The immune microenvironment of HPV-positive and HPV-negative oropharyngeal squamous cell carcinoma: a multiparametric quantitative and spatial analysis unveils a rationale to target treatment-naïve tumors with immune checkpoint inhibitors

Among head and neck squamous cell carcinomas (HNSCC), oropharyngeal squamous cell carcinomas (OPSCC) comprise cancers of the palatine tonsils, base of tongue, soft palate and posterior pharyngeal wall [1], and are usually associated with tobacco and alcohol consumption [2]. However, carcinogenic high-risk human papillomavirus (HPV) infection has emerged as an important risk factor causing an increase in the incidence of OPSCC over the past 20 years, and being responsible for 71 and 52% of all OPSCC in the USA and UK, respectively [3]. These tumors mainly arise from the reticulated epithelium lining the crypts of the palatine tonsils and the base of tongue which are the preferential target of HPV transforming infection [4]. HPV-positive cancers represent a separate entity characterized by a distinct genetic profile, a platform-independent better response to treatment and a higher chemo- and radio sensitivity, which result in a significantly longer overall survival compared with HPV-negative tumors [5, 6]. The immunological response against viral antigens may contribute to the more favorable clinical course, as the HPV-positive tumor immune microenvironment (TIME) is more enriched than the HPV-negative counterpart [7, 8]. In this regard, the presence of cytotoxic T lymphocytes (CTL) specifically directed against HPV16 E6 and E7 proteins has been reported in cervical carcinoma and in OPSCC patients, and correlated with an improved survival [9]. However, the high tumor HPV-antigen load results in a high expression of immune checkpoint genes on tumor cells (e.g., indoleamine 2, 3-dioxygenase 1, IDO-1), and in dysfunction of HPV-specific CTL [10]. In addition, the role of tissue-resident memory (T_RM) CD8+ T cells co-expressing the CD103 marker has recently emerged as a favorable prognostic indicator in many cancer types, included HNSCC [11‐14].

Immune checkpoint inhibitors (ICI) targeting the PD-1/PD-L1 pathway have been approved for recurrent and metastatic HNSCC patients in the first- and second-line settings [15‐17]. However, the role of ICI in OPSCC is still controversial, as only a small proportion of patients benefit from anti-PD-1 monotherapy or in combination with chemotherapy [18]. Therefore, several trials are currently ongoing to delineate new immunotherapy and combinatorial strategies effective for HNSCC patients [19]. Furthermore, the use of immunotherapy in a neoadjuvant setting is particularly attractive. In a TIME previously exposed to therapies and rich of tumor-derived antigens, immunotherapy may indeed enhance the efficacy of standard loco-regional treatments [11].

Overall, new immune-based therapies increasingly rely on an in depth characterization of the tumor-immune cell interactions [20]. Notwithstanding, little is known about the immune contexture diversity between primary tumors and matching metastasis from the same patient. The metastatic lesion hosts cancer cells with metastatic capacity, and thus the biomarker status at the metastatic location might give more relevant prognostic information.

Here, using a combination of immune-related gene expression profiling (GEP), quantitative multiplex immunofluorescence (mIF) and spatial proximity analyses, we provide insights about the TIME characterization of HPV-positive and HPV-negative OPSCC, both on primary and metastatic lesions. We advance a potential new rationale for the incorporation of ICI in the loco-regional therapy strategies for patients with heavily infiltrated treatment-naïve OPSCC, and for the combined use of ICI and tumor-specific T cell response inducers or tumor-associated macrophages (TAM) modulators for the non-inflamed counterparts.

PD-1 inhibitors pembrolizumab and nivolumab are approved by the Food and Drug Administration for treatment of recurrent or metastatic HNSCC in the first- and second-line settings [16, 28, 29]. Moreover, ICI therapies are currently being tested in earlier treatment situations, including neoadjuvant setting [30‐32]. Unfortunately, only 15–20% of patients with HNSCC benefit from ICI, with this poor outcome being increasingly ascribed to peculiar characteristics of the TIME [33]. Thus, the analysis of the TIME in pre-treatment HPV-positive and HPV-negative patients may critically contribute to rationally identify candidates to immunotherapy.

