol 2017;35:2117–24.

18. Summerhayes IC, Franks LM. Effects of donor age on neoplastic transfor-mation of adult mouse bladder epithelium in vitro. J Natl Cancer Inst1979;62:1017–23.

19. Vasconcelos-N�obregaC, Colaco A, Lopes C,Oliveira PA. Review: BBN as anurothelial carcinogen. In Vivo 2012;26:727–39.

20. Kobayashi T, Owczarek TB, McKiernan JM, Abate-Shen C. Modellingbladder cancer in mice: opportunities and challenges. Nat Rev Cancer2014;15:42–54.

21. Zhang ZT, Pak J, Shapiro E, Sun TT, Wu XR. Urothelium-specificexpression of an oncogene in transgenic mice induced the formationof carcinoma in situ and invasive transitional cell carcinoma. Cancer Res1999;59:3512–7.

22. Grippo PJ, Sandgren EP. Highly invasive transitional cell carcinoma of thebladder in a simian virus 40 T-antigen transgenic mouse model. Am JPathol 2000;157:805–13.

23. Rampias T, Vgenopoulou P, Avgeris M, Polyzos A, Stravodimos K, Vala-vanis C, et al. A new tumor suppressor role for the Notch pathway inbladder cancer. Nat Med 2014;20:1199–205.

24. Puzio-Kuter AM, Castillo-MartinM, Kinkade CW,Wang X, Shen TH,MatosT, et al. Inactivation of p53 and Pten promotes invasive bladder cancer.Genes Dev 2009;23:675–80.

25. Trotman LC, Niki M, Dotan ZA, Koutcher JA, Di Cristofano A, Xiao A, et al.Pten dose dictates cancer progression in the prostate. Plos Biol 2003;1:e9.

26. Liu X, Ory V, Chapman S, Yuan H, Albanese C, Kallakury B, et al. ROCKinhibitor and feeder cells induce the conditional reprogramming of epi-thelial cells. Am J Pathol 2012;180:599–607.

27. Wang K, Singh D, Zeng Z, Coleman SJ, Huang Y, Savich GL, et al.MapSplice: accurate mapping of RNA-seq reads for splice junction discov-ery. Nucleic Acids Res 2010;38:e178.

28. Bindea G, Mlecnik B, Tosolini M, Kirilovsky A, Waldner M, Obenauf AC,et al. Spatiotemporal dynamics of intratumoral immune cells reveal theimmune landscape in human cancer. Immunity 2013;39:782–95.

29. Iglesia MD, Vincent BG, Parker JS, Hoadley KA, Carey LA, Perou CM,et al. Prognostic B-cell signatures using mRNA-seq in patients withsubtype-specific breast and ovarian cancer. Clin Cancer Res 2014;20:3818–29.

30. Suzuki R, ShimodairaH. Pvclust: an R package for assessing the uncertaintyin hierarchical clustering. Bioinformatics 2006;22:1540–2.

31. W€olflM, Greenberg PD. Antigen-specific activation and cytokine-facilitat-ed expansion of naive, human CD8þ T cells. Nat Protoc 2014;9:950–66.

32. Rock JR, Onaitis MW, Rawlins EL, Lu Y, Clark CP, Xue Y, et al. Basal cells asstem cells of the mouse trachea and human airway epithelium. Proc NatlAcad Sci U S A 2009;106:12771–5.

33. Iglesia MD, Parker JS, Hoadley KA, Serody JS, Perou CM, Vincent BG.Genomic analysis of immune cell infiltrates across 11 tumor types. J NatlCancer Inst 2016;108:djw144.

34. Fantini D, Glaser AP, Rimar KJ, Wang Y, Schipma M, Varghese N, et al. ACarcinogen-induced mouse model recapitulates the molecular alterationsof human muscle invasive bladder cancer. Oncogene 2018;66:7.

35. Li S, Zhu M, Pan R, Fang T, Cao Y-Y, Chen S, et al. The tumor suppressorPTEN has a critical role in antiviral innate immunity. Nat Immunol2016;17:241–9.

36. Peng W, Chen JQ, Liu C, Malu S, Creasy C, Tetzlaff MT, et al. Loss of PTENpromotes resistance to T cell-mediated immunotherapy. Cancer Discov2016;6:202–16.


Saito et al.



molecular subtype-speciﬁc immunocompetent models of …007914; kppt model). in order to induce cre...

Documents