Polarity and Polarization of Fibroblasts in Culture

POLARITY AND POLARIZATION OF FIBROBLASTS IN CULTURE

Albert K. Harris

I. Introduction . . . . .................................. II. Description of Fib ity (Special Characteristics of Each Type of Margin) ..................... A. "Type A" Margins. . . . . . . . . . . . . . . . . ......................... B. "Type 6" Margins C. "Type C" Margins . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Contact Guidance. . . . . . . . . . . . . . . .................. IV. Contact Inhibition and Contact Paralysis. . . . . . . . . . . . . . . . . . . . . V. Calvanotaxis: Effects of Electric Fields on Fibroblast Polarity VI. Microtubules in the VII. Questions About the Autonomy of Fibroblast Polarity . . VIII. Conclusions and Prospects . . . .........................

Advances in Molecular and Cell Biology Volume 26, pages 201-252. Copyright 0 1998 by JAI Press Inc. All right of reproduction in any form reserved. ISBN: 0-7623-0381-6

201

202

206 210

,212

245

ALBERT K. HARRIS

202

1.

INTRODUCTION

Fibroblasts are the most familiar of tissue culture cells. They are the ubiquitous, nondescript, generally stellate cells that quickly come to dominate primary cultures derived from explants of almost any organ or tissue. Many of the classical studies of tissue cell locomotion, such as those of Michael Abercrombie and Joan Heaysman (1953, 1954), concentrated on fibroblasts from primary cultures, with chick embryo heart fibroblasts being the most frequently used. In addition, many “established’’ cell lines of fibroblasts have been developed and are very widely used in research; examples include the various 3T3 lines, among many others. To attempt any very precise definition of exactly what is, and what is not, a fibroblast would be premature. Molecular criteria will eventually be found, such as possession of intermediate filaments of the vimentin class (as opposed to those of the keratin type characteristic of epithelial cells), although vimentin has turned up in other cell types, including endothelia. The secretion of collagen and other fibers was once assumed to be a special property of fibroblasts, until it was shown that this is also done by epithelial cells, among others. The approximately synonymous term “mechanocyte” failed to catch on as a substitute for fibroblast, but would have been a more accurate reflection of what seems to be the real function and special role of these cells, namely the exertion of mechanical forces. Using flexible rubber substrata and gels of fibrin and reprecipitated collagen, it has been shown that fibroblasts exert pulling forces hundreds of times stronger than needed for their own propulsion, with these forces being capable of large scale reorganization of collagen fibers over distances as great as several centimeters (Stopak and Harris, 1982). The main function of the phenomenon we see as fibroblast ‘‘locomotion’’ may really be mechanical reorganization of extracellular matrix materials. Fibroblasts try to pull surrounding materials rearward past their cell bodies, for purposes believed to include wound closure. Because of the rigidity of glass, polystyrene and other artificial culture substrata, however. all we usually see is that the fibroblasts stretch themselves out flat and crawl around jerkily. The polarity of fibroblasts is primarily a matter of directional protrusion combined with the exertion of strong pulling forces in the direction opposite to this protrusion. This is easiest to visualize when a fibroblast happens to have only a single front end, advancing in front, with a single trailing margin at the opposite end. Unfortunately, such nice and simple “typical fibroblasts” tend to be more the exception than the rule. Even when a given fibroblast does happen to be monopodial in this way, this is usually only transient. In contrast to polymorphonuclear leucocytes, for example, where monopodial behavior is the norm, fibroblasts are neither consistently nor persistently monopodial, with distinct and long-lasting front and rear ends. Instead, most fibroblasts are irregularly multipolar. Two, three (or more) parts of a given cell’s margins typically crawl off in different directions at the same time, stretching the cell body between them. The usual stellate shape of fibroblasts results from their being stretched between several competing “front” ends, (Figures 1,2).

Fibroblast Polarity

figure 7.

203

3T3 fibroblast, seen by ordinary phase contrast.

The same 3T3 fibroblast seen by interference reflection microscopy, in which the areas of contact and adhesion between the substratum and the cell’s lower surface show up as grey and black areas. figure 2.

It should also be noted that even quite small fragments of fibroblasts, lacking nuclei, tend to adopt very much these same types of polarity (Gelfand et al. 1985). The very useful term “leading lamella” was introduced by Abercrombie et al. (1970a, 1970b) to refer to these “front” ends, and is equally applicable whether there happens to be only one of them (in a monopodial cell), or when a cell happens to have two or more leading lamellae at the same time. For example, the fibroblast shown in figures one and two happens to have four leading lamellae. It is not even

204

ALBERT K. HARRIS

unusual for a fibroblast to have one single broad leading lamella extending all the way, 360°, around its periphery. When trypsinized fibroblasts are plated out onto a new substratum, most of them will pass through such a stage. Freshly-plated cells spread initially along all parts of their periphery, thus passing through a characteristic “fried egg” morphology a few hours after first being plated out (see Witkowski and Broughton, 197I). Subsequently, this single leading lamella breaks up into two or three regions that continue to advance outward, separated by parts of the margin that retract inward. Another case where leading lamellae occupy the entire cell margin is when some kinds of fibroblasts are exposed to antimicrotubule drugs (Vasiliev et al. 1970), as is discussed more fully in a later section. Net locomotion, when it occurs, results from imbalances between leading lamellae, for example when those along one side of a given cell are stronger than those along the opposite side. Such a dominant leading lamella, or (often) groups of adjacent leading lamellae, therefore become the temporary front end of the cell, in the sense of pulling everything else along behind them. At the opposite, trailing side of the cell, one frequently observes parts of the cell margin progressively being pulled loose from their adhesions to the substratum, often breaking these adhesions a few at a time in a jerky fashion, sometimes with long strands of plasma membrane (called “retraction fibers”) being pulled out behind the cell as its margin passively retreats (Harris, 1990). As I am trying to emphasize, the essential feature of fibroblast polarity is the subdivision of each cell’s lateral margins into several relatively distinct kinds of behavior (three main kinds, I think) . Some parts of the margin advance actively outward. Other parts of the margin are dragged passively behind as a sort of tail. And, the third class of margins are those that sag inward in long smooth catenary-like curves along the sides. With the goal of streamlining discussion, I propose simply to call these types of cell margins “A,” “B,” and “C.” The “type A” margins are the ones that Advance outward; the type B margins are the ones trailing Behind; while the type C margins are the ones that sag inward like Catenary curves, (Figure 3). This nomenclature is meant only to be provisional, and to focus attention on such questions as which particular molecules need to be concentrated near a given part of a cell’s margin in order to cause that region to protrude forward as a “type A margin.” When the molecular basis of these three different behaviors becomes understood, then we can name them accordingly. For example, the type A margins may then be called “actin reassembly margins,” or perhaps “type I myosin margins.” Meanwhile, using simply A, B, and C is intended to minimize circumlocution. The kinds of questions we need to ask include the following. What special molecules (type I myosins, perhaps?), or molecular activities (such as assembly of gactin into fibers) areresponsible for causing a given one part of a cell’s margin to become “type A’? Likewise, why does treatment with microtubule poisons tend to cause type A activity to spread out, and why does this occur in certain cell lines much more than in those from primary cultures? In the phenomenon called “contact guidance,” why do ridges and grooves cause cells to concentrate type A activity to

Fibroblast Polar; ty

205

Figure3.

Diagram of a typical fibroblast, crawling on a flat glass or plastic substratum ir tissue culture, with the three characteristic types of cell margins that are described in the text as type A, type 6 , and type C. Type A margins advance forward, with protrusive activity, ruffling, actin assembly and type I myosin being concentrated there. Type E margins are those that retreat, but do retain adhesions to the substratum, while type C margins are mostly non-adherent and sag passively inward in smooth concave curves Fibroblast polarization is essentially a matter of separating these three sets of activities and properties into different parts of each cell’s margins.

the parts touching grooves and ridges? Conversely, in the twin phenomena of contact paralysis and contact inhibition, why is it that cell-cell contact so often causes type A margins to switch over to type B or even to type C behavior? With the molecular aspects of propulsion now becoming so much clearer, and with the accumulation of descriptions of new cytoskeletal proteins and their distributions, the time now seems to be ripe for explaining whole-cell behaviors in terms of molecules and forces. I hope that this review can help focus attention on solvable problems in this area. Some widely-used textbooks describe the locomotion of fibroblasts and other tissue cells as consisting of an alternating cycle of forward protrusion of filopodia, adhesion of these filopodia to the substratum, and contraction of these filopodia, so as to pull the cell body forward. Unfortunately, this “reach-grab-pull” concept is a misleading oversimplification, essentially an “Aesop’s fable,” especially as applied to fibroblasts. Notice that it would lead you to expect all of the following (none of which is true): (1) that cells lacking filopodia could not spread or crawl; (2) that cytoplasmic protrusions should be concentrated along the cell’s leading margins, and therefore that the presence of such protrusions along a certain part of a cell’s margin would be evidence that that must be the front end or leading edge; (3) that rearward force-exertion would be concentrated in the area right at the cell margin, between the tips and the bases of these filopodia; and (4) that force exertion would be pulsatory, varying with time in synchrony with the supposed cycle of protrusion, attachment and contraction. Not one of these predictions is true. (1) It is not unusual for actively crawling fibroblasts to lack filopodia. (2) Where concentrations of cytoplasmic extensions are observed, this most often corresponds to the rear end of the fibroblast (rather than the front end), with these extensions being retraction fibers in the process of being pulled out uassivelv from retreatinn (“tvue B”) cell margins.

ALBERT K. HARRIS

206

( 3 ) Most of the pulling forces exerted by fibroblasts are transmitted to the substratum 5, 10 or more millimicrons behind the advancing (type A) cell margins, being exerted tangentially through the plasma membrane. &furthermore, (4) even though the magnitude of these pulling forces can vary with time, such variations are not in any regular cycle correlated with filopodial extension and withdrawal. My own experience and belief is that filopodia have little or nothing to do with fibroblast locomotion, despite the undoubted importance of such protrusions in nerve growth cones. However useful for satisfying students’ curiosity, the “reach-grab-pull” idea makes too many incorrect predictions to be anything but an obstacle to understanding.

11.

