Short-term RANKL Exposure Initiates a Neoplastic Transcriptional Program in the Basal Epithelium of the Murine Salivary Gland

Abstract

Although salivary gland cancers comprise only ~3–6 % of head and neck cancers, treatment options for patients with advanced-stage disease are limited. Because of their rarity, salivary gland malignancies are understudied compared to other exocrine tissue cancers. The comparative lack of progress in this cancer field is particularly evident when it comes to our incomplete understanding of the key molecular signals that are causal for the development and/or progression of salivary gland cancers. Using a novel conditional transgenic mouse (K5:RANKL), we demonstrate that Receptor Activator of NFkB Ligand (RANKL) targeted to cytokeratin 5-positive basal epithelial cells of the salivary gland causes aggressive tumorigenesis within a short period of RANKL exposure. Genome-wide transcriptomic analysis reveals that RANKL markedly increases the expression levels of numerous gene families involved in cellular proliferation, migration, and intra- and extra-tumoral communication. Importantly, cross-species comparison of the K5:RANKL transcriptomic dataset with The Cancer Genome Atlas cancer signatures reveals the strongest molecular similarity with cancer subtypes of the human head and neck squamous cell carcinoma. These studies not only provide a much needed transcriptomic resource to mine for novel molecular targets for therapy and/or diagnosis but validates the K5:RANKL transgenic as a preclinical model to further investigate the in vivo oncogenic role of RANKL signaling in salivary gland tumorigenesis.

INTRODUCTION

Representing 3–6% of oropharyngeal cancers^1–4, salivary gland cancers are a rare and heterogeneous tumor type with at least 24 histologic subtypes of the malignant tumor class³. Although rare, salivary gland malignancies pose a significant public health concern as patients at advanced-stage have a poor prognosis in terms of their long-term survival. Unlike cancers of related exocrine tissues, prognostic and therapeutic management of salivary gland malignancies has not substantially improved in decades. For advanced tumors, radical surgical resection with subsequent adjuvant post-operative radiotherapy is usually the only treatment option^{2, 4}. Apart from facial disfigurement along with nerve damage that can occur with some surgeries (i.e. parotidectomy), sequelae from radiotherapy—xerostomia (dry mouth), taste loss, muscositis with recurrent oral infections, trismus, radiation caries, osteoradionecrosis, and dysphagia—can significantly reduce a patient’s quality of life^5–7. Because of their therapeutic ineffectiveness, current chemotherapeutic protocols are frequently repurposed for palliative rather than curative purposes for those patients with advanced-stage disease^{2, 8, 9}. With limited therapeutic options currently available, the consensus in the field is that significant advancements in this area must come from the development of new targeted therapies^10–13. Therefore, new oncogenic drivers and their downstream mediators that initiate and/or promote salivary gland tumorigenesis must be identified to achieve this goal.

A member of the tumor necrosis factor (TNF) superfamily of cytokines, receptor activator of NF-kB ligand (RANKL) signals through its receptor, RANK^14–16. Although RANKL and RANK interact as transmembrane homotrimers, RANKL can also act on RANK through its cleaved ectodomain¹⁷. Engagement of RANKL with its receptor results in recruitment of TNF-receptor associated factors (TRAFs) to the cytoplasmic region of RANK, which triggers induction and/or activation of distinct signaling cascades in a cell-context dependent manner¹⁸. Numerous studies have demonstrated that unscheduled activation of the RANKL/RANK signaling axis drives a plethora of clinicopathologies, including cancers; reviewed in¹⁹. In the case of the mammary gland, a tubuloacinar exocrine tissue like the salivary gland, aberrant RANKL/RANK signaling underpins proliferative, invasive, migratory, and metastatic colonization properties of tumor cells by regulating a broad array of cellular responses, which include (but are not limited to) the execution of the epithelial-mesenchymal transition (EMT) program, enlarging the cancer stem cell population, and recruiting infiltrating tumor associated macrophages to the tumor microenvironment to enable neoplastic expansion, invasion, and metastasis^19–21.

In the case of human head and neck cancers, aberrant RANKL/RANK signaling has been linked to a number of oropharyngeal subtypes, particularly oral carcinomas^22–31. Synthesized and secreted from oral squamous cell carcinomas, RANKL (as a paracrine signal) promotes osteolytic invasion of the mandibular bone^29–31, similar as in bone metastasis by other RANKL dependent cancers^{14, 32–35}. Recently, immunohistochemical studies have shown that the expression of RANKL and RANK is significantly higher in human salivary gland carcinomas as compared to adenomas³⁶. Importantly, significant expression of RANKL and RANK was observed in salivary duct carcinomas and mucoepidermoid carcinomas³⁶, the latter are the most common malignancy type of the salivary gland^37–41.

Using conventional mouse transgenics, we previously demonstrated that RANKL expression targeted to the salivary gland epithelium results in malignancies with a prevalent mucoepidermoid histopathology⁴². While these studies provide functional support for a causal role for aberrant RANKL/RANK signaling in salivary gland tumor initiation and progression, the early molecular signals that mediate this tumorigenic response were not addressed. To address this issue, studies described herein used transcriptomic analysis of early neoplastic salivary gland tissues derived from a new bi-transgenic mouse model that enables short-term RANKL expression in cytokeratin 5 (K5)-positive basal epithelial cells following doxycycline administration in the adult.

