Shapes of cells spreading on fibronectin: measurement of the stellation of BHK21 cells induced by raising cyclic AMP, and of its reversal by serum and lysophosphatidic acid

Fibroblasts in culture often change shape when their cyto-plasmic cAMP is elevated. Early work (reviewed by Will-ingham, 1976) on the effects of cAMP on cell shape was motivated by interest in malignant transformation. Raising cAMP caused morphological changes considered opposite to transformation: 3T3 cells and their transformed derivatives flattened, CHO cells elongated and formed parallel arrays. In contrast, many cells respond to cAMP elevation with a more dramatic shape change, known as arborization or stellation. Similar responses have been found in remarkably diverse cell types, including fibroblasts (Lamb et al., 1988), smooth muscle cells (Kreisberg et al., 1984; Chal-dakov et al., 1989), osteoblastic cells (Egan et al., 1991), melanoma cells (Preston et al., 1987), astrocytes (Baorto et al., 1992) and even epithelial cells (Ortie et al., 1973; Roger et al., 1988). They result from the formation, by outgrowth or cytoplasmic retraction, of narrow elongated processes. Very similar outgrowth of processes in response to cAMP elevation has been observed in various neurone-related cells (Miller and Ruddle, 1974; Heidemann et al., 1985; Rydel and Greene, 1988). In astrocytes and neurones the arborized shapes seem appropriate. In vascular smooth muscle cells (Nabika et al., 1988), they resemble shapes which may occur pathologically (Orekhov et al., 1986). We show here that cells of lines such as BHK21, 3T3 and CHO, which featured in the transformation-related work of the 1970s, also become stellate in response to elevation of cAMP, if bovine serum is omitted from the medium. In these cells the shapes are surely bizarre.

When stellation is induced in cells already spread in culture, elongated processes are formed chiefly by retraction of cytoplasm. However, rat astrocytes plated in the presence of dibutyryl cAMP extended narrow processes from a rounded cell body without prior spreading (Baorto et al., 1992). We show that in the presence of agents which raise cAMP, cells of fibroblastic lines likewise adopt stellate shapes directly during spreading on fibronectin. The effect can be quantified using the parameter dispersion (Dunn and Brown, 1986). We have found that stellation can be prevented by bovine serum and by the phospholipid mitogen, lysophosphatidic acid (Moolenaar, 1991).

Cell culture, preparation of fibronectin and spreading on fibronectin-coated coverslips were described previously (Edwards et al., 1991). After trypsinization, cells were pelleted twice from serum-free, Hepes-buffered, Hanks’ solution (HH). 5 × 10⁴ cells in 2 ml HH were added to each 22 × 22 mm fibronectin-coated, haemoglobin-blocked coverslip in a 35 mm diameter polystyrene dish. The cells were dispensed at 37°C, and incubated undisturbed, to avoid panning to the centre. In experiments with more than 0.5% serum (by volume), coverslips were quickly rinsed with warm HH before fixation to avoid precipitation of serum by the fixative. Cells were fixed by adding 2 ml 4% formaldehyde in phosphate-buffered saline (PBS), pH 7.2, stained for 15 min with Kenacid Blue (0.1% in water:methanol:acetic acid, 50:50:7 by volume), rinsed twice with water, dried and mounted in Gurr’s Clearmount.

Images from a ×50 objective of a Leitz Ortholux microscope were digitized to 512 × 256 pixels and 64 grey shades, usually with a Hamamatsu Vidicon C1000 camera and Archimedes digi-tizer (Watford). A simpler camera (National WV 1300AE/B, Mat-sushita) yielded indistinguishable results. A routine specially written in Acorn Risc Machine assembler for an Acorn Archimedes 310 microcomputer, was used for analysis. This subtracts background shading, identifies cell outlines and rejects objects smaller than unspread cells. It then calculates area, dispersion and elongation, as described by Dunn and Brown (1986). The system gave accurate values for these parameters from the model shapes in Brown et al. (1989). All cells found in a field-wide strip were analysed, unless in contact with others. The system stores parameters as fast as fields can be positioned and focussed.

For actin staining, cells were seeded on fibronectin-coated coverslips, 13 mm diameter. Spread cells were rinsed twice with 2 ml HH, fixed with 1 ml 4% buffered formaldehyde in PBS and permeabilized for 15 min with 1% Triton X-100 in PBS. After draining, each was incubated for 20 min at 37°C with 50 μl 0.2 μM TRITC-phalloidin. After two rinses with PBS, cells were mounted in 50% (by volume) glycerol/water, viewed and photographed with a ×50 objective and rhodamine optics (Vickers M 20 Photoplan).

