Intratumoral HPV16-Specific T Cells Constitute a Type I–Oriented Tumor Microenvironment to Improve Survival in HPV16-Driven Oropharyngeal Cancer

This article is featured in Highlights of This Issue, p. 503

The incidence of oropharyngeal squamous cell cancer (OPSCC) is rising, especially in younger adults (1). Classically, the development of OPSCC is related to p53 mutations, but currently, more than half of all OPSCCs are caused by a high-risk human papillomavirus (HPV), most often type 16 (HPV16; ref. 1). Although HPV-associated OPSCCs are more often diagnosed with TNM stage III to IV, consisting of an earlier T stage and more advanced N stage, than HPV-negative OPSCCs (2), they display a much better prognosis than HPV-negative tumors after (chemo)radiotherapy. This is independent of many common histopathologic parameters (2, 3) but associated with the presence of a strong adaptive immune response gene signature (4) and dense tumor infiltration by activated CD4⁺ and CD8⁺ T cells (3, 5, 6), suggesting a role for the adaptive immune system in the response to therapy. Notably, HPV-associated OPSCCs express viral proteins, and we have shown that they may function as tumor-specific antigens for OPSCC-infiltrating T cells (7). Clear evidence for a protective role of tumor-infiltrating, HPV-specific T-cells in OPSCC, however, is lacking. Hence, it is important to evaluate whether HPV-positive OPSCCs are commonly infiltrated by HPV-specific T cells and, specifically, how this pertains to the composition of the tumor microenvironment and survival. We focused purely on the analysis of HPV-specific T-cell reactivity within the tumor-infiltrating lymphocyte (TIL) population, as detection of circulating HPV-specific T cells might reflect a response to past infections (8), potentially even in other anatomic locations (8) and, thus, was less relevant to our study. In case of such a relation, reinforcement of HPV-specific T-cell reactivity becomes highly attractive for treatment of OPSCC.

Patients with histology-confirmed OPSCC were included after signing informed consent. This study is part of a larger observational study, P07-112 (7), approved by the local medical ethical committee of the Leiden University Medical Center (LUMC; Leiden, the Netherlands) and in agreement with Dutch law. Patient enrollment was from November 2007 to November 2015. Blood and tumor tissue samples were taken prior to treatment and handled as described previously (9) and in Supplementary Methods. Peripheral blood mononuclear cells (PBMC) and TILs were stored until use. HPV typing and p16^ink4a IHC staining was performed on formalin-fixed paraffin-embedded (FFPE) tumor sections at the LUMC Department of Pathology. Immunofluorescent staining of FFPE tumor sections for CD8 and Tbet was performed as described previously (10) and in Supplementary Methods. The patients received the standard-of-care treatment, which could consist of surgery, radiotherapy, chemotherapy, treatment with mAb, or combinations hereof. Staging of the tumor was done according to the National Comprehensive Cancer Network (https://www.nccn.org/professionals). Patient characteristics are provided in Supplementary Table S1.

The OPSCC cell lines were obtained from the University of Michigan (Ann Arbor, MI) and called UM-SCC. We obtained UM-SCC4 (passage 22), UMC-SCC6 (passage 33), UM-SCC19 (passage 17; all three HPV negative), UM-SCC47 (passage 98), and UM-SCC104 (passage 15; both HPV16 positive) in 2012. The cells were cultured in RPMI 1640 (Gibco/Thermo Fisher Scientific) with 10% FCS (PAA Laboratories) and penicillin/streptomycin (Thermo Fisher Scientific). Tumor cell supernatant (TSN) was prepared after 5 days of culture as described previously (11). Microsatellite analysis was performed in July 2016 by BaseClear to assure cell line authentication when the experiments were performed. Mycoplasma was tested on a monthly basis.

Clonal dilution was performed using the TILs from patient H68 as described previously (7). Their HPV specificity and cytokine production were determined. This resulted in multiple CD4⁺ Th cell clones of which Th1 (clones 78 and 97), Th2 (clone 133), and Th17 (clones 12 and 103) were selected for the experiments. T-cell supernatant was obtained after stimulation with cognate HPV peptide loaded on with Epstein–Barr virus (EBV)–immortalized autologous B cells for 3 days.

