HPV-associated head and neck cancer is characterized by distinct profiles of CD8⁺ T cells and myeloid-derived suppressor cells

Over the past decades, head and neck cancer (HNC), the worldwide 6th most common cancer, has experienced a significant change due to the identification of human papillomavirus (HPV)-associated disease, a molecular entity that is clearly distinct from HPV⁻ HNC [1]. In the oropharynx, the high-risk subtype HPV16 is–by far–the most frequent HPV subtype associated with the disease. Despite the fact that HPV⁺ HNC is usually diagnosed at an advanced nodal infestation, therapeutic response and survival is significantly better in comparison with HPV⁻ HNC patients [2‐4]. In some studies, the favorable prognosis was attributed to an increased sensitivity of HPV⁺ HNC to radiation [5]. However, recent evidence suggests that differences in immune profiles and response may contribute to the improved prognosis of HPV⁺ HNC patients [4, 6, 7]. In the peripheral blood, HNC is characterized by profound signs of immunosuppression associated with reduced T cells counts and induction of regulatory T cells as well as myeloid-derived suppressor cells (MDSC) [8‐11]. Although the prognostic relevance of circulating CD8⁺ T cells is less clear and demands further investigation, some studies already reported on the identification of CD8⁺ T cell subsets, which were associated with a more favorable outcome [12]. Therefore, further investigations exploring potential differences in T cell subsets of HPV⁺ and HPV⁻ HNC patients are required, including functionality, antigen specificity, phenotypes, and—of utmost importance—possible causes for these differences.

Over the past decade, MDSC, which are induced in HNC and many other types of cancer [8], have emerged as important regulators of CD8⁺ T cell-mediated immunity. MDSC are a heterogeneous group of myeloid cells, and several mechanisms have been identified how these cells regulate and inhibit the activity of other immune cells, particularly T cells [13, 14]. Originally identified in mice [15], more recently, MDSC and subsets thereof have also been described in humans [16‐18].

It was the aim of this study to investigate the frequency and subset distribution of CD8⁺ T cells and MDSC in patients with primary HNC. We focused our detailed longitudinal and comparative analysis of the CD8⁺ T cell compartment on a comparative analysis of HPV⁺ versus HPV⁻ HNC and hypothesized that differences in MDSC and CD8⁺ T cell subsets exist between these HNC subtypes. Our findings suggest different baseline levels and dynamics of MDSC and T cell phenotypes in HPV⁻ versus HPV⁺ HNC, which may contribute to differences in clinical outcome.

Various types of squamous cell carcinoma (SCC) induce systemic immune suppression, and alterations in the T cell compartment have been reported in patients with HNC [23, 24]. Therefore, we studied a series of 89 patients with localized or locally advanced HNC treated at the West German Cancer Center. Treatment strategies (surgery alone, surgery followed by adjuvant therapy and primary radiochemotherapy) were proposed by the multidisciplinary tumor board for head and neck cancers, and treatments were conducted using state-of-the art technique and facilities (Table 1). First, we analyzed the frequency, phenotype, and subset distribution of peripheral blood CD8⁺ T cells in our cohort of HNC patients compared to healthy donors (HD). We found that the relative frequency of CD8⁺ T cells within the T cell compartment was reduced in cancer patients (Fig. 1A). CCR7, CD62L, PD-1, and TIM-3 are important markers of T cell functionality. Analyzing expression of these markers by CD8⁺ T cells, we observed a significant relative reduction of CCR7⁺ and CD62L⁺ CD8 cells in cancer patients (Fig. 1B, C). Considering the expression of activation and exhaustion markers PD-1 and TIM-3, our analysis revealed no significant differences in CD8⁺ T cells between HNSCC patients and HD (Fig. 1D), with double negative cells representing the highest fractions in both groups. We next analyzed the subset composition of CD8 cells (naïve, central memory, effector memory, terminally differentiated effector cells) in HNC patients versus HD. Following the protocol delineated in Fig. 1E we applied CD62L and CD45RA as subset markers. While the frequency of T_CM, T_EM, and T_TD was similar in HNC patients and HD, there was a striking and statistically significant reduction of naïve CD8⁺ T cells in HNC patients (Fig. 1E–F). These data support previous reports on low frequencies of CD8⁺ T cell populations in HNC patients and reveal further differences between the phenotypic T cell properties of HNC patients and HD.

