OSI-774 | Nur77 Signal

Cancer Letters

Nerve growth factor (NGF)-TrkA axis in head and neck squamous cell carcinoma triggers EMT and confers resistance to the EGFR inhibitor erlotinib

Chengzhong Lin, Zhenhu Ren, Xi Yang, Rong Yang, Yiming Chen, Zheqi Liu, Zhenlin Dai, Yu Zhang, Youya He, Chunye Zhang, Xu Wang, Wei Cao, Tong Ji

Abstract

Understanding the molecular mechanisms regulating tumor dissemination and therapeutic resistance is of central importance for effective cancer therapies. Here, we report that nerve growth factor (NGF) and its receptor TrkA facilitate epithelial-mesenchymal transition (EMT) and EGFR inhibitor resistance via STAT3 activation in head and neck squamous cell carcinoma (HNSCC). Both NGF and TrkA expression were elevated in HNSCC, indicating poor clinical outcomes. NGF was highly expressed in cancer cells and nerves in perineural niche, whereas TrkA expression was higher in cancer cells with perineural invasion. The NGF/TrkA axis could promote HNSCC cell dissemination and trigger EMT via STAT3 activation. Moreover, we discovered that the NGF/TrkA axis conferred resistance to the EGFR inhibitor erlotinib via EMT processes in HNSCC cells. Blocking TrkA signaling markedly reversed EMT and sensitized HNSCC cells to erlotinib in both in vitro and in vivo models. Overall, our results demonstrate novel evidence that the paracrine NGF/TrkA axis favors EMT and confers EGFR-targeted therapeutic resistance in HNSCC.

Nerve growth factor (NGF)-TrkA axis in head and neck squamous cell carcinoma triggers EMT and confers resistance to the EGFR inhibitor erlotinib

Chengzhong Lin1,2,3#, Zhenhu Ren2,3,4#, Xi Yang2,3,4, Rong Yang2,3,4, Yiming Chen2,3,4, Zheqi Liu2,3,4, Zhenlin Dai2,3,4, Yu Zhang2,3,4, Youya He2,3,4, Chunye Zhang2,3,5, Xu Wang2,3,4*, Wei Cao2,3,4*, Tong Ji2,3,4*

1 The 2nd Dental Center, Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, 200011, China
2 National Clinical Research Center for Oral Disease, Shanghai, 200011, China

3 Shanghai Key Laboratory of Stomatology & Shanghai Research Institute of Stomatology, Shanghai, 200011, China
4 Department of Oral and Maxillofacial-Head and Neck Oncology, Ninth People’s Hospital, Shanghai Jiao Tong University, School of Medicine, Shanghai 200011, China
5 Department of Oral Pathology, Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200011, China

These authors contributed equally.

*Corresponding authors: [email protected] for Xu Wang;

[email protected] for Wei Cao; [email protected] for Tong Ji

Abbraviations: NGF: Nerve growth factor; TrkA: Tropomyosin receptor kinase A; HNSCC:

Head and neck squamous cell carcinoma; EMT: Epithelial-mesenchymal transition; EGFR: Epidermal growth factor receptor; TKI: Tyrosine kinase inhibitor; PNI: Perineural invasion; STAT3: Signal transducer and activator of transcription 3; WB: Western blot; H&E: Hematoxylin and Eeosin; IHC: Immunohistochemistry; IC50: Half maximal inhibitory concentration; EGF: Epidermal growth factor; BDNF: Brain derived neurotrophic factor; NRTN: Neurturin; ARTN: Artemin; NT-3: Neurotrophin 3; NSCLC: Non-small cell lung cancer; AXL: Anexelekto; TGF-β: Transforming growth factor-β; IGF1R: Insulin-like growth factor 1 receptor; TCGA: The Cancer Genome Atlas; ZEB1: Zinc finger E-box binding homebox 1.

Abstract

Introduction

Head and neck squamous cell carcinoma (HNSCC) is a common and aggressive malignancy with a poor survival rate of 45% at 5 years[1]. Treatment of HNSCC patients remains challenging despite the identification of molecular pathways with high therapeutic potential, such as that of the epidermal growth factor receptor (EGFR)[2, 3]. Recently, anti-EGFR targeted therapies, such as the EGFR-selective tyrosine kinase inhibitor (TKIs) erlotinib, have demonstrated remarkable effective outcomes in various preclinical models[4]. However, the clinical benefits deployed by inhibition of EGFR in HNSCC patients remain insufficient to date. Low response rates (4-15%) have been observed, and nearly all responding patients eventually develop resistance after a median period of approximately 10 months, implicating intrinsic and acquired resistance mechanisms[5, 6].
Multiple mechanisms underlying poor response rates and resistance to EGFR TKIs have been proposed. In the context of HNSCC, studies have demonstrated two main mechanisms involving the process of resistance: constitutive activation of bypass survival pathways via other RTKs or alternative downstream signaling and the epithelial–mesenchymal transition (EMT), a reversible process through which cells undergo morphologic changes from epithelial phenotype to mesenchymal phenotype[7-10]. Unfortunately, despite therapeutic approaches targeting these mechanisms and their use in combination with EGFR inhibitors demonstrating some positive outcomes, many patients fail to benefit from these therapeutic strategies. Consequently, there remains an urgent need to elucidate the precise mechanisms underlying EGFR-TKI resistance and develop effective treatment strategies.

Accumulating evidence has suggested that therapeutic resistance is highly dependent on complex dynamic interactions between tumor cells and the tumor microenvironment[11, 12]. Perineural invasion (PNI) is the presence of cancer cells spread in, around, or along nerve bundles[13]. The prevalence of PNI varies considerably among cancer types and reaches 30–82% in HNSCC, indicating poor prognoses[14, 15]. Recent studies have revealed that the perineural niche is abundant in various neurotrophic factors and chemokines that can induce remodeling of the tumor microenvironment. By binding to cancer cell membranous receptors, these cellular components can promote cancer cell proliferation, inflammatory response, dissemination, and even resistance to chemotherapies[16, 17]. Hence, it is of interest to investigate whether the neurotrophic factors in the perineural microenvironment participate in EGFR-targeted therapy resistance.
Nerve growth factor (NGF), of the neurotrophin family, and its high-affinity transmembrane receptor TrkA, has been extensively studied mostly with regard to its role in nervous system development[18, 19]. The binding of mature NGF to TrkA induces TrkA receptor dimerization and phosphorylation, which activates downstream signaling pathways, including Ras/MAPK, PLCγ, and PI3K/AKT[20, 21]. Emerging evidence has indicated that the NGF/TrkA axis is involved in the perineural growth of several cancers. In pancreatic cancer, overexpressed NGF and TrkA contribute to PNI by prompting the nerve hyperplasia and suppressing the apoptosis of tumor cells[22]. Aberrant activation of TrkA in breast cancer cells leads to constitutive activation of the PI3K/AKT and Ras/MAPK pathways, resulting in tumor cell proliferation, invasion, and even metastatic dissemination to the central nervous

system[23, 24]. The NGF/TrkA axis also promotes prostate cancer cell proliferation, migration and the EMT through its cross-talk with the androgen receptor (AR) in castration-resistant prostate cancer[25, 26]. In HNSCC, although high NGF expression has been reported to be correlated with positive PNI, advanced tumor stages, and poor outcomes, the effects of TrkA expression on HNSCC patients are conflicting according to the limited researches[15]. Information regarding the TrkA expression level and the potential role of the NGF/TrkA axis in the progression of HNSCC remains unclear.
In the present study, we demonstrated that the NGF/TrkA axis triggers EMT and confers resistance to the EGFR inhibitor erlotinib in HNSCC. We found that elevated NGF/TrkA expression in HNSCC is significantly associated with positive PNI, poor clinical outcomes and triggers EMT. We then discovered that the NGF/TrkA axis confers resistance to the EGFR inhibitor erlotinib via STAT3 activation and EMT in HNSCC. Targeting TrkA signaling can markedly reverse EMT and sensitize HNSCC cells to erlotinib. Hence, we provide a rationale and experimental evidence for TrkA as a novel biomarker and an adjuvant therapeutic target in combination with EGFR-targeted therapies for HNSCC patients.

