Cell density affects spreading and clustering, but not attachment, of human keratinocytes in serum-free medium

Attachment and spreading of cells on plastic surfaces seem to be the first steps in the onset of in vitro culture of several types of cells and in particular of keratinocytes. The definition of optimal conditions for the two phenomena and the understanding of the mechanisms underlying them are important prerequisites not only for successful in vitro cultures, but also for the analysis of cellular processes, such as morphogenesis, proliferation and differentiation, which are widely accepted as being influenced by the shape of cells (for reviews, see Ingber and Folkman, 1989; Watt, 1986).

In recent years much attention has been paid to the requirements of keratinocytes in terms of the nature, both chemical and physicochemical, of the substratum upon which these cells show optimal growth. The problem has been tackled with a variety of experimental systems, such as primary expiants and 2nd or 3rd passage keratinocytes from different organisms (man, guinea pig, rat). On the one hand, the effect of different synthetic membranes on growth and differentiation has been studied (Vaughan et al. 1986). On the other, a group of basement membrane associated molecules, including fibronectin, laminin, thrombospondin, types I, III and IV collagen, heparan sulphate and proteoglycans (Bemstam et al. 1986; Clark et al. 1985; Gilchrest et al. 1980; Junker and Heine, 1987; Kubo et al. 1987; Morykwas et al. 1989; O’Keefe et al. 1985; Stenn et al. 1983; Terranova et al. 1980; Toda and Grinnell, 1987; Varani et al. 1988), alone or in different combinations, have been demonstrated to be involved in in vitro adhesion and spreading of keratinocytes. Attachment of epithelial cells to natural extracellular matrices has also been tested (Furukawa et al. 1987; Tinois et al. 1987), as well as on other substrata, such as lectins (Toda and Grinnell, 1987), de-epidermized dermis (Grinnell et al. 1987; Régnier et al. 1988) and collagen-glycosaminoglycan substrata (Boyce and Hansbrough, 1988).

The results of these studies are partly contradictory, but allow it to be concluded that epithelial cells display several mechanisms of spreading and that none of the molecules or substrata tested is required by all spreading modes.

Moreover, serum components that mediate cell spreading effects and are distinct from fibronectin, laminin and chondronectin, have been isolated from human and calf serum (Barnes and Silnutzer, 1983; Stenn, 1981; Whateley and Knox, 1980) and have subsequently been identified as vitronectin (Hayman et al. 1983; Stenn et al. 1986). On the other hand, the presence in serum of a cell-spreading inhibitor has also been reported (Katayama et al. 1986).

The molecular aspects of cell-to-matrix and of cell-to-cell interactions through integral membrane glycoproteins have recently been studied in a variety of cell systems, ranging from human (Guo et al. 1990; Kajiji et al. 1989; Kaufmann et al. 1989; Staquet et al. 1990; Suzuki and Naitoh, 1990, for epithelial cells) to Dictyostelium discoideum cells (Kamboi et al. 1989).

Less attention has been paid, on the contrary, to the properties of the keratinocytes that adhere and spread. Watt (1984) showed that involucrin-positive keratinocytes have a lower adhesive affinity for the culture substratum than involucrin-negative cells (basal cells) and that the latter display more rapid attachment and spreading kinetics than the former. Moreover (Watt, 1987), involucrin-positive cells appear to adhere preferentially to the surface of spread basal cells rather than directly to the culture plastic.

In the present work we began a study of the cellular factors affecting attachment and spreading of human keratinocytes in serum-free medium, at the very beginning of a culture, before the onset of cell division. In particular, we took into consideration the effect of cell density and of time on these two processes. Our data show that attachment of keratinocytes to the plastic surface is not affected by cell density, that only a fraction of attached cells undergoes spreading and that some of the cells appear in clusters, suggesting a complex functional heterogeneity of the starting cell population. The percentage of spread keratinocytes is cell density-dependent and therefore spreading appears as an inducible phenomenon, due to a keratinocyte-secreted diffusible factor. The use of media conditioned by different types of cells shows that the spreading factor is specifically secreted by keratinocytes and seems to act directly on the cells. Data are presented showing that the rate of spreading is cell densityindependent and that both cluster formation and increase in cluster size occur in suspension.