The phenotype and composition of dispersed tumors (and expanded TILs) were analyzed by flow (9, 12–15) and time-of-flight mass cytometry (CyTOF; ref. 16; Supplementary Methods). Supplementary Table S2 shows the 36 markers used for CyTOF analysis. The reactivity of TILs was determined in a 5-day proliferation assay (9) and by intracellular cytokine staining (15). Supernatant from the proliferation test was subjected to cytokine analysis (15). The effect of TSN on dendritic cell (DC) differentiation was determined phenotypically and functionally (cytokine/chemokine production) upon LPS or agonistic CD40 antibody stimulation in the presence or absence of INFγ as described previously (11, 13) and in Supplementary Methods.

Tumor cells were seeded (15,000–27,500 cells/well) in a flat-bottom, 96-well plate (Costar/Thermo Fisher Scientific) and allowed to adhere overnight at 37°C. The next day, the cells were incubated with the indicated concentrations of IFNγ and/or TNFα for 48 hours at 37°C, followed by the MTT assay (Trevigen) according to the manufacturer's protocol to determine the percentage of proliferating cells compared with the untreated cells (set at 100%; ref. 13). Tumor cells (70,000–100,000) were adhered in a 24-well plate overnight as described above, followed by treatment for 24 hours with a fixed dose of cisplatin (15 μg/mL) in the presence or absence of indicated concentrations TNFα (0–30 ng/mL). The cells were harvested and analyzed for apoptosis by flow cytometry. In another experiment, tumor cells prepared in 24-well plates were treated for 24 and 48 hours with IFNγ (250 IU/mL; ImmunoTools) and TNFα (30 ng/mL) or 20% of supernatant obtained from Th1 (H68 clone 97), Th2 (H68 clone 133), or Th17 (H68 clone 103) cells with or without the addition of apoptosis inducer and cIAP1/2-interacting compound BV6 (5 μmol/L smac mimetic; ApexBio) and pan-caspase inhibitor zVADfmk (20 μmol/L FMK001; R&D Systems), together known to induce necroptosis (17–19). Necrostatin (Nec)-1s (2263-1; BioVision) was added to the conditions used for UM-SCC19 to inhibit necroptosis via inhibition of RIP1K (14). The treated tumor cells were harvested and subjected to SYTOX green staining to establish the percentage of dead cells and, in parallel, stained for flow-based apoptosis analysis using Annexin V (early apoptosis) and 7-AAD (late apoptosis). As indicated, tumor cells were also analyzed for RNA expression (quantitative PCR; ref. 14) and protein content (Western blot analysis; ref. 14; see also Supplementary Methods).

Unpaired parametric t test was used to determine the difference between various treatments of the cells from the UM-SCC tumor cell lines. Dates of two groups of patients were analyzed using the unpaired nonparametric analysis (Mann–Whitney). Fisher exact test was used to analyze categorical data in a contingency table. Data of the three groups of patients (p16⁻IR⁻, p16⁺IR⁻, and p16⁺IR⁺) were analyzed using the unpaired nonparametric one-way ANOVA (Kruskal–Wallis). HR with a 95% confidence interval (CI) was calculated to determine the difference in survival curves. The nonparametric log-rank test (Mantel–Cox test) was done to compare the survival distribution of the two groups of patients. In all cases, a P value of 0.05 and below was considered statistically significant (*), with P < 0.01 (**) and P < 0.001 (***) as highly significant.

To interrogate the role of HPV-specific T cells in OPSCC, we prospectively assembled a cohort of 97 patients with OPSCC, 57 of whom were HPV16 positive. Analysis of the patient characteristics showed the expected percentage of HPV-positive patients (2, 3) and the differences in smoking, N stage, and disease-specific survival when compared with HPV-negative OPSCC (Fig. 1A; Supplementary Table S1), indicating that our patient cohort does not differ from those reported in literature.

From each patient, freshly obtained and FFPE tumor material was stored (Supplementary Fig. S1). The presence, proliferation, and cytokine production of HPV16-specific and other OPSCC-infiltrating T cells in the dissociated OPSCC were analyzed either directly or following a 2- to 4-week expansion period (Supplementary Fig. S1). Reactivity to the HPV16 E6 and/or E7 oncoproteins was detected directly ex vivo in 6 of 24 samples, and in 29 of 45 of the expanded TIL HPV16-positive cases. All directly ex vivo detectable responses were confirmed in the expanded TILs. None of the 23 tested TIL cultures obtained from HPV-negative tumors displayed HPV-specific reactivity (Fig. 1B and C), showing the specificity of these types of TIL analyses (7) and demonstrating that HPV-specific T cells infiltrate only HPV-positive OPSCC.

