The Anatomical Location Shapes the Immune Infiltrate in Tumors of Same Etiology and Affects Survival

The tumor microenvironment is involved in tumorigenesis and tumor progression (1). Analyses of many tumor types revealed that a strong T helper 1 (Th1) cytotoxic microenvironment is associated with a more favorable prognosis and therapy responsiveness (2, 3). When the tumor metastasizes to other anatomical locations, the original immune contexture of the primary tumor is maintained (4), indicating that the immune contexture of the primary tumor determines the therapy response of later metastases (5). However, whether the microenvironment of the original tissue in which the primary tumor arises has an impact on the shape of the tumor microenvironment has largely been neglected. Perhaps because primary tumors of the same type can have different mutations and activated oncogenic pathways, some of which are known to affect the constitution of the tumor microenvironment. For instance, overexpression of BRAF, loss of PTEN, and activation of the WNT/β-catenin signaling pathway modulate the extent of the lymphocytic tumor infiltrate (6–8). In this context, it is hard to dissect the potential influence of the original tissue on the tumor microenvironment. An answer to what extent the immune contexture can be imprinted by the site or origin may come from studies on human papillomavirus (HPV)–induced tumors. Although these tumors can arise in different anatomical locations (cervix, vagina, vulva, anus, and oropharynx), they share the same virus-driven oncogenic pathway.

HPV is strictly epitheliotropic and infects basal epithelial cells in skin and mucous membranes. Integration of the viral genome into the host DNA leads to overexpression of the E6 and E7 oncoproteins and finally to transformation of the epithelial basal cells into cancer cells (9). Within the cervix, the cells close to the squamocolumnar junction, in what is called the transformation zone, are susceptible to HPV infection, transformation, and progression to cancer (10). Notably, within the genital tract it is the normal uterine cervix, including the transformation zone that contains the highest numbers of immune cells (11, 12). The oropharynx comprises the palatine tonsils, the soft palate, the tongue base, and the posterior pharyngeal wall, but HPV almost exclusively infects and transforms the highly specialized lymphoepithelium lining the tonsillar crypts (13). A direct comparison of the lymphocytic infiltrate in routinely removed nondiseased fresh tonsils and cervical tissue revealed differences in the CD4:CD8 T-cell ratio and the distribution of central memory (Tcm) and effector memory (Tem) CD4⁺ T cells between the 2 tissue types (14). The indication that these 2 issues at different anatomical locations already display a different immune contexture under noncancerous conditions provides an opportunity to assess if the original tissue microenvironment also bears impact on the tumor immune microenvironment.

To study the potential impact of the original tissue on the intratumoral immune contexture, we analyzed immune cell populations in a series of primary tumors, tumor-draining lymph nodes (TDLNs), and peripheral blood mononuclear cell (PBMC) samples, obtained from patients with either HPV-driven cervical carcinoma or oropharyngeal squamous cell carcinoma (OPSCC), by high-dimensional single-cell mass cytometry (CyTOF) and different functional analyses, including the assessment of HPV-specific T cells. Our data show that primary tumors of the same etiology, but arising in different anatomical locations, are infiltrated with distinct lymphocytic populations, and that the anatomical location affects the efficiency of tumor-specific T cells to infiltrate the tumor.

The authors acknowledge the reporting of Minimal Information About T-cell Assays.

Patients included in this study were part of 2 larger observational studies on cervical carcinoma and OPSCC. Women with histologically proven cervical carcinoma (International Federation of Gynecology and Obstetrics 1a2, 1b1/2) were included in the CIRCLE study investigating cellular immunity against anogenital lesions (15, 16). Patients with histology-confirmed OPSCC were included in the P07-112 study investigating the circulating and local immune response in patients with head and neck cancer (17, 18). Patients were included after signing informed consent. Both studies were conducted in accordance with the Declaration of Helsinki and approved by the local medical ethical committee of the Leiden University Medical Center (LUMC) and in agreement with the Dutch law. The patients received standard-of-care treatment consisting of surgery, radiotherapy, chemotherapy, treatment with monoclonal antibody, or combinations hereof. HPV typing and p16^ink4a IHC staining were performed on formalin-fixed paraffin-embedded tumor sections at the LUMC Department of Pathology as described (19). Tumor staging was done according to the National Comprehensive Cancer Network (https://www.nccn.org/professionals). An overview of patient characteristics and treatment is given in Supplementary Table S1.

