The shape of cultured epithelial cells does not depend on the integrity of their microtubules

When isolated fibroblasts and epithelial cells are migrating in culture the protrusive lamellar activity that leads to their movement is restricted to a limited region of the cell margin. As a result, moving cells of both types exhibit a characteristically polarized morphology (e.g. see Abercrombie, 1980; Brown & Middleton, 1985). The common use of the term ‘polarizeD′ in this context is perhaps unfortunate because it may be confused with the alternative use of the term to describe the apical-basolateral polarization of epithelial cell sheets. However, throughout this paper it will be used in the former sense to describe the morphology of cultured cells only.

In fibroblasts it is well established that the cells can adopt and maintain a polarized morphology only if they contain an intact system of microtubules (e.g. see Vasiliev & Gelfand, 1976). Treatment with microtubule-disrupt-ing drugs such as colcemid and nocodazole inhibits the polarization of spreading fibroblasts (Goldman, 1971; Ivanova et al. 1976) and causes loss of polarization in previously spread cells (Vasiliev et al. 1970; Goldman, 1971; Gail & Boone, 1971). Fibroblasts treated in this way are unable to migrate and they assume a rounded unpolarized morphology with protrusive lamellar activity more generally distributed around their margins (Vasiliev et al. 1970; Gail & Boone, 1971).

The polarization of epithelial cells is also generally considered to be dependent on an intact microtubule network. For example, cells of the epithelioid cell lines, IAR2 (Domnina et al. 1985) and McA-RH-7777 (Karavanova et al. 1985), respond like fibroblasts to microtubule disruption. However, contradictory reports have suggested that the morphology of epithelial cells, derived from a number of different embryonic chick tissues (DiPasquale, 1975; Chernoff & Overton, 1979) and from fish skin (Euteneuer & Schliwa, 1984), is unaffected by the disruption of their microtubules.

To clarify the position we have used a quantitative method of shape analysis to study the effects of microtubule-disrupting drugs on the polarization of fibroblasts derived from embryonic chick heart and compared this with the effects of the same drugs on the polarization of epithelial cells derived from embryonic chick skin and cornea. Our results suggest that, unlike the fibroblasts, those epithelial cell types that we have studied do not require intact microtubules either to adopt or to maintain a polarized morphology.

The medium throughout consisted of Dulbecco’s modified EaglE′s medium (DMEM) plus Ham’s F12 medium (1:1) supplemented with 10% foetal calf serum and containing 100unitsml⁻¹ penicillin and 100,ugml⁻¹ streptomycin (all from Gibco Ltd). Cultures were maintained at 37°C in an atmosphere of 5% CO₂ in air.

Suspensions of chick heart fibroblasts (HF) were obtained by culturing expiants of 12-day-old chick embryo ventricles in 25cm² tissue culture flasks (Falcon). After 8 or 9 days the outgrowths from these expiants were harvested with 0 ·025% (w/v) trypsin (Sigma Ltd) in calcium- and magnesium-free EarlE′s saline (Gibco Ltd) and the cells resuspended in medium. Suspensions of corneal epithelial (CE) and epidermal epithelial (EE) cells were prepared, as described (Brown & Middleton, 1981, 1985), from 12-day-old and 8-day-old chick embryos, respectively.

Cultures for routine histological examination and for cell shape analysis were prepared by plating out suspensions of the different cell types into 9 mm diameter glass rings waxed onto collagen-coated glass coverslips (see Middleton & Pegrum, 1976). To investigate the effects of microtubule-disrupting drugs on cell spreading, replicate samples (0 ·2 ml) of suspensions of HF, CE and EE cells, containing 0 ·5 ×10’ml⁻¹, 3 ·0×10’ml⁻¹ and 4 ·0 ×10⁵ml⁻¹, respectively, were plated out in this way in both control and drug-containing medium (see below). After 6h incubation the cultures were fixed in formol saline and stained with Harris’ haematoxylin.

To investigate the effects of these drugs on previously spread cells, replicate samples (0 ·2 ml) of the three cell types were plated out at the concentrations stated above in control medium and incubated for 6h. Half the cultures of each cell type were then transferred to drug-containing medium and half to appropriate control medium (see below). After a further 3h incubation the cultures were fixed and stained as described above.

Stock solutions of colcemid (Imgml¹, demecolcine, Sigma Ltd) in water and of nocodazole (1 mgml⁻¹, Janssen Pharmaceutical Ltd) in dimethyl sulphoxide (DMSO) were prepared and stored at —20°C. Before use these stock solutions were diluted with medium to a final concentration of 0 ·1 μg ml⁻¹ in the case of colcemid and of 1 ·0 μg ml⁻¹ in the case of nocodazole. Control medium was prepared by adding equivalent quantities of water or DMSO to medium as appropriate.

