Cholesteatoma-associated fibroblasts modulate epithelial growth and differentiation through KGF/FGF7 secretion

The keratinocyte growth factor (KGF/FGF7), produced by stromal cells, is a key paracrine mediator of epithelial proliferation, differentiation and migration. Expression of the growth factor is increased in wound healing and in hyperproliferative epithelial diseases, as a consequence of the activation of dermal fibroblasts by the inflammatory microenvironment. The middle ear cholesteatoma, an aural epidermal pathology characterized by keratinocyte hyperproliferation and chronic inflammation, may represent a model condition to study the epithelial-mesenchymal interactions. To develop an in vitro model for this disease, we isolated and characterized human primary cultures of fibroblasts associated with the cholesteatoma lesion, analyzing their secretory behaviour and degree of differentiation or activation. Compared to the perilesional or control normal fibroblasts, all cultures derived from cholesteatoma tissues were less proliferating and more differentiated and their highly variable activated phenotype correlated with the secretion of KGF as well as of metalloproteases 2 and 9. Culture supernatants collected from the cholesteatoma-associated fibroblasts were able to increase the proliferation and differentiation of human keratinocytes assessed by the expression of Ki67 and keratin-1 markers. The single crucial contribution of the KGF released by fibroblasts on the keratinocyte biological response was shown by the specific, although partial, block induced by inhibiting the KGF receptor or by immunoneutralizing the growth factor. Altogether, these results suggest that the activation of the stromal fibroblasts present in the pathological tissue, and the consequent increased secretion of KGF, play a crucial role in the deregulation of the epidermal proliferation and differentiation.


Introduction
The keratinocyte growth factor (KGF/FGF7) is a member of the Wbroblast growth factor family and a major paracrine mediator of the epidermal homeostasis through regulation of the keratinocyte proliferation, diVerentiation and migration (Finch et al. 1989;Marchese et al. 1990;Ceccarelli et al. 2007). Produced by stromal cells, KGF binds exclusively to the KGFR, a splicing transcript variant of the Wbroblast growth factor receptor 2 (FGFR2b) (Miki et al. 1992). In addition to its physiologic role, KGF is up-modulated in its expression in human wound healing during the process of re-epithelization (Marchese et al. 1995). In previous papers from our group, we have reported that KGF expression is also increased in hyperproliferative epithelial diseases, such as psoriasis and acanthoma, as a consequence of the activation of dermal Wbroblasts by the inXammatory microenvironment (Kovacs et al. 2005(Kovacs et al. , 2006. These results are consistent with the current knowledge that, in diVerent epithelial pathological conditions and in tumorigenesis, the activation of the stromal cells is able to sustain the keratinocyte hyperproliferation and altered diVerentiation through the secretion of soluble paracrine growth factors and cytokines (Räsänen and Vaheri 2010).
Among the hyperproliferative epithelial diseases, the middle ear cholesteatoma (CHO) is an aural lesion, which may grow and invade the surrounding stroma, including bone and middle/inner ear structures (Olszewska et al. 2004). Although diVerent studies have shown the invasive and hyperproliferative behaviour of the CHO epidermal tissue (Bujia et al. 1993;Vennix et al. 1996), very little is still known about the molecular mechanisms involved in the pathogenesis of the disease. Others and we have suggested that the proliferation and migration of the keratinocytes in CHO might be mediated by an altered release of autocrine and paracrine growth factors and a modulated expression of their receptors (Kojima et al. 1994;Tanaka et al. 1999;Yetiser et al. 2002). Among them, KGF and KGFR expression appear up-modulated in CHO tissue (Kojima et al. 1996;Ishibashi et al. 1997;Yamamoto-Fukuda et al. 2003Barbara et al. 2008;d'Alessandro et al. 2010) and the eYcacy of a common therapeutic treatment of the lesion is related to a down-modulation of the growth factor (Yamamoto-Fukuda et al. 2008). Therefore, it is possible that the activation of the stromal Wbroblasts associated to the lesion and the consequent release of KGF could play a role in the deregulation of the epidermal proliferation and diVerentiation, which characterizes the CHO tissue.
The development of in vitro models of epithelial-mesenchymal interaction may greatly contribute to elucidate the eVects mediated by the stromal Wbroblasts on the epithelial cells and the molecular mechanisms involved in the CHO pathogenesis. To this aim, in this study, we established and characterized human primary cultures of CHO-associated Wbroblasts analyzing their growth and secretory behaviour and their degree of diVerentiation/activation. In addition, we evaluated the ability of their culture supernatants to modulate the biological response of human keratinocytes, demonstrating the key contribution of the KGF secreted in the medium in modulating this response. Our results indicate that CHOassociated Wbroblasts might be able to sustain the growth and invasive behaviour of the lesion through the secretion of paracrine diVusible factors, such as KGF, and of pro-invasive molecules, such as metalloproteases.

