The human keratinocyte cell line HaCaT expresses essentially all epidermal differentiation markers but exhibits deficiencies in tissue organization as surface transplants in nude mice and even more so in organotypic co-cultures with fibroblasts. Whereas tissue differentiation by normal keratinocytes(NEKs) is regulated by stromal interactions, this mechanism is impaired in HaCaT cells. This regulatory process is initiated by interleukin-1 (IL-1)release in keratinocytes, which induces expression of keratinocyte growth factor (KGF/FGF-7) and granulocyte macrophage-colony stimulating factor(GM-CSF) in fibroblasts. Production and release of IL-1 is very low and,consequently, expression of the fibroblast-derived growth factors KGF/FGF-7 and GM-CSF is absent in HaCaT-fibroblast co-cultures. However, addition of KGF and GMCSF, respectively, is inefficient to improve stratification and differentiation by HaCaT cells due to the low expression of their cognate receptors. More importantly, expression and release of the autocrine keratinocyte growth factor TGF-α is dramatically decreased in HaCaT cells. Addition of TGF- α or EGF stimulated HaCaT cell proliferation but, even more effectively, suppressed apoptosis, thus facilitating the formation of a regularly stratified epithelium. Furthermore, TGF-αenhanced the expression of the receptors for KGF and GM-CSF so that addition of these growth factors, or of their inducer IL-1, further improved epidermal tissue differentiation leading to in vitro skin equivalents comparable with cultures of NEKs. Thus, supplementing TGF-α normalized epidermal tissue regeneration by immortal HaCaT keratinocytes and their interaction with stromal cells so that regular skin equivalents are produced as standardized in vitro models.
Regeneration and maintenance of epithelial tissue homeostasis requires a complex interplay with neighbouring cells and extracellular matrix of the adjacent stroma. This is well understood in skin based on data from wound healing, transplantation and cell culture studies, which strongly indicate that epidermal tissue regeneration is regulated by a network of cytokines and growth factors controlling functional behaviour of keratinocytes and dermal fibroblasts (Luger and Schwarz,1990; Fusenig,1994; Mackenzie et al.,1993; Werner et al.,1994; Breitkreutz et al.,1997; Boukamp et al.,1990; Stark et al.,2001). Comparably, keratinocyte proliferation in conventional cell culture is strongly enhanced by stromal cell interactions, a process routinely used in 2D feeder layer co-cultures with postmitotic fibroblasts(Rheinwald and Green, 1975; Limat et al., 1989). Accumulating data strongly indicate that this stromal cell support of epithelial cell proliferation is, for the most part, mediated by diffusible factors (Waelti et al., 1992; Smola et al., 1993; Maas-Szabowski et al., 1999). Importantly, the support of keratinocyte proliferation is not based on their passive utilization of growth factors constitutively produced by fibroblasts but represents an active interplay between both cell types in a double paracrine mechanism. Hereby, keratinocytes actively control the production of their growth factors in fibroblasts by release of interleukin-1 (IL-1) αand β, which induce the enhanced expression of growth factors in stromal cells such as keratinocyte growth factor/fibroblast growth factor 7(KGF/FGF-7) and granulocyte-macrophage colony-stimulating factor (GM-CSF). Both factors strongly stimulate keratinocyte proliferation in vitro and are upregulated in wound healing(Maas-Szabowski et al., 2001; Werner, 1998; Mann et al., 2001; Groves and Schmidt-Lucke,2000).
These regulatory mechanisms are also operative in the more in-vivo-like organotypic co-culture system representing an in vitro skin equivalent model consisting of a differentiated stratified epithelium on top of a dermal equivalent formed by fibroblasts in a collagen type I gel (for reviews, see Fusenig, 1994; Maas-Szabowski et al., 2002; Bell et al., 1981). Thus, in vitro regeneration of a structured epidermis is regulated by the same double paracrine mechanisms involving keratinocyte-released IL-1 as inducer, and KGF/FGF-7 as well as GMCSF as fibroblast-produced effector molecules to stimulate keratinocyte growth and differentiation(Maas-Szabowski et al., 2000; Maas-Szabowski et al., 2001; Szabowski et al., 2000). Obviously, these signaling molecules are only part of a more complex regulatory mechanism of epithelial-stromal interactions controlling tissue regeneration and homeostasis. In the absence of stromal cells both KGF and GM-CSF are unable to fully support epidermal reconstitution in organotypic co-cultures (Maas-Szabowski et al.,2001).
The study of the identity and mechanism of action of further players in this mechanism is often complicated by inter-individual variations of early passage keratinocytes as well as of fibroblasts when comparing cell strains isolated from different donors (Watt et al., 1987; Boukamp et al.,1990). Whether these differences are due to variations in the isolation technique, to variable contaminations of fibroblasts in the keratinocyte populations, varying composition of the dermis-derived cell populations or a combination of all three determinants, is difficult to discriminate when using cultures of early passaged cells. These variables may also obscur donor-age or body-site related differences in the regenerative capacity of keratinocytes and the supportive behaviour of fibroblasts. By contrast, keratinocytes of passages higher than three often exhibit reduced regenerative capacities (Gilchrest et al.,1983; Stanulis-Praeger and Gilchrest, 1986).