Suspensions of HF, CE and EE cells in control medium and in medium containing microtubule-disrupting drugs were plated out onto collagen-coated coverslips (see above). After incubating for 6 h the cultures were washed in phosphate-buffered saline (PBS) and fixed for 5 min with 0 ·25% (v/v) glutaraldehyde in PBS. The cells were permeabilized for 15 min with 0 ·5% (v/v) Triton X-100 in PBS, post-fixed for 5 min in 0 ·25% (v/v) glutaraldehyde in PBS and incubated for 5 min in each of three changes of sodium borohydride (1 mgml⁻¹) in PBS. The cultures were then incubated for 30 min in a rat monoclonal antibody against tubulin (YL 1/2, Scrotcc Ltd), diluted 1:30 (v/v) with Hepes-buffered EaglE′s minimum essential medium (H-MEM) supplemented with 10% foetal calf serum. After extensive washing with serum-free H-MEM the cultures were incubated for 30 min in FITC-conjugated rabbit anti-rat IgG (ICN Biomedicals Ltd) diluted 1:50 (v/v) with H-MEM plus 10% foetal calf serum. The cultures were then washed with serum-free H-MEM and mounted in PBS-bascd AF3 mountant (Citifluor Ltd).

These preparations were examined and photographed using a Leitz Diavert inverted microscope with epi-illumination and equipped for FITC excitation.

We quantified the effects of microtubule-disrupting drugs on the morphology of the different cells using two measures of cell shape, elongation and dispersion, which have been fully described by Dunn & Brown (1986). Both these measures are invariant to translation, rotation and scaling. Elongation takes a value of zero for shapes with no defined long axis. However, if such a shape is stretched along an axis the value of elongation increases without limit and the axis of stretch becomes the long axis. Dispersion, which is invariant to such a stretch transformation, takes a value of zero for any ellipse and increases as the shape deviates from an ellipse (Dunn & Brown, 1986). Thus freshly seeded cells, which tend to be rounded, have very low values for both elongation and dispersion, since both values are zero for a circle. As spreading cells polarize these values will increase, since polarized cells are both elongated and differ in shape from an ellipse. These shape factors, therefore, measure two different aspects of the polarized state. Examples of the way in which these measures vary in relation to changes in cell shape are shown by Dunn & Brown (1986).

Both the epithelial cell types used in this study exhibit contact-induced spreading (Brown & Middleton, 1981; Brown, 1983) and respond to cell-cell contacts by spreading more extensively on the substratum and becoming more polarized. However, HF do not exhibit similar behaviour and to prevent this difference between the cell types influencing our results we therefore restricted all our measurements to isolated cells. In fact the mean values for the elongation and dispersion of isolated epithelial cells in control medium were similar to, and in some cases exceeded, those for isolated HFs (e.g. see Fig. 3A,B). This is not surprising, since isolated epithelial cells spend a proportion of their time exhibiting well-polarized morphologies (Middleton, 1982) and isolated fibroblasts, although usually better spread than unpolarized epithelial cells, spend a proportion of their time exhibiting equally unpolarized morphologies.

Cells to be analysed were photographed onto 35 mm film and the cell outlines were traced by projecting the negatives onto paper with a photographic enlarger. The magnification of the cells obtained during photography and projection was adjusted so that the outlines of the different cell types were of similar size. In practice, the average area enclosed by the outlines was approximately 35 cm² The zero-, first- and second-order moments of the cell shapes were calculated from polygonal approximations of the cell outlines (Blair & Biss, 1967) entered into a PDP 11/44 computer by tracing round the outlines on a digitizing tablet (Summagraphics, Bitpad 2) linked to the computer. The outlines typically consisted of 100-250 coordinate pairs. Elongation and dispersion were calculated from the moments as described by Dunn & Brown (1986).

Immunofluorescence observations using anti-tubulin antibody demonstrated that heart fibroblasts (HF) and both the epithelial cell types contained abundant microtubules when cultured in control medium (Fig. 1A,C,E), but that these were absent from the cells when cultured in colcemid-containing medium (Fig. 1B,D,F).

The presence of nocodazole in the culture medium also eliminated any microtubules, detectable by immunofluorescence, from all three cell types (data not shown).

We investigated the effects of colcemid on cell spreading by plating out suspensions of HF, corneal epithelial (CE) and epidermal epithelial (EE) cells in control and colcemid-containing medium and incubating the cultures for 6h before fixation.

