Profiling dendritic cell subsets in head and neck squamous cell tonsillar cancer and benign tonsils

Dendritic cells (DCs) have a key role in orchestrating immune responses and are considered important targets for immunotherapy against cancer. In order to develop effective cancer vaccines, detailed knowledge of the micromilieu in cancer lesions is warranted. In this study, flow cytometry and human transcriptome arrays were used to characterize subsets of DCs in head and neck squamous cell tonsillar cancer and compare them to their counterparts in benign tonsils to evaluate subset-selective biomarkers associated with tonsillar cancer. We describe, for the first time, four subsets of DCs in tonsillar cancer: CD123+ plasmacytoid DCs (pDC), CD1c+, CD141+, and CD1c−CD141− myeloid DCs (mDC). An increased frequency of DCs and an elevated mDC/pDC ratio were shown in malignant compared to benign tonsillar tissue. The microarray data demonstrates characteristics specific for tonsil cancer DC subsets, including expression of immunosuppressive molecules and lower expression levels of genes involved in development of effector immune responses in DCs in malignant tonsillar tissue, compared to their counterparts in benign tonsillar tissue. Finally, we present target candidates selectively expressed by different DC subsets in malignant tonsils and confirm expression of CD206/MRC1 and CD207/Langerin on CD1c+ DCs at protein level. This study descibes DC characteristics in the context of head and neck cancer and add valuable steps towards future DC-based therapies against tonsillar cancer.

specializations of DC subsets that contributed to the overall outcome, an observation that remains to be confirmed with clinically relevant antigens 8 .
The plasticity of the DC subsets, their unique gene expression profiles and specialized functions, which depend on the local microenvironment, complicate targeting of DCs 12,13 . For instance, CD123 + plasmacytoid DCs (pDCs), CD1c + myeloid DCs (mDCs), and CD141 + mDCs, three well-characterized DC subsets in benign tonsils 14 , show an overlapping level of functionality in terms of cross-presentation [15][16][17] . For example, on one hand, TLR9-activated pDCs have been shown to initiate a systemic immune response through triggering sequential activation of NK cells, mDCs, and CD8 + T cells upon direct administration in mouse melanoma tumors 18 and, on the other hand, they have also been shown to contribute to tolerance development 19 . Similarly, mDCs have been shown to exhibit both tolerogenic and immunostimulatory properties depending on the prevailing micromilieu 20 . Thus, candidate targets that are DC subset specific with well-studied functional properties in the tumor environment are warranted to create desirable downstream immunological effects.
The aim of this study was to categorize DC subsets in tonsillar cancer, a HNSCC localized to the oropharynx, compared to benign tonsils, in order to outline immunological differences and to map out unique subset-specific expression of markers that may be potential targets in the development of future immunotherapeutics. We identified four DC subsets in tonsillar cancer through flow cytometry and compared the frequency of each tumor-resident subset to its counterpart in benign tonsils. We further took advantage of human transcriptome arrays to investigate gene expression in all the identified DC subsets, resulting in panels of surface markers selectively expressed by the individual DC subsets in malignant tonsillar tissue. Furthermore, we verified selective surface expression of CD206/MRC1 and CD207/Langerin on CD1c + mDCs from malignant tonsillar tissue. The data sheds light on DC subsets' characteristics in the cancer microenvironment and highlight markers that may be used for targeting specific DC subsets.
Subset transcriptome comparison in tonsillar cancer and benign tonsils. DC subsets from malignant (n = 4) and benign tonsillar tissue (n = 4) were sorted to a purity exceeding 95% (data not shown) and RNA from each DC population was hybridized to Affymetrix Human Transcriptome Pico Assay 2.0 microarrays. Pools of 4,817 and 1,538 differentially expressed transcripts among the four subsets in benign and malignant tonsillar tissues, respectively, were identified by the filtering strategy. Principal component analysis (PCA) of the differentially expressed transcripts demonstrated separate clustering of the four subsets in benign as well as in malignant tonsillar tissues (Fig. 2). CD123 + pDCs were observed to cluster further away from the other three subsets, likely due to their different origin ( Fig. 2A and B). Additionally, a lower number of differentially expressed transcripts observed in malignant tonsillar tissue indicated a higher degree of transcriptional similarity among DC subsets in tonsillar cancer as compared to benign tonsils, which may reflect that the tumor microenvironment has a strong influence on the expression profiles of the DC subsets.
