Cancer-associated fibroblasts promote oral squamous cell carcinoma progression through LOX-mediated matrix stiffness

Cancer-associated fibroblasts (CAFs), the most abundant cells in the tumor microenvironment, have prominent roles in the development of solid tumors as stromal targets. However, the underlying mechanism of CAFs’ function in oral squamous cell carcinoma (OSCC) development remains unclear. Here, we investigated the role of lysyl oxidase (LOX) expression in CAFs in tumor stromal remodeling and the mechanism of its effect on OSCC progression. Multiple immunohistochemistry (IHC) staining was performed to detect the correlation of CAFs and LOX in the stroma of OSCC specimens, as well as the correlation with clinicopathological parameters and prognosis. The expression of LOX in CAFs were detected by RT-qPCR and western blot. The effects of LOX in CAFs on the biological characteristics of OSCC cell line were investigated using CCK-8, wound-healing and transwell assay. CAFs were co-cultured with type I collagen in vitro, and collagen contraction test, microstructure observation and rheometer were used to detect the effect of CAFs on remodeling collagen matrix. Then, collagen with different stiffness were established to investigate the effect of matrix stiffness on the progression of OSCC. Moreover, we used focal adhesion kinase (FAK) phosphorylation inhibitors to explored whether the increase in matrix stiffness promote the progression of OSCC through activating FAK phosphorylation pathway. LOX was colocalized with CAFs in the stroma of OSCC tissues, and its expression was significantly related to the degree of malignant differentiation and poor prognosis in OSCC. LOX was highly expressed in CAFs, and its knockdown impaired the proliferation, migration, invasion and EMT process of OSCC cells. The expression of LOX in CAFs can catalyze collagen crosslinking and increase matrix stiffness. Furthermore, CAFs-derived LOX-mediated increase in collagen stiffness induced morphological changes and promoted invasion and EMT process in OSCC cells by activating FAK phosphorylation pathway. Our findings suggest that CAFs highly express LOX in the stroma of OSCC and can remodel the matrix collagen microenvironment, and the increase in matrix stiffness mediated by CAFs-derived LOX promotes OSCC development through FAK phosphorylation pathway. Thus, LOX may be a potential target for the early diagnosis and therapeutic treatment of OSCC.


