Loss of TP63 Promotes the Metastasis of Head and Neck Squamous Cell Carcinoma by Activating MAPK and STAT3 Signaling

This article is featured in Highlights of This Issue, p. 1233

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide, resulting in approximately 350,000 deaths every year (1). HNSCCs develop in the mucosal lining of the oral cavity, nasopharynx, hypopharynx, or larynx (2). Tobacco usage, alcohol consumption, and HPV infection represent the major risk factors for HNSCC development (3). There are various treatment options for HNSCC including surgical resection, chemotherapy, targeted therapy, immunotherapy, and radiation (4). These different therapies, alone or in combination, are often successful in eliminating the primary tumor, but treating HNSCC metastases has remained a major challenge (5). For this reason, the 5-year survival rate for HNSCC is only 40%–50% and has not improved over the last decade (6, 7). To achieve a better clinical outcome, studies focused on understanding the genetic and molecular mechanisms that contribute to HNSCC progression and metastasis are necessary.

TP63 (termed TRP63 in mice), an important regulator of stratified epithelial development and homeostasis (8), has been suggested to play a role in the development and progression of HNSCCs (9). However, whether TP63 functions as a tumor promoter or a tumor suppressor has remained controversial. A role for TP63 in tumor promotion has been suggested on the basis of findings that the TP63 gene is amplified in a subset of HNSCCs, leading to an overexpression of the ΔNp63α isoform of TP63 (10–12). Furthermore, forced overexpression of ΔNp63α in SCC cell lines leads to increased survival and proliferation (10, 13). Although TP63 is often overexpressed in primary tumors, TP63 loss has been associated with tumor progression and metastasis in some tumor types (14–17). However, whether this also occurs during HNSCC progression and metastasis has not been systematically investigated. Using archival paraffin-embedded HNSCC samples, we found that TP63 is lost in late-stage, but not in early-stage, HNSCCs. Furthermore, using genetically engineered mice as well as human orthotopic xenografts, we determined that loss of TRP63/TP63 promotes HNSCC progression and metastasis through activation of MAPK signaling. In the presence of TP63, this transcription factor directly induces expression of DUSP6, a suppressor of MAPK signaling. Upon TP63 loss, this suppression is relieved, leading to activation of MAPK signaling and subsequent STAT3 activation and MMP15 expression. The importance of our findings is underscored by our observation that pharmacologic inhibition of MAPK signaling prevents progression and metastasis of xenograft HNSCCs with low TP63 expression. Collectively, our study provides novel mechanistic insights into TP63 function in HNSCCs in vivo. Furthermore, we have identified key signaling pathways regulated by TP63 that can be therapeutically targeted to prevent HNSCC metastasis.

Krt14.CrePR1 (Jackson Laboratory) transgenic mice were crossed with Trp63^fl/fl mice (18) to obtain Krt14.CrePR1/Trp63^fl/+ (control) and Krt14.CrePR1/Trp63^fl/fl (Trp63-ablated) mice. To ablate Trp63, mice received RU486 (1 mg/mL) via oral gavage from P23 to P33. 4-NQO (Sigma-Aldrich) was added to the drinking water of control and Trp63-ablated mice at a final concentration of 100 μg/mL. After 16 weeks, mice were returned to normal drinking water and monitored weekly. Softened food was provided to mice with signs of food-intake difficulty. Mice were euthanized if there was a significant loss of body weight (>15% of original body weight). After euthanasia, tongue, cheek, and cervical lymph nodes were collected for histology, immunostaining, and RNA and protein extraction. Histopathology of all tumors generated was analyzed by W.A. High (University of Colorado School of Medicine, Aurora, CO), a board-certified pathologist. All animal experiments were approved by The University of Colorado Institutional Animal Care and Use Committee (IACUC).

FaDu and CAL27 cells were maintained in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS (Atlanta Biologicals), penicillin, and streptomycin (Thermo Fisher Scientific). UMSCCI, 584A2, and CCL30 cells were maintained in DMEM-F12 (Thermo Fisher Scientific) supplemented with 10% FBS, penicillin, and streptomycin. All cells were maintained at 37°C in 5% CO₂. FaDu and CAL27 cell lines were authenticated using short tandem repeat analysis at Molecular Biology Service center (UCD-AMC). The cell lines were routinely tested for Mycoplasma contamination using Venor TM GEM Mycoplasma Detection Kit (Sigma-Aldrich). To activate MAPK signaling, FaDu and CAL27 cells were maintained as described above and treated with 50 ng/mL of EGF (Sigma-Aldrich) for 3 hours in serum-free DMEM.

