J. Cell Sci. 69, 35-45 (1984) 35Printed in Great Britain © The Company of Biologists Limited 1984

MECHANICAL STRETCHING INCREASES THENUMBER OF EPITHELIAL CELLS SYNTHESIZINGDNA IN CULTURE

D.M.BRUNETTEFaculty of Dentistry, University of British Columbia, 2199 Wesbrook Mall, Vancouver,B.C., Canada V6T IZ7

SUMMARY

The influence of mechanical stretching on epithelial (E) cells was examined by culturing E cellsderived from the epithelial cell rests of Malassez on a flexible plastic substrate and stretching thesubstrate by means of an orthodontic screw. A significant increase in the number of E cells synthesiz-ing DNA was observed after just 30min of stretching. In 17 experiments the ratio of cells labelledwith tritiated thymidine in cultures stretched for 2h to the number of labelled cells in unstretchedcontrols was 1-92 ± 0-34. An increase in labelling as a result of stretching was found for E cellscultured at either high or low cell-population densities but the effect was most pronounced for E cellscultured at higher concentrations of foetal bovine serum. Morphometric analysis of electronmicrographs of stretched and unstretched cultures indicated that the stretched cultures had a highervolume fraction of filamentous structures and more desmosomes per unit length of cell membranethan unstretched cultures. The behaviour of E cells in response to stretching in vitro appears to besimilar to the response of the epithelial rests in vivo when the latter are exposed to tension as a resultof forces produced by orthodontic techniques.

INTRODUCTION

The importance of physical stresses on a variety of cellular activities has beenrecognized in many diverse systems including the developing frog embryo(Beloussou, Dorfman & Cherdantzev, 1975), moulting in the blood-sucking bugRhodnius prolixus (Wigglesworth, 1935), fibroblasts in culture (Curtis & Seehar,1978), smooth muscle cells (Leung, Glag & Mathews, 1976) and cranial sutures(Meikle, Sellers & Reynolds, 1980; Meikle, Reynolds & Dingle, 1979). For anumber of tissues in vivo, mechanical stress may lead to increased cellular division(Curtis & Seehar, 1978; Abercrombie, 1957), and Curtis & Seehar (1978) haveproposed that several types of experiments in vitro, which have been interpretedas evidence for the control of cell division by the action of diffusion boundary layers,could equally well be interpreted as being effects of mechanical action. Although thereis evidence that the stretching of mouse skin in vivo results in an increase in the mitoticindex (Squier, 1980), the interpretation of such in vivo experiments is limited by thefact that it is difficult to determine whether tension acts directly on the epithelium oraffects some other cell population whose products or activities result in epithelialproliferation. In this report I present evidence to show that mechanical stretchingcauses a rapid increase in the number of epithelial cells synthesizing DNA in cellculture.

36 D. M. Brunette

MATERIALS AND METHODS

Cell cultureEpithelial cells derived from porcine rests of Malassez were obtained from periodontal ligament

and cultured by the method described by Brunette, Melcher & Moe (1976). This technique resultsin the growth of both epithelial cells (E cells) and fibroblast-like cells (F cells). The two cell typescould be separated by the method of Kanoza, Brunette, Purdon & Sodek (1978)T using bacterialcollagenase applied to mixed cell populations growing on a collagen gel, or by the tendency offibroblasts to be less resistant to detachment by trypsin (Owens, 1974). The PPL-E cells were grownin alpha MEM (Flow, McLean, VA) supplemented with antibiotics (penicillin G (Sigma, St Louis,MO; 100/ig/ml), gentamicin (Sigma; 50/ig/ml), amphotericin B (Gibco, Grand Island, NY;3f/g/ml)) and 15% foetal bovine serum (FBS; Flow) on Petriperm (Tekmar, Cincinnati, OH)dishes at 37 °C in a humidified atmosphere of 95 % air/5 % COz. Unless otherwise noted, stretchingwas applied to cells that had been seeded at a density of 2x 104 cells/cm2. Tritiated thymidine([3H]dThd) (New England Nuclear, Boston, MA) at a concentration of 1-0/iCi/ml and a specificactivity of 72Ci/mmol was added for various periods of times as noted.

