Epiregulin reprograms cancer-associated fibroblasts and facilitates oral squamous cell carcinoma invasion via JAK2-STAT3 pathway

OSCCs, and matching normal oral biopsies were collected following ethical approval (2017NL-013(KS)) and written informed consent. 11 matching CAFs, and NFs were isolated according to a method previously described [16] and forwarded for morphologic examination. After appropriate NFs/CAFs were isolated and chosen, 5 pairs of matched NFs and CAFs were sent for library construction, and RNA sequencing was performed.

We performed the RNA-seq with the help of Novel Bioinformatics Co., Ltd. (Shanghai, China). Firstly, total RNA was extracted by Trizol reagent (Invitrogen) separately. The RNA quality was checked by Bioanalyzer 2200 (Aligent) and kept at − 80 °C.The RNA with RIN > 8.0 is right for cDNA library construction. Secondly, the complementary DNA (cDNA) libraries for single-end sequencing were prepared using Ion Total RNA-Seq Kit v2.0 (Life Technologies). The cDNA libraries were then processed for the Proton Sequencing process according to the commercially available protocols. Thirdly, Mapping of single-end reads. The clean reads were then aligned to human genome (version: GRCh38.p1) using the MapSplice program (v2.2.0). Finally, pathway analysis was performed and we also applied EBseq algorithm to filter the differentially expressed genes, the significant analysis and FDR analysis were performed. Besides, we presented gene co-expression Networks to find the relations among different mRNA and LncRNA. Results of LncRNA have been reported in a previous publication [17]. Results of mRNAs with the most significant changes are listed in Additional file 1: Table S1. The data in the list are the mean value of the 5 pairs of samples.

For the analysis of expression of EREG, 104 consecutive unselected T_1-4N₀M₀ OSCC patients treated in the Department of Oral and Maxillofacial Surgery at Nanjing Stomatology Hospital from 1 January, 2009 to 31 December, 2010 were included into the study. Including criteria: (1) diagnosed with primary OSCC by hematoxylin and eosin staining, and staged as T_1-4N₀M₀ according to the 8th version of American Joint Committee on Cancer (AJCC) TNM staging system [18] by experienced pathologists from the Department of Pathology at Nanjing Stomatology Hospital; (2) underwent primary surgical treatments in the Department of Oral and Maxillofacial Surgery at Nanjing Stomatology Hospital; (4) paraffin-fixed sample available; (3) with complete clinical-pathological information and follow-up data. Excluding criteria: (1) diagnosed with autoimmune or other malignant diseases and pregnant; (2) had adjuvant treatment before or after surgery. The ethical approval for this study was obtained from the Research Ethics Committee of Nanjing Stomatology Hospital (approval number: 2017NL-013(KS)). The clinical pathologic characteristics of the 104 OSCC patients are listed in Table 1.

Immunofluorescence of cryosections or immunohistochemistry of paraffin-embedded tissue was done as previously described [19]. Primary antibodies were rabbit anti-human EREG (Abcam), phospho-JAK2 (Proteintech), phospho-STAT3 (CST), SMA (Abcam).

EREG expression in CAFs was identified on the basis of SMA staining. CAFs were defined as SMA positive, spindled-shaped cells in the stroma (data of SMA staining not shown). EREG staining was assessed only in fibroblasts in tumor stroma by the evaluation of staining intensity and the percentage of EREG positive fibroblasts according to criteria modified from Zhang J et al. (Fig. 2a). Briefly, EREG expression was scored by a combination of staining intensity (score: 0 = no staining, 1 = weak staining, 2 = medium staining, 3 = strong staining) and percentage of EREG positive cells (score 0 = 0%~ 5% positive cells, 1 = 6%~ 33% positive cells, 2 = 34%~ 66% positive cells and 3 = 67%~ 100% positive cells) by a method modified from Zhang J, et al. (Fig. 2a) [1]. EREG staining index was taken as the product of staining intensity and percentage. The staining index was further divided into low and high: score 0–4 (negative to medium) was defined as EREG low, score 6 and 9 (strong) was defined as high EREG expression. Two observers examined the images independently without the knowledge of patients’ information.

EREG over expression plasmid was purchased from Genechem Co.,Ltd. (Shanhai, China). The plasmid served as vector was pcDNA3.1+ (Amp). Amplified DNA fragments were inserted into the EcoR I/Xhol sites of pcDNA3.1+ vector. The amplification sequences were: EREG-F: GAATTCatgaccgcggggaggaggatggag; EREG-R: CTCGAGtcagacttgcggcaactctggatc. Empty plasmid was used as control.