Background
Oral cancer is the sixth most common cancer in the world, and oral squamous cell carcinoma (OSCC) accounts for 90% of oral malignant neoplasms [1,2]. Although early diagnosis and treatment are developing and improving, the five-year survival rate of patients with OSCC remains low because of distant metastasis and recurrence, and the quality of life of patients has remained a serious problem [3,4]. Thus, further understanding of the complex pathogenesis of OSCC is needed to develop a more effective therapy. The tumor microenvironment (TME), a complex meshwork of extracellular matrix (ECM) macromolecules, plays an important role in tumor progression. As a main component of TME, cancer-associated fibroblasts (CAFs) are of great importance for ECM production and interstitial tissue remodeling in the disturbed homeostasis [5], which can induce stiffness and hypoxia in the matrix [6]. Matrix stiffness is one of the most important biomechanical characteristics of TME faced by cancer cells and stromal cells. Recently, there has been a consensus that mechanical stimulation in the TME, such as the alteration of matrix stiffness, can have an impact on the development of solid tumors by a variety of signaling pathways, and then promote invasion and metastasis [7]. Thus, the relationship between CAFs and matrix stiffness and the mechanism of CAFs remodeling ECM are especially worthy of further study.
Collagen is the most common ECM scaffold protein in the matrix [8], and its remodeling and cross-linking critically affect the tumor progression by enhancing ECM stiffness [9]. Previous studies reported that collagen is associated with the development of several malignancies such as breast, colon and liver cancer [7,10,11]. Increased collagen deposition and ECM density have been found in the tumor invasion frontier and have been shown to have a significant association with several important clinicopathological parameters [12]. Lysyl oxidase (LOX), a copper-dependent amine oxidase secreted by fibrogenic cells [13], can initiate covalent cross-linking of collagen and elastin in the ECM [14]. The elevation of LOX expression has been found in many malignant tumors [13,15,16] and correlates with enhanced cancer cell proliferation, metastasis, and angiogenesis in the TME [17,18]. LOX can be a molecular target for improving the efficacy of chemotherapies against different types of solid tumors [19], as it has been reported to promote tumor progression by regulating collagen cross-linking [17]. However, the relationship among CAFs, LOX, collagen and their mechanism in the remodeling matrix and influencing tumor progression in OSCC are still largely unknown.
A stiff microenvironment has an effect on many aspects of tumor cells, including internal molecular signaling pathways and cell behaviors [20]. Cellular interactions with the extracellular matrix, especially the state of cell adhesion, have changed in tumor initiation and progression. Focal adhesion kinase (FAK), a prominent signaling molecule in focal adhesions, is over-phosphorylated in multiple cancer progression and is responsible for transmitting extracellular mechanical force to changes in gene expression within cancer cells [14,21]. The potential of mechanical stimuli such as elevated collagen and tissue stiffness is responsible for the induction of β-catenin translocation and the expression of Wnt-associated proteins [22]. However, the molecular mechanisms driving tissue stiffness-related tumor progression have not been completely elucidated in OSCC.
In the present study, we sought to investigate the correlation between ECM remodeling mediated by CAFsderived LOX and OSCC development. The colocalization of CAFs and LOX in tumor stroma were explored in OSCC tissues. The expression of LOX in CAFs and its effect on biological behavior of OSCC cells were investigated in vitro. Then, we studied whether LOX expressed by CAFs can significantly catalyze the cross-linking of collagen and increase matrix stiffness, thereby regulating the morphology and invasion of tumor cells in a FAKdependent manner. Thus, our study suggests that matrix stiffness mediated by CAFs-derived LOX might present a novel therapeutic thought for OSCC patients.

Patients and specimens
Ninety six cases of primary OSCC tissue samples were recruited from 2010 to 2015 from patients treated with surgery at the Department of Oral and Maxillofacial Surgery, the Affiliated Stomatological Hospital of Nanjing Medical University: 59 were male (61.46%) and 37 were female (38.54%), with an average age of 60.4 years (range, 35-91 years). The primary tumors were located in the tongue (n = 32), gingiva (n = 21), buccal mucosa (n = 26), and other sites (n = 17), such as the palate, lower lip, jaw, soft palate, oropharynx, or floor of the mouth. All patients underwent extensive resection of the primary tumor and underwent classic radical neck dissection or Keywords: Cancer-associated fibroblasts, Epithelial-mesenchymal transition, Lysyl oxidase, Matrix stiffness, Oral squamous cell carcinoma

Evaluation of immunoreactivity
The staining intensity of tissue sections was individually evaluated by two pathologists who were blinded to the patients' clinical data. Immunoreactivity of LOX and α-SMA in tumor stroma was estimated according to the classification system described by Fuji et al. [23]: 0, negative (no staining); 1, sparse (a small population of discrete stained fibroblasts); 2, focal (irregular and intermittent areas of stained fibroblasts); and 3, abundant (widespread and consecutive areas of stained fibroblasts). The specimens were categorized into two subgroups according to immunoreactivity score: low expression (0-1) and high expression (2)(3).

Fibroblast isolation and preparation of conditioned media
Fresh oral cancer tissues and matched normal tissues were collected under sterile conditions. All samples were obtained with informed consent from the patients, and the fibroblast isolation protocol was approved by the Nanjing Medical University Ethics Committee. All tissue samples were snap-frozen in liquid nitrogen and stored at − 80 °C until fibroblast isolation. Fresh tissues were washed with PBS (penicillin 100 U/mL, streptomycin 100 μg/mL) three times and cut into approximately 1 mm 3 size. The tissues were spread evenly on the culture dish and placed upside down in a primary incubator for 4 h to adhere. The tissues were maintained in DMEM (Gibco, Grand Island, NY, USA) with 15% FBS (Gibco) for 1-2 weeks. Purified fibroblasts were obtained by differential enzyme digestion and maintained in DMEM with 10% FBS. Cells were seeded into T-25 culture flasks and grown in 5 ml serum-free media for 48 h until the cells were ~ 80% confluent. Conditioned media were collected and centrifuged at 2000 rmp for 30 min to remove cellular debris.