Materials and methods

The detailed methods were described in the Supplementary Materials and Methods, includ- ing patients and specimens, cell culture, RNA-Seq analysis, chemical compound and cytokine preparation, immunohistochemistry, Western blot analysis, Plasmid construction and cell transfection, cell viability assay, colony formation assay, real-time PCR, cell migration and

invasion assays, Immunofluorescence, Cell signaling array, animal experiments, and statistical analysis.

Results

Elevated NGF and TrkA levels indicate PNI and poor prognosis for HNSCC patients

To comprehensively investigate the NGF and TrkA status in HNSCC, we first assessed the mutation, deletion, amplification and expression levels of TrkA and NGF according to The Cancer Genome Atlas (TCGA) database. TrkA and NGF were detected with mutation or amplification in a relatively small proportion of HNSCC samples, whereas the TrkA and NGF genes were expressed at high levels compared with normal tissues (Fig. 1A, Additional file 1: Fig. S1A). Furthermore, when segregated by tumor PNI status, we found that NGF expression was elevated in PNI+ specimens and that there was no difference in TrkA expression between the PNI+ and PNI- specimens (Fig. 1B, Additional file 1: Fig. S1B). Thereafter, the expression levels of TrkA and NGF were investigated in 10 paired HNSCC tissue samples by western blotting. As shown in Fig. 1C, the expression levels of TrkA and NGF were obviously elevated in HNSCC tissues compared with the matched adjacent normal tissues.
To further explore the clinicopathological relevance of TrkA and NGF expression in HNSCC, we performed immunohistochemistry (IHC) assays to evaluate TrkA and NGF protein expression in 115 HNSCC tissues and 50 adjacent normal tissues. Notably, we observed that the expression levels of TrkA and NGF were significantly upregulated in HNSCC tissues compared with the matched adjacent normal tissues (Fig. 1D-F). In addition,

we also observed that the tumor cells around the nerve presented strong TrkA staining, whereas NGF was strongly stained in the nerves as well as the cancer cells (Fig. 1D). We then analyzed the correlations of TrkA and NGF expression with clinical and histopathological parameters (Additional file 2: Table S1). High TrkA expression levels were significantly correlated with larger tumor size (>4 cm, T3-T4, p=0.0327), lymph node metastasis (pN+, p=0.0054), and advanced tumor stage (III-IV, p=0.0214) (Fig. 1G), no obvious correlation with PNI status was found (Additional file 1: Fig. S1C). The high NGF expression levels were significantly correlated with positive PNI (p=0.0024), larger tumor size (>4 cm, T3-T4, p=0.0172), and higher pathological grade (p=0.0227) (Fig. 1H, Additional file 1: Fig. S1D). Moreover, Kaplan-Meier analyses and Cox regression analyses revealed that TrkA expression was significantly correlated with poor overall survival in HNSCC patients and was an independent predictor of overall survival in HNSCC (Fig. 1I, and K, Additional file 2: Table S2). While NGF could not be an independent predictor of survival in HNSCC, although it was strongly associated with poor survival of HNSCC patients (Fig. 1J, Additional file 2: Table S2). These data suggest that NGF/TrkA axis expression in the perineural niche may act as a potent chemoattractant for cancer cells in the PNI process and that TrkA can be an independent biomarker to predict the clinical outcomes of HNSCC patients.

TrkA is required for the maintenance of HNSCC cell viability in vitro and in vivo

To further explore the functional role of TrkA in HNSCC biological behavior, we investigated the effects of TrkA down-regulation on cell proliferation and metastasis. We first analyzed

TrkA protein expression in six HNSCC cell lines (HN4, HN6, HN13, HN30, CAL27 and SCC4) (Additional file 1: Fig. S2A). We then silenced TrkA expression with specific siRNAs in HN4 and HN6 cells which had relatively high expression of TrkA, and two siRNAs (siTrkA1 and siTrkA3) were able to efficiently suppress TrkA expression in both cell lines (Additional file 1: Fig. S2B). Knockdown of TrkA remarkably impaired cell proliferation and colony formation compared with the negative control in both HN4 and HN6 cell lines (Fig. 2A-C). Moreover, transwell migration and invasion assays revealed that the TrkA silencing significantly inhibited cell migration and invasion capability (Fig. 2D, and E). These results indicated that TrkA is implicated in the proliferation, migration and invasion of HNSCC cells. To confirm that TrkA silencing inhibits cell growth in vivo, subcutaneous tumor growth of the HN6 cells was monitored. HN6-shTrkA3 cells and-shNC cells were subcutaneously injected into 2 lateral croups of nude mice, and tumor growth were recorded over 23 days. Consistent with the in vitro findings, the tumor growth decreased in the TrkA-knockdown group compared with the negative control group. The tumor volume and weight were significantly decreased in TrkA-silenced xenograft mice (Fig. 2F, G). The transfection efficiency of TrkA was confirmed by TrkA immunohistochemistry assay in xenograft tissues (Fig. 3H). In addition, hematoxylin and eosin (H&E) staining and Ki67 staining were performed to verify the results. The percentage of Ki-67-positive stained cells was significantly decreased in the TrkA-knockdown xenograft tumor tissues compared with the negative control group (Fig. 3H). Taken together, these findings indicate that TrkA potentially
plays a significant role in promoting HNSCC development.
NGF induces EMT and cancer cell dissemination

Limited studies have suggested that soluble NGF can trigger EMT processes and promote cancer cell migration and invasion in castration-resistant prostate cancer[25]. Therefore, we determined the role of NGF/TrkA axis on HNSCC cell biological behaviors. Interestingly, the phase-contrast image analysis showed that HNSCC cell lines underwent a morphologic transition from epithelial-like to spindle-like phenotype in response to NGF stimulation (Fig. 4A). We next performed RNA-Seq analysis to identify the transcripts that were differentially regulated in HN4 cells after NGF incubation (GSE136896). Using the standard 2-fold change in expression as the threshold criterion, a total of 2530 genes was specifically changed in response to NGF (Additional file 2: Table. S3). Intriguingly, Gene Ontology (GO) enrichment analysis data revealed that the differentially expressed genes were significantly associated with extracellular functional pathways, including extracellular matrix organization, cell adhesion, regulation of cell proliferation, cell migration, and the epithelial to mesenchymal transition (Fig. 4B). Among them, 45 genes initially identified as significantly related to cell invasion and adhesion were also identified as belonging to one of the molecule clusters with the most alterations (Fig. 4C). As shown in Fig. 4D-F, real-time PCR and immunoblotting assays demonstrated that NGF stimulation strongly promoted the expression of mesenchymal genes such as N-cadherin, Vimentin, Twist1, Snail1 and Snail2 but attenuated the expression of the epithelial gene E-cadherin. In addition, immunofluorescence staining also indicated that NGF stimulation induced a change of decreased E-cadherin expression and increased

Vimentin expression in HN4 and HN6 cells, supporting the hypothesis that NGF induces the EMT (Fig. 4G). Moreover, the transwell assays revealed that NGF incubation significantly enhanced cell migration and invasion abilities in HN4 and HN6 cells (Fig. 4H, and I). These findings indicate that NGF induces the EMT and HNSCC cell dissemination.

