Tumour-derived TGF-β1 modulates myofibroblast differentiation and promotes HGF/SF-dependent invasion of squamous carcinoma cells

1Historically, most studies of neoplastic transformation and progression have focused on the tumour cell. However, in addition to transformed cells, tumours are also composed of host stromal tissue comprising fibroblasts, newly formed blood vessels, extracellular matrix and immune components. Although stroma was initially thought to support tumour development passively, there is increasing evidence to suggest that it actively contributes to malignant progression.

2A common finding in many types of solid tumours is that stromal fibroblasts become ‘activated’ and express a number of contractile proteins, particularly α-smooth muscle actin (SMA). These cells have been referred to as peritumour fibroblasts, carcinoma-associated fibroblasts and activated stroma, but are now more commonly called myofibroblasts (MF). The process of activation of fibroblasts is associated with increased proliferation, increased deposition of collagen and spliced-variant forms of fibronectin, assembly of vinculin-containing fibronexus adhesion complexes and acquisition of smooth muscle cell characteristics. In fact, it is the expression of SMA that is the hallmark of the myofibroblastic phenotype. The structural changes, such as assembly of fibronexi, and accumulation of cytoskeletal SMA, modulate MF contractility and reduce their migratory potential. Conversely, myofibroblasts upregulate the secretion of numerous growth factors, chemokines and cytokines, as well as extracellular matrix proteins and proteases.

3A number of cytokines including PDGF, IL-4, insulin-like growth factor II and TGF-β1 may be involved in the transdifferentiation of fibroblasts to myofibroblasts, and these can be derived from a number of different cell types. Among these cytokines, TGF-β1 is considered to have a central role in inducing the myofibroblastic phenotype, because it is capable of upregulating fibroblast SMA and collagen both in vitro and in vivo. Indeed, high levels of the cytokine are usually associated with MF-containing lesions. In many types of cancers, TGF-β1 is overexpressed by carcinoma cells, and it has been proposed previously that the expression of this cytokine by breast and prostate carcinoma cells induces reactive stroma. TGF-β1 has many effects: In addition to inhibiting epithelial cell proliferation, it also promotes the secretion of matrix proteins and proteases. Its powerful antiproliferative effect has led to it being thought of as a tumour suppressor in carcinomas. However, it is now apparent that TGF-β1 may be pro-oncogenic, driving malignant progression, invasion and metastasis. This is partly explained by observations of carcinomas, including oral SCC, which become refractory to the antiproliferative effect of TGF-β1. However, another mechanism by which TGF-β1 could promote tumour development is by inducing the transdifferentiation of stromal fibroblasts, producing an activated, myofibroblast-rich stromal microenvironment. For example, it has been shown that TGF-β1 produced by breast cancer cells activates normal breast stromal fibroblasts and promotes them to produce urokinase-type plasminogen activator, a serine protease important in cancer cell invasion and metastasis. Such changes have a potential role in tumorigenesis since, if tumour stroma becomes activated and immobilised in the vicinity of tumour cells, paracrine interactions may be established between the separate cellular compartments, some of which could encourage tumour development.

4To date, there has been little work investigating potential interactions between squamous carcinomas (SCC) and the surrounding stroma. showed that IL-1 produced by epidermal keratinocytes induced the expression of keratinocyte growth factor by dermal fibroblasts, which in turn stimulated keratinocyte proliferation. It has also been suggested that PDGF-activated stromal cells may maintain elevated keratinocyte proliferation via a paracrine mechanism. demonstrated that peritumour fibroblast-conditioned medium promoted SCC migration on tenascin, and that this effect could be partially inhibited by blocking EGF, TGF-β1 or hepatocyte growth factor/scatter factor (HGF/SF). In addition, paracrine interactions have been demonstrated between squamous carcinoma cells and other cell types found in stroma. found tumour-derived TGF-β1 and monocyte chemotactic protein-1 attracted and activated monocytes. They suggested that macrophages secreted TNF-alpha and IL-1, which in turn stimulated tumour cells to produce IL-8 and VEGF, the latter cytokine then inducing angiogenesis.