Thus, for studying in detail the complex basic mechanisms of epidermal-dermal cell interactions, a standardized and reproducible in vitro skin equivalent model would be advantageous. Obviously, established cell lines with maintained functional capacities would be the most suitable candidates to replace freshly isolated cells. Established cell lines, however, generally exhibit altered functional properties, although fibroblast cell lines, such as mouse 3T3 or comparably immortalized embryonic cell lines, seem to function similarly to adult primary fibroblasts when combined with normal human keratinocytes in organotypic co-culture(Kaur and Carter, 1992; Choi and Fuchs, 1994; Szabowski et al., 2000; Maas-Szabowski et al., 2001). Although these studies have demonstrated functioning of the epithelial-fibroblast interactions across species barriers, excessive growth of established fibroblasts in the collagen gel and/or other functional variations may lead to enhanced and atypical growth behaviour of keratinocytes(Kaur and Carter, 1992; Choi and Fuchs, 1994). Varying proliferative activities of fibroblasts in the collagen gel with the consequence of variable stromal cell numbers are another disturbing problem. However, this could, for the most part, be eliminated by using permanently postmitotic cells following γ-irradiation comparable with feeder layer fibroblasts. These irradiated fibroblasts maintained fairly constant cell numbers and functioned as well as proliferating cells to support keratinocyte growth and differentiation into 2D and 3D co-cultures(Maas-Szabowski and Fusenig,1996; Maas-Szabowski et al.,2000).
However, the use of human keratinocyte cell lines in organotypic cultures was less successful. They have been established mostly by immortalization with viral oncogenes and exhibited minor or major variations in their differentiation capacity (Blanton et al.,1991; Durst et al.,1989; Oda et al.,1996; Tsunenanga et al.,1994; Lechner and Laimins,1991). Otherwise, spontaneously immortalized human keratinocyte lines generally have maintained a rather high degree of differentiation potential (Boukamp et al.,1988; Baden et al.,1987; Allen-Hoffmann et al.,2000; Rice et al.,1993). This is best demonstrated with the HaCaT cell line, a non-tumorigenic keratinocyte population derived from adult trunk skin exhibiting a rather normal differentiation capacity despite multiple chromosomal alterations (Boukamp et al.,1988; Boukamp et al.,1997; Breitkreutz et al.,1997; Breitkreutz et al.,1998; Ryle et al.,1989). However, organotypic co-cultures of HaCaT cells with skin fibroblasts exhibit some stratification but usually lack typical criteria of an ordered structure and regular keratinization(Haake and Polakowska, 1993; Syrjänen et al., 1996; Steinsträsser et al.,1997; Boelsma et al.,1999). This deficiency, however, was not due to the permanent loss of essential differentiation functions. We have recently demonstrated that HaCaT cells under optimal environmental conditions (i.e. in surface transplants on nude mice) were able to reform a regularly structured differentiated epidermis, although with some delay and minor deficiencies compared with normal keratinocytes(Breitkreutz et al., 1997; Breitkreutz et al., 1998). Furthermore, we documented that improved organotypic co-culture conditions with increased numbers of supporting fibroblasts in the collagen gel,significantly enhanced stratification and differentiation of HaCaT epithelia(Schoop et al., 1999). Nevertheless, growth and differentiation of the squamous epithelia was delayed and incomplete, suggesting deficient signal transduction.
Here, we demonstrate that HaCaT cells exhibit distinct functional deficiencies that substantially reduce their interaction with stromal cells as well as their response to fibroblast-produced growth factors. The expression and release of the inducer molecule IL-1 is very low in HaCaT cells, with the consequence of a minimal induction of KGF and GM-CSF in fibroblasts. More importantly, expression of the receptors for both growth factors is strongly decreased on HaCaT cells so that signal transduction is significantly impaired. Finally, the level of the autocrine acting keratinocyte growth factor TGF-α is drastically reduced in HaCaT cells. By addition of TGF-α deficiencies in growth and differentiation of HaCaT organotypic cultures can be completely rescued. This is mediated by the direct stimulatory effect of TGF-α on HaCaT cell proliferation and survival, and, moreover,by the enhanced expression of IL-1 and the receptors of KGF and GM-CSF resulting in normalized interaction with stromal cells. Thus, following supplementation of one growth factor (TGF-α) skin equivalents with rather normal epidermal structures could be reproducibly obtained to serve as a standardized skin equivalent model system.