DESCRIPTION OF FIBROBLAST POLARITY (SPECIAL CHARACTERISTICS OF EACH TYPE OF MARGIN) A.

”Type A“ Margins

These are the parts of the cell margins that advance forward across substrata. Their forward progress is usually oscillatory, with periods of advance alternating with periods of retreat, in the category of “two steps forward and one step back,” with the durations of each type of behavior being on the order of a few minutes. Even during these small retreats, these type margins retain the peculiarity of being the only parts of the cell margin where stiff protrusions extend out in front of the most distal adhesions to the substratum. Thin forward protrusions, called lamellipodia, are the most characteristic feature of these margins. Lamellipodia are very thin, as little as 0.1 micrometers in thickness (Abercrombie et al., 1970a, 1971), but can be 5 , 10, 15 or more micrometers in width and extend out 5 or more micrometers in front of the cell, usually bending slowly backwards in a type of movement called “ruffling” (Abercrombie et al., 1970b), which is familiar to anyone who has ever seen a time lapse film of any type of crawling tissue cells. Lamellipodia are filled with a meshwork of actin fibers, and mechanically are surprisingly stiff, in the sense of resisting bending by glass microneedle. Lamellipodia may seem to be flopping around limply, but really they are stiff and their bending is an active process. Lamellipodia are packed tightly with meshworks of actin fibers, as seen in transmission electron microscopy sections (Heath and Dunn, 1978; Small, 1981) as well as when cells are stained with fluorescently labeled phalloidin, apeptide mushroom toxin which binds strongly to filamentous actin (as well as promoting assembly of actin into fibers). There is also evidence that the outermost parts of type A margins are sites of actin assembly (Wang, 1985; Svitkinaet al., 1986; Symons and Mitchison, 1991), as well as that actin flows centripetally from there (Fisher et al., 1988). Other cytoplasmic components are excluded from the lamellipodia, including not just mitochondria and vesicles but even microtubules and conventional type I1 my-

Fibroblast Polarity

207

osins. In contrast, type I myosins have recently been found concentrated specifically along the outer margins of fibroblast lamellipodia (Wagner et al., 1992). Two other classes of cell surface protrusion, in addition to lamellipodia, are often concentrated along the type A margins of fibroblasts. Filopodia are spike-like protrusions, which are likewise packed with fibrous actin, and likewise extend out ahead, fold back, and move centripetally across cell’s upper surface, often being intermingled with lamellipodia, like fingers extending from the hand. The third class of cell surface protrusion consists of hemispherical herniations in the plasma membrane, called blebs, ranging from diameters of as little as one micrometer up to 5 or even 10 micrometers. Blebs can bubble forth fromjust about any part of a fibroblast’s surface, but are often concentrated right along the advancing “type A” margins, intermingled with lamellipodia. If a cell is blebbing only slightly, at a few locations, these will almost always be concentrated right at the tips of the type A margins, although these parts of the cell surface were either under especially strong outward pressure, or were particularly weak in resisting widely distributed outward forces. Each bleb forms rapidly, in a few seconds, and then gradually shrinks and disappears over a period of a minute or more. Fibroblasts can be easily stimulated to switch from forming lamellipodia to forming blebs; one way to do this is to make the culture medium more hypertonic and another is to decrease the stretching tension in the cell body. A wave of spontaneous blebbing occurs each time a fibroblast undergoes partial detachment and retraction of one or more of its leading lamellae; not just the lamellipodia at the tip of the retracted lamella, but also those along the edges of this cell’s other lamellae, will be replaced with blebs until re-spreading has restored a state of tensile stress. Likewise, when trypsinized fibroblasts are plated out, they first bleb and only later ruffle. The typical sequence is for each new bleb to be a little smaller in diameter than those previously formed, until lamellipodia replace the (now very tiny) blebs. The mechanistic significance of this and other easily-observed transitions has yet to be discovered. Besides their special powers of outward protrusion, type A margins also have special adhesive properties, in particular the propensity to initiate new adhesions (Harris, 1973b; Heath and Dunn, 1978). The optical technique called interference reflection microscopy (IRM) makes visible the closeness of approach between the cell’s lower plasma membrane and glass, plastic, or even silicone rubber substrata. The closest contacts appear black by this technique, while slightly wider gaps appear gray (Abercrombie and Dunn, 1975; Izzard and Lochner, 1976, 1980). Both these types of cell-substratum contact are concentrated behind the type A margins of fibroblasts, the black “focal adhesion plaques,” appearing as many small spots elongated perpendicular to the nearby cell margin and around a micrometer in width and several micrometers in length, scattered within a much wider gray contact area. I once directly tested what had seemed to me the naive assumption that the black contacts were very strong adhesions while the gray contacts were weak adhesions. Using thin glass microneedles and a micromanipulator, I peeled fibroblasts

208

ALBERT K. HARRIS

away from glass, styrene and rubber substrata. As it turned out, however, the results indicated that the usual assumptions about relative adhesive strengths, however naive, seem to be absolutely correct! The gray areas can be easily peeled away from the substratum, while the black contacts cannot be detached without ripping away that part of the plasma membrane (except with silicone rubber substrata, in which case you can often pop loose adhesion plaques without ripping the membrane). In addition, I found that new black adhesions cannot be created simply by pushing down on the top of the cell until previously nonadherent areas of plasma membrane on the bottom are pushed close enough to the substratum to appear black by interference reflection. Even after being pushed close to glass substrata for as long as several minutes, such areas immediately moved back to their former separation distance once the pressure is removed. This supports the idea that there is something specially adhesive about the plasma membrane components at the focal contacts, and it has indeed been shown that fibronectin receptors (integrins) become concentrated there (Burridge and Fath, 1989). On the other hand, it remains to be clarified what is special about the membrane making up the broad gray contacts; one possibility is that these areas are likewise integrin-rich, but with a lesser degree of concen tration. Time lapse films of the interference reflection images of fibroblasts have to be made using long time exposures, at moderate light levels, since the intense light ordinarily used for IRM will radically change cell behavior. But with time exposures, one can easily observe a close correlation between type A margin behavior and the presence of the broad gray contact areas. These areas are very reminiscent of radar images shown on television newscasts of bands of thunderstorms. Wherever new contacts develop, this is a sure sign that a part of the cell margin that had previously been inactive (type B or C) is about to become a type A margin and begin protruding outward. Conversely, quiescence and shrinkage of these gray contact areas is the first indication that a type A cell margin is about to change into a type B region. The molecular basis of these changes deserves study. As regards force exertion, this aspect of fibroblast polarity has been studied by culturing the cells on flexible substrata of plasma clots, collagen gels and very thin sheets of silicone rubber (Figure 4). These forces are called “traction”, and should be not be confused with contraction, in that the parts of the cells exerting the traction forces do not shorten in length (do not contract), but rather become more and more elongated and stretched as a cumulative result of their exertion of these traction forces. The active displacement of visible particles and lectins (Abercrombie et al., 1 9 7 0 ~ 1972; ; Harris and Dunn, 1972) attached to the outer surface of the plasma membrane (sometimes called “capping”) seems to be caused by the same mechanism of force exertion, since the forces are exerted in the same directions and locations. Along type A margins, the traction forces are consistently directed inward, in the direction exactly opposite the direction of outward spreading. This is mostly perpendicular to the leading margin, but often with a degree of lateral convergence.

Fibroblast Polarity

209

Figure 4. Low magnification view of chick heart fibroblasts cultured on a silicone

rubber substratum. Note the wrinkles in the rubber layer, which are produced by the cells’ traction forces, by means of which the cells spread and crawl about.

This is shown by the patterns of wrinkles formed in silicone rubber substrata, where the compression wrinkles formed behind the type A margins are mostly oriented perpendicular to the advancing margin, but often with some wrinkling perpendicular to this as well. Particles attached to both the upper and lower surfaces of fibroblasts are actively transported directly rearward away from type A margins, although often with some tendency of their paths to converge somewhat. On the top, the particles are transported centripetally, usually accumulating near the nucleus, but sometimes also accumulating near the type C and type B margins. As regards cytoplasmic proteins, the concentration of actin as a meshwork along type A margins has already been mentioned. Based on time lapse videos made with very high contrast differential interference contrast optics by Heath and Holifield (I 991), a copy of which they kindly provided to me, it appears that this actin network flows steadily rearward away from fibroblast type A margins at several micrometers per minute (almost exactly as observed in neuronal bag cells by Forscher and Smith, 1988). This would necessitate a high rate of assembly of actin monomers into fibers, with this assembly apparently being concentrated right along the type A margins, apparently at the outer side of the lamellipodia. This concentration of actin assembly may be the most fundamental peculiarity of the type A margin, and be responsible both for the forward protrusion and mechanical stiffening of the lamellipodia, as well as for the traction forces exerted directly rearward away from the margin. Behind the type A margins, the so-called “stress fibers” are formed consisting of actin combined with type I1 myosin, tropomyosin, alpha actinin and other muscle-related proteins. These stress fibers terminate at the focal (black) adhesion plaques mentioned above, where they attach at the inner surface of the plasma membrane. The linking proteins talin and vinculin are also concentrated just inside the plasma membrane at these

ALBERT K. HARRIS

21 0

plaques. Talin is also concentrated along the lamellipodium itself, in other words right at the type A margin (Burridge and Connell, 1983). At least a few membrane proteins develop special geometric arrangements relative to type A margins, becoming progressively depleted in concentration there, so as to form dramatic concentration gradients detectable by staining with fluorescent antibodies (Holifield et al., 1990; Ishihara et al., 1990). The formation of these gradients by a given membrane protein is evidence that it is one of the “tine proteins” serving to transmit traction forces tangentially through the plasma membrane, from the rearward-flowing actin networks just inside the plasma membrane to adherent objects and substrata on the outside. B.

