CD8+CD103+ Tumor−Infiltrating Lymphocytes Are Tumor-Specific Tissue-Resident Memory T Cells and a Prognostic Factor for Survival in Lung This information is current as Cancer Patients of October 1, 2021. Fayçal Djenidi, Julien Adam, Aïcha Goubar, Aurélie Durgeau, Guillaume Meurice, Vincent de Montpréville, Pierre Validire, Benjamin Besse and Fathia Mami-Chouaib

J Immunol 2015; 194:3475-3486; Prepublished online 27 Downloaded from February 2015; doi: 10.4049/jimmunol.1402711 http://www.jimmunol.org/content/194/7/3475 http://www.jimmunol.org/ Supplementary http://www.jimmunol.org/content/suppl/2015/02/27/jimmunol.140271 Material 1.DCSupplemental References This article cites 50 articles, 23 of which you can access for free at: http://www.jimmunol.org/content/194/7/3475.full#ref-list-1

Why The JI? Submit online. by guest on October 1, 2021

• Rapid Reviews! 30 days* from submission to initial decision

• No Triage! Every submission reviewed by practicing scientists

• Fast Publication! 4 weeks from acceptance to publication

*average

Subscription Information about subscribing to The Journal of Immunology is online at: http://jimmunol.org/subscription Permissions Submit copyright permission requests at: http://www.aai.org/About/Publications/JI/copyright.html Email Alerts Receive free email-alerts when new articles cite this article. Sign up at: http://jimmunol.org/alerts

The Journal of Immunology is published twice each month by The American Association of Immunologists, Inc., 1451 Rockville Pike, Suite 650, Rockville, MD 20852 Copyright © 2015 by The American Association of Immunologists, Inc. All rights reserved. Print ISSN: 0022-1767 Online ISSN: 1550-6606. The Journal of Immunology

CD8+CD103+ Tumor–Infiltrating Lymphocytes Are Tumor-Specific Tissue-Resident Memory T Cells and a Prognostic Factor for Survival in Lung Cancer Patients

Fayc¸al Djenidi,*,†,‡ Julien Adam,†,‡,x Aı¨cha Goubar,†,‡,x,1 Aure´lie Durgeau,*,†,‡,1 Guillaume Meurice,†,‡,{ Vincent de Montpre´ville,*,‖ Pierre Validire,# Benjamin Besse,†,‡,** and Fathia Mami-Chouaib*,†,‡

We had previously demonstrated the role of CD103 integrin on lung tumor-infiltrating lymphocyte (TIL) clones in promoting spe- cific TCR-mediated epithelial tumor cell cytotoxicity. However, the contribution of CD103 on intratumoral distribution and functions and the prognosis significance of TIL subpopulations in non–small cell lung carcinoma (NSCLC) have thus far not been systematically addressed. In this study, we show that an enhanced CD103+ TIL subset correlates with improved early stage Downloaded from NSCLC patient survival and increased intraepithelial lymphocyte infiltration. Moreover, our results indicate that CD8+CD103+ TIL, freshly isolated from NSCLC specimens, display transcriptomic and phenotypic signatures characteristic of tissue-resident memory T cells and frequently express PD-1 and Tim-3 checkpoint receptors. This TIL subset also displays increased activation- induced cell death and mediates specific cytolytic activity toward autologous tumor cells upon blockade of the PD-1–PD-L1 interaction. These findings emphasize the role of CD8+CD103+ tissue-resident memory T cells in promoting intratumoral CTL responses and support the rationale for using anti–PD-1 blocking Ab to reverse tumor-induced T cell exhaustion in NSCLC http://www.jimmunol.org/ patients. The Journal of Immunology, 2015, 194: 3475–3486.

ytotoxic T lymphocytes play a major role in antiviral and essential in TCR-mediated target cell killing. Indeed, although antitumor immune responses through directional exocy- TCR engagement is necessary for inducing the cytotoxic activity + C tosis of cytotoxic granules, specialized secretory lyso- of specific CD8 T lymphocytes, integrins, mainly LFA-1 (aLb2), somes containing perforin and granzymes, into the specific target play a crucial role in triggering CTL effector functions. It is now leading to cell death (1). Integrins and their cognate ligands are well established that the interaction of LFA-1 with its ligand ICAM-1 is necessary for effective target cell lysis by the re- by guest on October 1, 2021 leased granules (2). CD103 (aEb7) integrin also plays a pivotal + *INSERM Unite´ 1186, 94805 Villejuif, France; †Institut de Cance´rologie Gustave role in epithelial target cell lysis by specific CD8 T cells. In- Roussy, 94805 Villejuif, France; ‡Universite´ Paris-Sud, 91400 Orsay, France; xINSERM { deed, the interaction of CD103 on tumor-specific tumor- Unite´ 981, Institut de Cance´rologie Gustave Roussy, 94805 Villejuif, France; Institut de infiltrating lymphocytes (TIL) with its ligand, E-cadherin on Cance´rologie Gustave Roussy, Plateforme de Bioinformatique, 94805 Villejuif, France; ‖Centre Chirurgical Marie-Lannelongue, Service d’Anatomie Pathologique, 92350 Le- tumor cells, is necessary for positioning the cytotoxic granules Plessis-Robinson, France; #Institut Mutualiste Montsouris, Service d’Anatomie Pathologique, near the interface and their delivery into the target, leading 75014 Paris, France; and **De´partement de Me´decine, Institut de Cance´rologie Gustave Roussy, 95805 Villejuif, France to lysis of the latter cells (3). CD103 is expressed at high + 1 levels by mucosal CD8 T lymphocytes, in particular intestinal A.G. and A.D. contributed equally to this work. intraepithelial lymphocytes (4), but it is also found on mucosal Received for publication October 27, 2014. Accepted for publication January 25, + + 2015. mast cells and dendritic cells (5), CD4 and CD8 regulatory + This work was supported by grants from INSERM, the Association pour la Recherche T cells (6), and on a large proportion of CD8 T cells infiltrating sur le Cancer, the Institut National du Cancer, the Ligue contre le Cancer, and the Site epithelial tumors, including pancreatic (7), colorectal (8), lung de Recherche Inte´gre´e sur le Cancer SOCRATE (Institut National du Cancer-Direction (3), and ovarian cancers (9). It has been recently reported that Ge´ne´rale de l’Offre de Soins Grant 6043). F.D. was supported by the Cance´ropoˆle Ile de France and the Universite´ Paris-Sud. J.A. was supported by the Institut The´matique CD103 expression on TIL is associated with increased survival Multi-Organismes Cancer and the Institut National du Cancer. in high-grade serous ovarian cancer (10). This integrin is induced + The microarray data presented in this article have been submitted to the European on CD8 T lymphocytes upon TCR engagement and exposure to Molecular Biology Laboratory European Bioinformatics Institute database (https:// TGF-b1, abundant within the tumor microenvironment, through www.ebi.ac.uk/arrayexpress) under accession no. E-MTAB-3266. binding of NFAT-1 and Smad2/3 transcription factors to the Address correspondence and reprint requests to Dr. Fathia Mami-Chouaib, INSERM Unite´ 1186, Integrative Tumor Immunology and Genetic Oncology, Institut de Can- promoter and enhancer elements of the ITGAE that encodes ce´rologie Gustave Roussy, 114 rue E´ douard Vaillant, 94805 Villejuif, France. E-mail CD103 (11). address: [email protected] CD103 ligation to E-cadherin not only promotes Ag recognition The online version of this article contains supplemental material. by favoring T cell adhesion to specific tumor cells (12), but it also Abbreviations used in this article: ADC, adenocarcinoma; AICD, activation-induced induces costimulation in activated CTL by triggering signals that cell death; CI, confidence interval; DFS, disease-free survival; HES, H&E with cooperate with TCR-mediated signaling (13). Emerging evidence saffron; HR, hazard ratio; MHC-I, MHC class I; NSCLC, non–small cell lung car- + + cinoma; OS, overall survival; PI, propidium iodide; qRT-PCR, quantitative RT-PCR; points to a role for CD8 CD103 T cells in transplant rejection SCC, squamous cell carcinoma; TIL, tumor-infiltrating lymphocyte; TRM, tissue- and in protecting prevalent infections posttransplantation (14–16). resident memory. This integrin serves as a marker of tissue-resident memory (TRM) Copyright Ó 2015 by The American Association of Immunologists, Inc. 0022-1767/15/$25.00 T cells, and accumulating evidence indicates that CD103 is www.jimmunol.org/cgi/doi/10.4049/jimmunol.1402711 3476 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE

