CD271 Confers an Invasive and Metastatic Phenotype of Head and Neck Squamous Cell Carcinoma Through the Upregulation of Slug

Abstract

Purpose:

Head and neck squamous cell carcinoma (HNSCC) is comprised of heterogeneous populations of cells, and CD271 (NGFR; p75NTR) has been associated with a tumor-initiating cell subpopulation. This study assessed the role of CD271 in modulating metastatic behavior in HNSCC.

Experimental Design:

CD271 was overexpressed in murine and human oral squamous cell carcinoma cells to assess the impact of CD271 activation on the invasive and metastatic phenotype of these cells, using in vitro and orthotopic in vivo modeling. Treatment with human nerve growth factor (NGF) to activate CD271, as well as shRNA knockdown of the CD271-upregulated Snai2 expression, was used to assess the mechanism of the CD271-induced invasive phenotype. Relevance of CD271 expression in human HNSCC was evaluated in patient derived xenografts (PDXs) and primary human oral cancers, annotated with clinical behavior characteristics and survival data.

Results:

Forced expression of CD271 resulted in a more invasive and metastatic phenotype. Slug, an epithelial-to-mesenchymal transition (EMT)-related transcription factor, encoded by Snai2, was highly expressed in MOC2-CD271 and HSC3-CD271, compared to respective parental cells. CD271 activation by NGF conferred enhanced invasiveness in CD271-overexpressing cells, which was abrogated by Snai2 knockdown. In PDXs and primary human HNSCC, CD271 expression correlated with higher Snai2 expression, greater nodal metastasis, and shorter disease-free survival.

Conclusions:

Activation of CD271 results in upregulation of Snai2/Slug, which, in turn, results in a more invasive phenotype and an enhanced capacity for metastasis to regional lymph nodes. These findings point to CD271 as a promising, therapeutic target for oral cancer metastasis.

INTRODUCTION

Head and neck squamous cell carcinoma (HNSCC) arises in heterogeneous mucosal sites of the upper aerodigestive tract, including the oral cavity, the pharynx, the larynx and the sinonasal cavity. Unfortunately, around two-thirds of newly diagnosed patients with HNSCC present with advanced stage disease, mostly affected by the lymphatic spread of tumor into the regional draining lymph nodes in the neck (1). Despite progress in combined treatment schemes, the clinical outcome of advanced keratinizing HNSCC remains unsatisfactory with overall 5-year survival rates of approximately 50~60%.

Heterogeneity of cells within tumors exist with a subset of cells, called tumor-initiating cells, having increased potential of self-renewal, invasion and resistance to chemoradiotherapy (2–8). Contrary to the physiological EMT process which allows cells to change morphology and lose polarity to become motile, EMT involved in cancer confers tumor cells with the ability to leave the primary site, enter blood or lymphatic vessels, and migrate to distant sites. These processes are thought to be key mediators of tumor-initiating cell invasiveness (9,10).

Previously, we identified CD271 (low-affinity nerve growth factor receptor; p75NTR) as a functional marker of a tumor-initiating cell subpopulation in HNSCC (11). We also have shown that CD44⁺CD271⁺ subpopulation contains the most tumorigenic cells, and that CD271 is a functional and targetable molecule. CD271+ cells, which comprise a subset of the CD44⁺ cells, signals through the MAPK pathway and promotes tumor formation. However, whether the CD44⁺CD271⁺ subpopulation has a more metastatic phenotype has not been determined.

Herein, we report that CD271 overexpression confers a significantly higher rate of regional lymphatic metastasis in a syngeneic murine oral cancer model and an increased invasive phenotype that is due to the induced expression of the transcription factor, Slug, encoded by Snai2. Furthermore, higher expression of CD271 in primary human oral squamous cell carcinoma tumors was associated with higher rate of pathologically discovered nodal metastases. Thus, our data provide insight into the contribution of the tumor-initiating cell marker CD271 to an invasive phenotype and enhanced capacity for metastasis to regional lymph nodes.

MATERIALS AND METHODS

Mice and Cell Lines

C57BL/6 mice (6–8 weeks old) were purchased from Taconic. Rag2^−/−γc^−/− mice (6–8 weeks old) in a Balb/c background were a kind gift from Dr. Irving L. Weissman (Stanford, CA) and were bred in our animal facility under specific pathogen-free conditions. All animal procedures were conducted under institutional guidelines that comply with national laws and policies. The murine oral carcinoma cell line MOC2 was developed and authenticated by exome sequencing (12). Cells were tested for mycoplasma by PCR and passaged for no more than 20 passages since 2012. The human oral carcinoma cell line HSC3 was obtained from the Riken BRC cell bank in Japan (where it was authenticated), immediately prior to use in these experiments, and cells were tested for mycoplasma by PCR. Cells were maintained in complete DMEM:F12 medium (DMEM:F12 1:1 with Glutamax (Gibco, Invitrogen, CA)) containing: 10% heat-inactivated FBS (Cellgro, MA), 100 IU/ml penicillin and 100 μg/ml streptomycin (Gibco, Invitrogen, CA). The HEK293T cell line was obtained from ATCC and maintained in complete DMEM medium.