At this stage cultures of HF in control medium contained well-spread cells, many of which had prominent leading lamellae and exhibited a clearly polarized morphology (Fig. 2A). In colcemid-containing medium, however, the majority of the cells lacked leading lamellae and had an unpolarized highly irregular morphology (Fig. 2B). In contrast, the presence of colcemid in the culture medium did not obviously affect the morphology of either of the epithelial cell types and the drug-treated cultures were indistinguishable from the controls (Fig. 2C-F). The majority of CE and EE cells exhibited prominent leading lamellae and had a well-polarized morphology that was not apparently affected by the presence of colcemid in the medium (Fig. 2C-F).

We repeated this experiment using nocodazole and found that its effects on spreading HF were similar to those of colcemid and that like colcemid it apparently had no effect on the morphology of either of the epithelial cell types.

In an attempt to confirm and further characterize the apparently different response of these fibroblasts and epithelial cells to microtubule-disrupting drugs we employed two measures of cell shape, elongation and dispersion. These measures quantify two different aspects of morphological polarization and both increase in value with increasing polarization (see Materials and methods). In this analysis we first used the cultures, described above, of cells fixed after spreading for 6h in control or colcemid-containing media. For each cell type we measured the elongation and dispersion of 35 randomly selected isolated cells in each of three replicate cultures in control medium and in each of three replicates in colcemid-containing medium. Fig. 3A,B demonstrates that HF that had spread in colcemid-containing medium had significantly lower values for both elongation and dispersion than did the cells in control medium. In contrast, the presence of colcemid in the culture medium did not significantly affect the elongation or dispersion of either spreading CE or EE cells (Fig. 3A,B).

We repeated this analysis, using the same technique, on cultures of HF and CE cells fixed 6h after plating out the cells in control or nocodazole-containing medium. It is clear from Fig. 4A,B that nocodazole, like colcemid, significantly reduced both the elongation and dispersion of spreading HF but that it did not have a significant effect on either of these measures for spreading CE cells.

We also examined the effects of microtubule-disrupting drugs on the morphology of cells of all three types that had previously settled and polarized in control medium. Suspensions of HF, CE and EE cells were plated out and cultured for 6h in control medium before replicate cultures of the different cells were transferred either to fresh control medium or to colcemid-containing medium for a further 3 h before fixation. The morphology of the HF was clearly altered by the transfer to drug-containing medium, since the well-polarized cells present in the controls (Fig. 5A) were absent from the drug-treated cultures (Fig. 5B) and after 3 h in the presence of colcemid the morphology of the cells closely resembled that of HF that had spread in the continuous presence of the drug (Fig. 2B). However, the same treatment had no obvious effects on the morphology of either of the epithelial cell types and cultures of both CE and EE cells transferred to colcemid-containing medium were indistinguishable from their respective controls (Fig. 5C-F). Substituting nocodazole for colcemid in this experiment again gave similar results to those obtained with colcemid for all three cell types (data not shown).

We quantified the differential response of spread chick fibroblasts and epithelial cells to microtubule-disrupting drugs using the cultures of HF and CE, described above, which were fixed 3 h after transferring them either to fresh control medium or to colcemid-containing medium.

As before we measured, for each cell type, the elongation and dispersion of 35 isolated cells in each of three replicate cultures transferred to fresh control medium and in each of three transferred to colcemid-containing medium.

The mean values of both elongation and dispersion ofHF transferred to colcemid-containing medium were significantly smaller than those of their equivalent controls (Fig. 6A,B). In contrast, the elongation and dispersion of CE cells transferred to colcemid-containing medium were not significantly different from those transferred to control medium (Fig. 6A,B).

Our qualitative observations suggested that drug-induced microtubule disruption reduced the polarization of spreading HF and reversed the polarization of previously spread cells of this type. Our quantitative analysis of cell shape supports this conclusion. Although the measures we have used do not completely describe polarization, they do quantify two aspects of cell shape, elongation and dispersion, which increase with increasing polarization (Dunn & Brown, 1986). Our data demonstrate that HF that have spread in the presence of microtubule-disrupting drugs have significantly lower values for both elongation and dispersion than they do when they have spread in control medium. Similarly, we found that colcemid-induced microtubule disruption significantly reduced both the elongation and dispersion of HF that had previously spread in control medium. Taken together these findings indicate that, as is the case in other fibroblastic cell types (see Introduction), chick HF require an intact system of cytoplasmic microtubules in order both to adopt and to maintain a polarized morphology.

However, in contrast, both our qualitative and quantitative observations suggest that microtubule-disrupting drugs have little effect on the polarization of the chick epithelial cell types that we have investigated. Despite the disruption of all microtubules detectable by immunofluorescence microscopy, CE and EE cells that had spread in the presence of these drugs appeared indistinguishable from their counterparts that had spread in control medium and the drug-treated cells had values of elongation and dispersion that were not significantly different from those of the cells in control medium. In addition, the morphology of CE cells that had previously spread in control medium appeared unaffected by subsequent exposure to colcemid, and again this treatment did not significantly change their values of elongation and dispersion. We have subsequently quantified the effects of colcemid on previously spread EE cells and obtained similar results (Middleton & Roberts, unpublished).