5The aim of the study was to investigate the role of squamous carcinoma cells in myofibroblast transdifferentiation, to determine the effect of such cells on SCC invasion and to elucidate the possible mechanisms involved in these processes.

6We show that myofibroblasts are commonly found within the stroma of squamous carcinoma in vivo, particularly at the invasive front. We demonstrate that squamous carcinoma cells may directly induce a myofibroblast phenotype in primary fibroblasts through the secretion of TGF-β1. Furthermore, such transdifferentiated myofibroblasts significantly upregulate the secretion of hepatocyte growth factor (HGF/SF), which promotes SCC invasion through basement membrane proteins. These in vitro data are consistent with the possibility that a similar double paracrine effect may also exist in vivo.

7Six monoclonal antibodies (mAbs) (all of mouse origin) were used in this study. Antibodies were purchased against human TGF-β1, HGF/SF (R&D Systems, Abingdon, UK), c-met and phosphorylated c-met (Upstate Ltd, Milton Keynes, UK) and SMA (Sigma, Dorset, UK; DAKO, High Wycombe, UK). W632 (anti-MHC class I) was a kind gift from W Bodmer (IMM, Oxford). FITC-conjugated rabbit anti-mouse immunoglobulin was purchased from DAKO (High Wycombe, UK). Recombinant human TGF-β1 was purchased from R&D Systems, Abingdon, UK. Matrigel was obtained from Becton Dickinson (Oxford, UK).

8Sections (3 μm) were dewaxed, brought to absolute alcohol and endogenous peroxidase neutralised with 0.5% methanolic hydrogen peroxide for 10 min. Sections were washed in water, followed by 0.05% Tween 20 in TBS pH 7.4 (TBS/Tween). Primary anti-SMA antibody was applied for 60 min at a dilution of 1: 150 (Dako, High Wycombe, UK). Sections were again washed in TBS/Tween and secondary antibody applied for 30 min (Dako K5001 ChemMate HRP/DAB kit). Sections were washed in TBS/Tween and peroxidase-labelled streptavidin was applied for 30 min (Dako K5001). The peroxidase was visualised using DAB (Dako K5001) for 7 min and counterstained in Mayer's haematoxylin. In all, 15 archival oral SCCs and 10 benign polyps showing fibroepithelial hyperplasia were chosen at random, stained for SMA and scored by two pathologists independently (PMS and GJT), according to the Quickscore method. Briefly, the staining intensity was scored out of 3 (1=weak, 2=moderate, 3=strong), and the proportion of the stroma in or adjacent to the tumour staining positively was scored out of 4 (1=<25%, 2=25–50%, 3=51–75%, 4=76–100%). The score for intensity was added to the score for proportion to give a score in the range of 0–7 and grouped as –(score=0), +(score=1–3), ++(score=4–5) or +++(score=6–7). Pathologists agreed completely in 11 of the 15 cases of SCC. The remaining four cases were reanalysed and a consensus score agreed. Staining of benign polyps for SMA-positive myofibroblasts was uniformly negative.

9Human primary oral fibroblasts (OF) had been established previously in the laboratory from redundant human tissue obtained during routine periodontal surgery at the Eastman Dental Hospital. Oral fibroblasts were maintained in fibroblast growth medium. This consisted of D-MEM (Life Technologies, Gibco BRL, Paisley, UK) supplemented with 10% foetal calf serum (FCS; PAA Laboratories, Yeovil, UK) plus penicillin (100 U ml−1) and streptomycin (100 μg ml−1) (Life Technologies). Cells were maintained in a humidified atmosphere of 5% CO2 at 37°C and routinely passaged using tryspin–EDTA (Life Technologies). A panel of three oral SCC cell lines were used. We generated the invasive VB6 cell line previously by transfection and retroviral infection of integrin subunits to express high levels of the integrin αvβ6. CA1 and 5PT were kind gifts from Professor IC Mackenzie (Cardiff dental School, UK). Cells were grown in standard keratinocyte growth medium (KGM) as described. Keratinocyte growth medium comprised α-MEM containing 10% FCS (Globepharm, Surrey) supplemented with 1.8 × 10−4 M adenine, 5 μg ml−1 insulin, 0.5 μg ml−1 hydrocortisone and 10 ng ml−1 epidermal growth factor (Sigma).