Cell culture
HNSCC cell lines Cal27 and HN6 were used. Cal27 and HN6 cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). All cancerous cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM)/F12 (Invitrogen) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin/ streptomycin at 37 °C in a 5% CO 2 -humidified incubator. All cells were routinely tested for mycoplasma at regular intervals throughout the whole course of the study. To inhibit of FAK phosphorylation, cells were treated with FAK inhibitor 14 (FAKi, Santa Cruz Biotechnology, Santa Cruz, CA, USA) at different concentrations.

Lentiviral construct and transfection
Human LOX-encoding lentiviral vectors were constructed by GeneChem Co., Ltd (Shanghai, China). CAFs cells were transfected with a specific shRNA lentivirus targeting LOX knockdown, and the scrambled sequence was co-transfected as a negative control. All transfection procedures were performed according to the manufacturer's instructions. Cells in the control group (CAFs-NC) and the experimental group (CAFs-shLOX) were cultured at 37 °C in a 5% CO 2 incubator and changed to complete medium 12-16 h after transfection. Stable cells were selected using 2 mg/ml puromycin. Fluorescence microscopy was used to observe transfection efficiency, and real-time polymerase chain reaction (PCR) and western blot were used to estimate the efficiency of LOX knockdown.

Protein extraction and nuclear, cytoplasmic protein fractionation
Protein extraction was performed using common methods. Briefly, cells were lysed with an appropriate radioimmunoprecipitation assay (RIPA) buffer (Beyotime, Shanghai, China) containing protease inhibitor cocktail and phosphatase inhibitor cocktail (Beyotime, Shanghai, China), incubated for 15 min at 4 °C, and scraped using a cell scraper. For cells cultured on collagen, cells were agitated on a rocking platform at 4 °C for 30 min and then removed the lysis buffer. Nuclear and cytoplasmic proteins were isolated using the Invent Nuclear and Cytoplasmic Extraction Reagents Kit (Invent Biotechnologies, Plymouth, MN, USA) according to the manufacturer's instructions.

Western blot assay
Western blot assays were performed according to the common methods. Briefly, equal amounts of protein samples were separated using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA). The membranes were blocked with 5% skimmed milk at room temperature for 2 h and incubated at 4 °C overnight with primary antibodies against β-actin as a control (1: (1:1000, CST) followed by incubation with anti-goat immunoglobulin G (IgG) horseradish peroxidase (HRP)conjugated secondary antibodies (Zhongshan Golden Bridge Bio, Beijing, China) for 45 min at room temperature. The protein bands were detected using ECL chemiluminescence reagents (Millipore) and visualized using the ImageQuant LAS4000 Mini Imaging System (General Electrics, Louisville, Kentucky). Analyses of the bands were performed using ImageJ software.

Cell proliferation assay
Cell proliferation was assessed by CCK-8 cell assay (Cell Counting Kit-8, Dojindo, Japan). Cells (3 × 10 3 cells/well) were seeded into 96-well plates, and 10 μl CCK-8 reagent was added to each well 2-4 h before testing at the same time. Cell viability was determined at 0, 1, 2, 3 and 4 days after cell attachment. Three parallel repeats were performed for each sample in each experiment and the results were expressed as the mean of three independent experiments.

