Tumoral HLA class I and APM expression and its effect on intratumoral T-cell reaction and outcome in HPV-negative oral squamous cell carcinoma Short running title: Tumor microenvironment and HLA-I expression in HPV-negative OSCC

Affiliations: 1 University Hospital Halle (Saale), Institute of Pathology, Halle (Saale), Germany 2 University Hospital Halle (Saale), Department of Oral and Maxillofacial Plastic Surgery, Halle (Saale), Germany 3 University Hospital Halle (Saale), Institute of Medical Immunology, Halle (Saale), Germany 4 Robert W. Franz Cancer Research Center, Earle A. Chiles Research Institute, Portland, Oregon, USA # These authors contributed equally to this work.


Introduction
As one of the eight "hallmarks of cancer", the interaction between immune-and tumor cells plays an integral role in controlling the initiation and progression of malignant diseases. A growing body of evidence of changes in the expression of immune modulatory molecules on tumor cells and on cells of the tumor microenvironment (TME) as well as of alterations in the immune contexture, in particular the nature, composition, density, localization, and function of immune cell subpopulations, soluble and physical factors, continuously leads to a better understanding of the tumor immune surveillance and immune escape [1]. The characterization of these mechanisms will give insights into the development of innate and acquired resistances to immunotherapies [2,3], which is crucial for improving the efficacy of immunotherapies for the treatment of cancer patients [4,5].
In murine experimental models as well as in human malignant tumors a number of immune escape strategies, resulting in T cell tolerance and loss of T cell recognition, has been described. These include (i) lack or downregulation of tumor antigen expression, (ii) loss or reduced expression of HLA-I surface molecules due to impaired expression of APM components, (iii) increased expression of immune suppressive molecules, e.g. the programmed death-like receptor ligand 1 (PD-L1) and the non-classical HLA-G and HLA-E antigens, (iv) downregulation of the interferon (IFN) signal pathways, (v) activation of oncogenic signaling and (vi) alterations in the expression of inflammatory cytokines, metabolites and pH as well as (vii) transcription factors [6][7][8][9]. Despite a loss of heterozygosity and somatic mutations of HLA-I antigens have been described in different tumor entities [10][11][12], their frequency is relatively rare suggesting that the impaired HLA-I surface expression of tumors appears to be mainly due to a deregulation of HLA-I APM components, such as the transporter associated with antigen processing (TAP) subunits and low molecular weight proteins (LMPs), mediated by epigenetic, transcriptional and/or post-transcriptional control [9,13]. This can occur at each step of the complex HLA-I APM and affect the anti-tumoral T cell responses [14][15][16][17]. Furthermore, an immune suppressive TME mediated by e.g. cellular and soluble components as well as by the interaction of the intra-and peritumoral immune cells negatively interferes with the immune control and could influence the prognosis of tumor patients and their response to (immuno)therapies [14,18].
Despite recent treatment advances, oral squamous cell cancer (OSCC) is with 700.000 new cases and 380.000 deaths worldwide [19] still one of the leading causes of cancer worldwide with an overall poor prognosis [20,21]. In industrialized countries, it is mainly attributed to smoking and alcohol abuse and is in about 5% of cases associated with human papilloma virus (HPV) infection [22]. Early disease is treated by either surgery or radiotherapy, while for recurrent/metastatic disease palliative chemotherapy is currently the standard of care [23].
Despite advances in cytotoxic therapies and surgical intervention, the prognosis of patients is still limited with only little improvement of their outcome over the last decades [24]. Cancer immunotherapies, in particular immune checkpoint inhibitors (CPI), have been shown to increase the overall survival (OS) of OSCC patients, which led to the FDA approval of the anti-PD-L1 antibody Nivolumab for their treatment [25,26]. Despite these promising results, only transiently responses or durable responses in a limited number of patients are generated, while the majority of OSCC patients progress and die of their disease [27]. Thus, there is an urgent need to identify prognostic and predictive markers and novel therapeutic targets. Emerging evidence of an enriched immune landscape with key immunological features and prognostic relevance in OSCC exists, which was confirmed by in silico analyses of TCGA data [28]. In colorectal cancer (CRC) the quantification of adaptive immune cells composed of CD3 + T lymphocytes with a cytotoxic (CD8) and memory (CD45RO) phenotype located at the invasive margin (IM) and tumor center (TC) called "immunoscore" was established as a risk predictor more powerful than TNM and histologic grading in this context. The use of the "immunoscore" as a consensus biomarker [29,30] and as a tool to distinguish between responders and nonresponders to CPI therapies [31] might be also useful for OSCC patients.
Concerning HPV -OSCC, our group recently proposed a more sophisticated immunoscore termed "cumulative suppression index" (CSI). By evaluating the density and geography of the TME, in particular the spatial distance of CD3 + /CD8 + T cells to CD3 + /CD8 -/FoxP3 + T cells, the CSI was of prognostic relevance in large tumors independent of well-established prognostic factors and it had a strong effect on the patient's OS [32]. However, OSCC cells have developed different strategies to counteract immune recognition involving both intrinsic cancer related factors and extrinsic mechanisms, like an immune suppressive TME and the downregulated expression of HLA-I APM surface antigen components recognized by tumorspecific T cells due to the reduced expression of HLA-I APM components [6]. Mapping these escape mechanisms might help to develop strategies to overcome this tolerance [33].
Therefore, the expression of HLA-I HC, β2-m and four distinct APM components (TAP1, TAP2, LMP2 and LMP10) was determined in 160 HPV -OSCC using immunohistochemistry (IHC) and qPCR, respectively. These data were correlated to tumor specific characteristics, the composition of the TME and the patients' outcome and were further validated by evaluation of corresponding data of The Cancer Genome Atlas (TCGA) from 103 HPV -HNSCC cases.