10Squamous carcinoma cells were grown to 70% confluence in KGM in 80 cm2 culture flasks, washed twice with phosphate-buffered saline (PBS; Life Technologies) and incubated for 72 h with 10 ml of α-MEM. The SCCM from each cell line was collected, clarified by centrifugation and the cells were detached with trypsin/EDTA and counted. A total of 1.5 × 103 OF cm−2 were plated in fibroblast growth medium in 80 cm2 culture flasks or on glass coverslips for 3 days, then washed twice with PBS and incubated for 72 h with α-MEM, SCCM (at equal volumes keratinocyte cell number−1) or α-MEM containing TGF-β1 (R&D Systems, Abingdon, UK; 10 ng ml−1). For blocking studies, anti-TGF-β1 antibody (R&D Systems; 1 μg ml−1) or a control antibody (W632; anti-MHC type 1; 1 μg ml−1) was added to the SCCM for 30 min prior to incubation with the fibroblasts. All experiments were performed in triplicate and the experiments were repeated four times.

11A total of 1.5 × 103 fibroblasts cm−2 were plated in fibroblast growth medium in 80 cm2 culture flasks for 3 days and then washed twice with PBS. To induce a myofibroblast, phenotype cells were incubated for 72 h with α-MEM containing recombinant TGF-β1 (R&D Systems, Oxford, UK), which was acid-activated prior to use (4 mM HCl/0.1% BSA). TGF-β1 was titrated at concentrations ranging from 0.5 to 10 ng ml−1, with maximum SMA induction observed at 10 ng ml−1. This concentration was then used routinely in all experiments. Control cells were cultured in α-MEM alone. Cells were also incubated for 72 h with SCCM (at equal volumes/keratinocyte cell number). The cells were washed twice with PBS and cultured for a further 72 h in α-MEM. The control fibroblast- (FCM) or myofibroblast-conditioned medium (MCM) was collected, clarified by centrifugation and the cells were detached and counted. The volumes of FCM and MCM were corrected for cell number, adjusted to a total volume of 500 μl and used in the lower chamber of a Transwell invasion assay as a chemoattractant, or assayed by ELISA for HGF/SF.

12Primary fibroblasts grown on glass coverslips were treated with SCCM and α-MEM (±TGF-β1) for 72 h. For blocking studies, anti-TGF-β1 antibody (R&D Systems; 1 μg ml−1) or a control antibody (W632; anti-MHC type 1; 1 μg ml−1) was added to the SCCM for 30 min prior to incubation with the primary fibroblasts. The cells were prefixed in 2% paraformaldehyde (BDH), rinsed in PBS, fixed with methanol (BDH) for 10 min at −20°C, permeabilised in 0.25% Triton (Sigma) in PBS and labelled by indirect immunostaining. Primary anti-SMA antibody (Sigma, clone IA4) was used at a concentration of 1: 1000, while the secondary antibody was FITC-conjugated rabbit anti-mouse immunoglobulin (Dako, High Wycombe, UK; 1: 500). Nuclei were visualised using DAPI (Sigma). Images were captured using a Cohu CCD camera attached to a Leica DM IRB microscope (Leica Microsystems (UK) Ltd, Milton Keynes, UK).