Patients and tissue samples
Tissue samples were obtained from middle ear cholesteatoma specimens (CHO-AFs) collected during surgical procedures performed on 11 consecutive patients. The control samples were obtained from skin of the medial external auditory canal (MEAC-Fs) during surgical procedure performed on a patient with otosclerosis and from non-auricolar skin (NAS-Fs) by punch-biopsy performed on a patient with abdominal aortic aneurysm. The clinical and demographic characteristics of the enrolled patients and site of sampling are described in Table 1. All patients were extensively informed and gave written consent for the investigations. The study was approved by the ethics committee of the University Hospital where it was conducted. Tissue samples were examined by toluidine blue semithin sections and assessed independently by two pathologists.

Cell cultures and treatments
Primary cultures of human Wbroblasts were obtained from middle ear cholesteatoma specimens (CHO-AFs). Control cultures were derived from perilesional tissue of cholesteatomatous patients (perilesional-Fs), from normal human skin of the medial external auditory canal (MEAC-Fs) and from non-auricolar skin (NAS-Fs).
Tissue samples were cut into small pieces, digested with dispase 0.1 mg/ml and collagenase I 0.35 % for 45 min at 37°C, pelleted, resuspended and maintained in Dulbecco's modiWed Eagle medium (DMEM) containing 10 % fetal bovine serum (FBS) and antibiotics. Primary cultures were obtained and propagated up to the second passage.
To induce Wbroblast diVerentiation, the cultures of CHO-AFs were maintained in DMEM without serum for 48 h in the presence of TGF 50 ng/ml (PeproTech Inc., Rocky Hill, NY, USA). Supernatants (SNs) obtained from primary cultured CHO-AFs, MEAC-Fs, NAS-Fs and perilesional-Fs kept in serum-free medium for 48 h were collected and frozen at ¡80°C until use. SNs were also obtained from two primary cultured CHO-AFs kept in serum-free medium in presence of TGF 50 ng/ml for 48 h.
The human keratinocyte cell line HaCaT was cultured in DMEM supplemented with 10 % FBS and antibiotics. Primary cultures of normal human keratinocytes (NHKs) were derived from skin biopsies as described previously (Belleudi et al. 2011)  To selectively block the KGF activity, we used a neutralizing rabbit polyclonal antibodies to KGF (C-19, Santa Cruz Biotechnology, Santa Cruz, CA, USA) (10 g/ml in culture medium).
The primary antibodies were visualized, after appropriate washing with PBS, by using goat anti-mouse IgG-FITC Fluorescence signals were analyzed by conventional Xuorescence or by scanning cells in series of 0.5 m sequential sections with an ApoTome System (Zeiss, Oberkochen, Germany) connected with an Axiovert 200 inverted microscope (Zeiss); image analysis was then performed by the Axiovision software (Zeiss).
The percentage of Ki67-or K1-positive cells was analyzed counting for each treatment a total of 500 cells, observed in ten microscopic Welds randomly taken from three diVerent experiments. Results have been expressed as mean values § SE.
Densitometric analysis was performed using Quantity One Program (Bio-Rad Headquarters, Hercules, CA, USA). BrieXy, the signal intensity for each band was calculated and the background subtracted from experimental values. The resulting values were then normalized and expressed as fold increase with respect to the control and mean values § SE from three diVerent experiments in triplicate.

Primers
Oligonucleotide primers for target genes and for the housekeeping gene were chosen with the assistance of the Oligo 5.0 computer program (National Biosciences, Plymouth, MN) and purchased from Invitrogen. The following primers were used: for KGF target gene: 5Ј-CACCAGGCAGA CAACAGACAT-3 (sense), 5Ј-GTAAGTTCAGTTGCT GTGACGCT-3Ј (anti-sense). For each primer pair, we performed no-template control and no-reverse-transcriptase control (RT negative) assays, which produced negligible signals.
RNA extraction and cDNA synthesis RNA was extracted using the TRIzol method (Invitrogen, Carlsbad, CA) according to manufacturer's instructions and eluted with 0.1 % diethylpyrocarbonate (DEPC)-treated water. Total RNA concentration was quantitated by spectrophotometry and the quality was assessed by measuring the optical density ratio at 260/280 nm. RNA samples were stored at ¡80°C. After denaturation in DEPC-treated water at 70°C for 10 min, 1 g of total RNA was used to reverse transcription using iScript™ cDNA synthesis kit (Bio-Rad) according to manufacturer's instructions.

PCR ampliWcation and real-time quantitation
Real-time PCR was performed using the iCycler Real-Time Detection System (iQ5 Bio-Rad) with optimized PCR conditions. The reaction was carried out in 96-well plate using iQ SYBR Green Supermix (Bio-Rad) adding forward and reverse primers for each gene and 1 l of diluted template cDNA to a Wnal reaction volume of 15 l. All assays included a negative control and were replicated three times. The thermal cycling programme was performed as follows: an initial denaturation step at 95°C for 3 min, followed by 45 cycles at 95°C for 10 s and 60°C for 30 s. Real-time quantitation was performed with the help of the iCycler IQ optical system software version 3.0a (Bio-Rad), according to the manufacturer's manual. The relative expression of the housekeeping gene was used for standardizing the reaction. The comparative threshold cycle (C t ) method was applied to calculate the fold changes of expression compared to control cells. Results are reported as mean § standard deviation (SD) from three diVerent experiments in triplicate.