The siRNAs were synthesized by Ribobio Co. Ltd. (Guangzhou, China) with following sequences: siNC: 5′-UUCUCCGAACGUGUCACGUdTdT-3′, siEREG-1: 5′-CCAGGAGAGUCCAGUGAUAdTdT-3′, siEREG-2: 5′-CCACCAACCUUUAAGCAAAdTdT-3′, siEREG-3: 5′-UACACUUUGUUAUUGACACUUdTdT-3′.

Plasmid-mediated overexpression of EREG in NFs, and siRNA-mediated interference of EREG in CAFs, were carried out using Lipofectamine® 2000 Transfection Reagent (Thermo Fisher Scientific) according to manufacturer’s instructions. Transient transfection was applied in all in vitro experiments.

Cells were lysed with TRIzol® reagent (Life Technologies), total RNA was extracted following the manufacturer’s instructions, and cDNA synthesis was conducted using High-Capacity cDNA Archive Kit system (Applied Biosystems). Quantitative PCR (q-PCR) was then conducted using iTaq™ Universal SYBR® Green Supermix (BIO-RAD). Comparative 2^-△Ct or 2^-△△Ct method was used to quantify the relative mRNA expression as indicated.

Western blots were performed using an SDS–PAGE electrophoresis system as described previously [19], employing rabbit anti-human EREG (Abcam), rabbit anti-human SMA (Abcam), rabbit anti-human E-cadherin (CST), rabbit anti-human N-cadherin (CST), rabbit anti-human vimentin (Proteintech), rabbit anti-human p-JAK2 (CST), rabbit anti-human JAK2 (CST), rabbit anti-human p-STAT3 (CST), mouse anti-human STAT3 (CST), rabbit anti-human p-NF-kB (CST), rabbit anti-human NF-kB (CST), rabbit anti-human p-p38 (CST), rabbit anti-human p38 (CST), rabbit anti-human p-akt (CST), rabbit anti-human akt (CST), rabbit anti-human p-erk1/2 (CST), rabbit anti-human erk1/2 (CST), rabbit anti-human snail (CST), antibodies. The blots were re-probed for rabbit anti-human GAPDH (Bioworld) to control for protein loading and transfer.

Immunofluorescence analyses were carried out as previously described [20]: primary antibodies, goat anti-human EREG (R&D), rabbit anti-human SMA (Abcam) and mouse anti-human CK (Abcam); secondary antibodies, AlexaFluor 647 donkey anti-goat IgG and AlexaFluor 488 goat anti-rabbit 594 goat anti-mouse IgG (Life Technologies).

NFs and CAFs were seeded into 12-well plates at identical density. NFs were transfected with control plasmid (pcDNA3.1) or EREG-overexpression plasmid (pcDNA3.1 EREG), and CAFs were transfected with control siRNA (siNC) or EREG siRNA (siEREG). The culture medium were collected 48 h after transfection, centrifuged at 3000RPM for 10 min to remove suspended cells, and stored in − 80 °C until Use.

Cells were transfected with over-expression plasmid or siRNA for EREG, followed with or without treatment of AG490 or rIL6 as indicated. After 48 h, the supernatants were collected for the measurement of IL6 with an ELISA kit (Dakewei, China) according to the manufacturer’s instructions.

Primary NFs and CAFs were transfected with pcDNA3.1-control, pcDNA3.1-EREG siNC or siEREG as indicated. The treated cells were either seeded in 96-well, or their culture medium was used to treat HSC3 cells seeded in 96-well. Cell viabilities were determined at 0, 24, 48, 72 and 96 h after transfection (in case of primary NFs/CAFs) or CM treatment (in case of HSC3 cells).

30/ml HSC3 cells were seeded in 12 well plates. Conditioned-medium from NFs + pcDNA3.1/pcDNA3.1-EREG, or CAFs+siNC/siEREG 24 h after transfection were used to treat HSC3 cells. The CM was changed every 3 days. After 1 to 2 weeks of cultivation, when visible colonies were formed, the medium was removed. Cells were 5% formalin-fixed after washed with PBS for 3 times, then stained with crystal violet.

HSC3 cells were seeded in 12-well plates until 100% fusion. After overnight starvation with serum-free DF12 medium, a scratch was made using a micropipette tip and cells were washed to remove detached cells and debris. Photographs of the same area of the wound were taken at 0 and 18 h for measuring the closure of the wound after the treatment of the indicated serum-free CM from NFs/CAFs.