Clinicopathologic data
A comprehensive overview of the clinicopathologic data and the OS of the 160 HPV -OSCC patients included in this study is provided in Table 1A. Gender and age had no influence on the survival, while the T stage and the presence of metastases significantly correlated with a poor survival. These data were confirmed by TCGA cohort analysis of 103 HPV -HNC, where no prognostic differences concerning gender and age of the patients were reported. Instead, a better survival was found for HPV -HNC patients at low N stage. Unexpected, but in line with other studies, the WHO tumor grading had no prognostic significance in both collectives (Table   1A) [34].
In Table 1B, information concerning therapy regimen is provided demonstrating that locally extended tumors received radiation therapy, while adjuvant standard chemotherapy was administered in all cases with tumor dissemination irrespective of tumor size, but in particular in patients with a T1/2 tumor. The frequency of radiation therapy (%) including the median dosage as well as the frequency of standard chemotherapy (%) of patients in different T stages is given.

HLA-I/APM expression in OSCC tumors
In order to determine the expression levels of HLA-I/APM components in OSCC, protein expression of HLA-I HC, β2-m and the APM molecules TAP1, TAP2, LMP2 and LMP10 were analyzed by IHC of formalin-fixed, paraffin embedded (FFPE) tumor lesions [35]. IHC staining revealed at least slightly reduced, but heterogeneous cytoplasmic, membranous and / or nuclear expression levels depending on HLA-I and APM components analyzed when compared to the surrounding non-malignant cells (data not shown). Phenotype III expressed discordant levels of HLA-I HC/β2-m and high levels of APM components (HLA-I discordant high/low /APM high ). Phenotype IV with discordant HLA-I and low APM expression was omitted from further analyses due to low the group size (Table 2A). It is noteworthy that the three HLA/APM phenotypes were uniformly distributed over the different T stages (Table 2B). As the HC10 antibody only recognizes the HLA-I HC, corresponding fresh frozen OSCC tissues from five FFPE OSCC lesions with phenotype I and five OSCC lesions with phenotype II were stained with the HLA-A/B/C antibody W6/32 recognizing the functional trimeric HLA-I HC/β2-m and peptide complex [36]. In these selected cases, the expression levels of HLA-I HC together with β2-m staining were comparable to those obtained by employing the W6/32 monoclonal antibody (mAb). These findings suggest expression of the functional trimeric HLA-I complex at least in the OSCC cases analyzed (Figure 1, Supplementary Table 1). As a further control, HLA-I HC and ß2-m mRNA expression was analyzed in a subset of OSCC specimen also demonstrating concordant expression profiles when compared to the protein data (data not shown).