PHIV-Luc-Zsgreen was a gift from Bryan Welm (Addgene, plasmid #39196). Doxycycline-inducible TP63 shRNA (1–3) and nonsilencing shRNA (NS shRNA) were obtained from Dharmacon, Inc.

PHIV-Luc-Zsgreen lentiviral vector was cotransfected with packaging vectors [pUVMC (plasmid #8449) and pMD2.G (plasmid #12259) from Addgene] into 293T cells. For lentiviral transduction, FaDu and CAL27 cells were seeded at 3 × 10⁵ cells/well of a 6-well plate for 24 hours prior to infection. One milliliter of viral supernatant was added to 1 mL of fresh complete medium plus polybrene (8 μg/mL; Sigma-Aldrich) and the cells were incubated overnight. The cells were then washed, fed with fresh medium, and cultured as above. FaDu and CAL27 cells were stably transduced with NS shRNA and TP63 shRNA plasmids as described above.

RNA was extracted using RNeasy Mini Kit (Qiagen). cDNA was synthesized using Verso cDNA Synthesis Kit (Thermo Fisher Scientific). TP63 and β-actin probes were obtained from Thermo Fisher Scientific. Data were analyzed using LC480 thermal cycler (Roche) and for relative quantification, the comparative cycle threshold (ΔΔC_t) method was used.

FaDu and CAL27 cells transduced with either NS shRNA or TP63 shRNA were fixed and stained with propidium iodide (Thermo Fisher Scientific) at room temperature for 20 minutes. The fixed and stained cells were stored at 4°C until further analysis. The cell-cycle distribution and data analysis were performed at the University of Colorado-Anschutz medical campus (UCD-AMC) flow cytometry core.

FaDu and CAL27 cells were grown to 80%–90% confluency in 12-well tissue culture plates. Confluent cells were treated with mitomycin C (10 μg/mL; Sigma-Aldrich) for 1 hour at 37°C to inhibit cell proliferation. To evaluate cell migration, a scratch was made using 20 μL pipette tip. Using a phase contrast inverted microscope (Nikon eclipse TE2000), images were obtained immediately after scratch (Time 0) until the closure of scratch area. Processing of images was done with the NIS Elements 3.2 imaging software (Nikon Instruments Inc). The percentage of the scratch closure area was quantified using ImageJ (NIH, Bethesda, MD).

Cell invasion was analyzed using Matrigel-coated 8.0-μm–pore polycarbonate membrane transwell inserts (BD Biosciences). HNSCC cells (2.5 × 10⁵) were suspended in DMEM and added to the insert. DMEM containing 10% FBS and penicillin and streptomycin was added to the BD Falcon TC companion plate and used as a chemoattractant. HNSCC cells were incubated for 36 hours and nonmigrating cells were carefully removed with a cotton swab. Migrating cells were stained with 0.5% crystal violet and counted under 40× magnification using Nikon Eclipse 90I microscope system. Image processing was done with the NIS Elements 3.2 imaging software (Nikon Instruments Inc). For each condition, five random images of invading cells were selected for quantification, and the percentage of invading cells/field was calculated.

Orthotopic xenograft SCCs were generated by implanting 5 × 10⁵ cells into the tongue of 4- to 5-week-old female athymic nude mice. For bioluminescence imaging of tumors, mice were anesthetized and injected intraperitoneally with 100 μL of d-luciferin (30 mg/mL in PBS). Mice were imaged using an IVIS-100 imaging system (Xenogen Corporation) and images were captured once weekly. For trametinib treatment, mice were treated once daily for 5 days by oral gavage with trametinib (1 mg/kg in 4% DMSO/Corn oil) or vehicle (4% DMSO/Corn oil). Mice were monitored daily and euthanized if there was a significant loss in body weight (>15% of original body weight). After euthanasia and necropsy, tumors were visualized using fluorescence microscopy. Tumors were excised for histology, immunostaining, IHC, and Western blot analysis. Histopathology of all tumors was analyzed by W.A. High (University of Colorado School of Medicine, Aurora, CO), a board-certified pathologist. All animal experiments were approved by The University of Colorado IACUC.

