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.

Primary keratinocytes and fibroblasts were isolated and cultured from BALB/c newborn pups less than 4 days old, as previously described (25, 26). Keratinocytes were cultured in EMEM (Lonza, Catalogue #: 06-174G) with 8% chelexed fetal bovine serum (FBS) (Gemini, Catalogue #: 100-106) at a final calcium concentration of 0.05 mM (low calcium EMEM). Fibroblasts were cultured for 8-9 days in DMEM (Lonza, Catalogue #: 12-733F) with 10% newborn calf serum (NBCS) (Gibco, Catalogue #: 16010-159), prior to use in grafting studies.

A Ψ² 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 ΔNp63α (LV-ΔNp63α) 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 3x10⁶ Titer Unit (TU) per 60 mm² dish (=1.4x10⁵ TU/cm², a total number of cells estimated to be 1-2 x10⁶ cells), and 4 μg/ml of polybrene with MOI of 1.5-3 (final volume of 0.5 ml/60 mm² 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 mm².

Primary murine newborn epidermal keratinocytes were transduced with v-ras^Ha and ΔNp63α or Stuffer as described above. After 9 days in culture (6 days post-transduction of keratinocytes with ΔNp63α), the keratinocytes and fibroblasts were trypsinized, collected, counted and aliquoted for grafting at 4x10⁶ keratinocytes and 8x10⁶ 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 one-week post-surgery, and the tumors and grafted sites were collected at the time points indicated.

The following primary antibodies and conditions were used for western blotting: 1) p63, BiocareMedical anti-p63 antibody clone 4A4, 1:500 in 5% milk overnight at 4˚C. This antibody is directed against amino acids 1-205 of ΔNp63 and the epitope has been mapped to sequences that are shared with p73 (29, 30); 2) Ras, Sigma-Aldrich anti-Ras antibody clone RAS10, 1:2,000 in 5% milk overnight at 4˚C; 3) Beta-actin, Cell Signaling Technology clone 8H10D10, #3700, 1:10,000 in 5% milk overnight at 4 ˚C. The blots were incubated with secondary antibodies and imaged using ECL reagent available from Kindle Biosciences, LLC (Catalogue # R1004) following the product manual. The blots were imaged using KwikQuant Imager (Kindle Biosciences, LLC, Catalogue # D1001).

Single cell suspensions generated via mechanical dissociation of spleen or digestion of tumor with the Mouse Tumor Dissociation Kit (Miltenyi) as per manufacturer recommendations were incubated with CD16/32 (FcR block) antibodies for 10 minutes. Cells were stained with primary antibodies for 30 minutes. Anti-mouse CD45.2 (clone 104), CD11b (M1/70), Ly-6C (HK1.4), Ly-6G (1A8), CD3 (145-2C11), CD8 (53–6–7), CD4 (GK1.5), CD25 (PC61.5.3), FoxP3 (FJK-16s), and NK1.1 (PK136) antibodies were purchased from Biolegend or eBioscience. 7AAD was used to determine cell viability and a “fluorescence minus one” method was used to determine antibody specificity. For intranuclear staining, cells were fixed and permeabilized using a FoxP3 staining kit (eBioscience) per manufacturer’s protocol. All samples were analyzed on a BD FACSCanto analyzer using FACSDiva software. Post-acquisition analysis was performed using FlowJo vX10.0.7r2.

Formalin-fixed paraffin-embedded sections (5 μm) were stained according to previously described methods (31). Antibodies used (target, clone or catalogue number, and sourcing) were as follows: mouse Ras (clone 18/Ras), BD Bioscience; ΔNp63 (Poly6190), Biolegend; CD4 (4SM95), CD8 (4SM15), and FoxP3 (FJK-16s), eBioscience; and Ly-6G (1A8), BD Pharmingen. Multispectral images were acquired using a Polaris system (Perkin Elmer/Akoya). Scanned images were digitized and individual cell types were quantified using inForm digital pathology analysis software (Akoya) per manufacturer recommendations.

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).

A Custom RT² 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.

Primary keratinocytes were sequentially transduced with viral vectors encoding v-ras^Ha, and either ΔNp63α or Stuffer, as described above. Three days post-transduction of lentivirus-ΔNp63 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.

Primary keratinocytes were sequentially transduced with viral vectors encoding v-ras^Ha, and ΔNp63α or Stuffer, as described above. At three days and 13 days post-transduction of lentivirus-ΔNp63 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 curve-fitting algorithm.

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 v-ras^Ha/ΔNp63 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).

Test of significance between pairs of data are reported as p-values, 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.

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 ΔNp63 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.