Application of mechanical stretchingA method similar to that adopted by Harell, Dehel & Binderman (1977) was used to apply

mechanical stress to the cultures. This was achieved by embedding an orthodontic screw (Westdent,Toronto; cat. no. 620-12 expansion 12 mm Leone) in acrylic resin moulded to fit a Petriperm culturedish. The Petriperm dish is used because the bottom is made of a flexible plastic membrane that canbe stretched by turning the screw. The amount of stretching was not uniform throughout the dish,but in the central area where observations are made there was a 4-2 % increase in length in the axisalong which the force was applied. Although the stretched Petriperm dish does not have ideal opticalqualities, phase-contrast micrographs of stretched and unstretched cultures indicated that a similarincrease in length of clusters of E cells occurred. Control dishes were treated identically except thatthe orthodontic screw was not turned.

Preparation of autoradiographsAfter incubation with [3H]dThd, the E cells were rinsed twice with phosphate-buffered saline

(PBS) and fixed with 3 % glutaraldehyde in PBS at 4 °C for 1 h. The fixative was then removed andthe culture was washed twice with distilled water and extracted three times for lOmin with 5 %trichloroacetic acid and finally washed twice more with distilled water.

A difficulty was encountered in using the Petriperm dishes, concerning the mounting of the plasticmembrane for preparation of autoradiographs. The following procedure was used. Glass slides wereattached to the bottom of the Petriperm dishes on the surface to which the cells were not attached(non-cell side) using Epon (Ladd, Research Industries, Burlington, UT). The dishes were thenplaced in an oven at 60 CC for 2 days until the Epon hardened and the slides were firmly attachedto the dishes. Then the part of the Petriperm dish attached to the glass slide was cut out using aheated scalpel. The slides were put into slide holders, dipped in emulsion (NTB2; Kodak, Roches-ter, NY), and processed by a method described previously (Gould, Melcher & Brunette, 1977).Following development the cells were stained with Gill's haematoxylin.

Electron microscopyE cells were processed for electron microscopy by a slight modification of the method of Ellisman

& Porter (1980), which is designed to demonstrate cytoskeletal architecture. Cultures were washedtwice with Millonig's (1961) phosphate buffer at room temperature and fixed for 30 min at 4 °C with2-5 % glutaraldehyde in phosphate buffer. After two further rinses with phosphate buffer the cellswere post-fixed with 1 % OsO* supplemented with 1 % tannic acid. After dehydration through agraded series of ethanol (to 70 %) the E cells were stained en bloc with 5 % uranyl acetate in 70 %ethanol and dehydrated again through 70% ethanol to absolute ethanol at room temperature.Subsequently the cells were infiltrated and embedded with an Epon/Araldite resin, which wasallowed to polymerize by incubation at 60°C for 48 h. Ultrathin sections were cut, mounted on

Epithelial cell response to stretching 37

carbon/colloidon-coated 200 mesh grids, double-stained with aqueous 5% uranyl for lOmin andlead citrate for 5 min, and subsequently observed with a Philips 300 electron microscope.

QuantitationNine fields were counted to determine the labelling index for each culture. The fields were

selected by a stratified random-sampling procedure in which the fields are selected blindly but covera specified spatial distribution. This is achieved by placing the slide on a mechanical stage andlocating the central area where the stretching is greatest. Then the microscope is focused and thenumbers of labelled and unlabelled cells are counted. The stage is then moved a predetermineddistance and another field counted and so forth. Statistical comparisons were made using theStudent's t-test.

Quantitation of the ultrastructural features of stretched and unstretched E cells was done in twoexperiments. In both experiments the areas to be photographed were selected at scanning magnifica-tion where details of cellular architecture were not apparent and the selection was made so that thecultures examined had a similar cell population density. In the first experiment, the number ofdesmosomes per unit length of membrane was determined by preparing micrographs (53 fromstretched cultures and 50 from controls) at a final magnification of 25 000. The numbers of desmo-somes in each micrograph were counted and the length of membrane profiles was determined bytracing their outline on the graphics tablet of a microcomputer (Apple, Cupertino, CA). In thesecond experiment, a point-counting method was used to determine the volume fraction ofmicrotubules and filaments. A lattice was printed directly onto the micrographs and the number ofpoints falling on microtubules or filaments was counted. It was found that, in some instances at themagnification used, it was difficult to distinguish unequivocally between microfilaments andtonofilaments, and the two types of filament were grouped in the same category. In order to obtaina realistic estimate of the precision of the volume fraction of these two categories, 300 points werescored on each micrograph so that counts of zero for either component were rare and each set of 300points was treated as a single observation.

