PD-1–Expressing Tumor-Infiltrating T Cells Are a Favorable Prognostic Biomarker in HPV-Associated Head and Neck Cancer

Recent studies have detected oncogenic HPV in 25.9% of head and neck cancers (mostly HPV16) and this prevalence increases to more than 50% for cancers of the oropharynx (tonsil, base of tongue, etc.) with a growing worldwide incidence (1).

Although these tumors frequently have aggressive histopathological features (2), the presence of HPV DNA is a favorable prognostic factor with regard to recurrence and survival (3–5). The mechanisms underlying the more favorable outcome of HPV-associated head and neck cancer have not been clearly elucidated.

These tumors are more responsive to chemotherapy and radiotherapy, but even patients with HPV-positive oropharyngeal cancers treated by surgery alone have a better prognosis than those with HPV-negative tumors after adjustment for tumor stage (6, 7).

HPV E6-E7 expressing tumors grew slower in immunocompetent mice but not in nude mice, than HPV-negative tumors suggesting a role of adaptive immunity and not simply an intrinsic biologic property of HPV-positive tumor cells to explain this difference (8).

In human, T cells directed against HPV derived proteins (E2, E6, and E7) have been described in the blood and in the tumor of patients with head and neck cancers (9–12). However, the capacity of tumor-infiltrating lymphocytes (TIL) to act as effector cells may be affected by the tumor microenvironment. Immunosuppressive cells or molecules (13, 14) or chronic inflammation in head and neck cancer predispose to a tumor promoting microenvironment (15, 16). Analysis of differences in the type of immunity within the tumor microenvironment of HPV-associated head and neck cancer compared with HPV-negative tumors may, therefore, provide new clues to explain the more favorable clinical behavior of this type of cancer. In the present study, we addressed the clinical significance of regulatory Foxp3⁺T cells and programmed death (PD-1)-positive T-cell infiltration in the tumor microenvironment of HPV-positive and HPV-negative head and neck cancers. These immunosuppressive cells were selected because regulatory CD4⁺CD25⁺Foxp3⁺T cells have emerged as the dominant T-cell population governing peripheral tolerance by inhibiting effector T cells. We and other authors have shown that the frequency of intratumoral Treg is increased in head and neck cancer patients and in HPV-associated cervical carcinoma (17, 18). PD-1, a member of the CD28 receptor family, is expressed by activated lymphocytes and inhibits their proliferation and effector functions after binding to PD-1 ligands such as B7-H1 (PD-L1; ref. 19). PD-1 expression is indicative of chronic antigen stimulation and contributes to T-cell dysfunction (exhaustion; ref. 19). Interference with PD-1–PD-L1 signaling, via either antibody blockade or PD-1 deficiency has been shown to improve clinical outcome and restore functional T-cell responses in chronic viral diseases and in several types of cancer (20). Preliminary results documented that more than 60% of freshly isolated human head and neck tumors express PD-L1 (21) and in early stage tongue cancer, intraepithelial T cells frequently express PD-1 (22). All these data, therefore, support a comprehensive analysis of PD-1–PD-L1 expression in head and neck cancers associated with chronic viral infection.

As expected, HPV-associated head and neck cancers were found to be associated with a good prognosis compared with HPV-negative tumors. HPV-positive tumors were more heavily infiltrated by regulatory T cells and PD-1 expressing T cells than HPV-negative tumors. Surprisingly, levels of PD-1⁺ T-cell infiltration positively correlated with a favorable clinical outcome in HPV-associated cancers and, in a preclinical HPV tumor model, predicted the clinical response to anti–PD-L1 antibody.

Sixty-four newly diagnosed untreated patients with primary histologically proven head and neck squamous cell carcinoma (HNSCC) were included in this study. Patient characteristics are presented in Supplementary Table S1. Each patient's disease was staged according to the 7th edition of the International Union Against Cancer/American Joint Committee on Cancer system for head and neck cancer. Treatment modalities consisted of chemoradiotherapy without surgery (organ preservation) or surgery combined or not with radiotherapy and chemotherapy. This study was conducted in accordance with French laws and after approval by the local ethics committee (ID RCB 2007-A01128-45).