Tumor growth and metastasis assay

For tumor growth assays, the indicated number of viable cells was injected subcutaneously in Rag2^−/−γc^−/− mice, after cell dissociation, washing, and suspension in PBS. Tumor growth was monitored twice a week and tumor diameters were measured using an electronic calliper to determine the tumor size. For the lymph node metastasis assay, MOC2 or MOC2 overexpressing CD271 (MOC2-CD271) cells were inoculated submucosally in the oral cavity of C57BL/6 mice and followed until tumor sizes reached 1cm in diameter. Then, mice were euthanized and dissected to harvest the neck lymph nodes, which were stained with hematoxylin & eosin to identify metastasis.

Xenograft Assays

Human tumor cell dissociation was performed as described previously (11). Briefly, tumors were minced and then digested with collagenase-hyaluronidase (Stem Cell Technologies) for 16 hrs at 37°C. Following this, the cells were treated for 3 min with trypsin-EDTA, washed, treated with dispase and DNase I (Stem Cell Technologies) for 1 min, washed, filtered, and resuspended in PBS or FACS buffer. Then, CD44⁺CD271⁺ and CD44⁺CD271⁻ tumor cells were sorted to high purity by FACS after gating out mouse MHC I–positive cells, and CD45-positive cells.

Quantitative RT-PCR

The relative abundance of CD271, Snai1, Snai2, NGF, BDNF and NT-5 mRNA was analysed by quantitative real-time reverse transcription-Polymerase Chain Reaction (RTPCR). Briefly, cell cultures were rinse with ice-cold PBS and then the cells were lysed and scraped off using Trizol (Invitrogen, CA). Afterwards, chloroform (Sigma, MO) was added and the mixture was centrifuge. RNA was recovered from the aqueous phase by precipitation with 2-propanol. The extracted RNA was treated with Turbo DNase to remove genomic DNA prior to reverse transcription with Superscript III Reverse Transcriptase (Invitrogen, CA).

Finally, real-time RT-PCR was performed using a specific Taqman Assay for the CD271, Snai1, Snai2, NGF, BDNF and NT-5 sequence. HPRT1 was amplified as control. The amount of each mRNA was expressed as arbitrary units defined as the n-fold difference relative to the control gene HPRT1 (2ΔCt ×100, where ΔCt represents the difference in threshold cycle between the control and target genes).

Lentivirus transduction

To establish the CD271 overexpressing MOC2 cell line (“MOC2-CD271”), the lentiviral construct pHIV-Zsgreen was kindly provided by Dr. Michael Clarke (Stanford, CA). A cDNA from the ATG codon to the stop codon of mouse CD271 was PCR cloned (forward primer: 5’-GCGGCCGCATGAGGAGGGCAGGTGCT-3’; reverse primer: 5’-GGATCCTCACACAGGGGACGTGGCA-3’) into the Topo-TA cloning vector (Invitrogen, CA) and NotI/BamHI inserted into the pHIV-Zsgreen vector. Packaging of VSV-G pseudotyped recombinant lentivirus was performed by transient transfection of 293T cells. One day prior to transfection, 293T cells were seeded in a 6 well plate at 5×105 cells/well in DMEM supplemented with 10% FCS. Cells were co-transfected with 2 μg pHIV-Zsgreen-mCD271 vector, 1 μg gag/pol packaging plasmid (pMDL-g/p-RRE), 0.5 μg rev expression plasmid (RSV-REV) and 0.7 μg VSV-G expression plasmid by Lipofectamin2000 (Invitrogen, CA). Medium was replaced with fully supplemented DMEM after 24 hours and supernatants were collected at 48 hours. Cell debris was removed by centrifugation at 1500 rpm for 5 minutes at 4°C, followed by passage through a 0.2 μm pore polyvinylidene fluoride Durapore filter (Millipore, Bedford, MA). Virus particles were added to the media for infecting target cells. Polybrene (8ug/ml) was added to enhance the lentiviral transduction efficiency. GFP-positive cells were sorted with BD FACSAria™ II cell sorter (BD Biosciences, CA) and propagated for further experiments.

