TERT drives broblasts reprogramming to promote the invasion of head and neck squamous cell carcinoma via transferring of exosomal MET protein

Head and neck squamous cell carcinoma (HNSCC) is characterized by the highly inltrative capacity and invariably aggressive feature, the oncogenic functions of TERT in cell transformation and tumor progression have well been elucidated; however, very little is known about the effect of TERT on the microenvironment during the tumorigenesis of HNSCC. An bioinformatics analysis of TERT was rstly performed from TCGA dataset, and the correlation of TERT activity with tumor microenvironment (TME) was assessed on the basis of the analysis of HNSCC patient specimen in vivo and in vitro co-culture experiments. We next conducted an cDNA microarray analysis and identied the TERT-driven key molecule MET in the functionally regulation of tumor-broblasts crosstalk. Furthermore, the phenotypic and functional alteration of primary broblasts were detected by a series of molecular biology experiments following treatment with TERT overexpressed/silenced HNSCC cell-derived exosomes or MET transfection, respectively, and the underlying intracellular signaling pathway was tested. Finally, the effect of programming broblasts on the aggressive behavior of HNSCC cells were evaluated in vivo and in vitro assays. level increased along


Abstract Background
Head and neck squamous cell carcinoma (HNSCC) is characterized by the highly in ltrative capacity and invariably aggressive feature, the oncogenic functions of TERT in cell transformation and tumor progression have well been elucidated; however, very little is known about the effect of TERT on the microenvironment during the tumorigenesis of HNSCC.

Methods
An bioinformatics analysis of TERT was rstly performed from TCGA dataset, and the correlation of TERT activity with tumor microenvironment (TME) was assessed on the basis of the analysis of HNSCC patient specimen in vivo and in vitro co-culture experiments. We next conducted an cDNA microarray analysis and identi ed the TERT-driven key molecule MET in the functionally regulation of tumorbroblasts crosstalk. Furthermore, the phenotypic and functional alteration of primary broblasts were detected by a series of molecular biology experiments following treatment with TERT overexpressed/silenced HNSCC cell-derived exosomes or MET transfection, respectively, and the underlying intracellular signaling pathway was tested. Finally, the effect of programming broblasts on the aggressive behavior of HNSCC cells were evaluated in vivo and in vitro assays.

Results
Here, we found that TERT was signi cantly elevated in HNSCC tissues in comparison with normal mucosa tissues, and its level was increased along with malignant progression of the tumor. Furthermore, the activated TERT enhanced cancer-broblasts interaction and promoted cancer-associated broblasts (CAFs) formation in TME. Mechanistically, TERT-driven normal broblast reprogramming into CAFs is exosomal MET dependent, activated TERT could physically interact with MET protein and increase exosomal MET protein released from HNSCC cells, which promotes the intercellular tra cking of MET into broblasts and consequently activated AKT and ERK signaling pathways, resulting in primary broblast reprogramming through increased cell proliferation, migration and elevation of the proin ammatory gene signature. More importantly, the programmed broblasts in turn promote the migration and invasion of HNSCC cells.

Conclusions
Our results unveil a novel role of TERT-driven modulation of the TME, which may offer new opportunities for potential therapeutic strategies targeting HNSCC development.

Page 3/31
Background Head and neck squamous cell carcinoma (HNSCC), which arises from the oral cavity, larynx and pharynx, ranks as the sixth most common malignancy with almost 835000 new diagnoses and 431000 deaths per year (1). The poor prognosis is closely associated with late detection, aggressive tumor feature, and poor response to available therapies, leading to the 5-year relative survival rate is still less than 65% now (2).
Telomerase reverse transcriptase (TERT) is a commonly mutate oncogene in human cancer, especially in melanoma (3), glioblastomas (4) and HNSCC (5). High level of TERT expression has been reported in most human cancer cells, while silencing in somatic cells, activation of TERT is a critical step for transformed cells to trigger in nite proliferation during carcinogenesis (6). Moreover, TERT possesses vital functions on proliferation (7), stemness (8), epithelial-mesenchymal transformation (EMT) (9) of cancer cells independent of the reverse transcriptase activity, which might contribute to the alteration of tumor microenvironment (TME) in the initial steps of tumorigenesis and cancer progression. Therefore, it is of great signi cance to understand whether the activated TERT affect the cross-talk between HNSCC and their TME.
In current study, on the basis of HNSCC patient specimen analysis and co-culture experiments, we classi ed that the activated TERT enhanced cancer-broblasts interaction and promoted cancerassociated broblasts (CAFs) formation in the microenvironment. Moreover, we showed that TERT increases exosomal MET protein released from HNSCC cells, which was transferred into broblasts and consequently activated AKT and ERK signaling, resulting in primary broblast reprogramming through increased cell proliferation, migration and elevation of the pro-in ammatory gene signature, in turn promotes migration and invasion of HNSCC cells.