The invasion and migration ability of HSC3 cells were tested in Transwell (boyden chamber) with an 8-um pore size of the polycarbonate membranes. Primary NFs and CAFs were placed in the lower chamber at a concentration of 3 × 10^4 cells/ml (0.6 ml/well). The NFs were then transfected with or without EREG-overexpression plasmids, and CAFs transfected with or without siEREG RNA. AG490 and recombinant human IL6 (Abcam) were added as indicated. After 5 h of incubation, the transfection mix was replaced by 0.5% FBS. Simultaneously, HSC3 cells, which had been previously starved overnight, were re-suspended with serum-free DF12 culture medium at a concentration of 10^6 cells/ml and then added to the upper chamber (200ul/well). For invasion assay, 40ul matrigel mix (matrigel: DMEM = 1: 2) was seeded at the bottom of Transwell chamber 1 h prior to the re-suspension of HSC3 cells to let the gel solidify. The number of migrated and invaded cells were qualified by blinded counting of migrated/invaded cells on the lower surface of the membrane, with five fields per chamber, at 24 h and 48 h, respectively.

3D culture invasion model was constructed and based on a method previously described by Igarashi et al. [21, 22]. Matrices containing fibroblasts were prepared by the addition of 1.4 parts of DF12 and 0.6 parts of FBS to 1 part of collagen type I solution (Thermo Fisher, 3 mg/ml), and neutralized with 1 M NaOH solution. Fibroblasts were then embedded into this solution to the final cell density of 5*10^5/ml, and seeded on top with HSC3 at the density of 4.5*10^5/cm² in triplicates. After 5 days of cultivation, 3D constructs were harvested, formalin-fixed, and paraffin-embedded, and sent for slide generation and HE staining.

To investigate the role of EREG in fibroblasts in vivo, stable EREG over-expression in NFs and EREG knockdown in CAFs were realized by lentivirus vector as previously decribed [17, 23]. Then 10⁶ conditional NFs/CAFs were subcutaneously co-injected with 10⁶ HSC3 in rear flank of four-week old male BALB/c-nu/nu T cells-deficient mice (Cavens, Changzhou, China). Tumor volumes (mm³) were measured at various time points until termination of the experiments. Tumors were collected for further experiments. Sixteen mice were used to construct xenografts with HSC3 and fibroblasts (four in each group). At the end of the study, mice were sacrificed and xenografted tumors were harvested. The experiments were executed in compliance with institutional guidelines and regulations.

All images represent at least three independent experiments. The results are presented as the mean ± SEM of at least three independent experiments. Statistical analysis was performed with SPSS® v 19 (SPSS, Chicago, IL, USA). Survival of patients was plotted using the Kaplan–Meier method and analyzed using the log-rank test, with patients censored at last follow-up. Kruskal–Wallis and Mann– Whitney U- and t-tests were used to compare groups, as appropriate. P < 0.05 was considered statistically significant for all tests (*p < 0.05, **p < 0.01, **p < 0.001, ***p < 0.0001).

As an abnormal activated form of NFs, CAFs differ from NFs with distinct phenotypes and functions. Exploration of the regulatory molecules that drive the transition from NFs to CAFs is urgently needed. Therefore, we carried out genomic analysis to identify differentially expressed genes between matching NFs and CAFs obtained from 5 OSCC patients as previously reported [17]. A series of differentially expressed genes were identified, among which Ereg was one of the genes with the most significant differential expression (Fig. 1a, Additional file 1: Table S1. The data are the mean value of the 5 matched samples).

To validate the results of the genomic studies, we then examined Ereg expression at the gene level in another 5 matching NFs and CAFs by quantitative PCR and found that Ereg was significantly higher in CAFs than the matching NFs (Fig. 1b). Furthermore, immunoblotting assays and immunofluorescence results also showed that EREG was significantly higher in CAFs than in matched NFs at the protein level, indicating that EREG may be involved in the transition from NFs to CAFs (Fig. 1c). Furthermore, the EREG expression profiles in normal mucosa tissue and tumor tissue were assessed by immunofluorescence staining in frozen sections. The results demonstrated that EREG expression was also significantly higher in fibroblasts from the cancer stroma (CAFs) than in fibroblasts from the stroma around normal mucosa (NFs; Fig. 1e). These results all proved that EREG expression was much higher in CAFs than in matched NFs.