MATERIALS AND METHODS

Generation of the K5:RANKL bigenic

The cytokeratin 5-reverse tetracycline transactivator (K5-rtTA) transgenic mouse was purchased from The Jackson Laboratory, Bar Harbor, ME (JAX Mice Stock number: 017519; allele type: Tg(KRT5-rtTA)T2D6Sgkd/J). As described previously^{43, 44}, the K5-rtTA transgenic mouse in the FVB/NJ inbred strain carries the bovine keratin 5 (KRT5) promoter, which drives expression of a nuclear localized rtTA gene. The K5-rtTA transgenic mouse enables the inducible expression of genes in K5 positive basal epithelial cells when administered doxycycline. The generation and characterization of our TetO-RANKL responder transgenic mouse was previously described⁴⁵. Crossing the K5-rtTA effector transgenic with our TetO-RANKL responder transgenic generated the bigenic K5-rtTA: TetO-RANKL mouse (abbreviated K5:RANKL hereon (Figure 1A)). The K5:RANKL bigenic mouse was designed so that doxycycline in the food and water will induce transgene-derived RANKL expression. To induce RANKL transgene expression in the K5:RANKL bigenic, adult (8–9 week old) mice (and monogenic TetO-RANKL control siblings) were switched to rodent chow containing doxycycline at 200mg/kg (Bio-Serv, Flemington, NJ (#53888)) and to water containing 0.2% doxycycline (Takara BIO Inc., Mountain View, CA (#631311)) supplied in light protected bottles^45–47. To reduce taste aversion, doxycycline fortified water was supplemented with 5% sucrose; doxycycline-supplemented water was changed every 3 days to maintain induction potency.

FIGURE 1 │.

Generation and characterization of the K5:RANKL bigenic mouse. (A) Breeding scheme to generate the K5:RANKL bigenic; the K5-rtTA and TetO-RANKL transgenic mice have been described^43–45. (B) Quantitative real time PCR analysis shows that Rankl (Tnfsf11) mRNA levels are significantly increased in salivary gland tissue of adult K5:RANKL mice administered doxycycline for 5-days as compared to similarly treated monogenic control siblings (RNA pooled from five mice per genotype and analysis performed in triplicate). (C) Western immunoblot analysis confirms significant induction of RANKL protein in the salivary glands of K5:RANKL mice that were treated with doxycycline for 5-days; β-actin serves as a loading control. Each lane (lanes 1–3) represents salivary gland protein isolate pooled from three mice per genotype (nine mice total per genotype). (D) Top panels: dual immunofluorescence for K5 and RANKL spatial expression in K5:RANKL salivary gland tissue demonstrates coincident expression of both proteins. Left bottom panel shows immunofluorescence detection of RANKL expression in basal epithelial cells of an interlobular salivary gland duct (white arrowhead). Right bottom panel shows immunohistochemical detection of RANKL expression in basal epithelial cells in ductal and mucinous acini of the K5:RANKL salivary gland (gray and white arrowheads respectively); black arrowhead points to abnormal accumulation of RANKL positive basal epithelial cells. Scale bars denote 10μm.

Mice used in these experiments were housed and maintained in an AAALAC accredited vivarium at Baylor College of Medicine. In temperature-controlled mouse rooms (22 ± 2°C) with a 12-hour lights-on:12-hour lights-off photocycle, mice were fed irradiated Tekland global soy protein-free extruded rodent diet (Harlan Laboratories, Inc., Indianapolis, IN) with free access to fresh water. Experiments on mice were performed according to guidelines detailed in the Guide for the Care and Use of Laboratory Animals (“The Guide” (Eighth Edition 2011)), published by the National Research Council of the National Academies, Washington, D.C. (www.nap.edu). The Institutional Animal Care and Use Committee (IACUC) at Baylor College of Medicine prospectively approved all animal procedures used in this study.

Histological analysis

Fixed overnight in 4% paraformaldehyde, salivary gland normal (monogenic control) and K5:RANKL tumor tissue were processed for embedding in paraffin as reported⁴². Sections (5μm) of salivary gland tissue and tumor were placed on Superfrost Plus glass slides (ThermoFisher Scientific, Waltham, MA) for immunohistochemical staining. Prior to staining, tissue sections were sequentially deparaffinized, rehydrated, and treated with an antigen unmasking solution⁴². After a blocking step, tissue and tumor sections were incubated with the appropriate primary antibody overnight. Primary antibodies used in these studies were: a rabbit polyclonal to human cytokeratin 5 (ab53121; Abcam Inc. (1:200), Cambridge, MA); a guinea pig polyclonal to bovine cytokeratin 8+18 (ab194130 (1:200); Abcam Inc.); a sheep polyclonal to 5-bromo-2-deoxyuridine (BrdU) (ab1893 (1:100); Abcam Inc.); a goat polyclonal anti-mouse RANKL (TRANCE; AF462 (1:200); R&D Systems, Minneapolis, MN); a rabbit polyclonal anti-human parathyroid hormone related peptide (PTHRP/PTHLH (parathyroid hormone like hormone); LS-B2325 (1:100); LifeSpan BioSciences Inc., Seattle, WA); a rabbit polyclonal anti-human secreted phosphoprotein 1 (SPP1 (osteopontin); ab8448; Abcam Inc.); and a goat polyclonal anti-mouse periostin (POSTN; AF2955 (1:100); R&D Systems). Following incubation with primary antibody, sections were incubated with the appropriate horseradish peroxidase conjugated secondary antibody (Vector laboratories Inc., Burlingame, CA) for 1 hour at room temperature followed by incubation with the R.T.U Vectastain Universal ABC reagent (Vector laboratories Inc.) for 30 minutes at room temperature. Immunopositivity was visualized in situ through incubation with 3, 3’-diaminobenzidine (DAB, Vector laboratories Inc.); slides were lightly stained with hematoxylin for contrast. Following a stepwise dehydration process, slides with stained sections were mounted with coverslips using permount solution (Fisher Scientific Inc. (SP15–500)).