Patients were divided in 2 groups depending on the presence or absence of oncogenic HPV. These 2 groups were matched for various parameters (gender, primary tumor site, tumor staging, lymph node involvement, presence of metastasis, and treatment modalities; Supplementary Table S1).

Six- to eight-week-old female C57BL/6 (H-2^b; B6) mice were purchased from Charles River Laboratories (L'Arbresle, France). All mice were kept under specific pathogen-free conditions at the INSERM U970 animal facility. Experiments were conducted according to institutional guidelines after acceptance by the Veterinary School of Maisons-Alfort ethics committee.

TC-1 cells are transformed murine (H2^b) epithelial cells cotransfected with HPV-16 E6/E7 genes and the activated human Ha-ras (G12V) oncogene DNA. They were obtained from Dr. TC Wu's laboratory (Department of Pathology, School of Medicine, Johns Hopkins University, Baltimore, USA). Cells were grown in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mmol/L L-Glutamine, antibiotics and 0.4 mg/mL G418 at 37°C with 5%CO₂.

Pool of 15 mer peptides spanning the entire HPV16 E7 protein were designed with a 4 mer overlap between each peptide. They were obtained from PolyPeptide Laboratories (Strasbourg, France), reconstituted in PBS and stored at −20°C.

The STxB-E7_43–57 vaccine was produced by chemical coupling between the N-bromoacetylated E7_43–57 peptide and the sulfhydryl group of STxB-Cys recombinant protein, as previously described (23). After purification, endotoxin concentrations determined by the Limulus assay test (Lonza) were less than 0.5 EU/mg.

The invariant natural killer T-cell ligand α-GalCer (KRN7000) was purchased from Funakoshi (Tokyo, Japan).

CT-011 is a humanized monoclonal anti-PD-1 antibody manufactured by CureTech LTD. This antibody has already been validated in ex vivo functional assays in humans (24).

Anti-mouse PD-L1 and isotype control antibodies were purchased from BioXcell. Mice received intraperitoneal injections of 200 μg mAb/mouse as previously described (25).

For immunofluorescence analysis, an anti-PD-1 mAb generated by D Olive's group was selected and used (see Supplementary Table S2) as previously described (26).

The presence of HPV in head and neck cancer biopsies was detected using the INNO-LIPA genotyping Extra assay (Innogenetics).

Immunofluorescence staining and flow cytometry analysis were conducted as previously described (16) and detailed in Supplementary Data. The antibodies used for the various immunofluorescence stainings are described in Supplementary Table S2.

C57BL/6 mice were immunized via the intraperitoneal route at days 0 and 14 with the STxB-E7 vaccine (20 μg) combined with 1 μg of αGalCer during the first immunization.

In a therapeutic setting experiment, 10⁵ TC-1 tumor cells were injected subcutaneously (s.c.) in the right flank of B6 mice. They were then vaccinated or not at days 9 and 18 after tumor graft in combination with anti-PD-L1 or isotype control antibodies. Mice were monitored every 48 to 72 hours for tumor growth.

Survival variables were estimated using the Kaplan–Meier method and compared by the log-rank test for categorical variables and the Cox model and associated Wald ×2 statistic for quantitative variables.

Overall survival was defined as the time from initial diagnosis until death or until last follow-up (right censored data).

Locoregional control was calculated from the end of treatment and defined as the absence or either persistent or recurrent disease at the primary site or in the cervical lymph nodes. Patients with persistent disease at the end of treatment were considered to have experienced failure at time 0. Patients with no signs of relapse were censored at the time of last-follow up or death. The median follow-up for the overall population was 26 months.

The Mann–Whitney U test was used to assess whether 2 samples of observations were derived from the same distribution. The χ² test with Yates correction was used to analyze the relationship between tumor cell infiltration and the tumor HPV status.

In a retrospective study, we divided 64 head and neck cancer patients into 2 groups differing according to presence or absence of oncogenic HPV (30 HPV16, 1 HPV18 and 1 HPV33, and 32 HPV-negative) in their tumors. These 2 groups were balanced for gender, primary tumor sites, tumor (T) staging, lymph node involvement, and treatment modalities (Supplementary Table S1). On the basis of Kaplan–Meier estimates, the overall survival for patients with HPV-positive cancer was significantly better (median survival not reached) than for patients with HPV-negative cancer (median survival: 16 months; P = 0.001; log rank test; Fig. 1). Patients with HPV-positive tumors also had a significantly better disease-free survival (P = 0.003; log rank test) than patients with HPV-negative cancer (Fig. 1).