Bioinformatics analysis of microarray data
Expression of targeted gene TERT and clinical data were downloaded from TCGA dataset. Data were normalized using the Bioconductor R package. The differential expression of TERT between HNSCC tissues and normal controls was compared by the empirical Bayes approach in linear models.
Additionally, gene expression pro les (GSE23558) was also obtained from Gene Expression Omnibus (GEO), including 5 normal and 27 HNSCC tissues, the Pearson r correlation was taken for correlation analysis of mRNAs expression. |logFC| >1 and p < 0.05 were selected as the cut-off criterion. of pCMV-TERT (Gene Copoeia) and pSIH-shTERT (5-GACGGTGTGCACCAACATCTA-3, System Bioscience) have been successfully constructed as previously described (10). As a comparison of the low TERT level in Cal27 cells, a high level of TERT was expressed in SCC25 cells, stably clones of pCMV-TERT Cal27 and shTERT SCC25 were generated on the basis of the endogenous cellular TERT levels, and was subjected to the following experiments, respectively. The human primary normal broblasts (NFs) or CAFs were isolated from fresh normal gingival or HNSCC tumor tissues with 2.5% collagenase (Sigma, USA). The cells were grown in DMEM/F12 supplemented with 10% FBS and 1% penicillin-streptomycin for less than 10 passages to ensure biologic similarity to the original specimen. All cell lines were cultured in a humidi ed incubator containing 5% CO 2 at 37 °C.
Co-culture experiments were performed in the 6-well transwell apparatus with 0.4 µm pore size (Corning Incorporated, USA), broblasts or HNSCC cells (1 × 10 5 ) were seeding in the lower or upper chamber sideby-side respectively. The cells were subjected to further analysis once they grew to 90% of con uence after 3-5 days of co-culture.

Lentivirus and RNA interference transfection
Lentivirus vectors containing full-lengh MET was constructed using pCDH-CMV-MCS-EF1-Puro (System Bioscience). Lentivirus was packaged in human embryonic kidney 293T cells and collected from the supernatant. The stably transfected NFs were selected with puromycin (1 µg/ml) for 5 days.
CAFs were transfected with small interfering RNA (siRNAs) targeting MET using Lipofectamine 3000 (Invitrogen, Carlsbad, CA) for 6 h. At 48 h after the transfection, the cells were used in further experiments.
The sequences of scramble and MET-speci c RNAi referred above were conducted by GenePharma (Shanghai, China) and listed in Supplementary Table 2. cDNA microarray analysis To compare relative gene expression pro les, total RNA from control and TERT-depleted SCC25 cells were extracted and puri ed according to the manufacturer's instructions. mRNA microarray analysis was performed at Capitalbio Corporation (Beijing, China) using the Human Genome U133 Plus 2.0 (Affymetrix). Brie y, biotinylated RNAs were prepared from 100 ng quantities of total RNA using the MessageAmp™ Premier RNA Ampli cation Kit (Ambion, USA). Following fragmentation, cRNA was hybridized in a Hybridization Oven 640 (Affymetrix) at 45 °C for 16 h, arrays were scanned using the GeneChip® Scanner 3000. Expression data were normalized through quantile normalization, and the Robust Multichip Average (RMA) algorithm, the differential expression of mRNAs were calculated on the basis of mean signal values ratio.
Isolation and identi cation of exosomes Exosomes were collected from HNSCC cell medium by differential ultracentrifugation according to the standard methods previous. Brie y, the conditioned medium was centrifuged at 300 g and 3000 g for 15 min to remove cells and other debris, the supernatant was centrifuged at 10000 g for 30 min at 4 °C to remove shedding vesicles. Finally, the exosomes were puri ed by centrifugation at 110000 g for 70 min, pelleted exosomes were collected and resuspended in PBS or other lysis buffers for validation or subsequent experiments. Exosomes were observed by Philips CM120 transmission electron microscope (FEI Company, USA). Size distribution and quanti cation of exosomes were analyzed with laser scatting microscopy (Particle Metrix, Germany).

