Granulocyte macrophage‐colony stimulating factor and keratinocyte growth factor control of early stages of differentiation of oral epithelium

Abstract Oral epithelial differentiation is known to be directed by underlying fibroblasts, but the responsible factor(s) have not been identified. We aimed here to identify fibroblast‐derived factors responsible for oral epithelial differentiation. Primary normal human oral keratinocytes and fibroblasts were isolated from healthy volunteers after informed consent (n = 5) and 3D‐organotypic (3D‐OT) cultures were constructed. Various growth factors were added at a range of 0.1‐100 ng/ml. 3D‐OTs were harvested after ten days and assessed histologically, by immunohistochemistry and the TUNEL method. Epithelium developed in 3D‐OT without fibroblasts showed an undifferentiated phenotype. Addition of granulocyte macrophage‐colony stimulating factor (GM‐CSF) induced expression of cytokeratin 13 in suprabasal cell layers. Admixture of GM‐CSF and keratinocyte growth factor (KGF) induced, in addition, polarization of epidermal growth factor (EGF) receptor and β1‐integrin to basal cell layer and collagen IV deposition. Terminal differentiation with polarization of TUNEL‐positive cells to superficial layers occurred only in the presence of fibroblasts in collagen gels either in direct contact or at distance from normal oral keratinocytes. Taken together, these results show that major aspects of oral epithelial differentiation are regulated by the synergic combination of GM‐CSF and KGF. However, the terminal stage seems to be controlled by other yet unidentified fibroblast‐derived diffusible factor(s).


INTRODUCTION
It is well established that the molecular interactions between epithelium and mesenchyme is essential for keratinocyte proliferation, differentiation and repair in both skin and oral mucosa [1][2][3][4]. Previous studies on three dimensional organotypic (3D-OT) in vitro models, including our own, demonstrated that fibroblasts were essential for the resemblance of the tissues reconstructed 3D in vitro with the in vivo human oral mucosa [1,5,6]. Several studies have tested various fibroblast-derived factors for their role on oral epithelium differentiation; keratinocyte growth factor (KGF) was found to stimulate proliferation of oral keratinocytes but not influence their differentiation when added alone to 3D-OT models constructed with keratinocyte only (3D-OT monocultures) [1]. This was in contrast to the effect of KGF on dermal keratinocytes in 3D-OT cultures [7]. The hematopoietic granulocyte macrophage-colony stimulating factor (GM-CSF) which is synthesized by macrophages, T cells, mast cells, natural killer cells, endothelial cells, and fibroblasts and which normally functions as a cytokine facilitating development of the immune system and promoting the defense against infections was found to also regulate dermal keratinocyte growth and differentiation [8]. Fibroblast-keratinocyte cocultures in fetal skin models strongly enhanced the expression of GM-CSF by fetal skin cells [9], while dermal keratinocytereleased interleukin 1α (Il-1α) induced the expression of both KGF and GM-CSF in dermal fibroblasts [10]. These studies indicated GM-CSF as a growth factor involved in epithelial-mesenchymal interactions, but its effect on oral epithelial morphogenesis has not been tested so far. Much of the knowledge on epithelial-mesenchymal interactions comes from studies on skin models, but there are distinctive signals for epithelial differentiation of oral and dermal fibroblasts [11] and there is a gap of knowledge on how fibroblasts regulate the differentiation of oral epithelium. The aim of this study was to identify the fibroblast-derived factors responsible for oral epithelial differentiation, and for this purpose several growth factors were tested, such as epidermal growth factor (EGF), KGF, GM-CSF, transforming growth factor α (TGFα), IL-1α, hepatocyte growth factor (HGF), and stem cell factor (SCF). The present study presents data in support for the control of oral epithelial differentiation by the underlying mesenchyme via soluble factors synthesized by oral fibroblasts. GM-CSF, alone or in combination with KGF, was able to control several steps of differentiation, except its terminal stages. This indicates that other yet unidentified fibroblast-derived soluble factor(s) may be responsible for regulation of terminal differentiation in oral epithelia.

Human donors
Eighteen samples of normal human oral mucosa were obtained from healthy donors undergoing wisdom tooth extraction (details in Table 1). Seven samples were snapfrozen in isopentane and six samples were formalin-fixed and embedded in paraffin. Cells successfully isolated and propagated from five samples were used for growing of 3D-OT cultures. The study was approved by the Ethics Committee of the University of Bergen (REK 2010/481) and the samples were collected after informed consent.