Next, we conducted a longitudinal follow-up analysis of patients and explored therapy effects on the peripheral CD8⁺ T cell compartment. As illustrated in Fig. 2A, the relative frequency of CD8⁺ T cells strongly increased in patients after therapy. Interestingly, the percentage of CD8⁺ T cells with CCR7 expression further decreased after therapy (Fig. 2B). The latter finding prompted us to investigate treatment effects on CD8⁺ T cell subsets. In agreement with the reduced expression of CCR7, we observed a significant treatment-associated relative increase in CD8⁺ T_EM and T_TD cells (left versus right columns in Fig. 2C). Surprisingly, no change in immune checkpoint expression (PD-1, TIM-3) was observed in pre- versus post-therapeutic samples (Fig. 2D). For the majority of patients, multiple post-therapeutic time points were available for analysis and representative patient trajectories suggesting durable subset shifts are shown in Fig. 2E–G.

HPV-positivity in HNC is associated with improved treatment response and better outcomes [25]. It has been suggested that improved anti-tumor immunity may contribute to this effect [26]. We, therefore, compared the treatment effect on CD8⁺ T cells and HPV-specific immune responses in HPV⁺ versus HPV⁻ patients. Therapy-associated relative increases in CD8⁺ T cells and reduction of CCR7⁺/CD8⁺ T cells could be confirmed for both HPV⁺ and HPV⁻ disease (Fig. 3A, B). However, a detailed analysis of the CD8⁺ subset composition revealed HPV-associated differences. A clear and statistically significant increase in peripheral blood CD8⁺ T_EM post-therapy could only be observed in patients with HPV⁺ tumors, whereas the frequency of CD8⁺ T_EM remained largely unchanged in patients with HPV⁻ disease (Fig. 3C). Interestingly, the frequency of PD-1⁺ and TIM3⁺ CD8⁺ T cells remained low after therapy for both patient groups. In sum, these data suggest a more pronounced treatment effect on CD8⁺ T cell phenotypes in HPV⁺ HNC.

In previous studies, T cells with specificity for HPV antigens have been reported in HPV-associated HNC. These HPV-specific T cells were mainly isolated from tissue samples [27, 28]. However, some studies reported the existence of HPV-specific T cells also in the peripheral blood, albeit at very low frequencies [29, 30]. We devised the highly sensitive dextramer technology as well as a multi-peptide ELISPOT to enumerate HPV-specific T cells pre- and post-therapy [31]. While CD8⁺ T cells with HPV16 E7-specificity could hardly be detected in HPV⁻ patients, the majority of patients with HPV16⁺ tumors displayed HPV-reactive cells at frequencies of more than 1 in 10,000 peripheral blood CD8⁺ T cells (Fig. 3E). Interestingly, associated with the removal of the primary tumor by curative therapy (primary or adjuvant radiotherapy), the frequency of these HPV16-specific CD8⁺ T cells decreased in all patients tested in our study (Fig. 3E, lower panel). Functional reactivity to HPV16 E6/E7 multi-peptides was tested by ELISPOT technology. The frequency of HPV-responsive IFN-γ producing cells was very low in pretherapeutic samples. Low-level specific responses could be detected in some patients after therapy, but no clear correlation with the HPV status was evident (Fig. 3F).

Recent molecular analyses suggested the importance of genetically HPV-driven oncogenesis (HPV RNA positivity) compared to the presence of HPV DNA only [32]. We used a method developed by Holzinger et al. [19] as a surrogate for RNA positivity. To this end, HPV DNA⁺ samples from patients with oropharyngeal SCC were stained for p16 and pRb, and tissues with high expression of p16 and simultaneous low expression of pRb were classified as HPV16 RNA⁺ (Fig. 4A). Using this classification, no significant differences between patient stratification according to DNA positivity (Fig. 3) and HPV RNA positivity (Fig. 4) were found. Thus, our data suggest that therapy-associated changes in CD8⁺ T cells occur in both HPV DNA⁺ and RNA⁺ patients and are independent of the mode of therapy (Fig. 4B, C).