Subsequently, supernatants taken from the HPV-reactive cultures were assessed for the presence of Th1 (IFNγ, TNFα, and IL2), Th2 (IL4, IL5, and IL10), and Th17 (IFNγ and IL17) cytokines, revealing a Th1/Th17-like profile (Fig. 1D). Flow cytometry analysis demonstrated that the population of activated and/or cytokine-producing, HPV-specific T cells frequently comprised both CD4⁺ and CD8⁺ HPV-specific T cells (Fig. 1E; Supplementary Fig. S2), which targeted multiple epitopes simultaneously (Fig. 1F), albeit the percentage of HPV-specific, cytokine-producing CD4⁺ T cells often was higher than that of CD8⁺ T cells (Fig. 1G). Thus, the majority of HPV16-positive OPSCC tumors are infiltrated by HPV16-specific CD4⁺ and CD8⁺ T cells with a Th1/Th17 profile.

The mechanisms underlying the failure to detect an intratumoral HPV16-specific response can be manifold, but the first requirement is the presence of sufficient quantities of antigen to stimulate T cells. The expression of p16^INK4a is a surrogate marker for overexpressed, functionally active E7 oncoprotein (20). Forty of the TIL-tested HPV16-positive OPSCC tumors could be analyzed for p16^INK4a overexpression, and, in contrast to immune responders (IR⁺), 7 of the 15 immune nonresponders (IR⁻) failed to show a positive staining (Supplementary Fig. S3A). Furthermore, tobacco smoking and, in particular, nicotine are known to impair the responsiveness of T cells to antigenic stimulation (21). Although many patients had more than 10 pack years of smoking (Supplementary Fig. S3B; ref. 2), this was not discriminative for the detection of HPV16-specific immunity (Supplementary Fig. S3C). Hence, the failure to produce a T-cell reaction to HPV in HPV16-positive OPSCC most likely is due to the limited quantities of viral proteins available to the immune system.

On the basis of the observation that the major component of OPSCC-infiltrating, HPV-specific T cells consists of CD4⁺ T cells, and the known activity of tumor-specific CD4⁺ T cells to recruit, activate, and sustain other immune cells (22, 23), we performed an in-depth analysis of the tumor microenvironment in the context of HPV-specific T-cell reactivity. As the absence of overexpressed p16^INK4a in HPV16-positive OPSCC may indicate that their development was not driven by the HPV oncoproteins (24), we separated the HPV16-positive patients into three groups: p16^INK4a-negative, IR-negative (p16⁻IR⁻); p16^INK4a-positive, IR-negative (p16⁺IR⁻); and p16^INK4a-positive, IR-positive (p16⁺IR⁺) patients.

An understanding of the general cytokine polarization in the tumors was obtained through analysis of cytokine production following the direct ex vivo activation of all tumor-infiltrating T cells using the mitogen phytohemagglutinin. Interestingly, the IFNγ/IL17 cytokine polarization of HPV-specific T cells was mirrored in the remainder of tumor-infiltrating cells (Supplementary Fig. S4). The production of IFNγ and IL17 was lower in the p16⁺IR⁻ and the p16⁻IR⁻ groups. Moreover, the production of IL5 was increased in the latter two groups, suggesting a shift toward a more type II cytokine profile.

In addition, we quantified the number of type I polarized immune cells in the HPV16-positive tumors using IHC for CD8 and then with IFNγ production–associated T-box transcription factor TBX21 (Tbet). The numbers of tumor-infiltrating Tbet⁺CD8⁺ T cells and Tbet⁺CD8⁻ T cells, based on our flow cytometry data, most likely CD4⁺ T cells, correlated with an improved survival (Fig. 2A) and were particularly high when the OPSCC contained HPV-specific T cells (Fig. 2B).