13ELISA kits for TGF-β1 and HGF/SF were purchased from R&D Systems (Oxford, UK). The assay is based on a two site ELISA ‘sandwich’ format. Cell supernatants were prepared as for conditioned medium and TGF-β1 activated by adding 0.1 ml of 1 M HCL for 10 min. This was neutralised with 100 μl of 1.2 M NaOH/0.5 M HEPES. Sample (200 μl) was added to each well and TGF-β1 or HGF/SF detected by a peroxidase-labelled FAb′ antibody directed to either cytokine. The reaction was stopped by the addition of an acid solution and the resultant colour change was read at 450 nm on a spectrophotometer. The concentration was determined by interpolation from a standard curve using known concentrations of TGF-β1 or HGF/SF standards as supplied.

14Cells of equal confluence were lysed with SDS lysis buffer containing protease inhibitors (1% SDS, 10 mM Tris, pH 7.4, leupeptin 100 μg ml−1, phenylmethylsulphonyl fluoride 100 μg ml−1, aprotinin 100 μg ml−1) and the protein was estimated using the BCA protein assay reagent (Pierce Warriner). Samples containing equal protein were boiled in reducing buffer (0.5 M Tris-HCl pH 6.8, 10% SDS, 10% glycerol, 0.4% bromophenol blue, 10% β-mercaptoethanol) and electrophoresed in 10% SDS–PAGE gel. Protein was electrotransferred onto nitrocellulose membranes (Hybond-C, Amersham, UK) in transfer buffer (20 mM glycine, 25 mM Tris, 0.6 mM SDS, 10% methanol) for 12 h at 26 mV. To prevent nonspecific binding, blots were blocked for 1 h at room temperature in 5% skimmed milk powder (Marvel®, Cadbury, UK) in PBS 0.1% Tween. Anti-SMA antibody (Sigma; clone 1A4; 1: 1000 dilution), anti-c-met antibody (Upstate Ltd, UK; 1: 750 dilution) and antiphosphorylated c-met (Upstate Ltd, UK; 1: 500 dilution) were used for immunoblotting. Horseradish peroxidase-conjugated anti-mouse was used as secondary antibody at a 1: 2000 dilution. Blots were developed with the ECL Western blotting detection kit system (Amersham, UK). Blots were also probed for β-actin as an additional loading control. These experiments were repeated a minimum of three times.

15Cell invasion assays were performed using Matrigel-coated polycarbonate filters (8 μm pore size, Transwell®, Beckton Dickinson) as described previously. Matrigel (70 μl; 1: 2 dilution in α-MEM) was added to the upper membrane and allowed to gel for 1 h at 37°C. Fibroblast conditioned medium or MCM was corrected for cell number, adjusted to a final volume of 500 μl with α-MEM and used as a chemoattractant in the lower chamber of the Transwell. For blocking experiments, the conditioned media were incubated with anti-HGF/SF antibody (R&D Systems; 10 μg ml−1) for 30 min at 4°C prior to placing in the assay. An irrelevant antibody (W632; anti MHC-type I; 10 μg μl−1) was used as a control. Squamous carcinoma cells were plated in the upper chamber of quadruplicate wells at a density of 5 × 104 in 200 μl of α-MEM and incubated at 37°C for 72 h. The cells in the lower chamber (including those attached to the undersurface of the membrane) were then trypsinised and counted on a Casy 1 counter (Sharfe System GmbH, Germany). Experiments were repeated four times in quadruplicate.

16Data are expressed as the mean±s.d. of a given number of observations. Where appropriate, one-way analysis of variance (ANOVA) was used to compare multiple groups. For comparisons between groups, Fisher's PLSD (set at 5% significance) was used. A P-value of <0.05 was considered to be significant.