MDSCs regulate the activation and function of T cells in multiple murine cancer models. While the identity and function of MDSC are less clear in cancer patients, a similar function for MDSC has also been proposed in humans. Employing a previously established protocol [16, 20], we determined the frequency of the three major human MDSC subsets in the peripheral blood of our patient cohort (Fig. 5A, B). In agreement with our previous studies [8], we found an increase in both monocytic MDSC (M-MDSC) and polymorphonuclear MDSC (PMN-MDSC) in HNC patients as compared to HD. While the frequency of PMN-MDSC was similar between patients with HPV⁺ and HPV⁻ tumors, high levels of M-MDSC (above 1%) were almost exclusively detected in patients with HPV⁻ tumors. This finding was further supported by a correlation of high M-MDSC frequencies and higher expression of CD62L on CD8⁺ T cells (Fig. 5C). Also, a high number of M-MDSC moderately correlated with reduced fractions of T_TD effector cells (Fig. 5D). These findings establish a correlation between MDSC frequencies and CD8⁺ T cells phenotypes in HNC and suggest a link between HPV status and M-MDSC expansion. Interestingly, frequencies of all circulating MDSC subsets decreased after therapy (Fig S2) and this decrease was observed independent of anatomical tumor localization (larynx or oropharynx).

In order to support and enhance our findings from the peripheral blood, we performed an analysis of the tumor microenvironment in a second cohort (Table 2) and investigated MDSC and T cells in tumor tissue islands as well as surrounding stroma. Because no robust consensus markers for tissue M-MDSC exist, only PMN-MDSC were investigated by co-staining of CD66b and LOX1 (lectin-type oxidized LDL receptor 1) expression [14]. To further investigate CD8⁺ T cell phenotypes, Granzyme B as major cytotoxic CD8⁺ T cell molecule was chosen for analysis of T cell functionality and Ki67 as surrogate for T cell proliferation activity. Of note, the frequency and density of LOX1⁺ PMN-MDSC was increased in tumor islands of HPV⁻ patients in comparison with the surrounding stroma, a phenomenon that was not observed in HPV⁺ patients (Fig. 6A). Conversely, the prevalence of CD8⁺ T cells that express Granzyme B⁺ and Ki67⁺ was notably greater within the tumor stroma of HPV⁺ patients compared to HPV⁻ patients (Fig. 6B). Consequently, the ratio of PMN-MDSC (CD66b⁺/LOX1⁺) and proliferating (Ki67⁺) CD8⁺ T cells as well as cytotoxic Granzyme B⁺ /CD8⁺ T cells, was higher in HPV⁻ patients (Fig. 6C) suggesting a higher level of intratumoral immunosuppression in those patients. Within the HPV⁺ group, patients exhibiting low levels of intratumoral CD66b⁺/LOX1⁺ PMN-MDSC (falling below the median) demonstrated significantly improved survival rates (Fig. 6D).

Finally, we analyzed potential differences in the CD8⁺ T cell infiltrate of HPV⁺ versus HPV⁻ tumors in patients. We focused on CD8⁺ and CD45RO⁺ cells as these markers are currently evaluated in the context of their immunoscore [33] and have been shown to be of prognostic relevance in HNC and other types of cancer [34‐36]. We found that CD8⁺ T cells were evenly distributed between the epithelial tumor islands (iTu, intratumoral) and the stromal regions (St, stromal) with the majority of patients showing less than 10% CD8⁺ T cells in these regions (score 1) (Fig. 7A). CD45RO⁺ T cells were more frequent than CD8⁺ T cells and preferentially located in the stroma. Interestingly, the intratumoral region was more densely infiltrated by T cells in HPV⁺ tumors compared to HPV⁻ (Fig. 7B–D). When we combined the CD8⁺ T cell score and the CD45RO⁺ T cell score in the intratumoral and stromal region into a comprehensive “immunoscore-like grade” (see material and methods for details), an increased infiltration of HPV⁺ tumors with CD8⁺ and CD45RO⁺ cells was confirmed (Fig. 7E). Collectively, these data show a denser T cell infiltrate in HPV⁺ tumors, which is mainly caused by a stronger infiltration of the epithelial tumor islands (iT region).