Materials and Methods
Normal human skin keratinocytes (NEKs) and dermal fibroblasts (HDFs) were derived from adult skin obtained from surgery, as previously described(Smola et al., 1993; Stark et al., 1999). HaCaT cells (passages 35-40) (Boukamp et al.,1988) and fibroblasts (passages 5 to 9) were grown in DMEM (Bio Whittaker, Serva, Heidelberg, Germany) supplemented with 10% FCS. For preparing fibroblast feeder cells (HDFis), trypsinized fibroblast suspensions(0.05% trypsin/0.025% EDTA, v/v) were γ-irradiated with 70 Gray. NEKs were expanded on X-irradiated feeder cells (5000 cells/cm2) in FAD medium (DMEM:Hams F12, 3:1) with 100 U/ml penicillin, 50 μg/ml streptomycin and supplemented with 5% FCS, 5 μg/ml insulin, 1 ng/ml recombinant human EGF, 10-10 M cholera toxin, 10-4 M adenine and 0.4μg/ml hydrocortisone (Sigma, Deisenhofen, Germany), as described(Smola et al., 1993); for RNA expression studies EGF was omitted. HaCaT cells were plated at 6000 cells/cm2 in DMEM with 10% FCS as monocultures; in 2D co-cultures 5000 cells/cm2 were plated on feeder cells (5000/cm2) in DMEM with 10% FCS. NEKs were grown at 7000 cells/cm2 on feeder cells (5000/cm2) in FAD medium without EGF. To determine the proliferation response to growth factors, HaCaT monocultures were grown for 24 hours in the respective serum-containing medium. Thereafter, medium was replaced by serum-free DMEM additionally containing one of the following additives: 10 ng/ml KGF (BTS, St Leon, Germany), 2 ng/ml EGF (Sigma,Deisenhofen, Germany), 2 ng/ml TGF-α, 100 ng/ml GM-CSF, 5 ng/ml IL-1α (all R&D Systems, Wiesbaden, Germany), and 5% FCS,respectively, as control. For cell number based calculations of cytokine concentration in supernatants, cells were trypsinized and counted at day 3, 5 and 7. Data are given as means ± standard deviation of duplicate measurements performed in 2-3 independent experiments.
Isolation of RNA, reverse transcription and PCR
Mono- and co-cultures of HDFs, NEKs and HaCaT cells were grown in the respective medium to subconfluence. Cells were lysed in guanidinium-isothiocyanate solution and total RNA was extracted(Chomczynski and Sacchi,1987). Furthermore, RNA was extracted from subconfluent HaCaT monolayer cultures, as controls or treated for 10 hours and 24 hours with 2 ng/ml TGF-α (in serum-free medium). RT-PCR was performed according to an established method with some modifications, as reported(Maas-Szabowski and Fusenig,1996). In brief, cDNA was transcribed at 42°C for 60 minutes in 50 μl final volume containing 5 μg total RNA, 5 μl 10×PCR-buffer, 10 μl MgCl2 (25 mM), 3 μl of each dNTP (10 mM),0.5 μl RNasin (0.5 U/μl), 0.5 μl reverse transcriptase (50 U/μl),1 μl oligo dT15 (50 μM) and 1 μl random hexamers (50 μM)(GeneAmp-PCR-Kit, Perkin Elmer, Weiterstadt, Germany). Four microliters of first-strand cDNA were added to the PCR-mixture up to a volume of 50 μl following the product description. The mixture was amplified in a thermal cycler (Biometra, Göttingen, Germany) at the indicated annealing temperature performing 24-30 cycles, such that the product yield was in the exponential range. Twenty-mer primers from separate exons flanking the regions of the following locations on cDNA were chosen: GAPDH 69-308 (62°C);IL-1α 84-504 (62°C); IL-1β (174-564; 64°C); TGF-α(240-448; 64°C); EGF (3800-4393; 63°C); KGF (208-474; 60°C); KGF-R(870-1022; 64°C); GM-CSF (66-388; 60°C); GM-CSF-Rα (83-524;56.5°C); and GM-CSF-Rβ (981-1318; 60.5°C). PCR fragments were separated on 1.5% agarose gels (Seakem; Biozym, Oldendorf, Germany),ethidiumbromide-stained, and identified by their running positions on the gel and by restriction mapping with two different enzymes. The mRNA amount of the house-keeping gene GAPDH was used as internal standard.
Protein determination by ELISA
Selected cytokines were quantified by enzyme-linked immunosorbent assays(ELISA) in aliquots of culture medium of feeder-layer co-cultures. Medium was collected at day 3, 5 and 7, always 48 hours after medium change. ELISA kits for IL-1 were purchased from Endogen (Eching, Germany), for TGF-α from Calbiochem (Bad Soden, Germany), and for EGF, KGF and GM-CSF from R&D Systems (Wiesbaden, Germany). Protein values were calculted as pg/105 cells and mean values ± standard deviation of data derived from duplicate measurements from 2-3 independent experiments are given.