We performed an IHC staining for p53 on the tissue sections of the OPSCC tumors as described previously (20). An experienced pathologist reviewed all slides and scored the specimen as “wild-type” (p53wt) when nuclei of the tumor cells stained weak to moderate, comparable with adjacent normal epithelium.

In addition, next-generation sequencing was performed as described previously (21). In brief, formalin-fixed paraffin-embedded tissue blocks of the OPSCC tumors were microdissected to get a tumor percentage of >70%. Tumor DNA was isolated using the Tissue Preparation System with VERSANT Tissue Preparation Reagents (Siemens Healthcare Diagnostics) after which somatic variant analysis of the TP53 gene was performed using the AmpliSeq Cancer Hotspot Panel 4 (Thermo Fisher Scientific) with a sequence coverage of 90%. Results of this analysis are shown in Supplementary Table S1.

Venous blood samples were drawn prior to surgery, and PBMCs were isolated using Ficoll density gradient centrifugation as described previously (18, 22). Cervical carcinoma and OPSCC tumor and cervical carcinoma TDLN material or biopsies were obtained and handled as described (16, 18). First, tumor material was cut into small pieces. One-third of the tumor pieces were incubated for 60 minutes at 37°C in Iscove's Modified Dulbecco's Medium (IMDM, Lonza) with 10% human AB serum (Capricorn Scientific) and supplemented with high dose of antibiotics [50 μg/mL Gentamycin (Gibco/Thermo Fisher Scientific, TFS), 25 μg/mL Fungizone (Invitrogen/TFS), 100 IU/mL penicillin (pen; Gibco/TFS), and 100 μg/mL streptomycin (strep; Gibco/TFS)], after which the tumor pieces were put in culture in IMDM supplemented with 10% human AB serum, 100 IU/mL penicillin, 100 μg/mL streptomycin, 2 mmol/L l-glutamin (Lonza; IMDM complete), and 1,000 IU/mL human recombinant IL2 (Aldesleukin, Novartis). Cultures were replenished every 2 to 3 days with fresh IMDM complete and IL2 to a final concentration of 1,000 IU/mL. When there were sufficient T cells, the cells were cryopreserved and stored in liquid nitrogen until use. Approximately two-thirds of the tumor pieces were incubated for 30 minutes at 37°C in IMDM dissociation mixture containing 10% human AB serum, high dose of antibiotics (as above) and 50 μg/mL DNAse I (Roche), and 1 mg/mL collagenase D (Roche), after which the tumor was dispersed to single cells using the GentleMACS dissociator (Miltenyi) with a company-installed program (h_tumor_02). Following centrifugation, cells were resuspended in PBS (B. Braun), put on a 70-μm cell strainer (Falcon) to obtain a single-cell suspension, counted using trypan blue exclusion (Sigma), and cryopreserved at approximately 2 million cells/vial. All tumor samples from OPSCC and part of the cervical carcinoma samples were dissociated as above. From patient C1016 onward, the approach to prepare single-cell suspensions was adjusted. The DNAse I/collagenase D enzymes were replaced by 0.38 mg/mL of the commercially available Liberase enzymes (Liberase TL, research grade, Roche), the incubation period was reduced to 15 minutes, and the GentleMACS dissociator was no longer used.

Single-cell suspensions from cervical carcinoma TDLN material were prepared either through DNAse I and collagenase D dissociation as described above with or without the use of the GentleMACS dissociator, or by scraping the cutting surface of the LN with a surgical scalpel blade, after which they were cryopreserved until further analysis. To generate TDLN-derived T-cell batches, ex vivo TDLN were expanded as described previously (22). In brief, ex vivo TDLN were thawed, washed, and seeded at a density of 0.4 to 1 × 10e6 cells per well in a 24-well plate in 1 mL of IMDM complete medium and stimulated with 5 μg/mL HPV16 E6 and E7 clinical grade long peptide pools in the presence of 150 U/mL recombinant human IL2. Cultures were replenished every 2 to 3 days with fresh IMDM complete and IL2, and after 22 to 28 days, the expanded T cells were cryopreserved and stored in the vapor phase of liquid nitrogen until further use.