RNA was extracted using the miRNeasy Mini Kit (Qiagen). FaDu-TP63 NS shRNA (n = 3) and FaDu-TP63 shRNA3 (n = 3) SCCs were used for this analysis. Upon library generation, libraries were sequenced (paired end read 150 cycles) using an Illumina 4000 HT sequencer at the microarray and genomics core in UCD-AMC (Aurora, CO). Raw sequencing reads were converted to FASTQ files and the quality of the reads was determined using the FASTQC tool. Low-quality reads and Illumina sequencing adapters from the 3′ end of the reads were removed using cutadapt tool. High-quality reads were further mapped onto the human genome hg19 (NCBI 37) using STAR (version 2.5.1a). Gene counts were determined using HTSeq tool. Normalization of reads, differential expression analyses, and principal component analyses (PCA) were carried out using the DESeq2 Bioconductor package within the R statistical programming environment. Genes that were differentially expressed (adjusted P <0.1) in TP63 shRNA3 SCCs compared with NS shRNA SCCs were used for pathway identification and Gene Ontology (GO) term analysis. DAVID bioinformatics resources (19) and Gene set enrichment analysis (GSEA; ref. 20) were used for pathway identification and GO analysis. The raw unprocessed RNA-seq data is deposited in Gene Expression Omnibus (GEO) under accession number GSE120689.

Flash-frozen HNSCCs were broken down into small pieces on dry ice and RIPA lysis buffer (Sigma-Aldrich) was added. Tumors were homogenized and sonicated for 30 seconds at low setting on ice. After lysis, tumor lysates were spun at 13,000 rpm for 10 minutes and the supernatant was transferred to a new tube. Protein concentration was determined using BCA method, and equal amounts of protein were loaded for immunoblot analysis. Immunoblots were probed with antibodies against TP63 (mouse monoclonal 4A4, 1:1,000, Santa Cruz Biotechnology), DUSP6, and MMP15 (mouse monoclonal, 1:1,000, Santa Cruz Biotechnology) at 4°C overnight followed by incubation for 2 hours at room temperature with goat anti-mouse secondary antibody conjugated to horseradish peroxidase (Vector Laboratories). GAPDH (mouse monoclonal, 1:10,000, EMD Millipore) was used as a loading control. Detection was performed using the Supersignal West Pico Plus Chemiluminescent Substrate (Thermo Fisher Scientific), and images were acquired using Odyssey Imaging Systems (LI-COR Biosciences).

At 70%–80% confluency, cells were washed with PBS and crosslinked with 1% formaldehyde for 10 minutes at room temperature. The crosslinking reaction was stopped with 2.5 mol/L glycine and the cells were washed twice with ice-cold 1× PBS. After the second wash, cells were lysed using ChIP-SDS lysis buffer [1% SDS, 10 mmol/L EDTS, 50 mmol/L Tris-HCl, pH 8.1, protease inhibitor cocktail (Roche)]. Lysates were centrifuged and the supernatant was discarded. The cell pellet containing 2 × 10⁶ cells was resuspended in 100 μL of ChIP lysis buffer and sonicated using Bioruptor sonicator (Diagenode) using a refrigerated water bath for 10 cycles at high power (30 seconds ON, and 30 seconds OFF). One milliliter of 1 mg/mL protein lysate was used for immunoprecipitation. ChIP was performed with TP63 (11H1) and IgG antibodies using magnetic beads (Magna ChIP protein A+G beads, Millipore) according to the manufacturer's instructions. Primers [P1-forward (5′-ACTATCTGGGCAGCTTCATTG-3′) and reverse (5′-TAAAGCCGGAGGTTCTCTCT-3′), P2- forward (5′-CGGCGCGTCCCAGCACACCT-3′) and reverse (5′-TAAAACAATGTAGAGTGCAGA-3′), P3- forward (5′- GCGGCTGCTGCTGATGGA-3′) and reverse (5′-CAGCAGCCCGAGCACCGACTC-3′)] spanning the upstream, downstream, and putative TP63 binding motif (21) within the DUSP6 promoter were used for ChIP-qPCR.