RESULTS

In the experiments shown in Fig. 1, [3H]dThd was present continuously in theculture medium. There was a significant rise, more than a doubling, in the numberof labelled cells in the stretched cultures at 2h and at all times the percentage oflabelled cells was higher in the stretched cultures than in the unstretched controls.However, the difference became progressively smaller and after 6 h was no longersignificant. In 17 experiments the ratio of the number of labelled cells in stretchedcultures to the number of labelled cells in the unstretched controls after 2h was1-92 ±0-34 (mean ± 9 5 % confidence limits). An additional control was done todemonstrate that the increase in the number of labelled cells caused by stretching wasnot artefactual (Table 1). Both the stretched and unstretched cultures exposed to[3H]dThd at 4°C showed no labelling. Thus, mechanical stretching did not cause anincrease in background as a result of pressure or stress effects on the emulsion (Rogers,1979).

In order to determine if the effect of stretching on E cells persisted after the forcewas removed, it was necessary to compare the labelling index in cultures that wereexposed to [3H]dThd while being stretched, to that of cultures that had beenstimulated by stretching but were no longer under tension when [3H]dThd was addedto the culture. This could be done by limiting the degree of stretching, which couldbe controlled by the number of turns on the orthodontic screw, to an amount which,while stretching the flexible substrate, did not irreversibly deform the frame of the

38 D. M. Brunette

60

50

81 30CD

. QCO

# 20

10

I-

5Time (h)

10

Fig. 1. Frequency of labelled cell nuclei when [3H]thymidine was present for the timesindicated on the abscissa in mechanically stretched ( • • ) and control ( • • )cultures.

Table 1. Effect of temperature and presence of label on number of E cells synthesizingDNA

Temperature°

Labelling index (%)

Expt 1 Expt 2

UnstretchedStretchedUnstretchedStretched and labelled*Stretched, then labelledf

44

373737

00

ll-6± 2-120-5 ±2-320-413-0

00

7-6 ± 1-415-5 ±1-315-511-3

•Stretched for 2h in the presence of l-0^Ci/ml tritiated thymidine.f Stretched for 2 h, then the stretching force was discontinued to allow the culture dish to resume

its shape, then 1-0/iCi/ml [3H]dThd was added and the cells incubated with the label for 2h.

dish. Culture dishes stretched in this limited fashion could resume their initial shapeand the E cells that had been stretched but were no longer under tension could beexposed to [3H]dThd. There was no significant difference between cultures that hadbeen stretched and simultaneously labelled for 2 h and those that had been stretchedfor 2h and then labelled for 2h in the absence of stretching (Table 1). Thus theincrease in the number of E cells synthesizing DNA that was caused by mechanicalstretching persisted after the force had been discontinued.

To investigate the time-course in a narrower time-range the cell cultures were giventhe isotope for 2 h but the stretching force was only applied for the last part of thelabelling period. Thus, at the time marked 15 min in Fig. 2, [3H]dThd was added tothe culture for 105 min before stretching and the label was left in the culture for a

Epithelial cell response to stretching 39

further 15 min while the culture was stretched. It can be seen that mechanical stretch-ing had to be applied to the culture between 15 and 30 min to cause a rise in thenumber of cells synthesizing DNA.

30 r -

60Time (min)

80 100 120

Fig. 2. Frequency of labelled cell nuclei when [3H]thymidine was given for 2 h and thecells were stretched only for the period of time indicated on the abscissa.

500

400

300

XI

^ 2 0 - 0

100

1015 30 60Time (min)

90 120

Fig. 3. Frequency of labelled cell nuclei when [3H]thymidine was added for the timesindicated on the abscissa to stretched (O, • ) or unstretched ( • , • ) cultures at twopopulation densities: 05X104 cells/cm2 (O, • ) and 4X104 cells/cm2 ( • , • ) .

40 D. M. Brunette

Zetterberg & Auer (1970) noted that, for E cells from kidney, cell proliferationdecreased dramatically with increasing local cell density. Thus one possible explana-tion of the data in Figs 1 and 2 is that stretching acted by simply increasing the areaof the dish and thus decreasing the cell density. This possibility was examined bycomparing the labelling index of stretched and unstretched cultures seeded at 0 • 5 X104

and 4x 104 cells per cm2. An increase in labelling as a result of mechanical stretchingoccurred at both cell population densities (Fig. 3).