Blood and matched tumor and/or LN samples from 20 patients with cervical carcinoma and 9 patients with OPSCC were analyzed by mass cytometry (CyTOF). Details on antibodies used are listed in Supplementary Table S2. Conjugation of the purified antibodies with the isotopic tags was done using the MaxPar X8 antibody labeling Kit (Fluidigm Sciences) according to the manufacturer's instructions, and following conjugation, the antibodies were stored at 4°C in Candor PBS Antibody Stabilization Buffer (Candor Bioscience GmbH). Mass cytometry antibody staining and acquisition procedures were performed as described (23). In brief, cell samples were thawed according to standard operation procedures and stained with 1 μmol/L Cell-ID intercalator-103Rh (Fluidigm Sciences) for 15 minutes at room temperature (RT) to identify dead cells. Following incubation, the cells were washed with MaxPar Cell stain buffer (Fluidigm Sciences) and incubated with Human TruStain FcX Fc Receptor Blocking Solution (Biolegend) for at least 10 minutes at RT. Next, cells were incubated with the metal-conjugated antibodies for 45 minutes at RT. After the cells were washed twice with the MaxPar Cell stain buffer, the cells were stained overnight at 4°C with 125 nmol/L Cell-ID intercalator-Ir in MaxPar Fix and Perm Buffer (Fluidigm Sciences). The next day, cells were washed and acquired by CyTOF 2 or helios-upgraded CyTOF2 mass cytometer (Fluidigm Sciences). Data were normalized by using EQ Four Element Calibration Beads (Fluidigm Sciences) with the reference EQ passport P13H2302.

Gating for single, live CD45⁺ cells for each PBMC, LN or tumor sample was done using the cloud-based Cytobank software (Fluidigm Sciences; example in Supplementary Fig. S1). The high-dimensional single-cell data were analyzed by analysis by tSNE analysis in Cytosplore with default parameters (perplexity, 30; iterations, 1,000; ref. 24) and by the fully automated hierarchical clustering (unsupervised) tool CITRUS using the cytobank software. The different cell populations were visualized and quantified.

To determine the specificity of T cells infiltrating the tumor and/or TDLN, cultured TIL and TDLN T-cell batches from HPV16⁺ cervical carcinoma and OPSCC tumors were analyzed for the presence of HPV16-specific T cells using 5-day [3H]-thymidine–based proliferation assay as described (18). T-cell responses against autologous HPV16 E6/E7 synthetic long peptide (SLP; 22-mers with 14 amino acids overlap) loaded monocytes were tested in triplicate. PHA (0.5 μg/mL; HA16 Remel; Thermo Fisher Scientific) was taken along as positive control, whereas unloaded monocytes served as negative control. At days 1.5 and 4, supernatant (50 μL/well) was harvested to determine cytokine production. During the last 16 hours of culture, 0.5 μCi/well of [3H]-thymidine was added to measure proliferation. The stimulation index was calculated as the average of test wells divided by the average of the medium control wells. A positive response was defined as a stimulation index of at least 3.

Antigen-specific cytokine production was determined by cytometric bead array (Th1/Th2 kit, BD Bioscience) according to the manufacturer's instructions. The cutoff value for cytokine production was 20 pg/mL, except for IFNγ for which it was 100 pg/mL. Positive cytokine production was defined as at least twice above that of the unstimulated cells.

The HPV16 reactivity of TIL was also determined by intracellular cytokine staining (ICS). ICS for the markers CD3, CD4, CD8, CD137, CD154, CD161, CD103, IFNγ, and TNFα was performed as described previously (18) following stimulation with HPV16 E6/E7 SLP-loaded autologous monocytes or Epstein-Barr virus (EBV)-transformed B-lymphoblastoid cell lines (BLCL). Unloaded autologous monocytes or BLCL cells were used as negative and Staphylococcal Enterotoxin B (SEB; 2 μg/mL; Sigma) was used as positive control. A positive response was defined as at least twice the value of the negative control and at least 10 events in the gate. Acquisition of cells was done on a LSRII Fortessa (BD Biosciences). Data were analyzed using DIVA software (version 6.2 or 8.02 BD Biosciences).