To establish the CD271 overexpressing HSC cell line (“HSC3-CD271”), the pLenti-GIII-CMV-Human-NGFR-GFP-2A-puro plasmid was purchased from ABM Inc, and packaging plasmids (pSAX2,pMD2G) were used to produce lentivirus particles. Subsequent steps for transduction and selection were similar to what was described for MOC2-CD271 establishment. To establish Snai2 knockdown MOC2-CD271 (shSnai2 MOC2-CD271), puromycin resistant vector (plasmid 40657: pLKO.1-sh-mSlug-3, plasmid 40648:pLKO.1-shmSlug-4, Addgene, MA) was used. Lentiviral transduction was performed as described previously, and shSnai2 MOC2-CD271 was selected by double positive for GFP (pHIV-Zsgreen) and puromycin (pLKO.1-sh-mSlug).

Invasion Assay

A total of 2.5×10⁵ cells were seeded on Matrigel-coated porous membrane (pore size, 8 μm; BD Biosciences) in upper chamber. Cells were allowed to invade Matrigel towards FBS gradients as well as fibronectin for 48 hours. The membranes were fixed with methanol and stained with crystal violet. The number of invaded cells was determined by counting six random fields under microscope.

Flow Cytometry Analysis

Cells were harvested from culture flasks or isolated from mouse xenografts after tumor dissociation, as described above. Single-cell suspensions were then pre-treated with IgG from mouse serum to block non-specific staining and incubated with the appropriate antibodies. DAPI was used to allow exclusion of non-viable cells; in the case of mouse xenografts, anti-mouse H-2K^d FITC antibody (SF1–1.1) was added to exclude mouse cells from the analysis.

Western Immunoblot

Whole cell lysates were collected, and proteins were separated by gel electrophoresis. Western analysis was performed with phospho-Erk, total Erk, Slug, α/β tubulin and β actin antibodies described above (11).

Immunohistochemistry

For immunohistochemical staining, the samples were prepared in formalin-fixed, paraffin-embedded blocks of human oral SCC primary tumors and metastatic lymph nodes. After sections were deparaffinized and rehydrated, endogenous peroxidase activity was blocked with 3% H₂O₂, followed by incubation with primary antibody for 15 minutes with p75 NGFR Ab-1 (MS-394, Thermo, 1:500) in Bond-max autoimmunostainer (Leica Biosystem, Melbourne, Australia) or 60 minutes with Slug (sc-15391, Santa Cruz, CA, 1:100) in room temperature. Bond™ Polymer refine detection, DS9800 (Vision Biosystems, Melbourne, Australia) for CD271 or DAKO Envision™ Detection kit for Slug was used for detection.

Images from slides were automatically scanned and reviewed using Aperio Image Scope (Aperio Technologies, Vista, CA). Tumor boundaries in primary samples as well as lymph nodes were delineated and DAB positive area fraction (%) within tumor was measured by Image J.

Statistical Analysis

Prism software (Graph Pad Software, Inc.) was used to analyse the in vitro and in vivo studies to determine statistical significance of the differences by unpaired Student’s t-tests or U-Mann Whitney tests. ANOVA was used for multiple comparisons. The Kaplan-Meier estimates and the log-rank test were done to assess the equality of the survival functions across variables in the DFS analysis. The Cox proportional hazard model was used in the multivariate comparisons and an estimated hazard ratio (HR) with 95% confidence interval (95% CI) was presented. All statistical tests were two-sided and assessed for significance at the level of 0.05 (p value).

RESULTS

CD271 expression in murine oral carcinoma confers a more metastatic and more invasive phenotype

We have previously shown that within the CD44⁺ population, CD271 marks a more tumorigenic subpopulation in human oral squamous cell carcinoma and that inhibition of CD271 renders cells less tumorigenic, indicating that this is a functional receptor in these cancer cells (11). Given that CD271 is known to be correlated with poor clinical outcome in head and neck squamous cell carcinoma, we extended previous work to assess the possible role of CD271 in the process of metastasis, a key negative prognosticator in this disease (13,14).

Here, we utilized a previously described murine oral cancer (MOC) cell line that can be used to model oral HNSCC (12). MOC2 cells have the ability to metastasize to regional lymph nodes when implanted orthotopically in the oral cavities of immunocompetent mice on the C57BL/6 background (11). Interestingly, a minor subset of MOC2 cells expresses CD271 (Fig. 1A), similar to human HNSCC (11).

Figure 1. CD271 expression in murine oral carcinoma confers a more metastatic and more invasive phenotype.