“Type B“ Margins

These are the retreating margins that form the rear ends of crawling fibroblasts. The process of their retraction has been analysed by Chen (198 1). Their two most characteristic features are the formation of retraction fibers, rather than lamellipodia or filopodia, and the locations of the black-appearing adhesion plaque cellsubstratum adhesions right at the extreme margins of the cell. This is in contrast to these adhesions’ locations along type A margins, where they are many micrometers behind the cell margin. Another difference is that the broad gray adhesions are usually absent behind type B margins. In other words, whereas behind type A margins, the black adhesions are well behind the edge of the cell, and are surrounded by gray adhesions, in the case of type B margins, the edge of the cell retreats right up to the black adhesions, and the gray-appearing ones are absent. In my experience watching time lapse films of fibroblasts’ IRM images, the disappearance of these broad gray contacts behind an actively spreading type A margin is a sure sign that it is about to become converted into a type B margin. Conversely, it sometimes happens that the latter become reactivated into type A margins, with one of the first signs being a reappearance of gray contacts. Especially in scanning electron micrographs, retraction fibers are often mistaken for filopodia. This can be very misleading indeed, given that retraction fibers are peculiar to trailing margins while filopodia are characteristic of advancing margins. To confuse the one for the other is to mistake the front end for the rear. Retraction fibers are attached to the substratum at their distal ends, which are often somewhat expanded where they adhere, somewhat resembling suction cups. Retraction fibers can be 5, 10,20 or more micrometers in length, in contrast to filopodia which are only a few micrometers long. Another difference is that filopodia are stiff, whereas retraction fibers are flexible, bend easily, and recoil if cut. It is also not unusual for blebs to form along type B margins of fibroblasts. But lamellipodia only form there rarely, and are a sign that the margin is about to become reactivated again as a type A margin. Type B margins are the rarest and most transient of our 3 categories, at least in fibroblasts. Note, for example, that the cell shown in Figures 1 and 2 has only type A

21 1

Fibroblast Polarity

and type C margins. Examples of type B margins can be seen in the cell at the extreme left of Figure 3, and also in Figure 18 (in a later section, on mirror image symmetry) at the extreme lower right end of the right-hand cell and the extreme lower left end of the left-hand cell. When the molecular basis of fibroblast polarity becomes better understood, it may be that “Type B” margins will be regarded as no more than an inactive form of a type A margin, or perhaps as essentially a transition stage in the conversion of type A margins to type C margins. On the other hand, there are some other cell types, polyrnorphonuclear leucocytes being a good example, for which it is normal to possess a well demarcated and more or less permanent “type B” retreating cell margin. This is accompanied by a higher degree of directionality and polarity than is characteristic of fibroblasts. Whether well-formed “type B” margins should be considered merely a side effect of greater polarization, or perhaps alternatively as part of the causation of that polarization, are questions that will have to be left to future studies.

C. “Type C” Margins These margins are concave, usually smoothly curved, have few if any adhesions to the substratum, and are quiescent, although sometimes with intermittent formation of small lamellipodia. The cytoplasm seems to be stretched tight along the axis parallel

Figure 5. Low magnification view of several chick heart fibroblasts cultured on a haptotactic substratum consisting of elongate rectangles (10 micrometers wide and 100 micrometers long) of vacuum-evaporated palladium metal overlying a nonwettable (and relatively nonadhesive) polystyrene culture dish. The cells elongate along these rectangles, and only extend short distances off of them. Had these adhesive islands been made smaller than the space occupied by a spread cell, cells would spread off them. In fact, when initially plated onto such rectangles, those cells that start out trying to spread prependicular to the long axes of the rectanglesoften do cross the gaps from one island to the next.

21 2

ALBERT K. HARRIS

Figure 6.

Low magnification view of a single 3T3 fibroblast cultured on a haptotactic substratum consisting of a series of small squares of vacuum-evaporated palladium metal overlying a nonwettable (and relatively non-adhesive) polystyrene culture dish. Because the individual 10 micrometer squares are much smaller than the area that such a cell would normally occupy on a homogeneously adhesive substratum, the cell eventually occupies 8 or 9 of them, becoming stretched between them into spindly shapes very unusual for these cells.

to the cell margin, which is often much thcker than either of the other two types of margin, often appearing phase dense and refractile in phase contrast microscopy (Harris, 1973b). Bundles of actin fibers are concentrated in these parts of the marginal cytoplasm, and microtubules are often found there, with both the actin fibers and the microtubules having their long axes oriented parallel to the margin (Figures 5, 6). Zand and Albrecht-Buehler (1989,1992) have made special studies of these marginal bundles of actin, reporting that they have properties somewhat different from the usual actin stress fibers in more central areas. In particular, they report that the bundles at the margins are more resistant than ordinary stress fibers to disruption, for example by the drug cytochalasin. They also studied the effects of using micromanipulation to disturb these margins (which they called “webbed edges”).

111.

CONTACT GUIDANCE

Fibroblasts, as well as other tissue cells, respond to fibrous or grooved substrata by aligning themselves along the axes of the fibers or grooves (Figure 7). This orientation response has been observed since the very earliest tissue culture studies (Harrison 1914), and was given the name contact guidance by Paul Weiss (1934, 1961), who studied the phenomenon intensively for many years and made it the basis of important theories about embryonic tissue development and the guidance of nerve

Fibroblast Polarity

213

fibers. When fibroblasts are cultured in or on gels made of clotted plasma (fibrin) or reprecipitated collagen, the cells’ individual morphologies become much more elongate than when flat glass and plastic surfaces are used as culture substrata (Figure 8). These differences in cell morphology are presumed to be due to contact guidance, and make cultured cells look much more like the mesenchymal cells one

Figure 7. Single frame from a time lapse film showing fibroblasts crawling about on a thin layer of clotted fibrin (within which small particles of carbon black had been placed in order to be able to trace the directions and relative strengths of the cells’ propulsive traction forces).

figure 8. Three chick heart fibroblasts cultured inside a gel made of reprecipitated collagen from a rat tail tendon. Notice how the cells adopt extremely elongated shapes when cultured in or on collagen substrata, in contrast to the flattened stellate and polygonal shapes these same cells would have adopted on a flat glass or plastic su bstratum.

21 4

ALBERT K. HARRIS

sees in histological sections. This implies that contact guidance must be a primary determinant of cell shape in vivo, where collagen (and, in wounds, fibrin) would form the substrata in which they spread and crawl . In collaboration with Garber, Weiss showed that the degree of directional polarization of fibroblasts plated out in plasma clots depends strongly on the conditions of clotting. When there are many small fibers, the cells become more stellate than when the plasma is clotted in such a way as to produce larger fibers. In the latter case, the fibroblasts become more bipolar (Weiss and Garber, 1952; Garber, 1952). It should be pointed out, however, that these authors’ interpretation of these results in terms of capillarity has long since been overtaken by demonstrations that fibroblast locomotion is an active process, not a form of passive wetting. One can produce contact guidance experimentally by plating fibroblasts out onto such things as diffraction gratings (see Figure 9), fish scales (at least some of which have aligned collagen fibres on one of their surfaces; see Weiss 1961), grooved surfaces (Rovensky and Slavnaja, 1974)), plastic surfaces scratched directionally with abrasives, thin glass fibers, and lengths of spiders’ web (Harrison, 1914) among many other materials. The cells’ dimensional range of sensitivity seems to span a surprisingly broad scale. At the large end of the scale, alignment will occur on cylindrical fibers as much as 200 micrometers in diameter (Dunn and Heath, 1976); while at the small end of the scale, they can also align in response to grooves almost too narrow and shallow to see with amicroscope (O’Hara and Buck, 1979). In either case, the net result can fairly be described as that of favoring development of what we are calling type A cell margins on those sides of each cell that are directed along the fibers or grooves. This results in polarization of the individual cell morphologies into bipolar shapes, each with its long axis oriented parallel to the fibers or grooves. Likewise, the cells’ direction of locomotion also becomes

Figure 9. Low magnification view of the alignment of chick heart fibroblasts along the grooves and ridges of a glass diffraction grating.

Fibroblast Polarity

21 5

strongly constrained to this axis, along which the individual cells shuttle back and forth. Weiss’s own mechanistic explanation for this phenomenon was based on erroneous notions about cell spreading being a passive wetting of the substratum by a sort of directional capillarity, rather than as an active process of locomotion. Nevertheless, many of Weiss’ proposals about control functions for contact guidance during embryonic development remain tenable. These proposals have been supported by observations of positive feedback cycles in which fibroblast traction pulls collagen fibers into alignment and this progressive alignment of fibers stimulates further alignment of fibroblasts by contact guidance (Stopak and Harris, 1982). In addition, Campbell and Marcum (1980) have documented an actual case of contact guidance of migrating cells inside living Hydra. On the other hand, it is interesting to note that in the development of arteries and arterioles, fibroblasts as well as smooth muscle cells line up circumferentially (rather than longitudinally) around the cylindrical shape of these vessels. Although this would seem to indicate some sort of “reverse contact guidance,” I have plated fragments of small arteries onto glass cylinders and found their cells to align parallel to these cylinders’ axes. The morphogenetic function of contact guidance thus confronts us with some interesting problems in the control of cell polarization. One attempt to account for contact guidance in terms of modern knowledge of cell locomotion was that of Dunn and Heath (1976). Their idea was that protrusion or contraction might inhibited in those parts of cells which happen to become bent by the curvature of the substratum. For example, bending might interfere with either the formation or the contraction of actomyosin fibers. On a cylindrical surface, the least bent sides of the cell would be those facing along the axis of the cylinder (curvature being zero in that direction). If the activities characteristic of type A margins were to become concentrated along these (least bent) sides of the cell, this would explain the bipolar elongation that one observes. This hypothesis was supported by elegant experiments in which fibroblasts were cultured on special substrata made to have sharp ridge lines (shaped like the roof of a house). When those cells that spanned this ridge line were closely examined, their cytoplasmic actin fibers were found not to cross from one side to the another. This indicates either that actin fibers cannot extend around bends, or that there is some kind of preferential elimination of those that are bent. Either should have the effect of biasing locomotory activities in the direction of least substratum curvature. Dunn and Brown (1986) have also developed geometrical criteria for describing and comparing degrees of cell alignment in response to contact guidance. A rather different explanatory hypothesis was proposed by O’Hara and Buck ( 1979). Their idea concerned cell-substratum adhesions rather than the cytoskeleton, and related more specifically to the situation on fibrous and finely grooved substrata, as opposed to ones that are smooth but curved. They reasoned that since the adhesion plaques formed by most fibroblasts are elongated in the direction of spreading, often being several microns long and only one micron or less in width,