Organotypic co-cultures (OTCs)
NEKs (passage 2) and HaCaT cells (passage 35-40) were seeded(1×106/insert) onto collagen type I gels (rat tail tendon)containing 3×105/ml postmitotic fibroblasts (HDFi) cast in cell culture filter inserts (pore size 3.0 μm, polycarbonat; Falcon, Becton Dickinson, Heidelberg, Germany) as described in detail previously(Stark et al., 1999). In HaCaT cultures medium was replaced after 24 hours by DME medium [10% FCS and 50μg/ml L-ascorbic acid (Sigma)] and cultures were raised to the air-liquid interface by lowering the upper medium level to the lower part of the collagen gel. Medium was replaced every 2 days with or without any of the following additives: 10 ng/ml KGF (BTS), 2 ng/ml EGF (Sigma), 2 ng/ml TGF-α, 100 ng/ml GM-CSF, 5 ng/ml IL-1α (all R&D Systems) and 1 μg/ml mouse monoclonal anti-human EGF-receptor antibody (C225, Imclone, New York, NY). NEK OTCs were cultivated in rFAD medium (FAD with 10% FCS, 5 μg/ml insulin, 0.4μg/ml hydrocortisone, 50 μg/ml L-ascorbic acid) as previously described(Stark et al., 1999). Before fixation organotypic cultures were incubated for 16 hours with 63 μM bromodeoxyuridine (BrdU) to label proliferating cells. Cultures were either fixed according to a standardized protocol in 3.7% phosphate-buffered formaldehyde for routine histology and staining in hematoxylin and eosin (H and E), or embedded in Tissue Tec OCT Compound (Medim, Gieβen, Germany)and frozen in liquid nitrogen vapor for cryosectioning.
Indirect immunofluorescence microscopy
Cryosections mounted on glass slides (Histobond, Medim, Gieβen,Germany) were fixed for 5 minutes in 80% methanol at 4°C followed by 2 minutes in acetone at -20°C, rehydrated in PBS, and blocked for 15 minutes in PBS with 1% BSA. First, antibodies were incubated overnight at 4°C in a moist chamber; after three washes in PBS the sections were incubated for 1 hour at room temperature with species-specific, fluorochrome-conjugated secondary antibodies (Dianova, Hamburg, Germany), as well as 0.5 μg/ml bisbenzimide (Hoechst No. 33258) DNA dye for nuclear counter staining. After three final washes with PBS, specimens were mounted with Mowiol (Medim,Gieβen, Germany) under a coverslip and examined and photographed using a Leica microscope (Leitz DMRBE, Bensheim, Germany) equipped with epifluorescence optics. To visualize proliferating cells, a monoclonal mouse anti-BrdU antibody (Progen, Heidelberg, Germany) was used and proliferation was quantitated by counting the ratio of labelled to total nuclei within the basal keratinocyte layer of the epithelium.
Apoptotic cells were detected by TUNEL staining. Paraffin sections mounted on glass slides (Histobond, Medim) were dewaxed in xylol and rehydrated in a graded series of ethanol. Sections were incubated with proteinase K as recommended in the protocol (In situ cell death detection kit TMR red; Roche,Mannheim, Germany). TUNEL reaction mixture was applied for 45 minutes at 37°C. Nuclei were counterstained for 15 minutes at RT by 0.5 μg/ml bisbenzimide (Hoechst No. 33258) DNA dye. Sections were washed twice in PBS and mounted with mowiol. Specimens were examined and photographed using a microscope equipped with epifluorescence optics. The percentage of TUNEL-positive nuclei to the total cell number within all layers of the epithelium was calculated.
Differentiated keratinocytes were characterized by labeling with specific antibodies to human transglutaminase and filaggrin (both mouse monoclonal,Cell Systems, St Katharinen, Germany), EGF-receptor (mAb425, mouse monoclonal,kind gift of U. Rodeck, Philadelphia, PA) as well as loricrin (rabbit polyclonal, kind gift of D. Hohl, Lausanne, Switzerland) and visualized by appropriate fluorochrome-coupled secondary antibodies.
Delayed tissue regeneration of HaCaT cells
In organotypic co-cultures with dermal fibroblasts embedded in a collagen type I gel, HaCaT keratinocytes show drastically reduced growth and differentiation potential compared with epidermal keratinocytes (NEKs)(Fig. 1). This deficiency is partially compensated by augmenting fibroblast numbers in the collagen matrix(Schoop et al., 1999)although, with proliferating cells, their actual numbers at distinct time points are difficult to control (Coulomb et al., 1989). Postmitotic (X-irradiated) fibroblasts, commonly used as feeder cells for keratinocytes(Rheinwald and Green, 1975),also function in organotypic co-cultures and maintain a stable number for at least 2 weeks (Maas-Szabowski et al.,2000). Thus, the fibroblast number required for optimal NEK and HaCaT cell growth and tissue restoration has been more accurately determined as 6×105 irradiated cells/ml collagen gel for HaCaT cell growth and 3×105 cells for NEKs [data not shown(Schoop et al., 1999; Maas-Szabowski et al., 2000)]. However, epidermal tissue formation by HaCaT cells was significantly delayed for 2 weeks and, even after 3 weeks, was incomplete compared with organotypic cultures of NEKs.
To study the molecular defects of HaCaT cells in their interaction with stromal cells, the fibroblast number was adjusted to be optimal for NEKs but suboptimal for HaCaT cells. Thus, with 2×105 irradiated fibroblasts per ml collagen gel, normal keratinocytes derived from adult skin grow to a typically structured epidermal tissue within 1 week, followed by increased differentiation and formation of a structured basement membrane after 2 and 3 weeks [Fig. 1(see also Stark et al.,1999)]. In contrast, HaCaT cells had formed only a 2-3 layered disorganized epithelium under the same conditions within 1 week, which increased in thickness after 2 weeks, but did not exhibit morphological features of differentiation. Thickness and differentiation of the HaCaT epithelia increased further after 3-4 weeks exhibiting a thin parakeratotic cell layer, but HaCaT epithelia were still less keratinized than NEK cultures after 2 weeks [data not shown (Schoop et al., 1999)].
