Oncogenic Ras and ΔNp63α cooperate to recruit immunosuppressive polymorphonuclear myeloid-derived suppressor cells in a mouse model of squamous cancer pathogenesis

Introduction Amplification of human chromosome 3q26-29, which encodes oncoprotein ΔNp63 among other isoforms of the p63 family, is a feature common to squamous cell carcinomas (SCCs) of multiple tissue origins. Along with overexpression of ΔNp63, activation of the protooncogene, RAS, whether by overexpression or oncogenic mutation, is frequently observed in many cancers. In this study, analysis of transcriptome data from The Cancer Genome Atlas (TCGA) demonstrated that expression of TP63 mRNA, particularly ΔNp63 isoforms, and HRAS are significantly elevated in advanced squamous cell carcinomas of the head and neck (HNSCCs), suggesting pathological significance. However, how co-overexpressed ΔNp63 and HRAS affect the immunosuppressive tumor microenvironment (TME) is incompletely understood. Methods Here, we established and characterized an immune competent mouse model using primary keratinocytes with retroviral-mediated overexpression of ΔNp63α and constitutively activated HRAS (v-rasHa G12R) to evaluate the role of these oncogenes in the immune TME. Results In this model, orthotopic grafting of wildtype syngeneic keratinocytes expressing both v-rasHa and elevated levels of ΔNp63α consistently yield carcinomas in syngeneic hosts, while cells expressing v-rasHa alone yield predominantly papillomas. We found that polymorphonuclear (PMN) myeloid cells, experimentally validated to be immunosuppressive and thus representing myeloid-derived suppressor cells (PMN-MDSCs), were significantly recruited into the TME of carcinomas arising early following orthotopic grafting of ΔNp63α/v-rasHa-expressing keratinocytes. ΔNp63α/v-rasHa-driven carcinomas expressed higher levels of chemokines implicated in recruitment of MDSCs compared to v-rasHa-initiated tumors, providing a heretofore undescribed link between ΔNp63α/HRAS-driven carcinomas and the development of an immunosuppressive TME. Conclusion These results support the utilization of a genetic carcinogenesis model harboring specific genomic drivers of malignancy to study mechanisms underlying the development of local immunosuppression.


Introduction
Human squamous cell carcinomas (SCCs) are derived from epithelial cells and share features across originating sites including head and neck, lung, esophagus, cervix and skin (1,2). Vast datasets such as those available from The Cancer Genome Atlas (TCGA) Research Network have enabled bioinformatic analyses of cancers arising in these different tissue types (https://www.cancer.gov/tcga (3,4)). Genomic and transcriptional analyses by Pan-TCGA revealed that chromosome 3q gain is a common molecular signature across all SCCs, estimated at up to 69% (2,5,6). At the heart of the amplified region of 3q26-3q29, TP63, a master transcriptional regulator of epithelial cell fate, is expressed predominantly as the DNp63 isoform in SCCs (2,5,6). The DNp63 isoform has been shown to play major roles in the establishment and maintenance of epithelial cell lineage, proliferation, and adhesion as well as the inhibition of differentiation and senescence. These activities are dependent on its expression level in a context-dependent manner (7)(8)(9). When overexpressed as in SCCs, the DNp63 isoform, lacking the Nterminal transactivation domain of p63, has a dominant-negative effect on p53 function and regulated genes, while promoting transactivation of a distinct gene repertoire through interaction with other transcription factors (9)(10)(11).
We previously established a murine genetic SCC tumor progression model utilizing primary epidermal keratinocytes that are transduced with retrovirus encoding Harvey rat sarcoma virus oncogene, v-ras Ha , with activating mutation at G12R alone or in the presence of lentiviral-driven DNp63a, and orthotopically grafted onto athymic nude mice hosts (12). We observed that the overexpression of DNp63a in combination with oncogenic vras Ha enhances malignant conversion, in contrast to the development of papillomas observed with Ras alone (12). The role of oncogenic v-ras Ha in neoplastic transformation has been attributed to activation of downstream effectors of receptor tyrosine kinases, which establishes a pro-inflammatory environment (13,14). The cooperation of v-ras Ha and DNp63a in malignant conversion can be explained, in part, by the role of DNp63a in overcoming v-ras Ha -induced senescence by inhibition of p16 ink4a and p19 arf expression (12,15,16). In addition to the anti-senescent role of DNp63a in driving malignancy, mounting evidence supports a role of DNp63a in orchestrating inflammation mediated by its interactions with NF-kB subunits (reviewed in (17)). We previously identified that the overexpression of DNp63a induces nuclear localization and activation of the NF-kB subunit, c-Rel, and regulates inflammatory response genes in primary murine keratinocytes [(18); King and Weinberg, unpublished results]. Furthermore, in human head and neck squamous cell carcinoma (HNSCC) cell lines, cREL and DNp63 form a complex in response to the inflammatory cytokine TNFa to activate NF-kB and AP-1 pathways (19)(20)(21). In addition, transgenic (TG) mouse models of overexpressed DNp63a in epidermis display hyperplasia, infiltration of immune and inflammatory cell populations (21)(22)(23), and enhanced malignant progression of chemically-induced tumors (24). The immune and inflammatory cells identified in DNp63a-overexpressing hyperproliferative epidermis included CD3 + T cells, CD4 + T cells, CD4 + /CD25 + /Foxp3 + regulatory T cells (Tregs), and M2 type macrophages, indicating that cell subsets implicated in both pro-inflammatory and immunosuppressive functions are recruited by prolonged overexpression of DNp63a in the epidermis. This appears to be mediated by increased levels of pro-inflammatory cytokines regulated by NF-kB (22). The activated NF-kB signaling and DNp63 expression levels showed positive correlation in HNSCCs, which are also enriched in immune components based on genomic analyses (2,5,6).
Considering the prevalence of human SCCs with elevated levels of DNp63 and increased immune infiltrates, and the co-activation of NK-kB/c-Rel with DNp63a, we investigated how DNp63a impacts the tumor microenvironment (TME) and its relationship to carcinoma formation. We adapted the orthograft model to evaluate the contributions of v-ras Ha /DNp63a in the athymic mouse background described above to immune-competent syngeneic hosts to characterize the complete composition of immune infiltrates in the v-ras Ha or v-ras Ha /DNp63a-induced TME. Our data suggest that DNp63a and oncogenic v-ras Ha cooperate to establish an immunosuppressive TME that promotes carcinogenesis.