Fig. 4 shows the increase in labelling index that occurred when the cells werestretched and labelled for 2h in medium that contained FBS at concentrations be-tween 0 and 15 % (v/v). As the number of cells synthesizing DNA is influenced bytheir history and includes such factors as number of times subcultured and populationdensity, the data are expressed as relative labelling index. In this paper the relativelabelling index is defined as the labelling index of a culture divided by the labellingindex found for unstretched E cells grown in 15% FBS. The increase in labellingindex caused by mechanical stretching was most marked at higher concentrations ofserum. In the absence of FBS, mechanical stretching was unable to stimulate anincrease in labelling index.

Electron-microscopic examination of control (Fig. 5) and stretched (Fig. 6) cul-tures was undertaken to determine if any obvious change in ultrastructure was in-duced by the stretching. When the cultures were fixed with glutaraldehyde and post-fixed with osmium the most striking difference was that specialized junctions,

1-5

10

0)

C

c

2

<r 0-5

10 15% Serum

Fig. 4. The relative labelling index of stretched (O O) or unstretched ( • • )E cell cultures in medium containing different concentrations of foetal bovine serum.Relative labelling index is defined as the % labelled nuclei in the test conditions dividedby the % labelled nuclei found in an unstretched culture with 15 % FBS.

Epithelial cell response to stretching

Fig. 5. Electron micrograph of unstretched E cell cultures, d, desmosomes; t, tonofila-mentB. X20 000.

Fig. 6. Electron micrograph of stretched E cell cultures, m, microtubules; d, desmosomes;/, tonofilaments. More desmosomes per unit length of membrane and a higher volumefraction of filaments and microtubules were found in the stretched cultures. X20000.

1-2253

24398-95 (6-58)66-7 (20-0)

0-8750

13755-38(6-31)50-1 (21-6)

42 D. M. Brunette

Table 2. Morphometric comparison of stretched and unstretched E cells

Stretched Unstretched

No. of desmosomes/100 ^mNo. of micrographs examinedTotal length of cell membrane (/im)Microtubules, vol. fraction X1000 (S.D.)Filaments, vol. fraction XlOOO (s.D.)

i.e. desmosomes, were more prominent in the stretched cells. The number ofspecialized junctions per 100/im of plasma membrane on 50 micrographs preparedfrom a control culture was found to be 0-87, whereas 1-22 desmosomes per 100[imwere present in the 53 micrographs prepared from a stretched culture. The differencein desmosome number per 100 jUm of cell membrane was statistically significant(P< 0-05). In order to examine the cytoskeletal features of E cells some cultures werealso processed by a modification of the method of Ellisman & Porter (1980). In thesecultures the most prominent difference was that the volume fraction of filamentousstructures was increased in the stretched cells (Table 2).

DISCUSSION

In this study mechanical stretching was found to affect the proliferation andultrastructure of E cells in vitro. A prominent ultrastructural alteration was that thenumber of specialized junctions (desmosomes) per unit length of membrane appearedto increase. Whether this measured increase was the result of a real increase in thenumber of desmosomes or was caused by stretching, making the desmosomes moreeasily demonstrated, cannot be stated with certainty. However, it is not unreasonableto suppose that alterations in the attachment of cells to each other would occur as aresult of physical forces. For example, in vivo desmosomes have been found to be verynumerous in stratified squamous epithelia that are subject to severe mechanical stress(Fawcett, 1980). Furthermore, in vitro ultrastructural alterations in the tightjunctions between mouse mammary cells have also been reported to occur as a resultof mechanical tension (Pitelka & Taggart, 1983). Finally, it is not unlikely that newdesmosome formation could occur within 2h; Overton & DeSalle (1980) noted thatin reaggregating embryonic chick corneal epithelial cells, new desmosomes began toappear after between 2 and 3 h in culture.