For BrdU immunohistochemical and immunofluorescence detection, mice received an intraperitoneal injection of BrdU (10mg/ml; Amersham Biosciences Corporation, Piscataway, NJ) at a dose of 1mg BrdU/20g body weight two hours prior to euthanasia. For immunofluorescence detection, the following Alexa Fluor-conjugated secondary antibody combinations were used: Alexa Fluor 488 goat anti-rabbit IgG (A-11034) and Alexa Fluor 594 goat anti-guinea pig IgG (A-11076) were used to detect K5 and K8+K18 respectively. Alexa Fluor 488 donkey anti-goat IgG (A-11055) and Alexa Fluor 546 donkey anti-sheep IgG (A-11016) were used to detect RANKL and BrdU respectively. Alexa Fluor-conjugated secondary antibodies were obtained from ThermoFisher Scientific Inc. Stained slides were mounted with coverslips using Vectashield mounting medium with DAPI (Vector Laboratories Inc., Burlingame, CA (H-1200)). Digital images of immunostained salivary gland tissue and tumor sections were captured using a color chilled AxioCam MRc5 digital camera attached to a Carl Zeiss AxioImager A1 upright microscope (Zeiss, Jena, Germany). For data display purposes, captured images were digitally collated and annotated with Photoshop and Illustrator (version 6) software (Adobe Systems Inc., San Jose, CA).

Quantitative real-time PCR analysis

Total RNA was extracted from salivary gland tissue and tumors using TRIzol reagent (ThermoFisher Scientific Inc.) before further purification with the RNeasy Plus Mini Kit (Qiagen Inc., Germantown Road, MD). Purified RNA was reversed transcribed into cDNA using the Superscript IV VILO Master Mix (ThermoFisher Scientific Inc.) before quantitative real-time PCR (qRT-PCR) amplification with the Applied Biosystems Step One Plus Real Time PCR System (ThermoFisher Scientific Inc.). Detailed information concerning the TaqMan gene expression assays used in these experiments is described in Table 1; 18S ribosomal RNA served as the internal control.

Table 1 │.

List of murine Taqman expression assays used in these studies.

Gene	ID	Catalog number
Aqp5	11830	Mm00437578_m1
Amy1	11722	Mm00651524_m1
Ccl8	20307	Mm01297183_m1
Ccl9	20308	Mm00441260_m1
Ccr1	12768	Mm00438260_s1
Cldn22	75677	Mm04209227_sH
Clec4n	56620	Mm00490934_m1
Ctsk	13088	Mm00484039_m1
Cxcl11	56066	Mm00444662_m1
Dcpp1	13184	Mm03019597_gH
Elf5	13711	Mm00468732_m1
Esp8	100126778	Mm04243104_m1
Esp18	100126774	Mm04279607_m1
Fscn1	14086	Mm00456046_m1
Krt17	1667	Mm00495207_m1
Mmp12	17381	Mm00500554_m1
Muc19	239611	Mm01306462_m1
Nfkb2	18034	Mm00479807_m1
Postn	50706	Mm01284919_m1
Prom2	192212	Mm00617472_m1
Pthrp/Pthlh	19227	Mm00436057_m1
Relb	19698	Mm00485664_m1
Relt	320100	Mm00723872_m1
Scgb2b26	110187	Mm01254729_m1
Smr3a	20599	Mm01964237_s1
Sox8	20681	Mm00803422_m1
Spp1	20750	Mm00436767_m1
Tfec	22797	Mm01161234_m1
Tnfaip2	21928	Mm00447578_m1
Tnfaip3	21929	Mm00437121_m1
Tnfrsf4	22163	Mm00442039_m1
Tnfrsf8	21936	Mm00437140_m1
Tnfrsf1b	21938	Mm00441889_m1
Tnfrsf11 (Rankl)	21943	Mm00441906_m1
Tnfrsf11a (Rank)	21934	Mm00437132_m1
Traf1	22029	Mm00493827_m1
18S rRNA	Thermo Fisher Scientific Inc.: 4352930E

Transcriptomic profiling

As with our previous RNA-seq studies⁴⁸, RNA purity was evaluated with a NanoDrop spectrophotometer (ThermoFisher Scientific Inc.) and RNA integrity determined using a 2100 Bioanalyzer with RNA chips (Agilent Technologies, Santa Clara, CA). Only RNA preparations that scored a RNA integrity number (RIN) above 8 were used for RNA-seq. Libraries for mRNA sequencing were prepared with the TruSeq Stranded mRNA Library Prep kit (Illumina, San Diego, CA) from 250 ng of RNA. Quality analysis of libraries was performed using Agilent 4200 TapeStation with D1000 ScreenTape assays (Agilent Technologies). Libraries were pooled and then quantified by two methods: 1) on a 2100 Bioanalyzer with DNA chips using High Sensitivity DNA Kit (Agilent Technologies); and 2) with KAPA Library Quantification Kit for Illumina platforms (KAPA Biosystems, Wilmington, MA). Sequencing of mRNA libraries was performed on an Illumina NextSeq 500 platform at mid-output of paired-ended 75 base pair sequencing reads.

Analysis of sequenced mRNA

For initial analysis, pair-ended reads were aligned to the mouse genome (UCSC mm10) using open source STAR software⁴⁹ with NCBI RefSeq genes as the reference; gene expression was measured in read counts for each gene. The R package DESeq2⁵⁰ was used to analyze the gene-based read counts to detect differentially expressed genes between monogenic control and K5:RANKL groups. The false discovery rate (FDR) of differentially expressed genes was estimated using the Benjamini and Hochberg method⁵¹; a FDR <0.05 was considered statistically significant. Because of the large number of differentially expressed genes with a FDR <0.05, a cut-off in absolute fold change |FC| >5 and a sum of average counts of control and mutant sample >100 was used for subsequent analysis. Raw mRNA sequencing data were deposited in NCBI/GEO with the super series accession code: GSE121954. Data were analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) version 6.8 Functional Annotation clustering tool and Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway Maps tool (http://david.ncifcrf.gov). For Gene Set Enrichment Analysis (GSEA; (http://software.broadinstitute.org/gsea/index.jsp)); mouse entrez gene identification numbers were matched to human homologs using their HomoloGene IDs.