Western blotting
Cells or exosomes were lysed wth RIPA lysis buffer and centrifuged at 12000 rpm for 15 min. Bicinchoninic acid (BCA) assay was performed to measure the protein concentrations. Proteins were separated by SDS-PAGE gel and transferred onto PVDF membranes. After blocked in 5% non-fat milk for 1 h, membranes were incubated overnight at 4 °C with various primary antibodies against α-SMA, CD63, CD9, MET, TERT, AKT, p-AKT, ERT, p-ERK, E-cadherin, N-cadherin, Slug and GAPDH, followed by incubation with horseradish peroxidase-linked secondary antibodies at 1:10000 (KangChen, China) for 1 h at room temperature, detection was performed using Chemiluminescent HRP Substrate (Millipore, USA), and signals were captured and observed using an Amersham Imager 600 (GE, USA).

Co-Immunoprecipitation assay
Total cell lysates were prepared in ice-cold IP lysis buffer (Beyotime, China), and the supernatant of cell lysates was collected by centrifugation at 13000 g for 15 min and transferred to new tubes for protein concentration measurement and immunoprecipitation. The protein concentration of the lysates was measured by BCA method and equal amounts of protein were used for immunoprecipitation. For immunoprecipitation, antibody against TERT was added to the lysates for incubation overnight at 4 °C, with rabbit IgG (1:100, Millipore, USA) as control antibody, followed by incubation with protein G Plus/

Wound-healing assay
Equal numbers of pre-treated cells were plated into six-well plates. Then the con uent cell monolayers were scratched with a pipette tip to draw a gap on the plates. The ability of cells to migrate into the cleared section was monitored under microscopy at the speci c time points.

Matrigel invasion assays
The effect of pre-treated broblasts on the invasion of HNSCC cells was determined by using matrigelcoated transwell inserts with 8 µm pores (Corning Incorporated, USA). Approximately, cancer cells were suspended in 200 µl (5 × 10 4 cells) of fresh medium containing 1% FBS and plated into upper chamber of 24-well plates, whereas equal number of pre-treated broblasts were seeded into the bottom chamber with 800 µl of culture media supplemented with 20% FBS. After 48 h, cells that across pores were xed with paraformaldehyde, stained with 1% crystal violet. For each chamber, three elds were randomly chosen and the invaded cells were counted.

Immuno uorescence
Cells were seeded onto the glass slides for 24 h, xed with 4% paraformaldehyde for 20 min, and permeabilized with 0.3% Triton X-100 for 10 min. After blocking with BSA, cells were incubated with primary antibodies speci c for α-SMA, Vimentin, MET and TERT overnight at 4 °C. After washing, the cells were incubated with secondary antibody for 1 h at room temperature. The secondary antibody was Cy3labeled Goat Anti-Rabbit IgG (1:500, Beyotime, China) and FITC-labeled Goat Anti-Mouse IgG (1:500, Beyotime, China). DAPI (Beyotime, China) was then used for counterstaining the nuclei. Immuno uorescence was detected by uorescence microscopy (Olympus, Japan).

Immunohistochemical analysis
Para n-embedded samples were sectioned at 4 µm thickness. The immunohistochemical procedure was as previously described (11). Antigen retrieval was performed by submerging the sections in 0.01M citrate buffer (pH 6.0) for 20 min to remove aldehyde links formed during initial xation of tissues. Specimens were incubated with antibodies speci c for α-SMA, MET, TERT, p-AKT and p-ERT overnight at 4 °C and the immunodetection was performed on the following day using GTvision Kit (GeneTech, China) according to the manufacturer's instructions. The intensity, percentage, and subcellular localization of the staining of each case were evaluated by two experienced pathologists on high power (200×) microscopic elds (HPF) in "hot-spots" (areas with high cellular density). The staining intensity was categorized: no staining as 0, weak as 1, moderate as 2 and strong as 3. We calculated the score of each sample by multiplying the staining intensity with the percentage of cells stained. To determine the number of broblasts, the elds of view (200×) per tissue section were counted.