Characterization of DC subsets in tonsillar cancer. Pathway enrichment analysis was performed on the transcripts differentially expressed by DC subsets in malignant compared to benign tonsils, and the top ten pathway enrichment hits were selected (Tables 1-4) revealing mechanistic information linked to ongoing immune responses. For instance, the transcripts expressed at a higher level in CD1c + mDCs in malignant compared to benign tonsils were enriched in pathways such as "Immune response-IL10 signaling pathway" and "Immune response-T cell co-signaling receptors" (including PD-L1, PD-L2, LAG3). CD123 + pDCs, CD141 + mDCs and the DN DCs in malignant tonsils display enriched transcripts in Interferon (IFN) α/β signalling via JAK/STAT pathway. Similarly, the "Immune response_Antiviral actions of Interferons" pathway was present in the higher expressed profiles of CD123 + mDCs and DN DCs. "PGE2 pathways in cancer" pathway was observed for the higher expressed profile of CD1c + mDCs and the DN DC population.

Subset selective gene expression and confirmation of CD206 and CD207 surface expression.
Plasma membrane and cell surface associated transcripts that were selectively expressed by individual subsets within malignant tonsils are visualized in heatmaps (Fig. 3). A variety of molecules including CLRs, TLRs, adhesion molecules, receptors, inhibitory molecules, enzymes, and transporters were identified, representing candidate molecules for selective targeting of DC subsets. Given the small sample sizes of malignant biopsies available and the complex multicolor staining for flow cytometry, a wider phenotypic analysis was not technically feasible, yet selective surface expression of CD206/MRC1 and CD207/Langerin by the CD1c + mDC subset was confirmed in malignant and benign tonsillar tissues (Fig. 4).

Discussion
DCs regulate both humoral and cellular functions of the immune system and are therefore valuable targets for therapeutic purposes. In a context of cancer, clonal expansion of CTLs is warranted, which can be facilitated by instructions to specific DC subsets to cross-present antigen to T-cells. The purpose of this study is to provide a descriptive account on DC subsets' characteristics in the tonsillar tumor microenvironment. Accordingly, we describe, for the first time, CD123 + pDCs, CD1c + mDCs, CD141 + mDCs, and DN mDCs in tonsillar cancer and compare their transcriptional profiles to each other and to their counterparts in benign tonsils. The results are of relevance to attempts to develop DC-targeting immunotherapies for tonsillar cancer.
In this study, flow cytometry analysis demonstrated that CD123 + pDCs as well as CD1c + and CD141 + mDCs, previously shown to be present in benign tonsillar tissue 14 , are also present in malignant tonsillar tissue. In addition, an HLA-DR + Lin − CD11c + DC subset that lacks expression of CD1c and CD141 was identified, here referred to as DN mDCs. Attention should be paid to the fact that the CD1c + mDCs in tissue may constitute a rather heterogeneic subset including migratory DCs, resident DCs and perhaps also inflammatory DCs derived from monocytes, and these are not further distinguished in this data set 21 . An overall higher frequency of DCs, out of CD45 + leukocytes, was observed in malignant as compared to benign tissue, suggesting that DC targeting may be feasible in tonsillar cancer. In addition, CD123 + pDCs were less frequent in malignant tissue, which could be advantageous to immunotherapy considering the role of this subset in tolerance development 19,22 . In contrast to CD123 + pDCs, the level of DN mDCs was higher in malignant as compared to benign tonsils. A similar DN mDC population, with dendritic cell-resembling profile, has been recently identified in human blood 23 . The   23 . Further investigations are warranted to understand the origin and functional role of DN DCs. Selective expression of several chemokine receptors was identified by our transcriptome profiling of the subsets in malignant tonsil, suggesting possible mechanisms for DCs to infiltrate the malignant tonsils. For example, CD123 + pDCs were found to selectively express CCR2, CMKLR1, and CXCR3, receptors known to be involved in DC maturation, trafficking and recruitment at tumor sites [24][25][26] whereas XCR1 and CX3CR1, were selectively expressed by CD141 + mDCs and CD1c + mDCs.