Cross species analysis using the cancer genome atlas

For cross-species comparison analysis, mouse sequencing reads were first adaptor trimmed using TrimGalore software⁵². Trimmed reads were mapped to the mouse reference genome using the HISAT spliced alignment program⁵³; transcript assembly and quantification were performed using StringTie software⁵³ against the Gencode gene model⁵⁴. Gene expression (FPKM) was quantile normalized using R statistical software. Differentially expressed genes between K5:RANKL salivary gland tumor and corresponding control samples were determined using the parametric t-test with a p-value <1.25. To compare differentially expressed genes in the K5:RANKL tumor dataset with multiple human cancers profiled by The Cancer Genome Atlas (TCGA), GSEA was applied to both up- and down-regulated genes in the K5:RANKL tumor dataset against rank files from multiple cohorts of TCGA human cancers. Results were ranked using a similarity score rewarding matching direction for normalized enriched scores for up- and down-regulated genes⁵⁵. Normalized enrichment scores (NES) for TCGA human cancer gene signatures and for our K5:RANKL tumor dataset were displayed in a circular combined format using Circos software (http://circos.ca/)⁵⁶.

Western immunoblot analysis

Protein was isolated from salivary gland tissue and tumors and Western immunoblotting was conducted as previously reported⁴⁷. Before transfer to polyvinylidene difluoride membranes (Bio-Rad Laboratories, Hercules, CA), protein (20 μg) was separated by electrophoresis on a 4–15% gradient sodium dodecyl sulfate-polyacrylamide gel. The following primary antibodies (at 1:1000 dilution) were used: goat polyclonal anti-mouse RANKL (AF462; R&D Systems) and a mouse monoclonal anti-human β-actin (Sigma-Aldrich, St. Louis, MO; A1978). After incubation with the appropriate secondary antibody, immunopositive bands were visualized using the SuperSignal West Pico Chemiluminescent Substrate kit (ThermoFisher Scientific Inc.). To enable re-probing of the same membrane with different antibodies, membranes were stripped of primary and secondary antibodies using Restore Western Blot Stripping buffer ((#21059) ThermoFisher Scientific Inc.).

Statistical evaluation

As required, data are displayed as means ± standard deviation. The significance of the difference between groups was determined by two-tailed Student’s t tests and one-way ANOVA with Tukey’s post-hoc multiple range tests using the GraphPad and Instat statistical analysis tools (GraphPad software Inc., La Jolla, CA). Unless indicated otherwise, at least three independent replicates were used, and no samples were excluded from any study. A p-value <0.05 was considered statistically significant; asterisks in histograms denote the level of significance: *p<0.05; **p<0.01; and ***p<0.001.

RESULTS

Short-term Doxycycline Administration Induces RANKL Expression in Basal Epithelial Cells of the K5:RANKL Salivary Gland

For transcriptomic profiling described in these studies, the K5:RANKL bigenic mouse (Figure 1A) was used to enable short-term induction of transgene-derived RANKL expression by doxycycline in a specific cell-type of the salivary gland (the K5 positive abluminal basal epithelial cell) at a predetermined age (9-weeks) and for a defined period of time (2-weeks). The above approach was feasible because: (a) RANK is ubiquitously expressed at low levels in the epithelial compartment of the mouse salivary gland⁴²; and (b) the TetO-RANKL effector transgene has proven highly efficient in targeting RANKL expression specifically in response to doxycycline administration⁴⁵. With just 5-days of doxycycline administration, RANKL was significantly induced at the RNA and protein level in salivary gland tissue of adult K5:RANKL mice (Figure 1B, C); similarly treated monogenic TetO-RANKL control mice did not exhibit this induction (Figure 1B, C). As expected, K5:RANKL mice without doxycycline in the food and water did not exhibit RANKL induction (data not shown). Immunohistochemical analysis showed that doxycycline-induced RANKL expression occurred in the K5 positive basal epithelial cell-type throughout the salivary gland (Figure 1d). Before this immunohistochemical analysis, we confirmed that dual immunofluorescence detection of K5 and K8+18 clearly defines the abluminal basal and luminal epithelial cellular compartments respectively of the normal salivary gland duct (Supplementary Figure 1).

The K5:RANKL Salivary Gland Tumor Exhibits a Predominant Mucoepidermoid Histopathology in Response to Short-Term RANKL Exposure

By two weeks of doxycycline administration, both female and male adult K5:RANKL transgenic mice exhibit a swollen neck region that is texturally hard and irregular by palpation (Figure 2A and Supplementary Figure 2). The salivary gland phenotype was 100% penetrant (n= 30 K5:RANKL females and n= 20 K5:RANKL males (compared with an equal number of similarly treated control mice)) and occurred in the three major salivary glands in the majority of K5:RANKL mice examined. Similar to our recent report⁴², the histopathology of both female and male K5:RANKL salivary gland neoplasms following 2-weeks of doxycycline administration was consistent with a high grade mucoepidermoid carcinoma with sparsely distributed mucinous cells and more frequently occurring epidermoid (squamoid) cells with pale to eosinophilic cytoplasm along with nondescript intermediate cells in varied proportions (Figure 2B,C and Supplementary Figure 3). Dual immunofluorescence detection of RANKL expression and BrdU incorporation revealed that tumor areas in the K5:RANKL salivary gland are highly proliferative and heterogeneous in terms of tumor cells that are double immunopositive for RANKL and BrdU or immunopositive for RANKL alone. However, approximately 35% of tumor cells are double positive for RANKL and BrdU, indicating a possible autocrine and/or paracrine signaling mode for RANKL action in this tumor tissue⁴². Not surprisingly, transgene-derived RANKL expression was detected in K5 positive basal cells in tissues other than the salivary gland (i.e. mammary gland, thymus, and skin (data not shown)). However, because of the short-term administration of doxycycline, the effects of transgene-derived RANKL expression in K5 positive basal cell types in other epithelial tissues was minimal with only marginal increased branching morphogenesis in the female mammary gland and a slight enlargement of the thymus in both sexes (data not shown).

FIGURE 2 │.