In vivo xenograft study
All experimental procedures were approved by the Institutional Animal Care and Use Committee of Shandong Provincial hospital. To verify the correlation of TERT with TME, the xenograft tumors previously established by subcutaneously injection of control-Cal27 or TERT-Cal27 cells (1 × 10 7 ) were used to examine for histologic and immunohistochemical evaluation.
To further assess the effect of programmed broblasts on tumorigenicity in vivo, the other HNSCC xenograft assay was used. Brie y, six weeks old female BALB/c nude mice purchased from the Center of Experimental Animal of Vital River Laboratories (Beijing, China) were randomly divided into 6 mice per group, 1 × 10 7 Cal27 cells alone or mixed with control or MET-overexpressed NFs (1 × 10 6 ) were injected subcutaneously into the right anks of nude mice respectively, and tumor sizes were measured using a Vernier calliper every 2 days when the tumors were readily visualized. The tumor volume was calculated according to the following formula: volume = 0.5 × length × width 2 .

Statistical analysis
Data analysis was performed using the SPSS 16.0 software package. Each experiment were carried out in triplicate at least and all results were presented as mean ± SEM. Quantitative data were compared by either one-way analysis of variance (ANOVA) (multiple groups) or unpaired student's t-test (two groups). The correlation between TERT and MET was determined by Pearson analysis. A p-value < 0.05 (two-sided) was considered statistically signi cant.