(A) Comparisons of CD271 mRNA expression by qRT-PCR analysis (left) and CD271 cell surface expression by flow cytometry (right) between MOC2 and MOC2-CD271 cells. Expression of mRNA was normalized to Hprt1. Mean±SD is shown, *p<0.05. (B) Comparison of in vivo tumor growth between MOC2 and MOC2-CD271 cells (1×10⁴ cells) implanted into the flanks of Rag−/−γc−/− mice (n=3 per group). Mean±SD is shown. (C) Comparison of early lymph node metastasis rate between MOC2 and MOC2-CD271 tumors following orthotopic intraoral implantation of tumor cells in C57BL/6 mice. Regional lymph nodes were assessed for the presence of tumor metastasis when the longest diameter of the primary tumor reached 1 cm. P-value by Fisher exact test. (D) Representative photomicrograph of H&E-stained section of lymph nodes from mice with an intraoral MOC2-CD271 tumor. Patterns of tumor metastasis shows a sinusoidal invasion only pattern (metastatic tumor resides in the sinusoid of the lymph node) or a mass formation pattern (metastatic tumor progresses to form mass within the lymph node). Scale bar, 100μm (left), 200μm (right). (E) Correlation of cancer cell invasiveness with CD271 expression. Growth factor-reduced Matrigel® Invasion chambers were used to assess and compare the invasive potentials of MOC2 and MOC2-CD271 cells. Cells were seeded on matrigel-coated porous membranes in upper chamber with fibronectin (10 ng/mL) as a chemoattractant in lower chamber for 48 hr. Cells that invaded and attached to the membrane were fixed and stained with crystal violet. To test the effect of CD271 inhibition on cancer cell invasion, MOC2-CD271 cells were pretreated with anti-mouse CD271 antibody (1:100 dilution) or isotype control antibody for 30 min before seeding invasion chambers. Mean±SD is shown, *p<0.05, **p<0.01.

To examine the phenotypic changes induced by CD271, we overexpressed CD271 in MOC2 cells (herein, called MOC2-CD271) by lentiviral transduction (Fig. 1A). When parental MOC2 and MOC2-CD271 were inoculated subcutaneously into the flanks of immunocompromised mice, we did not observe any difference in in vivo tumor growth between the two cell lines (Fig. 1B). However, when oral squamous cell carcinoma tumors were induced by orthotopic injection of the MOC and MOC-CD271 cells into the submucosal space in the oral cavities of syngeneic C57BL/6 mice, MOC2-CD271 cells exhibited a profoundly higher rate of regional metastasis even in the early phase of primary tumor growth (MOC2-CD271, 91.6%, 11/12 vs. parental MOC2, 30.0%, 3/10; p=0.006) (Fig. 1C). Consistent with the greater metastatic potential of the MOC2-CD271 cells, there was also a greater degree of invasion observed by MOC2-CD271 cells when assessed by in vitro invasion assays. Furthermore, this more invasive phenotype of the MOC2-CD271 cells was decreased when CD271 was blocked by a monoclonal antibody to CD271 (Fig. 1E). Thus, CD271 expression confers a more invasive and more metastatic phenotype in the murine oral squamous cell carcinoma cell line.

Activation of CD271 induces expression of Snai2/Slug in MOC2 cells and HSC3 cells

Having observed that forced overexpression of CD271 rendered MOC cells more invasive and metastatic, we next investigated the molecular basis for this phenotypic change. It is well known that epithelial-to-mesenchymal transition (EMT) is a key step in cancer cell migration as well as metastasis, and recently, its link with tumor-initiating cells has been established in various cancer types (5,9,10,15). Because CD271 expression has been correlated with the expression of the transcription factor Slug, encoded by Snai2 (16), we compared the expression of Snai2 between CD271⁻ and CD271⁺ subpopulations sorted to high purity from bulk parental MOC2 cells by FACS. We observed a significant difference in the level of Snai2 mRNA between the two subpopulations but very little difference in the expression of Snai1 and other EMT related genes (Fig. 2A and Supplementary Fig. 1). When we compared the expression level of Snai2 in MOC2 and MOC2-CD271 cells, we also found Snai2 expression to be significantly higher in the MOC2-CD271 cells compared to the parental MOC2 (Fig. 2C). Similarly, when we overexpressed CD271 in a human HNSCC cell line HSC3 (herein, called HSC3-CD271) by lentiviral transduction (Fig. 2B), we observed a significant increase in Snai2 mRNA expression (Fig. 2C).

Figure 2. Activation of CD271 induces expression of Snai2/Slug in murine (MOC2 cells) and human oral carcinoma cells (HSC3 cells).