Nonparametric (Wilcoxon signed-rank or Mann–Whitney test for 2 samples and Friedman or Kruskal–Wallis with Dunn multiple comparison test for multiple samples) and parametric (paired or unpaired t test for 2 samples or repeated measures one-way ANOVA or ordinary one-way ANOVA with Tukey multiple comparison test for multiple samples) tests were performed as appropriate. For survival analysis, patients were grouped into 2 groups according to the median (i.e., grouped into below or above the median of the total group for each parameter), after which survival was tested using the Kaplan–Meier method, and statistical significance of the survival distribution was analyzed by the log-rank testing. All statistical tests were performed at the 0.05 significance level, and differences were considered significant when P < 0.05, as indicated with an asterisk (^#, P < 0.1; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001). Statistical analyses were performed using GraphPad Prism 7.1.

First, the lymphocyte content was studied in a series of tumors, TDLN, and PBMC samples of patients with HPV-induced primary cervical carcinoma (Supplementary Table S1) using 36-parameter mass cytometry (Supplementary Table S2). tSNE analyses of all samples revealed different clusters of B cells, natural killer (NK) cells, and several populations of CD4⁺ and CD8⁺ T cells (Fig. 1A and B; Supplementary Fig. S1). Clearly, the naïve (Tn), Tcm, and Tem populations of CD4⁺ and the Tcm/Tem populations of CD8⁺ T cells did not differ between PBMC and LN (Fig. 1C). In contrast, the tumor showed a strong enrichment of CD8⁺ T cells, which were mainly effector T cells (Teff; Fig. 1B and C). Further analysis of the tSNE populations revealed 8 distinctive (groups of) lymphocyte populations (Fig. 1D) and demonstrated that cervical carcinoma TDLN contained more total B cells (both IgM⁻ and IgM⁺ B cells), whereas more NK cells were found in PBMC (Fig. 1D). Strikingly, the percentage of total CD4⁺ T cells was lower in tumors, albeit that those of CD4⁺ Treg-like cells was higher. CD8⁺ T cells were enriched in tumors, in particular the CD8⁺CD103⁺ T cells, also known as long-lived tissue-resident memory cells (25). Thus, although the T-cell populations in cervical carcinoma TDLN more or less resembled that of PBMC, cervical carcinoma tumors displayed a more prominent CD8⁺CD103⁺ T-cell infiltrate, with the CD8⁺ T cells in ratio of about 1:1 to CD4⁺ T cells.

Next, the lymphocyte content of the cervical carcinoma samples was compared with a series of tumor and blood samples from patients with HPV-induced OPSCC. Two-dimensional principal component analysis (2D-PCA; Fig. 2A) revealed clear phenotypic diversity between tumor, TDLN, and PBMC samples as well as between the groups of cervical carcinoma and OPSCC tumor samples (Fig. 2A). Combined tSNE analysis of paired blood and tumor samples resulted in the identification of 29 distinct immune populations (Fig. 2B and C). Subsequent comparison of the immune cell content showed that the lymphocytic composition of cervical carcinoma and OPSCC greatly differed with respect to the percentages of IgM⁺ B cells, CD4⁺ T cells, and CD8⁺ T cells (Fig. 2D). The lymphocytic composition of PBMC samples was quite similar for cervical carcinoma and OPSCC (Fig. 2E).