Animals
All animal work was performed in accordance with established NIH (National Institutes of Health) guidelines, following accepted standards of humane animal care under protocols approved by the Animal Care and Use Committee of the Center for Biologics Evaluation and Research of the Food and Drug Administration. Wild-type BALB/cAnNCr mice (BALB/c; strain code: 555) used to both establish syngeneic donor cell cultures and as grafting hosts were obtained from Charles River Laboratories, Kingston, NY.

Viruses and retroviral transduction of primary keratinocytes
A Y 2 retrovirus packaging cell line was used to introduce the Ha-MSV gene from Harvey murine sarcoma virus (single G12R mutation; v-ras Ha ) as previously described (27,28). Lentivirus construct encoding human DNp63a (LV-DNp63a) under the FerH promoter was described previously (12). The empty vector construct, also referred to as Stuffer control, contains the FerH promoter followed by multiple stop codons (with no start codons) and is thus unable to initiate transcription driven by the FerH promoter. The construct was purchased from Protein Expression Laboratory, Leidos Biomedical Research Inc., Frederick National Laboratory for Cancer Research (Construct name: 17506-M36-685). Lentiviruses were generated from the constructs and titered by Cellomics Technology, LLC. Three days post-plating (a day after the transduction of retrovirus encoding v-ras Ha ), the primary keratinocytes were incubated in fresh low calcium EMEM with lentivirus at 3x10 6 Titer Unit (TU) per 60 mm 2 dish (=1.4x10 5 TU/ cm 2 , a total number of cells estimated to be 1-2 x10 6 cells), and 4 mg/ ml of polybrene with MOI of 1.5-3 (final volume of 0.5 ml/60 mm 2 dish), for 3 hours at 37˚C with rocking every 20 minutes. Fresh medium was added at the end of the incubation to bring the total volume to 3.5 ml/60 mm 2 .

Grafting
Primary murine newborn epidermal keratinocytes were transduced with v-ras Ha and DNp63a or Stuffer as described above. After 9 days in culture (6 days post-transduction of keratinocytes with DNp63a), the keratinocytes and fibroblasts were trypsinized, collected, counted and aliquoted for grafting at 4x10 6 keratinocytes and 8x10 6 fibroblasts per mouse. The cells were deposited on the subcutaneous surface inside silicone domes that were implanted onto the mid-dorsum of the host (6-12 weeks old), as previously described (12). Both donor cells and hosts were of the wild-type BALB/cAnNCr strain. The domes were removed oneweek post-surgery, and the tumors and grafted sites were collected at the time points indicated.

Genomics analysis
The canSAR Black database was used to compare HRAS and TP63 isoform expression in different cancer types (32). Normalized isoform and gene expression data from The Cancer Genome Atlas (TCGA) were downloaded from firebrowse (http://firebrowse.org/), analyzed in R, and processed and visualized using Tidyverse (https://cran.r-project.org/web/packages/tidyverse/citation.html). Published single-cell RNA-seq (scRNA-seq) data were obtained from Puram et al. (33); processed expression data were downloaded from Gene Expression Omnibus (GSE103322) and subjected to log2 transformation after adding one to each value. Statistical analysis was perfomed in ggpubr (https://CRAN.R-project.org/ package=ggpubr). From this single-cell RNA-seq dataset, only tumors yielding 50 or more tumor cells were considered for analysis (10 tumors).

Cytokine expression by qPCR
A Custom RT 2 profiler PCR array (Qiagen) was used to profile mRNA expression of chemokines and their receptors in RNA samples isolated from tumors and from cultured primary keratinocytes. The assays were performed and analyzed according to the manufacturer's instructions.