Formalin-fixed paraffin-embedded tumors were sectioned at 4 μm by the University of Colorado Denver Morphology and Phenotyping Core. The sections were deparaffinized, subjected to antigen retrieval, and incubated with following primary antibodies: TP63 [mouse monoclonal (4A4), 1:100, Santa Cruz Biotechnology, and rabbit polyclonal, 1:100, Cell Signaling Technology Inc.)], KRT14, KRT1, KRT13 (guinea pig polyclonal, 1:1,000, rabbit polyclonal 1:100, rabbit polyclonal 1:5,000, kind gifts from Dennis Roop, University of Colorado Denver, Denver, CO), P-STA3 (Tyr705 and Ser727; rabbit monoclonal, 1:100, Cell Signaling Technology Inc.), and Ki67 (mouse monoclonal, 1:100, Cell Signaling Technology Inc.). Visualization was performed using Alexa Fluor–conjugated secondary antibodies (1:200, Invitrogen) followed by mounting in DAPI-containing mounting media (Vector Laboratories). Images were acquired using Nikon Eclipse 90I microscope system and processing was done with the NIS Elements 3.10 imaging software (Nikon Instruments Inc). IHC for P-EK1/2 signaling was performed as described previously using a rabbit anti-P-ERK1/2 antibody (rabbit monoclonal 1:100, Cell Signaling Technology Inc.; ref. 16).

To quantify TRP63/TP63-, Ki67-, KRT1-, KRT13, P-ERK1/2, P-STAT3 (Ser727)-expressing cells, five different images from five different HNSCCs (each image contains approximately 200 cells) were counted. Nuclear staining with DAPI was used to determine the total number of cells. Data are expressed as mean ± SD of the mean.

GraphPad (Prism 7) was used to perform statistical analysis. The overall survival for 4-NQO–treated mice and xenograft SCCs was estimated using the Kaplan–Meier analysis, and the statistical significance between groups was determined by performing log-rank test. Mann–Whitney U test was used to compare the number of tumors between groups. To compare among different groups, two-tailed unpaired Student t tests were performed. In all analyses, P < 0.05 was considered statistically significant.

To analyze TP63 expression levels in human HNSCCs, we performed immunofluorescence (IF) staining for TP63 and KRT14 on human archival HNSCC sections (22). Specifically, we analyzed HNSCCs originating from the tongue and larynx. KRT14 is expressed in stratified epithelia, including head and neck mucosa, and is used to identify the epithelial portion of the tumors. We only quantified expression of TP63 expression, and other markers, in cells expressing KRT14. We found that TP63 expression was dramatically downregulated in poorly differentiated tongue and laryngeal HNSCCs, but not in well-differentiated HNSCCs of the same origin (Fig. 1A–G; Supplementary Fig. S1A–S1G), suggesting a general role for TP63 loss in HNSCC tumor progression.

To determine the functional consequences of TP63 loss in HNSCCs, we initially generated inducible, epithelial-specific Trp63 knockout mice (Fig. 2A). Briefly, we bred Krt14.CrePR1 mice (Jackson Laboratory) with Trp63 floxed mice (Trp63^fl/fl) (18) to obtain control (Krt14.CrePR1/Trp63^fl/+) and Trp63-ablated mice (Krt14.CrePR1/Trp63^fl/fl). In the Trp63^fl/fl mice, the DNA-binding domain, common to all Trp63 isoforms, is flanked by LoxP sites. Thus, Cre-mediated recombination in these mice leads to elimination of all TRP63 isoforms. To ablate Trp63 from head and neck epithelia, mouse tongues were treated with RU486. Wild-type mouse tongue only expresses ΔNp63 isoforms (data not shown). IF staining for TRP63 and KRT14 on tongue sections obtained two weeks after completion of RU486 treatment demonstrated the efficient ablation of TRP63 from tongue epithelia in Trp63-ablated mice (Fig. 2B–D).

As spontaneous tumors did not develop in response to Trp63 ablation, we next treated control and Trp63-ablated mice with 4-Nitroquinoline-1-oxide (4-NQO), a chemical carcinogen that is widely used as a tobacco surrogate in mouse models (23). Importantly, 4-NQO treatment leads to oral SCC development by faithfully recapitulating human oral carcinogenesis, and the resulting oral tumors display the molecular and histologic features of human oral SCCs (24, 25). Mice were given 4-NQO in their drinking water for 16 weeks after which they were returned to normal drinking water for the remainder of the study (Fig. 2E). We found that Trp63-ablated mice developed more tumors than control mice, and had a significantly shorter survival time (Fig. 2F–H). We confirmed that TRP63 expression was significantly downregulated in SCCs fromTrp63-ablated mice (Fig. 2I and J). This downregulation was not associated with an increase in proliferation in these tumors (Supplementary Fig. S2A–S2C). Furthermore, 62.5% of Trp63-ablated tumors were classified as poorly differentiated SCCs, versus only 37.5% of control tumors (Supplementary Fig. S2D–S2H). The poor differentiation grade was further confirmed by the reduced expression of keratin 1 (KRT1) and keratin 13 (KRT13), keratins that are expressed in differentiated cells of normal oral epithelium, but not in malignant oral SCCs (refs. 26, 27; Fig. 2I and J). Collectively, these data demonstrate that loss of TRP63 leads to the development of more aggressive tumors. We could not analyze the metastatic potential of these tumors in vivo, as the mice had to be euthanized for humane reasons.