Human keratinocytes were obtained from skin biopsies and cultured as described by Rheinwald and Green (1975ft). Primary keratinocytes cultures were routinely grown on a lethally irradiated 3T3 feeder-layer in a humidified atmosphere of 5% CO2 and 95% air, at 37 °C, in a growth medium obtained by mixing Dulbecco-modified Eagle’s and Ham’s F12 media (3:1, v/v), containing 10% foetal bovine serum, 10⁻¹°M cholera toxin, 0.5 μgml^-1 hydrocortisone, 5 μgml^-1 insulin, 1.8×10⁻⁴M adenine, 5 μgml^-1 transferrin, 2×10⁻⁹M triiodothyronine, l0ngml^-1 EGF, l00 i.u.ml^-1 penicillin and l00 μgml^-1 streptomycin.

No fibroblasts were present in keratinocyte cultures, as evaluated morphologically and by Rhodanile Blue staining (Rheinwald and Green, 1975a).

Medium MCDB153, used in adhesion and spreading assays, was prepared as described by Boyce and Ham (1986), starting from MCDB151 (Sigma, St Louis, MO). This was supplemented with (final concentrations): 5iigml^-1 insulin, lOngml^-1 EGF, 5 μg ml^-1 transferrin, 0.5 μg ml^-1 hydrocortisone, 0.1 mw ethanolamine, 0.1 mM phosphoethanolamine. Final concentrations of CaCl₂ and FeSO₄ were 1×10⁻⁴M and 5×10⁻⁶M, respectively. Essential amino acids (7.5×10⁻⁴M isoleucine, 2.4×10⁻⁴M histidine, 9×10^−SM methionine, 9×l0^-5M phenylalanine, 4.5×10^−SM tryptophan and 7.5×10^−SM tyrosine) were added according to the method of Pittelkow and Scott (1986). Trace elements were added according to the method of Boyce and Ham (1986).

All chemicals were purchased from Sigma Chemical Co. (St Louis, MO, USA), except for Ham’s F12 medium, D-MEM medium and foetal bovine serum, which were obtained from Flow Laboratories (Irvine, Scotland, UK).

Cells from strain 1 and strain 2, used in these experiments, derived from single biopsies from two Caucasian individuals, female and male, respectively, both 50 years old.

Primary, semi-confluent keratinocytes cultures were removed from culture flasks by treatment with 0.25 % trypsin containing 0.025% EDTA for 5min. Trypsinization of keratinocytes was preceded by selective removal of 3T3 cells by EDTA treatment (0.025% for 5 min). Trypsin was neutralized with a solution containing 0.1% (w/v) soybean trypsin inhibitor and 4mgml^-1 bovine serum albumin. The mixture was pipetted vigorously to ensure single cell suspension; the absence of clusters was verified by microscopy. The cell suspension was pelleted by centrifugation at 1000 revs min^-1 for 10 min, washed twice with MCDB153 and counted in a hemocytometer.

Cells were plated in duplicate in 2 ml of MCDB153 at five different densities (0.5, 1, 2, 5, 10 and 20 ×10³ cells cm⁻²) into 12-well plates (Costar).

Plates were incubated at 37 °C in a humidified atmosphere of 5% CO₂ (v/v) in air. Unless otherwise stated, incubations were performed for 24 h. After incubation, plates were rinsed in PBS to remove non-adherent cells, fixed in methanol and stained with Giemsa.

Cell viability was determined by the Trypan Blue dye exclusion test.

Scoring of attached and spread cells was performed under a Leitz phase-contrast microscope, by counting 15 fields for each well, corresponding to 1/40 of the total area of the well. These fields were chosen according to the following criterion: one at the centre, 12 along two concentric circumferences and two at the border of the well. In the case where 10⁴ cells or less were plated, the number of counted fields was increased to 45, corresponding to 3/40 of the total area of the well.