directly involved in intraepithelial retention of TRM T cells (17). Apoptosis assay + Remarkably, CD103 TRM T cells play an essential role in pro- Activation-induced cell death (AICD) was measured by flow cytometry tecting human epithelial tissues against viral infections, and this using an annexin V apoptosis detection (BD Pharmingen). After cell subset has been shown to express inhibitory receptors such recovery of TIL and tumor cells by FACS based on cell size selection, as CTLA-4 and PD-1, associated with their capacity to maintain they were either kept in medium or stimulated overnight with rIL-2 (5– peripheral tolerance (18, 19). However, the presence of T 10 U/ml). TIL were then cultured for 4 h in the presence of autologous RM tumor cells. After 15 min incubation with 5 ml annexin V, labeling T cells in human epithelial tumors and their role in the antitu- buffer, and 5 ml propidium iodide (PI), cells were analyzed within 1 h mor immune response have thus far not been systematically by FACS. addressed. In this study, we sought to determine the prognostic value of CD103+ TIL in early stage non–small cell lung car- Cytotoxicity assay cinoma (NSCLC) and perform combined phenotypic and The cytotoxic activity of TIL was measured by a conventional 51Cr- functional analyses of freshly isolated CD8+CD103+ T cells release assay (3). Freshly isolated TIL were either kept in medium or infiltrating human lung tumor lesions. Our results indicate stimulated overnight with rIL-2 (5 U/ml). Autologous tumor cells that a strong tumor infiltration with CD103+ TIL correlates with were used as targets at a 10:1 E:T ratio. For cytotoxicity induction, TIL or tumor cells were preincubated for 1 h at 37˚C with neutral- NSCLC patient survival and increased intraepithelial lymphocyte izing anti–PD-1 or anti–PD-L1 mAb, respectively. Cytotoxicity in- infiltration, suggesting that CD103 promotes recruitment of TIL hibition was performed by preincubating effector or target cells for within epithelial tumor islets. Moreover, CD8+CD103+ TIL ex- 1 h with anti-CD103 or anti–MHC class I (MHC-I) blocking mAb, respectively. hibit TRM T cell features, express PD-1 and Tim-3, but not CTLA-4, and exert specific cytotoxic activity against autologous NSCLC patient cohort Downloaded from tumor cells upon neutralization of PD-1–PD-L1 interactions. We propose that this integrin can be used as a unique biomarker Paraffin-embedded primary tumor samples were obtained from patients of tumor-specific CD8+ T cells infiltrating human epithelial diagnosed with early stage NSCLC and who underwent curatively intended surgical resection between 1995 and 2002. A total of 101 tumors. tumor samples were included in this study and analyzed for CD3, CD8, and CD103 expression on TIL, located either in epithelial or stromal Materials and Methods regions, and their association with patient survival outcomes. All were in pathological stage I (33.7% were pT1N0 and 66.3% pT2N0) and none http://www.jimmunol.org/ TIL and patient PBMC of them had received neoadjuvant therapy at the time of surgery. Of Fresh NSCLC tumors were obtained from the Institut Mutualiste Mont- these, 44.5% were adenocarcinomas (ADC), 42.6% were squamous cell souris or the Centre Chirurgical Marie-Lannelongue and immediately carcinomas (SCC), 12.9% were mixed or large cell carcinoma, and 3% dissociated mechanically and enzymatically (dissociation kit from Miltenyi were other types. Median (range) age at diagnosis was 66.1 y (40.5– Biotec). Among a total number of 210 recruited tumors, only 60% were 83.6 y old); 68.3% of the patients were males; 89.1% were tobacco infiltrated with a sufficient number of TIL to perform ex vivo studies. TIL users; and median (range) tobacco consumption was 50 (5–100) pack were either directly analyzed by multicolor flow cytometry or isolated using years. a FACSAria cell sorter based on a small-size selection of mononuclear cells. PBMC from distinct consented NSCLC patients were obtained from Institut Immunohistochemical staining de Cance´rologie Gustave Roussy. All experiments were approved by the Serial sections from paraffin-embedded tumors were stained with H&E with by guest on October 1, 2021 Institutional Review Board. saffron (HES), anti-CD3, anti-CD8 (Thermo Fisher), or anti-CD103 (Abcam) mAb. Briefly, 4-mm-thick sections were mounted on poly-L-ly- Microarray analyses sine–coated slides, deparaffinized, and rehydrated through graded alcohol Total RNA from small size–selected TIL and PBMC were extracted using to water. Primary Ab were incubated for 1 h at room temperature. Im- TRIzol (Sigma-Aldrich) and then compared using an Agilent Technologies munostaining was visualized following incubation with goat anti-rabbit SurePrint G3 human GE 8 3 60K microarray (AMADID 39494) as de- HRP for 30 min at room temperature and then 3,39-diaminobenzidine scribed (3). Feature extraction software (Agilent Technologies, version substrate was added (PowerVision, Leica Biosystems). The slides were 10.7.3.1) was used to quantify the intensity of fluorescent images. Data counterstained with Mayer’s hematoxylin (VWR International). For all were normalized by a quantile method from the limma Bioconductor staining, whole slides were digitized using a slide scanner (VS120, 2 R package. Differential gene expression analysis was performed with Olympus). For each slide, three areas of 0.88 mm were selected as rep- limma’s moderated t test. Primary functional analyses were performed resentative of the potential heterogeneity of immune infiltration within the + + + using Ingenuity Pathway Analysis and the Database for Annotation, tumor. Quantification of the number of CD3 , CD8 , and CD103 cells Visualization andIntegrated Discovery bioinformatics. The microarray data was performed manually in each of the areas on serial sections for both are available at the European Molecular Biology Laboratory European intraepithelial and stromal compartments. For each slide, data were sum- Bioinformatics Institute database (https://www.ebi.ac.uk/arrayexpress) under marized over the three areas giving a score number of stained cells per accession no. E-MTAB-3266. square millimeter.

Quantitative RT-PCR analysis Statistical analysis cDNA were synthesized from 1 mg mRNA from small size–selected TIL, Distributions of CD3, CD8, and CD103 biomarkers according to each PBMC, or CD103+ TIL using an Applied Biosystems kit. Quantitative RT- clinical factor (age, histological type, pathological stage, smoking status, PCR (qRT-PCR) analysis was performed using a LightCycler and the and gender) were described separately by providing medians and ranges, LightCycler FastStart DNA Master SYBR Green I mix (Roche Applied and comparison of distribution between groups was performed using the Science) according to the manufacturer’s instructions. Expression of target Wilcoxon rank or Kruskal–Wallis test when appropriate. Disease-free was normalized to that of 18S. survival (DFS) was defined by the time from diagnosis to death or re- lapse, whichever occurred first, or to last follow-up date. Overall survival Ab and immunofluorescence analyses (OS) was defined as the time from diagnosis to death or date of last follow-up. Considering CD3, CD8, and CD103 as continuous variables, Anti-human CD3, CD8, CD4, CD103, CD107a, CD127, Bcl-2, KLRG1, their association with DFS and OS was evaluated using univariate and and granzyme B mAb as well as mouse and rabbit isotopic controls were multivariate Cox regression models adjusted for age, histological type, purchased from Ozyme. Anti-human CD69, CD62L, CD45RA, CD45RO, gender, stage, and smoking status. The required assumptions of CD27, and CD28 mAb were provided by Invitrogen. Anti-human PD-1, proportionality in multivariate survival analysis were checked by Tim-3, and CTLA-4 mAb were purchased from eBioscience. Phenotypic Schoenfeld’s test. analyses were performed by direct immunofluorescence using a FACS LSR For all other experiments, data were compared using the two-tailed II flow cytometer. Data were processed using FlowJo or FACSDiva software Student t test. Two groups were considered as significantly different (BD Biosciences). when p , 0.05. The Journal of Immunology 3477

Results the transcriptional profiles of freshly isolated NSCLC TIL, which CD103+ TIL were associated with increased survival of include 85% (mean, 82 6 5%) of CD3+, 51% (mean, 51 6 7%) of primary stage I NSCLC patients CD4+, and 37% (mean, 37 6 11%) of CD8+ T lymphocytes, re- spectively (Supplemental Fig. 1A). For this purpose, we cell To exert their lytic function, CTL must infiltrate the tumor tissue sorted TIL from 13 independent NSCLC specimens, including 5 and then interact with the target cell to finally trigger their func- SCC, 1 large cell carcinoma, and 7 ADC, and compared their tional activities. Thus, we first investigated the frequency and transcriptional profiles to PBMC from 19 independent NSCLC distribution of CD3+, CD8+, and CD103+ TIL in epithelial tumor patients (Supplemental Fig. 1A) by microarray analysis. Differential and stromal regions from tumor samples of 101 stage I NSCLC 2 gene expression analyses performed with a p value of #10 10 and patients. Serial tumor sections were stained with HES, anti-CD3, afoldchangeof$2 identified 487 genes, 233 of which were less anti-CD8, or anti-CD103 mAb, and intratumoral and stromal cell strongly expressed in TIL than in PBMC, and 254 genes that were localization was then quantified (Fig. 1A). Results indicated that more strongly expressed in TIL than PBMC (Supplemental Fig. the density of CD3+, CD8+, and CD103+ lymphocytes varied from 1B). Further filtering, acquired with an overall TIL signal inten- one tumor to another with a median range of 234 (5–1184), 135 sity $1000, identified an expression profile of 103 genes (Sup- (8–685), and 113 (2–697) infiltrations, respectively. Moreover, plemental Fig. 2A), including a cluster of 33 genes related to results showed that the large majority of CD3+,CD8+,and adhesion, exhaustion, activation, immunosuppression, and apopto- CD103+ TIL were located within the stroma (Fig. 1B, 1C). Indeed sis, which were more strongly expressed in TIL than in PBMC 89, 83, and 66% of patients had more than half of anti-CD3–, anti- (Supplemental Table II). The gene expression profile also included CD8–, and anti-CD103–stained cells in the stroma, respectively a cluster of 70 genes that, on the contrary, were less strongly (p , 0.002, see Table I). An increased intraepithelial lymphocyte Downloaded from expressed in TIL than in PBMC (Supplemental Table II). For ad- infiltration was particularly observed in tumors with a high density hesion and exhaustion, RGS1, RALGDS and ITGAE, and PDCD1, of CD103+ TIL (Fig. 1C), which was less frequently detected in LAG-3, CTLA-4, TNFRSF-18,and HAVCR2 were the most strongly older patients (p = 0.02) and in women compared with men overexpressed genes in TIL, respectively (Supplemental Table II). (p = 0.01; Table I). Moreover, intraepithelial CD103+ infiltration These results were further confirmed by qRT-PCR analyses was more present in ever-smokers who are generally male + (Supplemental Fig. 2B), which also showed a correlation with gene