To comprehensively analyze the composition and phenotype of intratumoral immune cells directly ex vivo, a validated panel of 36 antibodies adapted from a previous study (Supplementary Table S2; ref. 16) was used in combination with mass cytometry (CyTOF) in 13 freshly dissociated OPSCCs. This showed that the HPV16-positive OPSCCs from HPV16 immune responder patients were strongly infiltrated with CD4⁺ and CD8⁺ T cells (Fig. 2C and D) carrying an effector memory phenotype (Fig. 2E), whereas the HPV16-positive OPSCCs in which no HPV16-specific T-cell reactivity was detected displayed a strong influx with B cells (Fig. 2D). Natural killer (NK) cells, which may also infiltrate tumors and express Tbet, were virtually absent (Fig. 2D). To automatically discover stratifying biological signatures, we used the CITRUS algorithm with an FDR of 1%, resulting in five distinctive (groups of) populations of immune cells (Fig. 2F). It confirmed the differences in the percentages of tumor-infiltrating B cells and T cells (Fig. 2G) but also revealed the presence of three subsets of T cells that were present at significantly higher levels in HPV16 immune responders (Fig. 2H). Inspection of these subsets revealed two subsets of activated CD4⁺ T cells and a subset of tissue-resident effector memory CD8⁺ T cells expressing CD103 (Supplementary Fig. S5A and S5B). The two subsets of activated CD4⁺ T cells expressed CD38, HLA-DR, and PD-1 but were separated on the basis of CD161 expression (Supplementary Fig. S5A and S5B). The CD161⁻ subset of activated CD4⁺ T cells had a high expression of CD25 but also expressed CD127, whereas the CD161⁺ subset displayed an intermediate expression of CD25, making it unlikely that these two populations reflected regulatory T cells. Comparison of the tSNE plots of each patient clearly showed the almost exclusive presence of CD103⁺CD8⁺ T cells in the IR⁺ patient group (Supplementary Fig. S5C and S5D). Interestingly, part of the CD103⁺CD8⁺ T cells also expressed CD161. There was no difference between the different patient groups with respect to the percentage of central memory CD161⁺CD4⁺ T cells, but in each of the patients with an IR⁺ HPV16-positive OPSCC, a clearly visible effector memory CD161⁺CD4⁺ T-cell population was present (Supplementary Fig. S5C and S5D).

In parallel, we analyzed the tumor microenvironment in a cohort of 75 patients with HPV16-positive OPSCC present in the publicly available The Cancer Genome Atlas (TCGA) database (25) using our previously published analytic strategy to estimate subpopulations of tumor-infiltrating immune cells (26). As CD4⁺ T cells formed the major component in IR⁺ patients, a gene set enrichment analysis of the TCGA RNA-sequencing data was performed to determine which immune cells were relatively enriched or depleted in HPV16-positive OPSCC with a high versus low CD4 gene expression (Fig. 3A). The results confirmed the enrichment of activated and effector memory T cells but also pointed at a potential enrichment in NK cells and activated DC and B cells as well as a decreased presence of MDSCs in tumors with a high CD4 expression. Notably, an increased percentage of DCs/DC-like macrophages was observed among the HPV responders when the dissociated HPV16-positive OPSCCs of our cohort were analyzed by flow cytometry (n = 18) or CyTOF (n = 13; Fig. 3B and C). In vitro experiments suggest that the increased percentages of these antigen-presenting cells (APCs) are caused by the presence of the intratumoral, IFNγ-producing HPV-specific T cells. Analysis of the impact of two different HPV16-positive head and neck squamous cell carcinoma (HNSCC) cell lines (27, 28) on GM-CSF + IL4-driven differentiation of monocytes to IL12p70-producing DCs showed that tumor-secreted compounds skewed the monocytes toward type II–like macrophages instead (Fig. 3D), which had a low capacity to produce IL12p70 after CD40 ligation unless IFNγ was present (Fig. 3E). The resulting APCs also produced the T-cell–attracting chemokines CXCL9 and CXCL10 (Fig. 3F). Replacing IFNγ by the supernatant of genuine activated HPV-specific Th1 or Th17 T-cell clones (Supplementary Fig. S6A) also neutralized the M2-like macrophage skewing effect of the tumor cells (Supplementary Fig. S6B). A similar effect of HPV-specific Th1 and Th17 cytokines was observed on the direct M2-macrophage skewing effect of tumor cells (Supplementary Fig. S6C). In addition, the costimulatory molecules were upregulated.

Thus, the infiltration of OPSCC by HPV16-specific Th1/Th17 cells is associated with the presence of a highly active tumor microenvironment consisting of a dense type I–oriented immune cell infiltrate, known to favor immune-mediated control of cancer cells (29).