The percentage of adherent cells was calculated by dividing the number of adherent cells (number of counted adherent cells ×appropriate multiplication factor) by the number of plated cells, ×100.

As far as the evaluation of spreading is concerned, a cell was considered spread if it was flattened and its cytoplasm, nucleus and nucleoli were clearly visible.

Cells were classified as unspread when they were highly refractive and rounded, with no visible cytoplasmic organisation around the nucleus.

The percentage of spread cells was calculated by dividing the number of counted spread cells by the number of attached cells (unspread+spread cells), ×100.

For the production of keratinocyte-conditioned medium (KCM), 10^s second-passage human keratinocytes, prepared by trypsin-EDTA treatment after removal of the feeder cells with EDTA, were plated in 2 ml of MCDB153 into 12-well plates. After 24 h of incubation at 37 °C in a humidified atmosphere of 5 % CO₂ (v/v) in air, medium was removed, spun at 10 000 revs min^-1 for 10 min to remove any floating cells and either used directly, stored at 4 °C for up to two weeks, or frozen at −20°C for up to two months. Results using fresh KCM, KCM stored at 4°C or frozen and thawed were identical. Frozen KCM was routinely used. No viable cells were present in KCM, as evaluated by scoring wells in which 2 ml of KCM were incubated for 24 h.

Media conditioned by human diploid fibroblasts (5th-7th passage), by HeLa cells and control media, containing no cells, were prepared in the same way and processed as above.

Pre-coating of both tissue culture (Sterilin Cell-Cult, 50 mm) and bacteriological (Sterilin, 50 mm) dishes was performed by incubating at 37 °C in humidified atmosphere of 5% CO₂ (v/v) in air for 24 h fresh or conditioned media in the absence of cells, by removing media and by rinsing with PBS. Spreading of keratinocytes was tested separately on the pre-coated dishes in the presence of fresh MCDB153 medium and on untreated dishes in the presence of the media used for pre-coating. As an additional control (reconstruction experiments), spreading was evaluated on pre-coated dishes in the presence of the media used for precoating.

Regression analyses were performed to test relationships between the number of attached cells and cell density.

Differences in the evaluated parameters among the experimental groups were assessed by the multifactor analysis of variance (Sokal and Rohlf, 1981).

The effect of cell density on attachment of human keratinocytes, at the second in vitro passage, to the polystyrene surface of plastic dishes has been evaluated after 24 h incubation in serum-free medium.

Two different strains of cells have been analyzed, referred to as strain 1 and strain 2, respectively. Each strain derived from a single biopsy from a different individual.

Within the range of cell densities tested (5×10² to 2×l0⁴ cells cm^-2), the number of attached keratinocytes for each strain increased as a linear function of cell density (Fig. 1A). As a corollary, the calculation of the relative number of attached cells (expressed as the percentage over the total inoculum) showed that this parameter is independent of the density of the inoculum throughout the range of cell densities tested (Fig. 1B).

Multifactor analysis of variance, taking cell density and cell strain as sources of variation, showed that the percentages of attached cells were unaffected by cell density (P>0.05), whereas the percentage of attached cells (approximately 50% and 30%, for strain 1 and strain 2, respectively) was statistically different for the two strains of keratinocytes (P<0.01).

The percentage of attached cells appeared to be an intrinsic characteristic of each strain, as shown by the fact that this value could be reproducibly obtained by testing these strains in the attachment assay, even after several months of storage of cells in liquid nitrogen (Table 1).

Vital staining of unattached cells showed that this subpopulation was composed of approximately 63 % dead cells. Vital cells were large, apparently differentiated keratinocytes. Replating of this subpopulation did not result in any attachment.

Attached keratinocytes could be subdivided, on the basis of their organisation and morphology, into three classes (Fig. 2): (1) single unspread cells; (2) single spread cells; (3) clustered cells, both spread and unspread.

Clusters of different sizes were observed (Fig. 2B and C).

No mitotic plates were observed in approximately 5×l0⁵ cells scored in this study. Control experiments, however, demonstrated that these cells (analysed in the present experiments 24 h after plating) were able to proliferate in the same experimental conditions, provided that the incubation was prolonged.