(p = 0.01; Table I). The latter result was also observed for CD8 http://www.jimmunol.org/ chip expression profiles (Supplemental Fig. 2C) and did not reveal TIL (p = 0.04), which might also express CD103. In contrast, any differences between SCC and ADC histological types. These CD3+ TIL were not associated with patient clinical information. results suggest that a subset of TIL expresses higher levels of T We then tested whether a high density of CD103+ TIL correlated RM T cell marker genes than do PBMC. with a favorable clinical outcome. The median follow-up was 6.3 y. Next, we purified the CD103+ TIL subpopulation and analyzed Results revealed that CD103+ TIL significantly correlated with the expression of the selected 33 genes by qRT-PCR. Initial phe- favorable clinical outcome as a continuous variable in DFS and notypic analyses indicated that .80% of CD103+ cells were OS in univariate analysis (Table II). For each 50 CD103+ incre- CD3+ T lymphocytes, whereas ,20% were CD11c+CD83+ dendritic ment, the risk of relapse or death was reduced by 16% (unadjusted cells (Fig. 2A). Moreover, most CD3+CD103+ T lymphocytes were hazard ratio [HR] = 0.84, 95% confidence interval [CI], p = 0.01) by guest on October 1, 2021 CD8+, with 8–53% (mean, 27 6 13; n =21)ofCD8+ T cells and and a 12% reduction in risk of death (unadjusted HR = 0.88, 95% only 1–10% (mean, 7 6 6; n = 21) of CD4+ T cells (Fig. 2B). qRT- CI, p = 0.04). Notably, age, stage, and histology at diagnosis were PCR indicated that the CD3+CD8+CD103+ TIL subpopulation, iso- significantly associated with patient outcome, whereas for gender, lated from 10 NSCLC, expressed more strongly the T T cell a trend toward a significant association was observed. Multivariate RM markers PDCD1, CTLA4,andHAVCR2 genes, encoding PD-1, analysis was then carried out, including these four factors together CTLA-4, and Tim-3, respectively, as well as RGS1, HSPA1A, with CD103+ TIL. As summarized in Table II, CD103+ lympho- ICOS, NR4A2, EGR1,andBAG-3, also described as markers of T cytes maintained a strong prognostic value in the multivariate RM T cells (18, 19), than total PBMC (Fig. 2C). In contrast, SIP1 was model for DFS, but not for OS (HR = 0.85, 95% CI, p = 0.02 and less strongly expressed in CD3+CD8+CD103+ TIL than in total HR = 0.89, 95% CI, p = 0.09 for DFS and OS, respectively). PBMC. Moreover, all of these T T cell markers were Notably, none of the other parameters that were independent RM morestronglyexpressedinCD3+CD8+CD103+ TIL than in predictors of outcome remained significantly associated with DFS 2 CD3+CD8+CD103 PBMC (Fig. 2D). These results indicate (Table III). Similar results were obtained when we analyzed the that lung TIL often include a large subset of CD8+CD103+ Tcells association of CD103+ TIL in tumor regions with patient OS and suggest that, along with CD103 expression, this subset exhibits (Table II, Supplemental Table I). With regard to CD8+ TIL, no T Tcellfeatures. correlation with OS was found. However, an association with DFS RM + + + was observed in univariate analyses only for either CD8 total or CD8 CD103 T cells displayed a TRM phenotype, expressed for CD8+ lymphocytes within tumor islets (Table II, Supplemental PD-1 and Tim-3, and corresponded to a highly activated Table I). CD8+ TIL within epithelial tumor regions remained subpopulation significantly associated with DFS in multivariate analysis (Table II). To further characterize CD3+CD8+CD103+ TIL, we performed In contrast, the CD3 marker was never associated with patient multicolor immunofluorescent staining. Flow cytometry analy- outcome. These results suggested that CD103 promoted T cell ses of TIL from five tumors (two SCC and three ADC) showed infiltration within epithelial tumor islets and indicated that a high + 2 + a homogeneous T cell subpopulation with a CD69 CD62L density of CD103 TIL was associated with stage I NSCLC patient CD282CD27+CD45RA+CD45RO+CCR72 phenotype character- survival. istic of memory CD8+ T cells. Moreover, TIL from SCC and ADC tumors displayed similar T cell phenotypes. Results also indicated Phenotypic and transcriptional profiles of NSCLC TIL that these lymphocytes expressed receptors CCR5 and Little has been known about the phenotypic and functional features CCR6 (Fig. 3A), and the prosurvival family member Bcl-2, but of human epithelial TIL and their contribution in controlling tumor not KLRG1, and only low levels of CD127 (Supplemental Fig. progression. Experiments were therefore conducted to determine 1C). Remarkably, whereas CD3+CD8+CD103+ TIL expressed 3478 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 1. Distribution of CD3+, CD8+, and CD103+ cells in intraepithelial tumoral and stromal regions of NSCLC tumors. (A) Tumor sections from 101 early stage NSCLC patients were stained with anti-CD3, anti-CD8, or anti-CD103 mAb. Representative images of CD3, CD8, and CD103 immu- nostaining of the same tumor area and counterstained with HES (original magnification 3200) are shown. (B) CD3, CD8, and CD103 expression scatter plots in intraepithelial tumor and stromal regions among the 101 NSCLC tumors. Right panels, Bar plots of the total number (blue) and the percentage (yellow) of CD3+, CD8+, and CD103+ cells present in the tumor. (C) Mean percentages of CD3+, CD8+, and CD103+ lymphocytes within tumoral and stromal regions of the 101 NSCLC tumors. **p , 0.001. The Journal of Immunology 3479

Table I. CD3+, CD8+, and CD103+ TIL association with clinical characteristics in stage I NSCLC patients

Total Tumoral Stromal

Number (%) Median (Q1–Q3) p Value Median (Q1–Q3) p Value Median (Q1–Q3) p Value CD3+ Patient 93 234.7 (147.7–442.7) 35.7 (12.3–96.7) 185.7 (127.0–355.3) Histological subtype ADC 42 (45.16) 209.8 (151.9–402.0) 26.7 (7.9–81.5) 178.7 (132.3–353.9) SCC 40 (43.01) 198.3 (117.6–437.2) 40.8 (16.8–98.8) 158.7 (78.5–325.9) Other 11 (11.83) 474.0 (331.2–598.2) 0.09 85.7 (37.5–145.0) 0.07 303.7 (247.7–475.7) 0.07 Stage Ia 30 (32.26) 194.3 (159–391.5) 25.5 (6.2–79.5) 172.6 (129.5–320.8) Ib 63 (67.74) 277.7 (143.8–477.5) 0.45 49 (17.5–108.8) 0.06 199.7 (118.7–377.7) 0.69 Gender Male 30 (32.26) 190.5 (137.2–398.4) 32.2 (8.3–84.8) 169.8 (96.3–381.2) Female 63 (67.74) 266.3 (151.8–458.3) 0.48 39.7 (16.8–110.2) 0.27 202.7 (133.7–346.8) 0.67 Smoking status Ever 83 (89.25) 257.0 (145.8–442.2) 41 (18.5–103.5) 185.7 (118.7–352.5) Never 10 (10.75) 202.8 (154.9–502.0) 0.92 15.2 (1.3–68.2) 0.06 178.1 (136.7–429.6) 0.53 Age (y) ,65 63 (62.4) 277.7 (168.3–442.7) 47.3 (20–106) 209 (143.3–355.3)

$65 38 (37.6) 187.3 (117.6–439.7) 0.07 28 (8.4–87.2) 0.09 158.7 (68.0–327.2) 0.16 Downloaded from CD8+ Patient 98 135.0 (86.8–249.2) 24.3 (7.4–80.1) 106.8 (53.6–169.9) Histological subtype ADC 45 (45.92) 132.0 (83.0–225.7) 12.3 (2–48.7) 111.7 (65.7–170.7) SCC 40 (40.82) 115.3 (87.6–247.2) 27.6 (12.5–71.1) 82.7 (46.3–140.9) Other 13 (13.27) 331.3 (124.0–411.7) 0.03 85 (27.5–148.7) 0.02 173.7 (116.7–263.0) 0.01 Stage http://www.jimmunol.org/ Ia 32 (32.65) 124.0 (54.67–248.7) 16 (4.3–56.8) 111.5 (53.3–150.9) Ib 66 (67.35) 138.7 (91.08–249.3) 0.53 27.6 (8.6–80.1) 0.24 103.5 (60.9–174.4) 0.76 Gender 31 (31.63) 124.0 (84.17–209.3) 22.7 (3.2–50.5) 103.7 (57.4–170.2) Male 67 (68.37) 139.3 (91.5–251.5) 0.50 27.5 (7.83–90.5) 0.18 107.0 (52.0–168.7) 0.87 Female Smoking status 87 (88.78) 139.3 (85.83–255.3) 27.5 (7.8–87.7) 106.7 (50.7–170.3) Ever 11 (11.22) 124.0 (107–138.8) 0.50 8.7 (0–26.5) 0.04 108.3 (80.8 –123.5) 0.91 Never Age (y) ,65 63 (62.4) 144.7 (88.3–311.4) 27.8 (7.1–113.9) 111.5 (53.2–174.2)

$65 38 (37.6) 128.0 (82.3–194.6) 0.17 18.7 (7.4–52.7) 0.11 105.2 (56.5–167.9) 0.48 by guest on October 1, 2021 CD103+ Patient 98 112.5 (65.7–230.6) 36.8 (15.4–103.9) 70.1 (37.1–135.1) Histological subtype ADC 45 (45.92) 111.2 (63.4–204.3) 23.9 (1.0–72.0) 77.1 (48.1–143.93) SCC 40 (40.82) 99.4 (67.0–192.5) 46.0 (24.3–98.0) 46.3 (29.3–85.5) Other 13 (13.27) 264.3 (87.3–374.8) 0.13 90 (44–166.1) 0.01 155.2 (43.3–181.8) 0.02 Stage Ia 32 (32.65) 107.6 (61.0–203.8) 29.5 (8.5–95.7) 61.8 (36.7–104.7) Ib 66 (67.35) 127.6 (69.5–250.8) 0.24 40.7 (17.0–103.9) 0.35 76.4 (37.9–147.0) 0.36 Gender Female 31 (31.63) 104.1 (63.3–157.9) 23.9 (7.2–51.9) 71.3 (39.9–125.2) Male 67 (68.37) 128.2 (67.0–274.2) 0.25 55.2 (22.7–146.7) 0.01 69.6 (36.5–142.2) 0.86 Smoking status Ever 87 (88.78) 126.9 (67.0–254.8) 41.5 (21.0–121.3) 71.3 (37.4–140.2) Never 11 (11.22) 87.3 (62.4–118.3) 0.12 15.7 (0–35.7) 0.01 48.1 (32.9–90.6) 0.34 Age (y) ,65 63 (62.4) 136.1 (76.1–273.0) 51.9 (20.5–160.5) 81.4 (47.5–144.9) $65 32 (37.6) 97.4 (52.4–145.7) 0.02 26.6 (9.9–65.5) 0.02 64.1 (28.9–119.6) 0.02 The p value was determined using the Wilcoxon rank or Kruskal–Wallis test when appropriate. Bold type indicates significant (p ,0.05). Q1 and Q3, the first and the third quartile, respectively; stage Ia, pT1N0; stage Ib, pT2N0. high levels of PD-1 and Tim-3, only a small subset of CD3+CD8+ from two additional patients (Fig. 3C). These results further support the 2 + + CD103 TIL expressed these inhibitory receptors, and CTLA-4 hypothesis that CD8 CD103 TILbelongtoaTRM subset and that was slightly expressed by both subpopulations (Fig. 3B). they are rapidly activated following ex vivo stimulation with autolo- Next, we compared the activation status of CD8+CD103+ TIL gous tumor cells. + 2 to CD8 CD103 TIL. For this purpose, freshly size-selected TIL + + were stimulated with cell-sorted autologous tumor cells, and CD3+ CD8 CD103 TIL killed autologous tumor cells upon PD-1 CD8+ lymphocytes were gated and analyzed for CD103, annexin ligation V, and PI staining. As shown in Fig. 3C, annexin V+PI2 apoptotic The above results suggested that the CD8+CD103+ TIL subpop- cells comprised 19, 17, and 33% of CD103+ T cells from patients ulation was enriched with tumor-reactive T lymphocytes. To test 97, 98, and 99, whereas they comprised only 4, 1, and 6% of this hypothesis, we examined whether CD8+CD103+ T cells CD1032 T cells, respectively. Higher susceptibility of CD103+ T cells specifically recognized autologous tumor cells. To this end, to TCR activation-dependent apoptosis was further confirmed in TIL freshly isolated TIL from 11 of 18 tumors were isolated and in- 3480 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE

Table II. Association of CD3, CD8, and CD103 with prognosis in univariate and multivariate analyses