Here, we report a detailed TIME characterization of treatment-naive HPV-positive and HPV-negative OPSCC primary tumors and matched lymph node metastases, based on multiparametric approaches that define not only the immune gene signatures and the composition of the TIME, but also the cartographic assessment of cell-to-cell interactions. All these data were used to find specific immune signatures, cell populations and spatial interactions capable to stratify OPSCC patients with better outcome.

Previous studies have exploited gene expression analysis to describe the peculiarities of HPV-positive and HPV-negative TIME [34‐37]. In our study, the differences in immune gene signatures between the two groups of patients were complemented at a spatial level and validated directly in situ. Overall, markedly increased cytotoxic and immune activation signatures, together with a higher CTL infiltration and closer interactions between CD8+ T lymphocytes and tumor cells, characterize HPV-positive lesions. These results confirm the immune “hot” nature of HPV-positive tumors as compared to the immune “cold” HPV-negative OPSCC microenvironment [38, 39]. Moreover, HPV-positive tumor nests disclosed an increased infiltration by CD8 + CD103+ T_RM cells that associate with a better outcome, as previously described [14]. In this regard, T_RM lymphocytes have been reported to express molecules involved in cytotoxic activities, such as granzyme B, perforin, IL-2 and IFN-γ, as well as exhaustion markers such as PD-1 and CTLA-4 [40], likely representing tumor-specific effector cells induced by virus antigens [12, 41].

Additionally, the correlation found between the increased tumor-infiltrating CD8+ T cells and PD-1 expression may indicate that the local immune response also induces the PD1/PD-L1 checkpoint pathway, which in turn might limit the capacity of TILs to ultimately eliminate the tumor. Paradoxically but in agreement with Badoual et al. [42], a high density of PD-1+ T lymphocytes in OPSCC primary lesions and their close interactions with cancer cells or PD-L1+ macrophages, is associated with a better prognosis. Since the increase in PD-1 expression may be the result of T cell activation, PD-1 might remain upregulated in the context of a persistent antigen-specific immune stimulation. Furthermore, PD-1+ T cells include potentially tumor-specific T_RM cells, exerting anti-neoplastic effects [43]. All these findings support the idea that PD-1 expression should not be merely considered as an exhaustion marker but rather a reflection of the antitumor reactivity, and suggest that patients with high expression of PD-1 on T cells could be potential candidates for anti-PD-1/PD-L1 blockade. Moreover, recent studies demonstrated that the T_RM cell subset increases in responder patients with non-small cell lung cancer (NSCLC) and melanoma upon anti-PD-1 administration [44, 45]. Interestingly, data from HNSCC patients treated with neoadjuvant immunotherapy (nivolumab as monotherapy or in combination with ipilimumab) demonstrated that ICI-fostered early intra-tumoral responses are primarily mediated by pre-existing T cell populations with a T_RM gene program, which is characterized by tissue residency, cytotoxicity, effector functions and inhibitory receptors including PD-1 [46]. Moreover, authors showed that neoadjuvant ICI can enhance both local and systemic tumor immunity, as they found treatment-induced expansion of emergent T cell clones in tumors and in the peripheral blood, which were undetectable prior to therapy [46]. Overall, our data suggest that neoadjuvant ICI immunotherapy in HNSCC could be regarded as one of the most promising approaches to reactivate and enhance the cytotoxic potential of the tumor-specific T_RM cells in “hot” HPV-positive tumors, but also to induce the expansion of anti-tumor T cell clones in “cold” HPV-negative lesions.