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 ΔNp63 with immune deficient athymic nude mice as hosts (12), to evaluate the contribution of ΔNp63 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 ΔNp63 elevated expression of ΔNp63 mimics the gene amplification and overexpression of ΔNp63 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 ΔNp63α overexpression that were identified in this cutaneous model, such as activation of NF-κB/c-Rel, have been confirmed in human HNSCC samples and cell lines (18). The observation that overexpression of ΔNp63α 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 ΔNp63α-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 and a control lentiviral vector (Stuffer) along with primary dermal cells results in papilloma formation, while grafting of primary keratinocytes from BALB/c mice transduced with retroviral vectors encoding v-ras^Ha and ΔNp63α yields carcinomas, consistent with previous findings in athymic nude mice (12). No lesions were observed following grafting of control primary keratinocytes or keratinocytes overexpressing ΔNp63α alone ( Figure 3A ; Supplementary Figure 1 ), consistent with our previous findings (12). Western blot and immunofluorescence staining of the tissue sections confirm the expression of v-ras^Ha and ΔNp63α ( Figures 3B, C ). H&E staining confirms predominant papilloma vs carcinoma histology as early as 2 weeks post grafting in v-ras^Ha- and v-ras^Ha/ΔNp63α - expressing tumors respectively ( Supplementary Figure 2 ).

Chemokines are known to mediate immune cell trafficking in the tumor microenvironment (TME) and are secreted by both tumor and stromal cells. Our previous and on-going observations indicate that overexpression of ΔNp63α in primary murine keratinocytes promotes interactions with the c-Rel subunit of NF-κB and activates genes that are associated with inflammation (18), King and Weinberg, unpublished observations). In transgenic mice, elevated levels of ΔNp63α in the epidermis activate expression of pro-inflammatory chemokines that cooperate with NF-κB transcription factors to promote immunosuppressive type 2 chemokines and cytokines, consistent with the deregulated inflammatory response observed in human HNSCCs (21–23). 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” control), 2) empty vector in combination with v-ras^Ha (v-ras^Ha/Stuffer), 3) ΔNp63α alone, and 4) the combination of v-ras^Ha and ΔNp63α (v-ras^Ha/ΔNp63α) were screened for 96 mouse cytokines and chemokines using multiplexed mouse cytokine antibody array (RayBiotech). A qualitative analysis revealed upregulation of 16 secreted factors by v-ras^Ha/Stuffer- and v-ras^Ha/ΔNp63α-transduced cells, and 2 downregulated proteins ( Supplementary Figures 3A, B ; 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/Stuffer- and v-ras^Ha/ΔNp63α-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 (41–46). Notably, there was no significant change in the cytokine profile between the control and ΔNp63α- transduced 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 ΔNp63α 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 ΔNp63α either separately or together using RT²-custom PCR arrays ( Supplementary Figure 4 ). The data indicate that overexpression of v-ras^Ha upregulates Cxcl1, Cxcl2, and Cxcl5, while ΔNp63α 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 RT²-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/ΔNp63α expressing keratinocytes compared to tumors expressing v-ras^Ha in the absence of p63, or intact skin. In contrast, v-ras^Ha-initiated papillomas upregulated the expression of the Cxcr2 ligand Cxcl2 in comparison to intact skin or Ras/ΔNp63α 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/ΔNp63α 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/ΔNp63α carcinomas compared to v-ras^Ha/Stuffer tumors or normal skin ( Figure 4B ). These data support that ΔNp63α cooperates with v-ras^Ha in vivo to promote the production of chemokines implicated in driving the recruitment of cells with Treg and MDSC markers and immunosuppressive function 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/ΔNp63α-expressing primary keratinocytes were harvested at 2, 3, and 4 weeks post-grafting and immune infiltration profiles were determined by flow cytometry. The tumors were screened for the presence of (CD11b⁺Ly6G^hiLy6C^int) polymorphonuclear (PMN)-like myeloid cells, CD11b⁺Ly6G^loLy6C^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/ΔNp63α 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 v-ras^Ha and v-ras^Ha/ΔNp63α-induced tumors recruit more CD4⁺ T-cells 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/ΔNp63α 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/ΔNp63α (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/ΔNp63α-transduced keratinocytes also have significantly greater numbers of Ly6G^hiLy6C^int PMN-like myeloid cells compared to grafts from v-ras^Ha alone, control keratinocytes or intact skin, suggesting that the v-ras^Ha/ΔNp63α had a significant impact on the concurrent recruitment of these potentially immunosuppressive neutrophilic cells. Lower numbers of Ly6G^loLy6C^hi monocytic myeloid cells (~100s/10,000 cells) compared to Ly6G^hiLy6C^int PMN-like myeloid cells (300-1500s/10,000 cells) were observed across samples in both v-ras^Ha and v-ras^Ha/ΔNp63α tumors similar to control normal cell grafts. The number of NK cells recruited were also similar across primary keratinocytes, v-ras^Ha/Stuffer, and v-ras^Ha/ΔNp63α. 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/ΔNp63α 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/ΔNp63α tumors ( Figure 5C ). Together, these data indicate that v-ras^Ha/ΔNp63α-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 ΔNp63 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 demonstrate that the neutrophilic cells that were recruited into v-ras^Ha/Stuffer and v-ras^Ha/ΔNp63α tumors are PMN-MDSCs.