The aggregation of trypsinized BHK21 cells

An aspect of animal cell interactions which is beginning to receive increasing attention is that of cell-to-cell adhesion. The aim of these studies must be to relate the adhesive behaviour of cells to the molecular constitution of cell surfaces. It is apparent that for such studies there are a number of advantages in the use of cells of permanent lines grown in culture. For example, one can work with extremely homogeneous cell populations, manipulate culture conditions, and readily label surface macromolecules isotopically.

The work described in this paper stems from the observation, common in cellculture practice, that when cells grown on solid surfaces are dispersed with trypsin (for example, for re-seeding) they tend to adhere to one another in suspension, to form clusters. This is usually regarded as an inconvenient artifact, perhaps resulting from the action of the enzyme. We have been examining this phenomenon, using BHK21 cells (Stoker & Macpherson, 1964). We have adopted the attitude outlined by Curtis & Greaves (1965), namely that the kinetics of cell aggregation over short times can provide a measure of the instantaneous adhesiveness of the cells. Our observations suggest that the adhesiveness responsible for aggregation of these cells may not be an artifact of tryptic action, but could well be a property of the cell surface which survives the dispersal procedure. Since they may therefore have a bearing on the mechanism by which the cells adhere in vivo, we have investigated in detail 2 features of the aggregation in suspension: temperature dependence of the rate, and sensitivity of the adhesions formed to proteolytic enzymes.

Some earlier observations on the aggregation of cultured cells have been reported by Moskowitz, Ambroski &Wieker (1966). These are concerned with forms of aggrega tion observable only after much longer times and under quite different conditions from the work described here.

Cells of the hamster fibroblast line BHK21, clone C 13, were grown as monolayers (area 120 cm²) on glass. The growth medium was modified Eagle’s, 8 vol., tryptose phosphate broth, 1 vol., calf serum, 1 vol. To disperse the cells for aggregation experiments, freshly confluent monolayers were washed twice with 30 ml tris-saline (25 mM tris-HCl, pH 7·4; 0·14 M NaCl; S mM KC1; 0·7 mM Na₂HPO₄). The cells were then covered with 10 ml EDTA-trypsin, at room temperature (2·5 mg per ml Difco 1·250 trypsin in tris-saline, 1 vol.; 0·55 mM EDTA in phosphate-buffered saline (Dulbecco & Vogt, 1954), 4 vol.). This solution was assayed from time to time for tryptic activity by the method of Schwert & Takenaka (1955), and found to contain 250–300 BAEE units per ml. When the first cells started to detach (after about 2 min) the EDTA-trypsin was replaced with 20 ml tris-saline, and the cells shaken off into this medium. Subsequent operations were carried out at 2 °C. The cells were centrifuged, re suspended in tris-saline and pipetted to disperse residual aggregates. After a second centrifuga tion, the cells were resuspended in Hanks’s solution plus 5 mM tris-HCl, pH 7·2, at a density of 1 ·0 ×10⁶ cells per ml. Cells of the C13 derivative, PyY/AA/AAR/TG, were kindly provided by Professor Subak-Sharpe, to whom we owe the original observation that these cells adapt particularly readily to growth in suspension. These cells were either grown on glass and handled as described for C 13 cells, or grown in suspension culture in spinner flasks in 200 ml of the same medium except that foetal calf serum was used in place of calf serum. Except where otherwise described, cells grown in suspension were washed in tris-saline, pipetted to disperse any small clusters present, and finally resuspended in Hanks’s solution.

Four millilitres of cell suspension in 10-ml stoppered, siliconized conical flasks were shaken reciprocally with a displacement of 4 cm at 92 strokes per min (Curtis & Greaves, 1965), at 37 °C except where otherwise stated. Two methods were used to follow the progress of aggrega tion. For microscopic counting, samples were removed for counting the surviving single cells on haemocytometers (modified Fuchs-Rosenthal) using phase-contrast microscopy. To confirm that decreases in cell number were not merely due to loss of cells, cells were also counted in aggregates, which is possible except when very large aggregates have formed. For electronic counting 0·2 ml of sample were diluted 100 times in ice-cold 0·9 % saline and counted with a Coulter counter model A, using a 200- μ m aperture. (The use of the Coulter counter for similar purposes has been described by Orr & Roseman (1969). Unlike these workers, however, we made no attempt to discriminate single cells from clusters, but instead measured the total number of particles of all sizes (Curtis, 1969). Under some circumstances, changes in cell volume were observed, which would have interfered with measurements of aggregation if we had attempted to use critical pulse-height windows.) Standard coincidence corrections (always less than 10%) were applied.