Lack of IL-1 impedes stromal interaction
This discrepancy in epidermal tissue regeneration of HaCaT cells, compared with that of NEKs, was possibly due to deficiencies in epithelial-stromal interaction mechanisms [Fig. 2A(Maas-Szabowski et al.,2001)]. This double paracrine epithelial-stromal interaction is induced by keratinocyte-released IL-1α and -1β resulting in AP1-mediated enhanced expression of KGF and GM-CSF, two strong stimulators of keratinocyte proliferation (Szabowski et al., 2000). To identify any possible alterations in this interaction in HaCaT cells compared with that in NEKs, expression of growth factors was analyzed at the RNA- and protein level in mono- and co-cultures. Thus, RNA levels were determined in both 2D and 3D cultures with similar expression levels (Maas-Szabowski et al.,1999; Maas-Szabowski et al.,2000), whereas protein concentrations were quantitated in supernatants of 2D cultures (to avoid absorption to the collagen gel).
The expression of the signalling factors IL-1α and IL-1β,essential for the double paracrine mechanisms, was drastically decreased in HaCaT mono- and co-cultures when compared with that in NEKs(Fig. 2B). Comparably, the secretion of IL-1α into the culture supernatants remained extremely low over a 5-day-culture period and rose only slightly thereafter(Fig. 2C). Moreover, in contrast to NEKs, HaCaT cells were unable to store IL-1αintracellularly, since HaCaT cell lysates of 5-day cultures showed only minimal amounts (1.98±0.88 pg/105 cells) compared with a massive accumulation in NEKs (281±24.9 pg/105 cells). Similar protein values were measured in the culture supernatants for IL-1β (data not shown). As expected, and due to the lack of both inducer cytokines, RNA expression (data not shown) and protein secretion of KGF and GM-CSF by co-cultured fibroblasts were strongly reduced throughout the first week in culture in contrast to co-cultures with NEK[Fig. 3A(Maas-Szabowski et al., 2000; Szabowski et al., 2000)].
To compensate this deficiency of HaCaT cells in inducing their own growth factors in co-cultured fibroblasts, the addition of these factors or their inducer should restore their capacity to form differentiated epithelia. However, neither the continuous addition of IL-1α, KGF, GM-CSF or their combination (not shown) significantly induced HaCaT cell growth or tissue formation in organotypic co-cultures (Fig. 3B). This lack of response was most probably due to defects in signal reception and transduction in HaCaT cells, as shown by the reduced expression of the receptors for KGF and GM-CSF(Fig. 4A), and paralleled at the protein level by immunofluorescent staining with specific antibodies (data not shown). The expression of both receptors increased with culture time and cell density: a possible explanation for the delayed stratification of organotypic HaCaT epithelia at 3 and 4 weeks(Schoop et al., 1999).
Abrogated TGF-α expression impairs HaCaT cell proliferation and survial
More importantly, the expression of the transforming growth factor α(TGF-α), a known autocrine acting keratinocyte growth factor, was strongly reduced in HaCaT epithelia in contrast to that in NEKs(Fig. 4A). The reduced TGF-α production was more dramatically evident at the protein level in supernatants of HaCaT mono- or co-cultures(Fig. 4B). This was not due to a lack of cleavage of this factor from its membrane-bound form(Bosenberg et al., 1992), since HaCaT cell lysates showed similar low levels in the ELISA assay (data not shown).
Supplementation of TGF-α rapidly and efficiently restored the capacity of HaCaT cells to form structured epithelia in organotypic co-culture(Fig. 5A). The same normalizing effect was seen with EGF (data not shown), sharing the same receptor with TGF-α, the EGF receptor, which is constitutively expressed in HaCaT cells (Game et al., 1992; Stoll et al., 1998). The essential role of the EGF-receptor-mediated signal transduction for epidermal tissue formation was also demonstrated in organotypic cultures of TGF-α/EGF-substituted HaCaT cells as well as those of NEKs, by receptor blocking studies. Addition of the EGFR-blocking monoclonal antibody C225 drastically inhibited epithelial tissue formation by both cell types(Fig. 5A). EGF itself, although expressed at the RNA level in NEK, HaCaT and fibroblast culture(Fig. 5B), was not detected in the culture supernatants of any mono- or co-cultures (data not shown). Thus,TGF-α is the effective growth factor in skin cells acting in an autocrine manner on keratinocytes (Schulz et al., 1991; Carpenter,1993; Compton et al.,1995).