The selective expression of some inhibitory molecules by specific DC subsets from tonsillar cancer tissue suggests that these subsets may have an immunosuppressive profile. For instance, ILT4 (LILRB2), CD101, and PD-L1 (CD274) were shown to be selectively expressed by CD1c + mDCs and BTLA by CD141 + mDCs. BTLA engagement on DCs has been reported to actively tolerize T cell responses 27 . Additionally, CD101/ILT4 duet on CD1c + mDCs may have a critical role in diminishing T cell responses, as triggering CD101 results in IL-10 production that selectively upregulates ILT4, leading to T cell hypo-responsiveness 28,29 . In further support of an immunosuppressive profile, the IL-10 signaling pathway was an enrichment hit for CD1c + mDCs in tonsillar cancer. Furthermore, a higher expression level of PD-L1, PD-L2, and LAG3 in CD1c + mDCs from malignant compared to benign tonsils, as the corresponding pathway analysis suggests, may be an additional reflection of CD1c + mDC suppression, as a role of PD-L1 + DCs in induction of regulatory T cell and tolerance has been shown 30 .   In this study, higher expression of angiogenesis related transcripts in the malignant CD1c + mDC subset, such as VEGF-A, involved in the "Transcription_ HIF-1 targets" and in the "PGE2 pathway in cancer" pathways, may provide a further indication of immunosuppression produced by this particular DC subset. It has previously been reported in an in vivo ovarian cancer model that tumor infiltrating DCs have the ability to undergo endothelial-like differentiation, promoting tumor vasculogenesis, due to VEGF-A overexpression by DCs 31 . Further, in vivo studies of a breast cancer model have shown that a combinatorial treatment involving HIF-1 inhibitors, reducing the VEGF-A and FoxP3 expression, and a DC-based vaccine, gave rise to a strong anti-tumor cytotoxic T cell immune response with an increased IFN-γ production and reduced Treg cell activation 32 .
Despite observing a higher expression of immunosuppressive molecules by specific DC subsets in malignant compared to benign tonsillar tissue, it is important to highlight the higher expression of transcripts by CD123 + pDCs, CD141 + mDCs and the DNs, associated with IFN α/β signalling pathways, which may be linked to antiviral activity of these subsets towards HPV. It is well known that pDCs and mDCs produce Type I IFNs in response to viral infection 33 and treating tumors with IFNs has shown some promise 34 . Moreover, many of the transcripts expressed at lower levels in the DC subsets in malignant tonsillar tissue are mostly enriched in pathways related to development of an effective immune response. For instance, "Immune response. Antigen presentation by MHC class II" was enriched by transcripts with lower expression level in CD1c + DCs in malignant tissue. Interestingly, diminished Th1 response has been shown in patients with advanced HNSCC as compared to less advanced stages 35 . Also, complement C5a signaling was found to be expressed at lower levels, which may indicate a lower presentation capability of this subset 36,37 . Another interesting finding concerns IL4, as several transcripts involved in IL4 signaling were expressed to a lower degree in malignant tissue by CD1c + DCs and CD123 + pDCs compared to those in benign tissue. Previous studies have shown that IL4 can abrogate DC suppression in the tumor microenvironment 38 , as well as activate tumor-infiltration of DCs, in a STAT-6 dependent manner, to promote type-1 T cell responses 39 . These observations suggest an underlining immunosuppressive environment based on inhibitory molecules and a lower expression of genes involved in development of effector immune responses. Further, IFN related observations may describe an insufficient immune anti-tumor/viral response in the microenvironment.