HPV-positive head and neck cancers were more heavily infiltrated by CD8⁺ cells (P = 0.01), PD-1⁺CD4⁺ T cells (P = 0.045), and the total number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ T cells (P = 0.045). A trend was also observed between high levels of total PD-1⁺ cell (P = 0.079), PD-1⁺CD8⁺ cell (P = 0.078), and Foxp3⁺CD4⁺ T-cell (P = 0.059) infiltration and the HPV detection in the tumors (Fig. 2 and Table 1). PD-1 expression was higher in CD4⁺ T cells (median of PD-1⁺CD4⁺ cells = 5.5), than in CD8⁺ T cells (median of PD-1⁺CD8⁺ = 1).

As PD-1⁺CD4⁺ cells and PD-1⁺CD8⁺ cells could also represent macrophages and NK cells respectively, tumor dissociation was conducted for 14 tumors and cytometry analysis confirmed that more than 95% of PD-1⁺CD45⁺ cells were CD3⁺ T cells (Supplementary Fig. S1).

Thirty-three (51.5%) of the 64 head and neck tumors expressed significant levels of PD-L1 (score ++/+++) with homogeneous staining (Table 1 and Supplementary Fig. S2), but no correlation was observed between PD-L1 expression and HPV status of these head and neck cancer patients (Table 1).

Surprisingly, we showed that HPV-positive tumors, infiltrated by large numbers of PD-1–positive cells or total number of PD-1⁺CD4 and PD-1⁺CD8 T cells, were correlated with better overall survival of these patients. Indeed, patients with PD-1–positive tumor-infiltrating T cells above the median value (15 PD-1 cells/fields) had a 93.9% 60-month overall survival, whereas patients with low PD-1–positive cell infiltration had a 63.6% 60-month overall survival [P = 0.025; HR, 0.13; 95% confidence interval (CI), 0.02–0.067] (Fig. 3A). In line with these results and with the previous demonstration that most PD-1 cells are T cells (Supplementary Fig. S1), we showed that tumors infiltrated by high levels of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ cells had a better survival than tumors with low PD-1⁺ T-cell infiltration (P = 0.045; HR 0.15; 95% CI, 0.03–0.7). Patients with high infiltration (>median value) by PD-1⁺CD4⁺ T cells (P = 0.046) but not by PD-1⁺CD8⁺ T cells also had a better overall survival.

A correlation was also shown between the total number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ infiltrating T cells and 18-month overall survival in the whole population of head and neck patients whatever their HPV status (P = 0.009; Fig. 3B).

In contrast, the levels of infiltration by CD8⁺ T cells or CD4⁺ T cells or Foxp3 regulatory CD4⁺ T cells infiltration were not correlated with overall survival in HPV-positive head and neck tumors (Fig. 3A). However, as previously reported by us and other authors (17, 27), the number of Foxp3 regulatory T cells was correlated with better 18-month overall survival in the whole population of head and neck cancer patients (P = 0.0033; Supplementary Fig. S3).

PD-L1 expression was not correlated with overall survival or disease free survival in either the whole head and neck cancer cohort or in HPV-positive patients (Supplementary Fig. S2). In bivariate analysis including PD-L1 and various subpopulation of PD-1 cells each analyzed separately, we showed that only PD-1⁺CD4⁺ T cells, the total number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ T cells and the total number of PD-1 cells remained significant (data not shown).

PD-1⁺ T cells infiltrated both the stroma and the nest of tumor cells. As shown in Fig. 4A, close contact was shown between PD-L1 tumor cells and PD-1⁺ cells strongly suggesting that a negative signal was delivered to these PD-1⁺ tumor–infiltrating T cells. Interestingly, in the whole population, a correlation was observed between PD-L1 expression and the number of tumor infiltrating CD4⁺PD-1⁺ T cells (P = 0.029). A trend was also observed between PD-L1 expression and the number of PD-1⁺ CD4⁺ T cells and PD1⁺CD8⁺ T cells (P = 0.07) and also the total number of PD-1⁺ T cells (P = 0.089; Supplementary Fig. S4).