21 6

ALBERT K. HARRIS

then the formation of such adhesions would be hindered on substrata having grooves or fibers with lateral spacing of a few microns or less. Although the cells would be able to form adhesion plaques behind the parts of their margins facing along the axes of the grooves or fibers, formation of such adhesions would be difficult or impossible in directions oblique to this axis. As these authors showed, and I can attest from parallel studies of my own, fibroblasts will align strongly in response to grooves that are exceedingly narrow, shallow, and closely spaced. Carborundum and optical rouge are sold in a wide range of different grain sizes for the purpose of lens and mirror grinding by amateur astronomers: by directionally wiping these different abrasives across glass and polystyrene surfaces, one can easily make substrata having many parallel grooves, with the widths and depths of these grooves varying according to which abrasive is used. Fibroblasts orient quite strongly in response to extremely fine grooves with spacings and depths on the sub-micron scale. Buck (1980) also showed that when fibroblasts are cultured on rubber surfaces, and the rubber subjected to repeated cycles of stretching and relaxation, then the fibroblasts will line up perpendicular to the direction of this stretching. I would argue that this is simply due to differential breakage of cell-substratum adhesions. Those adhesions on the sides of the cells directed parallel to the stretching will be subject to much larger stress than those on the sides of the cells perpendicular to this direction. More recent studies of cellular responses to fine grooves have used ultrafine photoengraving techniques and photoresists, originally developed for the manufacture of printing plates, but now used to make computer chips. These methods allow construction of grooved surfaces with precisely consistent geometries and dimensions on the submicrometer scale. This technique has used extensively by Clark and his collaborators (1987, 1990, 1991, 1992, also see Britland et al., 1992, and Curtis Clark, 1990, and Dow et al., 1987). They found that some cells can orient in response to parallel arrays of grooves that are only about a tenth of a micrometer deep, and spaced only a quarter of a micrometer apart. As Dunn has discussed in arecent review (Dunn, 199l), this ability of fibroblasts to orient in response to extremely shallow grooves and ridges is difficult to explain by the actin bending theory that he and Heath had previously suggested. It is hard to see why such extremely fine order striations would be able to produce enough bending of cytoplasmic actin fibers to bias cell locomotion directionally. An important question here is the extent to which the cell’s lower surface bridges over the grooves, as opposed to extending down into them. Dunn and Brown (1986) found some evidence that cells do extend down into shallow grooves, as did Brunette (1986; see also Brunette et al., 1983; Oakley and Brunette, 1993). In these studies, fibroblasts were sometimes able to form focal adhesions to the bottoms and sides of such grooves. In addition, Clark et al. (1991) reported that some fibroblasts orient more in response to deep grooves than to shallow ones of the same width, which would be hard to explain if the cells were merely bridging over the tops. Adam Cur-

Fibroblast Polarity

21 7

tis has been developing a new type of hypothetical mechanism, based on the idea that sharp bends in substratum shape could locally favor either actin condensation or the formation of new substratum adhesions. This would fit in well with the recent discoveries by Depasquale and Izzard (1987,1991) on these marginal condensation events (see also Izzard and Izzard, 1987; Izzard, 1988). If high contrast DIC observations such as theirs were to be carried out on finely grooved substrata such as those of Clark et al., perhaps the underlying mechanisms would reveal themselves. Recent work by Oakley and Brunette (1993) employed confocal microscopy to study the sequences of rearrangements and realignments of cytoskeletal proteins in human gum fibroblasts plated out onto the surfaces of finely grooved titanium substrata. They report that microtubules are the first cytoskeletal components to align, that these become positioned along the bottoms of the grooves, and that this occurs as early as 20 minutes after the cells have been plated out. Not until 40 minutes to an hour after plating could aligned actin fibers be detected; and not until 3 hours after plating were aligned adhesion plaques apparent. They also note the interesting fact that the early position of the microtubule organizing center was essentially random relative to the nucleus and the direction of cell alignment, as if its role were merely passive. These exciting results seem to point to microtubules themselves as important in detecting substratum alignment as well as controlling morphological responses. Drawing upon the literature of plant cell wall morphogenesis, they suggest that the microtubules’ behavior may be a matter of minimizing shear stress. I believe this would be closely analogous to the hypothesis proposed by Dunn and Heath (discussed above), except that it would be the microtubules rather than the actin fibers that are being disturbed along the direction of substratum curvature. It is intriguing that the current hypotheses all look at contact guidance from the point of view of the inhibition of cell extension in the directions perpendicular to the fibers or grooves. Although such inhibition definitely occurs, perhaps more attention should be given to the converse, that is the question whether the process of cell extension may somehow favored along the axes of fibers and grooves. This is often what it looks like, at least when when cells are plated out on aligned collagen or on scratched plastic surfaces. It would seem to be a simple matter to plate cells out along a boundary between scratched and smooth areas, and simply observe the relative rates of spreading, ( I ) along the grooves, (2) across the grooves, and (3) onto the smooth areas. Certainly 1 would be faster than 2, but might it not also be faster than 3? Brunette (1988) has already demonstrated what happens when fibroblasts cross from an area of grooved substratum to an adjacent area in which the grooves were oriented in the perpendicular axis: their leading lamellae reorient even before their cell bodies do. It may be worth noting that fibroblasts respond very differently to fine-order surface roughness (pits and bumps) than they do to striations (grooves and ridges). Roughness actually inhibits fibroblast locomotion at least as strongly as striations seem to favor it (Rich and Harris, 1981). Nor do the dimensions of this surface roughness have to be very large; roughness having sub-micrometer dimensions is

21 8

ALBERT K. HARRIS

more inhibitory that larger-scale hills and valleys. If you drag the rounded end of a glass rod lightly across a polystyrene surface, you can produce bands a few micrometers in width whose roughness is barely visible by the highest resolution phase contrast microscopy, but which will totally block the locomotion of nearly all fibroblasts. We named this effect “rugophobia”, and it may contain clues as to the mechanism of contact guidance. The alternative hypotheses discussed above would seem to predict responses to fine-order surface roughness. Adhesion plaques wouldn’t be able to form, while actin fibers would be bent. In the case of Curtis’ forthcoming model, one could ask whether and why an elongate ridge should favor the condensation of actin or adhesion molecules, while spikes and pits having similarly microscopic dimensions seem to have the exact opposite effect. Presumably, the photoengraving techniques of Clark et al. could be used to produce substrata with two oblique or perpendicular sets of grooves. Would fibroblasts be as unable to spread on this as they are on a roughened surface? Or might it simultaneously favor spreading in both perpendicular directions? Incidentally, macrophages are very sensitive to contact guidance along fine grooves, but respond to roughened surfaces in exactly the opposite way as fibroblasts-macrophages actually accumulate on the roughened areas and refuse to crawl off of them (Rich and Harris, 1981)! As discussed below, macrophages also respond peculiarly to voltage gradients, where they move toward the positive pole rather than the negative one (Orida, 1980). These are a few of many puzzling cases in which the directional responses of macrophages are exactly the opposite to those of fibroblasts. It is hard to imagine what the causes might be, much less what single difference in cellular properties could explain all three differences in cell behavior. Nevertheless, for any possible molecular mechanism, either for contact guidance or these other two polarization phenomena, it may well be that the oppositeness of the macrophages’ responses can provide the proverbial “exception that proves the rule.” As regards the lower limit of a fibroblast’s sensitivity to the shape of its substratum, some of the most interesting studies were those of Murray Rosenberg (1962, 1963). These used the “Langmuir trough” method as described by Blodgett (1935) to coat glass surfaces with multiple monolayers of a long-chain fatty acid. The fatty acid is first allowed to form amonolayer on the surface of a small dish of water; then the slide or cover slip is dipped very slowly down into the water, lifted out again, dipped in again. If everything is done correctly, a single monolayer of the fatty acid, on the order of only 2 nanometers thick (one five-hundreth of amicrometer), will be deposited onto the glass surface each time it traverses the water surface. This is not quite as easy as it sounds, as the glass surface has to be extremely clean, the water has to be very pure, and the rate at which the glass is raised and lowered should be very steady and very slow (in the range of millimeters per minute: you need some kind of electric motor to do it right). My former colleague Mark Leader and I were allowed by the UNC Chemistry Department to use their facilities to replicate Rosenberg’s classic experiments. The

Fibroblast Polarity

219

most interesting of his experiments involved depositingjust a few layers of behenic acid, then scraping narrow lines (about 10 micrometers wide) through these layers, and then depositing more layers of behenic acid on top of the previous layers and the scratches. What made his results so interesting was his observation that when fibroblasts were then cultured on these surfaces, the fibroblasts accumulated at the sites of the scratches and lined up along them. The fibroblasts did this even when the scratches had been made after the deposition of only a very few layers of the fatty acid, and even though this scratching had been followed by the deposition of dozens of further layers. These results seemed to indicate either that the cells were responding to grooves only tiny fractions of a micrometer in thickness, or that the cells were capable of some kind of long range sensing of the slight differences in substratum thickness. Among researchers on cell motility, Rosenberg’s remarkable discoveries were sometimes compared with the fairy tale about “the princess and the pea”, in which true royalty is identified by the abiliiy to sense a single pea even through a dozen mattresses stacked on top of one another. How fibroblasts might be able to sense shallow grooves through many overlying monolayers was long a topic of inconclusive debates. Our own attempts (A. Harris and M. Leader, unpublished) to replicate these results pointed toward an explanation different than any considered previously. The answer hinges on a subtle aspect of the Langmuir trough method that seems not have occurred to anyone, namely that each successive layer of fatty acid is deposited with a reversed polarity, that between one dip and the next it is the hydrophobic face that forms the surface, and thus scratching down to the underlying glass creates a discontinuity, a hydrophilic line surrounded by a hydrophobic background. The first, third, fifth and successive odd-numbered layers are oriented with their hydrophilic, carboxylic acid surfaces facing toward the glass, and their hydrophobic faces outward. The second, fourth and other even numbered layers have the opposite orientation. Hydrophilic surfaces face toward other hydrophilic surfaces, and hydrophobic surfaces likewise face toward other hydrophobic surfaces. As pointed out by both Langmuir and Blodgett, because glass itself is hydrophilic, this means that the very first layer has its hydrophilic layer facing down toward the glass surface. This first layer is deposited when the glass surface is lifted out of the water after the first dip. The significance of all this for Rosenberg’s experiments may not yet be altogether obvious. The crucial point is that when you have deposited one or more layers of fatty acid onto glass, and then scratch this surface while it is out of the water, then the outermost layer will therefore always be oriented with its hydrophobic surface outward (Figure 10). This means that when you scratch down to the glass through however many fatty acid layers you have deposited, you create a hydrophilic line surrounded by a hydrophobic surface. This, by itself, accounts for the ability of fibroblasts to detect, accumulate and align along scratches. But what about the other experiments, in which Rosenberg made the scratches but then dipped the scratched monolayers in and out many more times? The point is that