3D-OT cell culture procedures
Simple collagen gels (700 μl for each culture) were prepared on ice by mixing 7 vol. (3.40 mg/ml) of rat tail collagen type I (Collaborative Biomedical), 2 vol. reconstitution buffer (261 mM NaHCO3, 150 mM NaOH, 200 mM HEPES) pH 8.15, 1 vol. Dulbecco's Modified Eagle Medium (DMEM) 10x (Sigma) and 1 vol. fetal calf serum. Fibroblast-containing collagen matrices were prepared by mixing 1 vol. fetal calf serum containing 0.5 × 10 6 /ml normal oral fibroblasts in passages 2-4. Seven hundred μl of the prepared matrix was pipetted in 24 well plates and let for 30 min in the incubator to turn to gel. Normal oral fibroblasts growth medium (1 ml/well) was then added over the matrices. After 24 h, the medium on top of the gels was removed and normal oral keratinocytes (0.5 × 10 6 cells/culture) at second passage were added in 1 ml of their growth medium [1]. After 24 to 48 h, the cultures were lifted on the air-liquid interface. The flow of procedures for construction of 3D-OT cultures is presented in Figure 1.
The suspended 3D organotypic cultures were grown in serum free medium comprising DMEM and Ham's F-12 nutrient mix in 3:1, supplemented with 1 μM hydrocortisone, 0.8 μM insulin, 0.25 mM transferrin, 0.25 mM L-ascorbic acid, 15-30 μM linoleic acid, 15 μM bovine serum albumin, 2 mM Lglutamine (all from Sigma). Sandwich models were manufactured by interposing a layer of collagen biomatrix (500 μl) between the epithelial compartment and the fibroblast containing matrix. Human growth factors (EGF, KGF, GM-CSF, TGFα, IL-1α, HGF, SCF -Sigma) were added to the culture media of some of the collagen simple matrix cultures at a range of 0.1-100 ng/ml, as summarized in Table 2. All cultures were maintained at 37˚C in 5% CO 2 incubators for the whole duration of the experiment. All cultures were harvested on day 10 of co-culture. One half of each culture was snap frozen in isopentane pre-chilled in liquid nitrogen and the other fixed in 4% buffered formalin pH 7.15 and embedded in paraffin. Experiments (run in duplicates) were repeated 5 times, each time with primary cells isolated from different patients (n = 5 donors).

ELISA
Conditioned media was collected from normal oral fibroblasts cells (n = 5 donors) maintained in monocellular (normal oral fibroblasts only) 3D cultures at similar passages, and analyzed for levels of various growth factors and cytokines by using the Widescreen Human Cancer Panel 2 (Novagen) with Luminex beads (R&D Systems).

F I G U R E 1
Step-by-step laboratory procedures for construction of three dimensional organotypic cultures using primary cells isolated from normal human oral mucosa

TUNEL method
Cell death was detected by the TUNEL method (terminal deoxynucleotidyl transferase-mediated dUTP in situ nick endlabelling) on formalin fixed paraffin embedded sections [12]. For positive controls, specimens were treated with 0.5 mg/ml DNase (Roche Diagnostics) in tris-buffered saline for 15 min at 37˚C prior to incubation with bovine serum albumin. The specificity of the TUNEL reaction was tested by substituting the biotinylated dUTP in the TUNEL labelling mixture with unbiotinylated dUTP (Roche) in excess. TUNEL positive keratinocytes found within the basal cell layer were considered spontaneously apoptotic cells, while TUNEL positive cells found at the superficial cell layer on top of the epithelium were considered terminally differentiated keratinocytes [13].

Evaluation of samples and statistical analysis
ELISA results are presented with values normalized for 10 6 cells; data were analyzed using t-test with a level of significance set at 5% (SPSS 11.0). The data is presented as mean +/-SD. Tissue sections (5μm) from paraffin embedded specimens, stained with Haematoxylin-eosin, were morphometrically analyzed by a computer-based optical image analyzer (analySIS 11.0 Pro Soft Imaging System). Ki-67 /proliferation index was determined as the percentage of positive cells among all cells of the basal cell layer per 400μm length of the epithelial-mesenchymal interface. The measurements and counts were done at 200 fold magnification on a standard microscope (LeikaDMLM) on six consecutive fields situated 200 μm apart. Statistical analysis was performed using Wilcoxon paired test with a level of significance set at 5% (SPSS 11.0).