(A) Comparison of mRNA expression of Snai1 and Snai2 genes in CD271⁻ and CD271⁺ MOC2 cells by qRT-PCR analysis. MOC2 cells were sorted to high purity by FACS prior to analysis. Data was normalized to Hprt1 expression. Mean±SD is shown, * p<0.05. (B) Establishment of CD271-overexpressing HSC3 cells (HSC3-CD271) by lentiviral transduction. Western immunoblot showing increased expression of CD271 in HSC3-CD271 cells as compared to HSC3 cells transduced by lentivirus with empty vector. Cell lysates were subjected to gel electrophoresis, and Western immunoblot analysis was performed with antibody specific for CD271 and actin. (C) Comparison of mRNA expression of Snai1 and Snai2 genes between MOC2 vs. MOC2-CD271 cells, and between HSC3 vs. HSC3-CD271 cells by qRT-PCR. Data was normalized to Hprt1 expression. Mean±SD is shown, **p<0.01. (D) Induction of Snai2 mRNA in MOC2-CD271 cells and HSC3-CD271 cells in response to recombinant human nerve growth factor-β (rhNGF-β). Snai2 mRNA was quantified by qRT-PCR analysis. Cells were cultured in low serum media (DMEM/F12 with 2% fetal bovine serum) for 24 hrs prior to treatment with rhNGF-β (100ng/mL) for 4 hr. Data was normalized to Hprt1 expression. Mean±SD is shown, *p<0.05. (E) Comparison of Snai2 protein (Slug) in MOC2 vs. MOC2-CD271 cells, and in HSC3 vs. HSC3-CD271 cells by Western immunoblot analysis. Cell lysates were quantitated and subjected to gel electrophoresis, and Western immunoblot analysis was performed with antibody specific for Snai2 protein (Slug) and tubulin (MOC2 cells) or actin (HSC3 cells). Densitometry was used to quantitate Slug using tubulin or actin as loading control. (F) Induction of Snai2 protein (Slug) in MOC2-CD271 cells and HSC3-CD271 cells in response to rhNGF-β treatment. Cells were cultured in low serum media (DMEM/F12 with 2% fetal bovine serum) for 24 hr, and then, treated with rhNGF-β (100ng/mL). Cells were collected over a time course, and lysates were measured for Slug by Western immunoblot analysis. Densitometry was performed for quantification. (G) To test the effect of CD271 inhibition on Slug expression in HSC3-CD271 cells, cells were cultured in low serum media (DMEM/F12 with 2% fetal bovine serum) for 24 hr, and pretreated with anti-human CD271 monoclonal antibody (1:100 dilution) or isotype control antibody for 30 min. Then, cells were treated with rhNGF-β (100ng/mL) and collected over a time course, and lysates were measured for Slug by Western immunoblot analysis. Densitometry was performed for quantification.

Concordantly, we found an increased level of Slug protein in MOC2-CD271 cells compared to parental MOC2 cells and in HSC3-CD271 cells compared to parental HSC3 cells, measured by Western immunoblot (Fig. 2E). Activation of CD271 by recombinant human nerve growth factor-β (rhNGF-β), resulted in a significantly increased expression of Snai2 mRNA and Slug protein, indicating that CD271 may promote invasion and metastasis through the induction of Slug expression (Fig. 2D, Fig. 2F). Furthermore, the effects of rhNGF-β on Slug expression in HSC3-CD271 cells was abrogated in the presence of antibodies to CD271 (Fig. 2G).

To address the in vivo source and expression profiles of specific ligands for CD271, orthotopic tumors of MOC2-CD271 were induced and dissociated for mRNA profiling. Higher expression of NGF mRNA was found in the cancer cells as well as stromal cells surrounding tumors, compare to CD45⁺ leukocytes (Supplementary Fig.2). From these findings, it is plausible that autocrine and paracrine signaling may be involved in the in vivo activation of CD271, resulting in an increased invasive behavior.

Activation of CD271 by nerve growth factor confers a more invasive phenotype in MOC2 cells and HSC3 cells by the induction of Snai2/Slug expression.

We next addressed the effect of CD271 activation by rhNGF-β on the invasive phenotype of MOC cells and HSC3 cells. When incubated with rhNGF-β in low serum media, both parental MOC2 and MOC2-CD271 showed enhanced invasion capacity compared to each control; however, MOC2-CD271 showed a significantly more robust response to rhNGF-β compared to MOC2 (Fig. 3A). A similar effect on invasion capacity was seen when HSC3-CD271 cells were cultured with rhNGF-β (Fig. 3B). Interestingly, the morphology of the cells treated with rhNGF-β took on a more spindle shape, consistent with EMT changes and a poorly differentiated phenotype (Supplementary Fig. 3). To investigate whether this enhanced invasive phenotype of MOC2-CD271 was dependent on Snai2 expression, we knocked down the expression of Snai2 in MOC2-CD271 by shRNA (MOC2-CD271 shSnai2) (Fig. 3C). A comparison of the invasive phenotype induced by rhNGF-β treatment in the MOC2-CD271 cells and the MOC2-CD271 shSnai2 cells demonstrated that the inhibition of Snai2 expression significantly reduced the ability of the cells to invade. Thus, the enhanced invasive phenotype conferred by CD271 activation was dependent on Snai2 expression (Fig. 3D).