220

ALBERT K. HARRIS

figure 10. Diagrammatic cross section of the sequential deposition of fatty acid monolayers onto a glass sheet by the Langmuir trough method. Because glass surfaces are hydrophilic, the first layer of fatty acid is deposited acid-end-downward when the glass is pulled out of the water the first time (c). When it is dipped the second time (d), a second layer of fatty acid is then deposited with its hydrophobic surface facing the hydrophobic surface of the first layer. Then when the glass is pulled back out the second time (d), a third layeroffattyacid isdeposited, like the first layer with its hydrophobicend facingout away from the glass. This means that scratches made between dips create a hydrophilic area (of uncoated glass) surrounded by the hydrophilic face of a fatty acid sheet. Nor can subsequent dips be expected to correct this situation. These considerations suggest a simple explanation for Rosenberg’s classical experiments showing that fibroblasts accumulate in the scratched areas.

monolayer deposition is not at all like applying a coat of paint. For one thing, it is much more sensitive to local surface properties. The reversal of polarity at the scratch site means that further monolayers cannot be expected to deposit over the hydrophilic strip you have created. The principles of monolayer formation predict that this scratched area will just be skipped over, in the same way that no monolayer is deposited when the glass first enters the water. In order to cover the scratch, the additional monolayers would either have had to reverse their polarities (flip upside-down) at the scratch site, or they would have to bridge passively across the gap. Such bridging would require the hydrophobic ends of the behenic acid molecules to stick to the hydrophilic glass. Even in the former case, you would still be left with a hydrophobic surface surrounding a hydrophilic strip where the scratch had been. I conclude from all this that the accumulation of cells at the scratch site is not due to any long range effects, nor to the depth of the groove, but merely to the scratched areas not really having been coated with any of the subsequent layers of fatty acid, and remaining as uncoated hydrophilic glass (Figure 11). Although our observa-

Fibroblast Polarity

221

Figure 71. Low magnification view of the accumulation and alignment of chick heart fibroblasts along a pair of prependicular scratches in a monomolecular layer of behenic acid deposited onto the surface of a glass cover slip. The fibroblasts are preferentially adhesive to the underlying glass, uncovered by the scratches, in preference to the intact layers of behenic acid.

tions matched Rosenberg’s (in the sense that fibroblasts did accumulate and align in the scratches), we believe this was because subsequent layers skipped the scratched areas. By watching the meniscus through a hand lens during deposition, we could see differences in its shape at the scratch sites. Likewise, observations of the coated surfaces with an interference reflection microscope indicated that the scratched regions remained uncoated even after many subsequent dips. I therefore conclude that these classical results were essentially artifactual.

IV.

CONTACT INHIBITION AND CONTACT PARALYSIS

When a leading lamella of one fibroblast crawls into contact with another cell, there is usually an inhibition of its further locomotion in that direction. The result is sometimes for the fibroblast’s locomotion to be stopped altogether, at least for a while; but more usually the result is a change in direction, away from the site ofcontact with the other cell (Figure 12). Especially when the collision is “head on,” the fibroblast’s leading margin stops its forward advance, becoming converted to an inactive state (converting what had been a type A margin into a type B margin). In these cases, the cell as a whole becomes temporarily immobile, but then usually reverses direction. All such events are examples of the phenomenon of contact inhibition, originally discovered by Abercrombie and Heaysman (1953, 1954). In addition to these changes in directions and speeds of locomotion, one usually can also observe inhibition of whatever ruffling movements or protrusion of blebs

222

ALBERT K. HARRIS

Figure 12. Part of a single frame from a 16-mm time lapse film, showing a high magnification view of the occurrence of contact inhibition (and also contact paralysis) beween the advancing margins of two 3T3 fibroblasts.

had been occumng along the parts of the two cells’ margins where the collision occurred (Abercrombie and Ambrose, 1958). This is called “contact paralysis”; or at least it should be; people too often lumpeverything together, using the term contact inhibition interchangeably for both the inhibition of ruffling and for the inhibition of locomotion itself. This term has even been applied to the inhibition of cell growth in division that often occurs in crowded cultures. Such usage implies that these different phenomena share common mechanisms, and reactions against this carelessness have led many to the opposite extreme, the assumption that the mechanisms are not related. We shall be careful with these distinctions here, despite my own expectation that close causal linkages will eventually be found, not only between contact inhibition of locomotion and contact paralysis of ruffling and other cell surface movements, but even between these phenomena and the inhibition of cell cycling in crowded cell populations. Abercrombie and Heaysman’s (1953, 1954) original discovery of contact inhibition was based on a long series of meticulous statistical analyses of cell positions and speeds of fibroblasts’ movement in time-lapse films. In particular, they studied how cell speeds varied as functions of local cell population densities. In one set of experiments, fibroblasts crawling radially outward from tissue explants were shown to be deflected laterally where two such explants were placed close to one another. In otherexperiments, overlaps between fibroblasts were shown to be much less frequent than they should have been if cell positioning were entirely random, and it was found that average speeds of fibroblast locomotion were inversely proportional to the number of other cells with which a fibroblast is in contact. These and related studies have been reviewed in detail by Abercrombie (1967, 1970), by Heaysman (1978), and by Harris (1 974).

Fibrobfast Polarity

223

It is important to emphasize that Abercrombie and Heaysman never claimed that complete stoppages of locomotion result from contacts, only that locomotion is reduced and its direction changed. For example, fibroblasts having contacts with as many as six immediate neighsors (most of which would therefore be completely surrounded by other cells) nevertheless continued to crawl at about half the speed of equivalent fibroblasts that had no cell-cell contacts. Unfortunately, their work has too often been misunderstood and over-simplified by others, who assumed that fibroblast locomotion is entirely prevented by cell-cell contact. The essential point is that locomotion is slowed and redirected: in other words, its polarity is changed. Subsequent studies showed that at least some cancerous cells have a much reduced sensitivity to contact inhibition (Abercrombie et al., 1957), suggesting that the invasiveness of cancerous cells may be partly attributable to an abnormally increased ability to continue locomotory activities (those characteristic of the type A margins) directly adjacent to cell-cell contacts. The general idea is that the function of contact inhibition would be to slow or redirect locomotion by normal cells in proportion to their contacts with other cells, thus allowing them to crawl into wounds or other gaps, but not otherwise to move excessively. Conversely, more recent studies by Paddock and Dunn (1986) suggest that the locomotory activities of some cancerous cells may actually be stimulated where they make contact with fibroblasts . An analogous sort of “reverse contact inhibition” also occurs with some macrophages, in that they accumulate preferentially under fibroblasts (Harris, 1974). Those lines or lunds of cells that are susceptible to contact inhibition are usually spoken of as being “contact inhibited,” and those with less susceptibility, such as various kinds of white blood cells and cancerous cells are said to be “non-contact inhibited.” The greatest obstacle to further progress has been not knowing exactly which intracellular events are being inhibited, or otherwise altered, so as to slow and redirect locomotion (much less why normal fibroblasts should be so much more sensitive to this inhibition than are either leucocytes or cancer cells). The formation and backfolding of lamellipodia (i.e., “ruffling”) are not the only types of cell surface movements that have been shown to be inhibited in areas of cell-cell contact. Bleb protrusion has also been found to undergo a similar paralysis, as has phagocytosis (Vasiliev et al., 1975), as well as the retrograde transport of such labels as Concanavalin A (Vasiliev et aI., 1976). &, as was first pointed out by Trinkaus et al., (1971), contact paralysis is usually very tightly localized to the actual region of cell-cell contact. Paralysis does not spread to the adjacent areas, which often continue to ruffle or bleb indefinitely; this localization can be taken as evidence that the mechanism of inhibition is mechanical in nature, rather than being based on some kind of diffusible chemical. At the time of the original demonstration of contact paralysis by Abercrombie and Ambrose (1958), it was strongly suspected that the cell surface movements being paralyzed represented some kind of peristalsis, such that these movements of the cell surfaces would be a necessary part of the mechanism for exerting propulsive forces. It thus seemed to make sense that inhibiting these surface move-