TGF-α enhances HaCaT cell proliferation in mono-culture with a sixfold higher cell number after 5-day treatment in serum-free medium and in 2D co-cultures (data not shown). Furthermore, TGF-α treatment strongly stimulates stratification in HaCaT mono-cultures on fibroblast-free collagen gels, indicating its major effect as a direct acting factor on HaCaT cells. Moreover, TGF-α induces the expression of IL-1α and, more importantly, the receptors of KGF and GM-CSF(Fig. 5B). The elevated RNA levels are not due to stabilization of mRNA, because the increase is abrogated by actinomycin D treatment. The upregulation of receptor expression is most probably indirect, because it was blocked by the addition of cycloheximide(data not shown). Although there is a basal expression of both receptors at the RNA level, the immunofluorescence data exhibit only background staining in control co-cultures (data not shown).
Based on the observed stimulation of HaCaT cell proliferation, we hypothesized that the effect of TGF-α on tissue reconstruction was mainly due to this effect. However, in organotypic co-cultures, TGF-αenhanced HaCaT cell proliferation only initially (at day 3), whereas at later culture timepoints BrdU-labelled cells were even more frequent in thin control epithelia than in TGF-α-treated multilayered cultures(Fig. 6A).
The explanation for this discrepancy was found when cultures were analyzed for the presence of apoptotic cells (Fig. 6). During the first week, the rate of TUNEL-positive cells was four- to five-times higher in control cultures compared with that in TGF-α-treated cultures (Fig. 6C). When differentiation began (day 9), the rate of apoptotic cells increased in TGF-α-treated cultures and after 12 days was higher than in controls. Whereas TUNEL-positive nuclei were seen throughout the whole epithelia in control cultures, in TGF-α-treated cultures they were exclusively localized in the upper flattened cell layers representing terminally differentiating cells (Fig. 6B). Nearly all nuclei of the parakeratotic superficial cell layers stained positively in the TUNEL assay, whereas the nuclei of the cuboidal cells of the 4 to 6 lower cell layers were unstained.
TGF-α normalizes HaCaT cell growth and differentiation
Replenishment of TGF-α in the deficient HaCaT organotypic cultures not only enabled these immortal keratinocytes to reform a structured squamous epithelium but also allowed a rather normal differentiation(Fig. 7). This is clearly visible in H-and-E-stained sections of 2-week-old cultures, and sections stained with antibodies to markers of early [e.g. keratin 1 and 10 (not shown)and transglutaminase] or later stages of keratinization, such as loricrin. The components are typically localized in the upper layers of the stratified epithelium comparable with, though still less regular than, those in cultures of NEKs. Furthermore, the stratum granulosum formation is rather poor, the stratum corneum is thin and mostly parakeratotic, as indicated by the remnant nuclei in the uppermost flattened cells.
This deficit in differentiation of HaCaT epithelia, however, is repaired to the most part by further supplementing the culture media with KGF, GM-CSF and IL-1α, respectively (Fig. 8). This resulted not only in improved morphologic organization of the reconstituted stratified epithelia but also in a more regular localization of the late differentiation products in the uppermost cell layers very similar to NEKs. Remarkably, as already seen earlier(Szabowski et al., 2000),GM-CSF not only exerts effects on keratinocyte proliferation but also enhances keratinization, demonstrated by the intensified staining with the late differentiation marker loricrin. This is similarly seen after addition of IL-1, which induces GM-CSF in fibroblasts(Maas-Szabowski et al.,2001).
Thus, by compensating for the missing expression of a single autocrine acting growth and survival factor, i.e. TGF-α, the delayed and deficient growth and differentiation capacity of the immortal HaCaT keratinocytes is restored. TGF-α induces proliferation with enhanced survival as well as the expression of IL-1 and the receptors of KGF and GM-CSF so that the HaCaT cells are able to respond typically to stromal interactions regulating keratinocyte growth and differentiation. This skin equivalent reconstructed by the immortal HaCaT keratinocytes represents a better standardized in vitro model to study further regulatory mechanisms in skin physiology and may serve as a highly reproducible test system for pharmakotoxicology.
The immortalized human keratinocyte line HaCaT exhibits distinct genetic features of cell transformation and represents an early stage in the skin carcinogenesis process (Boukamp et al.,1988; Boukamp et al.,1997; Fusenig and Boukamp,1998). However, under in vivo conditions, HaCaT cells are nontumorigenic but still respond typically to environmental control mechanims by reconstituting a rather normal stratified epithelium in surface transplants on nude mice (Breitkreutz et al.,1998). Although epidermal tissue reconstitution in vivo was delayed and showed some deficiencies, these data clearly demonstrated that HaCaT cells, although carrying severe chromosomal abnormalities, had not permanently lost their differentiation capacities. Owing to the well-maintained differentiation properties, also shown by the expression of many different keratins as well as other biochemical markers of differentiation in monolayer cultures(Ryle et al., 1989; Breitkreutz et al., 1997),HaCaT cells have become a paradigm for skin keratinocyte cultures.