Univariate Multivariate

HR (95% CI) for 50 HR (95% CI) for 50 No. of Patients No. of Events Increment Increase p Value Increment Increase p Value OS Total CD3 93 44 0.97 (0.91–1.04) 0.4 CD8 98 45 0.9 (0.80–1.02) 0.1 CD103 98 45 0.88 (0.77–0.99) 0.04 (0.78–1.02) 0.09 Epithelial tumor islets CD3 93 44 0.94 (0.76–1.15) 0.5 CD8 98 45 0.83 (0.63–1.09) 0.2 CD103 98 45 0.84 (0.67–1.06) 0.09 (0.66–1.05) 0.13 Stromal region CD3 93 44 0.97 (0.88–1.06) 0.5 CD8 98 45 0.88 (0.74–1.05) 0.2 CD103 98 45 0.78 (0.62–0.98) 0.03 (0.67–1.06) 0.14 DFS Total CD3 92 48 0.96 (0.89–1.02) 0.2

CD8 97 50 0.88 (0.78–0.99) 0.04 (0.78–1.02) 0.09 Downloaded from CD103 97 50 0.84 (0.74–0.96) 0.01 (0.74–0.97) 0.02 Epithelial tumor islets CD3 92 48 0.82 (0.65–1.03) 0.09 CD8 97 50 0.72 (0.54–0.96) 0.02 (0.51–0.97) 0.03 CD103 97 50 0.78 (0.63–0.96) 0.02 (0.61–0.96) 0.02 Stromal region CD3 92 48 0.96 (0.89–1.05) 0.4 CD8 97 50 0.88 (0.74–1.04) 0.1 http://www.jimmunol.org/ CD103 97 50 0.76 (0.61–0.94) 0.01 (0.64–1.01) 0.06 HR (95% CI) was calculated for an increment of 50 for CD3, CD8, and CD103. The p value was determined using the Wilcoxon rank or Kruskal–Wallis test when appropriate. Bold type indicates significant (p ,0.05). cluded in functional assays. Among these 11 TIL populations, 5 Discussion were tested for their capacity to degranulate after stimulation In this study, we show that NSCLC TIL include a homogeneous + with cell-sorted target cells. Results indicated that CD8 CD8+ T cell population that displays a phenotypic signature as- + by guest on October 1, 2021 CD103 TIL expressed much higher levels of the CD107a sociated with TRM cells, mainly expression of CD103 and CD69 + 2 molecule than did CD8 CD103 TIL following ex vivo stim- and the absence of CCR7. TRM T cells are highly activated ulation with autologous tumor cells in the presence of low T lymphocytes that reside within a variety of peripheral tissues, doses of rIL-2 (Fig. 4A). We then examined granzyme B ex- including intestine (17, 20), brain (21), skin (22, 23), and lung + + + 2 pression in CD8 CD103 and CD8 CD103 lymphocytes (24), and provide early responses to viral reinfection (25). Accu- unstimulated or stimulated with low concentrations of rIL-2. mulating evidence indicates that TGF-b1 plays an essential role in Intracytoplasmic immunofluorescence analyses indicated that the formation and maintenance of TRM T cells, at least in part via + nonstimulated CD8 T cell subsets did not express granzyme B, induction of CD103 and CD69. Although the natural ligand for but stimulation of size-selected TIL with rIL-2 induced a much CD69 has not been clearly defined, evidence for an interaction of + higher increase in serine protease expression in CD103 than in CD69 with S1P1, a receptor for sphingosine 1-phosphate that 2 CD103 T cells (Fig. 4B). mediates the egress of T cells from lymphoid organs (26), has + + + To assess whether CD8 CD103 granzyme B T cells were able been suggested to also play a role in retention of TRM T cells by to kill autologous tumor cells, which include 93 6 3% MHC-I+, modulating their capacity to exit nonlymphoid tissues (27). In- 6 + 6 + 6 + + + 85 5% Epcam ,70 21% E-cadherin , and 61 18% PD-L1 deed, downregulation of SIP1 was observed in CD103 CD8 TRM cells (Supplemental Fig. 1D), we performed a chromium release T cells (18) and appeared to be required for their retention in assay in the absence or presence of anti–PD-1 mAb combined or peripheral tissues (28). Our results revealed that CD103+ TIL not with anti–PD-L1. Results included in Fig. 4C show that TIL, display a transcriptomic signature characteristic of TRM cells, with previously stimulated with rIL-2, displayed a weak cytotoxic ac- downregulation of S1P1 and upregulation of RGS1 and ICOS as tivity toward autologous size-selected tumor cells. However, well as genes encoding PD-1 and Tim-3, along with Bag-3 and blocking of PD-1 or PD-L1 with neutralizing mAb induced a transcription factors EGR1 and Nr4a2 (18, 19). Microarray strong increase in target cell lysis. Moreover, anti–MHC-I (Fig. studies also showed downregulation of the ITGB2 gene in TIL 4C) and anti-CD103 (Fig. 4D) blocking mAb strongly inhibited (Supplemental Table II), suggesting that dampening of the aLb2 tumor cell killing by autologous TIL pretreated with anti–PD-1 integrin, most likely through TGF-b1 secretion, also participates and/or in the presence of anti–PD-L1 mAb, respectively. These in T cell residency within the tumor. Additionally, CD8+CD103+ results further emphasized that CD8+CD103+ TIL were exhausted TIL are able to mediate tumor-specific cytotoxic activity toward tumor-specific T lymphocytes and that a subset could be rescued autologous malignant cells after blockade of PD-1. These results by blocking PD-1 signals leading to T cell activation and autol- further support the notion that CD103 is induced on tumor-specific ogous tumor cell lysis. They also suggest that this T cell sub- T cells upon entry into a TGF-b1–rich tumor microenvironment population mediates in situ antitumor CTL response that leads to and interaction of TCR with the specific tumor peptide/MHC-I improved patient survival. complex (29). Persistent tumor Ag stimulation, together with h ora fImmunology of Journal The

Table III. Complete results of the univariate and the multivariate analyses for the DFS outcome

Univariate Multivariatea Multivariateb Multivariatec Multivariated Multivariatee

p p p p p p HR 95% CI Value HR 95% CI Value HR 95% CI Value HR 95% CI Value HR 95% CI Value HR 95% CI Value For an increase of 50 U CD3 total 0.96 (0.89–1.02) 0.2 CD3 tumor 0.82 (0.65–1.03) 0.09 CD3 stroma 0.96 (0.89–1.05) 0.4 CD8 total 0.88 (0.78–0.99) 0.04 0.89 (0.78–1.02) 0.09 CD8 tumor 0.72 (0.54–0.96) 0.02 0.70 (0.51–0.97) 0.03 CD8 stroma 0.88 (0.74–1.04) 0.1 CD103 total 0.84 (0.74–0.96) 0.01 0.85 (0.74–0.97) 0.02 CD103 tumor 0.78 (0.63–0.96) 0.02 0.76 (0.61–0.96) 0.02 CD103 Stroma 0.76 (0.61–0.94) 0.01 0.80 (0.64–1.01) 0.06 Age (y) ,65 1.00 1.00 1.00 1.00 1.00 1.00 $65 2.48 (1.38–4.44) 0.002 1.88 (1.01–3.5) 0.05 1.74 (0.93–3.24) 0.08 1.59 (0.84–3) 0.15 1.57 (0.83–2.98) 0.16 1.81 (0.96–3.38) 0.07 Gender Female 1.00 1.00 1.00 1.00 1.00 1.00 Male 1.34 (0.72–2.51) 0.4 1.59 (0.78–3.2) 0.20 1.55 (0.76–3.13) 0.23 1.63 (0.81–3.25) 0.17 1.71 (0.85–3.44) 0.13 1.56 (0.78–3.13) 0.21 Smoking status Ever 1.00 Never 1.09 (0.46–2.56) 0.8 Stage Ia 1.00 1.00 1.00 1.00 1.00 1.00 Ib 1.54 (0.81–2.92) 0.2 1.54 (0.75–3.16) 0.24 1.62 (0.79–3.32) 0.19 1.69 (0.82–3.47) 0.15 1.64 (0.8–3.35) 0.18 1.58 (0.77–3.24) 0.21 Histology ADC 1.00 1.00 1.00 1.00 1.00 1.00 SCC 1.89 (1.04–3.45) 0.04 1.38 (0.69–2.76) 0.36 1.53 (0.76–3.08) 0.24 1.40 (0.7–2.78) 0.34 1.55 (0.77–3.1) 0.22 1.29 (0.65–2.59) 0.47 Others 0.82 (0.28–2.42) 0.7 0.89 (0.31–2.58) 0.83 1.05 (0.36–3.08) 0.93 1.02 (0.35–2.95) 0.97 1.10 (0.38–3.24) 0.86 0.85 (0.3–2.42) 0.76 HR (95% CI) was calculated for an increment of 50 for CD3, CD8, and CD103. The p value was determined using the Wald test when appropriate. Bold type indicates significant (p ,0.05). a–eFive multivariate Cox regression models adjusted on the clinical factors (age, gender, smoking status, stage, and histology), and including CD8 totala, CD8 tumorb, CD103 totalc, CD103 tumord, and CD103 stromae. Stage Ia, pT1N0; stage Ib, pT2N0.

3481

Downloaded from from Downloaded http://www.jimmunol.org/ by guest on October 1, 2021 1, October on guest by 3482 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 2. Expression of CD103 integrin on TIL from NSCLC specimens. Freshly resected NSCLC tumors were dissociated and then directly analyzed by flow cytometry for CD103, CD3, CD8, CD4, CD83, and CD11c expression. (A) Surface expression of CD3 (black fill), CD11c, and CD83 on CD103+ TIL. An isotypic control mAb was included (gray fill). Three independent experiments are shown. Lower panel, Mean percentages of CD3+ T cells and CD11c+CD83+ dendritic cells among CD103+ TIL (n = 12). (B) Surface expression of CD103 on CD8+ and CD4+ TIL from three independent NSCLC. Percentages of positive cells are indicated. Numbers in parentheses correspond to mean fluorescence intensity. Right panel, Mean percentages of CD8+ CD103+ and CD4+CD103+ cells among CD3+ T cells (n = 21). (C) Genes differentially expressed in CD3+CD8+CD103+ TIL, cell sorted from 10 NSCLC with anti-CD103 mAb, and patient total PBMC. Genes described as markers of TRM T cells are shown in bold type. (D)TRM T cell marker genes dif- ferentially expressed in CD3+CD8+CD103+ TIL versus CD3+CD8+CD1032 PBMC. ***p , 0.0001.