Time-lapse video recording was with a phase-contrast ×40 objective, and ×24 time compression, using a Panasonic S-VHS time-lapse video cassette recorder. Cells were first seeded in HH, without additions, onto fibronectin-coated coverslips in dishes. After 5 min at room temperature, coverslips were inverted to form the upper face of a filming chamber, which contained HH with appropriate additions. The chamber was placed on the microscope stage, maintained at 37°C by an air-curtain incubator, and VTR recording started immediately.

Suramin was a gift from Dr J. Kusel, Department of Biochemistry, University of Glasgow. Invasive adenylate cyclase preparations, a gift from Dr R. Parton, Department of Microbiology, University of Glasgow, were crude urea extracts of Bordetella pertussis, over-producing strain BP348 (pRMB1), prepared as described by Brownlie et al. (1988), and extensively dialyzed against 100 mM NaCl, 1 mM CaCl₂, 10 mM Tris-HCl, pH 7.5, immediately before use. A similar extract of B. pertussis strain BP348, a transposon mutant with no detectable adenylate cyclase activity, was used as a control. Bovine serum albumin (Fraction V, essentially fatty acid free), bradykinin, cholera toxin, N⁶,2′-O-dibutyryladenosine 3′:5′-cyclic monophosphate (dibutyryl-cAMP), 3-isobutyl-1-methylxanthine (IBMX), insulin, 5′-deoxy-5′-methylthioadenosine (MTA) and vasopressin, from Sigma, were used from stocks in HH. Cytochalasin D, (Sigma), was dis-solved in ethanol, and lysophosphatidic acid (oleoyl, also from Sigma) in 50% (v/v) ethanol/water. Matching volumes of ethanol were used in controls. TRITC-phalloidin from Molecular Probes, was dissolved as described by the manufacturer. PDGF was from Life Technologies.

Time-lapse video microscopy showed that when BHK21 cells spread on fibronectin in the presence of agents which increase cAMP, stellate or arborized shapes were generated by outgrowth of a number of narrow processes (Fig. 1). Each was led by a fan of ruffled membrane, narrower but otherwise resembling the leading lamella of a locomoting fibroblast. These processes were sometimes reminiscent of neurites and their growth cones, as noted for similar processes induced by MTA (Smalheiser, 1989). The active lamellar margin appeared to divide into a variable number of discrete segments, instead of expanding essentially iso-diametrically as it does for example in the presence of serum. Branching was frequently observed, and rapid movements of particles sometimes seen in both directions along the narrow processes. Fig. 2 shows a BHK21 cell and other so-called fibroblastic cells fixed during spreading and stained with Kenacid Blue as described for shape measurement. The ruffles stained strongly for protein, and for actin with rhodamine-phalloidin (Fig. 3).

To investigate the effect of various agents on cell shape, particularly at varying concentrations, it is useful to have some means of measuring shapes. We measured the parameter dispersion (Dunn and Brown, 1986), using automated detection of cell outlines, rather than hand tracing. Examples of the detection and parameters for some typical cells are shown in Fig. 4. A difficulty is that as cells spread, their dispersion increases (as seen in Fig. 6), reflecting progressive shape changes occurring during spreading. Rather than use statistical methods to disentangle simultaneous effects on area and dispersion, we used conditions where agents caused increases in mean dispersion, accompanied by unchanged or, as typical for strong stellation, somewhat decreased mean areas. Fig. 5 shows typical examples of the distribution of dispersion versus area, for populations of cells in control, stellating and serum-inhibited conditions. In addition to showing the expected high dispersions corresponding to stellate shapes induced by raising cAMP, Fig. 5 also shows that serum reduced dispersion below the control. Here the value of measuring shapes was particularly evident. The latter shape change was more subtle than disappearance of obviously bizarre stellate shapes, and only with hindsight did the increased proportion of very compact shapes in the presence of serum become apparent to inspection. Dispersion/area distributions measured at intervals during spreading showed the increased dispersion of incompletely spread cells (Fig. 6), confirming that stellation occurred via a segmentation of the normal spreading process, and not from retraction of previously spread cytoplasm.

Measurements of mean dispersion, confirmed by visual observation, showed that a variety of agents expected to raise cAMP levels all induced stellation of BHK21 cells in the absence of serum (Table 1). Besides the expected dibutyryl-cAMP, the phosphodiesterase inhibitor IBMX and cholera toxin, these included the invasive adenylate cyclase from Bordetella pertussis, which crosses animal cell membranes, and MTA, as previously observed by Smalheiser (1989).