In these keratinocyte cultures melanocytes could also be observed. The percentages were significantly different for the two strains tested (4.1±0.17 and 1.4±0.13 for strain 1 and 2, respectively).

In order to understand the origin of such different morphological aspects, we tested the effect of cell density on the distribution of attached keratinocytes among the classes described above.

For each strain of keratinocytes, the percentages of unspread, total spread (single plus clustered, i.e. belonging to a cluster of cells), single spread, clustered (including spread and unspread) cells and of clusters, over the total number of attached keratinocytes, have been evaluated at different cell densities. Differences in these parameters, taking cell density and strain as sources of variation, were assessed by multifactor analysis of variance.

Although, as reported above, the two strains of cells differed significantly in the percentage of attached cells, no significant differences in the percentages of each class of attached cells were observed between the two strains (data not shown). We therefore plotted in Figs 3 and 4 the mean values of these percentages obtained from all experiments.

The percentage of total spread cells (single plus clustered) increased significantly (P<0.01) with increase of cell density, while that of unspread cells decreased (P<0.01) (Fig. 3). The increase in total spread cells resulted from significant increases in the percentages of both single spread cells (P<0.01) and clustered spread cells (P<0.01) at increasing cell densities.

The percentage of total clusters, as shown in Fig. 4, followed saturation kinetics at increasing cell densities: it increased up to cell densities of 10⁴ cells cm”² and remained constant at higher densities. Similar kinetics were observed for the percentage of clusters composed of spread and unspread cells (the latter reaching ‘saturation’ at lower cell densities). The percentage of total clustered cells (spread plus unspread), on the contrary, showed a linear increase at increasing cell density (Table 2).

Consequently, cell density of the inoculum also affected the size of the clusters (defined as the number of cells constituting them), as shown in Fig. 5. Up to cell densities of 2×l0³cellscm, almost all clusters were made up of two cells. Clusters resulting from more than 10 cells were detected only when cells were plated at the highest inoculum tested (2×10⁴ cells cm⁻²). No difference in the distribution of cluster sizes at any of the cell densities tested was observed between the two strains of cells.

In order to characterize attachment and spreading of keratinocytes further, time-course experiments were performed at three of the densities tested above.

As shown in Fig. 6A, the kinetics of attachment and spreading were identical at all cell densities tested. Attachment was completed, at each cell density tested, within 3 h after plating.

Spreading of keratinocytes began within the first hour of incubation and was completed by 16 h at all densities tested. Data depicted in Fig. 6 refer to cells from strain 1. Identical results were obtained for cells from strain 2.

The time-course of clustering is reported in Fig. 6B. The normalized number of total clusters appeared to be constant at each of the times tested for each cell density; on the contrary, a time-dependent increase in the relative number of clusters of spread keratinocytes was observed and reached the maximum value at 16 h. As far as cluster size is concerned, the percentages of total clustered cells (spread plus unspread) were constant at each of the times tested for each cell density (Table 2).

To test the hypothesis that cell density-dependence of spreading of keratinocytes is due to the secretion, by keratinocytes themselves, of a diffusible spreading factor, keratinocyte-conditioned medium (KCM) from both strains was prepared.

As shown in Fig. 7, a statistically significant increase CP<0.01) in the percentage of spread keratinocytes was observed when cells were plated in KCM. This increase is not due to incubation of the medium at 37 °C for 24 h (compare histograms B and C). Moreover, spreading of keratinocytes plated at 2 × 10³ cells cm^-2 in the presence of medium conditioned by 2×l0⁴cellscm^-2 reaches values of spreading identical to those observed for keratinocytes directly plated at 2×l0⁴ cells cm^-2 (histogram F). No difference was detected between KCMs from the two cell strains, or between the two strains in their sensitivity to KCMs.

Media conditioned by human diploid fibroblasts and by HeLa cells did not enhance the spreading of keratinocytes (histograms D and E).