Methyl cellulose (Methocel (Dow) Standard grade, 4000 cP (4N s m’⁻²)) in Hanks’s-tris, was prepared as described for growth medium by Stoker, O’Neill, Berryman & Waxman (1968).

To test the effectiveness of enzymes in dispersing aggregates, cell suspensions were shaken at 37 °C until the particle counts were stationary, normally after 40 min. The enzymes were then added, and shaking continued. Disaggregation could then be measured by following the in crease with time in single-cell or particle numbers. The eventual recovery of single cells was usually less than 80 %. This was largely because proteolytic enzymes lyse a fraction of the cells (10–15%), apparently the same cells which show low phase-contrast and stain with trypan blue.

Di-isopropylphosphoryl trypsin-³²P (DIP-trypsin) was prepared by incubating 3 × 10^{− 5} M trypsin (Sigma twice recrystallized, 11000 units/mg) for 2 h, at about 20 °C with 3 × 10⁻⁴ M di-isopropyl phosphofluoridate-³²P (DFP, The Radiochemical Centre, Amersham, England) in 0·1M phosphate buffer, pH 8·0. Protein-bound³²P was estimated by precipitation of samples of the protein with 5 % trichloroacetic acid (TCA) followed by collection of the precipitates on Millipore filters for liquid scintillation counting. The half-time for incorporation of³²P to the trypsin under these conditions was 10 min. Excess DFP was removed by gel-filtration on a column of Sephadex G25. The radioactivity of the DIP-trypsin corresponded to 0·65 mol phosphorus per mol trypsin as weighed and was all TCA-precipitable. Its enzymic activity was less than 1 unit/mg.

The cell-separating activity in Difco trypsin was purified as follows: 2·5 g Difco (1·250)trypsin was stirred at 2 ° C for 30 min in 20 ml 0·02 M sodium acetate buffer, pH 5·5. Undis solved material was removed by centrifugation, and the supernatant de-salted by gel-filtration on Sephadex G 15, pre-equilibrated with acetate buffer. The excluded fractions were pooled and run on to a carboxymethyl cellulose column (Whatman, CM 52; column dimensions 2 × 40 cm) also pre-equilibrated with the acetate buffer. The proteins were then eluted with 1 1. of this buffer containing a linear gradient 0–0·25 M sodium chloride.

The deoxyribonuclease (DNase) used was pancreatic Deoxyribonuclease 1 Worthington, 2000 units/mg, and was assayed by the method of Kunitz (1950). The chymotrypsin was Sigma β-chymotrypsin, 42 BTEE units/mg.

Fig. 1 shows that the electronic counting technique responds linearly to decreases in particle number as cells aggregate, over a much greater range than is used in these studies. This shows that cell clusters are not disaggregated by dilution in saline, neither do they give rise to multiple pulses in the sensing-zone of the instrument. The standard deviation for a set of 12 dilutions of a single aggregate suspension was 3 % (for microscopic counting about 7 %).

C13 cell-suspensions, prepared as described above, aggregate rapidly when shaken at 37 °C, to form a population of small clusters, with a stationary size-distribution. An example of the distributions obtained is shown in Fig. 2. The final extent of aggregation varies considerably between different cell suspensions (as discussed below) but for a given suspension the extent of aggregation is rather insensitive to the initial density in the range tested, i.e. from 0·0·10⁵ to 1·5 ×10⁶ cells per ml. The half-time for decrease of the particle count to its final value is approximately inversely pro portional to the initial density, as expected for a collision-dependent process. In other words, a suspension aggregated at an initial density of 10⁵ cells per ml will yield a similar size-distribution of clusters as at 10⁶, but will take about 10 times longer (Fig 3)

The extent to which cells grown on glass aggregate after trypsinization increases conspicuously as the density of the cells increases in the culture from which they were obtained. (This applies equally to CIJ cells (Fig. 4) and suspension-adapted PyY cells returned to growth on glass.) Much of the increase takes place as the cells are approaching confluency (at around 2 × 10⁵ cells per cm²). The cells continue to grow beyond this density, but it then becomes difficult to obtain single cell-suspensions with the standard dispersal procedure. Cells grown for several passages in suspension do not aggregate detectably at early times when washed and resuspended in Hanks’s solution. It was thus of interest to test the effect of the trypsinization procedure on these cells. Accordingly, suspension-grown cells were washed, placed in fresh medium and plated out on glass. After they had attached and spread they were dispersed and tested for ability to aggregate. Very little aggregation was detected (Fig. 5). However, cells from suspension culture seeded thinly and allowed to grow on glass regain the ability to aggregate during the first passage (Fig. 6).