Targeting TrkA reverses NGF-induced EMT in HNSCC cells

To gain further insight into the function of TrkA in NGF-induced EMT processes in HNSCC, we evaluated E-cadherin, Vimentin and Twist1 expression levels in TrkA-knockdown cells. As shown in Fig. 5A, genetic silencing of TrkA effectively increased E-cadherin expression and suppressed Vimentin and Twist1 expression in the absence or presence of NGF, as indicated in Fig. 4A. Similarly, when assessing the effects of TrkA inhibition using its specific inhibitor GNF5837 in HNSCC cells, we found that the expression of E-cadherin was obviously upregulated, whereas that of Vimentin and Twist1 was suppressed in response to NGF stimulation (Fig. 4B, Additional file 1: Fig. S3). Correspondingly, immunofluorescence staining also revealed that the TrkA inhibitor GNF5837 effectively increased E-cadherin expression while suppressing Vimentin expression in response to NGF, as indicated in Fig. 4C. In addition, we further investigated whether TrkA affected EMT-associated kinase expression in xenograft tumors. As shown in Fig. 4D, the shTrkA group of HN6 cells had a significantly upregulated expression level of E-cadherin and suppressed levels of Vimentin and Twist1.
We next asked whether ectopic expression of TrkA could induce EMT processes in response to NGF. We selected CAL27 cells, with relatively lower expression of TrkA, instead

of HN4 and HN6 cells for use in this experiment. As shown in Fig. 4E, ectopic expression of TrkA (TrkA-OE) was established in CAL27 cells. Indeed, in TrkA-OE cells, the expression of E-cadherin was suppressed, while the mesenchymal genes such as Twist1 and Vimentin were upregulated (Fig. 4F). Accordingly, immunofluorescence staining also revealed that TrkA overexpression reduced E-cadherin expression while increasing Vimentin expression in the absence or presence of NGF, as indicated in Fig. 4G. Collectively, these results indicate that TrkA mediates NGF-induced EMT in HNSCC cells.

NGF/TrkA axis confers resistance to the EGFR inhibitor erlotinib in HNSCC cells Several studies have revealed that EMT was associated with the sensitivity of cancer cells to EGFR-targeted therapeutic resistance in several types of cancer[7, 27]. Given the implication of NGF/TrkA axis in mediating EMT in HNSCC, we investigated whether the NGF/TrkA axis could rescue the growth-inhibitory effect of EGFR inhibitor erlotinib in cancer cells. First, we treated a panel of 6 HNSCC cell lines, shown in Fig. 2S, with various doses of erlotinib and measured erlotinib IC50 values, and we observed a significant correlation between TrkA expression and erlotinib IC50 (Fig. 5A). To further identify the role of NGF in mediating resistance to anti-EGFR therapy, we screened a panel of neural growth factors, including NGF, EGF, brain-derived neurotrophic factor (BDNF), neurturin (NRTN), artemin (ARTN), and neurotrophin 3 (NT-3), for their capacity to rescue HNSCC cell lines from erlotinib. Cell viability was measured to determine cell sensitivity to erlotinib, and the data revealed that, although EGF and BDNF tended to induce insensitivity to erlotinib in HNSCC cell lines,

NGF showed the strongest induction of resistance to erlotinib (Fig. 5B, Additional file 1: Fig. S4). These results suggest that the NGF/TrkA axis plays an important role in HNSCC cell resistance to erlotinib.
We then investigated whether TrkA inhibition could overcome NGF-induced erlotinib resistance in the HNSCC cells. As presented in Fig. 5C, it was found that HN6 cells cultured 72 hours in medium containing NGF (40 ng/ml) had decreased sensitivity to erlotinib. The IC50 of erlotinib was 5.47 µmol/L in HN6 siNC cells and 12.4 µmol/L in siNC cells treated with NGF. Interestingly, knocking down TrkA induced a 4- to 5-fold increase in erlotinib sensitivity and reversed NGF-induced erlotinib resistance. Similar results were observed with the HN4 cells (Additional file 1: Fig. S5). We also examined the erlotinib sensitivity of the TrkA-overexpressing CAL27 cells. Ectopically expressed TrkA decreased the sensitivity of CAL27 cells to erlotinib; moreover, the CAL27 TrkA-OE cells showed a strongly elevated IC50 value for erlotinib when treated with NGF(Fig. 5D).
In addition to the genetic regulation of TrkA protein expression, the synergistic effect of the TrkA inhibitor GNF5837 with erlotinib was tested. As shown in Fig. 5E, HN4 cell growth was significantly inhibited by the combined treatment of erlotinib and GNF5837. Erlotinib at doses of 0.03 µmol/L and 0.27 µmol/L inhibited cell growth by 6.4± 1.4% and 27.2 ± 3.1%, respectively, whereas GNF5837 alone at the IC25 and in combination with erlotinib inhibited cell growth by 14.2 ± 0.8% and 40 ± 2.1%, respectively. This finding represents an approximately 2-fold increase in sensitivity. A similar trend was observed with the addition of GNF5837 at the IC50 to erlotinib, and similar results were observed in HN6 cells (Fig. 5F).

These results indicate that the NGF/TrkA axis potentially mediates erlotinib resistance in HNSCC cells and that TrkA inhibition could reverse this drug sensitivity and enhance erlotinib efficacy.

Targeting TrkA sensitizes HN6 xenograft tumors to erlotinib in vivo

To confirm the validity of these results in vivo, we generated a tumor xenograft model using HN6 cells to determine the combined therapeutic effect of the TrkA inhibitor GNF5837 and erlotinib. When the tumor size reached 100 mm3, the xenograft-bearing mice were treated with the vehicle, erlotinib (50 mg/kg), GNF5837 (25 mg/kg), or a combination of erlotinib (50 mg/kg) and GNF5837 (25 mg/kg) every day for 25 days. A significant decrease in tumor volume was observed for each treatment group compared with the control group. Importantly, the combination treatment of erlotinib and GNF5837 resulted in substantial and significant inhibition of tumor growth compared with tumors treated with erlotinib or GNF5837 alone (Fig. 5G).. This combination therapy caused an 89.9% inhibition in tumor volume compared with the control group, whereas GNF5837 alone caused a 65.5% inhibition, and erlotinib alone caused a 74.5% inhibition, compared with the control, and both were accompanied by obviously decreased tumor weights for the treatment groups (Fig. 5H, and I). During the experiments, no obvious weight loss was observed in mice treated with GNF5837, erlotinib, or the combination of both (Additional file 1: Fig. S6). Furthermore, the mean number of proliferating and apoptotic cells in the xenograft tumors was evaluated by immunohistochemical analysis of Ki-67 staining and terminal deoxynucleotidyl transferase

dUTP nick end labeling (TUNEL) assays, respectively. Consistent with the inhibitory effect against tumor growth, an increase in apoptotic cells and a decrease in proliferating cells were observed for all treatment groups compared with the control (Fig. 5J). Moreover, the combined treatment group had significantly higher TUNEL staining levels and lower Ki-67 staining levels than the groups treated with erlotinib or regorafenib alone. Collectively, these findings in vivo recapitulated the in vitro results, indicating that concurrent inhibition of TrkA has the potential to significantly improve the efficacy of erlotinib treatment for HNSCC.

STAT3 activation contributes to the NGF/TrkA axis mediated erlotinib resistance of HNSCC cells
To clarify the underlying signaling pathways that NGF/TrkA axis induce to confer erlotinib resistance, we first determined whether there is a direct interaction between TrkA and EGFR. Cells were starved in growth factor-free medium overnight and subsequently stimulated with or without either NGF (40 ng/ml) or EGF (40 ng/ml) for 30 min. As shown in Fig. 6A, cells exhibited increased levels of TrkA phosphorylation and EGFR phosphorylation in response to NGF and EGF treatment, respectively, whereas no activated signaling was observed in the cross-examined samples. We then investigated whether turning on and off these two molecules can mediate the downstream signaling pathways. The HN4 and HN6 cells were deprived of serum overnight and incubated with NGF (40 ng/ml) or EGF (40 ng/ml) in the presence or absence of erlotinib for 30 min. The cell signaling pathways were then analyzed. The results showed that erlotinib decreased STAT3 phosphorylation in the absence of NGF,