Cell migration and invasion assays
Cell migration and invasion assays in vitro were performed using wound healing and transwell assays. For the wound healing assay, cells were plated in six-well plates and grown to 90% confluence. Artificial wounds were created with a 200 μl sterile pipette tip across the cell surface and then the cells were incubated with CM to migrate into the open area. Images of the same area of the wound were taken at 0, 12 and 24 h to determine the wound closure. Cell invasion assays were performed using 8 μm pore size transwell chambers (Miliipore) whose upper chamber was precoated with Matrigel (Corning, Bedford, MA, USA). For the migration assays, there was no matrigel in the upper chamber. Approximately 1 × 10 5 cells per well were seeded in the upper chambers and incubated with CM in the lower chamber. After 24 h, the transwell chambers were fixed with 4% PFA and stained with crystal violet (Sigma-Aldrich, St. Louis, MO, USA). Cells attached to the lower layer were imaged (Olympus, Tokyo, Japan) and counted in randomly selected fields.

Preparation of collagen matrix
For the fibroblast-modified collagen, fibroblasts were resuspended in FBS, and type I collagen, 5 × DMEM and reconstitution buffer (50 mmol/L NaOH, 260 mmol/L NaHCO 3 , and 200 mmol/L HEPES) were sequentially added and evenly mixed. The mixture was spread in a 12-well plate, and the plate was placed in a 5% CO 2 incubator at 37 °C for 30-60 min to solidify. Complete medium was added and the medium was changed every day. For the collagen treated with CAFs-CM, CM from CAFs-NC, CAFs-shLOX were collected to stimulate the formation of collagen with different stiffness. DMEM was used as blank control, and human recombinant LOX (huLOX; OriGene, Rockville, MD, USA) was used for the rescue of LOX knockdown. 12-well plates were coated with 1.5 mg/ml collagen and placed at room temperature for 30 min followed by incubation at 37 °C for 2 h, and then stimulated by DMEM, CM from CAFs-NC, CAFs-shLOX and CAFs-shLOX + huLOX for 5 days. For the collagen treated with drugs, huLOX was used to induce crosslinking of collagen. Ribose (Sigma-Aldrich, Dorset, UK), which induced crosslinking non-enzymatically, was used as a positive control and BAPN (Sigma-Aldrich, Dorset, UK) was used as an inhibitor of LOX. Then, PBS only (PBS), or PBS containing 150 ng/ml huLOX (huLOX), or 60 mM ribose (ribose), or huLOX combined with 10 mM BAPN (huLOX + BAPN) were added to each group and incubated at 37 °C for 5 days.

Determination of matrix stiffness
The relative stiffness of collagen gels was measured using a MARS60 microinfrared rheometer (Thermo Fisher, MA, USA) under the following conditions. Briefly, the collagen gels were measured over a range from 0.5-5% strain at a fixed angular frequency of 0.5 rad/s and temperature of 21 °C on MARS60 microinfrared rheometer. The output of G' was recorded for 20 min with a data point obtained every 60 s. The samples were found to be only minimally frequency dependent within the range of testing and showed a linear viscoelastic response within the strain range evaluated.

Three-dimensional co-culture system
Fibroblast-modified collagen was prepared as previously described. Cal27 cells were resuspended in the mixed medium and then seeded onto the surface of the solidified gel to create a 3D co-culture system. The culture medium was changed daily. After 3 days, the collagen gels were transferred to the scaffold in six-well plates for culture at the air-liquid interface. At the specific time point, the gels were fixed with 4% PFA, embedded in paraffin, and cut into 4 μm sections for H&E staining.

Statistical analysis
The correlation of α-SMA and LOX expression with clinicopathological parameters and prognosis was analyzed using ANOVA, the chi-square test, Fisher's exact test, Mann-Whitney test, COX regression analysis and Kaplan-Meier survival analysis. Spearman's correlation analysis was used to estimate the correlation between α-SMA and LOX in the tumor stroma. The results are represented as the means ± SEM or as specified in the figure legends. The data were analyzed using SPSS 18.0 and Prism 7 software (GraphPad Software, La Jolla, CA, USA). Comparisons between the two groups were analyzed using the 2-tailed, unpaired Student's t test. Comparisons among the three groups were performed using one-way ANOVA followed by Tukey's multiple comparisons test. Adjusted values of P < 0.05 were considered significant.