Impact of tumoral HLA-I/APM expression on the intratumoral immune cell reaction
To elucidate the impact of OSCC HLA-I/APM expression on the immune cell repertoire, the HLA-I HC and/or β2-m expression data was correlated to the frequency and composition of intratumoral T cell subsets recently acquired by multispectral imaging within the same sample collective [32]. In this approach, the intensity of tumoral HLA-I HC/β2-m surface expression correlated highly significant positively with the density of intratumoral CD8 + T cells (p=0.008 resp. 0.003), but negatively with the number of CD8 + T cells adjacent to FoxP3 + cells within a radius of 30 µm (p=0.009 resp. 0.007) (Table 3A). Furthermore, a positive correlation between the expression of LMP2 and the density of CD8 + T cells was detected (p=0.003). A similar distribution of the intratumoral T cell subsets was detected within the different tumor stages (Table 3B).
In silico analyses of TCGA data were in line with these findings as the mRNA expression levels of all HLA-I/APM genes were correlated positively with the quantity of the T cell infiltrate, in particular demonstrating a significant relation of CD8A with the HLA-I HC, β2-m, TAP1, TAP2, LMP2 and LMP10 (Table 3C, Suppl. Figure 1).    Here, an increase in HLA-I HC and β2-m expression negatively affected the OS independent of the tumor size. We next analyzed whether this effect was influenced by the T stage or, more specifically, the tumor mass. When considering T1/2 vs. T3/4 tumors, no significant effect was found (Table 4A). In contrast, by solely focusing on T2 tumors, a significant worse OS and 4.  (Figure 2).
Based on these data, we assume that the different results of T1 versus T2 tumors were due to the much better OS of patients with T1 tumors independent of the immune phenotype (Table   1).
In line with these data, in silico analyses employing the XENA database (https://xenabrowser.net/) revealed a reduced OS in HLA-I high /APM high HNSCC in comparison to HLA-I low /APM low HNSCC lesions. In this study, higher levels of TAP1 expression correlated with a better OS (Table 4B).

Discussion
It has become evident that an altered expression pattern of HLA-I antigens and HLA-I APM components in tumors represents an important immune escape mechanism from CTLmediated elimination in distinct tumor types, which is frequently associated with poor prognosis and resistance to immunotherapy, and has been linked to changes in the TME [37][38][39][40]. In HNSCC, an immune suppressive equilibrium has been identified, which is characterized by a low frequency of dysfunctional immune effector cells and an altered expression of immune modulatory molecules of HNSCC cells [41]. In HPV + HNSCC, the distinct underlying immune escape mechanisms have been well identified [7,42], while there is still a lack of data concerning immune evasion of HPV -OSCC lesions. However, recent evidence suggested that to revert HLA-I downregulation [5]. In this context, it is important to consider that a striking difference between the HLA-I/APM phenotypes was obvious for the FOXP3:CD8 ratio, which was more than threefold higher in phenotypes II and III compared to phenotype I.
Thus, our experimental data suggest that the immunological control of phenotype I tumors might be associated with patients´ outcome. Unexpectedly, the correlation of the HLA-I/APM phenotypes to clinical data revealed that the HLA-I high /APM high subgroup despite a high intratumoral CD8 + frequency and a rather low FOXP3:CD8 + ratio was linked to an adverse outcome, which was most significant in T2 disease, while due to their excellent prognosis, this effect was not detected in T1 tumors. This is in line with a number of studies demonstrating at least for high levels of β2-m a correlation with a poor prognosis [46,47]. Furthermore, experimental models suggest that β2-m overexpression is associated with a poor survival due to increased migration and invasion of OSCC cells [48].These data considerably extend our knowledge concerning the prognostic value of different components of the immune cell infiltrate. In this context, it is important to mention that next to melanoma, non-small cell lung cancer and CRC, in which a relationship between immune cell activity, HLA-I antigens, clinical outcome as well as response to CPI has been proven [18,37,49], our group recently showed a significant prognostic value of the presence and quantity of tumor infiltrating lymphocytes (TIL) in OSCC [32]. Taken the data of this study into account the established, prognostic relevant immune suppression index including the parameters CD8, FOXP3 distance to CD8 and PD-L1 expression has to be extended at least by the factor HLA-I expression, since a significant adverse outcome in tumors with a HLA-I high /APM high phenotype and high intratumoral CD8 + cell content was found.
Assuming that the high levels of HLA-I HC and β2-m in the phenotype I OSCC tumors might be often due to the IFN-γ secretion of T cells, an explanation for the adverse patients´ outcome might be caused by a systemic exhaustion leading to dysfunctional T cells both in blood and in the TME as described by others [50]. It is noteworthy that in our study the functionality of intratumoral T cells was not analyzed, since this study focused on the HLA-I/APM expression pattern. In contrast to the general dogma, patients with a HLA-I discordant high/low phenotype had a significant better prognosis compared to the HLA-I high phenotype, which is in line with a report in CRC, in which HLA-A2-positive patients had a poorer tumor differentiation [51].