PD-1 is a hallmark of both activated and exhausted T cells (28). We found that PD-1 positive T cells, either CD4 or CD8⁺ T cells, expressed higher levels of HLA-DR and CD38 than PD-1 negative T cells (Fig. 4B and C). No difference in this activation state appeared to occur between HPV-positive or HPV-negative tumors (data not shown). Tim-3 expression was also assessed, as it has been associated with an exhausted phenotype of PD-1⁺ T cells (29, 30). One-half of PD-1⁺ T cells in head and neck cancer were found to express Tim-3 (Fig. 4D and Supplementary Fig. S5A). The same number of PD-1⁺ T cells in HPV-positive and HPV-negative tumors presented hallmarks of exhausted T cells (Fig. 4D).

We also showed that PD-1–positive cells did not express the Foxp3 marker of regulatory T cells, despite the fact that, in some cases, PD-1–positive cells and Foxp3 regulatory T cells were in close contact (Supplementary Fig. S5B).

To assess whether the functionality of these T cells could be enhanced, fresh TIL were cocultured with tumor cells in the presence or absence of anti-PD-1 mAb and a pool of E7 peptides. In 3 (2 HPV negative and 1 HPV positive) of 4 patients, a significant increase of IFNγ positive T cells was shown by Elispot in the presence of CT-011, an anti–PD-1 mAb (Fig. 5). An isotype control antibody had no effect on enhanced IFNγ production (Fig. 5).

In 1 HPV-positive patient, E7 peptides slightly increased the frequency of CT-011 induced IFNγ secretion (Fig. 5).

To explain this paradoxical good prognostic value of PD-1⁺ T-cell infiltration in HPV-associated head and neck cancer, we modelized this phenomenon in a preclinical model.

When mice were grafted with an epithelial tumor cell line (TC-1) expressing the E7 protein derived from HPV, tumors were poorly infiltrated by CD8⁺ T cells that expressed low levels of PD-1 (6%; Fig. 6A).

When mice were vaccinated with a nonreplicative delivery vector targeting dendritic cells, the B subunit of Shiga toxin, coupled to E7, 44% of CD8⁺ T cells expressed PD-1 (Fig. 6A and B). In addition, 89% of specific anti-E7_49–57 CD8⁺ T cells, induced by the vaccine, also expressed PD-1 (Fig. 6A and B). The vaccine therefore appeared to elicit PD-1 expression in both total and E7_49–57 specific CD8⁺ T cells. The vaccine alone had a partial inhibitory effect on the growth of established TC-1 tumors (Fig. 6D). This effect was significantly enhanced by the addition of anti-PD-L1 (Fig. 6D). In contrast, the anti-PD-L1 antibody alone had no effect on tumor growth (Fig. 6D), which may be explained by the low (6%) PD-1 expression in infiltrating T cells in the absence of vaccination (Fig. 6A). Similar results were obtained, when overall survival or the percentage of tumor-free mice were selected as endpoints (Supplementary Fig. S6).

This study shows that HPV-positive head and neck tumors were heavily infiltrated by PD-1⁺ cells, as the number of PD-1⁺CD4⁺ infiltrating T cells and the total number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ T cells were significantly higher in HPV-positive than in HPV-negative head and neck cancers. The relevance of these results is strengthened by the matching between HPV-positive and HPV-negative tumors for various confounding parameters (tumor–node–metastasis, treatment modalities, etc).

Surprisingly, we found that high levels of PD-1⁺ cells and the total number of PD-1⁺CD4⁺ and PD-1⁺CD8⁺ T cells correlated with better survival compared with low levels of infiltration by these cells in primary HPV-positive head and neck cancers.