In order to automatically discover stratifying biological signatures within the OPSCC and cervical carcinoma blood and tumor samples, we made use of the automated and data-driven CITRUS platform, as an unbiased and thorough correlation-based tool for mining and inspection of cell subsets nested within high-dimensional datasets (26). The CITRUS analysis resulted in 30 distinctive (groups of) lymphocyte populations (Fig. 3A and B; Supplementary Fig. S2; and Supplementary Table S3). Closer inspection of the types of lymphocytes in cervical carcinoma and OPSCC revealed that the percentage of IgM⁺ B cells is higher in OPSCC and comprised 2 different populations (subsets 3 and 4; Fig. 3C) that differed with respect to surface levels of CD27 and HLA-DR (Supplementary Fig. S2). Major differences were observed with respect to infiltrating T cells (Fig. 3E and F; Supplementary Fig. S2; and Supplementary Table S3). This was not biased by the higher frequency of B cells in OPSCC, as similar differences were also observed within the total CD3⁺ T-cell population (Supplementary Fig. S3A and S3B). Cervical carcinoma displayed a slightly lower infiltration with CD8⁺ Tcm/Tem (subset 15) and a more dense infiltration of CD8⁺ effector memory RA⁺ T cells (Temra; subset 17). In addition, cervical carcinomas were highly infiltrated with CD8⁺CD103⁺CD161⁻ Teff (subsets 8 and 11) and CD8⁺CD103⁺CD161⁺ Teff (subsets 9 and 10) cells. Subset 8 Teff cells (CD27-HLADR-CD38dimPD1-) seem less activated based on their profile. A fifth population of CD8⁺CD103⁺ Teff (subset 12) and a population of CD8⁺ Teff, expressing intermediate levels of CD103, HLA-DR, CD27, CD38, and PD-1 (subset 14), were more prominent in OPSCC. At the CD4⁺ T-cell level, OPSCC contained more CD4⁺ Tn (subsets 19–21) and potentially CD4⁺CD161⁺ Tcm (subset 28). Levels of CD4⁺CD161⁺ Tem (subsets 22–24) were clearly higher in OPSCC and comprised populations of CD27⁺ and CD27⁻ cells. We also observed 2 populations of CD4⁺ cells with a Treg-like phenotype (subsets 26 and 27), but their levels were similar in cervical carcinoma and OPSCC. Notably, there were no gross differences in the composition of CD8⁺ and CD4⁺ T cells in the blood of patients with cervical carcinoma and OPSCC.

Thus, HPV-driven cervical carcinoma and OPSCC differ considerably with respect to their lymphocytic infiltrate. In OPSCC, more intratumoral IgM⁺ B cells as well as CD4⁺ Tn and CD4⁺CD161⁺ Tem were found, whereas cervical carcinoma contained higher numbers of CD8⁺CD103⁺ Teff and in particular CD8⁺CD103⁺CD161⁺ cells. Importantly, lymph node involvement or absence thereof in cervical cancer did not account for these observed differences between the cervical carcinoma and OPSCC tumor immune environment (Supplementary Fig. S3C and S3D).

Because the large majority of CD8⁺CD103⁺ Tem and CD4⁺ Tem coexpressed CD161, we next examined their functional properties. To this end, tumor-infiltrating lymphocyte (TIL) cultures containing HPV-specific T cells were stimulated with cognate antigen, and cytokine production was analyzed in CD103 and CD161 double-negative, single-positive, and double-positive CD4 (n = 28) or CD8 (n = 9) responding T-cell populations (Fig. 4A–C). CD8 reactivity was predominantly found in CD103⁺CD161⁺ and CD103⁺CD161⁻ T cells, with more than half of the responding CD8⁺ T cells expressing CD161. About 40% of the cytokine-producing CD4⁺ T cells expressed CD161 (Fig. 4B and C). Moreover, the expression of CD103 and/or CD161 within the CD4⁺ and CD8⁺ T-cell populations is not biased by changes in their expression following antigenic stimulation, as CD103 and CD161 expression was shown stable following HPV16 and SEB stimulation (Supplementary Fig. S4). Importantly, the level of cytokine production—based on the mean fluorescence intensity of cytokine staining (27, 28)—was higher in CD161⁺CD4⁺ or CD8⁺ T cells (Fig. 4A and D), suggesting that CD161⁺ T cells are highly activated. To substantiate this notion, the staining intensity of surface markers expressed by CD161⁺ and CD161⁻CD4⁺ and CD8⁺ T cells infiltrating cervical carcinoma and OPSCC tumors was analyzed (Fig. 4E). In both types of tumors, CD4⁺CD161⁺ T cells expressed higher levels of CD103, PD-1, and CD127, whereas levels of CD25, CD27, and CD45RA were lower. CD8⁺CD161⁺ T cells displayed higher levels of CD103, HLA-DR, and PD-1, whereas levels of CD45RA and CCR7 were lower (Fig. 4E). Collectively, this classifies CD4⁺CD161⁺ and CD8⁺CD103⁺CD161⁺ TIL as highly activated effector T cells.