224

ALBERT K. HARRIS

ments would directly cause inhibition of locomotion: like stopping the wheel of a car from turning. Certainly most people who are shown time lapse films of contact paralysis get a subjective impression that the phenomenon represents some kind of “turning off’ of the cellular motors responsible for locomotion. As naive as the peristalsis idea has proven to be, this subjective impression may not be so wrong after all. What is actually being inhibited is more likely to be polymerization of actin monomers into networks, or rearward transport of tine proteins, or something along those lines. A frequent result of contact inhibition is that fibroblasts tend to become monopolar along boundaries between high and low population densities. In other words, when a fibroblast has many neighbors on one side, and few or no neighbors on the opposite side, the contact inhibition of its leading lamellae along the crowded side tends to polarize locomotion in the direction of the fewest neighbors. This occurs along the boundaries of explants and where cells have been scraped away from monolayers that had previously been confluent. The effect is the locomotion of fibroblasts and other cells into areas of lower population density. Inside the body, this would help to fill in wounds, which is presumed to be one of the main reasons for the evolution of this property. Anyone wanting to study fibroblasts that are nicely polarized and monopolar would do well to use cells with strong susceptibilities to contact inhibition, and then concentrate their attention along boundaries between areas of high and low population densities. Contact inhibition can also contribute to the alignment of bipolar fibroblasts into tracts within which all the cells are nearly parallel to one another. Fibroblasts explanted from certain tissues (the kidneys), including those of certain established cell lines (BHK), are especially prone to this type of alignment, which has been most extensively described in the studies of Elsdale (1968). The experiments of Erickson (1978a) showed how this alignment could be explained in terms of contact inhibition. As she demonstrated, when one fibroblast advances obliquely into contact with the side of another, then one side of its leading lamalla contacts the other cell first, and is therefore inhibited, while the other side of this same leading lamella continues its outward advance. This inhibition of one side of the leading lamella, but not the other, tends to make the advancing cell turn slightly away from the contact, thus bringing itself more nearly parallel with the side of the other cell. When the contacted side of the leading lamella ceases forward spreading, but the noncontacted side continues to crawl forward and pulls somewhat sideways, away from the contact site, then the result tends to rotate the axis of the crawling cell toward the direction parallel to the axis of the cell that it has contacted, as diagrammed in Figure 13. This makes mechanical sense, especially if we imagine that each side of a leading lamella pulls slightly laterally, somewhat like the side horses of a Russian troika, so that weakening one side will rotate the direction of maximum force exertion, away from the weakened, inhibited side. Indeed, subsequent observations with silicone rubber substrata do indicate that individual leading lamellae of fibroblasts frequently exertjust this sort of convergent force; in particular, some of their

Fibroblast Polarity

225

Figure 13. How contact inhibition can cause progressive alignment between fibroblasts, based on the observations of Carol Erickson. When one cell collides with another at too large an angle (top left), it reverses direction (top right). But when the collision occurs at a shallow angle (bottom left), the result is inhibition of its leading lamella on one side more than the other, so that the colliding cell turns slightly into a parallel direction (bottom right).

compression wrinkles are oriented perpendicular to the nearest advancing cell margin. As already mentioned, the mechanistic relation between contact paralysis and contact inhibition remains very uncertain. We very much need to find out exactly what molecular events are blocked near the contacts, what mechanism blocks them, and how cell-cell contact is able to switch what we are calling type A margins into inactive or retreating (type B or even type C ) types of behavior. As already mentioned, a likely explanation is that actin assembly somehow becomes locally blocked; but experimental proof of this is not yet available. Michael Abercrombie himself favored the idea that cell-cell contact induces some kind of increased cytoplasmic contractility, so that the cells are induced to pull apart. This would explain the frequent retraction of cells soon after contact (called “contact retraction”), which is usually away from the site of the cellkell contact, but is sometimes directly toward the contact site. Using flexible substrata and other criteria, I was never able to find direct evidence for such any consistent strengthening of fibroblasts’ longitudinal contractility following contact with another cell, and proposed the alternative hypothesis that contact retraction may instead be caused by induction of weakening of cell-substratum adhesions near sites of cell-cell contact (Harris, 1973b). The idea was that this weakening of cell-substratum adhesions would permit the fibroblast’s pre-existing (and ever-present) contractility to cause retraction. However, this would not explain why cells sometimes contract toward the site of contact inhibition, and when Abercrombie and Dunn (1975) used interference reflection microscopy to look for evidence of the inhibition of cell-substratum adhesions that I had postulated, they failed to find it. They used mercury arc illumination, however, the extreme brightness and near-ultraviolet components of

226

ALBERT K. HARRIS

which interfere with the normal formation and breakage of cell adhesions. (This is itself a phenomenon worthy of more study: for example, leucocytes illuminated by an arc become immobilized and trapped in the light beam, because they cannot detach their old adhesions!) My own experience is that when interference reflection is carried out with sufficiently dim illumination from ordinary tungsten incandescent lamps, you can often see marked inhibition of cell-substratum contact areas just distal to sites of contact inhibition. Obviously, this is a topic deserving further study. A more important and controversial question concerns overlapping between cells, and the extent to which increased overlapping among cancerously transformed fibroblasts is really a consequence of decreased sensitivity to contact inhibition. It had been long assumed that overlapping between adjacent cells could only be the result of active locomotion of one cell up onto the surface of its neighbor, which would necessarily constitute a failure of contact inhibition. However, it was eventually realized that only a small fraction of the lower surfaces of most fibroblasts is actually adhering to a glass or plastic substratum (Harris, 1973a), with what we are calling the “type C” margins offering relatively unobstructed avenues to the type A margins of neighboring cells to crawl underneath (Boyde et al., 1969). Detailed analyses of time lapse films of polyoma virus transformed fibroblasts revealed that most apparent overlaps are really cases of “under-lapping”, in which the advancing cell has continued to use the glass or plastic as its substratum, and has merely crawled into the unoccupied area underneath its neighbor’s type C margins (Guelstein et al., 1973;Di Pasguale and Bell, 1974;Bell, 1977; Erickson, 1978b). In those cases where an advancing type A margin of one cell encountered a type A margin of one of its neighbors, both ruffling and further spreading were found to be inhibited, preventing all but slight overlapping. Bell (using 3T3 fibroblasts) and Erickson (using BHK fibroblasts) did find somewhat more apparent overlapping (“crisscrossing”) between virus transformed fibroblasts than among untransformed ones, although with more nuclear overlapping between the transformed 3T3 cells than between the transformed BHK cells. In both cases, however, the increased overlaps seemed to be due to underlapping, and also to retraction clumping which occurs when one cell pulls loose from its substratum adhesions and retracts onto the surfaces of its neighbors. Erickson, for example, found not even a single case of true overlapping, in the sense of one cell crawling up onto and across the upper surface of another, using the other cell as a substratum, neither among the transformed or the untransformed BHK cells. Such true overlap does seem quite rare, but can occur, at least with very highly transformed fibroblasts; I have seen L cells and sarcoma- 180cells crawl across the upper surfaces of untransformed fibroblasts (see Figure 9 in Harris, 1982). Erickson also demonstrated, however, a very dramatic difference between the transformed and the untransformed BHK cells in the outcome of what one could call “front to side” collisions (where type A margins encounter type C margins). In the case of such encounters between untransformed cells, fully 87% of the contact events resulted in adhesion between the cells, with 64%changing or reversing direction, and some de-

Fibroblast Polarity

227

gree of underlapping in 36%. This was in contrast to the front to side encounters between her transformed BHK cells, in which only 9% resulted in adhesion, 7% resulted in a change in direction, and 93% resulting in underlapping! These results obviously cast a shadow over the naive but appealing view that transformed cells overlap more because of their reduced sensitivity to contact inhibition, and that this is part of the behavioral basis of their greater invasiveness inside the body. On the other hand, we should not be too quick to discard that view altogether. Even if transformed cells have not actually acquired an ability to use the surfaces of other cells as substrata, the work of both Bell and Erickson seems to show big increases in these cells’ ability to continue crawling forward in areas where they are closely underlapping and touching other cells. After all, equivalent nontransformed cells would quickly have undergone both contact paralysis and contact inhibition. The difference may lie only in a reduced ability to form new cell-cell adhesions, or perhaps in an increased ability to continue “type A” margins adjacent to cell-cell contacts, but these or other changes could also contribute to increased invasiveness inside the body. This is one of many examples where an increased understanding of the molecular nature of the events occurring along type A margins should soon provide a new perspective on a large body of older knowledge.

V.

GALVANOTAXIS: EFFECTS OF ELECTRIC FIELDS ON FIBROBLAST POLARITY

Fibroblasts react to voltage gradients by slowly elongating in the direction perpendicular to the axis of the gradient, and by reorienting in this direction if they had previously had a long axis in some other direction (Figure 14). This perpendicular response is in contrast to the behavior of at least some epithelial cells, which respond to electric fields by crawling toward the negative electrode, not to mention nerve cells, whose response is extension of growth cones toward the negative electrode. It is also in contrast to macrophages, which despite resembling miniature fibroblasts in shape and most behavior, have been shown to respond to voltage gradients by crawling directionally toward the positive electrode (Orida, 1980). Another cell type showing this type of directionally-reversed galvanotaxis is the osteoclast, which serves to break down bone; this is in contrast to the bone-depositing osteocytes, which migrate toward the cathode (Femer et al., 1986). It is not known whether these differences have anything to do with the opposing roles of these two cell types in skeletal morphogenesis, nor if there is any relation to the widelyhypothesized role of electric fields in bone growth. Note, however, that osteoclasts are believed to be formed by the fusion of macrophages with one another. It would be interesting to know how many of the different macrophage-like cells of the body, such as microglia in the nervous system and Kupfer cells in the liver, also move preferentially toward the positive ends of electric gradients.

228

ALBERT K. HARRIS

Chick heart fibroblasts, cultured on a silicone rubber substratum, that have oriented in response to a voltage gradient. The axis of the voltage gradient is perpendicular to the direction of cell alignment.

Figure 14.

It remains to be discovered exactly what differences in membrane or cytoskeletal components, or other differentiated cell characteristics, cause such differences in &rectionality of response. Besides fibroblasts, it has been found that neural crest cells, among others, respond by perpendicular elongation. To a variable degree, fibroblasts also crawl slowly toward the negative electrode. In my experience, however, the perpendicular orientation is by far the more dramatic and consistent effect. The fibroblasts’ contractility likewise becomes directed perpendicular to the voltage axis. The morphological reorientation occurs gradually over 20 to 30 minutes, and has now been observed in cultured fibroblasts from a wide variety of sources, including established cell lines and cells explanted from embryos. In most studies of this phenomenon, the steepness of the voltage gradients has been in the range of a volt per millimeter (equal to a millivolt per micrometer), which is considerably higher than is found within the tissues of developing embryos. Nevertheless, Nuccitelli andErickson (1983) have shown that fibroblasts can also orient in response to voltage gradients only a tenth this steep, and that is in the range of field strengths actually found in developing embryos. Their results thus suggest that galvanotaxis could be a functional mechanism of tissue morphogenesis. It is important to stress that these galvanotactic orientations depend on reasonably steady, long-lasting voltages, acting in a certain direction and not reversing polarity. Thus, while much larger transient voltages occur in many tissues, for example as a result of rapid stressing of cartilage, tendons and bone, fibroblasts are only found to reorient in response to sustained uni-directional (direct current) voltages. As is discussed below, fibroblasts do not seem to show any response to alternating current voltages, which has important implications not only for the possible functional significance of galvanotaxis. but also for its subcellular mechanism.