Early morphological and neoplastic cellular changes in the K5:RANKL salivary gland with short-term RANKL expression. (A) Following two weeks of doxycycline administration, female and male K5:RANKL mice show clear evidence of gross morphological changes in their salivary glands (white arrowheads). Compared to controls, the K5:RANKL salivary gland neoplasms are significantly enlarged and ill-defined with a firm consistency. (B) Section stained with hematoxylin and eosin (H&E) from salivary gland neoplastic tissue of a K5:RANKL transgenic mouse shows evidence of squamous differentiation (arrows). (C) Focal glandular differentiation with areas of intracellular mucous is evident (arrows). (D) Dual immunofluorescence detection of BrdU (red) and RANKL (green) in a salivary gland tumor tissue section from a doxycycline-treated K5:RANKL mouse. Note the presence of tumor cells positive for RANKL and BrdU (white arrowhead) as well as tumor cells positive for RANKL alone (black arrowhead). (E) Dual immunofluorescence detection of BrdU and RANKL in salivary gland tissue derived from a similarly treated control mouse. Note the absence of immunopositivity for BrdU and RANKL in the tissue section. Scale bar (100μm) in (B) applies to (C); scale bar (100μm) in (D) applies to (E).

Transcriptional Reprogramming by RANKL Drives Salivary Gland Tumorigenesis

To identify downstream contributory signals that may underpin early neoplastic cellular changes in the K5:RANKL salivary gland, mRNA-seq analysis was performed using salivary gland tissue derived from K5:RANKL female mice and corresponding monogenic controls, which were administered doxycycline at 9-weeks of age for two weeks. Triplicate mRNA samples from K5:RANKL and control groups (n= 5 mice per replicate; 15 mice total per genotype) were sequenced at an average of 20 million read-pairs per sample. Using a FDR <0.05 and a |FC| >5 (and sum of control and K5:RANKL counts >100), a total of 471 genes (375 up-regulated and 96 down-regulated) were differentially expressed between the K5:RANKL and control groups (Supplementary File 1); Table 2 and 3 display the top 50 genes up- and down-regulated respectively. This degree of stringency was chosen to enable subsequent gene enrichment analysis. Differentially expressed genes analyzed with DAVID for functional enrichment and for enrichment in biological pathways cataloged in KEGG respectively revealed a significant representation in genes involved in signal peptide/disulfide bond, extracellular matrix (EM)-receptor interactions, immune response/chemotaxis, tumor necrosis factor signaling, NF-kB signaling, and osteoclast differentiation (Figure 3 and Supplementary File 1). Collectively, these genes serve critical functions in neoplastic development and progression, including cellular proliferation, survival, promotion of EMT, migration, invasion into the tumor microenvironment, and metastasis. Validation at the transcript level of a selection of these differentially expressed genes is shown (Figure 4 and Supplementary Figure 4). In keeping with the established mediator role for RANKL^{18, 19}, members of the NF-kB signaling cascade are significantly represented in the K5:RANKL tumor signature as well as pro-inflammatory cytokine/chemokine family members (Figure 3, 4). Not surprisingly, many genes (i.e. amylase 1 (Amy1) and aquaporin 5 (Aqp5)) involved in normal salivary gland function are significantly down-regulated in the K5:RANKL salivary gland tumor compared to normal salivary gland tissue (Supplementary Figure 4). Although salivary gland excrinopathy is predicted by the down-regulation in the expression of these genes, the K5:RANKL mice did not exhibit overt signs of salivary gland dysfunction (i.e. absence or reduced saliva).

Table 2 │.

Top 50 genes upregulated in the K5:RANKL salivary gland tumor.

GENE	GENE ID	GENE NAME	K5:RANKL/CONTROL
Ccl20	20297	chemokine (C-C motif) ligand 20	577.6
Alox12e	11685	arachidonate lipoxygenase, epidermal	468.7
Abca13	268379	ATP-binding cassette, sub-family A (ABC1), member 13	269.7
Dsg1a	13510	desmoglein 1 alpha	215.4
Msln	56047	mesothelin	176.0
Gm8909	667977	predicted gene 8909	152.5
Hal	15109	histidine ammonia lyase	144.4
Tnfsf11	21943	tumor necrosis factor (ligand) superfamily, member 11	127.7
Eno1b	433182	enolase 1B, retrotransposed	126.0
Calcb	116903	calcitonin-related polypeptide, beta	116.7
Irg1	16365	immunoresponsive gene 1	113.2
Fcgbp	215384	Fc fragment of IgG binding protein	90.8
Plb1	665270	phospholipase B1	87.9
H2-Ea-ps	100504404	histocompatibility 2, class II antigen E alpha, pseudogene	86.5
Slc30a2	230810	solute carrier family 30 (zinc transporter), member 2	84.2
Alox15	11687	arachidonate 15-lipoxygenase	83.7
Spp1	20750	secreted phosphoprotein 1	79.6
Serpina3h	546546	serine (or cysteine) peptidase inhibitor, clade A, member 3H	78.3
Gad1	14415	glutamate decarboxylase 1	72.4
Spink5	72432	serine peptidase inhibitor, Kazal type 5	65.2
Dsg1b	225256	desmoglein 1 beta	64.0
Thbs4	21828	thrombospondin 4	60.3
Plekhs1	226245	pleckstrin homology domain containing, family S member 1	60.2
Sox8	20681	SRY (sex determining region Y)-box 8	60.2
Mmp12	17381	matrix metallopeptidase 12	59.6
Cd207	246278	CD207 antigen	59.5
Mmp9	17395	matrix metallopeptidase 9	59.5
H2-Q1	15006	histocompatibility 2, Q region locus 1	57.6
Sectm1a	209588	secreted and transmembrane 1A	57.0
Serpina3n	20716	serine (or cysteine) peptidase inhibitor, clade A, member 3N	54.3
Gdpd3	68616	glycerophosphodiester phosphodiesterase domain containing 3	53.4
Tnfrsf11b	18383	tumor necrosis factor receptor superfamily, member 11b (osteoprotegerin)	52.0
Fcamr	64435	Fc receptor, IgA, IgM, high affinity	52.0
Adamts13	279028	a disintegrin-like and metallopeptidase (reprolysin type) with thrombospond	48.9
Foxn1	15218	forkhead box N1	48.1
Wnt7b	22422	wingless-type MMTV integration site family, member 7B	47.6
Tnfaip2	21928	tumor necrosis factor, alpha-induced protein 2	46.4
Cxcl11	56066	chemokine (C-X-C motif) ligand 11	46.3
Gm1821	218963	ubiquitin pseudogene	46.3
Serpina3i	628900	serine (or cysteine) peptidase inhibitor, clade A, member 3I	44.8
Col8a1	12837	collagen, type VIII, alpha 1	44.1
Cyp2a5	13087	cytochrome P450, family 2, subfamily a, polypeptide 5	40.3
Aadac	67758	arylacetamide deacetylase (esterase)	39.5
Ctsk	13038	cathepsin K	39.1
Ccl9	20308	chemokine (C-C motif) ligand 9	37.0
S100a14	66166	S100 calcium binding protein A14	34.9
Gm8615	667410	glucosamine-6-phosphate deaminase 1 pseudogene	34.7
Ccnb1ip1	239083	cyclin B1 interacting protein 1	33.1
C920025E04Rik	667803	RIKEN cDNA C920025E04 gene	32.9
Spib	272382	Spi-B transcription factor (Spi-1/PU.1 related)	32.2