A recent publication on flow cytometric analyses of the percentages of CD4⁺ and CD8⁺ subpopulations in routinely removed nondiseased fresh tonsils and cervical tissue (14) prompted us to compare this with the composition of these populations in OPSCC and cervical carcinoma.

Interestingly, the CD4:CD8 ratios in cervical carcinoma tumors resembled that of normal cervical epithelium, whereas the CD4:CD8 ratios in cervical carcinoma TDLN and OPSCC were more similar to normal tonsil (Fig. 5A). These data are consistent with our calculations showing that the median CD4:CD8 tumor-infiltrating T-cell ratio in cervical carcinoma is not different from normal cervical epithelium (0.41 vs. 0.51, respectively) as measured by IHC (29) and is 3-fold lower than the CD4:CD8 ratio (1.56) in OPSCC (30). Furthermore, the composition of CD8⁺ Tn, Tcm, Tem, and Temra within cervical carcinoma and cervical carcinoma TDLN showed similarity to normal cervical tissue and tonsils, respectively. This was not the case for OPSCC and normal tonsils (Fig. 5B). The composition of CD4⁺ T cells was different in OPSCC and cervical carcinoma compared with normal tissue (Fig. 5C). The CD4⁺ and CD8⁺ T-cell populations in cervical carcinoma TDLN paralleled that of PBMC (Fig. 5B and C). Thus, cervical carcinoma and OPSCC strongly differ with respect to the composition of CD8⁺ and CD4⁺ T-cell subsets and CD4:CD8 ratio, the latter of which is more comparable with the tissue of origin. The observation that the CD4:CD8 ratio and composition of CD4 and CD8 Tn, Tcm, Tem, and Temra of HPV-driven OPSCC is similar to that of OPSCC with a nonviral etiology (Supplementary Fig. S5) also suggests an impact of the location on the tumor immune microenvironment.

To determine whether the differences in CD4⁺ T-cell infiltration between cervical carcinoma and OPSCC were also visible at the tumor-specific T-cell level, we compared the detection rate of HPV16-specific CD4⁺ T-cell responses in cervical carcinoma (n = 48), TDLN of cervical carcinoma (n = 18), and OPSCC (n = 53) as analyzed and reported by us earlier (15–18, 22). Indeed, HPV16-specific T-cell reactivity significantly differed between the groups and was detected in 35% of cervical carcinoma, 100% of cervical carcinoma TDLN, and 63% of OPSCC (Fig. 6). The detection of HPV16-specific CD4⁺ T cells in all cervical carcinoma TDLN, including those of patients in which the matched TIL samples did not show reactivity (n = 2), and in only 35% of HPV16⁺ cervical carcinoma tumors suggests that the infiltration of cervical carcinoma by tumor-specific CD4⁺ T cells is hampered.

Previously, we showed that a strong infiltration with CD4⁺ T cells and CD4⁺CD161⁺ T cells is positively associated with survival in patients with HPV⁺ OPSCC (18). Consequently, the lower CD4⁺ T-cell infiltration observed in cervical carcinoma would predict that here CD4⁺ T-cell infiltration would have less impact on clinical outcome. First, we analyzed the effect of CD4 gene expression in a group of 214 squamous cervical carcinoma patients within the publicly available cancer genomic atlas (TCGA) database (31). Patients with higher than median CD4 expression in cervical carcinoma tumors displayed no survival benefit (Fig. 7A), irrespective of molecular classification as CD4⁺ Tcm or Tem (32). In addition, we performed follow-up analysis of a group of 38 patients with cervical carcinoma from whom we had previously analyzed the tumors with respect to T-cell infiltration by IHC (19). This analysis confirmed that there is no significant impact on survival when the patients were divided based on the median number of tumor-infiltrating CD4⁺ T cells in cervical carcinoma (Fig. 7B). Thus, the number of CD4⁺ T cells is generally lower in cervical carcinoma than OPSCC, and this is reflected by a different impact on survival of CD4⁺ T cells in these 2 tumor types.