These results were unexpected because of the known inhibitory function of PD-1 on T cells and immune cells (20, 31). In addition, in renal cell carcinoma, nasopharyngeal cancer and Hodgkin lymphoma, PD-1 expression on immune cells was correlated with poor prognosis and was associated with shorter survival (32–34). However, Hsu and colleagues showed that the prognostic value of PD-1 differed between CD4⁺ and CD8⁺ T cells (33). The other studies did not conduct double immunofluorescence staining or cytometry analysis to more precisely characterize PD-1 positive cells that can be expressed by many types of infiltrating cells (NK, B cells, macrophages, etc.). In the present study, we showed that most (>95%) PD-1 positive cells corresponded to T cells (Supplementary Fig. S1) and that the prognostic value of PD-1 was associated with the total number of CD4⁺ and CD8⁺ T cells expressing this marker. It should also be mentioned that in contrast to many tumors, CD8⁺ T cells are associated with bad prognosis in renal cell carcinoma and hodgkin lymphoma supporting the fact that the clinical significance of molecules expressed by these cells could also vary depending on the tumor histology (35, 36).

Interestingly, recent studies have reported that follicular lymphomas infiltrated by PD-1⁺ T cells were associated with a lower risk of transformation, a higher progression-free survival and a better overall survival than tumors not infiltrated by PD-1⁺ cells (37). PD-1 was preferentially expressed by infiltrating CD4⁺ T cells in these follicular lymphomas, as also observed in our study. In line with our results, PD-1 mRNA expression determined by reverse transcriptase-PCR (RT-PCR) in colorectal carcinoma was also associated with a good prognosis (38). It is therefore noteworthy that the 3 tumors (lymphoma, colorectal cancer, and head and neck cancer), in which PD-1 expression was significantly associated with a favorable outcome (37, 38), were also those with an unexpected clinical benefit of Treg infiltration (17, 39, 40). The present study, conducted in an independent series of patients, confirms our previous findings that high levels of Treg in head and neck cancer patients were also positively correlated with a better prognosis (17). These results concerning the good prognostic value of intratumor regulatory T cells in head and neck cancer have subsequently been reproduced by various groups (27, 41). In addition, the frequency of regulatory Foxp3⁺CD4⁺ T cells has been shown to be higher in patients with HNSCC with no evidence of disease after oncologic therapy, than in patients with active disease (42). The plasticity of regulatory T cells, transient expression of Foxp3 by activated T cells and a possible role of regulatory T cells in the inhibition of deleterious cancer inflammation could explain the paradoxically good clinical prognosis associated with regulatory T cells in some circumstances (43, 44).

Several hypotheses can also be proposed to explain this paradoxical favorable outcome associated with PD-1 expression on T cells. PD-1 is upregulated on lymphocytes following T-cell receptor activation and remains elevated in the context of persistent antigen-specific immune stimulation (45). The absence of co-expression of Tim-3 and PD-1 in one-half of PD-1–positive cells strongly suggests that some PD-1⁺ T cells in the tumor microenvironment of HPV-positive head and neck cancers are not exhausted T cells (Fig. 4). We did not assess the expression of other inhibitory receptors (Lag-3, CTLA-4…) on PD-1⁺ T cells, but Tim-3 is the most frequent coinhibitory receptors associated with PD-1⁺ T cells in tumor and an hallmark of exhausted T cells (29, 46). It has been reported that high PD-1 levels on T cells more closely reflect their “exhausted status” than intermediate PD-1 levels. However, in our study, PD-1 expression was homogeneous and it was, therefore, difficult to divide T cells into 2 groups. Some of our PD-1 infiltrating T cells remained responsive to anti-PD-1 in vitro (Fig. 5). From these results, PD-1 expression also indicates, that T-cell activation has occurred during the development of HPV-associated head and neck cancers and PD-1 could, therefore, also be viewed as a witness of the antitumor immune response.

Indeed, we showed that PD-1–positive T cells expressed higher levels of the HLA-DR and CD38 activation markers, than PD-1–negative T cells (Fig. 4). Although this response was not effective to clear the tumor, it may have slowed tumor growth compared with a clinical setting associated with absence of host response to the tumor. Although the tumor specificity of PD-1 expressing T cells could not be shown because of the small number of cells present in tumor biopsies, it has been previously shown that tumor-reactive CD8⁺ T cells generated after in vitro culture of dissociated biopsies of melanoma were more frequently derived from PD-1 expressing T cells present at the beginning of culture (47). In preclinical tumor models, the CD137⁺CD8⁺ T cells were enriched in antitumor effector T cells (48).