This study identified selectively expressed surface markers on specific DC subsets in tonsillar cancer, and CD206/MRC1 and CD207/Langerin on CD1c + mDCs were confirmed at protein level. Targeting these particular   receptors may efficiently induce cross-presentation of antigen and activation of CD4 + and CD8 + T cells 9,40 . Furthermore, out of the surface markers selectively expressed by CD141 + DCs, XCR1 is considered a promising target for antibody-based DC vaccines 41 . As examples of markers expressed by CD123 + DCs, targeting EPHA2 may produce anti-tumor effects 42,43 , while blocking of IL28RA result in augmented vaccine responses 44 . Information on DNs is lacking in the context of immunotherapy, but our analysis on their selective markers introduces candidates that, following functional studies, may emerge as immunotherapy targets. Taken together, selectively expressed molecules presented from this study may be considered as potential targets in future immunotherapy strategies 9,45,46 . Other receptors such as CLEC4a, DEC-205, and DC-SIGN, although not exclusively expressed on specific DC subsets, may still represent attractive targets 41,47,48 , depending on therapeutic strategy. Further studies are warranted to assess candidate marker protein expression and their efficacy in priming immune responses in tonsillar cancer.
In conclusion, this study describes four DC subsets in tonsillar cancer and their selective expression of surface targets. Our observations suggest overall immunosuppressive DC features, based on expression of inhibitory molecules and as well as of low expression of genes involved in development of immune effector responses, but observations on IFN may nevertheless suggest an existing yet insufficient immune response in tonsillar cancer.

Materials and Methods
Patients and sample preparation. Biopsies from tonsillar cancer lesions were obtained at Skåne University Hospital (Lund, Sweden). Benign tonsils were obtained from patients with tonsillar hypertrophy or recurrent tonsillitis (without signs of ongoing infection) undergoing tonsillectomy at Ängelholm and Helsingborg Hospitals (Sweden). It is confirmed that all research was performed in accordance with relevant guidelines/regulations, approved by the Regional Ethics Committee (EPN-Regionala Etikprövningsnämnden i Lund) and informed consent was obtained from all participants and/or their legal guardians. The tonsillar cancers were all tested positive for p16 expression, suggesting an association with human papilloma virus (HPV). The biopsies were cut into small pieces in RPMI 1640 medium (Thermo Fisher Scientific, Bremen, Germany) supplemented with 0.1 mg/ml gentamycin (Sigma Aldrich, St Louis, MO). Enzymatic digestion was then performed by incubating the tissue suspension at 37 °C for 20 minutes with Collagenase IV (Sigma Aldrich) (2 mg/ml) and DNase I (Sigma Aldrich) (200 Kunitz units/ml). Thereafter, single cell suspensions were prepared by filtering through 70 µm cell strainers (BD Biosciences, San Jose, CA). Cell viability was evaluated through Trypan blue exclusion.
The gating strategy is shown in Fig. 1A. The four identified subsets were then sorted out of malignant (n = 4) and benign (n = 4) tonsillar tissue (purity was routinely ensured to be >95%) and populations were collected in 1 ml of 100% FBS. Different sorted subset fractions (n = 32), ranging from 32 to 57035 cells, were then kept in Trizol (Life Technologies, Carlsbad, CA) and stored at −20 °C before being subjected to RNA extraction procedure.
cRNA preparation, hybridization and upstream microarray data analysis. Total RNA amplification and labeling was performed using the Ovation Pico WTA system V2 (NuGEN, San Carlos, CA) and Encore Biotin Module Kit (NuGEN), respectively. 5 µg of the labeled RNA from the respective samples was hybridized to GeneChip Human Transcriptome Pico Assay 2.0 microarrays (Affymetrix, Santa Clara, CA) and generated signals were finally analyzed through Affymetrix GeneChip Scanner 3000 7 G platform. Microarray signal intensities were normalized through log2-transformed Robust Multi Array (RMA) algorithm on Expression Console 1.4.1.46 (Affymetrix). Two samples (one tonsillar cancer CD141 + mDCs and one benign tonsil CD1c + mDCs) were removed from the data set due to low amount or diminished integrity of RNA.