Results
Ectopic expression of TERT in HNSCC is associated with the accumulation of broblasts.
To evaluate the impact of TERT in the development of HNSCC, we analyzed its expression from TCGA, the mRNA level of TERT was signi cantly elevated in HNSCC tissues in comparison with normal mucosa tissues, and TERT expression increased along with malignant progression of the tumor, implying TERT is correlated with HNSCC oncogenesis (Fig. 1a-b). As previous reported the signi cance of CAFs in angiogenesis, lymphangiogenesis and invasion of HNSCC (12), we isolated and identi ed a more elongated, mesenchymal morphology and higher levels of the broblast-speci c marker α-SMA in primary CAFs than NFs (Fig. 1c). To investigate the association of TERT activity with TME over the course of the progression from precursor lesions to high-grade lesions, we initially employed clinical model to examine the level of TERT and number of activated broblasts. As we have previously described that the TERT expression in epithelium was upregulated from NOM to carcinoma in situ and HNSCC (9). A signi cant difference of TERT expression was in the cytoplasm rather than nucleus among three groups.
Concurrently, an elevated number of broblasts were detected in the process of carcinogenesis, broblasts were scarce in the stroma except for some vessels in NOM. In the pre-invasive lesion of carcinoma in situ, a signi cantly increased number of activated broblasts accumulated in the stroma. As the tumor progresses and invades the dermis, more broblasts lled the stroma (Fig. 1d).
Next, we further assess the clinical relevance of TERT on the distribution of CAFs in HNSCC. Remarkably, a gradual increase of TERT expression was in the tumor nests from well, moderately to poorlydifferentiated stage as previously mentioned (9). Meanwhile, an elevated density of CAF was also observed in the stroma (Fig. 1e). Additionally, HNSCC patients with local recurrences after chemoradiotherapy also had a higher TERT expression than those with primary surgeon. Importantly, CAF was signi cantly increased in recurrent dermal region (Fig. S1). These observations suggest that TERT might be implicated in the interplay between tumor cells and broblasts, and is positively associated with the accumulation of broblasts during the initiation and progression of HNSCC.
TERT drives the reprogramming of primary broblast into activated CAFs.
To verify whether TERT affect the biological features of broblasts, we co-cultured the primary NFs with cancer cells. Interestingly, as compared to the broblasts cultured alone, the broblasts co-cultured with Cal27 cells, especially overexpressing TERT, showed an elongated phenotypic change (Fig. 2a). In cocultured conditions, the proliferation of broblasts was signi cantly increased after co-culture with TERToverexpressed Cal27 cells compare with those co-cultured with control Cal27 cells (Fig. 2b). Woundhealing assays further con rmed that overexpression of TERT in Cal27 cells could remarkably improve the broblasts migration ability (Fig. 2c). More importantly, broblasts educated by TERT-overexpressed Cal27 cells resulted in a clear upregulation in the expression of pro-in ammatory genes IL-1β, IL-6 and IL-8 (Fig. 2d). In contrast, the elongated phenotype of broblasts was abolished followed by co-culture with TERT-depleted SCC25 cells, a decrease in broblasts proliferation and migration was also validated experimentally, as well as the downregulation of pro-in ammatory genes, compared with those co-cultured with control SCC25 cells (Fig. 2e-h). These data suggest that TERT reprograms primary broblast into activated CAFs.
To further investigate the mechanism of TERT on the broblasts programming, we co-cultured broblasts with HNSCC cells and examined the expression of TERT and α-SMA in broblasts. As is shown in Fig. 2i-j, the levels of TERT and α-SMA were both increased in broblasts after co-culture with HNSCC cells compared with those cultured alone. Notably, α-SMA expression was much increased in the broblasts upon co-culture with TERT-overexpressed Cal27 cells than those co-cultured with control Cal27 cells, whereas the level of TERT exhibited no statistical difference. Conversely, broblasts co-cultured with TERT-depleted SCC25 cells exhibited slightly decrease of α-SMA expression in comparison to those cocultured with control SCC25 cells, while TERT level was not altered. These results indicate that enhanced TERT activity in HNSCC cells acts the promoting effects on the reprogramming of broblasts via an indirect manner.
MET is the downstream target of TERT and upregulated in the activated CAFs.
To identify the potential key molecular effectors for these functionally regulation, we performed microarray analysis to compare mRNA pro les from control and TERT-depleted SCC25 cells. As is shown in Fig. 3a, the gene oncology (GO) function analysis of differentially expressed genes (DEGs) indicated that TETR was mainly involved in in ammatory response and extracellular stimulus associated with the alteration of TME. KEGG pathways enrichment further revealed that the inhibition of pathways in cancer and cytokine-cytokine receptor interaction (Fig. 3b). Among these DEGs, ve major downregulated genes (IL-8, TIMM8A, IL1β, MET and CCL20) involved in in ammatory or cancer processes were evaluated by qRT-PCR, the results showed that the mRNA levels of these selected genes were decreased in TERTdepleted SCC25 cells compared with controls, con rming the validation of microarrays (Fig. 3c-d); however, we next tested that only TIMM8A and MET expression were upregulated in TERT-overexpressed Cal27 cells (Fig. 3e). Furthermore, we found that the changed level of MET was more than 2 times than TIMM8A in response to the overexpression or depletion of TERT in HNSCC cells. Thus, the role of MET induced by TERT seems worthy of further investigating.