In organotypic co-cultures with fibroblasts, conditions that induce normal keratinocytes to form stratified epithelia with all features of a normal epidermis (Smola et al., 1998; Stark et al., 1999), HaCaT cells have been described to be deficient in this complex function of tissue regeneration (Haake and Polakowska,1993; Syrjänen et al.,1996; Steinsträsser et al., 1997; Boelsma et al.,1999). We have been able to improve their epidermal tissue formation and differentiation in this model by intensifying stromal influences by elevating the number of supporting fibroblasts(Schoop et al., 1999). However, tissue architecture and keratinization was delayed and stayed deficient, indicating major defects in transduction of the major signals mediating epithelial-stromal interplay.
Here we demonstrate that the deficiency of HaCaT cell interaction with fibroblasts is based on their very low constitutive expression of the prime signalling cytokine interleukin-1 (IL-1) by which normal epidermal keratinocytes (NEKs) induce expression of their growth factors in fibroblasts(Maas-Szabowski et al., 2000). Moreover, the response of HaCaT cells to these stromal-cell produced factors KGF and GM-CSF is abolished due to the low level of expression of the respective receptors. More importantly, production of transforming growth factor α (TGF-α), known as a potent autocrine acting stimulator of keratinocyte proliferation, is barely detectable in HaCaT cells. Since its cognate receptor, the EGF receptor, is expressed, addition of TGF-α or EGF rapidly re-established the capacity of HaCaT cells to grow to a stratified epithelium and to exhibit typical differentiation markers of the epidermis. Furthermore, with substitution of TGF-α, the effect of stromal cell-derived factors, in particular of GM-CSF, on keratinocyte differentiation is restored, resulting in improved differentiation in HaCaT organotypic cultures comparable with skin equivalents of NEKs(Maas-Szabowski et al.,2001).
This remarkable effect of one single growth factor on the complex mechanism of tissue regeneration was mediated in HaCaT cells by a coordinated function:TGF-α-upregulated the expression of IL-1, and receptors for KGF and GM-CSF; it exhibited an autocrine stimulation of cell proliferation; and, last but not least, it suppressed cell apoptosis. As shown earlier(Maas-Szabowski et al., 1999; Maas-Szabowski et al., 2001),both IL-1α and IL-1β are released by keratinocytes in culture following stress comparable with conditions of injury to the skin in vivo(e.g. Kupper and Groves, 1995; Wood et al., 1996). Whereas in skin, IL-1 signalling is mostly understood as part of a proinflammatory cytokine cascade to induce multiple effects on fibroblasts, endothelial cells and inflammatory cells (for a review, see Dinarello, 1997), we have demonstrated that it has additional major signaling functions in epidermal tissue regeneration.
In keratinocyte co-cultures with fibroblasts, IL-1 upregulates of growth factors in stromal cells to stimulate keratinocyte proliferation, such as KGF and GM-CSF (Chedid et al.,1994; Maas-Szabowski and Fusenig, 1996; Maas-Szabowski et al., 1999; Maas-Szabowski et al., 2000). Thus, deficiency in the constitutive production of IL-1 in HaCaT cells, as already noticed earlier(Ruhland and de Villiers,2001), with the consequence of inefficient induction of KGF and GM-CSF in fibroblasts, could to some extent be overcome by increasing the number of `producer' cells in the collagen gel(Schoop et al., 1999). However, the improvement in tissue organization was best noticed in later culture periods of 2-3 weeks. Upregulation of IL-1 by EGF and TGF-α as already noticed earlier both in normal keratinocytes(Lee et al., 1991) and in HaCaT cells (Philips et al., 1995), was also demonstrated here in HaCaT co-cultures. In addition to the proliferation stimulating effect of KGF and GM-CSF on keratinocytes, GM-CSF also enhanced epidermal differentiation in organotypic co-cultures (Szabowski et al.,2000). This differentiation-stimulating effect of GM-CSF was also noticed in HaCaT organotypic co-cultures; however, only after enhanced receptor expression by TGF-α. The same tendency to a morphologically better differentiated HaCaT epithelium was noticed after addition of IL-1, the inducer of both KGF and GM-CSF in the co-cultured fibroblasts.
Most astonishingly, however, was the crucial role of the EGF receptor and its ligands TGF-α and EGF in regulating epidermal tissue regeneration by HaCaT cells. A very low release and expression of TGF-α by HaCaT clones,compared with ras-transfected tumorigenic HaCaT cells, had already been observed earlier (Game et al.,1992), although this fact had not been associated with their reduced tissue regeneration. Here we clearly demonstrate that the low production rate of TGF-α is due to reduced RNA expression and not a failure of release of the active compound from its membrane-bound precursor. In transplants in vivo, the deficiency of HaCaT cells in TGF-αexpression may be compensated by EGF provided by the circulation(Derynck, 1988). As shown earlier, both high and low affinity EGF receptors are functional on HaCaT cells (Game et al., 1992) and this has been confirmed recently (Kaufmann and Thiel, 2002). Thus, EGF and TGF-α stimulate HaCaT cell proliferation and migration similar to that observed in normal keratinocytes in culture (Pittelkow et al.,1989; Ju et al.,1993), whereas EGF receptor inhibition induces growth arrest(Peus et al., 1997) and consequently blocks stratification in skin equivalents.