Expression of LOX in the stroma of OSCC is associated with CAFs
The fibroblast marker FAP, the epithelial marker EPCAM, and LOX were stained by multiplex immunohistochemical (IHC) staining to evaluate the expression distribution of LOX in OSCC tissues. We found that LOX was highly expressed both in the epithelium and stroma of OSCC (Fig. 1A). To investigate the expression of LOX in tumor stroma, we also evaluated the expression of LOX and  expressed at low levels in the region with the low expression of α-SMA and FAP, suggesting the colocalization between LOX and CAFs (Fig. 1B). Then, the sections of 96 OSCC clinical samples were subjected to IHC staining to evaluate the expression of α-SMA and LOX in the stroma. α-SMA and LOX were separately stained in serial sections of OSCC samples. The results showed that in 96 cases of OSCC, 56.25% (54/96) exhibited relatively high expression of α-SMA (Additional file 1: Fig. S1A1-3), and 43.75% (42/96) exhibited low expression (Additional file 1: Fig. S1C1-3). Meanwhile, 64.58% (62/96) showed the high expression of LOX (Additional file 1: Fig. S1B1-3), whereas the remaining 35.42% (34/96) showed the low expression (Additional file 1: Fig. S1D1-3). The association between these two parameters was analyzed, as shown in Table 1, LOX expression in the stroma of OSCC was positively associated with α-SMA expression (P = 0.008), further verifying the correlation of the two.

Association of clinicopathological parameters with α-SMA and LOX expression in the stroma of OSCC
The staining results of 96 OSCC clinical samples were scored based on the IHC staining scoring standards, and their relationship with the clinicopathological parameters was evaluated. Notably, α-SMA expression in CAFs significantly correlated with tumor size ( Table 2, P = 0.028), and the higher expression of α-SMA was observed in patients with advanced clinical stages ( Table 2, P = 0.012). In addition, LOX expression in the stroma of OSCC was significantly associated with the advanced clinical stage ( Table 2, P = 0.042). The correlations between α-SMA or LOX expression with patient age, sex, tumor location, lymph node metastasis, distant metastasis and pathological grade could not be identified.
Additionally, the COX proportional hazards regression model and the Kaplan-Meier survival analysis were used to further investigate the prognostic significance of LOX and α-SMA expression in the stroma of OSCC. As shown in Tables 3, 4, the univariate and multivariate COX analyses indicated that LOX expression in tumor stroma could serve as an independent prognostic factor for the overall survival (OS) of OSCC patients. The Kaplan-Meier survival analysis showed that higher expression of LOX in tumor stroma had an adverse prognostic impact on the OS of OSCC patients (Fig. 1C, P < 0.001). Based on these results, it was concluded that the stromal expression of LOX was significantly associated with the malignant grade and poor prognosis of OSCC.