We analyzed public gene expression pro les GSE23558 to nd out the relationship between TERT and MET expression using the Pearson r correlation test. The results showed that the upregulated TERT and MET in HNSCC tissues had a signi cant positive correlation (r = 0.4826, p = 0.0051) (Fig. 3f). More importantly, immuno uorescent staining indicated that the overexpression of TERT signi cantly enhanced the level of MET protein (Fig. 3g), we also con rmed that the expression of MET protein was promoted in the TERT-overexpressed Cal27 cells by immunoblotting, while it was reduced upon TERT knockdown in SCC25 cells (Fig. 3h-i).
Next, to examine the mechanism whereby TERT regulated the activity of secreted protein MET, we detected whether exists the physical interaction of TERT and MET in HNSCC cells. As shown in Fig. 3j, immuno uorescence imaging showed co-localization of TERT and MET in the cytoplasm and nucleus of Cal27 or SCC25 cells. In addition, co-immunoprecipitation (Co-IP) further revealed that endogenous TERT was able to physically interact with MET. Therefore, MET is a direct downstream target of TERT (Fig. 3k).
To determine the role of MET mediated by TERT on the broblasts programming, we test the protein level of MET in primary NFs co-cultured with HNSCC cells. As is shown in Fig. 3l, the broblasts had elevated expression of MET in co-cultured condition compared with those cultured alone, and the level of MET expression much increased in broblasts co-cultured with TERT-overexpressed Cal27 cells than those cocultured with control Cal27 cells. In contrast, co-culture of broblasts with TERT-depleted SCC25 cells could abrogate the level of MET, as compared with those co-cultured with control SCC25 cells (Fig. 3m).
Previously, we reported that TERT enhanced HNSCC growth in vivo xenograft model (10). Consistent with the above nding, our immunochemical staining results also exhibited that a signi cantly elevated number of activated broblasts presented at the interface with tumor epithelial islands overexpressing TERT compared with controls. More importantly, increased TERT promotes a higher expression level of MET in both tumor and stromal broblasts (Fig. 3n). Therefore, given that TERT regulates the reprogramming of broblasts, we hypothesized that intercellular transfer of MET might strongly contribute to this effect.
Exosomes are cell-derived vesicles that serve as mediators of intercellular communication (13). The transfer of HNSCC-derived exosomes to broblasts is still poorly characterized. To investigate whether exosomes-derived HNSCC cells participated in the activation of broblasts, we rst managed to isolate and identify exosomes from the conditioned media of HNSCC cells, the cup-shaped structure ranging in diameter from 70 to 160 nm of vesicles were con rmed using transmission electron microscopy as well as Nanosight particle tracking analysis, indicative of exosomes (Fig. 4a). We next examined whether these exosomes could be internalized by broblasts, exosomes were labeled with DiI (red). After incubation, immuno uorescence imaging showed the presence of red spots was increased in recipient broblasts in a time-dependent manner, suggesting that these labeled exosomes release by HNSCC cells could be direct uptake by broblasts (Fig. 4b). Moreover, the differential levels of MET regulated by TERT in HNSCC cell-derived exosomes were checked using immunoblotting with the exosome speci c markers CD9 and CD63. As is shown in Fig. 4c, MET was detected in the exosomes of HNSCC cells, and the level of exosomal MET was strongly increased in TERT-overexpressed Cal27 cells compared with controls, whereas a signi cant decrease of MET level was observed in TERT-depleted SCC25 cell-derived exosomes, suggesting that TERT could enhance the enrichment of exosomal MET (Fig. 4d).
We further tested the functionally effect of tumor-derived exosomes on primary broblasts. Similar to the condition co-cultured with HNSCC cells, NFs treated with exosomes-derived from HNSCC cells exhibited CAF-liked features compared with untreated controls (Fig. 4e). More importantly, the broblasts treated with exosomes from TERT-overexpressed Cal27 cell exhibited a more markedly increased proliferation than those treated with the corresponding exosomes from control Cal27 cells (Fig. 4f). Also, TERToverexpressed Cal27 cell-derived exosomes improved the migration ability and elevated the expression of pro-in ammatory genes IL-1β, IL-6 and IL-8 in broblasts, as veri ed by the upregulated expression of α-SMA and MET detected by immunoblotting analysis (Fig. 4g-i). Conversely, TERT-depleted SCC25 cellderived exosomes decreased the effect on broblast migration and pro-in ammatory gene expression, as well as downregulated expression of α-SMA and MET (Fig. 4j-n).
To further analyze the mechanisms by which exosomal MET affects the programming of activated broblasts, we rst established stable overexpression of MET in primary NFs by lentivirus infection (Fig. 5a). The upregulation of MET increased the levels of α-SMA in broblasts, as compared with controls, leading to CAFs-liked phenotypic changes (Fig. 5b-c). As is shown in Fig. 5d, the upregulated MET dramatically accelerated the proliferation of NFs. Meanwhile, the activated broblasts with MET overexpression exhibited a promotion on motility and expression of pro-in ammatory genes IL-1β, IL-6, IL-8 ( Fig. 5e-f).
In parallel, the MET expression was knocked down in CAFs with siRNA and the effect was con rmed by immunoblotting (Fig. 5g). As expected, the silencing of MET decreased the levels of α-SMA in CAFs, leading to the inhibition of cell proliferation, signi cant abrogation of mobility ability and downregulation of these pro-in ammatory genes ( Fig. 5h-l). Collectively, these results indicate that HNSCC cell-derived exosomal MET mediated by TERT reprograms primary broblasts into CAFs.