Figure 3. Activation of CD271 by nerve growth factor confers a more invasive phenotype in MOC2 cells and HSC3 cells by the induction of Snai2/Slug expression.

(A-B) Comparison of cancer cell invasiveness in response to recombinant human nerve growth factor-β (rhNGF-β) treatment. Growth factor reduced Matrigel® Invasion Chambers were used to assess and compare the invasive potential of MOC2 vs. MOC2-CD271 cells (A), and that of HSC3 vs. HSC3-CD271 cells (B). Cells were cultured in low serum media (DMEM/F12 with 2% fetal bovine serum) for 24 hrs, and then, treated with rhNGF-β (100ng/mL). Cells were collected and incubated in invasion chambers with fibronectin (10ng/mL) as a chemoattractant for 48 hrs. Quantitation of invasion assay was performed by counting invaded cells in 6 random fields per chamber under light microscopy. Mean±SD is shown, **p<0.01. Representative photomicrographs of invaded cells (from wells cells treated with rhNGF-β after 48 hrs) stained with crystal violet are shown. (C) Establishment of Snai2 knockdown MOC2-CD271 cells (MOC2-CD271 shSNAI2) by lentiviral transduction. Comparisons of Snai2 mRNA by qRT-PCR and Snai2 protein (Slug) by Western immunoblot analysis between empty vector-transduced MOC2-CD271 cells and MOC2-CD271 shSNAI2 cells. Mean±SD is shown, *p<0.05. (D) Comparison of invasiveness after 48 hrs treatment with rhNGF-β of MOC2-CD271 cells and MOC2-CD271 shSNAI2 cells, measured by invasion chamber assays. Cells were cultured in low serum media (DMEM/F12 with 2% fetal bovine serum) for 24 hrs, and then, treated with rhNGF-β (100ng/mL) for 48 hr. Cells were collected and incubated in invasion chambers for 48hr. Quantitation of invasion assays was performed by counting invaded cells in 6 random fields per chamber under light microscopy. Mean±SD is shown, **p<0.01.

As it has been shown that CD44 mediated oral cancer aggressiveness is regulated by ERK1/2 in the MOC model (12), we checked the level of phosphorylated ERK1/2 in MOC2-CD271 to assess intracellular signaling related to CD271 activation. Consistent with previous studies of NGF inducing ERK phosphorylation through CD271 (11,17), MOC2-CD271 showed profoundly higher levels of phospho-ERK1/2 in the presence of rhNGF-β (Supplementary Fig. S4A). Next, to investigate whether increased expression of Snai2 by CD271 activation is dependent on the ERK1/2 pathway, a selective inhibitor of MEK1/2, U0126, was incubated with the MOC2-CD271 cells in the presence of rhNGF-β. Consistent with the increased phosphorylation of ERK1/2 when CD271 is activated, U0126 inhibition of MEK1/2 abrogated the ability of rhNGF-β to increase Snai2 mRNA expression (Supplementary Fig. S4B). Thus, the CD271-mediated induction of Snai2/Slug expression is mediated through the activation of ERK1/2.

CD271 expression is correlated with higher Snai2 gene expression and greater regional nodal metastasis rates in human oral squamous cell carcinoma

Given that an enhanced metastatic phenotype is conferred by increased CD271 expression in the murine MOC2 and human HSC3 models, we next tested the clinical relevance of these findings. Using enzymatically dissociated patient-derived xenografts derived from primary oral cancer specimens, the CD44⁺CD271⁺ and CD44⁺CD271⁻ subpopulations were sorted to high purity by FACS, and Snai2 mRNA expression was measured in these subsets. Consistent with what was observed in the MOC2 cells (Fig. 2A), there was a higher level of Snai2 expression in the CD44⁺CD271⁺ cells compared to CD44⁺CD271⁻ cells (Fig. 4).

Figure 4. CD271 expression in a subset of HNSCC cells within a tumor is correlated with higher Snai2 gene expression.