17Of the 15 oral SCCs examined, 11 (73%) contained a significant proportion of strongly SMA-positive stromal cells, indicating myofibroblastic differentiation (Table 1 ). Four tumours contained focal areas of strong stromal SMA expression. No tumour was completely SMA negative, and myofibroblastic stromal differentiation was seen in all tumour grades and stages. Although variation in myofibroblast distribution was seen between different tumours, this did not correlate with tumour grade or architecture. Figure 1 shows the prominence of myofibroblasts in the tumour stroma. Such cells were usually concentrated at the invasive margin of the tumour, directly abutting malignant epithelial cells (Figure 1C). Stromal SMA expression often demarcated the margin of the tumour (Figure 1C) and was only observed in close proximity to the tumour mass (Figure 1D), even in tumours containing a diffuse inflammatory infiltrate. This suggests that in oral SCC myofibroblasts may be induced primarily by tumour cells. In addition, 10 benign mucosal polyps were stained for SMA expression to determine whether hyperplastic (but non-malignant) squamous epithelium could also induce SMA induction in adjacent fibroblasts. Although positive staining of blood vessel smooth muscle was observed (Figure 1I arrow), no polyp contained SMA-positive myofibroblasts in the connective tissue (Figure 1(G, H and I).

18Human gingival fibroblasts expressed low levels of SMA in culture (Figure 2A, 3A). Immunostaining showed occasional cells with weak, diffuse cytoplasmic expression of the protein (Figure 3A). Treatment of primary fibroblasts with exogenous-activated TGF-β1 (at concentrations ranging from 0.5 to 10 ng ml−1) produced a significant increase in SMA expression (data not shown). Maximum SMA expression was observed at a concentration of 10 ng ml−1 (Figures 2A and 3B). Figure 3B demonstrates increased intensity of SMA staining and shows that the protein is now associated with cytoplasmic stress fibres. SMA upregulation was also observed when primary fibroblasts were cultured in conditioned medium from VB6, CA1 and 5PT cell lines squamous carcinoma cell lines (SCCM) (Figures 2A and 3C–H).

19ELISA on SCCM confirmed that the SCC cell lines produced TGF-β1 (Figure 2E), the highest levels secreted by 5PT cells. In order to determine the proportion of activated TGF-β1 in the SCC supernatants, we repeated the ELISA comparing acid-activated conditioned medium (in which all TGF-β1 is in active form) with untreated samples. In conditioned media from VB6, CA1 and 5PT cells, we found that the proportion of activated TGF-β1 relative to total TGF-β1 was 87, 81 and 59%, respectively. In addition, however, fibroblasts probably activate TGF-β1, and thus the initial amount of activated cytokine in the SCC medium, before it is placed onto the fibroblasts, may not be relevant to its eventual biological effect.

20To demonstrate that the generation of a myofibroblastic phenotype was TGF-β1-dependent, we carried out blocking studies using a TGF-β1 inhibitory antibody that blocks the biological activity of activated TGF-β1. Figures 2B–D and 3I–K demonstrate that when TGF-β1 inhibitory antibody was added to SCCM from VB6, CA1 and 5PT cells prior to fibroblast treatment, the induction of SMA expression was reduced significantly (as determined by densitometric scanning; by 65%, P=0.0297; 75%, P=0.0028; 82%, P=<0.0005, respectively). These data were confirmed by immunofluorescence, which showed only a weak diffuse cytoplasmic staining for SMA with no stress fibre formation when TGF-β1 was inhibited (Figure 3I–K).

21To determine whether myofibroblasts secrete factors, which stimulate the invasion of squamous carcinoma cell lines, we carried out transwell assays through Matrigel. Myofibroblasts were generated using either TGF-β1 or SCCM from each cell line, and then cultured for 72 h in α-MEM. The MCM was used as a chemoattractant in the lower chamber of the Transwell and SCC cells were allowed to invade towards this stimulus for 72 h before being counted. Primary FCM was used for comparison. We demonstrate that MCM significantly promoted invasion of VB6, CA1 and 5PT cells compared with FCM (Figure 4A; P=<0.0001, <0.0001, 0.0005, respectively). Myofibroblasts-conditioned medium from myofibroblasts generated using exogenous recombinant TGF-β1 (10 ng ml−1) produced a similar level of invasion as MCM from myofibroblasts generated by SCCM from each of the cell lines (Figure 4A). If TGF-β1 was inactivated using a blocking antibody added to SCCM, no transdifferentiation of myofibroblasts was seen (Figure 3I–K), and conditioned medium from such cells (which remained fibroblasts) no longer promoted invasion (Figure 4B). To ensure that altered cell invasion was not simply due to increased cell proliferation, growth assays were performed in which SCC cells were grown in MCM or FCM for 72 h. The cell proliferation was low (due to the absence of serum in FCM and MCM) and no differences in growth rate were observed.