Trypsinized C13 suspensions aggregate very slowly, if at all, at 2 °C (Fig. 7). Sincethe cell-suspensions are more viscous at 2°C, it was possible that the difference in aggregation behaviour between the 2 temperatures might reflect the altered hydro dynamic conditions (Curtis, 1969). To test this the viscosity of a cell suspension was measured at both temperatures; the values found were essentially those of water. A concentration of Ficoll was found which when added to a suspension at 37 °C increased its viscosity to the value for 2 °C. Aggregation at 37 °C was then measured in the presence of this concentration of Ficoll. It is clear from Fig. 8 that the viscosity difference cannot account for the failure of the cells to aggregate at 2 °C, provided the possibility is discounted that Ficoll itself has a very large effect on cell adhesiveness. The same conclusion is reached using methyl cellulose to adjust the viscosity. We cannot detect a delay in aggregation after shaking starts at 37 °C. Decrease in single cell number is readily measurable after 2 min; shorter times would in any case fall within the interval of temperature equilibration. Furthermore, aggregation stops immediately on transfer to 2 °C. To test whether pre-incubation at 37 °C confers on cells the ability to aggregate at 2 °C, ideally we should have incubated cells at 37 °C without permitting aggregation. This has not so far proved possible, because there is some aggregation in both very rapidly shaken and stationary flasks. However, by keeping cells stationary in a siliconized tube, or shaking them at 140 cycles/min in an aggregation flask, one observes in 10 min aggregation equivalent to only about 2 min at 92 cycles/min. Figs. 9 and 10 show that there is little further aggregation of such cells when transferred to the cold. These experiments also show that there is no net disaggregation of cells at 2 °C. Conversely, cells shaken at 2 °C do not lose the ability to aggregate when transferred to 37 °C.

Another approach to this has been the use of methyl cellulose to suppress the aggregation of the cells. If cells are pre-incubated for 30 min at 37 °C in Hanks’s solution containing 1·6%(w/v) methyl cellulose, washed twice and resuspended in Hanks’s solution, the temperature-dependence of subsequent aggregation is un affected (Fig. 11). Experiments of this type, however, are perhaps less useful than the former because they do not exclude the possibility that the 37 °C-dependence is for a rapid recovery from the trauma of centrifugation and resuspension, or for an altera tion to the medium which occurs only at 37 °C.

To investigate the effect of trypsin on aggregates we had to add DNase to cell suspensions. Threads of gel-like material can arise in suspensions of trypsinized cells, apparently as a result of interaction between trypsin and the DNA of lysed cells (Steinberg, 1963). We sometimes saw this in our suspensions, but only when further trypsin, without DNase, was added after the washing procedure. It is not likely that such material is responsible for the aggregation of the cells, since the extent of aggrega tion was unaffected by the presence of 100 μg /ml DNase. When DNase was present, disaggregation by trypsin was detectable with as little as 1 to 3 BAEE units per ml of cell suspension (Figs. 12, 13). The sensitivity of aggregates to trypsin was very similar in terms of BAEE units (as opposed to total protein concentration) for Difco 1·250, twice recrystallized trypsin, and trypsin prepared by chromatography of Difco 1·250 (Fig. 14). This was in spite of the fact that Difco trypsin normally had 20 times lower specific activity against BAEE than the 2 purified trypsins (600 compared with 10000 to 12000 units per mg). This suggests that its effectiveness in separating cells in this system can be adequately accounted for by its trypsin content only. In chromatographic purification, most of the cell-separating activity eluted in close correspondence with the major peak of activity against BAEE. There was also some activity in the material not bound to the column, possibly due to chymotrypsin, which also hydrolyses BAEE (Fig. 15). Recrystallized chymotrypsin was also tested against aggregates, and found to have similar activity to trypsin at equal enzyme molarities (Fig. 16). The aggregates were also dispersed by pronase (personal communication from M. Vicker).

We found that DIP-trypsin would not disperse the aggregates, even when added in much larger amounts than required for the active enzyme (Fig. 17). This experiment was performed 3 times, with 2 separate preparations of DIP-trypsin.