Snai2 mRNA profiles of patient derived xenografts by qRT-PCR analysis. Primary tumors were dissociated into single cells, and the CD44⁺CD271⁺ and CD44⁺CD271⁻ tumor cell subsets were sorted to high purity by FACS after gating out mouse MHC-I⁺ cells, and CD45⁺ cells. Data was normalized relatively to HPRT. Mean±SD is shown, * p<0.05.

Next, clinicopathologic data of patients with similar pathologic T stage (pT3–4) and different pathologic N stage (24 pN0, 45 pN+) were retrieved from the cancer registry at Samsung Medical Center (Supplementary Table 1). Tumor samples (primary tumors from 24 pN0 patients; and primary tumors and metastatic lymph nodes from 45 pN+ patients) from formalin-fixed paraffin-embedded blocks were assessed for CD271 and Slug expression by immunohistochemistry. The pattern of CD271 expression on the cancer cell membrane/surface was diverse (Fig. 5A). When measured quantitatively (Supplementary Fig. 5), oral tumors with pN+ patients were found to have significantly higher levels of CD271 expression compared to oral tumors without pathologically proven nodal metastases (Fig. 5B). To test whether increased CD271 expression is predictive for a worse prognosis in patients with oral SCC, disease-free survival data was correlated with clinicopathologic parameters, including CD271 expression levels in primary tumors. The median disease-free survival of patients with high CD271 expression (dichotomized by the median value of percent CD271 positivity within primary tumors) was 9.0 months compared with 64.0 months for patients with low CD271 expression, which was a statistically significant difference (Fig. 5C, Log rank, p = 0.01). More interestingly, even within the pN+ group, patients with high CD271 expression showed a significantly worse disease-free survival compared to patients with low CD271 expression by Kaplan-Meier analysis (Fig. 5D, median 17.6 months vs. 39.0 months, Log Rank, p = 0.03). In an adjusted proportional hazard model of pN+ patients, CD271 expression was strongly predictive of increased risk of recurrence [HR, 2.18; 95% confidence interval, 1.11–4.27; p = 0.02] (Supplementary Table 2).

Figure 5. High CD271 expression is correlated with greater regional nodal metastasis rates and poor prognosis in human oral squamous cell carcinoma.

Comparisons of CD271 expression was performed in human oral SCC patient tumors by immunohistochemical analysis. (A) Representative photomicrographs of oral squamous cell carcinoma tumors showed negative (left) or positive (right) CD271 expression by immunohistochemistry. Immunohistochemistry was performed using anti-CD271 monoclonal antibody on sections of formalin-fixed paraffin-embedded blocks of human oral squamous cell carcinoma primary tumors of pT3–4 stage. Scale bar, 100μm. (B) CD271 expression was quantified by Image J, and comparisons were performed between pN0 and pN+ cohorts. Pathologic staging of the cervical nodes was determined by analysis of neck dissection specimens obtained from all the patients in both cohorts. * p<0.05. (C-D) Disease-free survival was compared between oral cancer patients stratified by the level of CD271 expression (C, entire group of patients, n=69; D, pN+ patients, n=45). High and low expression was dichotomized by the median value of percent CD271 positivity within primary tumors. (E) Representative photomicrographs of metastatic tumor cells within a lymph node shows positive expression of CD271 and Slug by immunohistochemistry. Sections for immunohistochemistry using anti-CD271 and anti-SLUG monoclonal antibody were prepared from formalin-fixed paraffin-embedded blocks. Scale bar, 100μm. (F) A significant correlation is observed between CD271 and Slug expression within metastatic tumors (Pearson’s coefficient, R²=0.35, p=0.02).

In metastatic lymph nodes, there was widespread membraneous CD271 expression on tumor cells as well as cytoplasmic Slug expression in a large majority of the tumors (Fig. 5E). Interestingly, CD271 expression levels within the metastatic tumors showed a correlation with the level of Slug expression (Fig. 5F. Pearson’s coefficient, R²=0.35, p=0.02), consistent with the role of CD271 activation on Snai2/Slug expression and the metastatic process. From these data, it appears that CD271 expression is greater in oral squamous cell carcinomas with associated regional nodal metastases and that degree of CD271 expression in the primary tumor correlates with poor disease-free survival.

DISCUSSION

The current study indicates that activation of the tumor-initiating cell marker CD271 in oral squamous cell carcinoma results in upregulation of Snai2/Slug, which, in turn, results in a more invasive phenotype and the enhanced capacity for metastasis to regional lymph nodes. These findings define one mechanism by which CD271 mediates a more aggressive tumor behavior and point to CD271 as a promising target for therapeutic in this disease.