22Previously, several studies have demonstrated that myofibroblasts may secrete HGF/SF. This cytokine acts to promote epithelial cell growth and migration and has been shown to stimulate invasion in prostate carcinoma cells. To determine whether induction of a myofibroblastic phenotype was associated with increased production of HGF/SF, we examined conditioned medium by ELISA. Myofibroblasts were generated using either exogenous TGF-β1 or SCCM from each SCC cell line. Untreated primary fibroblasts were used as a control. Figure 5 demonstrates that MCM contains significantly higher levels of HGF/SF compared with FCM (up to 35-fold higher). Myofibroblasts that had been generated using conditioned medium from all three SCC cell lines consistently showed a significant upregulation of HGF/SF secretion when compared with primary fibroblast controls (VB6, P=0.0063; CA1, P=0.0003; 5PT, P=<0.0001). Secreted HGF/SF levels were generally higher in myofibroblasts, which had been generated with conditioned medium from the 5PT cell line. This was probably due to the higher levels of TGF-β1 produced by this line (Figure 2E).

23To determine whether HGF/SF promoted invasion in the Transwell assays, we carried out blocking experiments using inhibitory antibodies directed against HGF/SF, which were added to the MCM. Figure 6A confirms that the three SCC cell lines expressed the HGF/SF receptor c-met. Figure 6D demonstrates that inactivation of HGF/SF significantly reduced the invasion of VB6, CA1 and 5PT cells through Matrigel (P=0.0014, 0.0159, 0.0012, respectively). Following HGF/SF inhibition, the level of invasion was similar to that produced by FCM, suggesting that the invasion-promoting effect of MCM was mediated by HGF/SF.

24These data are the first to show that invasion of SCC may be due partly to a double paracrine effect, resulting in proinvasive release of HGF/SF from stromal myofibroblasts.

25Accumulation of fibroblast-like cells, including myofibroblasts, is frequently observed associated with the edge of an actively expanding tumour mass. Such a phenomenon has been demonstrated, to different extents, in a variety of tumours and there is increasing evidence that tumour stroma may promote tumour progression. Interactions between epithelial cells and fibroblasts have a major role in many biological processes and it follows that the interactions between tumour cells and neighbouring myofibroblasts may be biologically significant, probably mediated by soluble factors such as growth factors and cytokines. This has been demonstrated previously in breast cancer where TGF-β1 produced by breast cancer cells activates normal breast stromal fibroblasts and promotes them to produce proteases. Similar interactions have been shown in prostatic carcinomas, and in the fibrosis observed in organs such as the kidney and liver.

26In the present study, we examine potential interactions between squamous carcinoma cells and primary fibroblasts. We show that stromal cells in SCC in vivo often express SMA, indicating a myofibroblastic phenotype (Table 1; Figure 1). Such cells are most commonly found at the invasive margin, directly abutting tumour cells but are absent in areas distant from tumour. Furthermore, myofibroblasts were not detected in benign mucosal polyps. These data are consistent with the possibility of a tumour-derived, diffusible factor that promotes fibroblast-to-myofibroblast transdifferentiation.