LOX-positive CAFs enhance proliferation, migration, invasion and EMT process in OSCC cells
Primary CAFs and normal fibroblasts (NFs) were obtained from fresh human oral cancer tissues and paired adjacent non-tumor tissues. The morphology of purified NFs and CAFs were shown in Additional file 2: Fig. S2A. The expression of CAFs classical markers in three paired tissues were verified by RT-PCR and western blot (Additional file 2: Fig. S2B, C). The study found that LOX was highly expressed in CAFs compared with matched NFs ( Fig. 2A, B), which was in line with the clinical results and proved the correlation between LOX and CAFs. To characterize the biological roles of LOX-positive CAFs in OSCC, a loss-of-function assay was performed by knocking down LOX expression through lentivirus transfection. As shown in Additional file 2: Fig. S2D, E, the protein and RNA expression levels of LOX significantly decreased in CAFs transfected with shRNA lentivirus (CAFs-shLOX), compared with that in the control group (CAFs-NC). Then, CAFs-NC and CAFs-shLOX conditioned medium were harvested to stimulate oral cancer cells Cal27 and HN6. The CCK-8 assay showed that the downregulation of LOX in CAFs inhibited the proliferation and viability of Cal27 and HN6 cells (Fig. 2C). Meanwhile, after LOX knockdown in CAFs, the invasive ability of both cell lines was significantly reduced as measured by the transwell assay (Fig. 2D). The scratch test and the migration assay showed that the migratory ability of oral cancer cells was weakened upon LOX knockdown ( Fig. 2E; Additional file 2: Fig. S2F). Changes in the expression of EMT markers were also related to tumor progression in vitro. As shown in Fig. 2F, G, the expression of epithelial marker E-cadherin was upregulated in OSCC cells upon LOX suppression in CAFs, while the expression of mesenchymal markers vimentin and N-cadherin were downregulated. To further demonstrate the occurrence of EMT in OSCC cells, we performed immunofluorescence assay and found it consistent with the above-mentioned results (Additional file 2: Fig. S2G). These results confirmed that LOX expression in CAFs promoted the proliferation, migration, invasion and EMT of OSCC cells.

Effect of LOX-positive CAFs on matrix collagen stiffness
To investigate the effect of LOX-positive CAFs on matrix collagen stiffness, type I collagen was used for co-cultured with NFs and CAFs cells. The contractility of collagen in the CAFs-treated group was significantly stronger than that in the NFs-treated group. However, the contractile capacity of collagen was reduced when LOX was knocked down by lentivirus or inhibited by BAPN in CAFs (Fig. 3A). The changes in collagen diameter in different groups were consistent with its contractility (Fig. 3B). Meanwhile, the pore size of collagen in the CAFs group was significantly smaller than that in the NFs group, whereas the downregulation of LOX reversed this effect (Fig. 3C, D). These results indicated that LOX expression in CAFs could directly promote the crosslinking of collagen, resulting in an increase in contractile capacity and a reduction in the pore size of collagen.
To detect collagen stiffness, a MARS60 microinfrared rheometer was used to evaluate the elastic modulus of collagen. The elastic modulus of collagen in the CAFs group significantly increased compared with that in the NFs group, while the elastic modulus reduced upon LOX inhibition (Fig. 3E). Collectively, these data confirmed that LOX expressed in CAFs played an important role in ECM remodeling, leading to further changes in matrix collagen stiffness.

Effect of matrix stiffness on the biological behavior of OSCC cells
The interaction between cancer cells and the microenvironment can affect tumor growth and progression [7]. To explore the effects of LOX-mediated matrix stiffness on the biological behavior of oral cancer cells, CM extracted from CAFs-NC and CAFs-shLOX cells was added to type I collagen, on which oral cancer cells were cultured. The relative stiffness of stimulated collagen was measured by rheology. The results showed that collagen in each group reached a steady state in ~ 15-20 min (Fig. 4A), indicating that collagen substrates with different stiffness were successfully constructed. CM from CAFs-NC led to an increase in the stiffness of collagen, which was inhibited by CM from CAFs-shLOX, and this inhibition could be rescued by huLOX (Fig. 4B). Similarly, type I collagen was modified by huLOX, ribose or BAPN to construct cell culture substrate, and the relative elastic modulus was measured (Additional file 3: Fig. S3A), in order to further verify the effect of matrix stiffness on the progression of OSCC. Cell morphology is related to cell motility, and abnormal changes are important factors influencing cell biological behavior. Therefore, Cal27 cells were seeded onto collagen with different stiffness. Immunofluorescence staining for cytoskeletal protein F-actin showed that Cal27 cells became slender in the CAFs-NC group compared with the control group, and returned to normal morphology in the CAFs-shLOX group (Fig. 4C), suggesting that matrix stiffness mediated by CAFs-derived LOX had a certain effect on the morphological changes in oral cancer cells. Furthermore, the 3D collagen co-culture system showed that the number and depth of tumor cells invading into the collagen substrate were higher in the CAFs-NC group than in the NFs group, which were reduced by LOX knockdown in CAFs (Fig. 4D, E), proving the effect of matrix stiffness on the invasion ability of tumor cells. Meanwhile, rescue experiment were carried out for further confirmation, and we found that huLOX can reversed the morphological change and invasion ability of Cal27 cells in the CAFs-shLOX group. Also, immunofluorescence staining and transwell assays with drug-modified collagen showed that LOX-mediated high collagen stiffness led to deformation and promoted cancer cell invasion (Additional file 3: Fig. S3B, C). Furthermore, western blot assay showed that the increase in matrix stiffness might induce LOX expression, promote N-cadherin and vimentin expression, and reduce E-cadherin expression, suggesting that the increase in matrix stiffness promoted EMT in OSCC cells ( Fig. 4F; Additional file 3: Fig. S3D). Collectively, these results suggested that CAFs-derived LOX-mediated changes in the   biomechanical properties (stiffness) of the cellular environment directly affected the cellular behavior.