Although lymphatic metastasis in patients with oral squamous cell carcinoma is known to be major prognosticator for worse prognosis, preclinical research has been hindered by the lack of proper in vivo models. Xenografts implanted into immunodeficient mice cannot fully recapitulate host-tumor interactions, and other transplantable syngeneic models have been shown to have innate limitations for the study of lymphatic metastasis. (18,19) The MOC murine model utilized in the study presented here is capable of displaying several aspects of human oral cancer, including lymphatic metastases in immunocompetent C57BL/6 mice.(12) Using orthotopic implantation of the tumor cells, we observed that CD271 overexpression in MOC2 cells had a significantly enhanced lymphatic metastatic capacity compared to parental MOC2 cells, even at early stages of tumor progression (91.6% vs. 30.0%). In addition, we identified an association between CD271 activation and an increased expression of Snai2/Slug. This finding is consistent with previous studies that have shown that Slug controls metastatic the potential of cancer stem cell populations in lung cancer and breast cancer (20,21). Also, it is notable that when Snai2 was inhibited by shRNA transduction, the enhanced invasiveness induced by activation of CD271 (by rhNGF-β) was significantly diminished. It has been shown that the induction of EMT promotes not only tumor cell invasion and metastasis, but also, the cells with the phenotype and properties of tumor-initiating cells (22–24). Some studies demonstrated that tumor cells capable of undergoing EMT may resemble cancer stem cells, and more importantly, these cells are more drug resistant and have a greater propensity for metastasis (25–27).

Slug is a member of the Snail family of transcription factors and plays a role in normal EMT during processes like development and wound healing. In some solid malignancies, such as HNSCC, Slug has been observed to be overexpressed, and the degree of expression in the primary tumor has also been observed to correlate with recurrence of disease, risk of metastasis, and survival (28–31). In modeling experiments of the effects of Slug on human HNSCC, forced expression on Slug resulted in a more mesenchymal phenotype, a switch in cadherins, and an increased cell motility, consistent with our findings (32). While the role of Slug in this disease is becoming more evident, an understanding of the upstream events leading to Slug expression in subsets of HNSCC tumors has not been delineated. In our study, we provide evidence for the role for the activation of CD271, a marker of tumor-initiating cells in HNSCC (11), in the induction of Slug expression.

Studies investigating CD271 as a marker of tumorigenicity have been somewhat conflicting, and in some studies of melanoma, the expression of CD271 has not been found to be stable. In experiments using patient-derived melanoma xenografts, CD271⁻ and CD271⁺ melanoma cells had similar tumorigenic capacities (33,34). Furthermore, variable CD271 expression patterns were observed in sibling PDX models, findings which were supported by large difference in copy number variation even in phenotypically identical CD271⁻or CD271⁺ melanoma cells. On the contrary, studies examining antibody targeting of CD271⁺ melanoma cells found this to be a powerful therapeutic approach against metastatic melanoma (35). Studies regarding the therapeutic role of targeting CD271 in HNSCC are sparse, and further studies are necessary to determine the clinical utility in this disease.

From surgically resected samples of human oral squamous cell carcinoma, we observed that CD271 expression among a clinicopathologically homogenous group of advanced stage primary tumors was predictive of disease-free survival. This finding is consistent with the observations by Soland et al., in which CD271 positivity at the invasive front of tumor as well as the pattern of invasion were significant prognostic factors of disease-free survival in early stage oral cancer (13). In our study, entire tumor areas were comprehensively analyzed to measure CD271 positive regions in this TNM stage-matched cohort to correlate CD271 expression with nodal metastasis. Interestingly, metastatic tumor cells within the involved lymph nodes showed distinctive CD271 expression, and there was positive correlation with statistical significance between Slug and CD271. These data support the use of CD271 as a marker of prognosis and potential predictor of cervical lymph node metastases in oral squamous cell carcinoma. In addition, these finding point to the potential utility of targeting CD271 in this disease.

TRANSLATIONAL RELEVANCE.

Clinical behavior of head and neck squamous cell carcinoma (HNSCC) is variable, and metastasis to regional lymph nodes in this disease portends a significantly worse outcome. Understanding drivers of this process would provide strategies for prediction of metastatic potential, treatment, and prevention. CD271 (the low affinity receptor for nerve growth factor) is a marker of a subpopulation of cells within HNSCC that have a greater capacity for tumor-initiation. In this study, we observed that activation of CD271 has a functional consequence of increased invasive and metastatic behavior by HNSCC cells. This behavior was dependent on the Snai2/Slug transcription factor, which is involved in epithelial-to-mesenchymal transition. Importantly, CD271 expression correlates with increased metastasis to regional lymph nodes in both murine and human oral squamous cell carcinoma. Thus, CD271 may be useful as both a marker and therapeutic target of metastatic HNSCC.