More than 50% of HNSCC patients present with metastasis to regional lymph nodes at the time of diagnosis, a feature associated to poor survival and a major prognostic factor for determining the appropriate treatment [47]. Since T cell composition can also vary in the metastatic setting determining different responses to adjuvant therapy, a comprehensive assessment of the TIME in both primary and secondary lesions can provide a more informative view. Irrespective of HPV status and in agreement with previous observations [48, 49], we found that metastases “phenocopied” the originating tumors in terms of immune infiltration and immune checkpoint expression, highlighting the possibility to evaluate the lymph node metastasis specimen if the primary tumor sample is not available for pathological analysis, a not uncommon situation as the presence of neck metastases from hidden HPV-positive OPSCC is a possible clinical manifestation of these malignancies [50].

PD-L1 is considered a predictive marker for response to PD-1/PD-L1 blockade therapies [31]; however, PD-L1 negative tumors sometimes respond to ICI treatment, suggesting the existence of other mechanisms [15]. Differently from Succaria et al. [51], we identified a significantly higher percentage of PD-L1+ cancer cells and macrophages in both HPV-positive primitive and secondary lesions, as compared to HPV-negative samples. In this regard, two mechanisms have been proposed for PD-L1 upregulation [52]. In the innate expression response, PD-L1 upregulation depends on dysregulated oncogenic signalling pathways, and chromosomal alterations and amplifications in the tumor. Thus, PD-L1 expression in cancer cells does not correlate with the nature or the intensity of the local immune response. Conversely, in the adaptive expression response, it is the IFN-γ secreted by activated CTL to induce the upregulation of PD-L1 in tumor cells. Under these latter conditions, therefore, PD-L1 expression is considered a marker of an active host antitumor immune response. Accordingly and regardless the HPV status, in our patient cohort PD-L1 expression was apparently a consequence of an active inflammatory anti-tumor microenvironment involving both myeloid and lymphoid cell populations. Consequently, PD-L1 expression on macrophages and cancer cells was higher in the immune “hot” HPV-positive samples than in the immune “cold” HPV-negative tumors. Our data are consistent with retrospective studies carried out in Merkel cell carcinoma, NSCLC and HNSCC where tumor PD-L1 expression is a positive prognostic factor [53, 54]. That PD-L1 expression is likely an adaptive response in HPV-positive tumors is further supported by the close proximity of PD-L1+ TAM and cancer cells to PD-1+ CTL, which may reflect a potentially active host immunological response otherwise blocked by immune checkpoint interactions. Thus, in a “hot” HNSCC setting, PD-L1 blockade immunotherapy appears particularly promising as a strategy to allow tumor-specific T cells to overcome the shield formed by PD-L1+ tumor cells and to exert their effector activity. Accordingly, an association between PD-1/PD-L1 proximity and better response to anti-PD-1 treatment was reported in Merkel cell carcinoma [55]. On the other hand, these results provide also the rationale to adopt combinatorial therapies [18] to enhance TIL infiltration prior or concurrent with PD-1/PD-L1 blockade immunotherapy, as a good strategy to treat non-inflamed HPV-negative patients expressing low amounts of PD-L1 molecule.

The myeloid cell compartment constitutes another major player in the TIME, and the inflammation associated with tumor and metastasis recruits high amounts of macrophages in the stroma, forming a sort of barrier to obstacle lymphocyte infiltration within the tumor mass [56]. In this regard, evidences in lung squamous-cell carcinoma showed a poor invasion of CD8+ T cells within tumor nests due to long-lasting interactions with TAM in the stroma. Depletion of such TAM restored CD8+ T cell infiltration into tumor islets improving the efficacy of anti-PD-1 immunotherapy [56]. Accordingly, we observed a higher density of macrophages in the stromal compartment and an elevated percentage of CD163+ TAM in close contact with CD8+ T lymphocytes in HPV-positive lesions. However, the prognostic role of such population appears different according to HPV status. Unexpectedly, in HPV-positive lesions the presence of TAMs within the tumor regions and their interactions with CTL has a positive role on patient DFS, probably reflecting an active host anti-tumor immunological response, or a direct role for macrophages in antitumor defence, as already reported in colon cancer [57]. Conversely, higher levels of TAMs and interactions between CD8+ T cells and TAMs negatively associated with the prognosis in HPV-negative patients, in line with previous studies [58, 59]. Taken together, our results highlight the rationale of combining approaches targeting TAMs [60] and immune checkpoint molecules to increase tumor surveillance by CD8+ T cells, and make HNSCC more responsive to anti-PD-1 treatment, particularly in HPV-negative patients.