The OPSCC-infiltrating, HPV-specific CD4⁺ T cells produced IFNγ and TNFα, known to drive tumor cell senescence (30) and to synergize with platinum-based therapy to kill tumor cells (31). We, therefore, studied whether similar mechanisms could play a role in controlling oropharyngeal cancer cell growth by HPV-specific CD4 T cells in vitro. We used our collection of three HPV-negative and two HPV16-positive HNSCC cell lines to analyze the expression of proteins involved in proliferation, apoptosis, and necroptosis following stimulation with IFNγ and TNFα. All cell lines expressed IFNGR and TNFR1 (and were responsive to IFNγ, evidenced by the phosphorylation of STAT1, and to TNFα, evidenced by RelA phosphorylation; Supplementary Fig. S7A–S7C). Furthermore, they expressed the proteins required for apoptosis and necroptosis, although the HPV16-positive tumor cells lacked expression of the necroptosis-essential protein RIPK3 (Fig. 4A). Stimulation of the tumor cells with IFNγ and/or TNFα or culture supernatant from antigen-stimulated HPV-specific Th1 or Th17 cells revealed a reduction in their proliferation (Fig. 4B and C) and an increase in the expression of the IFNγ-responsive genes IFITM1 and RARRES. Both genes are known to stop the proliferative process in cells (refs. 32, 33; Fig. 4D and E), albeit these effects differed per cell line tested. Expression analysis of the relation between IFNγ, IFITM1, and RARRES in the TCGA cohort of HPV16-positive patients showed that especially IFNγ and IFITM1 were coexpressed (r = 0.475; P = 0.00060), suggesting that IFNγ-induced arrest in proliferation occurs in vivo. In line with the RIPK3 expression, only the HPV-negative cell lines were sensitive to necroptosis (Fig. 4F). As cisplatin is the chemotherapeutic compound of choice for the treatment of OPSCC, the induction of cell death by increasing doses of TNFα in the presence of cisplatin was tested. The combination of TNFα and cisplatin resulted in an increased percentage of apoptotic tumor cells at 24 hours, specifically in the HPV-positive cell lines, as no synergistic effect was observed in the HPV-negative cell lines (Fig. 4G), and resulted in a high percentage of dead tumor cells at 48 hours (Supplementary Fig. S7D). These effects did not depend on necroptosis, as inhibition with necrostatin-1s did not prevent cell death (Supplementary Fig. S7D). Thus, apart from their role in changing the microenvironment, IFNγ and TNFα may also synergize with standard therapy in controlling tumor cell growth and form one of the underlying mechanisms explaining the good response rate of HPV-responding patients to (chemo)radiotherapy (2, 3).

CD161⁺CD4⁺ T cells are the dominant subtype of T cells present in inflammatory diseases where CD4⁺ T cells have an important role to drive acute inflammatory processes (34). Hence, a similar role may be expected in the rejection of cancer cells. First, CD161 expression among freshly and in vitro expanded TILs was analyzed. A large proportion of our fresh and in vitro expanded TILs expressed CD161. Importantly, in vitro expansion did not induce CD161 expression (Supplementary Fig. S8A). Subsequently, a flow cytometric analysis of eight in vitro expanded TILs was performed to assess the HPV-specific component among these cells. On average, the percentage of CD161⁺CD4⁺ T cells was 29% (Fig. 5A). The number of HPV-specific T cells producing TNFα (Fig. 5B) was a bit higher than those producing IFNγ (Supplementary Fig. S8B), and, on average, 31% of the HPV-specific CD4⁺ T cells expressed CD161 (Fig. 5B). This indicates that there was a sizeable CD161⁺ T-cell fraction among HPV-specific CD4⁺ T cells in most of the patients and also that the distribution of CD161⁺ cells among these HPV-specific T cells is similar to that of the total population.

Subsequently, we analyzed the survival of the 75 patients with HPV16-positive OPSCC in the publicly available TCGA database focusing on the expression of CD4, CD8, CD103, and CD161. A high expression of CD4, CD8, or CD161 was associated with better overall survival, but this was not the case for CD103 expression (Fig. 5C–F), albeit the combined high expression of CD103 with CD8 resulted in a better segregation of the survival curves (Fig. 5G). This fits with the observation that the expression of CD8 and CD103 was not strongly correlated (r = 0.2559; P = 0.0267) within this cohort. A high expression of CD161 with either high CD4 or CD8 expression was also associated with better survival (Fig. 5H and I). Notably, the populations of patients within the group seem to overlap completely. Indeed, these markers were highly coexpressed (CD4 and CD161: r = 0.8351, P = 0.00E00; and CD8 and CD161: r = 0.8363, P = 0.00E00), suggesting that they predominantly single out the same patients. As the HPV-specific T cells predominantly produced IFNγ, TNFα, and IL17 (Fig. 1D), we also analyzed the contribution of the respective gene expression levels to survival. Specifically, a high expression of IFNγ was associated with better survival, whereas a similar trend was visible for IL17 (Fig. 5J and K). Combinations of two to three cytokines did not result in better separation of the survival curves (Supplementary Fig. S8C–S8G).