TGF-b1 secretion, are likely required for maintaining high ex- Ag was present (21). Moreover, influenza virus–specific T lymphocytes, pression levels of CD103 on TIL. Indeed, frequent stimulation of but not EBV- and CMV-specific T lymphocytes, express the integrin, CD8+CD103+ T cell clones (3) with autologous tumor Ag– further emphasizing that a sufficient Ag amount is needed to maintain expressing feeder cells is required to maintain high surface ex- CD103 expression (24). pression levels of CD103. Consistently, CD103 was maintained on The present study also suggests that CD103 is involved, at least vesicular stomatitis virus-specific CD8+ T cells only as long as the in part, in recruitment of CD8+ T cells in lung tumor islets and The Journal of Immunology 3483 Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 3. Characterization of T cell surface markers of CD3+CD8+CD103+ TIL. (A) TIL from freshly isolated NSCLC tumors were analyzed by flow cytometry using indicated mAb (black fill) or corresponding isotypic controls (gray fill). Two representative experiments from five are shown. (B) Surface expression of PD-1, Tim3, and CTLA-4 among CD103+ and CD1032 TIL. Three representative experiments are included. Right panel, Mean percentages of PD-1+, Tim-3+, and CTLA-4+ cells among CD3+CD8+ TIL expressing (or not) CD103 (n = 12). (C) AICD of CD3+CD8+ TIL expressing (or not) CD103. Percentages of apoptotic cells among CD3+CD8+CD103+ and CD3+CD8+CD1032 TIL. Size-selected freshly isolated TIL were cultured with cell-sorted autologous tumor cells and then analyzed for apoptosis induction. Three independent NSCLC tumors are included. Annexin V+PI2 apoptotic cells are shown in the lower right quadrant. Right panel, Mean percentages of annexin V+PI2 apoptotic lymphocytes among CD3+CD8+ TIL expressing (or not) CD103 (n = 5). *p , 0.05, ***p , 0.0001. provides evidence that a high density of CD103+ TIL is associated both the TGF-b1 receptor and TCR via an active form of TGF-b1 with early stage NSCLC patient survival. In our series of NSCLC and the tumor peptide/MHC-I complex, respectively, is required patients, the density of CD103+ cells ranged from 2 to 697/mm2 of for CD103 induction (11). Thus, TGF-b1, considered as an im- tumor area. The essential cause of this variation is not clearly munosuppressive , can control immune responses to viral understood, but it may reflect interpatient variations in the intra- infections and tumors via induction of CD103 and formation of + + 2 tumoral level of TGF-b1 or factors involved in its activation, such CD8 CD103 TRM T cells. The CD103 lymphocytes observed in as aV integrins (30, 31), the number of tumor Ag-specific T cell several NSCLC tumors might represent either nonspecific T cells precursors, or both. Indeed, as mentioned above, engagement of passing through the lung tumor lesion or recently recruited tumor- 3484 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE Downloaded from http://www.jimmunol.org/ by guest on October 1, 2021

FIGURE 4. Functional characterization of freshly size-selected TIL. (A) CD107a induction on CD3+CD8+ TIL expressing (or not) CD103. TIL, cultured in medium or in the presence of rIL-2, were stimulated with autologous tumor cells, and then CD3+CD8+CD103+ and CD3+CD8+CD1032 lymphocytes were analyzed for surface expression of CD107a. Two representative experiments are included. Right panel, Mean percentages of CD107a+ lymphocytes among CD3+CD8+ T cells expressing (or not) CD103 (n = 5). (B) Expression of granzyme B in CD3+CD8+CD103+ and CD3+CD8+CD103- TIL, unsti- mulated or stimulated with rIL-2. Two representative experiments are included. Right panel, Mean percentages of granzyme B–expressing CD3+CD8+ CD103+ and CD3+CD8+CD1032 lymphocytes (n = 3). (C) Cytotoxic activity of freshly isolated TIL toward autologous tumor cells. TIL were preincubated or not with anti–PD-1 blocking mAb and then cultured with autologous tumor cells, untreated or pretreated with anti–PD-L1, in the absence or presence of anti–MHC-I mAb. Anti–MHC-I alone and IgG1 isotypic control were included. (D) Inhibition of TIL-mediated tumor cell killing with neutralizing anti- CD103 mAb. TIL were preincubated with anti-CD103 mAb and then cultured with target cells pretreated with anti–PD-L1 mAb or IgG1 isotypic control. *p , 0.05, **p , 0.001, ***p , 0.0001. specific T cells that do not yet express the integrin. Indeed, mice, was needed for CD103 induction (29). The notion that a prolonged period (8–10 d) of residency of human tumor-specific CD103 is a marker of Ag-reactive T lymphocytes is further sup- T cells within the cognate tumor, engrafted in immunodeficient ported by studies of antiviral and antitumor CTL responses in The Journal of Immunology 3485 which CD103 is expressed on CD8+ T cells specific to influenza reactive CD8+ TIL. Indeed, CD103 delineates a highly activated virus in the lung (24), EBV in the tonsil (32), and cancer testis Ag tumor-specific T cell subset that is able to kill malignant cells NY-ESO-1 in ovarian cancer (9). In the latter and present studies, upon neutralization of the immune checkpoint receptor PD-1. CD103+ T cells expressed high levels of PD-1, which has been Thus, investigating freshly isolated patient TIL contributes defined as a biomarker of tumor-specific T cells (33). More im- not only to our understanding of the local tumor-specific CTL portantly, we show in the present study a correlation between function in NSCLC, but may also provide insight into the role + dense tumor infiltration by CD103 TIL, without the need to of TRM T cell populations in antitumor immune response. By discriminate their location in epithelial versus stromal regions, and characterizing TRM TIL and identifying the mechanisms in- NSCLC patient DFS. Similar results were obtained in ovarian volved in their retention within the tumor ecosystem and cancer, and CD103 expression on TIL was also associated with tumor-induced dysfunctions, we hope that manipulation of this increased patient survival (10). These results suggest that CD103 tumor-specific T cell subset will lead to enhanced immune pro- alone may be used as a prognostic biomarker in epithelial tumors. tection and improvement in current immunotherapy approaches for Conversely, no correlation between total CD8+ TIL and NSCLC cancer treatment. patient survival was observed unless they were located in intra- epithelial tumor regions, in which case improved DFS outcome was observed. Thus, CD103 identifies a subpopulation of CD8+ Acknowledgments T cells that more accurately predicts patient outcome. We thank Yann Lecluse and Philippe Rameau for help with FACS anal- Immune checkpoints are important for maintaining self- yses. We also thank Olivia Bawa for help with immunohistochemistry tolerance and regulating immune responses in peripheral tissues staining. We are grateful to Ste´phanie Corgnac for critical reading of (34). Among these inhibitory receptors, CTLA-4, PD-1, and Tim- the manuscript. Downloaded from 3 appeared to be associated with T cell exhaustion in chronic viral + infections (35–38) and tumor progression (39). Exhausted CD8 Disclosures T cells frequently coexpress PD-1 and Tim-3, and triggering of The authors have no financial conflicts of interest. both molecules has been reported to transmit a death signal to T cells (40, 41). It has also been suggested that tumor cells use the PD-1/PD-L1 pathway to escape CD8+ T cell immunity by in- References http://www.jimmunol.org/ ducing T cell apoptosis (42). Our data indicate that CD103+ TIL 1. Bossi, G., C. Trambas, S. Booth, R. Clark, J. Stinchcombe, and G. M. Griffiths. frequently express PD-1 and Tim-3, and that this T cell subset is 2002. The secretory synapse: the secrets of a serial killer. Immunol. Rev. 189: 2 152–160. much more highly susceptible to AICD than CD103 T cells. 2. Anikeeva, N., K. Somersalo, T. N. Sims, V. K. Thomas, M. L. Dustin, and Moreover, anti–PD-1 neutralizing mAb promotes tumor cell kill- Y. Sykulev. 2005. Distinct role of lymphocyte function-associated antigen-1 in + mediating effective cytolytic activity by cytotoxic T lymphocytes. Proc. Natl. ing by autologous TIL, supporting the conclusion that CD103 Acad. Sci. USA 102: 6437–6442. + TIL correspond to an activated CD8 T cell subpopulation, and 3. Le Floc’h, A., A. Jalil, I. Vergnon, B. Le Maux Chansac, V. Lazar, G. Bismuth, that blocking PD-1–PD-L1 interactions enhances antitumor ef- S. Chouaib, and F. Mami-Chouaib. 2007. aEb7 integrin interaction with E- cadherin promotes antitumor CTL activity by triggering lytic granule polariza- fector functions. It is now admitted that binding of PD-1, fre-