Sexual dimorphism has been recently ascribed as a relevant factor for cancer incidence and survival [61], even though the role of sex hormones in HNSCC is still controversial and a topic of debate [62]. Evidences highlight the importance of patient sex in modulating the molecular mechanisms that drive the anti-tumor immune response [63]. Accordingly, TIME and levels of immune cell infiltration may differ in males and females with HNSCC, leading in turn to different responses to immunotherapy [64]. In this regard, we found that women had a stronger and more structured immune response in metastatic lesions, as highlighted by the higher abundance of CD8+ T_RM cells as well as by the higher percentage of contacts between tumor cells and CTL. On the other hand, OPSCC arising in women apparently develop also complex mechanisms of resistance to counteract such more efficient initial immune recognition and response, as revealed by the higher abundance of Treg cells, the higher expression of multiple checkpoint molecules, and the enrichment of interactions between inhibitory elements and CD8+ lymphocytes. Conversely, we found that the TIME of men secondary lesions was characterized by lower amounts of tumor cells expressing HLA class I, which could lead to a less efficient presentation of tumor neoantigens and potentially explain the poorer immune infiltration. Overall, our results are in line with previous studies performed in different type of malignancies [64‐66], and show meaningful sex-based differences in the landscape of OPSCC, as well as in mechanisms exploited by tumors to evade immune response. Importantly, we found remarkable sex-specific differences also in the prognostic value of TIME. As already demonstrated in other tumors [27, 66, 67], our findings suggest that a significant sex-based heterogeneity of response to different type of immunotherapy strategies could be observed in patients with OPSCC, and therefore sex may represent a critical variable in the choice of the optimal treatment for patients with this malignancy.

Finally, some considerations about the limitations of this work. First, there are potential biases due to the retrospective nature of the study, and results must be considered as hypothesis-generating only. Second, the sample size of our cohort is relatively small, and therefore further studies with more OPSCC patients stratified by HPV status are warranted to fully validate our findings. In this regard, however, we want to stress that such patient cohort is quite homogenous in terms of HPV status, sex and anatomical sub-sites of primary and secondary lesions. The third limitation is that the quantity of smoking/drinking is not defined, since we collected data from electronic medical records that usually do not report such data. Fourth, since the main aim of the study was to characterize the TIME both in primary lesions and related lymph node metastases, the present series consists only of patients undergoing upfront surgery that allowed to collect both types of specimens. While this makes the study population more homogeneous, the results relating to the impact of TIME on the outcome can not be extended to patients undergoing upfront (chemo)-radiotherapy. Finally, since only FFPE material was available, the evaluation of the HPV status was based on the double positivity for high-risk HPV-DNA and p16. Although detection of E6 and E7 mRNA would have provided more robust data, double positivity for HPV-DNA and p16 was observed to be the strongest surrogate marker for transforming HPV infection [68, 69].

Our study clearly establishes that remarkable differences exist in the immune infiltrate between “T-cell-inflamed” HPV-positive and “T-cell-non-inflamed” HPV-negative OPSCC, and that these features are conserved both in primary and metastatic lesions. Moreover, although TAMs and checkpoint molecules expression are generally regarded as immunosuppressive elements, they are not necessarily synonymous of tumor immune evasion and may reflect an ongoing antitumor immune response. Furthermore, we highlight sex-specific differences in the TIME composition of OPSCC and in their prognostic value. Based on these observations, our study provides the rationale for the integration of ICI in the loco-regional therapy strategies for patients with heavily infiltrated treatment-naïve OPSCC, and at the same time advances the notion that approaches combining ICI with tumor-specific T cell response inducers or TAM modulators might be beneficial also for the HPV-negative “cold” counterparts.