Interestingly, in organotypic cultures of HaCaT cells, the stimulating effect of TGF-α on cell proliferation was prominent only at early culture time points, whereas at later stages the rate of DNA synthesis was comparable with untreated control cultures. By contrast, the percentage of apoptotic cells in the non-differentiated epidermal layers was drastically decreased in TGF-α-treated cultures from the beginning. This indicated that the enhancement of survival of HaCaT cells was a major function of TGF-α in the developing epidermal tissue and, consequently, a major contribution to the increased number of cell layers. With the beginning of keratinization in the TGF-α-treated HaCaT epithelia, the number of TUNEL-positive cells also increased here; however, apoptotic cells were located in the upper parakeratotic cell layers. As shown earlier(Schoop et al., 1999), the remnant nuclei of the incompletely differentiated cells stain positive with the TUNEL reaction owing to incomplete DNA degradation(Bernerd and Asselineau,1997).
An important role of EGF receptor signalling for cell survival in epithelial cells has been postulated earlier. EGFR activation in keratinocytes increases the expression of bcl-XL, a member of the anti-apoptotic bcl-2 family of proteins, and human keratinocytes exhibit enhanced apoptosis when EGFR-signalling is inhibited (Rodeck et al., 1997a; Rodeck et al.,1997b; Stoll et al.,1998; Sibilia et al.,2000). Similarly, blockade of EGFR decreased bcl-XL expression and enhanced apoptosis in HaCaT cells following UV-B irradiation(Jost et al., 2001). Thus,EGFR-signalling is important for both keratinocyte proliferation and survival and the latter function seems to play a major role in organotypic cultures when cells have to detach from their specific extracellular matrix, the basement membrane, and migrate upwards to form a multilayered tissue.
On the other hand, stable transfection and overexpression of bcl-2 into HaCaT cells reduced their apoptotic rate in organotypic cultures but did not show noticeable effects on their tissue organization and epidermal differentiation features (Delehedde et al., 2001). Thus, the combined effect of TGF-α on cell proliferation and survival is required for regular stratification and differentiation of HaCaT cells.
Furthermore, a major component of the TGF-α effect on HaCaT cells is the enhanced expression of IL-1 and of the receptors for KGF and GM-CSF,rendering HaCaT cells reactive and responsive for stromal growth regulatory signals. In particular, the induction of the KGF receptor seems to be an essential step to gaining sustained keratinocyte proliferation both in wound healing in vivo as well as in cell culture (for a review, see Werner, 1998; Marchese et al., 1997; Werner et al., 1992). It had been noticed that expression of the KGF receptor is deficient in sparse HaCaT cultures but upregulated upon confluence and the onset of differentiation(Capone et al., 2000). Moreover, the induction of KGFR in keratinocyte cultures grown at high calcium concentration has been interpreted in that its signalling plays a possible role in keratinization control (Marchese et al., 1997). We have shown recently, however, that addition of KGF effects mainly proliferation in organotypic cultures of normal keratinocytes co-cultured with AP-1-defective fibroblasts, whereas differentiation was not affected(Szabowski et al., 2000). Keratinization and formation of specific differentiated strata (i.e. stratum granulosum) was enhanced by GM-CSF and a regular tissue structure required the combined action of both growth factors.
Clearly, the complex process of epidermal tissue regeneration,differentiation and homeostasis is unlikely to be regulated only by the four factors identified so far, i.e. TGF-α, IL-1, KGF and GM-CSF. Further regulating factors have to be identified and their molecular mechanisms of action elucidated. For these investigations, in-vivo-like but reproducible and simple-to-handle skin equivalent in vitro models are required. Owing to the rather faithful mimicry of the basic functions of epidermal proliferation,differentiation and tissue reorganization by the immortal HaCaT cells in TGF-α-supplemented organotypic co-culture with fibroblasts, this reproducible skin equivalent represents a biologically relevant model for such studies. Furthermore, as an immortalized cell line, HaCaT cells can be genetically modified by overexpression and/or blockade of specific genes so that their consequences can be studied in a tissue context. Finally, with further improvement of the quality of structural organization, differentiation and barrier functions of the stratified epithelia formed by HaCaT cells, such skin equivalents may become highly standardized in vitro tissue models for routine testing in pharmacology and toxicology.
Critical discussion and helpful comments of Hans-Jürgen Stark, Dirk Breitkreutz, Axel Szabowski and Petra Boukamp, as well as the technical assistance of Iris Martin, Eva Goedecke, Silke Haid and Angelika Krischke, are gratefully acknowledged. We thank Daniel Hohl (Lausanne) for kindly providing the loricrin-specific antibody, Peter Bohlen (ImClone, New York, NY) for the gift of the EGFR blocking antibody C225 and Ulrich Rodeck (Philadelphia, PA)for the EGF receptor antibody for immunostaining (mAB 425). Work was supported by grants from the Deutsche Forschungsgemeinschaft DFG-Schwerpunkt`Epitheliale Differenzierung' (Fu 94-5), the Israel-DKFZ Cooperation Program by the German Ministry of Research (Ca 94) and Beiersdorf AG, Hamburg.