To determine the relevance of our findings for human HNSCC tumorigenesis, we analyzed the consequences of TP63 downregulation in human HNSCC cell lines. FaDu and CAL27 cells were selected as they expressed both TP63, specifically the ΔNp63α isoform, and KRT14 (Fig. 3A; Supplementary Fig. S3A–S3C and data not shown). TP63 was downregulated through lentiviral transduction of a doxycycline-inducible TP63 shRNA [three shRNAs were tested (1–3)]. A NS shRNA was used as a control. All viral constructs also encode RFP to enable the localization of tumors by fluorescence microscopy. After quantification of remaining TP63 expression, we selected FaDu and CAL27 cells transduced with TP63 shRNA2 and TP63 shRNA3 for subsequent studies as they were most efficient at downregulating TP63 expression (Fig. 3B–D). These cells will be referred as FaDu-NS shRNA, FaDu-TP63 shRNA2, FaDu-TP63 shRNA3, CAL27-NS shRNA, CAL27-TP63 shRNA2, and CAL27-TP63 shRNA3. Using cell-cycle analysis and Ki67 staining of the SCCs, we did not observe differences in proliferation in response to TP63 downregulation (Supplementary Fig. S4A–S4H).

Next, we assessed the effects of TP63 loss on migration and invasion of HNSCC cell lines in vitro. For these experiments, cells were treated with mitomycin-C to block proliferation. Using a scratch assay, we observed that TP63 downregulation increased the migration of FaDu and CAL27 cells (Fig. 3E and F; Supplementary Fig. S5A and S5B). Furthermore, TP63 downregulation increased the invasion of FaDu and CAL27 cells in a transwell assay (Fig. 3G; Supplementary Fig. S5C and S5D). Taken together, these findings demonstrate that downregulation of TP63 increases the migratory and invasive properties of two HNSCC cell lines without affecting proliferation.

To enable in vivo imaging, we introduced a luciferase-expressing lentivirus into all cell lines used in this study. We then generated orthotopic xenograft HNSCCs by implanting these cells into the tongue of athymic nude mice. Such orthotopic grafts are a well-established method for studying HNSCCs and these models are considered highly physiologically relevant (28). Mice were fed a doxycycline diet to achieve TP63 knockdown. Strikingly, within 2–3 weeks after implantation, all TP63 knockdown SCCs had metastasized to the cervical lymph nodes (Fig. 4A–E; Supplementary Fig. S6A–S6F; Supplementary Fig. S7A–S7E). Furthermore, all mice carrying TP63 knockdown SCCs developed multiple tumors leading to rapid weight loss. In contrast, even after 8–9 weeks most FaDu-NS shRNA (5/7) and CAL27-NS shRNA (7/7) SCCs remained restricted to the tongue (Fig. 4A–E; Supplementary Fig. S6A–S6F; Supplementary Fig. S7A–S7E).

Consistent with the higher rate of metastasis, primary TP63 knockdown SCCs displayed a poorer differentiation grade (Fig. 4F–I; Supplementary Fig. S7F–S7J). Furthermore, IF staining for TP63 and KRT14 on FaDu-TP63 shRNA3 and CAL27-TP63 shRNA3 SCCs confirmed significant downregulation of TP63 (Fig. 4J–L; Supplementary Fig. S7K–S7M). Collectively, these findings demonstrate that loss of TP63 promotes the progression and metastasis of human xenograft HNSCCs in vivo.

It has been suggested that TP63 downregulation leads to epithelial to mesenchymal transition (EMT; ref. 29). As SCCs that developed in our models retain KRT14 expression, EMT is unlikely to have contributed to tumor progression. This is further supported by our observation that TP63 knockdown SCCs did not exhibit other hallmarks of EMT including loss of E-cadherin or gain of vimentin (Supplementary Fig. S8A–S8F and data not shown; ref. 30).