The increase in the number of labelled E cells could be the result of either the entryof quiescent cells into the cell cycle, such as occurs with the addition of serum toserum-starved cultures, or a general speeding up of the cell cycle. Unlike thefibroblast cultures commonly used to investigate the regulation of the cell cycle, theE cells used in this study are not truly quiescent at either high or low populationdensities. At high population densities the E cells tend to form multilayers. While theuppermost cells in the layer occasionally assume a squame-like appearance, similar tokeratinizing epithelium, and detach from the cell layer into the medium, some of the

Epithelial cell response to stretching 43

cells, particularly those that are attached directly to the substrate, synthesize DNA.At low cell densities, when the cells form a non-confluent monolayer, the number ofcells synthesizing DNA is higher. Thus under the culture conditions used, for eitherlow or high population densities there may not be many quiescent cells available forrecruitment to proliferation and the more likely alternative is that, as suggested byCurtis & Seehar (1978) for chick embryo fibroblasts, mechanical stress causes ageneral speeding up of the cell cycle. In support of this hypothesis it was found thatmechanical stretching had the largest effect at high serum concentrations where therewas greater proliferation and, presumably, less-quiescent cells. A definitive answer tothe question of whether mechanical stretching stimulates quiescent E cells to enter thecell cycle or simply speeds up the cell cycle may require the development of cultureconditions that render the E cells quiescent but allow proliferation in response tovarious stimuli.

In his experiments on the blood-sucking bug Rhodnius prolixus, Wigglesworth(1935) found that stretching of the abdomen produced moulting. This effect, how-ever, appeared to be indirect and Wigglesworth concluded that nervous impulses tothe brain, due to stretching of the abdomen, provoked the secretion of a moultinghormone. Using in vitro systems to investigate mechanisms is a more direct approachbut the mechanism underlying the response to tension is still not known. Curtis &Seehar (1978) postulated that the cellular microfilament system might be involved inthe regulation of the cell cycle, as the extension of cells and the generation of tensionwithin them depend on the microfilament system. Morphometric analysis of stretchedand unstretched cultures of E cells indicated that the volume fraction of filamentousstructures was significantly greater in the stretched cultures. However, in preliminarystudies, cytochalsin B at 2-5^g/ml, an agent that disrupts microfilaments, had noeffect on the response of the E cells to mechanical stretching.

Application of mechanical stress for 15min has been found to cause a transientincrease in cyclic AMP levels in cultured bone cells (Harell, Dehel & Binderman,1977). A rapid response of this type would be required to explain one of the moststriking features of the response of E cells to mechanical stretching; namely, anincrease in the number of cells synthesizing DNA was observed after just 30min ofmechanical stretching. Moreover, although the effect of cAMP on cell growth appearsto vary with cell type, and for a number of cell lines increased cAMP levels areinhibitory (Pastan, Johnson & Anderson, 1975), stimulation of growth by choleratoxin (which increases the level of intracellular cAMP) has been observed for a numberof epithelia, including E cells derived from mammary tissue (Taylor-Papadimitriou,Purkis & Fentiman, 1980) and keratinocytes (Green, 1978). Thus, another possiblemechanism is that mechanical stress causes a rapid rise in intracellular cAMP in theseE cells, which results in increased proliferation.

The E cells used in this study were derived from the epithelial cell rests of Malassez,which are found in the periodontal ligament. Although little is known of their functionor cellular kinetics, the cell rests of Malassez were found to be stimulated to a moreproliferative state in vivo as a result of orthodontic tooth movement. This stimulationwas found only in the area that was under tension, epithelial cell rests were found to

44 D. M. Brunette

be reduced or absent on the portion of the periodontal ligament that was undergoing

pressure (Gilhous-Moe & Kvam, 1972). The data presented here indicate that

mechanical stretching in vitro increases the number of E cells synthesizing DNA.

Thus, the behaviour of the E cells in vitro reported here appears to be similar to the

response of the epithelial cell rests in vivo.

The author is indebted to Holly Maledy and Andre Wong for their excellent technical assistanceand to Linda Skibo for her care and patience in preparing the manuscript. The work was supportedby the Medical Research Council of Canada.

REFERENCES

ABERCROMBIE, M. (1957). Localized formation of new tissue in an adult mammal. Symp. Soc. exp.Biol. 11, 235-254.

BELOUSSOU, L. V., DORFMAN, J. G. & CHERDANTZEV, V. G. (1975). Mechanical stresses andmorphological patters in amphibian embryos. Embryol. exp. Morph. 34, 559-574.