Matrix stiffness promotes OSCC progression via FAK signaling
Considering the critical role of FAK in regulating the biological behaviors of tumor cells, such as adhesion, invasion and EMT [21,24], we speculated that matrix stiffness might promote tumor progression by activating the FAK signaling pathway. Thus, the FAK phosphorylation level at tyrosine 397 was detected in OSCC cells cultured on collagen with different stiffness. The western blot assay showed that the FAK phosphorylation was upregulated in the CAFs-NC group and downregulated  19:513 in the CAFs-shLOX group (Fig. 5A). Meanwhile, the same activation of FAK phosphorylation was found on the drug-modified collagen (Additional file 4: Fig. S4A). FAK inhibitor 14 (FAKi), a selective small-molecule inhibitor, could repress the phosphorylation of FAK at tyrosine 397. Here, FAK phosphorylation was suppressed after treatment with FAKi, but the total protein level of FAK showed no difference (Additional file 4: Fig. S4B). The increase in matrix stiffness caused by CAFs-derived LOX promoted the proliferation and invasion ability of OSCC cells, which could be reversed by FAKi (Fig. 5B-D). Furthermore, FAKi inhibited the matrix stiffness-induced EMT process in OSCC cells (Fig. 5E). Meanwhile, we found that the expression of β-catenin and its downstream proteins c-myc and cyclin D1 was increased in the group with the high stiffness of collagen, indicating that the Wnt/β-catenin signaling pathway might be activated (Additional file 4: Fig. S4A). The cell fraction assay showed that the increase in β-catenin was observed both in the cytoplasm and nucleus of OSCC cells cultured on collagen with high stiffness, while FAKi restrained matrix stiffness-stimulated nuclear β-catenin enrichment (Additional file 4: Fig. S4C). These results indicated that the matrix stiffness increased by CAFs-derived LOX induced an increase in the FAK phosphorylation level, which might impact tumor development. Moreover, the activation of the FAK phosphorylation pathway might further promote the nuclear accumulation of β-catenin. Further investigation should be performed to explore the underlying mechanism leading to this change in the expression of β-catenin in the cytoplasm and nucleus mediated by matrix stiffness.