Standard and multiplex immunohistochemistry
Immunohistochemical analyses of tissue samples were performed as recently described [32].
Briefly, FFPE tissue probes were incubated with the following primary mAbs: NAMB (β2-m; room temperature with a washing step in between. Visualization was done with DAB and hemalaun as described above. Immunohistochemistry (IHC) results were semi-quantitatively evaluated utilizing the immune reactive score (IRS) as described by Remmele et al. [35]. In short, for evaluation of membranous, cytoplasmic and nuclear staining intensity (0-negative, 1-low, 2-moderate, 3strong positive) as well as percentage of stained cells (0-negative, 1-<10%, 2-10-50%, 3-51-80%, 4->80%) were evaluated and the IRS then calculated as the product of both ranging from 0-12. .Ten percent of all cases plus all cases presenting any technical challenges were independently evaluated by two pathologists (CW, DB) and co-reviewed to harmonize and ensure reproducibility of the scoring. One pathologist (DB) scored all the remaining cases.
For correlation analyses of the expression of HLA-I and APM components with the frequency and localization of immune cell subpopulations within the tumor, multispectral imaging data from the same patient cohort (n=108 with immune cell data) previously published by us were employed [32].

Bioinformatics
In silico analysis was performed using the R2 web tool (http://r2.amc.nl) in order to predict the correlation of HLA-I APM components, CD4, CD8, CD274 and FoxP3 with the mRNA expression of the different genes in HPVhead and neck cancer (HNC) patients. For this, the TCGA data set "Tumor head and neck squamous cell carcinoma" was chosen (n=520). HPV + HNC patients and patients with unknown HPV status were omitted from the calculations, while HPV -HNC patients (hpv_status_p16-negative, 103 patients) with data available for the respective markers were analyzed. The 2log expression ratio was compared and a linear regression was calculated. A p-value ˂ 1e -3 was considered as significant.
For the OS analyses, the XENA database (https://xenabrowser.net) was employed for the TCGA Head and Neck Cancer study (n=604). Only primary tumor samples from HPVpatients (hpv_status_p16-negative, 103 patients) were included. A p-value ˂ 0.1 was considered as significant.

Statistical analyses
To

RNA isolation, semi-quantitative and quantitative PCR
For evaluation of HLA-I/APM expression by the tumor cell lines as well as by a subcollective of OSCC with availability of FFPE and fresh frozen sections and specimen of normal oral mucosa total RNA was extracted using the NucleoSpin RNA extraction kit (Macherey-Nagel) according to the manufacturer's instructions and converted to cDNA using the cDNA synthesis kit and oligo-dT primer from Thermo Scientific. Target-specific primers (Supplementary Table   2) have been already described [8]. PCR was performed in a Rotorgene and using the

Conclusions
Immunotherapy is a promising tool for the treatment of different malignancies. However, the survival benefit is limited to a subset of patients. A better understanding of the complex interactions between tumor cells and the tumor microenvironment is urgently needed to improve overall response for cancer patients and improve prognosis towards long-term outcome under immunotherapy.
In this study, of the HLA-I and APM component expression was analyzed in a large cohort of OSCC tissues and correlated to the intratumoral immune cell infiltrate. These findings were associated to the clinical data of the patients. Unexpectedly, despite a high intratumoral CD8 + T cell content, tumors with high expression levels of HLA-I and APM components presented a rather unfavorable outcome, which was most significant in smaller (T1/T2) tumors. We conclude that an effective immunomodulatory treatment strategy in OSCC has to include the blockade of more than one immune regulatory axis.

Supplementary Material
Supplementary Staining patterns of HLA-I HC and β2-m on FFPE specimen with known HLA expression status (concordant membranous positivity) were compared to HLA-I surface expression on fresh frozen tissues using the W6/32 antibody recognizing the trimeric HLA-I complex.

Acknowledgements
We would like to thank Nicole Ott and Maria Heise from the Institute of Immunology for excellent secretarial help. Claudia Wickenhauser and Daniel Bethmann would like to thank the members of the Institute of Pathology`s, histology laboratory for their comprehensive help in this study.

Conflict of interest statement
BAF discloses consulting activities and/or research support related to multiplex IHC from Akoya Biosciences/PerkinElmer, Bristol-Myers Squibb, Definiens/AstraZeneca, Macrogenics, OncoSec and Ultivue. The remaining authors declare that there exist no conflicts of interest for any of the authors.

Ethics approval
The study was approved by the Ethics Committee of the Medical Faculty of the Martin Luther University Halle-Wittenberg, Halle, Germany (#2017-81).

Funding
This work was sponsored by the Deutsche Forschungsgemeinschaft project DFG 585/22-1 (BS) and the Mildred Scheel Foundation grants 110703 and 111091 (BS).