Flow cytometry for cell cycle
For each culture, the Wbroblasts were plated sparsely, trypsinized, pelleted and resuspended in 70 % ethanol in PBS and stored at 4°C overnight. Cells were washed with PBS, resuspended in propidium iodide (PI) staining solution (50 g/ml) and RNAse A (100 Kunitz/ml) (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) and incubated in the dark for 40 min at room temperature. At least 20,000 cells for each culture were collected and analyzed with MACS-Quant ® Analyzer Xow cytometer (Miltenyi Biotec GmbH). Cell cycle distribution was calculated with MACSQuantify ® software (Miltenyi Biotec GmbH).

Morphological analysis
For the Wbroblast morphological analysis, three diVerentiation stages were evaluated: FI = small spindle shaped cell, FII = small epithelioid cells, FIII = large epithelioid cells (Bayreuther et al. 1988). The L:E diVerentiation index was calculated analyzing the Wbroblasts in early (E: FI + FII) and late (L: FII/FIII + FIII) diVerentiation state (Herskind and Rodemann 2000). The culture samples were observed on a Zeiss Axiovert 200 inverted microscope equipped with diVerential interference contrast (DIC) optics. Quantitative analysis was performed by counting, for each cell culture, a total of at least 250 cells observed in Wve microscopic Welds randomly taken from three diVerent experiments.

Statistical methods
Mann-Whitney non-parametric test was used to compare variables that do not assume Gaussian distribution. Student's t test was used to compare variables that assume Gaussian distribution. All the correlation measures were evaluated by the Pearson test (r) and by linear analysis of regression curve. P values < 0.05 were assumed as statistically signiWcant.

Establishment and characterization of primary cultures of Wbroblasts derived from cholesteatoma lesions and assessment of their diVerentiated phenotype
To analyze the behaviour of the stromal Wbroblasts associated with the cholesteatoma (CHO) hyperproliferative epidermal disease, we isolated and propagated in culture primary Wbroblasts obtained from middle ear CHO specimens (CHO-AFs). For control samples, we established Wbroblast cultures from the perilesional tissue of cholesteatomatous patients (perilesional-Fs) as well as from normal human skin of the medial external auditory canal (MEAC-Fs) and from non-auricolar skin (NAS-Fs). All the cells were isolated and cultured as described in "Materials and methods". The number of the cultures analyzed throughout the study and the clinical and demographic characteristics of the enrolled patients and site of sampling are described in Table 1.
The mesenchymal phenotype of the primary cultures and their purity were Wrst conWrmed by immunoXuorescence assessment of the expression of vimentin, a component of the intermediate Wlaments widely used as a Wbroblastic marker: the analysis showed that, in all cultures, cells were positive for vimentin staining, displaying a signal compatible with the structure and localization of perinuclear cytoplasmic bundles of Wlaments (Fig. 1a). To rule out the presence of epithelial cells in the cultures, cells were incubated with anti-cytokeratins antibodies which recognize the intermediate Wlaments of epithelial cells: no signal was detected for cytokeratins (not shown) conWrming the mesenchymal phenotype of all the isolated primary cultures.
To analyze possible real diVerences in the biological characteristics of our cultures, which may reXect the in vivo behaviour, as well as to avoid in vitro diVerentiation, we used all cultures at the second passage (P2). To evaluate the degree of diVerentiation of the cells, Wrst we utilized and adapted a morphological characterization (Bayreuther et al. 1988) proposed to distinguish in vitro diVerent diVerentiation stages based on distinct morphological features: the three stages (FI-FII-FIII) used in our classiWcation were deWned as follows: FI = small spindle shaped cell, FII = small epithelioid cells, FIII = large epithelioid cells. The L:E diVerentiation index was calculated analyzing the Wbroblasts in early (E: FI + FII) and late (L: FII/FIII + FIII) diVerentiation stage (Herskind and Rodemann 2000). The quantitative analysis revealed that the percentages of cells of the FII and FIII groups, corresponding to more diVerentiated stages, was highly variable among the CHO-AFs (FII + FIII ranging from approximately 14 to 97 %), while more homogeneous percentages were found in the controls perilesional-Fs (ranging from approximately 58 to 78 %) and in MEAC-Fs (40 %) or NAS-Fs (61 %). Consistent with these percentages, also the L:E ratio was highly variable in CHO-AFs, as reported in Table 1 (representative examples are shown in Fig. 1b).
To evaluate the growth rate of the Wbroblast cultures, we analyzed the cell cycle distribution by cytoXuorimetry as described in "Materials and methods". The results showed that all the CHO-AFs cultures were characterized by a lower percentage of cells in G2/M phase compared to control cultures (Table 1 and representative examples in Fig. 1c) and that a reduction of the mitotic ability was frequently related to a higher L:E ratio (Table 1).