The spreading activity of conditioned media was stable to heating at 70°C for 10min, but was destroyed by heating at 100 °C for 10 min and by incubation with 50 μgml^-1 (final concentration) trypsin for Ih at 37°C, followed by neutralisation of the enzyme by 50 μgml^- (final concentration) soybean trypsin inhibitor (data not shown).

As shown in Fig. 8, tissue culture dishes pre-coated for 24 h with keratinocyte-conditioned medium supported only a limited increase in spreading of keratinocytes when the assay was performed with cells plated in fresh MCDB153 (histogram D). On the contrary, most of the spreading activity was detected when the assay was performed in untreated dishes with cells plated in the conditioned medium used for pre-coating (histogram E).

Full spreading activity was observed in reconstruction experiments, when cells were plated in KCM-pre-coated dishes in the presence of the conditioned medium used for pre-coating (histogram F). The percentages of attached cells were identical in pre-coated and untreated dishes (data not shown). Similar results were obtained when precoating was performed on bacteriological dishes.

Different aspects of proliferative heterogeneity of cultured keratinocytes have been pointed out (Albers et al. 1987; Barrandon and Green, 1985; Barrandon and Green, 1987).

Our data demonstrate the functional heterogeneity of cultured human keratinocytes in so far as attachment, spreading and clustering are concerned.

Clearly, this population includes two subpopulations, one represented by cells able to attach to the plastic surface of flasks (‘attachable’ cells), the other by those unable to do so. The fact that the percentage of attachable cells is not affected by cell density and is constant, for a given strain, when cells are tested for attachment after several months of storage, suggests that attachable and non-attachable cells form non-interconvertible subpopulations. This is not surprising, since: (1) among dissociated cells a certain percentage of dead cells is present; and (2), although keratinocytes in culture are less differentiated than in epidermis, nevertheless to some extent they undergo differentiation and consequently may partly lose their ability to attach to plastic surfaces (Watt and Green, 1982).

Our values for attachment fall within the ranges reported by different authors (Clark et al. 1985; Grinnell et al. 1987; Kubo et al. 1987) for human keratinocytes. Our kinetics of adhesion show that attachment is completed within 3 h after plating, in agreement with that reported for mouse transformed epithelial cells (PAM212) plated on coated polystyrene (Terranova et al. 1980) and for human keratinocytes seeded on dermal matrix (Grinnell et al. 1987), and shorter than that reported by Watt (1987), who observed maximum attachment of keratinocytes on plastic dishes 8h after plating. This could be explained on the basis of different culture conditions and/or on the different proportion of involucrin-negative cells (Watt, 1984).

But our data suggest, in addition, that attachable cells are not a homogeneous group. Different subpopulations can be identified: cells able to attach, but unable to undergo spreading spontaneously, and cells able to attach and to spread spontaneously, singly or as a part of a cluster of cells. The finding that the percentage of spread keratinocytes increases and that of unspread decreases at increasing cell density suggests that the latter can be converted into the former as a function of the number of cells present on the plate: in other words, spreading of keratinocytes on uncoated surfaces is an inducible phenomenon. Melanocytes, on the contrary, do not show this property (Tenchini and Malcovati, unpublished data).

The increase in the percentage of spread keratinocytes results from significant increases in both single and clustered cells. In particular, the increase in single spread cells, since by definition such cells are not in contact with each other, can only be explained by the hypothesis that induction is brought about by some diffiisible factor produced by keratinocytes themselves and acting on keratinocytes. The use of keratinocyte-conditioned medium strongly supports this idea. This phenomenon, which we termed ‘autocrine induced spreading¹ is different from the contact-induced spreading observed in corneal and pigmented retina epithelial cells from chick embryos (Middleton, 1977; Brown and Middleton, 1981). The latter phenomenon might explain the slightly steeper response to cell density in spreading by clustered cells in comparison to single keratinocytes.