It is well recognized that CAFs can promote the progression of tumors. However, the clinical outcomes of EREG expression in fibroblasts remain unclear. Considering the high level of EREG in CAFs, we first analyzed the correlation between EREG expression and the clinicopathological characteristics of patients with OSCC (n = 104). Chi-square analysis revealed that high EREG expression in CAFs was closely related to higher T staging (p = 0.0041), deeper depth of invasion (DOI, p < 0.0001) and inferior worst pattern of invasion (WPOI, p < 0.0001) (Table 1). Then, the prognostic value of EREG expression in fibroblasts was assessed. Kaplan–Meier analysis showed that high EREG expression in CAFs was significantly correlated with inferior overall survival (OS), and disease-free survival (DFS) (Fig. 2b, p = 0.0398, and p = 0.0314). The results demonstrated that high EREG expression in fibroblasts correlated with poor prognosis in OSCC patients.

In addition, we ascertained the role of EREG during the NF-CAF transition. First, we modified EREG expression in NFs and CAFs separately and evaluated changes in the cell phenotypes. High N-cadherin, vimentin, and SMA expression was used to mark the CAF phenotype. Three siRNAs (siEREG-1, siEREG-2, and siEREG-3) were designed, and their interference rates were 64.9, 58.3, and 48.9%, respectively. Then, siRNA-1, which had the highest interference rate, was chosen for subsequent experiments (hereafter referred to as siEREG; Additional file 2: Figure S1B). Plasmid overexpression of EREG in NFs was also successful (Additional file 2: Figure S1A). In addition to the confirmation of EREG expression at the gene level, we also confirmed the overexpression and interference of EREG by ELISA at the protein level (Additional file 2: Figure S1C). Overexpression of EREG in NFs significantly augmented the expression of N-cadherin, vimentin, and SMA, while siRNA modulation of EREG levels in CAFs resulted in reduced N-cadherin, vimentin, and SMA expression (Fig. 3a&b), indicating that EREG was involved in the transition from NFs to CAFs. However, the specific mechanism underlying EREG participation in the NF-CAF transition remains to be further elucidated.

To ascertain the pathway of the EREG-regulated NF-CAF transition, major pathways, including the JAK2-STAT3 pathway, NF-kB pathway, p38-MAPK pathway, PI3K/AKT pathway and erk1/2 pathway, were tested in unmodified NFs and EREG-overexpressing NFs. The results showed that both the levels of phospho-JAK2 and phospho-STAT3 and the ratios of phospho-JAK2/JAK2 and phospho-STAT3/STAT3 were significantly increased by the overexpression of EREG in NFs, indicating that the JAK2-STAT3 pathway was the most activated pathway after EREG overexpression (Fig. 3c-e). Although a close relationship between the erk1/2 pathway and EREG has been reported [13], our results demonstrated that erk1/2 did not play a significant role in the EREG-mediated NF-CAF transition (Fig. 3c-e).

To further confirm the role of the JAK2-STAT3 pathway in the EREG-mediated NF-CAF transition, we treated EREG-overexpressing NFs with AG490, an inhibitor of the JAK2-STAT3 pathway, and found that AG490 could reverse EREG-mediated CAF marker augmentation (Fig. 3f-h). Moreover, AG490 reduced EREG-mediated CAF activation (Additional file 3: Figure S2A). IL-6 is an important downstream factor of the JAK2-STAT3 pathway [24]. To further confirm the role of the JAK2-STAT3 pathway, we added recombinant human IL-6 (rIL-6) into the culture medium to construct a rescue assay. The results showed that while siEREG significantly lowered the expression of N-cadherin, vimentin and SMA, rIL-6 upregulated their expression (Fig. 3f-h). IL-6 levels were also assessed by ELISA. The results demonstrated that the IL-6 level was increased after EREG overexpression and was downregulated after EREG interference (Fig. 3i). In clinical OSCC samples, we also found that high EREG expression in CAFs was correlated with higher phospho-JAK2 and phospho-STAT3 expression (Fig. 3j-k). These results all demonstrate that EREG activates NFs in a JAK2-STAT3-dependent way. In addition to the change in protein markers, in OSCC, it has been reported that CAFs promote tumor migration [25] and invasion [7, 26] and are related to poor prognosis [27]. However, the effects of the EREG-mediated NF-CAF transition during OSCC progression are still unknown.