Since it is well established that the activation of quiescent Wbroblasts leads to the acquisition of a more diVerentiated phenotype and to an increased secretion of extracellular matrix components and proteases (for a recent review, see Räsänen and Vaheri 2010), we quantiWed by ELISA test, in cellular homogenates of CHO-AFs and control Wbroblasts, the expression levels of the metalloproteases MMP-2 and MMP-9, which are known to be modulated in cholesteatoma tissues (Schönermark et al. 1996;Shibosawa et al. 2000). The concentrations of both MMP-2 and MMP-9 obtained from CHO-AFs were clearly higher compared to MEAC-Fs, NAS-Fs and perilesional-Fs homogenates ( Fig. 1d; Table 1) and the CHO-AFs expressing increased amounts of MMPs correspond to less proliferating cultures (Table 1).
Because the expression of -smooth muscle actin ( -SMA) is the most commonly used marker for activated and diVerentiated Wbroblasts (Räsänen and Vaheri 2010), to better evaluate the diVerentiated phenotype of the CHO-AFs, we analyzed by quantitative immunoXuorescence the percentage of cells positive for -SMA: as shown in Fig. 2a, the immunoXuorescence signal appeared compatible with actin bundles in the peripheral areas of the cytoplasm and was enhanced in most of CHO-AFs. In addition, the percentage of the -SMA-positive cells was higher in the CHO cultures compared to perilesional-Fs and control cells (Fig. 2a). The percentages of -SMA-positive cells were correlated with the parameters of activation/diVerentiation used above: the linear regression analysis showed a strong positive correlation Fig. 1 Characterization of the primary cultures of cholesteatomaassociated Wbroblasts. a ImmunoXuorescence analysis of expression of the mesenchymal marker vimentin on representative example of diVerent cultures of human Wbroblasts isolated from non-auricolar skin (NAS-Fs), from normal skin of the medial external auditory canal (MEAC-Fs), from middle ear cholesteatoma samples (CHO-AFs) and from perilesional tissue of cholesteatomatous patients (perilesional-Fs). in all cultures, the cells are positive for vimentin staining, which appears as perinuclear cytoplasmic bundles of Wlaments. Nuclei are stained with DAPI. Bar 20 m. b Morphologycal analysis of the primary cultures by diVerential interference contrast microscopy. Three diVerentiation stages were deWned: FI, small spindle shaped cell (blue); FII: small epithelioid cells (red); FIII: large epithelioid cells (green). The Late:Early (L:E) ratio was calculated. The quantitative analysis indicate that among the CHO-AFs the percentages of cells in the more diVerentiated stages FII, FIII and the L:E ratio are highly variable. In contrast, in the control perilesional-Fs and in MEAC-Fs or NAS-Fs the percentages are homogenous. Bar 50 m. c Fibroblast cultures were stained with PI and analyzed for cell cycle distribution with Xow cytometry. The percentage of cells in G2/M results lower in CHO-AFs compared to control cultures and the more quiescent phenotype frequently corresponds to a higher L:E ratio. d MMP-2 and -9 quantitation by ELISA test on cellular homogenates from CHO-AFs, perilesional-Fs, MEAC-Fs, NAS-Fs. The MMP-2 and -9 levels in CHO-AFs were detected in various amounts but are higher respect to the cellular homogenates from control Wbroblasts. Results reported in graph represent the mean values § SE. Mann-Withney test was performed and signiWcance level has been deWned as described in "Materials and methods".   (Fig. 2b). Taken together, these results indicate that, although showing a great variability in morphological shape and in biological behaviour, the CHO-AFs display a more diVerentiated and activated phenotype than the perilesional-Fs, NAS-Fs and MEAC-Fs. Interestingly, some of the CHO-AFs showing the diVerentiated/activated phenotype were isolated from cholesteatoma tissue samples characterized by an extended dermal inXammatory inWltrate (Table 1), assessed as previously described (d'Alessandro et al. 2010).
Supernatants from CHO-AFs enhance keratinocyte proliferation and diVerentiation through KGF/FGF7 release It is well known that there is a close interplay between the stromal cells and the epithelial counterpart; this interaction is mediated by soluble paracrine signals and secreted extracellular matrix and proteases released from mesenchymal cells which regulate the epithelial proliferation, diVerentiation and migration (Räsänen and Vaheri 2010). In order to analyze the possible biological activity of soluble factors produced by the cultured CHO-AFs on a cellular epithelial model, we treated the human keratinocyte HaCaT cell line with the culture supernatants (SNs) collected from CHO-AFs, NAS-Fs, MEAC-Fs and perilesional-Fs, to recreate in vitro the interaction that occurs in vivo between derma and epithelium. To Wrst analyze the proliferative eVect, HaCaT cells were serum starved, treated with SNs from CHO-AFs or from control cells for 48 h at 37°C and then Wxed and stained with antibodies against Ki67 antigen, which identify cycling cells (Fig. 3a). Quantitative immunoXuorescence analysis indicated that the percentage of cells presenting positive nuclei was higher in keratinocytes treated with SNs collected from CHO-AFs than in those treated with SNs from NAS-Fs or MEAC-Fs. Moreover, the SNs from perilesional-Fs induced a lower proliferating rate than SNs collected from the corresponding lesional cultures, but similar to SNs of control Wbroblasts (Fig. 3a). The growth rate induced by the diVerent SNs was correlated with the percentage of -SMA expression in the corresponding primary Wbroblast cultures and the linear regression analysis revealed a signiWcant positive correlation (r: +0.66; f: p < 0.01) (Fig. 3a).