Table 3 │.

Top 50 genes downregulated in the K5:RANKL salivary gland tumor.

GENE	GENE ID	GENE NAME	K5:RANKL/CONTROL
BC018473	193217	cDNA sequence BC018473	−564.7
Dcpp3	620253	demilune cell and parotid protein 3	−89.6
Pgr	18667	progesterone receptor	−74.1
Hapln4	330790	hyaluronan and proteoglycan link protein 4	−62.2
Cldn22	75677	claudin 22	−49.5
Crisp1	11571	cysteine-rich secretory protein 1	−43.1
Bpifa2	19194	BPI fold containing family A, member 2	−42.2
Nxpe4	244853	neurexophilin and PC-esterase domain family, member 4	−41.6
Smr3a	20599	submaxillary gland androgen regulated protein 3A	−41.0
Prb1	381833	proline-rich protein BstNI subfamily 1	−33.8
Dnase1	13419	deoxyribonuclease I	−31.1
Azgp1	12007	alpha-2-glycoprotein 1, zinc	−28.3
St6galnac1	20445	ST6 (alpha-N-acetyl-neuraminyl-2,3-beta-galactosyl-1,3)-N-acetylgalactosam	−26.3
Prpmp5	381832	proline-rich protein MP5	−24.2
Ggh	14590	gamma-glutamyl hydrolase	−22.9
2610507I01Rik	72203	RIKEN cDNA 2610507I01 gene	−19.4
Amy1	11722	amylase 1, salivary	−19.1
Rab6b	270192	RAB6B, member RAS oncogene family	−18.9
Esp18	100126774	exocrine gland secreted peptide 18	−18.3
1700066N21Rik	73471	RIKEN cDNA 1700066N21 gene	−18.3
Shisa7	232813	shisa family member 7	−16.1
Ret	19713	ret proto-oncogene	−14.7
Lipo1	381236	lipase, member O1	−14.2
Gm4736	114600	predicted gene 4736	−13.7
H2-Eb1	14969	histocompatibility 2, class II antigen E beta	−12.5
Mup6	620807	major urinary protein 6	−11.8
Ttr	22139	transthyretin	−11.5
Gprc6a	210198	G protein-coupled receptor, family C, group 6, member A	−11.0
Fgf1	14164	fibroblast growth factor 1	−10.9
Cst10	58214	cystatin 10 (chondrocytes)	−10.3
Chia1	81600	chitinase, acidic 1	−10.0
Agt	11606	angiotensinogen (serpin peptidase inhibitor, clade A, member 8)	−10.0
B4galnt3	330406	beta-1,4-N-acetyl-galactosaminyl transferase 3	−9.7
Rbm20	73713	RNA binding motif protein 20	−9.6
9130230L23Rik	231253	RIKEN cDNA 9130230L23 gene	−9.6
Tcea3	21401	transcription elongation factor A (SII), 3	−9.2
AI463170	100504549	expressed sequence AI463170	−8.9
Isoc2b	67441	isochorismatase domain containing 2b	−8.7
Unc5a	107448	unc-5 netrin receptor A	−8.5
Atp6v1c2	68775	ATPase, H+ transporting, lysosomal V1 subunit C2	−8.4
Ttc25	74407	tetratricopeptide repeat domain 25	−8.4
2310057J18Rik	67719	RIKEN cDNA 2310057J18 gene	−8.3
Esp8	100126778	exocrine gland secreted peptide 8	−8.2
4833423E24Rik	228151	RIKEN cDNA 4833423E24 gene	−8.2
Insig1	231070	insulin induced gene 1	−8.0
Rap1gap	110351	Rap1 GTPase-activating protein	−7.9
Muc19	239611	mucin 19	−7.9
Aldh3b2	621603	aldehyde dehydrogenase 3 family, member B2	−7.8
Scgb2b26	110187	secretoglobin, family 2B, member 26	−7.8
Gpr37l1	171469	G protein-coupled receptor 37-like 1	−7.7

FIGURE 3 │.

Differentially expressed genes between K5:RANKL salivary gland tumors and control salivary gland tissue analyzed by DAVID. (A) DAVID functional annotation clustering analysis for enrichment of differentially expressed genes in functionally-related gene groups/gene ontology terms. The enrichment score is based on the EASA score of each term. (B) DAVID KEGG pathway analysis of biological/signaling pathways enriched in the differentially expressed gene dataset between K5:RANKL and control salivary gland group.

FIGURE 4 │.