It has been suggested that immune-mediated mechanisms contribute to the differential prognosis of patients with HPV⁺ versus HPV⁻ HNC. In accordance with this concept, several differences between the immune status of HPV-related and non-related HNC have been described [6, 27, 37]. Herein, we carried out the first detailed longitudinal analysis of T cells and MDSC in HPV-related HNC and compared pre-and post-therapy samples from 89/61 patients from whom material was available for this type of analysis. We found that pre-therapy levels of circulating monocytic MDSC were higher in HPV⁻ disease. Importantly high numbers of MDSC correlated with low fractions of terminally differentiated CD8 effector T cells in the cohort of HPV⁻ patients. HPV⁺ tumors showed an increased infiltrate with CD8⁺/CD45RO⁺ T cells, Granzyme B⁺/CD8⁺, and Ki-67⁺/CD8⁺. This increase in CD8 effector cells was accompanied by reduced infiltration of tumor tissue by PMN-MDSC. Therapy of HPV⁺ patients was associated with an increase in CD8⁺ T_EM in the peripheral blood, which was not evident in patients with HPV⁻ disease.

In contrast to peripheral blood T cells, the clinical relevance of tumor-infiltrating CD8⁺ T cells is better established. A number of studies have associated high TIL counts with favorable prognosis in HNC [38‐40]. Controversial data exist, however, on a potential correlation between TIL frequency and HPV status. While in some studies such a correlation was found [27], it was missing in other studies [40]. In our study we found a stronger infiltration of HPV⁺ tumors with CD8⁺ and CD45RO⁺ T cells. Considering the different biology of TIL in the tumor stroma versus tumor cell islands, we separately enumerated CD8⁺ T cells in stromal versus tumor regions and included markers originally used in the immunoscore [22]. Combining both markers and intratumoral localizations, the HPV⁺ tumors displayed a higher immune cell infiltrate. This was mainly caused by a stronger infiltration of the tumor cell islands in HPV⁺ tumors. One could speculate that this deeper infiltration into the tumor center is associated with improved anti-tumor immune activity. However, in a study by Oguejiofor only CD8⁺ stromal and not CD8⁺ T cell infiltration of the tumor area was associated with improved clinical outcome [39].

Another novel aspect of our study is the longitudinal analysis which allowed us to examine therapy-associated changes. In a previous study of 20 patients with HPV⁺ HNC, Parikh et al. found a decrease in HPV-peptide-induced IFN-γ in bulk cultures of PBMC when comparing pre- and post-therapy samples [41]. While the authors did not determine the frequency of HPV-specific T cells in their system, they suggested that reduced IFN-γ was caused by immune inhibitory mechanisms at post-therapy time points and induced via chemoradiotherapy. Our data show a distinct reduction of HPV-specific T cells in post-therapy samples. This reduction of HPV-specific immunity may be caused by a reduction of tumor burden and viral load. Similar observations have been made for anti-HPV serum antibodies [42]. Interestingly, in our study, this reduction in HPV-specific T cells was paralleled by an increase in the fraction of CD8⁺ T effector memory cells (T_EM) at post-therapy time points, which is indicative of a shift in the functional, anti-tumor effective T cell compartment of these patients.

Heusinkveld et al. reported that the frequency and functionality of HPV-specific T cells are higher in the tumor as compared to the peripheral blood [28]. The exploitation of HPV-specific immunity may require therapeutic costimulation in order to achieve clinically meaningful responses [43]. Alternatively, recent advances in ex vivo expansion of HPV-specific T cells may even make adoptive T cell transfer possible [44]. We found a relatively low frequency (0.5–2 in 10.000 CD8⁺ T cells) of HPV-specific T cells in the peripheral blood of HPV⁺ cancer patients. This frequency further declined after therapy.