To evaluate the ability of the SNs to induce epithelial diVerentiation, we analyzed the expression in keratinocytes of the early diVerentiation marker keratin 1 (K1). To this aim, HaCaT cells, which are known to undergo diVerentiation following cell conXuence and stratiWcation (Ryle et al. 1989;Capone et al., 2000), were let to grow up to conXuence, serum starved and treated with SNs as above before Wxation and staining with anti-K1 antibodies (Fig. 3b). Quantitative immunoXuorescence analysis showed that SNs isolated from CHO-AFs induced a higher percentage of K1 positive HaCaT cells than the SNs from NAS-Fs and MEAC-Fs. Again, cells treated with SNs from perilesional-Fs appeared less diVerentiated than the cells treated with the SNs from the corresponding lesional cultures and similar to SNs from control cells (Fig. 3b). The percentage of K1 positive HaCaT cells induced by the diVerent SNs was correlated with -SMA expression as above and the linear regression analysis showed a signiWcant positive correlation (r: +0.67; f: p < 0.0001) (Fig. 3b). Altogether, these results indicate that SNs from CHO-AFs have a greater ability to induce epithelial cell proliferation and diVerentiation in comparison with SNs from control Wbroblasts.
To verify whether the eVects induced by SNs treatment on the HaCaT keratinocytes could be due to KGF released by the Wbroblasts, the addition of the supernatants to the cells was performed in the presence of the speciWc FGFR tyrosine kinase inhibitor SU5402, similar to our previous in vitro study (Visco et al. 2009). Treatment with recombinant KGF 20 ng/ml in the presence or not of SU5402 was used as control. Cells were then Wxed and stained with anti-Ki67 (Fig. 4a, c). The quantitative immunoXuorescence analysis of cycling cells as above showed that, as expected (Belleudi et al. 2007), KGF treatment induced an increase in the percentage of Ki67 positive cells that was signiWcantly reduced by the presence of the KGFR inhibitor. Similarly, SU5402 was able to inhibit the proliferative eVects for all the SNs; however, the inhibition was much greater in the keratinocytes treated with the SNs isolated from CHO-AFs (Fig. 4a, c), suggesting that the proliferative activity of the SNs derived from CHO-AFs could be most attributable to an increased release of KGF in the culture medium. Comparable results were obtained by inhibiting the proliferative eVect of the SNs through immunoblock of KGF with neutralizing anti-KGF antibodies (Fig. 4b, c), unequivocally indicating the relevance of the growth factor in the process.
Since it is known that KGF promotes keratinocyte diVerentiation (Capone et al. 2000;Belleudi et al. 2011) and we have reported that in cholesteatoma tissue KGF up-modulation is associated with keratinocyte diVerentiation (d' Alessandro et al. 2010), to evaluate if the eVects of the SNs on the expression of K1 could be ascribed to KGF, HaCaT cells were treated as above in the presence of SU5402, Wxed and stained with anti-K1 antibodies (Fig. 5a, c). The quantitative analysis showed that the percentages of K1-positive cells in response to treatment either with the growth factor or the diVerent SNs was reduced in the presence of the receptor inhibitor (Fig. 5a, c). The diVerentiative eVects were more reduced by the inhibitor when the stimulation was obtained with the SNs collected from CHO-AFs (Fig. 5a, c) compared to the treatments with SNs from control Wbroblasts, further indicating that KGF might represent the major paracrine eVector secreted by the activated Wbroblasts associated with the cholesteatoma lesions. Again, the immunoblock of KGF by the addition of neutralizing   antibodies led to an inhibition of the diVerentiative eVect totally comparable to that obtained through the use of SU5402 (Fig. 5b, c), conWrming the key role played by KGF. In order to demonstrate eYcient secretion of KGF in the medium from the Wbroblast cultures, we performed Sandwich ELISA assay to analyze the KGF concentration in the culture supernatants. The results indicate that all Wbroblast cell cultures were able to produce KGF in various amounts (Table 1; Fig. 6a). The concentrations of KGF obtained from CHO-AFs were clearly higher compared to perilesional-Fs, MEAC-Fs and NAS-Fs SNs (Fig. 6a). Interestingly, among the CHO-AFs, the #1410 whose SN contained high amounts of KGF and showed the greater ability to induce epithelial cell proliferation and diVerentiation, corresponded to activated and diVerentiated Wbroblasts characterized by the highest percentage of SMA-positive cells and the greatest MMP levels. In contrast, the #1211 SN, which contained lower amounts of KGF, was collected from Wbroblasts with a quiescent phenotype like control cells.
The relationship between the diVerentiation and activation degree of the Wbroblast cultures and their ability to secrete KGF was demonstrated by linear regression analysis (Fig. 6a) that showed a positive correlation between the -SMA expression rate of the cultures and the KGF levels in diVerent SNs (r: +0.50; f: p < 0.05). In addition, the role of KGF in the HaCaT cell stimulation induced by SNs treatment was assessed by regression analysis (Fig. 6b): in fact, the KGF levels in the SNs showed a positive correlation with HaCaT proliferation (r: +0.58; f: p < 0.05) and diVerentiation rates (r: +0.68; f: p < 0.05).