Detection of HPV-specific T cells in OPSCC is a strong predictor for patient survival in OPSCC (18). Hence, we analyzed the value of this in patients with cervical carcinoma from whom we cultured TIL and were able to test HPV specificity. In contrast to a group of 51 patients with OPSCC, the detection of HPV-specific T cells among ex vivo–expanded TIL was not able to predict survival in a set of 42 patients with cervical carcinoma (Fig. 7C and D). Most likely, because the total CD4⁺ T-cell number and thus also that of HPV-specific T cells—before the in vitro expansion—are too low to mediate any strong effect in vivo. Interestingly, whenever an HPV-specific CD4⁺ T-cell response was detected in the tumor, the percentage of CD4⁺CD161⁺ Tem was higher in both OPSCC and cervical carcinoma (Fig. 7E). As we showed that the intratumoral CD4⁺CD161⁺ T-cell population is highly activated, we assessed whether high numbers of CD4⁺CD161⁺ effector cells would have the same positive association with survival in cervical carcinoma as previously reported by us for OPSCC (18). Indeed, the group of patients with cervical carcinoma with high expression levels of CD4 and CD161 in the tumor showed a better outcome (Fig. 7F).

We exploited the fact that HPV16 can cause tumors to arise in different tissues to study the extent to which the anatomical location contributes to the immune contexture of a developing primary tumor. Application of our high-dimensional single-cell mass cytometry–based approach with 36 markers in freshly digested tumor samples and PBMC of patients with cervical carcinoma or OPSCC identified 30 distinctive clusters of lymphocytes. These findings are the first line of evidence showing that tumors of the same etiology, but arising in a different tissue, have a different immune contexture. In a direct comparison, OPSCC contained more B cells and showed a specific enrichment with subpopulations of CD4⁺CD161⁺ Tem, whereas several subpopulations of CD8⁺CD103⁺ Teff were enriched in cervical carcinoma. In addition, the CD4:CD8 ratio was 3-fold higher in OPSCC than cervical carcinoma. Finally, HPV-specific TILs were more frequently detected in OPSCC than cervical carcinoma.

Interestingly, the CD4:CD8 ratio in OPSCC closely resembled the ratio found in normal tonsils (14), and the CD4:CD8 ratio in cervical carcinoma is highly similar to that found in normal cervical epithelium as measured by flow cytometry (14, 33) and IHC (11, 12, 29). Furthermore, the percentages of CD8⁺ Tn, Tcm, Tem, and Temra found in cervical carcinoma are comparable with previous reported findings in normal cervical tissue (14). Moreover, the numbers of cervical carcinoma resident (CD8⁺CD103⁺) T cells display the same mean and variability as found in normal cervix (34), suggesting that these cells are already present in normal tissue and thus not appear due to malignant transformation. It seems that the different immune contextures developed in primary OPSCC and cervical carcinoma reflect the immune contexture found in the tissue of origin. This notion is sustained by the observation that OPSCC of nonviral etiology displays a similar general (CD4:CD8 ratio and percentages of Tn, Tcm, Tem, and Temra) immune infiltration as HPV-driven OSPCC. Evidently, differences in the distribution of naïve and effector memory CD4⁺ and CD8⁺ T cells between healthy tonsil (and TDLN) and cervical tissue are explained by the fact that tonsil and TDLN are secondary lymphoid organs where naïve T cells recirculate (35).