while administration of NGF promoted potent STAT3 activation, which was resistant to erlotinib. In contrast, the ability of erlotinib to suppress other key signaling molecules, such as ERK1/2 and AKT, was not affected by the activation of the TrkA pathway (Fig. 6B). Consistently, western blotting also revealed that administration of NGF induced STAT3 activation regardless of erlotinib treatment (Fig. 6C), suggesting that the NGF/TrkA axis functions as the dominant regulator of the STAT3 signaling pathway in HNSCC. Moreover, we found that blocking TrkA signaling by its inhibitor GNF5837 significantly suppressed STAT3 activity in HN4 and HN6 cells, underscoring the role of TrkA signaling in mediating STAT3 activity in HNSCC (Fig. 6D). Thus, these data suggest that the NGF/TrkA axis induces sustained activation of STAT3 signaling in erlotinib-treated HNSCC cells.
To further evaluate the role of STAT3 in mediating the response of HNSCC to EGFR inhibitor erlotinib, we assessed whether genetic or pharmacologic inhibition of STAT3 in HNSCC cell lines could restore sensitivity to erlotinib. As presented in Fig. 6E, STAT3 expression was silenced in HN4 and HN6 cells by specific siRNAs, according to the previous study. We found that downregulation of STAT3 completely abrogated the protective effect of NGF in erlotinib-treated HN4 and HN6 cells, reversing NGF-induced erlotinib resistance. In addition, STAT3 silencing alone induced a 2- to 3-fold increase in erlotinib sensitivity (Fig. 6F, G). Besides we observed a similar increase in sensitivity to erlotinib when using the STAT3 inhibitor cryptotanshinone, which has been demonstrated to have an obvious inhibitory effect on STAT3 signaling in HNSCC cells. Cryptotanshinone on its own discernibly inhibited the growth of HN4 and HN6 cells at the indicated concentrations, and it remarkably sensitized

the cells to erlotinib even in the presence of NGF (Fig. 6H, I). Collectively, our results indicated that STAT3 activation contributes, at least in part, to the NGF/TrkA axis-mediated protection from EGFR inhibitor erlotinib in HNSCC.

NGF/TrkA axis induces the EMT and promotes erlotinib resistance via STAT3 activation
To demonstrate whether there is a causal link between TrkA-STAT3 activation, EMT, and EGFR inhibitor resistance, we evaluated E-cadherin, Vimentin and Twist1 expression levels in erlotinib pre-treated TrkA knockdown cells in response to NGF stimulation. As shown in Fig. 7A, western blot analysis results revealed that TrkA silencing in HN6 cells abrogated the induction of EMT and partially attenuated the downregulation of E-cadherin and the upregulation of Vimentin and Twist1 following NGF treatment. In addition, erlotinib treatment had no obvious diminish effect on the NGF-induced EMT process in both NC- and TrkA-knockdown cells.. We then further examined the influence of TrkA overexpression on NGF induced EMT and erlotinib resistance. Obviously, increased TrkA overexpression in CAL27 cells promoted the NGF-stimulated induction of EMT. In addition, E-cadherin expression was decreased, and the Vimentin and Twist1 expression were enhanced in TrkA-OE cells exposed to erlotinib when compared to NC cells exposed to erlotinib. Moreover, erlotinib treatment had less diminish effect on NGF-induced EMT in TrkA-OE cells compared to NC cells (Fig. 7B). These results suggested the involvement of EMT in NGF induced erlotinib resistance.

To investigate the requirements for inducing EMT processes during NGF-induced erlotinib resistance, we assessed whether genetic inhibition of Twist1 or Slug in HNSCC cells would restore cell sensitivity to erlotinib. Twist1 and Slug expression was suppressed with specific siRNAs in HN4 and HN6 cells (Additional file 1: Fig. S7A-B). Twist1 knockdown could enhance the inhibition effect of erlotinib and effectively abrogate the protective effect of NGF in cells treated with erlotinib (Fig. 7C-D). While single Slug knockdown could also increase HN4 and HN6 cell sensitivity to erlotinib, but partially abrogated the protective effect of NGF (Fig. 7E-F). Thus, these observations indicated that EMT is required for the NGF/TrkA axis-induced resistance to erlotinib, and Twist1 is a key transcriptional factor for this process.
We next examined whether the inhibition of STAT3 would suppress the NGF-mediated EMT process. As shown in Fig. 7F, the results demonstrated that STAT3 silencing in HN6 cells partially attenuated the downregulation of E-cadherin and the upregulation of Vimentin and Twist1 in response to NGF treatment. We then further analyzed the EMT markers and p-STAT3 levels in the context of combined GNF5837 and erlotinib therapy in HN6 xenografts. Immunohistochemistry staining showed a significant decrease in the phosphorylation level of TrkA and STAT3 in the GNF5837 treatment group alone or in the combined treatment group, and the expression of E-cadherin was relatively high and the expression of Vimentin and Twist1 was low (Additional file 1: Fig. S8). Taken together, these findings suggest that the paracrine effect of NGF/TrkA axis induces the EMT and enhances erlotinib resistance via STAT3 activation in HNSCC (Fig. 7F).

EMT markers are coexpressed with TrkA and predict poor outcomes for HNSCC patients
To characterize the correlations of the EMT markers and the NGF/TrkA axis, first, the expression levels of E-cadherin, Vimentin, and Twist1 were investigated in 10 paired HNSCC tissue samples by western blotting. As shown in Fig. 8A, the expression levels of Vimentin and Twist1 were obviously elevated, whereas the expression of E-cadherin was decreased in HNSCC tissues compared with the matched adjacent normal tissues. Quantification using relative band intensity revealed a significant correlation of TrkA with Vimentin and Twist1 and a significant correlation between NGF and Twist1 (Fig. 8B-D, Additional file 1: Fig. S9A-C). To further validate the overall correlation of the EMT markers with NGF and TrkA, consecutive sections of all specimens from the patient cohort shown in Table S1 were stained for E-cadherin, Vimentin, and Twist1 expression (Additional file 1: Fig. S9D). Spearman correlation analysis of the IHC scores confirmed the obvious positive correlation of TrkA with Vimentin and with Twist1, whereas a significantly negative correlation between TrkA and E-cadherin (Additional file 2: Table S4). We also found a positive correlation of NGF with Twist1 and with Vimentin (Additional file 2: Table S5). Concurrent expression of TrkA, E-cadherin, Vimentin, and Twist1 in HNSCC is shown in Fig. 8E. The correlations of the EMT markers with the clinical and histopathological parameters were also analyzed. Low E-cadherin expression levels significantly correlated with large tumor size, lymph node metastasis, and advanced tumor stage, and high Vimentin expression levels significantly correlated with poor pathological grade, lymph node metastasis, and advanced tumor stage

(Additional file 2: Table S6). In addition, the Kaplan-Meier analysis revealed that low expression of E-cadherin or high expression of Vimentin was associated with HNSCCs with poor overall survival, while no significant correlation was observed between Twist1 expression and survival rate. Hence, NGF/TrkA represents a significant signaling axis that mediates the cell EMT processes, promotes cell resistance to the EGFR inhibitor erlotinib, and indicates the HNSCC patients with poor clinical performance.