Stellation of BHK21 cells was completely abolished by colchicine (Fig. 7A) and vinblastine (not shown), confirming the expected dependence of the elongated protrusions0020on microtubules. These agents also substantially decreased the projected areas attained during spreading on fibronectin, suggesting that microtubule polymerisation contributes to the outward drive of cell margins during spreading, but does so in co-operation with other forces. In many similar systems, treatment with cytochalasins has been reported to mimick the effect on shapes of raising cAMP levels. We were unable to obtain this effect with BHK21 cells spreading on fibronectin: cytochalasin D at concentrations greater than 1 μg/ml completely inhibited spreading (not shown), but even at partially inhibitory concentrations no characteristic stellate shapes were observed, or higher values of dispersion obtained than expected for the reduced areas (Fig. 7B).

Earlier studies with CHO cells described a different shape response to elevation of cAMP (long-term, during culture), sometimes described as “reverse transformation”. We wanted to resolve whether the difference between this and the stellation seen with BHK21 cells was a matter of cell type, or of conditions. We found that CHO cells also became stellate during spreading on fibronectin in the absence of serum, as did 3T3 cells. The increase in mean dispersion of CHO cells was measured above a significantly lower control, reflecting the epithelioid appearance of these cells on fibronectin (Table 2).

Bovine sera (both new-born and foetal) at the concentration usually used in culture media (10% by volume) abolished stellation induced by raising cAMP levels (and further reduced the dispersion of cells in HH only). The activity of serum could be detected at 0.05% by volume (Fig. 8). Serum acts from solution, rather than as a component of the substratum-adsorbed protein: if fibronectincoated, haemoglobin-blocked surfaces were pre-incubated with serum (even 10%) and then rinsed, subsequent stellation on them was little inhibited (Fig. 9). The activity is macromolecular, since it is retained on dialysis (not shown).

We speculated that the stellation-reversing activity of serum might be due to a mitogenic hormone or growth factor. The activity of a number of growth factors such as PDGF can be inhibited in cell cultures by the antifilarial agent suramin, which apparently inhibits binding of growth factors to their receptors (Betsholz et al., 1986). Suramin did indeed competitively inhibit serum from preventing stellation (Fig. 10). An initial screen of various mitogens failed to identify one which reproduces the effect of serum (Table 3). Vasopressin reversed stellation by IBMX in rat mesangial cells (Glass et al., 1988) but appeared inactive with BHK21 cells. However, following the description of its activity in stimulating stress fibres in 3T3 cells (Ridley and Hall, 1992) we found that lysophosphatidic acid, in the presence of bovine serum albumin, was as active as serum in preventing cAMP-induced stellation and inducing a compact morphology indistinguishable from that induced by serum (Table 3).

This work has two distinguishing features. Firstly, we have measured the shape change objectively using the parameter dispersion. This is particularly advantageous with nonextreme shapes and when the response varies between cells in a population. Secondly, we have used the most defined medium and substratum practicable: a salts solution with buffer and glucose, and adsorbed fibronectin. This prohibited longer-term experiments, but allowed detection of the effect of small amounts of serum.

Stellation by cholera toxin and the invasive adenylate cyclase of B. pertussis strongly confirm that dibutyryl-cAMP and phosphodiesterase inhibitors such as IBMX affect shape by elevating cAMP levels. This activity of the adenylate cyclase promises to provide a useful assay for its ability to penetrate animal cell membranes. MTA was previously shown to induce neurite-outgrowth from neural hybrid NG108-15 cells (Smalheiser and Schwartz, 1987), and similar processes from 3T3 cells and human foreskin fibroblasts (Smalheiser, 1989). MTA raised cAMP levels in S49 mouse lymphoma cells (Riscoe et al., 1984), apparently acting as a phosphodiesterase inhibitor, and could act in the same way to stellate fibroblasts. However stellation by 3 mM MTA is not prevented by 2% serum, unlike both dibutyryl-cAMP and B. pertussis cyclase (data not shown). This may explain why it could cause stellation in growth medium (Smalheiser, 1989), and suggests it may act by a different mechanism, such as by inhibiting protein kinase C (Smalheiser, 1990).

Similar shapes to those caused by cAMP elevation can be induced by cytochalasins (Lawrence et al., 1979; Chaldakov et al., 1989; Bliokh et al., 1980; Phaire-Washington et al., 1980), and by lovastatin (Fenton et al., 1992). Stellation by elevation of cAMP itself is accompanied by depletion of microfilaments, particularly loss of stress fibres (Westermark and Porter, 1982; Nabika et al., 1988; Chaldakov et al., 1989; Baorto et al., 1992). These observations have led to the view that stellate shapes result from depletion of actin microfilaments. In agreement with Baorto et al. (1992) for astrocytes, we find that outgrowth of processes in stellate spreading of BHK21 cells is accompanied by strong staining for F-actin at the tips of growing processes. Thus any effect of cAMP elevation on actin must be selective for different subsets of microfilaments. Loss of F-actin selectively, eg. from the cortical web, could leave a plasma membrane which weakly resists stretching.