Then we investigated whether EREG expression in fibroblasts affected their tumor supportive functions. Ereg-expression level was assessed in 4 human OSCC cell lines and one normal epithelium cell line HACAT by qPCR. All 4 OSCC cell lines showed higher Ereg expression level compared with HACAT, and HSC3 was chosen for further experiments as it expressed the lowest level of Ereg among OSCC cell lines (Additional file 2: Figure S1D). Then, CM obtained from NFs/CAFs were added in to HSC3, and we found that compared with CM obtained from NFs, CM obtained from CAFs significantly improved the cell viabilities and promoted proliferation of HSC3 (Fig. 4a&b). However, modulation of EREG-expression in NFs/CAFs (transfected with plasmid or siRNA as indicated) did not significantly affect their supporting role of HSC3 viabilities and proliferation (Fig. 4a&b).

In our retrospective study of immunohistochemistry EREG staining in CAFs of OSCC, we already found that high EREG expression in CAFs of OSCC was correlated to a deeper invasion depth and inferior WPOI (Table 1). To confirm these results in vitro and to study the EREG-mediated NF activation in terms of their invasion-supportive abilities, in transwell migration assays, we found that CAFs with EREG interfered had an attenuated effect in inducing HSC3 migration, while up-regulation of EREG in NFs augmented their migration inducing ability (Fig. 4c&e). Moreover, the inhibition of JAK2-STAT3 pathway by AG490 could attenuate the migration-supportive role of EREG-high NFs, while rIL6 could rescue the migration-supportive role of EREG-low CAFs. The invasion abilities of HSC3 were assessed by transwell invasion assays. EREG-overexpressed NF increased HSC3 cell invasion more than two-folds than NF-control, while AG490 inhibited this invasion-supportive ability. Invasion-supportive ability was attenuated in CAF-siEREG, but could be rescued by rIL6 (Fig. 4c&f). Again, AG490 treatment attenuated the EREG-mediated migration and invasion-supportive abilities of CAFs (Additional file 3: Figure S2B).

To further confirm the invasion-inducing ability of EREG, we constructed a 3D invasion model with conditional NFs/CAFs, and found that augmented EREG expression in NFs led to deeper invasion. CAFs supported HSC3 invasion most significantly, with deeper invasion and more diffused spreading. This ability was attenuated after EREG interference, but could be restored by rIL6 (Fig. 4d&g). Meanwhile, overexpression of EREG in NFs significantly increased their invasion-supportive ability, but could be interfered after inhibition of JAK2-STAT3 pathway through AG490 (Fig. 4d&g).

To further reveal the mechanism underlying the EREG-mediated invasion-supportive role of fibroblasts, we analyzed the epithelial-mesenchymal transition (EMT) markers of HSC3 cells after treatment with CM from conditional NFs/CAFs. The results showed that compared with NFs, CAFs could significantly promote EMT in HSC3 cells in an EREG- and JAK2-STAT3-dependent way (Fig. 4h&i).

Together, these results indicate that the NF-CAF transition, which is induced by EREG, affects the migration and invasion of OSCC cells through EMT in a JAK2-STAT3-dependent way.

To further examine the tumorigenic role of EREG in vivo, we established a xenograft model in which 1*10⁶ HSC3 cells were subcutaneously coinjected with 10⁶ conditioned fibroblasts into the flank of nude mice, which were allowed to develop measurable tumors. One of the most important functional characteristics of CAFs is their tumor-supportive nature in vivo. Indeed, HSC3 cells coinjected with CAFs demonstrated rapid in vivo tumor growth, but this tumor-supportive effect could be attenuated by EREG knockdown (Fig. 5a-c). On the other hand, NFs demonstrated an in vivo tumor-supportive role after EREG overexpression (Fig. 5a-c). EREG knockdown and EREG overexpression were confirmed by qPCR and IHC (Fig. 5d & h). The EREG-mediated NF-CAF transition was confirmed in vivo, with IHC showing a positive correlation between EREG expression and SMA expression in fibroblasts (Fig. 5h). Moreover, EREG modulated the NF-CAF transition through JAK2-STAT3 pathway activation (Fig. 5e-h). In addition, WB confirmed that EREG regulated the EMT-promoting abilities of fibroblasts in vivo (Fig. 5i&j). Thus, these data provide novel insights into the involvement of EREG in the EREG-JAK2-STAT3-IL-6 axis in fibroblasts, which supports OSCC invasion through induction of EMT in the tumor microenvironment.