In vitro chemokine and cytokine protein expression by dot blot array
Primary keratinocytes were sequentially transduced with viral vectors encoding v-ras Ha , and either DNp63a or Stuffer, as described above. Three days post-transduction of lentivirus-DNp63 or Stuffer, the cell culture medium was replaced with fresh medium; 24 hours later the culture supernatant was collected and immediately incubated with a dot blot antibody array at 4°C overnight (Mouse Cytokine Array C1000, Raybiotech) according to the manufacturer's instructions. The image was developed using Amersham ImageQuant LAS 4000.

In vitro cytokine quantitative immunoassay
Primary keratinocytes were sequentially transduced with viral vectors encoding v-ras Ha , and DNp63a or Stuffer, as described above. At three days and 13 days post-transduction of lentivirus-DNp63 or Stuffer, the cell culture medium was replaced with fresh medium and 24 hours later the culture supernatant was collected after centrifugation and frozen at -80°C until the day of the ProcartaPlex assay. Three independent biological experiments were performed. Cytokine assays were performed using multiplex bead-based kits for the indicated mouse cytokines per the manufacturers' instructions (ProcartaPlex Immunoassays, ThermoFisher Scientific, CA). A total of 4 cytokines were assessed: CXCL1, CXCL5, CCL2, CCL20. Fluorescence of beads was measured using a Luminex Bioplex 200 analyzer (Bio-Rad Laboratories, Hercules, CA, USA), and data analysis was performed using the BioPlex Manager software (BioHercules, CA. USA) based on a five-parametric logistic nonlinear regression curvefitting algorithm.

T cell proliferation assay
A T lymphocyte proliferation assay was performed as previously described (34). CD4 + and CD8 + T cells were isolated from naïve B6 spleens using the Pan T-Cell Kit (Miltenyi Biotec, negative selection) on an autoMACS Pro Separator (Miltenyi Biotec), labeled with a fluorescence dye 5 (6)-Carboxyfluorescein diacetate N-succinimidyl ester (CFSE, Sigma), and stimulated with plate-bound anti-CD3 (clone 145-2C11, eBioscience) and -CD28 (Clone 37.51, eBioscience) antibodies. T cells were co-cultured at a 1:2 ratio with putative MDSCs isolated from spleens or harvested from tumors derived from v-ras Ha /Stuffer (= empty vector) or vras Ha /DNp63 expressing keratinocytes. Granulocytic myeloid cells were isolated from spleens using the Anti-Ly6G Microbead Kit (Miltenyi Biotec, positive selection). To enrich tumor-infiltrating granulocytic myeloid cells, a 40/80% isotonic Percoll (Sigma) gradient (centrifuged at 325 × g for 23 minutes at room temperature) was followed by positive selection using the anti-Ly6G Microbead Kit. Flow cytometry was used to quantify CFSE dilution at 72-hours. Proliferation was quantified as the average number of divisions of all cells in the culture (division index) using commercially available FlowJo software v10.8.2 (35).

Statistics
Test of significance between pairs of data are reported as pvalues, derived using a student's t-test with a two tailed distribution and calculated at 95% confidence. Comparison of multiple sets of data was achieved with analysis of variance (ANOVA) with Tukey's multiple comparisons. All error bars indicate standard error. Statistical significance was set to p < 0.05. All analyses were performed using GraphPad Prism v7 unless otherwise indicated.

Upregulation of HRAS and TP63 expression in human squamous cell carcinoma
Large databases allow for analysis of common pathways and oncogenes aberrantly expressed across diverse cancer types. Using the canSAR database (http://cansarblack.icr.ac.uk (36,37)), which includes data from The Cancer Genome Atlas (TCGA), we analyzed TP63 and HRAS gene expression in multiple cancer types. Study of the TCGA data corroborates that both HRAS and TP63 expression are significantly elevated in advanced stage HNSCCs and early-stage lung squamous cell carcinoma (LSCC) compared to normal tissue ( Figures 1A, B). The expression of each of these genes in HNSCCs ranked the highest among the major cancer types analyzed, suggesting oncological significance. We further demonstrate that DNp63 isoforms are expressed to a greater degree than TAp63 isoforms in these cancer types ( Figure 1C), consistent with earlier reports (2,5).
A limitation of the application of bulk genomic data from TCGA is the inability to distinguish the heterogeneity that exists in gene expression across different cell populations within the TME. To evaluate the expression of TP63 and HRAS within individual cell types, we utilized previously published scRNA-seq data generated from primary HNSCC tumors (33). Data presented in Figure 2 indicate that TP63 and HRAS expression is generally greater in malignant epithelial cells compared to non-malignant cell populations, such as immune cells and stromal cells. These data indicate that the increased expression of TP63 and HRAS observed in bulk genomic data is likely due to increased expression in tumor cells, with limited contribution from immune or stromal cells.