Discussion

Although anti-EGFR therapies have achieved significant clinical success in several epithelial cancers, outcomes in HNSCC are still disappointing[28]. Understanding the molecular mechanisms regulating tumor dissemination and therapeutic resistance is of central importance for developing effective cancer therapies. Here, we provide clinically novel insight into these processes by revealing that the heterogeneous expression of nerve growth factor (NGF) and its high-affinity transmembrane receptor TrkA in HNSCC can trigger the epithelial-mesenchymal transition (EMT) and confer resistance to the EGFR inhibitor erlotinib.
Our study demonstrated novel evidence that the NGF/TrkA axis promotes cancer cell dissemination in the perineural niche in HNSCC with perineural invasion (PNI). PNI is a mode of cancer progression in which cancer cells spread in, around, or along the nerves[13]. Studies have revealed that the perineural niche, the microenvironment in the space surrounding a nerve, is abundant in multiple neurotrophic factors. These neurotrophins and

their receptors are critical signaling molecules that are involved in mediating cell-to-cell communication in tumor microenvironment and promoting a wide range of complex functions for cancer cells, including proliferation, differentiation, motility, and metabolism[17, 29, 30]. The NGF/TrkA axis, which has been studied most extensively for its role in nervous system development, has been implicated in tumor growth and metastasis in several cancers[31-33]. Here, we demonstrated that NGF/TrkA expression was significantly associated with positive PNI, larger tumor size, lymph node metastasis, advanced tumor stage, and reduced survival for patients with HNSCC. TrkA can be an independent biomarker to predict the clinical outcomes of HNSCC patients. Moreover, we revealed that NGF and TrkA expression were heterogeneous in HNSCC tissues with PNI, in which NGF expression was elevated in the nerves as well as the cancer cells, while TrkA expression was higher in cancer cells in the perineural niche. These data suggest that NGF/TrkA axis expression in the perineural niche may act as a potent chemoattractant for cancer cells during the PNI process. However, the precise mechanism remains to be elucidated.
Our study underscores the role of NGF/TrkA axis in mediating epithelial-mesenchymal transition (EMT) and cell dissemination in HNSCC. EMT is a reversible process in which epithelial cells lose their polarity and adhesive abilities and shift toward the mesenchymal phenotype. The molecular hallmarks of the EMT are downregulation of E-cadherin accompanied by the upregulation of mesenchymal markers such as N-cadherin and Vimentin. Twist and Snail are two major transcriptional factors that contribute to the induction of the EMT process[34]. Multiple mechanisms underlying EMT processes have been demonstrated,

including the upregulation of the AXL, TGF-β, and IL-6 signaling axes[35-37]. In the current study, we found that NGF/TrkA axis promoted the expression of mesenchymal genes such as N-cadherin, Vimentin, Twist1, Snail1 and Snail2 but attenuated the expression of the epithelial gene E-cadherin in HNSCC cell lines. Concomitantly, HNSCC cells underwent a morphologic transition from an epithelial-like to a spindle-like phenotype in response to NGF stimulation. Furthermore, inhibiting TrkA through specific siRNAs or a selective tyrosine kinase inhibitor (TKI) obviously reversed the NGF-induced EMT-associated kinase switching in HNSCC cells. We also discovered that the NGF/TrkA axis induced EMT processes mainly via STAT3 activation. Indeed, inhibiting STAT3 obviously abolished the NGF/TrkA axis-mediated upregulation of Vimentin and Twist1 and downregulation of E-cadherin. Hence, we provide clinically novel evidence that the NGF/TrkA axis triggers the EMT in HNSCC, which is essential for cancer cell invasion and dissemination in the perineural niche.
Efforts to identify drivers of poor response rates and resistance to EGFR TKIs have revealed multiple mechanisms of resistance. In the context of HNSCC, unlike non-small cell lung cancer (NSCLC), where EGFR mutations confer sensitivity to EGFR-TKIs, the prevalence of EGFR-activating mutations in HNSCC are quite low[38]. Coactivation of other RTKs and downstream signaling have been implicated in the limited response of HNSCC cells to TKIs[9, 39]. However, the precise mechanisms mediating EGFR TKI sensitivity in HNSCC remain to be further elucidated. There is growing appreciation for the need for combined targeted therapy studies. In the current study, we demonstrated that elevated NGF/TrkA axis conferred resistance to erlotinib in HNSCC. Inhibiting TrkA in HNSCC cells

increased cell sensitivity to erlotinib in vitro and in vivo. Our data further revealed that STAT3 was a primary downstream signaling of the cross-talk between NGF/TrkA axis and EGFR signaling pathway. Studies have demonstrated that STAT3 is activated in many cancers via both EGFR-dependent and -independent pathways, contributing to EGFR inhibitor resistance[40]. Our results also revealed that genetic or pharmacologic inhibition of STAT3 in HNSCC cells could restore cell sensitivity to erlotinib. Therefore, concurrent inhibition of both EGFR and TrkA is necessary to sufficiently attenuate STAT3 signaling and induce a significant response in HNSCC. These findings collectively provide a rationale for the combined use of TrkA- and EGFR-targeted therapeutics.
In addition to the coactivation of downstream STAT3 signaling pathway, NGF/TrkA axis can induce EMT, which may also play a significant role in mediating resistance to EGFR inhibitor treatment in HNSCC. EMT has been linked to EGFR-targeted therapeutic resistance, previous studies have demonstrated that upregulation of AXL, VGF and IGF1R signaling axes are drivers of EMT-mediated acquired resistance to EGFR TKIs[27, 41, 42]. EMT transcription factors, which induce global transcriptional changes leading to EMT, specifically Snail2, and Twist1, have been found to confer resistance to EGFR TKIs[43, 44]. Here, we established that the NGF/TrkA/STAT3 signaling pathway induced EMT phenotypic alterations in HNSCC cell lines, orchestrating the EMT-associated kinase switch and conferring resistance to erlotinib in HNSCC cells. Importantly, our findings revealed that Twist1 expression was required for the NGF/TrkA axis-induced resistance to the EGFR inhibitor erlotinib. Silencing Twist1 effectively abrogated the protective effect of NGF and restored erlotinib sensitivity in the HNSCC cell lines. These results suggest that the NGF/TrkA signaling-induced EMT processes play critical roles in EGFR-TKI resistance.
Taken together, our findings were the first to reveal that NGF/TrkA axis is significantly elevated in HNSCC perineural niche and favors the EMT and EGFR-targeted therapeutic resistance, thereby predicting poor clinical outcomes. As TrkA inhibition can markedly reverse EMT processes and sensitize HNSCC cells to erlotinib, the combined treatment of TrkA and EGFR is worthy of future preclinical investigation in erlotinib resistance settings. In summary, our findings provide evidence for the role of TrkA, OSI-774 a neurotrophic tyrosine kinase receptor, to serve as a novel biomarker and therapeutic target for overcoming erlotinib resistance in HNSCC patients.

Additional files

Additional file 1: Supplementary Methods and Figures. Additional file 2: Supplementary Tables.

Availability of data and materials

The dataset used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Conflict of interest

The authors declare that they have no competing interests.

Authors’ contributions
CL, XW, WC, and TJ designed this study. CL, ZR, XY, ZL, XZ, RY, ZD, and YZ performed experiments. CL, ZR, and XY analyzed the data . CL, ZR, and XW wrote the manuscript. XW, WC, and TJ supervised the study. All authors read and approved the final manuscript.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (81702675, 81872189, 81672745, 81772886, 81972589), Science and Technology Commission of Shanghai Municipality (YDZX20173100004422), and Shanghai Youth Medical Talents-Specialist Program.

References

1. Ferlay J, Soerjomataram I, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray F: Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer 2015, 136(5):E359-386.
2. Kalyankrishna S, Grandis JR: Epidermal growth factor receptor biology in head and neck cancer. J Clin Oncol 2006, 24(17):2666-2672.
3. Cramer JD, Burtness B, Le QT, Ferris RL: The changing therapeutic landscape of head and neck cancer. Nat Rev Clin Oncol 2019.
4. Calvo E, Malik SN, Siu LL, Baillargeon GM, Irish J, Chin SF, Santabarbara P, Kreisberg JI, Rowinsky EK, Hidalgo M: Assessment of erlotinib pharmacodynamics in tumors and skin of patients with head and neck cancer. Ann Oncol 2007, 18(4):761-767.
5. Martins RG, Parvathaneni U, Bauman JE, Sharma AK, Raez LE, Papagikos MA, Yunus F, Kurland BF, Eaton KD, Liao JJ et al: Cisplatin and radiotherapy with or without erlotinib in locally advanced squamous cell carcinoma of the head and neck: a randomized phase II trial. J Clin Oncol