In contrast to CD8⁺ T cells, the composition of the CD4⁺ T-cell population in OPSCC and cervical carcinoma was different from healthy tissue. A strong infiltration of OPSCC with CD4⁺ T cells was found to be beneficial for survival in 3 independent cohorts, and this was correlated to their functional activity, as indicated by strong expression of HLA-DR, CD38, and PD-1 (36) with or without CD161 and by expression of Tbet (18). Activated CD4⁺ Teff cells producing IFNγ and TNFα were shown to induce permanent growth arrest in the Simian virus 40 large T antigen–driven β-cancer cell model (37, 38). CD4⁺ T cells were also responsible for control of spontaneous cervical tumor outgrowth in genetically engineered K14-HPV16 transgenic mice. Within this model, a strong reduction of progressive precursor lesions was found when the HPV-specific CD4⁺ T-cell levels and activity were boosted by vaccination (39). Hence, tumor-specific CD4⁺ T cells actively participate in tumor control. However, here we showed that the relative CD4⁺ T-cell numbers are lower in cervical carcinoma than OPSCC. In addition, comparison of median CD4⁺ T-cell infiltration numbers in 2 previous studies in cervical carcinoma and 1 in OPSCC performed with similar techniques (18, 19, 29) showed that the absolute median number of CD4⁺ T-cell infiltration per square millimeter of tumor was 2-fold lower in cervical carcinoma than OPSCC. Moreover, this difference was also reflected in their association with patient's survival. Although division of patients with OPSCC based on their median infiltration with CD4⁺ T cells showed a beneficial clinical effect, this was not observed in patients with cervical carcinoma. Similar to our previous findings in OPSCC (18), CD4⁺ T-cell–mediated survival benefit is found for patients with cervical carcinoma with high levels of highly active CD4⁺ T cells, which are identified by CD161. Moreover, CD161⁺ Teff produced the highest amounts of type 1 cytokines upon activation with their cognate tumor antigen. Intermediate expression (as compared with MAIT cells) of CD161 has identified CD4⁺ and CD8⁺ mucosal Teff and Tcm cells as highly functional type 1/17 cytokine-producing cell (40–44). Acute Graft-versus-Host Disease (GvHD) is associated with high CD4⁺CD161⁺ to CD8⁺CD161⁺ ratio (45) validating a role in tissue rejection specifically for the CD4⁺ T-cell population. However, high levels of CD4 and CD161 were only observed in a small group of cervical carcinoma patients. The numerical and relative lower abundance of CD4⁺ T cells in cervical carcinoma, which is also observed in normal cervical tissue, suggests that the lack of sufficient attraction of CD4⁺ T cells is intrinsic to the location. This notion is sustained by our observation that in all patients tested, HPV-specific CD4⁺ T-cell responses are detected in the TDLN. However, HPV-specific CD4⁺ T cells were only found in a minor fraction of the large set of patients with cervical carcinoma analyzed.

Thus, the strong differences in lymphocytic infiltrate between oncogenic HPV-driven primary cervical carcinoma and OPSCC indicate a role for the originating tissue in shaping the immune contexture. Our results imply that the problem of CD4⁺ T-cell attraction in cervical carcinoma will continue to exist throughout the progression of disease and suggests that it should form a point of focus for future immunotherapeutic approaches aiming to treat progressive cervical carcinoma.

No potential conflicts of interest were disclosed.

Conception and design: S.J. Santegoets, S.H. van der Burg

Development of methodology: S.J. Santegoets, M.J.P. Welters, S.H. van der Burg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.J. Santegoets, V.J. van Ham, I. Ehsan, L.-A. van der Velden, S.L. van Egmond, K.E. Kortekaas, P.J. de Vos van Steenwijk, M.I.E. van Poelgeest

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.J. Santegoets, V.J. van Ham, I. Ehsan, P. Charoentong, C.L. Duurland, V. van Unen, K.E. Kortekaas, M.J.P. Welters, S.H. van der Burg

Writing, review, and/or revision of the manuscript: S.J. Santegoets, C.L. Duurland, L.-A. van der Velden, K.E. Kortekaas, P.J. de Vos van Steenwijk, M.I.E. van Poelgeest, M.J.P. Welters, S.H. van der Burg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Höllt, M.J.P. Welters

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

We thank the patients for participating in this study.

This study was financially supported by grants from the Dutch Cancer Society 2014-6696 (to S.H. van der Burg, L.-A. van der Velden, and M.J.P. Welters) and 2016-10726 (to S.H. van der Burg, M.J.P. Welters, and S.J. Santegoets).

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.