Genes involved in chemokine, NFkB, and TNF signaling are significantly expressed in the K5:RANKL salivary gland tumor. Quantitative real-time PCR demonstrates that members of the chemokine/chemokine receptor (i.e. chemokine (C-C motif) ligand 8 (Ccl8); chemokine (C-C motif) ligand 9 (Ccl9); C-C chemokine receptor type 1 (Ccr1); and C-X-C motif chemokine 11 (Cxcl11)); NFkB (i.e. nuclear factor kappa B subunit 2 (Nfkb2); and RelB proto-oncogene, Nfkb subunit (RelB)); and TNF (i.e. TNF alpha induced protein 3 (Tnfaip3); TNF receptor superfamily member 1B (Tnfrsf1b); TNF receptor superfamily member 4 (Tnfrsf4); TNF receptor superfamily member 11a (Tnfrsf11a (Rank)); and TNF receptor associated factor 1 (Traf1)) signaling pathways are highly expressed in K5:RANKL salivary gland tumors. Data are represented as the average expression fold change relative to control group ± standard deviation (n=4 mice per control and K5:RANKL groups).

The K5:RANKL Salivary Gland Tumor is Most Closely Related to Human Head and Neck Squamous Cell Carcinoma Subtypes

To determine the translational significance of our molecular findings in the K5:RANKL salivary gland tumor, we assessed the degree of molecular similarity between our K5:RANKL tumor transcriptomic dataset with transcriptomic signatures of 25 human cancers profiled in the TCGA. Combined NES for all TCGA tumor development gene signatures and our K5:RANKL tumor dataset were visualized using Circos software (Figure 5A⁵⁶). Of the 25 human cancers analyzed, head and neck squamous cell carcinoma (HNSCC) subtypes were listed within the top four cancers—HNSC (mesenchymal subtype); kidney renal clear-cell carcinoma (KIRC); HNSC (basal); and HNSC (atypical)—that scored the closest molecular similarity to the K5:RANKL tumor transcriptomic signature (Figure 5A,B). Many of the pathways enriched between the HNSC (mesenchymal) and our K5:RANKL tumor dataset are associated with tumor cell proliferation, migration, and metastasis (Figure 5C and Supplemental File 2). All gene expression changes, which were selected from this gene list (Supplementary File 2), validated at the transcriptional level (Figure 6). Using PTHRP, SPP1 and POSTN as examples from this gene list (Supplementary File 2), our immunohistochemical analysis confirmed significant expression of these three genes at the protein level in K5:RANKL salivary gland tumors as compared with control salivary gland tissue (Figure 7).

FIGURE 5 │.

The molecular signature of the K5:RANKL salivary gland tumor most closely resembles human head and neck squamous cell carcinoma subtypes. (A) Normalized enrichment scores for each of the 25 human cancer signatures compared to corresponding K5:RANKL/control were calculated. Cancer types were ranked by similarity to the K5:RANKL salivary gland tumor signature in descending order and in a clockwise direction on a Circos Plot. The outside ring displays enrichment of up-regulated transcripts while the inside ring shows the enrichment of down-regulated transcripts. Circular presentation of the data clearly shows that the K5:RANKL signature is most closely related to mesenchymal and basal subtypes of the head and neck squamous cell carcinoma (HNSCC). Note: Head and neck squamous cell carcinoma–mesenchymal subtype (HNSC-mesenchymal); kidney renal clear cell carcinoma (KIRC); kidney renal papillary cell carcinoma (KIRP); stomach adenocarcinoma (STAD); esophageal adenocarcinoma (ESCA AD); esophageal squamous cell carcinoma (ESCA SCC); cutaneous squamous cell carcinoma (SCC); breast cancer luminal A (lum A) subtype; breast cancer her2 subtype; breast cancer basal subtype; liver hepatocellular carcinoma (LIHC); thyroid carcinoma (THCA); bladder urothelial carcinoma (BLCA); colon adenocarcinoma (COAD); cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC); prostate adenocarcinoma (PRAD); uterine corpus endometrial carcinoma (UCEC); lung adenocarcinoma (LUAD); breast cancer luminal B subtype (BRCA-lumB); rectum adenocarcinoma (READ); lung squamous cell carcinoma (LUSC); and pancreatic adenocarcinoma (PAAD). (B) Enrichment plots of the TCGA tumor datasets that most positively and inversely correlated with the K5:RANKL gene expression signature. (C) Pathway analysis by DAVID KEGG pathway tool of genes shared between the HNSCC GSEA datasets and the mouse K5: RANKL signature.

FIGURE 6 │.

Validation of gene expression changes in the K5:RANKL salivary gland tumor that are also changed in human head and neck squamous cell carcinomas. Quantitative real time PCR results show significant expression changes of genes (i.e. C-C motif chemokine receptor 1 (Ccr1); Claudin 22 (Cldn22); C-type lectin domain family 4, member n (Clec4n); Cathepsin K (Ctsk); Fascin 1 (Fscn1); Keratin 17 (Krt 17); Periostin (Postn); Parathyroid hormone related peptide (Pthrp); Receptor expressed in lymphoid tissue (Relt); Secreted phosphoprotein 1 (Spp1); Transcription factor EC (Tfec); and Tumor necrosis factor receptor superfamily member 8 (Tnfrsf8)) in the K5:RANKL salivary gland tumor that are also markedly changed in expression levels in human head and neck squamous cell carcinomas (Supplementary File 2). Data are represented as standard deviation of the average fold change (n=4 mice per control and K5:RANKL groups).

FIGURE 7 │.