Effects of KGF and GM-CSF, alone or in combination, on epithelial cell proliferation and thickness of in vitro reconstructed normal human oral epithelium
The oral mucosa formed by growing primary normal human buccal keratinocytes on simple collagen gels in absence of fibroblasts displayed a thin epithelium (Figure 2A) with low cell proliferation ( Figure 3). Presence of fibroblasts in the collagen matrix, either in direct contact with keratinocytes ( Figure 2J) or at distance (in the 'sandwich models'- Figure 2K) induced an increase in cell proliferation ( Figure 3). No differences between cell proliferation indices or the phenotypes of reconstructed oral epithelia could be detected between cultures with direct keratinocytefibroblast contact and cultures with keratinocytes at distance from fibroblasts ( Figure 3). Analysis of conditioned medium from 3D gels populated with fibroblasts showed that fibroblasts secreted HGF, KGF, GM-CSF, and IL-1α when grown in 3D cultures in vitro ( Figure 4). Both KGF and  Other three dimensional models were constructed by seeding normal oral keratinocytes on top of human fibroblast-containing collagen gels (J-L) either in direct contact (J and L) or at distance through a layer of simple collagen layer (sandwich models -K). An antibody against GM-CSF (L) was added to three dimensional cultures with fibroblast-containing collagen type I gels. All cultures were harvested on day 10 of co-culture. One half of each culture was fixed in 4% buffered formalin pH 7.15 and embedded in paraffin. Sections of representative cultures stained with haematoxylin & eosin are shown. Scale bar = 100 μm GM-CSF at concentrations higher than 1 ng/ml, either alone or in combination, increased cell proliferation in the basal cell layer (Figure 3). EGF, TGFα, IL-1α, HGF, or SCF did not alter epithelial thickness ( Figure 2B, E-H), or epithelial cell proliferation in 3D-OT monocultures of keratinocytes ( Figure 4).

Effects of GM-CSF alone or in combination with KGF on oral epithelial differentiation of in vitro reconstituted normal human oral epithelium
When grown in 3D monocultures on simple collagen gels, normal oral keratinocytes formed an epithelium with an undifferentiated phenotype ( Figure 5, Table 2). Immunohistochemistry for various differentiation markers of these cultures revealed a weak, scattered expression of cytokeratin 13 (CK13, Figure 5A), strong expression of β1-integrin ( Figure 5F), and EGF receptor (EGF-R, Figure 5K) throughout all cell layers with no deposition of collagen IV at the epithelium-matrix interface ( Figure 5P). The presence of fibroblasts either in direct contact or at distance from the epithelium promoted formation of a fully maturated human buccal epithelium (Table 2) similar to the in vivo oral mucosa as judged after the panel of differentiation markers used in this study: uniform and strong expression of CK13 throughout all suprabasal epithelial cell layers ( Figure 5D,E), polarization to the basal cell layer of β1 intergin ( Figure 5I,J), and EGF-R ( Figure 5N,O), as well as  Figure 5R). This analysis shows that GM-CSF alone or in combination with KGF was able to induce major aspects of oral epithelial differentiation of in vitro reconstituted normal human oral epithelium. None of the other growth factors tested in the study (EGF, TGFα, IL-1α, HGF, SCF) did influence, when added, the phenotype of the epithelium grown on simple collagen gels.

Effects of KGF and GM-SCF on terminal differentiation of in vitro reconstituted normal human oral epithelium
The 3D monocultures of oral keratinocytes displayed TUNEL positive cells randomly distributed within the epithelium ( Figure 5U). There was no polarization of TUNEL positive cells to the superficial layer, suggesting that cells did not complete the terminal stages of epithelial differentiation in these cultures. Similar pattern of distribution of TUNEL positive cells was also observed in the 3D monocultures of oral keratinocytes supplemented with GM-CSF alone or in  Figure 5V,X respectively). Polarization of TUNEL positive cells to the superficial cell layer was observed only when fibroblasts were present in the connective tissue equivalent, either in direct contact or at distance from the epithelium (Figure 5Y,Z). Addition of an anti-GM-CSF antibody to the culture medium of fibroblast-containing cultures did not impair cell growth or the terminal differentiation of the reconstituted oral epithelium ( Figure 2L). None of the other growth factors tested in the study (EGF, TGFα, IL-1α, HGF, SCF) did influence, when added, the distribution of TUNEL positive cells within the epithelium grown on simple collagen gels. This analysis shows that terminal differentiation of in vitro reconstituted normal human oral epithelium was induced by underlying fibroblasts through diffusible factors, but not by the combination of KGF and GM-CSF.