Effects of cAMP on microfilaments may result from dephosphorylation of myosin light-chain kinase (MLCK), causing relaxation of myosin-dependent tension in smooth muscle, and perhaps destabilizing microfilaments (Lamb et al., 1988; Egan et al., 1991; Baorto et al., 1992). Inhibitors of MLCK can themselves induce stellation (Baorto et al., 1992). Dephosphorylation of a 19 kDa actin depolymerizing factor may also be involved (Baorto et al., 1992).

Bliokh et al. (1980) and Baorto et al. (1992) observed outgrowth of narrow processes from cells spreading in the presence of cytochalasin, suggesting that actin polymerization at the ruffling tips, although it accompanies outgrowth, is not essential for it. Under our conditions, with and without elevation of cAMP, spreading was totally inhibited by cytochalasin D. Our failure to observe stellate spreading in cytochalasin may be because such outgrowth needs a longer time, or requires components we excluded from the medium or substrate.

In addition to the strong correlation between stellation and disappearance of sub-membranous actin, the likely importance of microtubule stabilization has also been emphasized, particularly in relation to neurite outgrowth (Heidemann et al., 1985). However, extracellular matrix allowed neurites to elongate spontaneously from PC12 cells in the absence of microtubules (Lamoureux et al., 1990).

Bovine serum has occasionally been found to oppose the formation of narrow cell processes, either formed spontaneously or induced by elevation of cAMP, particularly by glial or neural-related cells (Lim et al., 1973, Edstrom et al., 1974; Moonen et al., 1975; Ghahary et al., 1989). It reversed a novel dendritic morphology formed by spreading amnion epithelial cells (Campbell et al., 1984). Induction of stellate shapes by elevation of cAMP levels in glomerular mesangial cells (Kreisberg et al., 1984) and vascular smooth muscle cells (Nabika et al., 1988; Chaldakov et al., 1989) was studied in serum-free media, whereas with many other cell types, stellation has been obtained in the presence of serum (Lawrence et al., 1979; Westermark and Porter, 1982; Preston et al., 1987; Lamb et al., 1988; Baorto et al., 1992). We have found that serum suppresses cAMP-induced stellation of BHK21, 3T3 and CHO cells, which may explain at least in part why stellation was not reported in the earlier transformation-related studies of cAMP effects.

The identity and mode of action of the factor or factors present in serum which inhibit stellation of fibroblasts on fibronectin are presently unknown. The observation that serum can be replaced by lysophosphatidic acid in the presence of serum albumin strongly suggests that the activity of serum may itself be due to endogenous phospholipids bound to serum albumin. One feature we have identified is reversal of the effect of serum by suramin. Suramin reverses the effects of both lysophosphatidic acid and phosphatidic acid as mitogens (van Corven et al., 1992). Lysophospha-tidic acid lowers cAMP accumulation in intact cells (van Corven et al., 1989). However the level at which signals from lysophosphatidic acid oppose the effects of cAMP to suppress stellation, presumably via effects on the actin cytoskeleton, remain to be investigated. Identification of the stellation-inhibiting factor in serum, and of its mode of action, will be of considerable interest in relation to such phenomena as control of locomotion in general, and wound-healing in particular.

Our results confirm and extend a consistent pattern (summarized in Fig. 11) of effects of cAMP on the outgrowth of narrow elongated processes from cells in vitro. Raising cAMP favours such outgrowth and is often antagonized by the presence of bovine serum. Sometimes obtained in response to hormones, stellation should probably be viewed as an abnormal response to a normal stimulus (Westermark and Porter, 1982). The diversity of responsive cell types suggests the involvement of a fundamental mechanism of regulation of filament/membrane events, perhaps involving relaxation of cortical actin, which seems to operate in vitro regardless of the appropriateness of the resultant shapes. Stellation may be important for what it reveals of the interplay of these structures in normal spreading and locomotion.

We thank Amanda Morton and Fahd Al Mulla for data obtained during their final year Honours projects in Molecular Biology; Martin O’Hare for help with LPA experiments, and Dr Roger Parton, Department of Microbiology, University of Glasgow, for B. pertussis adenylate cyclase. The work was supported by University of Glasgow general funds and by a vacation award to M. Carr from the Yamanouchi Research Institute, Oxford.