The K5:RANKL salivary gland tumor expresses high levels of PTHRP, POSTN, and SPP1 protein. (A) and (B) Immunohistochemical staining for PTHRP in control salivary gland tissue and K5:RANKL salivary gland tumor tissue respectively. Note the conspicuous expression of PTHRP (white arrowhead) in the K5:RANKL tumor tissue compared to control. (C) and (D) Immunohistochemical detection of POSTN in control salivary gland tissue and K5:RANKL salivary gland tumor tissue respectively. While there are low levels of POSTN expression in control salivary gland tissue, note the marked increase in POSTN expression in the extracellular matrix of the K5:RANKL salivary gland tumor tissue (black arrowhead). (E) and (F) Immunohistochemical staining for SPP1 in control salivary gland tissue and K5:RANKL salivary gland tumor tissue respectively. Note the prominent expression of SPP1 in the epithelial (white arrowhead) and stromal (black arrowhead) compartments of the K5:RANKL salivary gland tumor tissue (inset shows a higher magnification of the stroma area that expresses SPP1 (black arrowhead)). Scale bars indicate 100μm (scale bar in (A) applies to (B-F (excluding the inset)).

DISCUSSION

Advancement in the development of effective molecular diagnostics and targeted therapeutics for the clinical management of salivary gland malignancies has been hampered by the limited availability of suitable experimental models, particularly in vivo models such as the mouse. As clearly demonstrated for cancers of other glandular tissues (i.e. the mammary gland), mouse models for these cancers have been successfully used not only for gaining essential insights into the molecular mechanisms that underpin all stages of tumor development and progression but also to test investigational and repurposed drugs as possible novel anti-neoplastic therapies.

In transgenic mouse studies described here, we demonstrate that targeting RANKL to K5 positive basal cells of the salivary gland epithelium elicits rapid tumorigenesis. The predominant histopathology of a high-grade mucoepidermoid carcinoma exhibited by these tumors is clinically significant as this tumor subtype is the most common salivary gland malignancy^{40, 41, 57} and has recently been shown to express RANKL and RANK³⁶. Moreover, patients with high-grade mucoepidermoid carcinomas have low survival rates and high morbidity following treatment due in part to tumor resistance to conventional platinum-based chemotherapy and ionizing radiation^58–60. Recent studies have also shown that mucoepidermoid carcinomas express high basal levels of NFkB, which may contribute to their chemoresistance and radioresistance^{58, 60}. In the same studies, targeting NFkB signaling in a combined therapy protocol was shown to be effective in markedly reducing the resistance phenotype of human mucoepidermoid carcinoma cell lines in culture^{58, 60}. Because many activators and mediators of the NFkB pathway are markedly up-regulated in the salivary gland tumor of the K5:RANKL mouse (Figures 3, 4 and Supplementary Table 1), this transgenic represents a potentially important preclinical model in which to test the efficacy of these new mechanism-based therapeutic approaches in an in vivo context.

Using the K5:RANKL mouse, together with genome-wide RNA profiling, we disclosed a complex transcriptomic signature that underlies salivary gland tumorigenesis in response to short-term RANKL exposure. Despite this complexity, gene set enrichment analysis and pathway analytics revealed a significant enrichment of transcriptional programs that exert important functions in cellular proliferation, communication, migration, and signaling to the extracellular matrix or tumor microenvironment. These transcriptional programs are reflective of an aggressive tumor phenotype, with potential for locoregional invasion and distant metastasis. Through cross-species analysis, we showed that the transcriptomic signature of the K5:RANKL salivary gland tumor is most closely related to mesenchymal/basal subtypes of human head neck squamous cell carcinomas; mucoepidermoid carcinomas are not profiled by TCGA. Lending strong translational support for our mouse studies, the genes stratified in this analysis represent evolutionary conserved molecular targets that may prove important in the development of new molecular diagnostics and/or therapeutics in the future clinical management of salivary gland malignancies. As a corollary, the K5:RANKL transcriptomic dataset is also closely related to the human kidney renal clear cell carcinoma signature (Figure 5A (KIRC)). This finding is interesting as RANKL/RANK signaling has recently been linked to poor prognosis in patients diagnosed with renal clear cell carcinoma⁶¹ and supports the K5:RANKL transcriptomic signature as a powerful informational resource with which to molecularly understand RANKL-dependent cancers outside the head and neck cancer group.

In addition to RNA expression, we validated at the protein level the upregulation of three of these genes (PTHRP; POSTN; and SPP1) in the K5:RANKL salivary gland tumor. Apart from its role in humoral hypercalcemia and bone resorption⁶², PTHRP is implicated in promoting metastasis of oral squamous cell carcinomas to bone^63–65. Furthermore, studies support a role for PTHRP in conferring malignant potential to human mucoepidermoid carcinomas of the head and neck region⁶⁶, suggesting that PTHRP may be used as a prognostic factor for this malignancy type. Interestingly, previous studies have shown a close correlation between PTHRP and RANKL expression in a number of cancers^67–70, particularly in oral cancers that invade bone^70–73. A secreted extracellular matrix protein, POSTN overexpression is also known as a prognostic indicator of poor survival for many cancers⁷⁴, including head and neck cancers ^(75–79). An extracellular matrix protein, SPP1 is involved in tumor cell proliferation, invasion, and chemoresistance metastasis in a number of cancer types, such as nasopharyngeal and esophageal malignancies ^(80–82). Taken together, our immunohistochemical analysis reinforces the robustness of the K5:RANKL salivary gland tumor model with which to study both established and novel oncogenic drivers of human salivary gland tumorigenesis in an in vivo context.

Apart from providing a new genome-wide transcriptomic resource to gain better insight into the key molecular mechanisms that drive development and progression of this understudied salivary gland malignancy, further mining of this resource may reveal novel molecular vulnerabilities that could be exploited for future targeted therapies. In future studies, the K5:RANKL mouse will be used to determine whether RANKL-dependent enlargement and activation of the cancer stem cell population drives the development of this tumor-type and whether these tumor cells exhibit intrinsic metastatic potential. Noteworthy, enlargement and intrinsic activation of the cancer stem cell population by RANKL has been shown to be a common cellular underpinning for the development and metastatic properties of other glandular tissue cancers such as the mammary gland⁸³. Finally, the K5:RANKL mouse represents an attractive preclinical in vivo model with which to assess the preventative and/or treatment efficacies of RANKL inhibitors in combination with other anti-neoplastic agents in the clinical management of this malignant neoplasm.