Since the aggregates are sensitive to less than [EQ] the level of trypsin used to prepare the initial suspension, the removal of trypsin by the washing procedure is likely to be fairly complete. This is of some importance in considering whether the aggregating cells are extensively modified by bound trypsin. To investigate this point further, the trypsin used at the preparative stage was labelled by addition of a trace of DIP-trypsin-³²P. The radioactivity of the cell suspension indicated that the level of trypsin was reduced at least 1000-fold (i.e. to less than 0·3 BAEE units/ml, Table 1). DIP trypsin presumably differs very little from trypsin in general physical properties (Cunningham, 1954) so that this allows us to place an upper limit on the trypsin carry-over resulting from general (e.g. electrostatic) interactions, but there could conceivably also be active trypsin bound by an interaction depending on the integrity of its active site.

These observations show that the extent to which the cells aggregate after subjection to a standard dispersal procedure varies strikingly with their circumstances of growth. Cells of the derivative PyY/AA/AAR/TG grown on glass aggregate after trypsiniza tion as do C13, but the Py Y-derived cells grown in suspension, plated at high density on glass and trypsinized a few hours later do not. This argues against the idea that the adhesiveness is (at least in any simple way) generated by the dispersal procedure. It is important to emphasize that the clear-cut difference between the aggregation of suspension- and glass-grown cells is observed only under appropriate conditions. The suspension-grown cells will aggregate very extensively in the medium in which they have grown, and also in simple media such as Hanks’s after a delay, normally of more than i h. This aggregation shows very much less temperature dependence, and may conceivably depend on a different adhesive mechanism.

Many explanations are possible for the effects of growth density on the aggregation of both kinds of cell grown on glass. Since the aggregates formed are so readily sepa rated by proteolytic enzymes (see below), it is tempting to explain these effects in terms of protection from the action of trypsin. This would suppose that with glass grown cells, the surfaces are of uniform high adhesiveness as growth proceeds. The regions of cell-to-cell contact, presumably increasing in extent as confluency is reached, might be relatively inacessible to trypsin during the short exposure employed in the dispersal procedure. A separate reason would have to be found for the non adhesiveness of the suspension-grown cells before trypsinization.

Another explanation in terms of degree of cell contact, which would also explain the results with suspension-grown cells would suppose that most of the surface of both kinds of cells is non-adhesive (as judged by short-term aggregation in Hanks’s medium) whether exposed to trypsin or not, but that surface originating from scission of contacts established during growth is adhesive. If this is correct, it would mean that the aggregation in suspension can be regarded as reconstruction of pre-existing adhesions.

Alternatively, the growth effects may be only indirectly related, or indeed unrelated, to the extent of pre-existing cell-to-cell contacts. For example, the aggregation may depend on synthesis by cells of a specialized product, the normal function of which may not necessarily be an adhesive one. Work is now in progress which we hope will allow us to distinguish between these possibilities.

In the work of Moscona (1962) the re-aggregation of embryonic cells was examined after fairly long periods and it was found that aggregation was inhibited at low temperatures. This was explained by the proposal that the cells became adhesive only after the resynthesis of a macromolecular component which had been degraded during trypsinization. Support for this was found in the observation that pre-incubation of cells in the warm enhanced subsequent aggregation at lower temperatures.

Using cell suspensions prepared by the use of EDTA alone, Curtis & Greaves (1965), have shown that the temperature dependence of aggregation is much less marked if serum is omitted from the re-aggregation medium. This is explained by the presence in serum of a protein which inhibits aggregation at 2 °C but is inactivated at37 ° C (Aggregation Inhibiting Protein, AIP). Curtis (1967) has extended these observa tions to the behaviour of trypsinized cells by suggesting that the trypsin used may contain AlP-like activity as a contaminant, which is carried through from the trypsinization to the aggregating suspension. In these systems, cells on first experi encing 37 °C after dispersal should be of low adhesiveness, as shown by a delay before aggregation. Furthermore cells incubated at 37 °C for periods comparable with the delay should then be capable of enhanced aggregation at low temperatures.

Both Steinberg (1962) and Jones & Morrison (1969) have found that aggregation of embryonic cells is continuously temperature dependent, in that the rate drops on cooling, irrespective of prior incubation of the cells at 37 °C, so that a process of synthesis of macromolecules is insufficient to explain the behaviour.

In the aggregation of trypsinized C13 cells, there is no detectable delay before single cells start to form aggregates, and no enhancement of aggregation at 2 °C after incubation at 37 °C. The temperature dependence is thus of a continuous type, and cannot be explained solely in terms of phenomena such as those postulated by Moscona (1962) and Curtis & Greaves (1965). Instead, the aggregation responds to temperature in the manner described by Steinberg (1962) and Jones & Morrison (1969). Thus, if particular macromolecular species are involved in the adhesion of these cells, they are presumably already synthesized at the beginning of aggregation. It is the expression of their adhesive property which is temperature dependent.