Fibroblast Polarity

229

Galvanotaxis by tissue culture cells, specifically neural growth cones, was first reported by S. Ingvar in 1920. However, Paul Weiss (1934) was unable to repeat those results, and pronounced that whatever effects Ingvar had observed must really have been secondary consequences of reorientation of extracellular fibers of plasma clot around the electrodes. Weiss proposed that contact guidance, rather than galvanotaxis, must have caused Ingvar’s cells to align along these extracellular fibers. Considerably later, Weiss and Scott (1963) reported their unsuccessful attempts to produce any changes in fibroblast locomotion by means of imposed voltages, even though these appear to have been well within the range of field strengths subsequently been shown to produce alignment. Such was Weiss’ influence, however, that the whole topic of tissue cell galvanotaxis went into suspended animation until the early 1980s,when it was more or less simultaneously rediscovered by several independent groups of researchers (Hinckle et al., 1981; Luther et al., 1983; Nuccitelli and Erickson, 1983; Cooper and Keller, 1984; Erickson and Nuccitelli, 1984; reviewed by Robinson, 1985; see also Soong et al., 1990). Among other things, these workers found that it is essential to keep the culture chamber quite thin; otherwise too much heating will be produced by the electrical current, thus killing the cells before they can reorient. This heating effect may explain Weiss’ inability to confirm what Ingvar had discovered. Fibroblast reorientation seems to occur predominantly by the withdrawal of any leading lamellae that are oriented parallel to the voltage gradient, followed by the extension of leading lamellae along the axis perpendicular to this gradient (Harris et al., 1990).Enlargement of preexisting leading lamellae oriented in this perpendicular direction also occurs. Variations are sometimes seen in the extent of this reorientation, however, sometimes for no very obvious reasons. I have seen whole cultures in which practically no orientation could be produced, as well as groups of recalcitrant cells within other cultures in which the remainder had all reoriented; my impression is that published papers in this field tend to understate the degree of such unexplained variations. Apparently, the cells’ locomotion has to be reasonably active for reorientation to occur. Unusually immobile cells, and especially quiescent cultures often won’t respond. I have also noted special insensitivity on the parts of those fibroblasts that happen to be adhering to one another more than to the glass or plastic substratum. Such unexplained variations are annoying, of course, but they may also serve as clues for the design of future experiments. As to the mechanism of fibroblast galvanotaxis, we need to ask two distinct classes of question. First, how does the cell “feel” the electric field: for example is it the electrophoresis of charged molecules in the plasma membrane, or is it the distortion of the transmembrane “resting potential” difference in voltage between the cytoplasm and the surrounding medium? The second type of question concerns the cells’ responses: for example, do the leading lamellae contract more strongly on the sides facing up and/or down the voltage gradient (and thus preferentially pull themselves loose from the substratum)? Alternatively, are the cell-substratum adhesions differentially weakened on the sides of the cell facing up and down this gradient?

230

ALBERT K. HARRIS

Or, as a third alternative, is the outward spreading of leading lamellae somehow favored on the sides of the cells not facing either pole of the electric field? It is essential to realize that these two classes of question are logically distinct. A cell’s preferential withdrawal of extension of lamellae could result from either electrophoretic effects or changes in transmembrane voltage. Likewise, it might be useful to reflect that differentiated cell types might differ from one another in the mechanisms of their responses to electric fields, or even that mixtures of effects may occur. Nerve cells might use one mechanism, and macrophages another. Only fibroblasts will be considered here, however. Note also that I am assuming that cells detect voltage gradients, rather than detecting the actual flow of electrical current that results from this voltage. Some who have written on galvanotaxis, Weiss in particular, seem to have been making the opposite assumption. But my reasoning is that the permeability of the plasma membrane to electric current is so much lower than that of either tissue culture medium or cytoplasm, that very little current can be expected to flow through the cell bodies. Likewise, when you feel the wind blowing past you, or when a house blows down in a hurricane, what is felt are the differences in air pressure (equivalent to voltage differences) and not the cubic volumes of air moving past (equivalent to electrical current amperage). In the case of a cell culture, the amount of current flowing has more to do with the electrical resistance of the surrounding medium (including its dimensions) than with any property of the cells. I do not believe it would be difficult to make tissue culture media whose electrical conductivities differed over a fairly wide range, so someone might wish to test the effects of different amounts of current with the same rate of change of voltage, as compared with the converse. It is also important to realize that both the electrophoresis of membrane components and also the biasing of the transmembrane potential will necessarily occur when you put a fibroblast in a voltage gradient. Neither effect can help occurring to some degree. Thus, the question is not which one occurs, but which one the cells are responding to when they reorient their axes. It is inevitable that any charged material on the cell’s surface will feel a displacement force proportional to its own charge and to the steepness of the voltage gradient being imposed. The extent to which this force will be resisted by the viscosity of the membrane, or overcome by other forces, are separate questions. But there will be a force. Likewise, because the electrical resistance of the plasma membrane is enormously larger than the resistances of either the cytoplasm or the surrounding medium, we can also be certain that only a negligible voltage gradient can be maintained between one end of a cell and the other, and similarly that very little current will flow there. This implies that we can discount the electrophoresis of any cytoplasmic materials. Just as in nerve and muscle cells, the transmembrane voltage of fibroblasts results primarily from the differences in the concentrations of potassium ions (high inside the cell and low outside the cell), combined with an appreciable permeability of the plasma membrane to potassium ions leaking through it out of the cell. Because none of the other ions are as free to leak though the membrane as is potassium,

Fibroblast Polarity

231

and because potassium ions are positively charged, the net result is a voltage (outside positive, inside negative) that is proportional in strength to the log of the inside/outside ratio of potassium concentrations. This voltage difference tends to be smaller in fibroblasts than in nerves or muscles; and is also not subject to active reversal by sodium-driven action potentials. In an externally-imposed voltage gradient, the much lower conductivity of the plasma membrane relative to either the medium or the cytoplasm means that the voltage gradient inside the cell will be much less steep than the one outside. This leads us to the important conclusion that the transmembrane voltage will have to vary from point to point along the cell surface, as you move along the axis of the external voltage gradient. Note that a fibroblast’s transmembrane potential of 30-40 millivolts or so is in the same range of magnitude as the voltage difference that is produced across the 30-40 micrometer width of a fully spread fibroblast by an imposed voltage gradient of about one millivolt per micrometer. On the other hand, it would be reassuring if someone were actually to measure the transmembrane voltages of different fibroblasts and fibroblast lines, to see how closely these match the steepnesses of the voltage gradients imposed on them. It seems obvious that when the voltage is initially lower inside a cell than outside, and an external voltage gradient is then imposed on the surroundings, but (because of the relative lack of permeability of the plasma membrane) the cytoplasmic voltage remains nearly constant throughout the cell, then the result must be for the end of the cell facing the positive electrode to become hyperpolarized, while the opposite side of the cell (in the direction facing the negative electrode) becomes depolarized (see diagram Figure 15). It would be useful for someone to use the fluorescent dye methods referred to above to confirm that the plasma membrane of a fibroblast subjected to a voltage

-

Outside cell Inside cell

........

Voltage

-

.:............... .:.:: . .:.:.:.:.:<->; j:i ...................................

:

. . .

...,!

Outside cell Irside cell In a gradient

figure 75. Diagram of the expected distribution of relativevoltages, inside and outside a fibroblast, when it is not in a voltage gradient (top), as compared with when a voltage gradient i s imposed on the culture medium around it (bottom). The point i s that the end of the cell toward the negative electrode will become depolarized while the end toward the positive electrode will become hyperpolarized.

232

ALBERT K. HARRIS

gradient really does vary according to this expected pattern. It would also be useful to know (from direct measurements, rather than theoretical calculations) which part of the cell remains at its original voltage, whether this is something that varies from one cell (or from one cell type) to another, and what kinds of ion fluxes occur as aresult of these imposed voltage gradients. For example, are we to suppose that potassium just continues to leak out through the depolarized parts of the plasma membrane, or that equal amounts of this ion are drawn in through the hyperpolarized areas? An especially important question is to what extent fibroblast membranes are like those of muscles in responding to depolarization by locally increased permeability to calcium ions. In a voltage gradient, this would be expected to occur at the end of the cell facing the negative electrode. Such localized changes in ion concentrations in the cytoplasm have been suggested as possible causes for morphological alignment. But although several laboratories have tested this possibility using calcium-sensitive probes, I believe with negative results, I am not aware that any such results have yet been published. One experimental approach to the question of whether the cells are responding to depolarization or to electrophoresis is to compare the effects of alternating currents, direct currents, and intermittent currents. Nuccitelli and Erickson (1983) discovered that cells pay no apparent attention to alternating current such as that from the power mains, which reverses its polarity 60 times per second. This approach was extended by Harris et al., (1990) who used an electronic timing device to reverse the direction of the voltage gradient at longer intervals: specifically once per second, once per ten seconds, and once per minute. Our idea was that 60 times per second might somehow be too fast for the cells to notice, or for cytoplasmic changes to take effect, but that longer durations would allow sufficient time for the opposite sides of the cells to complete their various hyperpolarizations, depolarizations and ion leakages (if any). Fibroblasts might then be expected to reorient their morphological axes, especially since this reorientation is essentially symmetrical with respect to the voltage axis. In other words, even though the cells would first be depolarized at one end, and then hyperpolarized at that same end and depolarized at the opposite one, again and again, time after time, they might still react in the same way as to adirect current, by withdrawing the leading lamellae at both the hyperpolarized and the depolarized sides. After all, if withdrawal is the response both to depolarization and to hyperpolarization, then why shouldn’t the response be the same to a regular alternation between these two states? In fact, what we found was that fibroblasts did not align in response to even these very slow alternating voltages. Instead, they continued to crawl about for 2 hours or more, with apparently normal morphology and behavior. These same cells then required only the usual 20 to 30 minutes to align once the reversals of polarity were stopped and the direction of the voltage was held constant. As a response to Cooper’s helpful suggestion that the ion pumps at each extremity of a fibroblast might effectively be “bailing out” whatever ions had leaked in during the periods of opposite polarity, we then did equivalent experiments with the same timer in which the volt-