Adaptation of the immune deficient orthotopic mouse model of SCC to a syngeneic immune competent host
We previously described an orthotopic murine graft model that uses primary epidermal keratinocytes transduced with retroviral vectors to overexpress oncogenic v-ras Ha and wildtype DNp63 with immune deficient athymic nude mice as hosts (12), to evaluate the contribution of DNp63 and v-ras Ha to squamous cancer pathogenesis. Overexpression of v-ras Ha in this model mimics the RAS activation in human SCCs ( Figure 1) by oncogenic mutation or overexpression of wild type gene. Likewise, lentiviral-driven DNp63 elevated expression of DNp63 mimics the gene amplification and overexpression of DNp63 observed in human SCCs. Mouse cutaneous SCC (cuSCC) models have been described to harbor molecular similarities and parallels not only to human cuSCCs but to SCCs arising from other tissues as well (38,39). Our orthograft model reflects the genetic alterations observed in human cancers of head and neck and lung (Figures 1, 2), and has served as a useful tool to decipher the implications of these genetic changes. Indeed, events associated with DNp63a overexpression that were identified in this cutaneous model, such as activation of NF-kB/c-Rel, have been confirmed in human HNSCC samples and cell lines (18). The observation that overexpression of DNp63a can induce an immune response in mice (21,22) further suggested that this orthograft system could be adapted to explore the full complement of immune components modulated during v-ras Ha -initiated tumorigenesis and DNp63a-dependent malignant conversion, as a model of human HNSCCs that frequently harbor amplified p63 and are often heavily infiltrated by inflammatory cells (40). We therefore adapted the athymic mouse model to an immunocompetent syngeneic background with BALB/c mice as hosts. As shown in Figure 3, orthotopic grafting of BALB/c primary epidermal keratinocytes that had been transduced with a retroviral vector encoding v-ras Ha   inflammatory response observed in human HNSCCs (21)(22)(23). To gain insight into whether these chemokines are differentially regulated in epithelial cell populations that give rise to benign versus malignant tumors, we used a commercially available cytokine array to examine the chemokines and cytokines produced in vitro by these keratinocyte populations. In this experiment, supernatants from cultured primary murine keratinocytes transduced with 1) empty vector alone ("Stuffer" B C A D FIGURE 3 In vivo grafts derived from engineered keratinocytes retain expression of exogenous v-ras Ha Table S1). Specifically, increased levels of CXCL1, CXCL2, CXCL5, CXCL7, CXCL16, CCL2, CCL20, IGFBP-3, MMP-3, and OPN were observed in the supernatant of v-ras Ha /Stufferand v-ras Ha /DNp63a-transduced cells. Many of these chemokines and cytokines are known to play a role in chemotaxis of immune and immunosuppressive cells, including myeloid-derived suppressor cells (MDSC), tumor associated macrophages (TAM), monocytes, and neutrophils (41)(42)(43)(44)(45)(46). Notably, there was no significant change in the cytokine profile between the control and DNp63atransduced cells. Relative protein levels of CXCL1, CXCL5, CCL2, and CCL20 were further evaluated using the ProcartaPlex method, with similar findings (Supplementary Figure 3C). To rule out the possibility that the method may not be sufficiently sensitive to detect small changes by DNp63a alone, we evaluated the mRNA levels of MDSC-and Treg-recruiting chemokines Cxcl1, Cxcl2, Cxcl5, Ccl1, Ccl17, and Ccl22 in keratinocytes expressing v-ras Ha and DNp63a either separately or together using RT 2 -custom PCR arrays (Supplementary Figure 4).
The data indicate that overexpression of v-ras Ha upregulates Cxcl1, Cxcl2, and Cxcl5, while DNp63a downregulates the expression of these chemokines (Supplementary Figure 4). Taken together, these data suggest that the enhanced chemokine/cytokine production was mainly driven by v-ras Ha expression in the in vitro setting.
In light of this finding, we evaluated whether these chemokines are similarly deregulated in vivo in the murine tumor context, using the same RT 2 -custom PCR arrays. We specifically focused on the expression of genes involved in immunosuppression during early establishment of the tumor (2 weeks post-grafting). As shown in Figure 4A, the mRNA levels of chemokine receptors on MDSCs, Cxcr1 and Cxcr2, as well as corresponding ligands, Cxcl1 and Cxcl5, are significantly increased in the tumors derived from v-ras Ha / DNp63a expressing keratinocytes compared to tumors expressing v-ras Ha in the absence of p63, or intact skin. In contrast, v-ras Hainitiated papillomas upregulated the expression of the Cxcr2 ligand Cxcl2 in comparison to intact skin or Ras/DNp63a carcinomas. The Cxcl1 and Cxcl5 expression levels were also increased in v-ras Ha tumors compared to control but to a lesser degree, with lower statistical significance compared to v-ras Ha /DNp63a tumors. Additionally, the mRNA levels of Treg chemokine receptors Ccr4, Ccr8, and Ccr10 as well as their ligands, Ccl1, Ccl17, and Ccl22 are significantly upregulated in the v-ras Ha /DNp63a carcinomas compared to v-ras Ha /Stuffer tumors or normal skin ( Figure 4B). These data support that DNp63a cooperates with v-ras Ha in vivo to A B FIGURE 4 In Vivo chemokines and chemokine receptors are highly expressed in Ras/DNp63 carcinomas relative to Ras initiated papillomas. RNA isolated from skin or lesions derived from grafted primary keratinocytes encoding v-ras Ha and Stuffer control (v-ras Ha /Stuffer) or v-ras Ha in combination with DNp63 (v-ras Ha /DNp63) at 2 weeks post grafting were analyzed for the expression of chemokines and chemokine receptors involved in the trafficking of myeloid cells (A) or Tregs (B) by qRT-PCR using a custom RT 2 qPCR profiler. Three tumors per group were tested and analyzed. Quantified as fold change relative to normal skin. *, p < 0.05; **, p < 0.01; ***, p < 0.001. Stuffer = Empty Vector.
promote the production of chemokines implicated in driving the recruitment of cells with Treg and MDSC markers and immunosuppressive function into the TME.