In the preceding discussion we assumed that a spreading factor is produced by keratinocytes. Actually, melanocytes are also present in our cultures and might be responsible for this phenomenon. We excluded this possibility on the basis of the postulate that the extent of induction of spreading should be proportional to the number of ‘spreading factor’-secreting cells present in the culture. In fact, the extent of induction of spreading was identical in both strains tested, although they differed significantly in the percentage of melanocytes present. Moreover, the lack of spreading activity in media conditioned by other human cell types strongly supports the idea that such a ‘spreading factor’ is specifically produced by keratinocytes.

The fact that the spreading activity of KCM is destroyed by boiling and proteolytic treatment suggests that the spreading factor is a protein.

The observations that most of the spreading activity can be recovered in the conditioned media used for pre-coating and that full activity can be recovered in reconstruction experiments seem to suggest that the spreading factor acts directly on cells and not through a modification of the plastic surface of the dishes. Indirect evidence that this factor may not be a molecule of the extracellular matrix, such as fibronectin, laminin, collagen and thrombospondin, known to be secreted by keratinocytes (O’Keefe et al. 1984; Pruniéras et al. 1983; Varani et al. 1988), comes from the observation that identical adhesion values were observed on untreated and KCM-pre-coated dishes. On the contrary, dishes pre-coated with such cell-adhesive proteins have been shown to enhance keratinocyte attachment (Clark et al. 1985; Kubo et al. 1987; Varani et al. 1988).

The time-course experiments show that although the extent of spreading of keratinocytes, as discussed above, is cell density-dependent, the rate of spreading is not affected by cell density, being the same at all the densities tested and being completed at 16 h.

The presence of ‘small aggregates’ of in vitro cultured keratinocytes, which seem homologous to our clusters, has been noticed by Furukawa et al. (1987). Similar structures can be observed in the papers of Clark et al. (1985) and Watt and Green (1981). ‘Clustering’ associated with morphological alteration was also described by Bemstam et al. (1986). The clusters we observed do not arise from incorrect trypsinisation of primary cultures, since in this case their percentage should have been constant at all cell densities and different from one experiment to another.

Time-course experiments show that the percentages of total clusters (both spread and unspread) and of total clustered cells are constant in time but are cell densitydependent. It can therefore be concluded that cluster formation and increase in cluster size take place in suspension.

The percentage of total clusters follows saturation kinetics at increasing cell density, while the total number of clustered cells increases proportionally to cell density.

Two hypotheses could explain these results. According to the first hypothesis, clustering could result from random collisions between cells. The chance of collision would increase as cell density increases: this would explain the first part of the saturation curve depicted in Fig. 4. But in order to explain the lack of increase in the percentage of clusters at the highest cell densities (when saturation is reached) some restriction must be introduced in this hypothesis: for example, if clusters are dynamic structures (as observed by Brown and Middleton, 1981, in the case of corneal epithelial cells), it could be postulated that the rate of cluster disaggregation increases with cell density and that at the highest cell densities an equilibrium is reached between the rates of cluster formation and disaggregation.

According to the second explanation, the population of cells could contain a definite and constant proportion of ‘centres’ able to originate clusters of cells (‘clusterogenic’ cells). Clusters would then include two types of cells: clusterogenic and ‘gregarious’ cells. The saturation kinetics of cluster formation could be explained by the finding that the more cells are close to each other, the higher will be the chance of gregarious cells colliding with a clusterogenic cell; above a given cell density, all clusterogenic cells would have had the chance of generating a cluster and the percentage of the latter could not increase any more, while their size could increase, as observed, in proportion to cell density. According to this hypothesis, the maximum (saturation) percentage of clusters would be a measure of the percentage of clusterogenic cells in the population. The value observed (8 %) seems to rule out the possibility that this class of cells coincides with involucrin-negative cells, whose percentage, according to Watt (1987), is of the order of 60 % of the keratinocytes population.

In conclusion, our data allow us to identify two specific phenomena (regulation of spreading and of clustering), which can be studied separately. Moreover, their deeper understanding would be relevant to the technology of keratinocyte cultures.

This work was supported by grants from Consiglio Nazionale delle Ricerche, Rome, Italy, contract no. 88.03121.26 and by Progetto Finalizzato ‘Biotecnologie e Biostrumentazione’, contract no. 89.00240.70.