tion and exocytosis. J. Exp. Med. 204: 559–570. by guest on October 1, 2021 quently upregulated on tumor-specific T cells, to its ligand PD-L1 4. Kilshaw, P. J., and S. J. Murant. 1990. A new surface antigen on on malignant cells results in inhibition of T lymphocyte activation intraepithelial lymphocytes in the intestine. Eur. J. Immunol. 20: 2201–2207. 5. Sung, S. S., S. M. Fu, C. E. Rose, Jr., F. Gaskin, S. T. Ju, and S. R. Beaty. 2006. A (41). Targeting of the PD-1/PD-L1 pathway to enhance antitumor major lung CD103 (aE)-b7 integrin-positive epithelial dendritic cell population immunity is supported by the recent success of anti–PD-1 and expressing Langerin and tight junction . J. Immunol. 176: 2161–2172. 6. Annacker, O., J. L. Coombes, V. Malmstrom, H. H. Uhlig, T. Bourne, anti–PD-L1-based cancer immunotherapy in patients with various B. Johansson-Lindbom, W. W. Agace, C. M. Parker, and F. Powrie. 2005. Es- types of advanced solid tumors, including NSCLC (43, 44). sential role for CD103 in the T cell-mediated regulation of experimental colitis. We previously reported that TIL entry into the tumor site J. Exp. Med. 202: 1051–1061. 7. French, J. J., J. Cresswell, W. K. Wong, K. Seymour, R. M. Charnley, and requires, at least in part, expression of CCR5, and that CCR6 is J. A. Kirby. 2002. T cell adhesion and cytolysis of pancreatic cancer cells: a role induced within the tumor and likely participates in intratumoral for E-cadherin in immunotherapy? Br. J. Cancer 87: 1034–1041. + T cell migration (29). In the present study we show that freshly 8. Quinn, E., N. Hawkins, Y. L. Yip, C. Suter, and R. Ward. 2003. CD103 intraepithelial lymphocytes—a unique population in microsatellite unstable isolated TIL also express high levels of both chemokine receptors, sporadic colorectal cancer. Eur. J. Cancer 39: 469–475. + 9. Webb, J. R., D. A. Wick, J. S. Nielsen, E. Tran, K. Milne, E. McMurtrie, and suggesting that they may be involved in the formation of CD8 + + B. H. Nelson. 2010. Profound elevation of CD8 T cells expressing the CD103 TRM T cells in human lung tumors. This is consistent intraepithelial lymphocyte marker CD103 (aE/b7 integrin) in high-grade serous with our previous findings demonstrating that CD103-mediated ovarian cancer. Gynecol. Oncol. 118: 228–236. recruitment of CCR5 at the immune synapse between TIL and 10. Webb, J. R., K. Milne, P. Watson, R. J. Deleeuw, and B. H. Nelson. 2014. Tumor- infiltrating lymphocytes expressing the tissue resident memory marker CD103 specific cancer cells contributes, at least in part, to T cell retention are associated with increased survival in high-grade serous ovarian cancer. Clin. within the tumor by inhibiting lymphocyte responsiveness to Cancer Res. 20: 434–444. CCL5 (29). CCR5 ligand CCL5 was one of the major 11. Mokrani, M., J. Klibi, D. Bluteau, G. Bismuth, and F. Mami-Chouaib. 2014. Smad and NFAT pathways cooperate to induce CD103 expression in human CD8 involved in regulating antitumor immune responses (45). In- T lymphocytes. J. Immunol. 192: 2471–2479. creased levels of CCR5 ligands have been demonstrated to im- 12. Franciszkiewicz, K., A. Le Floc’h, M. Boutet, I. Vergnon, A. Schmitt, and F. Mami-Chouaib. 2013. CD103 or LFA-1 engagement at the immune synapse prove antitumor functions by inducing effector T cell infiltration between cytotoxic T cells and tumor cells promotes maturation and regulates (46). Indeed, intratumoral production of CCL5 and CCL3 was T-cell effector functions. Cancer Res. 73: 617–628. reported to promote recruitment of CD8+ T cells and, thereby, 13. Le Floc’h, A., A. Jalil, K. Franciszkiewicz, P. Validire, I. Vergnon, and F. Mami- Chouaib. 2011. Minimal engagement of CD103 on cytotoxic T lymphocytes CTL-dependent control of tumor progression (47). Moreover, with an E-cadherin-Fc molecule triggers lytic granule polarization via a phos- CCR5 deficiency was correlated with reduced CD8+ T cell infil- pholipase Cg-dependent pathway. Cancer Res. 71: 328–338. tration, poor prognosis (48, 49), and reduced efficacy of cancer 14. Turner, D. L., C. L. Gordon, and D. L. Farber. 2014. Tissue-resident T cells, in situ immunity and transplantation. Immunol. Rev. 258: 150–166. immunotherapy (50). 15. Feng, Y., D. Wang, R. Yuan, C. M. Parker, D. L. Farber, and G. A. Hadley. 2002. Overall, our results demonstrate a major role for CD8+CD103+ CD103 expression is required for destruction of pancreatic islet allografts by CD8+ T cells. J. Exp. Med. 196: 877–886. T cells infiltrating human epithelial tumor lesions in the antitumor 16. El-Asady, R., R. Yuan, K. Liu, D. Wang, R. E. Gress, P. J. Lucas, immune response and identify CD103 as a biomarker of tumor- C. B. Drachenberg, and G. A. Hadley. 2005. TGF-b-dependent CD103 expres- 3486 ROLE OF CD8+CD103+ TIL IN ANTITUMOR CTL RESPONSE

sion by CD8+ T cells promotes selective destruction of the host intestinal epi- 35. Trautmann, L., L. Janbazian, N. Chomont, E. A. Said, S. Gimmig, B. Bessette, thelium during graft-versus-host disease. J. Exp. Med. 201: 1647–1657. M. R. Boulassel, E. Delwart, H. Sepulveda, R. S. Balderas, et al. 2006. Up- 17. Zhang, N., and M. J. Bevan. 2013. Transforming -b signaling regulation of PD-1 expression on HIV-specific CD8+ T cells leads to reversible controls the formation and maintenance of gut-resident memory T cells by immune dysfunction. Nat. Med. 12: 1198–1202. regulating migration and retention. Immunity 39: 687–696. 36. Freeman, G. J., E. J. Wherry, R. Ahmed, and A. H. Sharpe. 2006. Reinvigorating 18. Mackay, L. K., A. Rahimpour, J. Z. Ma, N. Collins, A. T. Stock, M. L. Hafon, exhausted HIV-specific T cells via PD-1-PD-1 ligand blockade. J. Exp. Med. J. Vega-Ramos, P. Lauzurica, S. N. Mueller, T. Stefanovic, et al. 2013. The 203: 2223–2227. developmental pathway for CD103+CD8+ tissue-resident memory T cells of 37. Kaufmann, D. E., D. G. Kavanagh, F. Pereyra, J. J. Zaunders, E. W. Mackey, skin. Nat. Immunol. 14: 1294–1301. T. Miura, S. Palmer, M. Brockman, A. Rathod, A. Piechocka-Trocha, et al. 2007. 19. Wakim, L. M., A. Woodward-Davis, R. Liu, Y. Hu, J. Villadangos, G. Smyth, Upregulation of CTLA-4 by HIV-specific CD4+ T cells correlates with disease and M. J. Bevan. 2012. The molecular signature of tissue resident memory CD8 progression and defines a reversible immune dysfunction. Nat. Immunol. 8: T cells isolated from the brain. J. Immunol. 189: 3462–3471. 1246–1254. 20. Masopust, D., D. Choo, V. Vezys, E. J. Wherry, J. Duraiswamy, R. Akondy, 38. Blackburn, S. D., H. Shin, G. J. Freeman, and E. J. Wherry. 2008. Selective J. Wang, K. A. Casey, D. L. Barber, K. S. Kawamura, et al. 2010. Dynamic T cell expansion of a subset of exhausted CD8 T cells by aPD-L1 blockade. Proc. Natl. migration program provides resident memory within intestinal epithelium. Acad. Sci. USA 105: 15016–15021. J. Exp. Med. 207: 553–564. 39. Fourcade, J., Z. Sun, M. Benallaoua, P. Guillaume, I. F. Luescher, C. Sander, 21. Wakim, L. M., A. Woodward-Davis, and M. J. Bevan. 2010. Memory T cells J. M. Kirkwood, V. Kuchroo, and H. M. Zarour. 2010. Upregulation of Tim-3 and persisting within the brain after local infection show functional adaptations to PD-1 expression is associated with tumor antigen-specific CD8+ T cell dys- their tissue of residence. Proc. Natl. Acad. Sci. USA 107: 17872–17879. function in melanoma patients. J. Exp. Med. 207: 2175–2186. 22. Gebhardt, T., and F. R. Carbone. 2009. Immunology: a helpers’ guide to in- 40. Jones, R. B., L. C. Ndhlovu, J. D. Barbour, P. M. Sheth, A. R. Jha, B. R. Long, fection. Nature 462: 418–419. J. C. Wong, M. Satkunarajah, M. Schweneker, J. M. Chapman, et al. 2008. Tim-3 23. Jiang, X., R. A. Clark, L. Liu, A. J. Wagers, R. C. Fuhlbrigge, and T. S. Kupper. expression defines a novel population of dysfunctional T cells with highly + 2012. Skin infection generates non-migratory memory CD8 TRM cells provid- elevated frequencies in progressive HIV-1 infection. J. Exp. Med. 205: ing global skin immunity. Nature 483: 227–231. 2763–2779. 24. Piet, B., G. J. de Bree, B. S. Smids-Dierdorp, C. M. van der Loos, 41. Sakuishi, K., L. Apetoh, J. M. Sullivan, B. R. Blazar, V. K. Kuchroo, and E. B. Remmerswaal, J. H. von der Thusen,€ J. M. van Haarst, J. P. Eerenberg, A. C. Anderson. 2010. Targeting Tim-3 and PD-1 pathways to reverse T cell A. ten Brinke, W. van der Bij, et al. 2011. CD8+ T cells with an intraepithelial exhaustion and restore anti-tumor immunity. J. Exp. Med. 207: 2187–2194. Downloaded from phenotype upregulate cytotoxic function upon influenza infection in human lung. 42. Dong, H., S. E. Strome, D. R. Salomao, H. Tamura, F. Hirano, D. B. Flies, J. Clin. Invest. 121: 2254–2263. P. C. Roche, J. Lu, G. Zhu, K. Tamada, et al. 2002. Tumor-associated B7-H1 25. Mackay, L. K., A. T. Stock, J. Z. Ma, C. M. Jones, S. J. Kent, S. N. Mueller, promotes T-cell apoptosis: a potential mechanism of immune evasion. Nat. Med. W. R. Heath, F. R. Carbone, and T. Gebhardt. 2012. Long-lived epithelial im- 8: 793–800. munity by tissue-resident memory T (TRM) cells in the absence of persisting 43. Topalian, S. L., F. S. Hodi, J. R. Brahmer, S. N. Gettinger, D. C. Smith, local antigen presentation. Proc. Natl. Acad. Sci. USA 109: 7037–7042. D. F. McDermott, J. D. Powderly, R. D. Carvajal, J. A. Sosman, M. B. Atkins, 26. Cyster, J. G., and S. R. Schwab. 2012. Sphingosine-1-phosphate and lymphocyte et al. 2012. Safety, activity, and immune correlates of anti-PD-1 antibody in

egress from lymphoid organs. Annu. Rev. Immunol. 30: 69–94. cancer. N. Engl. J. Med. 366: 2443–2454. http://www.jimmunol.org/ 27. Cauley, L. S., and L. Lefranc¸ois. 2013. Guarding the perimeter: protection of the 44. Champiat, S., E. Ileana, G. Giaccone, B. Besse, G. Mountzios, A. Eggermont, mucosa by tissue-resident memory T cells. Mucosal Immunol. 6: 14–23. and J. C. Soria. 2014. Incorporating immune-checkpoint inhibitors into systemic 28. Skon, C. N., J. Y. Lee, K. G. Anderson, D. Masopust, K. A. Hogquist, and therapy of NSCLC. J. Thorac. Oncol. 9: 144–153. S. C. Jameson. 2013. Transcriptional downregulation of S1pr1 is required for the 45. Mule´, J. J., M. Custer, B. Averbook, J. C. Yang, J. S. Weber, D. V. Goeddel, establishment of resident memory CD8+ T cells. Nat. Immunol. 14: 1285–1293. S. A. Rosenberg, and T. J. Schall. 1996. RANTES secretion by gene-modified 29. Franciszkiewicz, K., A. Le Floc’h, A. Jalil, F. Vigant, T. Robert, I. Vergnon, tumor cells results in loss of tumorigenicity in vivo: role of immune cell sub- A. Mackiewicz, K. Benihoud, P. Validire, S. Chouaib, et al. 2009. Intratumoral populations. Hum. Gene Ther. 7: 1545–1553. induction of CD103 triggers tumor-specific CTL function and CCR5-dependent 46. Uekusa, Y., W. G. Yu, T. Mukai, P. Gao, N. Yamaguchi, M. Murai, T-cell retention. Cancer Res. 69: 6249–6255. K. Matsushima, S. Obika, T. Imanishi, Y. Higashibata, et al. 2002. A pivotal role 30. Munger, J. S., X. Huang, H. Kawakatsu, M. J. Griffiths, S. L. Dalton, J. Wu, for CC 5 in T-cell migration to tumor sites induced by in- J. F. Pittet, N. Kaminski, C. Garat, M. A. Matthay, et al. 1999. The integrin avb6 terleukin 12 treatment in tumor-bearing mice. Cancer Res. 62: 3751–3758.