DNp63a expressing carcinomas have increased numbers of PMN-MDSCs recruited into the TME
To investigate whether these cytokine expression patterns correspond to distinct host immune profiles in papillomas relative to carcinomas, grafts of v-ras Ha -and v-ras Ha /DNp63a-expressing primary keratinocytes were harvested at 2, 3, and 4 weeks postgrafting and immune infiltration profiles were determined by flow cytometry. The tumors were screened for the presence of (CD11b + Ly6G hi Ly6C int ) polymorphonuclear (PMN)-like myeloid cells, CD11b + Ly6G lo Ly6C hi monocytic myeloid cells, CD8 + T-cells, CD4 + T-cells, and CD4 + CD25 + FOXP3 + regulatory T-cells (Tregs).
The immune landscape changed over a 4-week period during the development of the tumors, and there were notable differences in the immune profiles between tumors arising from grafts of v-ras Ha -expressing keratinocytes compared to those expressing v-ras Ha / DNp63a at 2, 3, and 4-weeks post-grafting (data not shown). The most notable differences in immune cell components were seen at week 2 post-grafting, therefore the experiment was repeated with tumors harvested at this peak timepoint. CD4 + and CD8 + T-cells can exert effector function and regulate tumor growth and are typically associated with good prognosis (47). As shown in Figure 5A, both vras Ha and v-ras Ha /DNp63a-induced tumors recruit more CD4 + Tcells and CD8 + T-cells compared to normal keratinocyte controls (P ≤0.01), suggesting that an immunoregulatory and effector T cell response is triggered by oncogenic v-ras Ha expression. A significantly increased number of CD8 + T-cells was observed in v-ras Ha /DNp63a tumors (P ≤0.05), albeit with a high degree of variability. However, Regulatory T-cells (Tregs) are recruited at~2-fold higher number to v-ras Ha (P ≤0.05) and v-ras Ha /DNp63a (P ≤0.001) tumors compared to normal keratinocyte grafts or intact skin, suggesting that oncogenic v-ras Ha can concurrently drive recruitment of Tregs implicated in immunosuppression into the TME. Further, the tumors arising from v-ras Ha /DNp63a-transduced keratinocytes also have significantly greater numbers of Ly6G hi Ly6C int PMN-like myeloid cells compared to grafts from v-ras Ha alone, control keratinocytes or intact skin, suggesting that the v-ras Ha /DNp63a had a significant impact on the concurrent recruitment of these potentially immunosuppressive neutrophilic cells. Lower numbers of Ly6G lo Ly6C hi monocytic myeloid cells (~100s/10,000 cells) compared to Ly6G hi Ly6C int PMN-like myeloid cells (300-1500s/10,000 cells) were observed across samples in both v-ras Ha and v-ras Ha /DNp63a tumors similar to control normal cell grafts. The number of NK cells recruited were also similar across primary keratinocytes, v-ras Ha /Stuffer, and vras Ha /DNp63a. Consistent with their role in innate immunity, NK cells were recruited to the wounding of the graft procedure independent of the oncogenes expressed. Multiplex immunofluorescence (multiplex IF) staining was used as an orthogonal method to visualize and determine the level of immune infiltrates in the tumors ( Figure 5B). Consistent with the flow cytometry results ( Figure 5A), v-ras Ha /DNp63a tumors have significantly increased numbers of Ly6G hi neutrophilic myeloid cells ( Figure 5C). Tumors stained for CD4 + , CD8 + , and FoxP3 + (Tregs) show no significant differences between the v-ras Ha and v-ras Ha /DNp63a tumors ( Figure 5C). Together, these data indicate that v-ras Ha /DNp63a-induced carcinomas recruit increased numbers of Ly6G hi PMN-like cells in the TME.
Together, the flow cytometry and IF data suggest that expression of DNp63 in SCC supports the induction of cells that bear PMN-MDSC markers as well as CD4 + and CD8 + T cell markers.
In mice, the phenotype of neutrophils is very similar to that of immunosuppressive neutrophilic myeloid derived suppressor cells (PMN-MDSCs), and the distinction between PMN-MDSCs and neutrophils requires functional assays (45). In order to distinguish PMN-MDSCs from neutrophils in this context, we determined whether neutrophilic populations isolated from tumors and spleen are capable of inhibiting T-cell proliferation. At the base line, cells without stimulation result in a single peak, indicating the absence of proliferation ( Figure 6A, top panel). When stimulated with antibodies to CD3 and CD28, CD4 + and CD8 + T-cells proliferate in the presence of non-specific control PBMCs (from splenocytes), indicated by the progressive dilution of CSFE dye after a few days ( Figure 6A, middle panel). However, in the presence of Ly6G hi cells isolated from the tumors, the extent of proliferation was significantly inhibited upon stimulation ( Figure 6A, bottom panel). Quantitation of the suppressive activity of tumor-and spleen-derived Ly6G hi cells indicate that both populations inhibit proliferation but to a different degree ( Figure 6B). The Ly6G hi cells from the tumors inhibited T cell proliferation to a significantly greater degree than peripheral Ly6G hi cells ( Figure 6B). These data B A FIGURE 6 Lesion infiltrating Ly6G hi cells are PMN-MDSCs. Ly6G hi myeloid cells were isolated from the spleens or tumors of mice bearing lesions derived from primary keratinocytes encoding v-ras Ha alone (v-ras Ha /Stuffer) or in combination with DNp63 (v-ras Ha /DNp63) and assessed for their ability to suppress the proliferation of CD3/28 stimulated wild-type CFSE-labelled CD4 + and CD8 + T cells in comparison to total splenocytes (PBMC). Proliferation was assessed by flow cytometric analysis. (A) Representative CFSE histograms of unstimulated T cells (top panels), T cells co-cultured with splenocytes (middle panels), or Ly6G hi cell isolated from a v-ras Ha  demonstrate that the neutrophilic cells that were recruited into vras Ha /Stuffer and v-ras Ha /DNp63a tumors are PMN-MDSCs.