In combination with the above, these data suggest that a dense infiltration of HPV16-positive OPSSC with IFNγ/IL17-oriented CD4⁺ and/or CD8⁺CD161⁺ T cells, including the HPV16-specific T cells, is important for superior disease control in HPV-driven OPSCC. Therefore, we analyzed the disease-specific survival of HPV-specific T-cell responders within the group of patients with HPV16-positive OPSCC. Patients with HPV-positive OPSCC displaying an HPV-specific T-cell reaction had a 37.8-fold (95% CI, 7.1–199.9) higher chance to respond to therapy when compared with patients with HPV16-positive OPSCC lacking such a T-cell reaction (Fig. 6A). Especially in stage III to IV HPV16-positive OPSCC, the local presence of an HPV16-specific T-cell response was a better prognostic parameter for long survival after therapy than staging (Fig. 6B). The differences in survival between these two groups could not be attributed to a different cancer treatment (Supplementary Table S3). Intriguingly, also the T and N stage were, on average, lower in the immune responders (Fig. 6C and D), suggesting that HPV16-specific T cells were especially present in patients with a better control of tumor growth.

The improved clinical response of patients with OPSCC to (chemo)radiotherapy has been associated with HPV and with a dense, activated T-cell infiltrate, but the role of the immune response against HPV in this was still not completely understood. Our findings demonstrate that the virally derived E6 and E7 antigens make HPV-associated OPSCC highly visible to the immune system and unleash an intratumoral, HPV-specific T-cell response. These cells are polyfunctional, detected among TILs in many of the patients, and have the CD161⁺ phenotype often found in acute rejection processes. They may locally facilitate the development of a clinically favorable tumor microenvironment because their presence is associated with a stronger influx of type I–oriented CD4⁺ and CD8⁺ T cells, as well as DCs and DC-like macrophages. Moreover, they produce cytokines that synergize with the platinum-based chemotherapy used to treat these patients, and their detection is highly predictive for the response of patients to (chemo)radiotherapy.

HPV-specific CD4⁺ and CD8⁺ T cells were detected in 64% of the TILs derived from HPV16-positive OPSCC, with a predominance of HPV-specific CD4⁺ T cells—a result that closely matches an earlier study (35). We show that these HPV-specific tumor-infiltrating T cells as well as the other TILs predominantly produced IFNγ and IL17, suggesting the presence of Th1 and Th17 cells. In view of the accepted roles of Th1/Th17 CD4⁺ and CD8⁺ T cells in tumor control (36, 37), the detection of these cells in HPV16-positive OPSCC is likely to favor tumor control. Indeed, a high expression of IFNγ and, to a lesser extent, IL17 in HPV16-positive OPSCC was associated with superior survival. Furthermore, the detection of HPV-specific T cells singled out immunologically “hot” tumors, with higher numbers of CD4⁺ and CD8⁺ T cells expressing Tbet, effector memory T cells, DCs, and DC-like macrophages when compared with HPV16-positive OPSCC without HPV-specific T cells. A dense tumor infiltration by T cells (38) and DCs (39) and a predominant adaptive immune gene signature (4) have been associated with better survival in head and neck cancer, indicating that HPV-specific T-cell–infiltrated tumors possess the right type of inflammation. Last but not least, a dense infiltrate with T cells is found more often in patients with superior local disease control (40), fitting with our observation that the group of patients with a tumor-specific immune response presented with lower T and N stages.