To better deWne the molecular mechanisms leading to the increased KGF release in the culture supernatants from CHO-AFs and to ascertain if the variable amounts of KGF secreted in the medium would correspond to enhanced expression of the growth factor at both transcriptional and protein levels, we performed both RT-PCR and Western blot analysis on a selection of cultured Wbroblasts characterized by highly diVerent amounts of KGF secretion: the results shown in Fig. 6c and d demonstrated, compared to MEAC-Fs and perilesional-Fs, a clear increased mRNA and protein expression in those CHO-AFs #2810 and #0811, which displayed higher concentrations of the growth factor detected in ELISA test (Fig. 6c, d). In accordance with the conclusion that the enhanced KGF secretion corresponds to a real transcriptional up-regulation, very low levels of mRNA and protein expression were observed in CHO-AFs #1211 characterized by a quiescent phenotype and by a low amount of the growth factor released in the medium (Fig. 6c, d).
To further demonstrate that the proliferative and diVerentiative eVects induced by the SNs collected from the activated lesional Wbroblasts represent a physiological/ pathological phenomenon occurring in middle ear epidermis, we performed a parallel set of experiments using human primary cultured keratinocytes (HKs). As observed for HaCaT cells, the SN from lesional Wbroblasts, but not that from the perilesional cells, was able to increase either the proliferation or the diVerentiation of the HKs (Fig. 7a, b). The addition of anti-KGF neutralizing antibodies during the incubation with the SNs as above caused a reduction of the percentage of cycling and K1-positive cells to the untreated levels ( Fig. 7a, b), conWrming the key role played by KGF in the keratinocyte stimulation.
To prove that the secretion of the KGF from the primary cultures is related to the Wbroblast diVerentiation degree, we used TGF treatment to induce in vitro diVerentiation of the CHO-AFs cultures. In fact, it is well documented that TGF promotes the diVerentiation and activation of quiescent stromal Wbroblasts (Serini and Gabbiani 1999;De Wever et al. 2008). We selected two CHO-AFs moderately diVerentiated, #2010 Fig. 3 Biological eVects induced by SNs from the Wbroblast cultures on keratinocyte proliferation and diVerentiation. a HaCaT cells were serum-starved, treated with SNs of diVerent primary cultures of Wbroblasts for 48 h at 37°C, Wxed and immunostained with anti-Ki67 polyclonal antibodies, which identiWes cycling cells. Quantitative immunoXuorescence analysis indicates that the percentage of cells presenting positive nuclei was higher in HaCaT treated with SNs collected from CHO-AFs respect to that treated with SNs from control cells. The SNs from perilesional-Fs (#0811p and #1011p) induce a lower proliferating rate respect to SNs from the corresponding lesional tissue (#0811and #1011), but similar to SNs of control Wbroblasts NAS-Fs and MEAC-Fs. At linear regression analysis, the proliferation rate induced by the diVerent Wbroblast derived SNs on HaCaT cells shows a positive correlation with -SMA expression in the corresponding primary Wbroblast cultures (r: +0.66; f: p < 0.01). Results reported in graph represent the mean values § SE. Nuclei are stained with DAPI. Bar 20 m. Student's t test was performed and signiWcance level has been deWned as described in "Materials and methods". Statistics: p < 0.05 versus untreated and p = NS versus MEAC-Fs; *p < 0.05 versus untreated; **p < 0.05 versus MEAC-Fs; ***p < 0.01 versus MEAC-Fs, ****p < 0.01 versus MEAC-Fs and p < 0.01 versus corresponding perilesional-Fs; *****p < 0.01 versus MEAC-Fs and p < 0.05 versus corresponding perilesional-Fs. b ConXuent HaCaT cells were serum-starved, treated with SN as above, Wxed and immunostained with anti-K1 polyclonal antibodies, which identify early diVerentiated cells. Quantitative immunoXuorescence analysis reveals that SNs collected from CHO-AFs induce a higher diVerentiation respect to SNs from control cells. Cells treated with SNs from perilesional-Fs (#0811p and #1011p) appear less diVerentiated respect to those treated with SNs from the corresponding lesional tissue (#0811and #1011) and similar to SNs of NAS-Fs and MEAC-Fs. At the linear regression analysis, the percentage of K1 positive keratinocytes in response to the diVerent SNs shows a positive correlation with -SMA expression of the corresponding Wbroblast (r: +0.67; f: p < 0.0001). Nuclei are stained with DAPI. Bar 50 m. Results reported in graph represent the mean values § SE. Student's t test was performed and signiWcance level has been deWned as described in "Materials and methods". Statistics: *p < 0.05 versus untreated and p = NS versus MEAC-Fs; **p < 0.05 versus untreated; ***p < 0.05 versus untreated and p = NS versus MEAC-Fs, ****p < 0.01 versus untreated and p < 0.01 versus MEAC-Fs; *****p < 0.05 versus MEAC-Fs and p < 0.05 versus corresponding perilesional-Fs and #2810, and we treated them with TGF 50 ng/ml for 48 h before Wxation and immunostaining with anti -SMA antibody as above. The quantitative immunoXuorescence analysis showed that TGF treatment induced an increase in the number of -SMA positive cells in both cultures (Fig. 8a). To verify if the most diVerentiated phenotype induced by TGF