Discussion
Among 33 cancer types analyzed by Pan TCGA, the TP63 and HRAS genes are significantly overexpressed in human head and neck and lung SCCs compared to normal tissues (Figures 1A, B). Both genes are overexpressed more significantly in the advanced stage HNSCCs, in which expression levels of both genes rank the highest among the major cancer types analyzed, suggesting pathological significance. The elevated TP63 gene expression resulting from 3q amplification and preferential expression of DNp63 isoforms are consistent with previous reports (2,5). RAS genes are frequently mutated across cancer types; however it has been reported that HRAS mutations are relatively low frequency overall, and associated primarily with squamous cancers. The HRAS mutation frequency varies depending on studies, with up to 20% reported in cutaneous SCCs, and up to 6% reported in HNSCCs (1,3,4,6,(48)(49)(50). Based on our analysis of the TCGA data, the degree of overexpressed HRAS in HNSCCs is significant (Figures 1A, B). This is consistent with a previous finding that wild-type overexpressed HRAS is at a significantly higher frequency in human HNSCCs than that of the mutated HRAS (49). Interestingly, the prognosis of HNSCCs may be different based on the mutational status of HRAS (51). Further analyses will be needed to investigate the role of RAS mutational status/expression level in the immune TME and how it relates to the clinical outcome of SCCs. Nonetheless, this information emphasizes that the activation of the Ras pathway, whether achieved by the overexpression of wildtype gene or oncogenic gain-of-function mutation, plays an important role in the pathogenesis of SCCs. As suggested by analyses of single cell RNA-seq data from human HNSCCs, the contribution of RAS and DNp63 oncogenes are derived from epithelial cells and not from the stromal or immune cells within the tumors ( Figure 2).
As reported here, in this mouse model, significantly elevated levels of immunosuppressive PMN-MDSCs are recruited along with CD8 + T cells into v-ras Ha /DNp63a-driven carcinomas compared to v-ras Ha -initiated tumors by 14 days. We established that cells from these tumor-bearing hosts expressing these PMN-MDSC markers functionally suppress T cells. These results are consistent with the secretion of chemokines transcriptionally inducible by DNp63 and implicated in the recruitment of PMN-MDSCs to the TME (21,22). Notably, these findings are consistent with bioinformatic analyses from an earlier Pan-Squamous TCGA study showing that human SCC that express DNp63 are concurrently infiltrated with immune cells bearing CD4 + and CD8 + T cell markers along with cells expressing MDSC and immunosuppressive markers (2).
The contribution of MDSCs to tumor progression has been studied extensively (45). Although MDSCs are practically undetectable in healthy individuals, increasing numbers of circulating MDSCs correlate with stage, metastasis, tumor burden, and a worse prognosis in various cancers. MDSCs exert an immunosuppressive function locally within the TME as well as systemically throughout the host (45). Proposed mechanisms of this immune suppression by MDSCs include depletion of local nutrients required for T-cell function, such as L-arginine and tryptophan, by producing enzymes such as arginase-1 (ARG1), nitric oxide synthase (NOS2), and indole amine 2,3 dioxygenase (IDO) (41). In the current study, we demonstrate that Ly6G hi neutrophilic populations isolated from v-ras Ha /DNp63a tumors are PMN-MDSCs ( Figure 6). Our data indicate that PMN-MDSCs in both premalignant and malignant tumors were more immunosuppressive than peripheral PMN-MDSCs, suggesting that the TME significantly polarizes recruited neutrophilic cells toward a more immunosuppressive state, consistent with previously reported studies (reviewed in (41), Figure 6). Levels of monocytic myeloid cells, which may include monocytic MDSCs and macrophages, were consistently low in the tumor site. This may reflect differential chemokines required for trafficking of monocytic cells into tumors.