binds and activates latent TGF b1: a mechanism for regulating pulmonary in- 47. Jo¨hrer, K., L. Pleyer, A. Olivier, E. Maizner, C. Zelle-Rieser, and R. Greil. 2008. by guest on October 1, 2021 flammation and fibrosis. Cell 96: 319–328. Tumour-immune cell interactions modulated by chemokines. Expert Opin. Biol. 31. Henderson, N. C., T. D. Arnold, Y. Katamura, M. M. Giacomini, J. D. Rodriguez, Ther. 8: 269–290. J. H. McCarty, A. Pellicoro, E. Raschperger, C. Betsholtz, P. G. Ruminski, et al. 48. Zimmermann, T., M. Moehler, I. Gockel, G. G. Sgourakis, S. Biesterfeld, 2013. Targeting of av integrin identifies a core molecular pathway that regulates M. Muller,€ M. R. Berger, H. Lang, P. R. Galle, and C. C. Schimanski. 2010. Low fibrosis in several organs. Nat. Med. 19: 1617–1624. expression of chemokine receptor CCR5 in human colorectal cancer correlates 32. Woodberry, T., T. J. Suscovich, L. M. Henry, M. August, M. T. Waring, A. Kaur, with lymphatic dissemination and reduced CD8+ T-cell infiltration. Int. J. Co- C. Hess, J. L. Kutok, J. C. Aster, F. Wang, et al. 2005. aEb7 (CD103) expression lorectal Dis. 25: 417–424. identifies a highly active, tonsil-resident effector-memory CTL population. 49. Man˜es, S., E. Mira, R. Colomer, S. Montero, L. M. Real, C. Go´mez-Mouto´n, J. Immunol. 175: 4355–4362. S. Jime´nez-Baranda, A. Garzo´n, R. A. Lacalle, K. Harshman, et al. 2003. CCR5 33. Gros, A., P. F. Robbins, X. Yao, Y. F. Li, S. Turcotte, E. Tran, J. R. Wunderlich, expression influences the progression of human breast cancer in a p53-dependent A. Mixon, S. Farid, M. E. Dudley, et al. 2014. PD-1 identifies the patient-specific manner. J. Exp. Med. 198: 1381–1389. CD8+ tumor-reactive repertoire infiltrating human tumors. J. Clin. Invest. 124: 50. Ugurel, S., D. Schrama, G. Keller, D. Schadendorf, E. B. Bro¨cker, R. Houben, 2246–2259. M. Zapatka, W. Fink, H. L. Kaufman, and J. C. Becker. 2008. Impact of the 34. Pardoll, D. M. 2012. The blockade of immune checkpoints in cancer immuno- CCR5 gene polymorphism on the survival of metastatic melanoma patients re- therapy. Nat. Rev. Cancer 12: 252–264. ceiving immunotherapy. Cancer Immunol. Immunother. 57: 685–691. A C CD3+/CD8+/CD103+T cells TIL PBMC BCL-2 CD127 KLRG1 250 100 100 83% 51% 98% 27% 1% 200 (3069) (4944) 80 (7814) (10001) 80 150 CD3 100 60 60 50 40 Patient 81 40 Patient 83 0

20 20 250 53% 200 40% 0 0 (20897) (16848) 150 CD4 100 number 100 number 70% 6%

1% 100 80 (4724) (8426) 80 50 Cell Cell 0 60 60 250 Patient 82 Patient 84 36% 15% 40 40 200 (16560) (7304) CD8 20 150 20 100 0 0 50 102 103 104 105 102 103 104 105 102 103 104 105 0 102 103 104 105 102 103 104 105 Fluorescence intensity Fluorescence intensity

Fluorescence D intensity Size selected cells

100 MHC-I Epcam E-cadherin PD-L1 TIL 250 cells 90% 91% 74% 57% 80 200 PBMC (19848) (11441) (3048) (7513) 150

60 100 Patient 103 50 selected

40 0

number 250 89% 80% 52% 76% 20 200 (32218) (35051) (4870) (10298) among Cell 150 Patient 104

% 0 100 CD3 CD4 CD8 50 0 102 103 104 105 102 103 104 105 102 103 104 105 102 103 104 105

B Fluorescence intensity

Volcanoplot TIL versus PBMC 100 cells

30 80 ) 25 60 selected 20 40 p-value 15 size- 20 10 0

5 among Log10(Adj. % 0 MHC-I Epcam PD-L1 -5 0 5 E-cadherin Log2(FC) FC > 2 Underexpressed: 233 P-value ≤ 10-10 Overexpressed : 254

Supplemental Figure 1 A. Expression of CD3, CD4 and CD8 T-cell markers on TIL isolated from a NSCLC tumor on the basis of small cell size selection and NSCLC patient PBMC. Upper panel: TIL and PBMC were analyzed by flow cytometry using anti-CD3, anti-CD4 or anti-CD8 mAb (black fill) or an isotypic control (gray fill). Lower panel: Mean percentages of CD3+, CD4+ and CD8+ T cells among size-selected TIL (n=13) and PBMC (n=19). B. Global gene expression analyses of patient TIL and PBMC. Differential analysis of gene expression profiles in NSCLC patient TIL (n=13) and PBMC (n=19) are shown. Genes of interest were selected using a FC > 2 and a p-value ≤ 10-10 (487 genes). A subset of 103 genes with intensity higher than 1,000 in TIL samples was selected for further analysis. C. Expression of Bcl-2, CD127 and KLRG1 in CD3+/CD8+/CD103+ freshly size-selected NSCLC TIL. Flow cytometry analysis of intracytoplasmic (Bcl-2) and surface (CD127 and KLRG1) expression was performed using specific Ab (black fill) or corresponding isotypic controls (gray fill). Two representative experiments are included. D. Expression of MHC-I, Epcam, E-cadherin and PDL-1 on cell-sorted tumor cells. Percentages of positive cells are indicated; mean fluorescence intensities are in parentheses. Lower panel: Mean percentages of MHC-I+, Epcam+, E-cadherin+ and PD-L1+ cells among large size-selected cells (n=8).

A Z-score B

-2,5 2,7 TIL PBMC PBMC TIL 400 200 Fc 300

Fc 150 RGS1 200 RALGDS 100 ITGAE 50 100 ITM2B ITGB2 6

PLEK RGS-1 10 ARRB2 4 PXN 5 2 APLP2 TNFRSF-4 0 VCL 0 Adhesion PTGER2 400 ITGAM 60 CD33 300 C5AR1 Fc 40

Fc 200

CD36 CSF3R 100 VCAN 20 S100A9 12 GP1BB 6 ICOS 9 PPBP

4 RALGDS PF4 Activation 6 TNFRSF9 2 3 TNFRSF4 ICOS 0 0 IL15 100 Activation IL17RA Adhesion Fc PDCD1 600 80 LAG3 300 CTLA4 60 TNFRSF18 Fc 40 HAVCR2 Exhaustion IL6 60 20 IL21 6 Inhibition HSPA1B 40 TNFRSF-9

HSPA1A ITGAE 4

CYR61 20 HSPD1 2 GADD45G 0 BAG3 0 TNFSF9 CCR5 RASD1 BIRC3 320 JUN

EGR1 150 240

BIK Fc 100 Fc 160 Apoptosis TNFAIP3 50 NR4A2 80 TFRC 30 EGR2 25 12 SERPINB9 20 9 BAG-3

IFIH1 PDCD1

15 ST6GAL1 10 6 BNIP2 5 3 TSPO CDKN2D 0 AIF1 LYN 40 NLRP3

FGR 80

LST1 Fc

LILRB2 60 30 CFP 20 Fc 40 S100A8 S100A9 20 15 FCN1 S100A12 6 EGR-1 10 Inflammation PPBP 4 LAG-3 5 MAP3K3 Apoptosis KLF13 2 TKT 0 0 KCNQ1 ZNF385A LMO2 100

ASGR2 80 KLF13 400

Metal Ion SMARCD3 Fc 60 200

SPI1 Fc 40 binding MAFB 20 MTPN 20 ARRB2 15 PTGER2 6 NR4A2

SLC7A7 Exhaustion 10 Regulation ACRBP 4

CHP1 CTLA-4 5 2 DNA SH3BGRL PTP4A2 0 0 VAMP3 Proliferation HSD17B11 LAMTOR1 P34 P35 P36 P37 P38 P39 P40 P41

DENND5A FBXL5 200 100 PTPLAD2 Fc C10ORF54 25 SLC16A5 THEMIS2 20 FAM101B 15 AP1S2 C NACC2 10 AHNAK KBTBD11 5 MPEG1 250 Other TNFRSF-18 R² = 0.811 CPVL 0

functions LRRC25 200 PRAM1

PLBD1 20 Fc RGS18 150

NRGN Fc 15

10 100 Chip 6 4 50

HAVCR-2 2

0 0 100 200 300

P34 P35 P36 P37 P38 P39 P40 P41 qRT-PCR Fc P 34 P 35 P 36 P 37 P 38 P 39 P 40 P 41 P 42 P 43 P 44 P 45 P 46 P 47 P 48 P 49 P 50 P 51 P 52 P 53 P 54 P 55 P 56 P 57 P 58 P 59 P 60 P 61 P 62 P 63 P 64 P 65

Supplemental Figure 2 A. Heat map of gene expression of the 103-gene subset. Data were bi-clusterized by genes and by samples using Pearson’s coefficient as the distance metric and the Ward method for clusterization. The magnitude of relative expression of a particular gene is shown in color (LogRatio). Microarray analysis was performed using Ingenuity Pathway Analysis (IPA) and DAVID Bioinformatics. B. qRT-PCR analyses of selected genes. qRT-PCR validation of genes differentially expressed in TIL and patient PBMC. The expression of genes differentially expressed in TIL from 8 NSCLC tumors (4 SCC, 1 LCC and 3 ADC) and PBMC from 8 NSCLC patients (3 SCC, 1 LCC and 4 ADC). C. Correlation between microarray analysis and qRT-PCR in TIL and PBMC.