However, we believe that caution should be exercised in attributing this temperature dependence directly to formation of bonds between surface macromolecules, as Jones & Morrison (1969) have suggested. A wide variety of temperature-dependent pro cesses, such as for example the maintenance of an intracellular pool of a substance having a regulatory effect on adhesion, might confer temperature dependence on the rate of aggregation. Moreover, unless such indirect effects of temperature on cell aggregation can be ruled out, temperature dependence cannot be used as an argument to favour a particular mechanism of adhesion. In particular, it cannot be held to invalidate the theory that interactions in the primary or secondary minimum of the lyophobic colloid theory determine the kinetics of cell aggregation (Curtis, 1967).

The technique we have used for observing the effects of proteolytic enzymes on aggregated suspensions, namely measurement of the rate of disaggregation of the shaken suspension, provides a useful new assay for the sensitivity of adhesions to enzymes, which can clearly be used to test a wider variety of enzymes. We have been concerned particularly with proteolytic enzymes, because of the use of trypsin in preparing the cell suspensions. Aggregates of BHK21 cells, originally dispersed by the use of trypsin, are themselves sensitive to redispersal by trypsin. The ineffective ness of DIP-trypsin indicates that it is in fact enzymic activity which results in redispersal. The same conclusion was reached by Easty & Mutolo (1960) for the dispersal of adult rat-liver and Walker ascites-tumour cells by trypsin. On the other hand, these workers found that trypsin inactivated with DFP was capable of dis persing ‘artificial tissues’ prepared by agglutinating erythrocytes with synthetic basic polymers and protamine. They attributed this to a non-proteolytic interaction between trypsin and the agglutinating polymer, and it was important to exclude the possibility of such an effect in our system.

The similar effectiveness of trypsin and chymotrypsin rules out the possibility that the cell separation results from hydrolysis of a particular trypsin-sensitive bond. The levels of tryptic activity which disperse the BHK aggregates are similar to those found by Weiss (1963) to facilitate separation of various cultured cell types from glass, on the assumption that it was in fact the trypsin component of Difco trypsin which was responsible for the effects he observed.

The observation that the aggregates are dispersed by low levels of trypsin somewhat increases our confidence that the mechanism of adhesion by which they are formed is not an artifact of the original use of trypsin to disperse the cells from the monolayer. Such an artifact is a genuine possibility because in spite of the general efficacy of such enzymes in separating cells, proteolytic enzymes are also known to be able to increase the adhesiveness of cells in some circumstances. For example, they promote the agglutination of human erythrocytes sensitized with incomplete antibodies (Morton & Pickles, 1947; Pollack et al. 1965). In this case they probably act by removing surface sialate residues, thereby reducing the net negative surface-charge. It is also possible to argue that the traces of trypsin carried over to the aggregating suspension could cause the aggregation at 37 °C by being bound to the cells, and acting as a bridging reagent. It would be necessary to postulate further that the binding is rapidly reversed at 2 °C, which though unlikely is possible in principle. The sensitivity of the aggre gates to trypsin is most easily explained on the assumption that it reflects an adhesive property of the surface of the cells before dispersal which was always potentially trypsin sensitive, but survived, at least in part, the initial trypsin treatment.

Irrespective of this problem, the question arises whether the sensitivity of these aggregates to proteolysis is consistent with their adhesion being solely by interactions of the kind postulated in the lyophobic colloid theory of cell adhesion (Curtis, 1967; Brooks, Millar, Seaman & Vassar, 1967).

The dispersal of adhering cells by proteolytic enzymes is often presented as evidence that intercellular protein bridges are responsible for cell adhesion. While this may be the case, it should be pointed out that proteolytic enzymes could increase the electro static repulsion between cells. For example, they could unmask buried sialate residues (Kraemer, 1967a) either by removing other materials or by bringing about configura tion changes similar to those postulated to explain the cell-cycle dependent variation in electrophoretic mobility (Kraemer, 19676). Such increases could account for the dispersal of aggregates in terms of the lyophobic colloid theory.

We are indebted to Professor H. Subak-Sharpe of the Institute of Virology, University of Glasgow, for providing clone C13 and PyY\AA\AAR\TG cells, and to Professor A. S. G. Curtis and Mr M. Vicker for many discussions and for reading this manuscript. We thank Miss F. McFarlane and Mr P. Hannah for excellent technical assistance. This work was supported by the Science Research Council (grant 4909).