The accumulation of inflammatory cells including MDSCs in cancer is attributed to the production of cytokines such as GM-CSF, M-CSF, CCL2, CXCL2, and CXCL5 (41,42). In the current study, our data demonstrate that DNp63a further promotes accumulation of tumor PMN-MDSCs within the tumor tissue, which correlated with the increased levels of chemokines Cxcl1 and Cxcl5. This observation is in line with previous studies demonstrating the dependence of tumor growth on host immune cells (52,53). Early work by Pekarek et al. demonstrated the role of granulocytes in rapid growth of tumor cells in vivo (52), and overexpression of CXCL1 (KC, Gro-a/Gro1) in the PAM 212 murine SCC cell line yields larger and more aggressive tumors upon subcutaneous transplantation (53). This is linked to enhanced inflammatory and angiogenic responses, dependent on infiltration of CXCR2 expressing granulocytes from the host (52,53). Similarly, the role of DNp63 in the recruitment of tumor PMN-MDSCs has been demonstrated in a mouse tumor model of triple negative breast cancer (TNBC), a disease which shares common genetic and molecular features of squamous-like cancer subtype, including overexpression of DNp63 (5,54). In the syngeneic mouse model of TNBC, a mammary epithelial cell line expressing oncogenic vras Ha and DNp63-induced the recruitment of PMN-MDSCs to the primary tumor and metastatic sites. Chemokines, CCL22 and CXCL2 were identified as important effectors of MDSC recruitment into these DNp63 expressing TNBC tumors (54).
Our in vitro chemokine data indicated that v-ras Ha alone induces a significant level of chemokines (CXCL2, CXCL5, CXCL7) that recruit inflammatory cells (Supplementary Table S1). This is consistent with previous observations that overexpression of vras Ha in keratinocytes activates EGFR signaling leading to activation of IL-1a, NF-kB, and CXCR2 ligands, important mediators of tumorigenesis (13,14). Interestingly, our data indicate that in vitro, DNp63a overexpression alone resulted in a minimal impact on chemokine production compared to v-ras Ha or Stuffer control primary keratinocytes (Supplementary Figure 3). This pattern was confirmed by three independent methods: cytokine array, Bioplex (Supplementary Figure 2C), and the same custom RT 2 profiler array used for in vivo chemokine detection (Supplementary Figure 4). These in vitro results may partly explain why the overexpression of DNp63a by itself is not sufficient to initiate tumors and that DNp63a requires cooperation with additional oncogene(s) (i.e., RAS) to promote malignant conversion (12,15,21,22,55). We have shown that DNp63 cooperates with NF-kB to promote cytokine gene expression (21), and Ras has been shown to be an inducer of NF-kB (56). It is possible that the expression levels of DNp63 in tumors is significantly higher relative to what occurs in vitro, which may be explained by the paracrine and autocrine signaling in the TME. In an autochthonous murine model of p63induced SCC tumors, DNp63 expression was significantly higher in tumors compared to cultured cells (57). In addition, paracrine signaling between tumor cells and surrounding cells such as fibroblasts and macrophages, which can activate chemokines within the TME, may play a role in DNp63-dependent tumorigenesis. Such paracrine signaling between DNp63-overexpressing cells has been reported in other tumor models (54,57). Likewise, the expression of DNp63 has been shown to be induced by TGF-b via Smad2 and IKKa in the A431 epidermoid carcinoma cell line (58). Preferential expression of DNp63 by hypomethylation of the DNp63 transcriptional start site is also observed in SCCs (2), supporting that other factors (i.e. TGF-b, epigenetic regulation) may contribute to increased DNp63 expression independently of 3q amplification. This underscores the dynamic interplay between the DNp63 and the TME.
Taken together, our data suggest that DNp63a in cooperation with v-ras Ha promotes an immunosuppressive TME through production of immune cell chemokines and recruitment of PMN-MDSCs and Tregs. Our previous studies have demonstrated crosstalk between v-ras Ha , DNp63a, and NF-kB signaling pathways implicated in squamous tumorigenesis (18,21,22) and highlight a potential role of NF-kB/c-Rel signaling together with DNp63a in the recruitment of PMN-MDSCs. Moreover, NF-kB, which has been shown to be essential in two-stage skin carcinogenesis (59), imparts survival of mutant Ras-expressing MEFs from macrophageinduced apoptosis and overcomes immune surveillance via regulation of gene expression that enriches the MDSC population, thereby facilitating a tumorigenic phenotype (60). PMN-MDSCs are increasingly recognized as an important target within the TME for their overarching role in cancer progression and that has been targeted in clinical trials in cancer patients (61). The data presented here enhance our understanding of the link between underlying genomic alterations commonly present within carcinomas and the development of an immunosuppressive TME. This engrafted keratinocyte model adapted to a syngeneic murine background may serve as a valuable tool in future interventional studies aimed at abrogating tumor immunosuppression.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Ethics statement
The animal study was reviewed and approved by Animal Care and Use Committee of the Center for Biologics Evaluation and Research of the Food and Drug Administration.