2013, 31(11):1415-1421.
6. Massarelli E, Lin H, Ginsberg LE, Tran HT, Lee JJ, Canales JR, Williams MD, Blumenschein GR, Jr., Lu C, Heymach JV et al: Phase II trial of everolimus and erlotinib in patients with platinum-resistant recurrent and/or metastatic head and neck squamous cell carcinoma. Ann Oncol 2015, 26(7):1476-1480.
7. Pan M, Schinke H, Luxenburger E, Kranz G, Shakhtour J, Libl D, Huang Y, Gaber A, Pavsic M, Lenarcic B et al: EpCAM ectodomain EpEX is a ligand of EGFR that counteracts EGF-mediated epithelial-mesenchymal transition through modulation of phospho-ERK1/2 in head and neck cancers. PLoS Biol 2018, 16(9):e2006624.
8. Brand TM, Hartmann S, Bhola NE, Li H, Zeng Y, O’Keefe RA, Ranall MV, Bandyopadhyay S, Soucheray M, Krogan NJ et al: Cross-talk Signaling between HER3 and HPV16 E6 and E7 Mediates Resistance to PI3K Inhibitors in Head and Neck Cancer. Cancer research 2018, 78(9):2383-2395.
9. Lin C, Lu W, Ren Z, Tang Y, Zhang C, Yang R, Chen Y, Cao W, Wang L, Wang X et al: Elevated RET expression enhances EGFR activation and mediates EGFR inhibitor resistance in head and neck squamous cell carcinoma. Cancer Lett 2016, 377(1):1-10.
10. Keysar SB, Le PN, Anderson RT, Morton JJ, Bowles DW, Paylor JJ, Vogler BW, Thorburn J, Fernandez P, Glogowska MJ et al: Hedgehog signaling alters reliance on EGF receptor signaling and mediates anti-EGFR therapeutic resistance in head and neck cancer. Cancer research 2013, 73(11):3381-3392.
11. De Palma M, Biziato D, Petrova TV: Microenvironmental regulation of tumour angiogenesis. Nat Rev Cancer 2017, 17(8):457-474.
12. Boilly B, Faulkner S, Jobling P, Hondermarck H: Nerve Dependence: From Regeneration to Cancer. Cancer Cell 2017, 31(3):342-354.
13. Liebig C, Ayala G, Wilks JA, Berger DH, Albo D: Perineural invasion in cancer: a review of the literature. Cancer 2009, 115(15):3379-3391.
14. Zhu J, Zhou R, Wang Y, Yu M: Perineural invasion as a prognostic factor in head and neck squamous cell carcinoma: a systematic review and meta-analysis. Acta Otolaryngol 2019:1-6.
15. Shen WR, Wang YP, Chang JY, Yu SY, Chen HM, Chiang CP: Perineural invasion and expression of nerve growth factor can predict the progression and prognosis of oral tongue squamous cell carcinoma. J Oral Pathol Med 2014, 43(4):258-264.
16. Amit M, Na’ara S, Gil Z: Mechanisms of cancer dissemination along nerves. Nat Rev Cancer 2016, 16(6):399-408.
17. Lin C, Cao W, Ren Z, Tang Y, Zhang C, Yang R, Chen Y, Liu Z, Peng C, Wang L et al: GDNF secreted by nerves enhances PD-L1 expression via

JAK2-STAT1 signaling activation in HNSCC. Oncoimmunology 2017,
6(11):e1353860.
18. Denk F, Bennett DL, McMahon SB: Nerve Growth Factor and Pain Mechanisms. Annu Rev Neurosci 2017, 40:307-325.
19. Tomlinson RE, Li Z, Li Z, Minichiello L, Riddle RC, Venkatesan A, Clemens TL: NGF-TrkA signaling in sensory nerves is required for skeletal adaptation to mechanical loads in mice. Proc Natl Acad Sci U S A 2017, 114(18):E3632-E3641.
20. Wong HLX, Qin HY, Tsang SW, Zuo X, Che S, Chow CFW, Li X, Xiao HT, Zhao L, Huang T et al: Early life stress disrupts intestinal homeostasis via NGF-TrkA signaling. Nat Commun 2019, 10(1):1745.
21. Gudasheva TA, Povarnina PY, Antipova TA, Firsova YN, Konstantinopolsky MA, Seredenin SB: Dimeric dipeptide mimetics of the nerve growth factor Loop 4 and Loop 1 activate TRKA with different patterns of intracellular signal transduction. J Biomed Sci 2015, 22:106.
22. Li J, Ma J, Han L, Xu Q, Lei J, Duan W, Li W, Wang F, Wu E, Ma Q et al: Hyperglycemic tumor microenvironment induces perineural invasion in pancreatic cancer. Cancer Biol Ther 2015, 16(6):912-921.
23. Pundavela J, Roselli S, Faulkner S, Attia J, Scott RJ, Thorne RF, Forbes JF, Bradshaw RA, Walker MM, Jobling P et al: Nerve fibers infiltrate the tumor microenvironment and are associated with nerve growth factor production and lymph node invasion in breast cancer. Mol Oncol 2015, 9(8):1626-1635.
24. Yang X, Shen H, Buckley B, Chen Y, Yang N, Mussell AL, Chernov M, Kobzik L, Frangou C, Han SX et al: NTRK1 is a positive regulator of YAP oncogenic function. Oncogene 2019, 38(15):2778-2787.
25. Di Donato M, Cernera G, Migliaccio A, Castoria G: Nerve Growth Factor Induces Proliferation and Aggressiveness In Prostate Cancer Cells. Cancers (Basel) 2019, 11(6).
26. Singh R, Karri D, Shen H, Shao J, Dasgupta S, Huang S, Edwards DP, Ittmann MM, O’Malley BW, Yi P: TRAF4-mediated ubiquitination of NGF receptor TrkA regulates prostate cancer metastasis. J Clin Invest 2018, 128(7):3129-3143.
27. Weng CH, Chen LY, Lin YC, Shih JY, Lin YC, Tseng RY, Chiu AC, Yeh YH, Liu C, Lin YT et al: Epithelial-mesenchymal transition (EMT) beyond EGFR mutations per se is a common mechanism for acquired resistance to EGFR TKI. Oncogene 2019, 38(4):455-468.
28. Sacco AG, Cohen EE: Current Treatment Options for Recurrent or Metastatic Head and Neck Squamous Cell Carcinoma. J Clin Oncol 2015, 33(29):3305-3313.
29. Zareba P, Flavin R, Isikbay M, Rider JR, Gerke TA, Finn S, Pettersson A, Giunchi F, Unger RH, Tinianow AM et al: Perineural Invasion and Risk of

Lethal Prostate Cancer. Cancer Epidemiol Biomarkers Prev 2017,
26(5):719-726.
30. Xia Q, Bai QR, Dong M, Sun X, Zhang H, Cui J, Xi H, Hu XL, Shen Q, Chen L: Interaction Between Gastric Carcinoma Cells and Neural Cells Promotes Perineural Invasion by a Pathway Involving VCAM1. Dig Dis Sci 2015, 60(11):3283-3292.
31. Griffin N, Faulkner S, Jobling P, Hondermarck H: Targeting neurotrophin signaling in cancer: The renaissance. Pharmacol Res 2018, 135:12-17.
32. Demir IE, Tieftrunk E, Schorn S, Friess H, Ceyhan GO: Nerve growth factor & TrkA as novel therapeutic targets in cancer. Biochim Biophys Acta 2016, 1866(1):37-50.
33. Dudas J, Dietl W, Romani A, Reinold S, Glueckert R, Schrott-Fischer A, Dejaco D, Johnson Chacko L, Tuertscher R, Schartinger VH et al: Nerve Growth Factor (NGF)-Receptor Survival Axis in Head and Neck Squamous Cell Carcinoma. Int J Mol Sci 2018, 19(6).
34. Brabletz T, Kalluri R, Nieto MA, Weinberg RA: EMT in cancer. Nat Rev Cancer 2018, 18(2):128-134.
35. Suresh D, Zambre A, Mukherjee S, Ghoshdastidar S, Jiang Y, Joshi T, Upendran A, Kannan R: Silencing AXL by covalent siRNA-gelatin-antibody nanoconjugate inactivates mTOR/EMT pathway and stimulates p53 for TKI sensitization in NSCLC. Nanomedicine 2019, 20:102007.
36. Syed V: TGF-beta Signaling in Cancer. J Cell Biochem 2016,
117(6):1279-1287.
37. Rokavec M, Oner MG, Li H, Jackstadt R, Jiang L, Lodygin D, Kaller M, Horst D, Ziegler PK, Schwitalla S et al: IL-6R/STAT3/miR-34a feedback loop promotes EMT-mediated colorectal cancer invasion and metastasis. J Clin Invest 2014, 124(4):1853-1867.
38. Zhou Q, Zhang XC, Chen ZH, Yin XL, Yang JJ, Xu CR, Yan HH, Chen HJ, Su J, Zhong WZ et al: Relative abundance of EGFR mutations predicts benefit from gefitinib treatment for advanced non-small-cell lung cancer. J Clin Oncol 2011, 29(24):3316-3321.
39. Stabile LP, He G, Lui VW, Thomas S, Henry C, Gubish CT, Joyce S, Quesnelle KM, Siegfried JM, Grandis JR: c-Src activation mediates erlotinib resistance in head and neck cancer by stimulating c-Met. Clin Cancer Res 2013, 19(2):380-392.
40. Sen M, Joyce S, Panahandeh M, Li C, Thomas SM, Maxwell J, Wang L, Gooding WE, Johnson DE, Grandis JR: Targeting Stat3 abrogates EGFR inhibitor resistance in cancer. Clin Cancer Res 2012, 18(18):4986-4996.
41. Hwang W, Chiu YF, Kuo MH, Lee KL, Lee AC, Yu CC, Chang JL, Huang WC, Hsiao SH, Lin SE et al: Expression of Neuroendocrine Factor VGF in Lung Cancer Cells Confers Resistance to EGFR Kinase Inhibitors and