27A number of cytokines including PDGF, IL-4, insulin-like growth factor II and TGF-β1 may be involved in the transdifferentiation of fibroblasts to myofibroblasts, and these can be derived from several cell types. In addition, mast cell-derived histamine and tryptase has been reported to induce SMA expresson in fibroblasts. However, it is generally accepted that TGF-β1 has a key role in inducing myofibroblast differentiation, and high levels of the cytokine are usually associated with MF-containing lesions. TGF-β1 is frequently detectable in SCC, particularly in the more advanced stages of tumour progression, and relatively high concentrations of TGF-β1 are usually found in tumour stroma. Recently, showed that keratinocytes genetically modified to produce activated TGF-β1 induced collagen type I gene expression in dermal fibroblasts in a coculture system. The role of TGF-β1 in SCC is complex and studies suggest that TGF-β1 has biphasic actions on tumour cells, having an important negative growth effect in the early stages of carcinogenesis, but at later stages enhancing invasion and metastasis through epigenetic mechanisms. However, most studies have concentrated on the direct effect of TGF-β1 on tumour cells. Our data suggest that a possible indirect tumour-promoting effect of SCC-derived TGF-β1 may be in generating a myofibroblastic stroma, which in turn modulates invasion in a paracrine manner.

28Myofibroblasts may promote tumour progression in a number of different ways. They upregulate the expression of serine and matrix metalloproteinases, which degrade and remodel extracellular matrix, possibly potentiating cell invasion and migration. In addition, showed that peritumour FCM upregulated the expression of the integrin αvβ6 in SCC cells, and we have previously demonstrated that de novo expression of this integrin promotes invasion of oral carcinoma. Myofibroblasts also secrete interstitial matrix, as well as numerous soluble mediators of inflammation and growth factors, including HGF/SF. The latter cytokine was originally identified as a potent mitogen for hepatocytes, but was also identified independently as a scatter factor (SF), a secretory protein of fibroblasts and smooth muscle cells that dissociates and induces motility of epithelial cells. Scatter factor and HGF were later found to be identical, hence the current name HGF/SF. The cytokine modulates its effects through the c-met tyrosine kinase receptor and misregulated expression of both cytokine and receptor is a common finding in many tumour types. Although c-met is generally expressed in oral SCC in vivo and in vitro, it is not commonly mutated to a constitutively active form, and is not tumour-promoting per se in the absence of ligand. Hepatocyte growth factor/scatter factor may induce invasive growth by affecting the activity and expression of cadherins, integrins and matrix metalloproteinases. This results in disruption of intercellular junctions, dissolution of epithelial basement membrane and altered integrin interactions with extracellular matrix. Fibroblast-derived HGF/SF has been shown to stimulate invasion and migration in a number of tumour types including squamous cell carcinoma, and we have demonstrated previously that exogenous HGF/SF induces expression of the type IV collagenases MMP-2 and -9 in squamous carcinoma cells. The latter observation suggests a possible mechanism for the HGF/SF-dependent invasion through basement membrane-like Matrigel (which comprises predominantly type IV collagen) described in this study. In addition, we have shown more recently that HGF/SF regulates integrin function in oral SCC cells.

29Several other paracrine interactions between keratinocytes and fibroblasts have been demonstrated previously. For example, it has been suggested that PDGF-activated stromal cells maintain elevated keratinocyte proliferation via a paracrine mechanism, and showed that IL-1 produced by epidermal keratinocytes induced the expression of KGF by dermal fibroblasts, which in turn stimulated keratinocyte proliferation. Paracrine interactions have also been demonstrated between squamous carcinoma cells and other cell types. found tumour-derived TGF-β1 and monocyte chemotactic protein-1 attracted and activated monocytes. They suggested that macrophages secreted TNF-alpha and IL-1, which in turn stimulated tumour cells to produce IL-8 and VEGF, the latter cytokine-inducing angiogenesis.

30In conclusion, this study shows for the first time that a double paracrine interaction between SCC cells and fibroblasts can exist that results in enhanced tumour invasion. We show that SCC-derived TGF-β1 induces a myofibroblastic phenotype and that such cells secrete significantly higher levels of HGF/SF compared with primary fibroblast controls. In turn, HGF/SF promotes invasion of SCC cells through basement membrane proteins. We also confirm that the myofibroblast population is usually located adjacent to the invasive front of SCC. These clinical observations are consistent with the suggestion that the paracrine interactions observed in vitro between SCC and fibroblasts may also occur in vivo, and emphasises the importance of the stromal contribution to tumour development.