To investigate the mechanism by which loss of TP63 promotes metastasis, we performed RNA-Sequencing (RNA-Seq) on FaDu-NS shRNA and FaDu-TP63 shRNA3 SCCs. We identified 176 differently expressed genes (P < 0.001 and FDR ≤ 0.1; Supplementary Table S1) in FaDu-TP63 shRNA3 compared with FaDu-NS shRNA SCCs. Pathway analysis using NIH DAVID Bioinformatics Resources and Gene Set Enrichment Analysis (GSEA) indicated an upregulation of genes associated with EGFR/RAS/MAPK and JAK-STAT3 signaling in TP63 knockdown SCCs (Fig. 5A and B; Supplementary Fig. S9A and S9B). Furthermore, gene ontology (GO) term functional analysis revealed changes in biological processes related to proliferation, migration, and adhesion (Supplementary Fig. S9C and S9D). These data suggest that genes previously associated with migration, invasion, and metastasis in other tumor types (31–33), might be under the control of TP63 in HNSCCs.

To validate the RNA-Seq results, we determined the levels of P-ERK1/2, indicative of activate MAPK signaling. We observed upregulation of P-ERK1/2 in TP63 knockdown SCCs (FaDu and CAL27) compared with their respective controls (Fig. 5C and D). Furthermore, we performed IF staining for two active forms of STAT3, P-STAT3 (Tyr705) and P-STAT3 (Ser727; ref. 34). Interestingly, we observed upregulation of P-STAT3 (Ser727), but not P-STAT3 (Tyr705), in TP63 knockdown SCCs (Fig. 5E and F; Supplementary Fig. S10A–S10D). Finally, we assessed expression of P-ERK1/2 and P-STAT3 (Ser727) in primary human HNSCCs. In agreement with our data on the xenograft SCCs, we found that loss of TP63 correlated with an upregulation of P-ERK1/2 and P-STAT3 (Ser727). Specifically, poorly differentiated SCCs exhibited TP63 loss and induction of P-ERK1/2 and P-STAT3 (Ser727), whereas well-differentiated SCCs expressed TP63, but exhibited low levels of P-ERK1/2 and P-STAT3 (Ser727) expression (Fig. 5G–L; Supplementary Fig. S11A–S11F).

Finally, consistent with findings that MAPK activation promotes tumor progression and metastasis, in part, by activating MMPs (matrix metalloproteinases; ref. 31), our RNA-seq studies identified significant upregulation of MMP15 (MT2-MMP) in FaDu-TP63 shRNA3 SCCs (Supplementary Table S1). Immunoblot analysis confirmed the upregulation of MMP15 in FaDu-TP63 shRNA3 and CAL27-TP63 shRNA3 SCCs compared with their respective controls (Fig. 5M). Taken together, these findings indicate that loss of TP63 upregulates MAPK and STAT3 signaling leading to induction of MMP15 in HNSCCs.

Our RNA-seq data, validated by Western blot analysis, indicated that Dual Specificity Phosphatase 6 (DUSP6), a negative regulator of MAPK signaling (35), is downregulated in TP63 knockdown SCCs (Fig. 5A and N). Furthermore, using ChIP, we found that TP63 directly interacts with a consensus TP63 binding motif (21) within the DUSP6 promoter (P2), but not with upstream or downstream sequences (P1 and P3), suggesting that TP63 suppresses MAPK signaling by inducing DUSP6 expression (Fig. 5O–R).

As we observed both activation of MAPK and JAK-STAT3 signaling in our system, we wanted to determine whether cross-talk between these two signaling pathways occurs. To this end, we treated FaDu and CAL27 cells with EGF, the upstream ligand for MAPK signaling, and observed a robust activation of P-ERK1/2 and P-STAT3 (Ser727), but not P-STAT3 (Tyr705), in FaDu and CAL27 cells (Supplementary Fig. S12A–S12D). EGF treatment also led to an upregulation of MMP15 (Supplementary Fig. S12E and S12F). These data provide a direct link between MAPK activation, STAT3 signaling, and MMP15 expression in HNSCC.

To determine the significance of MAPK activation in TP63 knockdown SCCs, we treated TP63 knockdown cells with the MEK inhibitor trametinib (36). The IC₅₀ values of FaDu-TP63 shRNA3 and CAL27-TP63 shRNA3 cells treated with trametinib are shown in Supplementary Fig. S13A and S13B. Trametinib treatment led to inhibition of P-ERK1/2 in FaDu-TP63 shRNA3 and CAL27-TP63 shRNA3 cells (Supplementary Fig. S13C and S13D). Consistent with our finding that MAPK functions upstream of P-STAT3 (Ser727) in HNSCC cells, trametinib treatment also led to inhibition of P-STAT3 (Ser727), but not of P-STAT3 (Tyr705; Fig. 6A and B). Furthermore, trametinib treatment prevented the migration and invasion of FaDu-TP63 shRNA3 and CAL27-TP63 shRNA3 cells (Fig. 6C–H). Thus, the consequences of TP63 loss in HNSCC cells are reversed by inhibition of MAPK signaling.