BRUNETTE, D. M., MELCHER, A. H. & MOE, H. K. (1976). Culture and origin of epithelium-likeand fibroblast-like cells from poricine peridontal ligament explant and cell suspensions. Archs oralBiol. 21, 393-400.

CURTIS, A. S. G. & SEEHAR, G. H. (1978). The control of cell division by tension or diffusion.Nature, Lond. 274, 52-53.

ELLISMAN, M. H. & PORTER, K. R. (1980). Microtrabecular structure of the axoplasmic matrix:visualization of cross-linking structures and their distribution.^. Cell Biol. 87, 464—479.

FAWCETT, D. W. (1980). In The Cell, 2nd edn, pp. 160. Philadelphia: Saunders.GILHOUS-MOE, O. & KVAM, E. (1972). Behaviour of the epithelial remant of Malasaez following

experimental movement of rat molars. Ada Odont. scand. 30, 139-149.GOULD, T. R. L., MELCHER, A. H. & BRUNETTE, D. M. (1977). Location of progenitor cells in

periodontal ligament of mouse molar stimulated by wounding. Anat. Rec. 188, 133-141.GREEN, H. (1978). Cyclic AMP in relation to proliferation of the epidermal cell: A new view. Cell

18, 801-811.HARELL, A., DEHEL, S. & BINDERMAN, I. (1977). Biochemical effect of mechanical stress on

cultured bone cells. Calc. Tiss. Res. 22, 202-207.KANOZA, R. J. J., BRUNETTE, D. M., PURDON, A. D. & SODEK, J. (1978). Isolation and identi-

fication of epithelial-like cells in culture by a collagenase-separation technique. In Vitro 14,746-753.

LEUNG, D. Y. M., GLAG, S. & MATHEWS, M. B. (1976). Cyclic stretching stimulates synthesis ofmatrix components by arterial smooth muscle cells in vitro. Science, N.Y. 191, 475-477.

MEIKLE, M. C., REYNOLDS, J. J. & DINGLE, S. T. (1979). Rabbit cranial sutures in vitro: A newexperimental model for studying the response of fibrous joints to mechanical stress. Calc. Tiss.Int. 28, 137-144.

MEIKLE, M. C , SELLERS, A. & REYNOLDS, J. J. (1980). Effect of tensile mechanical stress on thesynthesis of metalloproteinases by rabbit coronal sutures in vitro. Calc. Tiss. Int. 30, 77-82.

MILLONIG, G. (1961). Advantages of a phosphate buffer for osmium tetroxide solutions in fixation.J. appl. Phys. 32, 1637.

OVERTON, J. & DESALLE, R. (1980). Control of desmosome formation in aggregating embryonicchick cells. Devi Biol. 75, 168-176.

OWENS, R. B. (1974). Glandular epithelial cells from-mice: A method for selective cultivation. J.natn. Cancer Inst. 52, 1375-1378.

PASTAN, I. H., JOHNSON, G. S. & ANDEKSON, W. B. (1975). Role of cyclic nucleotides in growthcontrol. A. Rev. Biochem. 44, 491-522.

PITELKA, D. & TAGGART, B. N. (1983). Mechanical tension induces lateral movement of intra-membrane components of the tight junction: Studies on mouse mammary cells in culture. J. CellBiol. 96, 606-612.

ROGERS, A. W. (1979). In Techniques of Autoradiography, 3rd edn, p. 111. Amsterdam: Elsevier.

Epithelial cell response to stretching 45

SQUIER, C. A. (1980). The stretching of mouse skin in vivo: Effect on epidermal proliferation andthickness, J. invest. Derm. 74, 68-71.

TAYLOR-PAPADIMITRIOU, J., PURKIS, P. & FENTIMAN, I. S. (1980). Cholera toxin and analoguesof cyclic AMP stimulate growth of human mammary epithelial cells. J. Cell. Physiol. 102,317-321.

WIGGLESWORTH, V. B. (1935). The physiology of ecdysis in Rhodnius prolixus (Hemiptera). II.Factors controlling moulting and 'metamorphosis'. Q.Jl Microsc. Sci. 77, 191-223.

ZETTERBERG, A. & AUER, G. (1970). Proliferative activity and cytochemical properties of nuclearchromatin related to local cell density of epithelial cells. Expl Cell Res. 62, 262-270.

{Received 11 November 1983-Accepted, in revised form, 8 February 1984)