Transcriptional characterization of subsets in tonsillar cancer and benign tonsils. Distinctive pools of differentially expressed transcripts were identified among DC subsets in benign and malignant tonsils through applying a strategy to reduce the false discovery rate (FDR). As a brief example, significantly different transcripts exhibiting ≥1.2-fold change (FC) in expression levels (based on log2-transformed data, corresponding to ≥2-fold of non-transformed data) were identified through two-tailed t-tests comparing each subset's transcriptome in malignant tonsils against all other three subsets' in the same tissue. Filtered transcripts were subsequently added together to form a pool of differentially expressed transcripts among DC subsets. To remove background noise, an intensity cut-off in expression levels was applied (raw signal 100, corresponding to 6.64 in log2-transformed intensities). Two-tailed t-tests (p < 0.05) and FC filtering were performed in Omics Explorer 3.2 (Qlucore, Lund, Sweden). Thereafter, Principal Component Analysis (PCA) was performed on Qlucore Omics Explorer 3.2 to visualize the replicate similarities and their inter-subset relationships. Profiling of DC subsets in tonsillar cancer. To identify the transcripts selectively expressed at a higher level in individual subsets, the transcriptome of each subset in malignant tissue samples was compared to the other three subsets. Common transcripts among all the comparisons that were significantly different (one-tailed t-tests, p < 0.05) and over-expressed (FC ≥ 1.2) were considered as selectively expressed to a higher level by individual subsets. Transcripts were subsequently annotated using "Affymetrix tag ID (exon)" and transferred into MetaCore, version 6.28 build 68602, (Thomson Reuters, New York, NY). The lists of transcripts were subjected to localization gene ontology (GO) enrichment and all transcript annotations enriched in "Plasma membrane" and "Cell surface" localizations were individually reviewed. All transcripts associated with either endocytosis, cross-presentation, or signaling activities were then selected and consequently resulted in lists of markers specific for different DC subsets in malignant tonsillar tissue, which could be potential molecules for selective targeting of a specific DC subset in tonsillar cancer. The same analysis was performed to provide the selectively lower expressed transcripts by individual subsets in malignant tonsil tissue.
Transcriptional comparison of DC subsets in malignant and benign tonsillar tissue. First, differentially expressed transcripts in DC populations in tonsillar cancer, compared to the DC counterparts in benign tonsils, were filtered using one-tailed t-test (p < 0.05) and ≥1.2-fold change in expression levels based on log2-transformed data. The lists were then imported into MetaCore and annotated using "Affymetrix tag ID (exon)" as described previously. Pathway enrichment analysis was performed on the corresponding transcripts and the top ten pathways enriched were reviewed. A q-value of 0.05 was considered as the significance cutoff to control the FDR of the enrichment analysis (pathways with FDR > 0.05 were still extracted and is marked with an asterisk in Table 2-4).

Confirmation of marker expression using flow cytometry.
To evaluate the validity of some of the selectively expressed markers, cell surface expression levels of CD206/MRC1 and CD207/Langerin were assessed using flow cytometry through a similar gating strategy to the one presented in Fig. 1. Fluorescence minus one (FMO) controls were applied and net percentage positive cells was calculated by subtracting positivity for the FMO control. Results were analyzed using FCS Express v4.0 (De Novo Software, Los Angeles, CA).
Statistical analysis of flow cytometry data. Differences in mean percentages DCs out of CD45 + leukocytes, as well as in mean percentages of DC subsets out of DCs, between malignant and benign tonsillar tissue were assessed using Mann-Whitney test. Wilcoxon matched-pairs signed rank tests were used to analyze the differences between pDC and mDCs frequencies in tonsillar cancer and benign tonsils, respectively. Moreover, the frequencies of subsets expressing CD207/Langerin or CD206/MRC1 were separately compared through Wilcoxon matched-pairs signed rank test in benign and malignant tonsils. The statistical analysis was performed in GraphPad prism 7.01 (GraphPad Software, La Jolla, CA) and p-values < 0.05 were considered significant.  Table 5. Antibody (Ab) panels applied for DC subset identification and PRR expression analysis.