treatment could correspond to a greater capacity of the Wbroblasts to stimulate the epithelial cells, HaCaT keratinocytes were serum starved and incubated with the SNs collected from the TGF treated cultures before Wxation and staining with anti-Ki67 or with anti-K1 antibodies. The quantitative analysis revealed that SNs obtained from CHO-AFs stimulated with TGF had a greater capacity to induce HaCaT cell proliferation and diVerentiation than SNs Fig. 4 Reduction of the SN proliferation eVects by inhibition of the KGFR activity or block of the KGF binding to the receptor. HaCaT cells were serum-starved, treated with SNs or with KGF for 48 h at 37°C in presence of the speciWc FGFR tyrosine kinase inhibitor SU5402 (a) or of the neutralizing anti-KGF antibodies (b), Wxed and immunostained with anti-Ki67 polyclonal antibodies. Quantitative immunoXuorescence analysis indicates that inhibition of the KGFR activity as well as block of the KGF binding to the receptor induce a reduction in the percentage of cycling cells in response to the diVerent SNs. The reduction is more evident after treatment with SNs from CHO-AFs, as well as after KGF treatment, respect to SNs from control cells. Nuclei are stained with DAPI. Bar 20 m. c Results reported in graph represent the mean values § SE. Student's t test was performed and signiWcance level has been deWned as described in "Materials and methods": Statistics: (left): ^p = NS versus corresponding untreated; *p < 0.05 versus corresponding untreated; **p < 0.01 versus corresponding untreated; (right): ^ p = NS versus corresponding untreated; *p < 0.01 versus corresponding untreated; **p < 0.05 versus corresponding untreated; ***p < 0.01 versus corresponding untreated 123 isolated from unstimulated CHO-AFs (Fig. 8b). Again, this induction was reduced in the presence of the KGFR inhibitor SU5402, suggesting that the eVects on the keratinocytes would be mostly mediated by the paracrine release of KGF (Fig. 8b). In agreement with this hypothesis, sandwich ELISA assay performed on these SNs demonstrated that the enhancement in proliferation and diVerentiation observed with the SNs from stimulated CHO-AFs is probably mediated by an increased release of KGF in the medium (Fig. 8c).
Altogether, these results suggest that the CHO-AFs are able to sustain the cholesteatoma growth and altered diVerentiation through the secretion of paracrine diVusible factors and that, among them, KGF might represent one of the major eVectors.

Discussion
In the last few years, it has been widely recognized that the dermal Wbroblasts in an inXammatory microenvironment play a crucial role in the deregulation of the keratinocyte growth leading to epidermal hyperproliferation. Due to the diYculties in studying these complex epithelial-mesenchymal interactions in vivo, many eVorts have been devoted to develop in vitro models for the analysis of the molecular mechanisms and autocrine or paracrine loops involved. The cholesteatoma of the middle ear, being characterized by keratinocyte hyperproliferation and chronic inXammation, represents a model pathology to study the pathogenesis of epithelial non-neoplastic diseases, in which activation of the stromal cells sustains the hyperproliferation and metaplastic altered diVerentiation. Similarly to the experimental studies conducted for other pathological conditions, diVerent in vitro models have been developed also for the cholesteatoma disease: however, most of them aimed to culture and analyze cholesteatoma-derived keratinocytes, which appear to retain in vitro the altered behaviour observed in vivo (Cheshire et al. 1995;Albers-op t' Hof et al. 2002;Hilton et al. 2011). The only 3D co-culture model, developed using both primary keratinocytes and Wbroblasts from a b c the cholesteatoma tissue, was useful to investigate the cytokeratin protein proWle and metaplastic behaviour of the lesional epidermal cells and to ascertain the importance of the Wbroblasts in the optimization of the in vitro system (Raynov et al. 2008); however, the study was not attempted to address the potential activation of the stromal components. Therefore, to investigate if also the cholesteatoma-associated Wbroblasts might possess biological characteristics, which distinguish them form the normal quiescent Wbroblasts and which could be retained at early passages in culture, in this study, we focused our attention to these cells, analyzing their secretory behaviour and their degree of diVerentiation or activation. We found that, compared to the perilesional or control normal Wbroblasts, all cultures derived from cholesteatoma samples were less proliferating and more diVerentiated; moreover, although very variable among the cultures, the diVerentiated phenotype well correlated with an activated state, shown by MMP-2 and MMP-9 secretion, demonstrating that also in a non-neoplastic context, such as the cholesteatoma disease, the associated Wbroblasts might exert a pathological role through biological functions which are maintained in culture. Among the growth factors and cytokines which may play a role in the development of the cholesteatoma disease, as well as in other hyperproliferative epithelial pathologies, KGF appears a key molecular eVector (Kojima et al. 1996;Ishibashi et al. 1997;Yamamoto-Fukuda et al. 2003Barbara