Triggers Epithelial-to-Mesenchymal Transition. Cancer research 2017,
77(11):3013-3026.
42. Tulchinsky E, Demidov O, Kriajevska M, Barlev NA, Imyanitov E: EMT: A mechanism for escape from EGFR-targeted therapy in lung cancer. Biochim Biophys Acta Rev Cancer 2019, 1871(1):29-39.
43. Yochum ZA, Cades J, Wang H, Chatterjee S, Simons BW, O’Brien JP, Khetarpal SK, Lemtiri-Chlieh G, Myers KV, Huang EH et al: Targeting the EMT transcription factor TWIST1 overcomes resistance to EGFR inhibitors in EGFR-mutant non-small-cell lung cancer. Oncogene 2019, 38(5):656-670.
44. Chang TH, Tsai MF, Su KY, Wu SG, Huang CP, Yu SL, Yu YL, Lan CC, Yang CH, Lin SB et al: Slug confers resistance to the epidermal growth factor receptor tyrosine kinase inhibitor. Am J Respir Crit Care Med 2011, 183(8):1071-1079.

Figure legend

Figure 1. NGF and TrkA analysis in head and neck squamous cell carcinoma (HNSCC).

(A) NGF and TrkA gene expression levels in the HNSCC samples according to The Cancer Genome Atlas (TCGA) database.
(B) NGF gene expression level in PNI-positive or PNI-negative tumor specimens according to the TCGA database.
(C) TrkA and NGF protein levels were determined in ten pairs of HNSCC tissues and matched adjacent normal tissues using western blotting (N, adjacent normal tissue; T, tumor tissue). GAPDH was used as a loading control.
(D) Immunohistochemical assays were performed to detect NGF and TrkA expression in HNSCC tissues and adjacent normal tissues. IHC staining indicated that, in perineural invasion cases, NGF was strongly stained in the nerves and the cancer cells, while TrkA was expressed in tumor cells around the nerve fibers (N, nerve fibers).
(E-F) Relative scores of TrkA and NGF staining in 50 pairs of human HNSCC tissues and matched adjacent normal tissue (p < 0.001, representatively).
(G) High TrkA expression levels were correlated with large tumor size, lymph node metastasis and advanced tumor stage in HNSCC.
(H) High NGF expression levels were correlated with large tumor size and positive PNI in HNSCC.
(I) Kaplan-Meier survival analysis of all HNSCC patients stratified by TrkA expression level (n = 115; p < 0.001).

(J) Kaplan-Meier survival analysis of all HNSCC patients stratified by NGF expression level (n = 115; p < 0.0225).
(K) Multivariate analysis of the hazard ratios (HR) showed that the upregulation of TrkA may be an independent prognostic factor for the overall survival rate. The HR is presented as the median (95% confidence interval, 95% CI).
Figure 2. TrkA is required for HNSCC cells to retain their proliferation and migration phenotype in vitro and in vivo.
(A-B) Cell viability of HN4 and HN6 cells after transfection with siTrkA1, siTrkA3 or siNC for 24 hours as determined using CCK-8 assays.
(C) Colony formation assay revealed the suppressive effect of TrkA knockdown in HN4 and HN6 cells.
(D-E) Cell migration and invasion abilities of HN4 and HN6 cells after TrkA knockdown were determined using Transwell assays.
(F) HN6 TrkA-knockdown cells (shTrkA3) or negative control cells (shNC) were subcutaneously injected into 2 lateral croups of BALB/c nude mice. The time course tumor growth curves were analyzed.
(G) Representative images of xenograft tumors isolated from the shNC and shTrkA3 groups after 23 days of treatment are shown. The weights of the tumors removed from mice in the shNC and shTrkA3 groups were measured.
(H) H&E staining and IHC assays for TrkA and Ki-67 staining in the shNC and shTrkA xenograft tumor tissues.

(*p< 0.05; **p < 0.01; *** p < 0.001 compared with the control group).

Figure 3. NGF triggers EMT-associated kinase switch and promotes HNSCC cell dissemination.
(A) The morphology of HN4 cells after treatment with the vehicle control or NGF (40 ng/ml) for 48 hours, as shown by bright field microscopy.
(B) The transcripts were differentially regulated in HN4 cells incubated with the vehicle control or NGF (40 ng/ml) for 24 hours, according to the RNA-Seq analysis. Results from the GO enrichment cluster analysis based on the genes significantly changed.
(C) Heat map showed the significantly upregulated and downregulated genes related to cell invasion and adhesion abilities.
(D) The relative mRNA levels of the indicated genes related to EMT (E-cad, N-cad, VIM, Twist1, Snail1 and Snail2) were measured by real-time PCR in HN4 and HN6 cells incubated with the vehicle control or NGF (40 ng/ml) for 24 hours.
(E-F) The indicated protein expression levels were analyzed by western blot in HN4 and HN6 cells after incubation with vehicle control or NGF (40 ng/ml) for 48 hours. The relative band intensity was assessed. Tubulin was used as a loading control.
(G) Expression of E-cadherin and Vimentin in HN4 and HN6 cells after incubation with the vehicle control or NGF (40 ng/ml) for 48 hours, as detected by immunofluorescence staining. (H-I) Cell migration and invasion abilities of HN4 and HN6 cells after treated with vehicle control or NGF (40 ng/ml) for 48 hours, as determined using transwell assays.
(*p < 0.05; **p < 0.01; ***p < 0.001; ns, no significant difference).

Figure 4. Targeting TrkA reverses the NGF-induced EMT in HNSCC cells.

(A) Expression levels of EMT markers (E-cad, Vimentin and Twist1) were analyzed by western blot (WB) in HN4 and HN6 cells pre-transfected with siRNAs for 48 hours, flowed by treatment with vehicle control or NGF (40 ng/mg) for 48 hours. Tubulin was used as a loading control.
(B) EMT marker expression levels were determined by WB analysis in HN4 and HN6 cells pretreated with vehicle control, TrkA inhibitor GNF5837 (2 µmol/L) for 1 hour, flowed by treatment with vehicle control or NGF (40 ng/mg) for 48 hours. Tubulin was used as a loading control.
(C) Expression of E-cadherin and Vimentin was determined by immunofluorescence staining.