Our preclinical model showed that grafting of the E7-expressing TC-1 epithelial cell line, which constitutively expresses PD-L1 led to tumor growth without therapy. Analysis of its tumor microenvironment showed the absence of PD-1 expression by CD8⁺T cells. In contrast, activation of an anti-HPV response by an anti-E7 vaccine, elicited PD-1 expression by CD8⁺T cells and E7-specific T cells, which were associated with partial regression of the tumor. In humans, it has been shown that TIL derived from spontaneously regressing melanomas also express substantial levels of PD-1 (49).

Another important result of this study is that the efficacy of anti-PD-L1 mAb depended on the presence of PD-1 on specific T cells induced after vaccination. In preclinical models, immunotherapy approaches that should lead to upregulation of PD-1 following activation of the immune system, such as recombinant cytokines or stimulatory antibodies or transfer of activated adoptive T cells, synergized with blockade of the PD-1–PD-L1 pathway (21, 25, 50).

Consistent with these results, a recent report showed that expression of PD-L1, the ligand of PD-1, which was considered to be an immunosuppressive molecule, positively correlated with better survival in melanoma (51). In melanoma, PD-L1 is mainly induced by IFNγ produced by infiltrating T cells and the distribution of T cells within the tumor was correlated with that of PD-L1 (51). In line with these results, in patients treated with anti-PD-1 mAb, tumor cell surface PD-L1 expression appeared to be correlated with the likelihood of response to treatment, possibly because of a reprogramming of infiltrating antitumor PD-1⁺T cells (52). PD-1 and PD-L1 expression may, therefore, represent surrogate markers of endogenous antitumor immune response, explaining their unexpected association with good prognosis in some tumors in which PD-L1 expression is not oncogene driven. Indeed, in glioma, loss of tumor suppressor PTEN function increased PD-L1 (53). In our study, PD-L1 was not associated with a good prognosis in head and neck cancer patients, regardless of HPV status, but low PTEN expression was detected in the majority of these tumors (54).

This study, therefore, revisits the significance of PD-1–positive tumor-infiltrating T cells in cancer. The inhibitory function linked to PD-1 should not mask the fact that its detection could also reflect a past antitumor immune response ready to be reprogrammed by PD-1–PD-L1 blockade.

W.H. Fridman is a consultant/advisory board member of CureTech LTD. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Badoual, S. Hans, E. Tartour

Development of methodology: C. Badoual, N. Merillon, C. Van Ryswick, N. Benhamouda, E. Levionnois, A. Si-Mohamed, N. Besnier, A. Gey, H. Pere, C.L. Guerin, A. Chauvat, C. Alanio, D. Olive, D. Brasnu, E. Tartour

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Badoual, S. Hans, N. Merillon, N. Benhamouda, E. Levionnois, M. Nizard, A. Si-Mohamed, N. Besnier, A. Gey, H. Pere, A. Chauvat, E. Dransart, C. Alanio, S. Albert, B. Barry, F. Sandoval, F. Quintin-Colonna, P. Bruneval, W.H. Fridman, F.M. Lemoine, S. Oudard, L. Johannes, D. Olive, D. Brasnu, E. Tartour

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Badoual, S. Hans, N. Merillon, C. Van Ryswick, P. Ravel, N. Benhamouda, E. Levionnois, A. Si-Mohamed, A. Gey, R. Rotem-Yehudar, H. Pere, C.L. Guerin, A. Chauvat, C. Alanio, F. Sandoval, F. Quintin-Colonna, P. Bruneval, W.H. Fridman, F.M. Lemoine, S. Oudard, L. Johannes, E. Tartour

Writing, review, and/or revision of the manuscript: C. Badoual, S. Hans, R. Rotem-Yehudar, F. Sandoval, D. Olive, D. Brasnu, E. Tartour

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Badoual, S. Hans, E. Tartour

Study supervision: C. Badoual, E. Tartour

This work was supported by grants from the Ligue contre le Cancer, the Agence Nationale de la Recherche (ANR), the Association pour la Recherche contre le Cancer (ARC), Labex, Immuno-Oncology, Canceropole Region Ile de France, Fonds d'amorçage AP-HP, and Institut National du Cancer (INCA).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.