et al. 2008;d'Alessandro et al. 2010) and a possible target or biomarker for therapeutic intervention (Yamamoto-Fukuda et al. 2008). Many studies have reported that the expression of KGF is up-modulated in Fig. 8 Evaluation of biological eVects induced on HaCaT cells by SNs from CHO-AFs stimulated with TGF . a Two selected cultures of CHO-AFs, #2010 and #2810, were treated with TGF for 48 h at 37°C and then Wxed and stained with anti--SMA. Quantitative immunoXuorescence analysis of the percentage of -SMA positive cells shows that, respect to untreated cells, TGF treatment induces an increase of the -SMA positive cells in both cultures. Nuclei are stained with DAPI Bar 20 m. Results reported in graph represent the mean values § SE. Student's t test was performed and signiWcance level has been deWned as described in "Materials and methods": Statistics: *p < 0.05 versus untreated. b HaCaT cells were serum-starved, treated for 48 h at 37°C with the SN from #2010 and #2810 cultures treated with TGF as above. The treatments were performed also in the presence of the KGFR inhibitor SU5402. The cells were then Wxed and immunostained with anti-Ki67 or with anti-K1 polyclonal antibodies. Quantitative immunoXuorescence analysis of the Ki67 or K1 positive cells shows that the SNs from cultures treated with TGF induce an enhance the proliferation and diVerentiation of HaCaT cells respect to the SNs from untreated cultures. Inhibition of the KGFR activity reduces the eVects of SNs. Nuclei are stained with DAPI. Bars 20 m (left panel) and 50 m (right panel). Results reported in graph represent the mean values § SE. Student's t test was performed and signiWcance level has been deWned as described in materials and methods. Statistics for proliferation analysis: *p < 0.05 versus SNs from unstimulated CHO-AFs, **p < 0.05 versus SNs from unstimulated #2010 and p < 0.01 versus SNs from TGF stimulated #2010, ***p < 0.05 versus SNs from unstimulated and TGF stimulated #2810. Statistics for diVerentiation analysis: * p < 0.01 versus SNs from unstimulated CHO-AFs, **p < 0.05 versus SNs from unstimulated CHO-AFs and p < 0.01 versus SNs from TGFb stimulated CHO-AFs. c KGF-quantitation by ELISA test on SNs from #2010 and #2810 cultures treated with TGF as above. In both cultures the amount of KGF released in the medium is higher in SNs from TGF treated cells respect to the SNs from untreated cells. Results reported in graph represent the mean values § SE. Mann-Withney test was performed and signiWcance level has been deWned as described in "Materials and methods". Statistics: *p < 0.05 and **p < 0.01 versus untreated Fig. 9 Schematic drawing of the proposed role of the cholesteatomaassociated Wbroblasts (CHO-AFs) in the deregulation of epidermal proliferation and diVerentiation through a paracrine loop involving secretion of KGF: in the pathological framework (right panel), the CHO-AFs, activated by growth factors and cytokines (such as TGF ) released by inXammatory cells, produce an increased amount of KGF compared to the physiological conditions (left panel). Through binding to its receptor KGFR, the growth factor increases keratinocyte proliferation and diVerentiation. The activated CHO-AFs release in the stroma also enhanced levels of MMPs which contribute in the remodeling and degradation of the extracellular matrix cholesteatoma, but none of them was aimed to analyze in detail the contribution of the KGF-secreting Wbroblasts to the disease. Here, through the evaluation of the capacity of the culture supernatants to modulate the biological response in human keratinocytes in the presence of the KGFR inhibitor, we demonstrated the single key contribution of the KGF released in the medium in deregulating the proliferative and diVerentiative response. Moreover, the variable amount of KGF detected in the diVerent supernatants well correlated with the quiescent or diVerentiated/activated phenotype of the cultured Wbroblasts; this correlation is further demonstrated by the increase in KGF production observed following the widely used triggering of diVerentiation by TGF treatment. Therefore, as depicted in the schematic drawing in Fig. 9, the activation of the stromal Wbroblasts present in the pathological tissue and the consequent increased secretion of KGF appear to play a crucial role in the deregulation of the epidermal proliferation and diVerentiation which characterizes the cholesteatoma tissue. Interestingly, we observed that the CHO-AFs showing the diVerentiated/activated phenotype were isolated from cholesteatoma tissue samples characterized by an extended dermal inXammatory inWltrate, assessed as described in our previous study (d'Alessandro et al. 2010), suggesting that the variability of the phenotypic changes in the diVerent Wbroblast populations would reXect the level of tissue inXammation. Therefore, we propose that this model might be very useful to evaluate the unknown pathogenetic role of the stroma perimatrix component of the cholesteatoma lesion and might contribute in general for a better understanding of the epithelial-mesenchymal interactions. Moreover, it is possible to assume that this model could be utilized for the in vitro assessment of the eYcacy of therapeutic strategies directed to interfere with the pathological paracrine loops.