We next determined whether trametinib treatment could prevent metastasis of HNSCCs that lack TP63 expression. To this end, we engrafted mice with FaDu-TP63 shRNA3 or CAL27-TP63 shRNA3 cells into the tongue. Two weeks after grafting, mice were treated with either vehicle or trametinib once daily. They were also fed a doxycycline diet to downregulate TP63 expression. Whereas all vehicle-treated mice developed cervical lymph node metastases, only 2 of 7 trametinib-treated FaDu-TP63 shRNA3 mice and 3/7 trametinib-treated CAL27-TP63 shRNA3 mice developed cervical lymph node metastases (Fig. 7A–C; Supplementary Fig. S14A–S14D; Supplementary Fig. S15A–S15C). Moreover, the tumor burden was lower in trametinib-treated mice compared with vehicle-treated mice (Fig. 7D; Supplementary Fig. S15D). Most importantly, trametinib treatment significantly improved the survival of mice compared with vehicle treatment (Fig. 7E; Supplementary Fig. S15E). The reduced metastasis was not caused by a possible antiproliferative effect of trametinib as we did not observe a difference in proliferation between trametinib- and vehicle-treated tumors (Fig. 7F–H; Supplementary Fig. S15F–S15H).

Histopathologic analysis confirmed that trametinib treatment prevented the progression to poorly differentiated SCCs. Specifically, whereas all vehicle-treated SCCs were classified as poorly differentiated, 71.4% of FaDu-TP63 shRNA3 SCCs and 43% of CAL27-TP63 shRNA3 SCCs were classified as moderately differentiated (Fig. 7I–K; Supplementary Fig. S15I–S15K). Finally, trametinib treatment reversed the induction of P-ERK1/2, P-STAT3 (Ser727), and MMP15 in both FaDu-TP63 shRNA3 SCCs and CAL27-TP63 shRNA3 SCCs (Fig. 7L–R; Supplementary Fig. S15L–S15Q). Collectively, these findings demonstrate that trametinib treatment prevents the metastasis of HNSCCs with low TP63 expression through downregulating the levels of P-ERK1/2, P-STAT3 (Ser727), and MMP15.

TP63 is frequently overexpressed in primary human SCCs, including HNSCCs, and is often believed to function as an oncogene during early stages of tumorigenesis (10, 11, 13, 17). Furthermore, loss of TP63 during later stages of tumorigenesis has been linked to tumor aggressiveness and metastasis in skin, breast, prostate, and bladder cancers (14–16, 37). However, the role of TP63 in metastasis of HNSCCs has not been investigated. In this study, we demonstrate that TP63 expression is specifically downregulated in poorly differentiated HNSCCs (Fig. 1; Supplementary Fig. S1). Furthermore, using genetically engineered mice as well as xenograft models, we determined that TRP63/TP63 downregulation accelerates HNSCC progression and metastasis. The role for TP63 in suppressing tumor metastasis is mediated by induction of DUSP6, a negative regulator of RAS-MAPK signaling (Fig. 5O–R). Upon TP63 downregulation, activation of MAPK signaling leads to STAT3 phosphorylation and MMP15 induction, providing a mechanistic explanation for the high metastatic potential of TP63 knockdown SCCs (Fig. 7S). Consistent with these findings, we previously analyzed publicly available patient datasets (TCGA) and observed a correlation between low TP63 expression and an overall shorter survival time in patients with HNSCC (16).

The TP63/TRP63 gene can produce at least six isoforms by using two different promoters and alternative splicing (38). Consistent with previous reports on TP63/TRP63 expression in stratified epithelia and squamous tumors (8), we found that both the mouse tongue and the human HNSCC lines used in this study only express ΔNp63α. Furthermore, our strategies were designed to eliminate all TRP63/TP63 isoforms. Thus, even if other TRP63/TP63 isoforms are expressed in these cells or tissues at low levels, they would have been eliminated in our experiments.