Author contributions
NS, VG, RP and WW, optimized and performed the in vivo grafting studies. PC and BW performed the tumor processing and characterization of immune subsets, multiplex staining, analysis of chemokines from tumors. NS performed the in vitro studies with primary keratinocytes including analysis of chemokine secretion. CS performed the computational analyses of isoform and gene expression data from TCGA and scRNA-seq data. ZC contributed to the data analysis, graphic presentation, and interpretation of TCGA results across comparison of 33 cancer types. NS and PC wrote the first draft of the manuscript. All authors contributed to the article and approved the submitted version. WW, CW, and CA, oversaw the study.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Author disclaimer
This publication reflects the views of the authors and should not be construed to represent FDA's views or policies.

SUPPLEMENTARY FIGURE 1
Grafting of keratinocytes overexpressing DNp63 alone does not give rise to tumors. Primary control keratinocytes transduced with viruses encoding DNp63 alone or in combination with v-ras Ha were combined with cultured primary dermal cells and grafted onto the dorsal surfaces of wildtype BALB/c mice. The pictures were taken 4 weeks post grafting. Grafts of v-ras Ha /DNp63 expressing cultures are included as positive controls. These results are consistent with results previously described in immune deficient mice (12).

SUPPLEMENTARY FIGURE 2
H&Es of lesions. Representative photomicrographs (10X lens, Zeiss Axio Vert) H&E sections of 2 and 4-week-old lesions derived from grafted primary keratinocytes that had been transduced with virus encoding v-ras Ha alone (vras Ha /Stuffer) or in combination with DNp63 (v-ras Ha /DNp63). Stuffer = Empty control vector. Each image represents an independent tumor from a different animal.

SUPPLEMENTARY FIGURE 3
In Vitro Chemokine and Cytokine Assays. Cultured primary keratinocytes overexpressing Ras release chemokines involved in recruiting infiltrating immune cells. in vitro Supernatants from primary keratinocytes following transduction with retrovirus encoding v-ras Ha (v-ras Ha /Stuffer), empty vector (Stuffer), DNp63a (DNp63), or combination of v-ras Ha and DNp63a (v-Ras Ha / DNp63) were collected for chemokine and cytokine detection. (A, B) Three days following the final transduction as described in Methods, the cell culture medium was replaced with fresh medium and 24 hours later the supernatant was collected and used immediately to probe for chemokines and cytokines (as indicated in the table, upper panel). Culture supernatants were overlaid on array membranes (bottom panels) C3 (A) and C4 (B) and incubated overnight at 4˚C. The chemiluminescence signal was detected using Amersham ImageQuant LAS 4000. Results shown are representative of 3 independent experiments. The dot blot results were qualitatively examined across triplicate results, and the chemokines significantly up-regulated by v-ras Ha (and vras Ha /DNp63) are marked by red boxes, and modestly upregulated chemokines are marked by green boxes. Modest down-regulation by vras Ha was seen in some chemokines, indicated by blue boxes. A list of the modulated chemokines is provided in Supplementary Table S1. (C) ProcartaPlex in vitro quantitative immunoassay was performed according to manufacturer's instructions to measure the relative levels of chemokines present in the supernatants. The supernatants were collected 4-and 14-days post-final transduction. Each sample was tested in duplicate and averaged. Results presented represent the mean and standard deviation from 3 independent experiments. The mean of the results were compared to Stuffer and analyzed using 2way ANOVA using GraphPad Prism 9.4.0 (*, p < 0.05; **, p < 0.01; ***, p < 0.001).

SUPPLEMENTARY FIGURE 4
In vitro chemokine expression analysis by RT 2 profiler PCR analysis. Custom RT 2 qPCR profiler (also used for the detection of chemokine and receptor expression in tumors presented in Figure 4) was used to assess the chemokine gene expression in the RNA extracts of cultured BALB/c newborn primary keratinocytes transduced with retrovirus encoding v-ras Ha (v-ras Ha ) followed by empty vector (Stuffer) or DNp63a (DNp63) as described in Methods. Four days following the first transduction (v-ras Ha ), the RNA was collected and RT 2 qPCR was performed according to the manufacturer's protocol. Values were normalized to GAPDH as a housekeeping gene. The results were analyzed using the RT 2 profiler PCR data analysis online tool available on Qiagen's website (https://geneglobe.qiagen.com/us/analyze). Each sample was tested in duplicate and averaged, and the fold change compared to Stuffer control was plotted using GraphPad Prism 9.2.0. The results presented are from three independent experiments. The mean of the results were compared to Stuffer and analyzed using 2-way ANOVA using GraphPad Prism 9.4.0 (*, p < 0.05; **, p < 0.01; ***, p < 0.001).