Our observations and quantitative measurements suggest that these chick epithelial cell types, unlike fibroblasts, do not require an intact system of cytoplasmic microtubules in order to adopt or maintain a polarized morphology. Whether or not this is also the case for other epithelial cells remains to be established, but qualitative results obtained in some other systems have suggested a similar conclusion. The morphology of isolated fish epidermal kératinocytes has been found to be unaffected by microtubule disruption (Euteneuer & Schliwa, 1984). Similarly, DiPasquale (1975) reported that the morphology of chick embryo gut epithelial cells, as well as of CE and EE cells, was unaltered by microtubuledisrupting drugs and Chernoff & Overton (1979) found that the same was true for epithelial cells derived from early chick blastoderms. In some cases, however, the polarization of epithelial cells has been shown to be microtubule-dependent (Domnina et al. 1985; Karavanova et al. 1985), but it may be significant that in these cases the results were obtained using epithelioid cell lines that may not necessarily have retained the characteristics of the parent epithelium. In view of this we are currently using quantitative methods, similar to those used here, to investigate the relationship between the time for which cells have been in culture and the extent to which their polarization is sensitive to microtubule disruption.

The reasons underlying the contrasting effects of microtubule disruption on the polarization of fibroblasts and epithelial cells that we have observed are unclear. One factor that may be important is the way in which the microtubules of these different cell types interact with their intermediate filaments. The vimentin filaments of fibroblasts collapse when their microtubules are disrupted (Hynes & Destree, 1978; Wang & Goldman, 1978; Franke et al. 1978; Henderson & Weber, 1981) but the cytokeratin filaments of epithelial cells generally do not (Osborn et al. 1977; Franke et al. 1978; Henderson & Weber, 1981; Karavanova et al. 1985). If the effects of microtubule-disrupting drugs on cell shape are mediated via the intermediate filaments this could explain the differential response of the cell types to these drugs. Such evidence as there is does not appear to support this suggestion, since the selective collapse of cytokeratin intermediate filaments, without simultaneous disruption of microtubules, has no effect on the morphology of either PtKl or PtK2 epithelial cells (Eckert, 1985; Klymkowsky et al. 1983). However, these experiments again involved cell lines and it would be of interest to repeat them on cells more closely related to their parent tissue.

Another factor that may be relevant to our results is the respective sizes of the different cells involved. As is clear from the scale bars in Figs 2 and 5, the HF were significantly bigger than either of the epithelial cell types. In fact the mean spread area (area of substratum occupied) of HF was approximately 1300 firn², while that of the CE and EE cells was approximately 200 ftm² and 350 um², respectively. Thus the differential effects of microtubule-disrupting drugs on fibroblasts and epithelial cells that we have observed could be explained if the polarization of smaller cells is less dependent on microtubules than that of larger cells. Some support for this suggestion comes from the observation that the polarization of small fragments of fibroblastic cells, unlike that of their intact counterparts, is unaffected by the disruption of their microtubules (Gelfand et al. 1985)

In culture there is a close association between the ability of fibroblasts and epithelial cells to adopt a polarized morphology and their ability to migrate. Isolated cells of either type that cannot polarize do not show significant migration (Vasiliev et al. 1970; Gail & Boone, 1971; Brown & Middleton, 1985). Since the polarization of CE and EE cells appears to be microtubule-independent, it is possible that the migration of these epithelial cells, in contrast to that of fibroblasts, may not be inhibited by microtubule disruption. We have not established whether this is the case for these cells, but some evidence in favour of this suggestion is available from studies of other epithelial cell types. Euteneuer & Schliwa (1984) have shown that microtubule-disrupting drugs do not affect either the speed or directional persistence of migration of isolated fish kératinocytes, and DiPasquale (1975) found that such drugs did not reduce the speed of spreading of sheets of epithelial cells derived from a number of different chick embryonic tissues.

Cells in culture become polarized when their protrusive lamellar activity is restricted to a limited region of the cell margin. Our results and those from previous investigations clearly imply that in fibroblasts this restriction involves cytoplasmic microtubules. However, the microtubule-independent polarization that we have described in epithelial cells suggests that the restriction of lamellar activity in these cells may involve a different mechanism.

We are grateful to Mr P. Drake and Miss A. K. Kernaghan for their excellent technical assistance and to Miss L. Cockroft for typing the manuscript. C. A. Middleton thanks Professor B. Boycott, F.R.S. and Dr G. A. Dunn for enabling him to work in the MRC Cell Biophysics Unit whilst on research leave from the University of Leeds.