The mechanism underlying TP63 loss in HNSCCs is not known. Although induction of EMT has been suggested to occur in response to TP63 loss, we did not observe EMT in our tumors. Furthermore, mutations in TP63 gene are rarely found in HNSCCs (39, 40). However, silencing of TP63 gene expression through promoter methylation has been reported in lung and bladder cancer (41, 42). In addition, in some nonsquamous tumor types, TP63 expression is downregulated by miRNAs (43, 44). However, whether these mechanisms contribute to TP63 loss in HNSCCs remains to be investigated.

Our conclusion that TRP63/TP63 functions as a tumor suppressor gene in oral tumorigenesis appears to contradict previously published work that concluded that ablation of TRP63 in established mouse oral SCCs leads to tumor regression (45). In this study, TRP63 was ablated in SCCs that developed in response to DMBA treatment. Importantly, the oral tumors that developed in these mice were located on the outer lip and facial skin rather than within the oral cavity. Thus, these tumors do not mimic the intra-oral SCCs that developed in our model. Furthermore, the use of DMBA, rather than 4NQO, as a tumor initiator complicates the interpretation of the findings as it is well-established that only 4NQO leads to the development of oral SCCs that mimic human SCCs at the molecular and cellular level (46, 47). Our analysis of 4NQO-induced oral tumors as well as xenograft tumors clearly establishes a critical role for TRP63/TP63 loss in tumor progression and metastasis.

Using RNA-sequencing followed by functional studies, we determined that upregulation of MAPK and STAT3 in TP63 knockdown HNSCCs is responsible for the high metastatic potential of these tumors. Previous studies have demonstrated that a crosstalk between MAPK and STAT3 signaling pathways occurs in cancer cell lines (48). However, a crosstalk between these two signaling pathways has not been investigated in HNSCC progression and metastasis in vivo. Here, we report that the serine phosphorylation of STAT3 in HNSCCs is dependent on MAPK activation (Supplementary Fig. S12). Furthermore, MAPK inhibition by trametinib treatment led to a downregulation of P-STAT3 (Ser727) levels and prevention of tumor metastasis (Figs. 6 and 7). Thus, we have identified a novel functional link between MAPK and P-STAT3 (Ser727) in HNSCC progression.

A potential limitation of this study is the use of HNSCC cell lines harboring TP53 mutations (49). A previous study using breast cancer cells found that some mutant TP53 proteins can directly interact with TP63 affecting the ability of TP63 to transactivate genes that are involved in suppression of metastasis (50). It is possible that these interactions also occur in HNSCCs carrying TP53 mutations. However, TP53-mediated inactivation of TP63 does not seem to occur in the HNSCC cell lines used in our study as we observed a clear difference in phenotype between cells and xenografts with or without TP63 expression. Nevertheless, future studies using HNSCC cell lines without TP53 mutations would be informative in providing further insights into the specific role of TP63 in HNSCC progression and metastasis.

In summary, we have identified a novel role for TP63 loss in HNSCC progression and metastasis. Our study has uncovered the upregulation of the MAPK-P-STAT3 (Ser727)-MMP15 axis as the mechanism by which TP63 promotes metastasis. Furthermore, therapeutic inhibition of this axis prevents HNSCC metastasis caused by loss of TP63. Our data demonstrating that metastasis of HNSCCs that have lost TP63 expression can be prevented by inhibiting MAPK-P-STAT3 signaling has a direct translational impact relevant to the treatment of these HNSCCs.

No potential conflicts of interest were disclosed.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: M.I. Koster, S. Lakshmanachetty

Development of methodology: M.I. Koster, S. Lakshmanachetty

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.I. Koster, S. Lakshmanachetty, V. Balaiya, W.A. High

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.I. Koster, S. Lakshmanachetty, V. Balaiya, W.A. High

Writing, review, and/or revision of the manuscript: M.I. Koster, S. Lakshmanachetty, W.A. High

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.I. Koster, S. Lakshmanachetty, V. Balaiya

Study supervision: M.I. Koster

Other (histologic services): W.A. High

The authors would like to thank Drs. Peter J. Koch and Shi-Long Lu (University of Colorado Anschutz Medical Campus, Aurora, CO) for helpful comments on this manuscript. We also would like to thank Dr. James K. Wahl, III (University of Nebraska Medical Center, Omaha, NE) for generating the monoclonal anti-TP63 antibody (11H1). This work was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) of the NIH under award number R01 AR061506 (to M.I. Koster). This work was also supported by a research grant from the Cancer League of Colorado.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.