Concomitant with the detection of HPV16-specific TILs, we found increased frequencies of CD161⁺ effector memory CD4⁺ and CD8⁺ TILs as well as CD8⁺CD103⁺ TILs. The intratumoral presence of CD8⁺CD103⁺ T cells is a beneficial prognostic factor in a number of cancer types (41), and this would fit with the fact that we detected a high frequency of these cells specifically in T-cell inflamed tumors as well as with our analysis of the TCGA database, showing a survival advantage for patients with HPV16-positive OPSCC with a strong expression of both CD8 and CD103. Earlier reports showed that CD161⁺ is predominantly detected on effector and central memory T cells that produce IFNγ and/or TNFα (42), Th17 cells (43), and regulatory T cells (44). CD4⁺CD161⁺ T cells can drive acute inflammatory processes (34), suggesting an important and similar role for them in cancer. Indeed, CD161 was among the top 10 of tumor leukocyte-associated genes associated with positive prognosis for many human tumors (45). In our study, CD161 was expressed by tumor-specific IFNγ- and/or TNFα-producing CD4⁺ T cells; higher frequencies of CD161 expressing CD4⁺ T cells were detected in T-cell inflamed tumors; and, finally, in the TCGA database, the expression between CD161 and CD4 or CD8 was highly correlated and a high expression of these three genes was associated with a survival advantage for patients with HPV16-positive OPSCC. Interestingly, mass cytometry showed that part of the CD8⁺CD103⁺ T cells also expressed CD161.

In a large meta-analysis in head and neck cancer (MACH-NC) patients treated with radiotherapy alone had an overall 5-year survival of 27.2%, whereas in patients receiving concomitant cisplatin chemotherapy and radiotherapy, an improvement in overall survival of 6.5% was achieved (46). Potentially, this is explained by studies showing that platinum-based chemotherapy synergizes with immune cell–produced IFNγ and TNFα in killing tumor cells (31), including OPSCC cells (this study). Because of the described cisplatin toxic side effects, dose reductions in cisplatin of 30% to 69% are often required for sustained concurrent (chemo)radiotherapy treatment (47, 48), and deintensification protocols for these patients are being discussed. This should not pose a major problem, as lower doses of cisplatin still synergize with T-cell responses in animal tumor models (31).

Finally, the question surfaces whether reinforcement of HPV16-specific T-cell reactivity in patients with HPV16-positive OPSCC is warranted, not only to convert nonresponders to HPV responders but also to boost existing responses. Clearly, the HPV16-positive OPSCCs infiltrated by HPV16-specific T cells meet the criteria of the cancer immunogram for immunotherapy (49). The percentages of HPV-specific T cells among TILs are respectable, however, not enough to mediate full tumor regression. In parallel to melanoma, where treatment with increased numbers of tumor-specific T cells can mediate clinical responses, therapeutic vaccination is expected to increase the number of HPV16-specific T cells and may result in clinical benefit for patients with OPSCC. In view of the expression of PD-1 by the effector memory CD4⁺ and CD8⁺ T cells (this study and ref. 6) and of PD-L1 (50) in tumor tissue, a combination of therapeutic vaccination and PD-1/PD-L1 blocking is expected to have the best outcome.

S.H. van der Burg reports receiving other commercial research support from ISA Pharmaceuticals and Kite Pharma EU, is listed as a co-inventor on a provisional patent application on the use of synthetic long peptide vaccines for therapeutic vaccination against chronic infections and cancer including HPV that is owned by the Leiden University Medical Center and licensed to ISA Pharmacueticals, Leiden, the Netherlands, and is a consultant/advisory board member for ISA Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.J.P. Welters, L.-A. van der Velden, S.H. van der Burg

Development of methodology: M.J.P. Welters, W. Ma, E.S. Jordanova

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.J.P. Welters, W. Ma, S.J.A.M. Santegoets, R. Goedemans, I. Ehsan, E.S. Jordanova, V.J. van Ham, F. Koning, S.I. van Egmond, L.-A. van der Velden, S.H. van der Burg

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J.P. Welters, W. Ma, S.J.A.M. Santegoets, E.S. Jordanova, V. van Unen, F. Koning, P. Charoentong, Z. Trajanoski, S.H. van der Burg

Writing, review, and/or revision of the manuscript: M.J.P. Welters, W. Ma, E.S. Jordanova, S.I. van Egmond, Z. Trajanoski, L.-A. van der Velden, S.H. van der Burg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.J.P. Welters, L.-A. van der Velden

Study supervision: M.J.P. Welters, S.H. van der Burg

Other (mass cytometry analysis): F. Koning

The authors thank the patients for participating in this study. The authors also acknowledge C.E. van der Minne for her technical assistance.

This study was financially supported by a grant from the Dutch Cancer Society 2014-6696 (to S.H. van der Burg, L.-A. van der Velden, and M.J.P. Welters).

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.