Exosomal MET induces CAFs formation via activation of AKT and ERK signaling.
To unravel the signaling pathways activated in MET-overexpressed broblasts, we detected the downstream status of PI3K/AKT and MEK/ERK1/2 of MET, which were known to be associated with cell proliferation and migration (14). We found that an elevated levels of the phosphor-AKT and AKT in METoverexpressed broblasts, as compared with controls; similar results were obtained from the broblasts treated with exosomes from TERT-overexpressed Cal27 cells (Fig. 5m-n). In contrast, the activity of phosphor-AKT and ERT were suppressed in MET depletion of CAFs, this effect was also abrogated in broblasts treated with TERT-depleted SCC25 cell-derived exosomes (Fig. 5o-p).
To further evaluate the status of AKT and ERK signaling of broblasts in vivo, we employed a xenograft mouse model co-injected with Cal27 cells. Immunohistochemistry results showed a higher MET expression accompanied with an elevation in the phosphor-AKT and ERT levels was observed in the METoverexpressed broblasts (Fig. 5q). These results suggest that MET promoted broblasts programming by mediating its downstream AKT and ERK signaling activation.
MET activation of broblasts reversely promotes EMT of HNSCC cells.
CAFs are known to enhance invasive behavior and chemoresistance of cancer cells in the TME (15). To determine whether broblasts educated by MET contribute to the promotion of tumor characteristics, HNSCC cells were mono-or co-cultured with broblasts in vitro. Subsequently, the impact of activated broblasts on biological behavior of HNSCC cells was performed using wound healing and transwell assay, respectively. As is shown in Fig. 6a-b, Cal27 cells co-cultured with broblasts exhibited signi cantly enhanced in migration and invasion ability compared with those cultured alone. Speci cally, in cocultured conditions, the Cal27 cells co-cultured with broblasts overexpressing MET were more activated than those co-cultured with control broblasts. More importantly, changes in the expression of EMTpromoting signals were greater in Cal27 cells co-cultured with MET-overexpressed broblasts by immunoblotting, as characterized by downregulation of epithelial markers E-cadherin, upregulation of mesenchymal markers N-cadherin and transcripts for Slug (Fig. 6c). Similar results were found in SCC25 cells co-cultured with MET-overexpressed broblasts (Fig. S2a-c).
In contrast, we next test the functionally change of HNSCC cells under the co-cultured condition with METdepleted CAFs. As expected, a decrease in migration and invasion ability and reversion of EMT were observed in Cal27 and SCC25 cells after co-culture with MET depletion of CAFs, as compared with those co-cultured with control CAFs (Fig. 6d-f, Fig. S2d-f).
To verify the above ndings, the effect of MET activated broblast on HNSCC cells was tested in vivo. Interestingly, consistent with in vitro results, the results showed that co-injection of control or METexpressed broblasts enhanced the tumorigenicity upon transplantation of Cal27 cells, as compared with Cal27 cells alone (Fig. 6g-j). Importantly, in co-grafted mice, more aggressive HNSCC growth was observed when Cal27 cells were injected with MET-overexpressed broblasts. Furthermore, the tumors coinjected with MET-overexpressed broblasts exhibited a larger extensive stroma with signi cant higher expression of α-SMA, rather than those with control broblasts (Fig. 6k). Taken together, our data showed that exosomal MET induced by TERT is uptaken by broblasts, which activates AKT and ERK signaling and triggers NFs reprogramming into CAFs (Fig. 7).