Discussion
Recently, it has become clear that TME promotes tumor progression [25]. The components and mechanical properties of ECM have aroused concern as important determinants of cancer cell behavior and disease progression. As the major TME cellular component, CAFs play a key role in matrix remodeling [6]. However, whether LOX participates in extracellular matrix remodeling induced by CAFs and affects the progression of OSCC remains unclear. Our study revealed that LOX is highly expressed in CAFs, and LOX can mediate CAFs-induced matrix remodeling by catalyzing collagen cross-linking. The increase in matrix stiffness can, in turn, promote tumor invasion and progression in an FAK-dependent manner. CAFs are responsible for the secretion of multiple growth factors, kinases, cytokines and chemokines into the TME to facilitate tumor progression. Research identified high expression of LOX in several malignant tumors [26]. The secretion of LOX was reported to induce the formation of a pre-metastatic niche in preclinical breast cancer models [27]. LOX has a significant effect on ECM remodeling in the lungs and promotes lung metastasis [28] and correlates with increased staging in renal cell carcinoma [13]. In previous studies, the expression of LOX was usually stronger at the tumor invasive front than in the tumor center in OSCC, and higher expression of LOX was associated with advanced pathological grade [12]. Here, we found that LOX was highly expressed in the same region of CAFs distribution in tumor stroma (Fig. 1), and there was a statistical correlation between the two (Table 1). Notably, higher expression of LOX in the tumor stroma was associated with an advanced clinical stage and poor prognosis (Tables 2, 3, 4). Moreover, we found that the expression level of LOX was much higher in CAFs than in NFs, and further confirmed that LOX expressed in CAFs functioned as an oncogene in OSCC (Fig. 2). These results suggest that the expression of LOX in tumor stroma may be a key molecule affecting the progression of OSCC.
There has been extensive evidence to prove that biophysical factors, including matrix stiffness and mechanical forces, play critical roles in the initiation and progression of malignant tumors [29][30][31]. CAFs can produce large quantities of ECM molecules to remodel stroma, and LOX expression and collagen content have been observed to be clinically elevated in various malignant tumors [13,[32][33][34][35]. In Kurie's review, the deposition of a cross-linked collagen matrix by CAFs leads to intratumoral fibrosis [36]. Xu et al. summarized the most abundant matrix protein polymers are collagens, whose crosslinking by LOX can create a stiffer tumor microenvironment, regulate tumor immunity, and promote metastasis [37]. In present study, in order to investigate the effect of LOX in CAFs on matrix remodeling, we created a co-culture system and found that the contractile capacity of collagen was enhanced by CAFs, leading to an increase in collagen density and a decrease in pore size. Meanwhile, the elastic modulus of collagen in the CAFsco-cultured group was elevated (Fig. 3). Thus, we innovatively explored the relationship between CAFs and LOX, finding that LOX may be a key factor in CAFs to remodel ECM through collagen crosslinking, and then promote cancer progression.
In the development of malignant tumors, the complexity of the epithelial-stromal interaction is due to the fact that the ECM is a dynamic mechanical microenvironment. ECM remodeling leads to functional changes in the biomechanical properties of the ECM, resulting in the activation of pathogenic signaling pathways involved in various processes such as cell morphology, proliferation, differentiation, migration and etc. [30,38,39]. In breast carcinoma, mammary epithelial cell cultured within high-density collagen form larger, less organized structures and proliferate faster than that those cultured in low-density collagen [40]. LOX has been identified as playing a critical role in establishing and mediating a prometastatic microenvironment within fibrotic tissues [35]. In the present study, type I collagen was stimulated with CM from CAFs or drugs to form different stiffness, and the morphology of oral cancer cells cultured on high-stiffness collagen became slender. Moreover, a 3D collagen co-culture model was successfully established, in which oral cancer cells were plated on collagen with different stiffness mediated by CAFs. Our results showed that an increase in collagen stiffness could promote the depth of tumor cell invasion into the collagen matrix (Fig. 4). Such changes in the mechanical properties of the ECM arising from LOX-mediated collagen cross-linking play a critical role in the transformed phenotype of oral cancer cells.
Tumors respond to changes in ECM stiffness by integrin activation and subsequent phosphorylation of FAK, leading to the activation of protein-protein interactions and adhesion signaling to promote tumorigenesis [41,42]. FAK is required for the activation of the Wnt/βcatenin pathway through phosphorylation of GSK3β Y216 , which is crucial for intestinal tumorigenesis [43]. In Xue's study, LOXL2 secreted by CAFs activated the FAK signaling pathway and then induced EMT in human colon cancer LOVO cells, thereby promoting the invasion and metastasis of colon cancer (44). By seeding OSCC cells on the collagen with different stiffness, we found that collagen with a stiffer surface mediated by CAFs-derived