Supplemental Table I: Complete results of the univariate and the multivariates analyses for the overall survival outcome a b c Univariate Multivariate Multivariate Multivariate

HR 95% CI P-v HR 95% CI P-v HR 95% CI P-v HR 95% CI P-v CD3 Total 0.97 (0.91-1.04) 0.40

CD3 Tumor 0.94 (0.76-1.15) 0.50 CD3 Stroma 0.97 (0.89-1.06) 0.50 CD8 Total 0.90 (0.8-1.02) 0.10 CD8 Tumor 0.83 (0.63-1.09) 0.20 CD8 Stroma 0.88 (0.74-1.05) 0.20 CD103 Total 0.88 (0.77-0.99) 0.04 0.89 (0.78-1.02) 0.09 CD103 Tumor 0.84 (0.69-1.03) 0.09 0.84 (0.66-1.05) 0.13

CD103 Stroma 0.78 (0.62-0.98) 0.03 0.84 (0.67-1.06) 0.14 for an increase of 50 units Age <65 1.00 1.00 1.00 1.00 ≥65 2.99 (1.62-5.51) 0.001 2.17 (1.11-4.21) 0.02 2.15 (1.1-4.22) 0.03 2.36 (1.23-4.53) 0.01 Gender Female 1.00 1.00 1 1.00 Male 1.70 (0.86-3.37) 0.10 1.85 (0.85-4.01) 0.12 1.89 (0.87-4.12) 0.11 1.79 (0.83-3.89) 0.14 Smoking status Ever 1.00 Never 0.52 (0.16-1.69) 0.30 Stage Stade Ia 1.00 1.00 1 1.00 Stade Ib 1.67 (0.84-3.32) 0.10 1.57 (0.72-3.44) 0.26 1.54 (0.7-3.36) 0.28 1.50 (0.69-3.26) 0.31 Histology ADC 1.00 1.00 1 1.00 SCC 2.58 (1.35-4.92) 0.004 1.62 (0.76-3.45) 0.21 1.75 (0.82-3.75) 0.15 1.52 (0.71-3.25) 0.28 Others 1.15 (0.38-3.47) 0.80 1.26 (0.42-3.77) 0.68 1.31 (0.43-4) 0.64 1.10 (0.37-3.23) 0.87 Stage Ia: pT1N0 and stage Ib: pT2N0. HR: hazard ratio, CI: confidence interval. The HR (95% CI) were calculated for an increment of 50 for CD3, CD8 and CD103. P-v: Wald test p-values. a-c: three multivariate cox regression models adjusted on the clinical factors (age, gender, smoking stauts, stage and histology), and including CD103 Total (a), CD103 Tumor (b) and CD103 Stroma (c).

Supplemental Table II: Differential gene expression profiles of NSCLC TIL and patient PBMC

Genes more strongly expressed in TIL than in PBMC Sequence name Accession N° Sequence description FC Int TIL Int PBMC Adhesion RGS1 NM_002922 Regulator of G-protein signaling 1 24 40364 1700 RALGDS NM_001042368 Ral guanine nucleotide dissociation stimulator 2 2472 1293 ITGAE NM_002208 Integrin, alpha E (Ag CD103) 2 1132 312 Exhaustion PDCD1 NM_005018 Programmed cell death 1 13 7880 620 LAG3 NM_002286 Lymphocyte activation gene 3 10 9284 890 CTLA4 NM_005214 Cytotoxic T-lymphocyte-associated protein 4 9 3786 428 TNFRSF18 NM_148901 receptor superfamily, member 18 8 1465 180 HAVCR2 NM_032782 Hepatitis A virus cellular receptor 2 2 2974 1690 Immunosupression IL-6 NM_000600 6 (, beta 2) 11 1976 178 IL-21 NM_021803 (IL21), transcript variant 1 8 109855 9136 Activation TNFRSF9 NM_001561 Tumor necrosis factor receptor superfamily, member 9 17 1198 70 TNFSF4 NM_003327 Tumor necrosis factor receptor superfamily, member 4 13 7100 546 ICOS NM_012092 Inducible T-cell co-stimulator 8 10073 1306 Apoptosis HSPA1B NM_005346 Heat shock 70kDa protein 1B 233 75665 325 HSPA1A NM_005345 Heat shock 70kDa protein 1A 110 156625 1416 CYR61 NM_001554 Cysteine-rich, angiogenic inducer, 61 27 19617 4852 HSPB1 NM_002156 Heat shock 60kDa protein 1 (chaperonin) 12 5256 1822 GADD45G NM_006705 Growth arrest and DNA-damage-inducible, gamma 10 3282 309 BAG-3 NM_004281 BCL2-associated athanogene 3 10 5256 1822 TNFSF9 NM_003811 Tumor necrosis factor (ligand) superfamily, member 9 8 1632 199 CCR5 NM_000579 Chemokine (C-C motif) receptor 5 8 7020 910 RASD1 NM_016084 RAS, dexamethasone-induced 1 6 3266 493 BIRC3 NM_001165 Baculoviral IAP repeat containing 3 5 9536 1882 JUN NM_002228 Jun proto-oncogene 4 10209 2217 EGR1 NM_001964 Early growth response 1 4 19617 4852 BIK NM_001197 BCL2-interacting killer (apoptosis-inducing) 3 2974 1687 TNFAIP3 NM_006290 Tumor necrosis factor, alpha-induced protein 3 3 23395 7931 NR4A2 NM_006186 Nuclear receptor subfamily 4, group A, member 2 3 5256 1822 TFRC NM_003234 Transferrin receptor (p90, CD71) 3 5486 1940 EGR2 NM_000399 Early growth response 2 3 2738 968 SERPINB9 NM_004155 Serpin peptidase inhibitor, clade B (ovalbumin), member 9 2 2948 1109 IFIH1 NM_022168 Interferon induced with helicase C domain 1 2 1712 672 ST6GAL1 NM_173216 ST6 beta-galactosamide alpha-2,6-sialyltranferase 1 2 6789 3307

Genes less strongly expressed in TIL than in PBMC Adhesion ITM2B NM_021999 integral membrane protein 2B -2 5272 2634 ITGB2 NM_000211 integrin, beta 2 -3 19363 6067 PLEK NM_002664 Pleckstrin -4 1862 471 ARRB2 NM_004313 arrestin, beta 2 -4 8774 2186 PXN NM_002859 Paxillin -5 10915 2370 APLP2 NM_001642 amyloid beta (A4) precursor-like protein 2 -5 7460 1465 VCL NM_014000 Vinculin -6 1554 254 PTGER2 NM_000956 prostaglandin E receptor 2 (subtype EP2), 53kDa -6 1567 252 ITGAM NM_000632 integrin, alpha M -7 2077 314 CD33 NM_001772 CD33 molecule -15 1204 80 C5AR1 NM_001736 complement component 5a receptor 1 -16 1428 89 CD36 NM_001001547 CD36 molecule (thrombospondin receptor) -16 1102 68 CSF3R NM_156039 colony stimulating factor 3 receptor -22 5015 229 VCAN NM_004385 -32 2784 87 S100A9 NM_002965 S100 calcium binding protein A9 -38 10434 272 GP1BB NM_000407 Ib (platelet), beta polypeptide -64 2389 37 PPBP NM_002704 pro-platelet basic protein (chemokine (C-X-C motif) ligand 7) -284 1100 4 PF4 NM_002619 -412 1723 4 Activation IL-15 NM_172175 -3 1046 304 IL-17RA NM_014339 receptor A -4 2589 603 Inflammation AIF1 NM_004847 allograft inflammatory factor 1 -4 1128 316 LYN NM_002350 v-yes-1 Yamaguchi sarcoma viral related oncogene homolog -4 6558 1798 NLRP3 NM_001243133 NLR family, pyrin domain containing 3 -8 1074 139 FGR NM_001042747 feline Gardner-Rasheed sarcoma viral oncogene homolog -8 17139 2158 LST1 NM_007161 leukocyte specific transcript 1 -9 4485 507 LILRB2 NM_005874 leukocyte immunoglobulin-like receptor, subfamily B member 2 -13 3371 254 CFP NM_002621 complement factor properdin -22 9319 426 S100A8 NM_002964 S100 calcium binding protein A8 -27 42026 1543 S100A9 NM_002965 S100 calcium binding protein A9 -38 10434 272 FCN1 NM_002003 ficolin (collagen/fibrinogen domain containing) 1 -58 14482 251 S100A12 NM_005621 S100 calcium binding protein A12 -91 5163 57 PPBP NM_002704 pro-platelet basic protein (chemokine (C-X-C motif) ligand 7) -284 1100 4 Apoptosis Bnip2 NM_004330 BCL2/adenovirus E1B 19kDa interacting protein 2 -3 4631 1662 Tspo NM_000714 translocator protein (18kDa) -3 36434 10732 Cdkn2d NM_001800 cyclin-dependent kinase inhibitor 2D (p19, inhibits CDK4) -4 12890 2869 Metal ion binding MAP3K3 NM_203351 mitogen-activated protein kinase kinase kinase 3 -2 5330 2299 KLF13 NM_015995 Kruppel-like factor 13 -3 11788 4340 TKT NM_001064 Transketolase -5 27727 5571 KCNQ1 NM_000218 potassium voltage-gated channel, KQT-like subfamily, member 1 -6 2024 358 ZNF385A NM_015481 zinc finger protein 385A -6 1622 286 LMO2 NM_005574 LIM domain only 2 (rhombotin-like 1) -12 3449 282 ASGR2 NM_080912 receptor 2 -29 1028 35 Regulation DNA KLF13 NM_015995 Kruppel-like factor 13 -3 11788 4340 SWI/SNF related, matrix associated, actin dependent regulator of -4 3089 775 SMARCD3 NM_003078 chromatin, subfamily d, member 3 SPI1 NM_001080547 spleen focus forming virus (SFFV) proviral integration oncogene -9 3399 370 MAFB NM_005461 v-maf avian musculoaponeurotic fibrosarcoma oncogene homolog B -13 12102 920 Proliferation MTPN NM_145808 Myotrophin -2 2405 1139 ARRB2 NM_004313 arrestin, beta 2 -4 8774 2186 PTGER2 NM_000956 prostaglandin E receptor 2 -6 1567 252 SLC7A7 NM_001126106 solute carrier family 7 -8 2963 369 ACRBP NM_032489 acrosin binding protein -21 1077 51

Gene expression profiles of freshly isolated TIL were compared with those of PBMC from NSCLC patients (p-value < 10-10, Fold change (FC) > 2 and intensity expression > 1,000). The selected 33 genes, which were more strongly expressed in TIL than in PBMC, and the 70 genes, which were less strongly expressed in TIL than in PBMC, were classified according to their presumed biological functions by ingenuity software (www.ingenuity.com). Int TIL: intensity of transcripts obtained in TIL; Int PBMC: intensity of transcripts obtained in PBMC.