Interaction of PD-1 with its ligand PD-L1 has been implicated in the regulation of the immune response in HNC and other types of cancer. Our data show a low expression of PD-1 on CD8⁺ peripheral blood T cells in HNC patients. Low expression of PD-1 on peripheral blood has also been reported by Lyford-Pike et al. and by Partlova et al. [27, 45]. This is in contrast to the higher expression of PD-1 on tumor-infiltrating lymphocytes noted in those studies. Although not statistically significant in the entire cohort, we noted a small increase in the percentage of PD-1⁺ CD8⁺ T cells in the course of therapy in some patients (Fig. 2F). Originally considered an exhaustion marker, under certain conditions, PD-1 expression may also represent antigen-experienced T cells with the unique capacity to recognize tumor-specific neoantigens [46, 47]. Thus, an increase in the course of therapy, as observed in our study, could be indicative of post-therapeutic immune activation.

Recent evidence suggests that MDSC significantly contribute to the regulation of anti-tumor immunity and in particular regulation of T cell activity in murine and human malignant disease. For the first time, we comprehensively compared levels of MDSC subsets in HPV⁺ versus HPV⁻ human HNC. Interestingly the frequency of M-MDSC was significantly higher in the peripheral blood of HPV⁻ patients. In addition, in HPV⁻ disease, high levels of MDSC were associated with lower fractions of terminally differentiated peripheral blood effector T cells. In murine models of vaccination against HPV⁺ tumors, MDSC partially controlled the activity of anti-tumor CD8⁺ T cells [48]. Some earlier studies suggested that PMN-MDSC induced antigen-specific tolerance, while monocytic MDSC is involved in non-specific suppression [49]. Interestingly, in a recent multicenter analysis of circulating MDSC subsets across multiple diseases within the Mye-EUNITER (www.​mye-euniter.​eu) consortium, programmed death-ligand 1 (PD-L1) was found to be primarily expressed in M-MDSC [50]. As the PD-1 and PD-L1 interaction results in T cell anergy and inhibits T cell proliferation, the increased PD-L1 expressing M-MDSC fractions in HPV⁻ HNC patients potentially contribute to our observed differences in T cell phenotypes with lower T_EM cells and lower immunoscores for these patients. Accordingly, Spehner and colleagues identified M-MDSC levels in another HPV-dependent squamous cell carcinoma entity as the major influence on TH1-adaptive immune responses and patients' survival [51].

Additionally, our investigations show that differences between HPV⁺ and HPV⁻ tumor patients are also present in the tumor tissue. In previous work we provided evidence for in situ intratumoral activity by LOX1⁺ PMN-MDSC [14]. These findings are in line with reports by Chai and colleagues who identified ARG1/iNOS-dependent inhibition of T cell proliferation by LOX1⁺ PMN-MDSC in glioblastoma [52]. In this study we report reduced Ki67 and Granzyme B expression in CD8⁺T cells of HPV⁻ patients that also have higher PMN-MDSC / CD8⁺Granzyme B⁺ ratios. In HPV⁺ patients, we observed improved clinical outcome in patients with lower PMN-MDSC fractions. An association of MDSC and T cell frequencies was also observed by Liang and colleagues who reported—in cervical cancer patients—a negative correlation of PMN-MDSC and densities of CD8⁺ T cells, influencing clinicopathologic parameters and prognosis [53]. Although the immunoregulatory mechanisms of MDSC in cancer patients are still far from being understood, these data support a link between MDSC expansion and T cell differentiation and function, which could further explain the somewhat poor prognosis of patients with HPV⁻ disease.

In sum, our analyses revealed important differences in the MDSC and T cell compartment in HPV⁺ versus HPV⁻ disease. Patients with HPV⁺ HNC showed a post-therapeutic expansion of CD8⁺ T_EM. These patients also demonstrated higher immune cell infiltration of their tumors as defined by the immunoscore markers CD8 and CD45RO. In contrast, patients with HPV⁻ tumors had higher pre-treatment levels of M-MDSC and this correlated with a reduced frequency of CD8⁺ T effector cells in the blood. HPV⁻ patients also showed higher frequencies of intratumoral PMN-MDSC and lower frequencies of proliferating and cytotoxic T cells. In HPV⁺ patients, lower intratumoral PMN-MDSC levels were associated with an improved survival. These differences in MDSC and T cell functionality may further contribute to and explain differences in clinical outcome of patients with HPV⁺ versus HPV⁻ HNC.