Discussion
HNSCC is characterized by the highly in ltrative capacity and invariably aggressive feature (2). As a dynamic network orchestrated by intercellular communications, the TME in the role of HNSCC progression provokes growing interests. Starting from clinical evidence, our present study revealed that aberrant activity of TERT accompanied with increased accumulation of broblasts in the TME on the initial steps of tumorigenesis and cancer progression. Moreover, further studies illustrated that TERT act as a driver enhanced the cancer-broblasts crosstalk, contributing to cancer progression.
Although the function of TERT have been investigated comprehensively in cancer cells, its effect on TME have not previously been characterized. Increasing evidences demonstrated that TERT reactivation in several tumor types lies in gene mutations (16), chromosomal re-arrangements(17) and altered methylation pattern (18) of TERT promoter. In HNSCC, ectopic TERT expression was dominantly associated with the mutations of TERT promoter (-124 G > A or -146 G > A) (5,19). Here, we initially found that TERT mRNA was consistently upregulated in analyzed expressing pro ling datasets and positively correlated with malignant progression of the tumor. Similar to the loss of p53 driving neuron reprogramming (20), our immunohistochemical staining statistics also revealed that TERT was signi cantly associated with the number of activated broblasts in the primary and recurrent HNSCC samples. Furthermore, our previous in vivo experiments showed that aberrant activated TERT enhanced broblast accumulation in the stroma, further experiments demonstrated TERT could promote the broblasts reprogramming into CAFs in co-cultured conditions. Therefore, these ndings indicated a so far unknown effect of oncogenic TERT on the distribution of activated broblasts, which may account for why cancer cells with ectopic expression of TERT have more aggressive properties in TME.
Exosomes-mediated intercellular communication in the microenvironments is a key event of malignant transformation and progression in many cancers(21). Recent studies showed that oncogenic exosomes can deliver functional mRNAs(22), miRNAs(23) and proteins(24) to surrounding normal cells and alter the cellular environment for favoring tumor growth, contributing to the immune escape, drug resistance and angiogenesis of tumors. In our study, we identi ed MET was the direct downstream target of TERT in cancer cell by mRNA array analysis. Mechanically, TERT can physically interact and increase the expression of MET protein. It is known that MET is a transmembrane receptor tyrosine kinase and ectopic expressed in various cancer types including HNSCC, activation of MET pathway promotes self-renewal and tumorigenecity (14). Consistent with previous reports describing that the intracellular tra cking of MET oncoprotein via exosomes induces a more pro-malignant phenotype of recipient cells(25), our results showed that TERT-driven the increased level of exosomal MET could be transferred from cancer cell to broblasts, contributing to the formation of CAFs. As a contrast, the decrease of exosomal MET resulted from TERT-depletion in HNSCC cells impaired the transformation of educated broblasts. More convincing data can be provided by MET-overexpression of primary broblasts or silencing of MET in CAFs. These ndings indicated that aberrant activity of TERT enhances the communication of cancer cells and broblasts in TME and that targeting MET could inhibit the tumorigenic action of oncogenic TERT on broblasts. However, the effects of TERT in cancer cells on proliferation and pro-in ammatory genes expression of broblasts treated with exosomes were not as large as those of co-culture treatment, indicating that TERT might mediate additional factors to involve in CAF formation apart from exosomal MET.
As the predominant component of the tumor stroma, CAFs are known to be regulators of tumor invasion, metastasis and chemoresistance via activating the crosstalk with cancer cells (26, 27). The positive expression of α-SMA in stromal broblasts represents a suitable immunophenotype of CAFs(28). Indeed, our present study found the proportional number of broblasts in the stroma was positively associated with the progression of HNSCC tumor. Furthermore, consistent with previous reports in most solid human cancers, MET could regulate the cellular proliferation and survival in activated broblast through triggering the activation of AKT and ERK signaling pathways, promoting broblast reprogramming into CAFs. Based on the reciprocal interaction between the tumor and broblasts, our results also indicates that the cancer cells exhibited a more aggressive property in co-cultured with ectopic MET expression of broblasts in vivo and vitro. Therefore, the results from the current study support the hypothesis that the TERT-driven programming broblasts induce a more malignant phenotype of HNSCC cells.

Conclusions
In summary, our study uncovered a novel role of TERT in driving broblasts programming, this promoting effect was mainly attributed to the intercellular tra cking of MET from tumor cells to broblasts via exosomes and converted broblasts to CAFs by activating AKT and ERK signaling pathways. Although further studies will be required to elucidate the molecular events that link the education of broblasts to cancer progression, our study provide the possibility that preventing this transfer is likely a new strategy for the treatment of HNSCC. Ethics approval and consent to participate

Abbreviations
The present study was authorized by the Ethical Committee of the Shandong Provincial Hospital. All procedures performed in studies were in accordance with the ethical standards. Written informed consent was obtained from all patients prior to study.
All animal experimental procedures were approved by the Institutional Animal Care and Use Committee of Shandong Provincial hospital.

Consent for publication
Not applicable.

Con ict of interest
The authors declared that they have no con ict of interest.