DISCUSSION
Formation and maintenance of mature oral epithelium rely on a tightly balanced process of keratinocyte proliferation and terminal differentiation [14], but the knowledge about the specific factors involved is limited. Previously, we have developed a highly standardised serum free organotypic 3D-OT model of human oral mucosa [15] and showed that fibroblasts are essential for differentiation of oral epithelium [1]. Data presented here further demonstrate that fibroblast-derived diffusible factors are able to fully restore the differentiated phenotype of in vitro oral epithelium, including the fine-tuned terminal stage of epithelial differentiation. From all the growth factors tested in the present study, alone or in various combinations, only GM-CSF, alone or in combination with KGF, had a significant effect on the phenotype of oral epithelium. Previous reports from similar '3D organotypic' models of skin morphogenesis and homeostasis [3,16,17] identified also GM-CSF, alone or in combination with KGF, as a factor that induces a significant effect on the phenotype of epithelium. Of note, the skin 3D organotypic cultures supplemented with KGF only, displayed delays in expression of differentiation markers [18,19]. Other reports showed that dermal keratinocytes treated with KGF exhibited increased proliferation as well, and inhibited differentiation, while reduced KGF levels restored the expression of differentiation markers [2]. One possible explanation is that the secretion of high doses of KGF by fibroblasts might influence the choice between proliferation and differentiation [19]. A limitation of the present study is that we did not test the effect of neutralizing antibodies for KGF, such that we could not infer more on the importance of this growth factor for the fine tuning of differentiation in oral epithelium.
The hematopoietic growth factor GM-CSF is another growth factor that was found to regulate dermal keratinocyte growth and differentiation [8], playing an important role during the process of wound healing [20]. Fibroblastkeratinocyte interactions in skin models strongly enhanced the expression of GM-CSF [9], KGF and its receptor [21], while dermal keratinocyte-released IL-1 induced the expression of both KGF and GM-CSF [10]. The regulatory mechanism of these two factors in skin homeostasis is a feedback loop within the multiple other epithelialmesenchymal interactions: skin keratinocytes release IL-1α and IL-1β, which stimulate the release of KGF and GM-CSF by dermal fibroblasts. Then in turn, these two growth factors synthesized by dermal fibroblasts act on skin keratinocytes regulating their differentiation and proliferation [3,7,10].
In contrast to these reported observations from skin models proving that a combination if KGF and GM-CSF can substitute for the dermal fibroblasts and provide sufficient support for both growth and differentiation of skin keratinocyte in their absence [3,16,17], our current study shows that the final stages of oral epithelial maturation could not be restored by KGF and GM-CSF only. These differences between skin and oral mucosa morphogenesis might be due to the fact that oral mucosal fibroblasts and adult skin fibroblasts have different origin (the former originates from the neural crest and the latter from the mesoderm) and gene expressions, and consequently, different phenotypes and functions [22]. Oral fibroblasts were proven to express higher levels of KGF and to accelerate much faster the collagen gel contraction than dermal stromal cells [23,24]. The results presented here, based on analysis of conditioned medium collected from oral fibroblasts maintained in 3D monocellular cultures containing fibroblasts collagen matrices, show that oral fibroblasts synthesise considerable amounts of KGF, GM-CSF, and HGF. These results corroborate with previous literature showing the oral fibroblasts to be the major producers of these growth factors [24][25][26], although oral keratinocytes have also been proven to synthesise GM-CSF [27].
That the mesenchymal cell source has a significant influence on the thickness and ultrastructure of the epithelium has been previously shown [28]. Moreover, cytokeratin expression of the epithelial component was also proven to be strongly influenced by the origin of fibroblasts [29].
The data presented in this study indicate that in contrast to skin, other soluble factors than KGF and GM-CSF released from fibroblasts exert the final tuning of oral epithelial differentiation. In support for this conclusion comes also the observation that addition of neutralizing antibodies against human GM-CSF, previously shown to reduced keratinocyte proliferation and differentiation in skin models [7], did not impair cell proliferation or differentiation in our oral mucosa models. Taken together, the results of this study indicate that major aspects of oral epithelial differentiation are regulated by GM-CSF in combination with KGF, but its terminal stage is controlled by another yet unidentified fibroblast-derived diffusible factor.

A C K N O W L E D G E M E N T S
The authors thank Mrs. Gunnvor Øijordsbakken (Gade Laboratory for Pathology) for support with histology and immunohistochemistry.