HN4 and HN6 cells pretreated with the TrkA inhibitor GNF5837 (2 µmol/L) for 1 hour, followed by treatment with vehicle control or NGF (40 ng/ml) for 48 hours.
(D) Immunohistochemical analysis for E-cad, Vimentin, and Twist1 staining in the shNC and shTrkA xenograft tumor tissues.
(E) Relative TrkA mRNA and protein expression levels were evaluated for TrkA-OE or null vector transfection in CAL27 cells by WB.
(F) Expression levels of EMT markers were analyzed by WB in CAL27 cells pretransfected with TrkA-OE or null vector for 48 hours, flowed by treatment with vehicle control or NGF (40 ng/mg) for 48 hours. Tubulin was used as a loading control.
(G) Expression of E-cadherin and Vimentin in CAL 27 TrkA-OE or null vector cells was determined by immunofluorescence staining.

Figure 5. NGF/TrkA axis confers resistance to the EGFR inhibitor erlotinib in vitro and

in vivo.

(A) TrkA levels in a panel of cells from 6 HNSCC cell lines were significantly correlated with erlotinib IC50 values.
(B) The summary of the response to recombinant nerve growth factor family ligands at 40 ng/ml. A panel of cells from 6 HNSCC cell lines were exposed to serially diluted erlotinib in the presence or absence of NGF (40 ng/ml), EGF (40 ng/ml), BDNF (40 ng/ml), NRTN (40 ng/ml) or NT-3 (40 ng/ml) for 72 hours, and cell viability was measured via CCK-8 assay.
(C) TrkA-knockdown HN6 cells (siTrkA) and control cells (siNC) were treated with serially diluted erlotinib in the presence or absence of NGF (40 ng/ml). Cell viability was assessed using a CCK-8 assay after 72 hours of treatment.
(D) TrkA-overexpressed CAL27 cells (OE) and null vector cells were treated with serially diluted erlotinib in the presence or absence of NGF (40 ng/ml). Cell viability was assessed using a CCK-8 assay after 72 hours of treatment.
(E-F) GNF5837 increased the sensitivity of HNSCC cells to erlotinib. The HN4 and HN6 cells were treated with the indicated doses of erlotinib and GNF5837 for 72 hours, and cell viability was assessed using a CCK-8 assay.
(G) Tumor growth curve of subcutaneous HN6 xenografts in nude mice. The xenograft-bearing mice were treated with the vehicle, GNF5837 (25 mg/kg/day), erlotinib (50 mg/kg/day), or a combination (regorafenib and erlotinib) for 25 days, and the tumor size was calculated.

(H) Representative images of xenograft tumors isolated from the treatment groups. Tumor growth inhibition compared to the control group were evaluated.
(I) The weights of tumors removed from the mice were measured.

(J) Immunohistochemical analysis of Ki-67 staining and TUNEL assays of the xenograft tumor tissues.
(*p< 0.05; **p < 0.01; *** p < 0.001 compared with the control group)

Figure 6. STAT3 activation contributes to the NGF/TrkA axis-mediated erlotinib resistance of HNSCC cells.
(A) HN4 and HN6 cells were starved in growth factor-free medium overnight and stimulated with the vehicle control, NGF (40 ng/ml), EGF (40 ng/ml), or a combination for 30 min. Cell lysates were then harvested, and pY, and total EGFR and TrkA were analyzed by WB.
(B) HN4 and HN6 cells were starved in growth factor-free medium overnight and pretreated with erlotinib (2 µmol/L) for 1 hour, followed by stimulation with 40 ng/ml NGF, EGF or the vehicle control for 30 min, and then harvested for cell signaling analysis.
(C) HN4 and HN6 cells were starved in growth factor-free medium overnight and pretreated with erlotinib (2 µmol/L) for 1 hour, followed by stimulation with 40 ng/ml NGF, EGF or the vehicle control for 30 min. Cell lysates were then harvested, and pY and total EGFR, TrkA, AKT, ERK, and STAT3 were analyzed by WB.
(D) HN4 and HN6 cells were starved in growth factor-free medium overnight and pretreated with GNF5837 (2 µmol/L) for 1 hour, followed by stimulation with 40 ng/ml NGF or the vehicle control for 30 min, Cell lysates were then harvested, and pY and total TrkA and

STAT3 were analyzed by WB.

(E) STAT3 protein expression after siRNA transfection was evaluated by WB analysis.

(F-G) STAT3-knockdown HN4 and HN6 cells (siSTAT3) and control (siNC) cells were treated with serially diluted erlotinib in the presence or absence of NGF (40 ng/ml). Cell viability was assessed via CCK-8 assay after 72 hours.
(H-I) Inhibiting STAT3 reversed NGF-TrkA-induced erlotinib resistance in HNSCC cells. HN4 and HN6 cells were treated with the indicated doses of erlotinib and STAT3 inhibitor cryptotanshinone for 72 hours, and cell viability was assessed using a CCK-8 assay.
Figure 7. NGF/TrkA axis induces the EMT and promotes erlotinib resistance via STAT3 activation.
(A) TrkA-knockdown HN6 cells (siTrkA) and control cells (siNC) were stimulated with NGF (40 ng/ml) or vehicle control in the presence or absence of erlotinib (2 µmol/L) as indicated for 48 hours. Cell lysates were then harvested, and the E-cadherin, Vimentin, and Twist1 levels were analyzed by WB.
(B) TrkA-overexpressed CAL27 cells (OE) and null vector cells were stimulated with NGF (40 ng/ml) or vehicle control in the presence or absence of erlotinib (2 µmol/L) as indicated for 48 hours. Cell lysates were then harvested, and the E-cadherin, Vimentin, and Twist1 levels were analyzed by WB.
(C-D) Twist1-knockdown HN4 and HN6 cells (siTwist1) and control cells (siNC) were treated with the indicated doses of erlotinib in the presence or absence of NGF (40 ng/ml). Cell viability was assessed via CCK-8 assay after 72 hours.

(E-F) Slug-knockdown HN4 and HN6 cells (siSlug) and control cells (siNC) were treated with the indicated doses of erlotinib in the presence or absence of NGF (40 ng/ml). Cell viability was assessed via CCK-8 assay after 72 hours.
(G) STAT3-knockdown HN4 and HN6 cells (siSTAT3) and control cells (siNC) were treated with the vehicle control or NGF (40 ng/mg) for 48 hours. Cell lysates were then harvested, and the E-cadherin, Vimentin, and Twist1 levels were analyzed by WB.
(H) Proposed model of the paracrine effect of NGF/TrkA axis, which triggers the EMT and confers resistance to the EGFR kinase inhibitor erlotinib via STAT3 signaling in HNSCC cells.
Figure 8. EMT markers are coexpressed with TrkA and predict poor survival for HNSCC patients.
(A) E-cad, Vimentin and Twist1 protein levels were determined in ten pairs of HNSCC tissues and matched adjacent normal tissues using western blotting (N, adjacent normal tissue; T, tumor tissue). GAPDH was used as a loading control.
(B-C) Scatter plots generated from the relative band intensity displaying correlations between TrkA and Vimentin and between TrkA and TWIST1 (Spearman correlation analysis).
(D) Scatter plots generated from the relative band intensity displaying correlations between NGF and TWIST1 (Spearman correlation analysis).
(E) HNSCC tumors were stained for the expression of E-cad, Vimentin, and Twist1 in consecutive paraffin-embedded sections. Two examples are shown: high TrkA and NGF expression (case 1) and low TrkA and NGF expression (case 2).

(F-H) Kaplan-Meier survival analysis of all HNSCC patients stratified by E-cad, Vimentin, and Twist1 expression levels (n = 115; p =0.0048, p =0.0001, p =0.1893, respectively)
Highlights

1. Paracrine NGF/TrkA axis triggers EMT in HNSCC perineural niche.

2. NGF/TrkA axis confers resistance to the EGFR kinase inhibitor erlotinib via EMT and STAT3 signaling in HNSCC cells.
3. TrkA could be a novel therapeutic target for overcoming EGFR-targeted therapeutic

resistance in HNSCC patients.

Conflict of interest

The authors declare that they have no competing interests.