Fibroblast Polarity

233

age was turned on for one minute, then off for the next minute, then on again with the same original polarity, then off again, and so forth. This would allow the “bailing out” to occur, even though subjecting the cell only to voltages of one of the two polarities. Nevertheless, the fibroblasts did reorient to these intermittent voltages, and in little more than their usual 30 minutes. Taken together, these results indicate that reversing the direction of a voltage has the consequence of undoing its previous effects (i.e. not just producing a different effect, but actually reversing that produced earlier by the voltage having the opposite polarity). This result, of course, is exactly what would be predicted by the hypothesis that the galvanotactic signal is electrophoretic in nature. Conversely, I do not see any obvious way to explain these responses in terms of ion leakages. If you had a ship and first let water leak in one side, then let water leak in the other side, and repeating this indefinitely, you could keep the ship on an even keel, but would eventually sink it. More research is needed, but the current evidence seems to indicate that at least the perpendicular orientation aspect of fibroblast galvanotaxis is most likely to result from the electrophoretic displacement of one or more membrane components into the ends of the cells that face the electrodes. These membrane components then somehow promote retraction of the cell extremities into which they have been accumulated. The second of our two classes of question concerns the physical nature of the fibroblast’s response. Three specific alternatives have been considered: (a) localized or directional strengthening to the contractility of leading lamellae, so that they pull loose from the substratum; (b) localized weakening of the adhesiveness of these leading lamellae to the substratum, so that their existing contractility becomes sufficient to retract them; or (c) favoring the protrusion of those leading lamellae that are oriented perpendicular to the voltage gradient. As a test of the first of these three possibilities (and to some extent also as a test of the third) fibroblasts were cultured on silicone rubber substrata, allowed to spread sufficiently to produce many compression wrinkles under themselves, and then subjected to a voltage gradient. Not only the wrinkles in the rubber, but also markers on its surface were observed to see whether any changes in the strength or directionality of cell contraction occurred in response to the imposed voltage. In the most sensitive of these experiments, tiny rectangular islands of silicone rubber were cut out of the sheet, with only one or a few cells per island, and left floating on the silicone fluid surface. The idea was that even the tiniest changes in direction or strength of cellular forces would be visible as changes in the shapes and dimensions of these islands. But the results were negative. The anticipated increase in contractility along the axis of the voltage gradient simply did not occur. Instead, there was a gradual weakening of the cells’ contractile force in this axis, beginning after 10 or 20 minutes and continuing until the cells had completely reoriented and were exerting little or no tension in this direction. Accompanying and following this slow reorientation, there was a gradual increase in substratum wrinkles running transverse to the cells’ new axis, indicating a gradual increase in the contractile force being exerted in the direction perpendicular to

234

ALBERT K. HARRIS

the voltage gradient. Because this increased force in the perpendicular direction developed so gradually, it was judged to be the effect, rather than the cause, of the morphological reorientation. The results of this experiment thus seemed to point to the second of the three hypotheses: achange in the spatial distribution of each cell’s adhesiveness. Interference reflection microscopy (Izzard and Lochner, 1976) offers a more direct approach to this question of whether fibroblasts respond to voltage gradients by changing the spatial distribution of their substratum adhesions. In fact, the study by Harris et al., (1990) originally included the use of this optical technique, with observations of cell-substratum contacts disappearing on the sides of fibroblasts facing the positive electrode and the apparent formation of broad dark contact areas on the sides of the fibroblasts facing the negative electrodes. In other words, the voltage gradients seemed to favor the formation of new adhesions on the negative side and to favor the breakage of adhesions on the positive side. This process was recorded by time-lapse video, and observed throughout entire cultures in several replicate experiments. Nevertheless, in later experiments, the formation of these broad, dark contact areas inexplicably failed to occur, even though the fibroblasts did undergo the usual alignment perpendicular to the voltage gradient. We were unable to determine what difference in the cells or the culture conditions was responsible and omitted the interference reflection observations from the final version of the paper. I mention this in the hope that some reader of this review may be stimulated to reexamine this phenomenon. As was for so long true of galvanotaxis itself, there is a tendency just to ignore the very existence of phenomena so long as some of the variables that control them remain unknown.

VI.

MICROTUBULES IN THE CONTROL OF FIBROBLAST POLARITY

Several lines of evidence indicate that microtubules must play some important role in governing the polarization of fibroblasts. For one thing, cytoplasmic microtubules are generally oriented along a fibroblast’s long axis, when it has one; and reorientation of a fibroblast’s axis (for example in response to voltage gradients, see Harris et al., 1990) is accompanied by reorientation of the microtubules. Furthermore, when fibroblasts are crawling in a definite direction, such as along the edge of a scraped area “wound” or other boundary between high and low population densities, their microtubule organizing centers (MTOCs) usually become positioned between the nucleus and the leading margin (Singer and Kupfer, 1986). The Golgi apparatus also becomes localized in this area. Likewise, there is evidence of preferential stabilization of those microtubules directed toward the cells’ leading margins (Gunderson and Bulinski, 1988). The question is whether these phenomena should be regarded as actual causes of the cells’ directionality and polarity, as opposed to being consequences or byproducts. In other words, does the end of the cell that

Fibroblast Polarity

235

faces toward the lower population density become the front end because the MTOC is located there? Or does the MTOC move there because this end has become the front, for example because of the new patterns of contractile forces within the cytoplasm? For some years, many people in this field seem to have tended to the former conclusion: that MTOC position actually caused directionality, rather than the other way around. More recently, however, observations on the time course of changes in MTOC position have led to a general reversal of opinion (Euteneuer and Schliwa, 1992). In careful studies of the time course of changes during reorientations of fibroblasts along the edges of scraped regions, it was found that the MTOCs do eventually move to the side of each cell between its leading edge and its nucleus; but this change in MTOC position often occurs after the change in direction of movement, implying that it more a result than a cause. In addition, it has been found that when fibroblasts are cultured in a collagen gel, then the position of the MTOC bears no particular geometric direction to the leading edge or direction of the cells’ movement, very much in contrast to the situation when equivalent cells are grown on flat glass or plastic (Schutze et al., 1991). This last observation is quite puzzling from any point of view, but certainly doesn’t support the idea of MTOCs as controlling directionality. Unfortunately, there doesn’t seem to be any way to do the converse types of experiments and force microtubules to reorient in a chosen direction. We might, for example, imagine putting an iron filing inside a MTOC, and then using a magnet to pull it around from one side of a fibroblast to another. Would this cause a corresponding redirection of the fibroblast’s direction of locomotion and morphological polarity? No one knows. On the other hand, it is not difficult to disrupt or even eliminate cytoplasmic microtubules using spindle poisons such as colchicine, vinblastine or nocodazole, and the long term morphological effects of these drugs can be quite dramatic. As Vasiliev and his colleagues first demonstrated, exposure to these drugs tends to reduce or eliminate fibroblast polarity ( et al., 1970; Vasiliev and Gelfand, 1976; see also Vasiliev and Gelfand, 1977; see also Vasiliev. 1991).Fibroblasts treated with these drugs sometimes lose their polarity so completely that they revert to a “fried egg” morphology, in which most or all of the cell margin switches to what we are calling the type A class of behavior, extending 360 degrees around the periphery. An example of such a cell is shown in Figure 16; before treatment with vinblastine this cell had been elongate and bipolar with a narrow leading lamella at each end. The implication of such morphological changes seems to be that fibroblasts need their microtubules to subdivide or polarize different parts of their margins into distinct types of behavior. This result is very much as if intact cytoplasmic microtubules were somehow required to maintain the category of margins we are calling type C (and perhaps also the type B margins). Sensitivity to this morphological effect varies widely among fibroblasts, with those from primary cultures (from explants) not being susceptible, in other words

236

ALBERT K. HARRIS

Figure 76. High magnification view of a single vinblastine-treated BHK fibroblast.

Notice the cell’s nearly circular “fried egg” shape, which is very much in contrast to the usual elongate shape of BHK cells. Notice also the formation of a ring of small blebs, or lobopodia, all concentrated along the cell margin. A few of the characteristic vinblastine-induced crystals of microtubule protein are also visible in the cytoplasm surrounding the nucleus.

undergoing little or no morphological depolarization in response to any of the antimicrotubule drugs. No one knows why fibroblasts recently derived from explants should fail to undergo the morphological depolarization, nor why the effect is so dramatic in certain established lines (BHK cells are particularly good). One plausible hypothesis had been that it might reflect the possession (by insensitive cells) of larger sub-populations of especially long-lived and otherwise stable microtubules, for example those made of detyrosinated tubulin. Recent observations by Middleton et al. (1989) do not seem to support this type of explanation, in that fibroblasts from primary cultures were not found to contain any greater proportions of tubules containing these modified tubulins. In fact, given that fibroblasts from primary cultures continue not to undergo the morphological depolarization, even after prolonged exposures (of days) to substantial concentrations of any of the anti-microtubule drugs, the conclusion seems inescapable that the difference must lie in how the cells respond to their lack of microtubules, as opposed to how completely the cells are deprived of them. In other words, fibroblasts from primary culturesjust must not need microtubules to maintain their polarity. In contrast, fibroblasts from many established lines seem unable to maintain their polarity without microtubules. Maybe polarity is normally maintained by Some kind of combined effort of several cytoplasmic elements,one of which is microtubules, but the other elements are able to do the job by themselves in fibroblasts from primary cultures. Cell lines that have spent much longer in culture, becoming aneuploid and otherwise jaded, apparently lack these other contributors to polarity.

Fibroblast Polarity

237

What seems to be a related effect occurs when fibroblasts treated with antimicrotubule drugs are cultured on “haptotactic substrata”, in particular those substrata having adhesive metal-coated “islands” separated by lines where the substratum is less adhesive. Normal fibroblasts conform only very roughly to the dimensions and boundaries of the adhesive islands, extending lamellae beyond their edges in some places, while retracting well inside these boundaries in other places. But when the same fibroblasts are grown in media containing microtubule poisons, they change behavior so as to conform much more exactly to the adhesive boundaries (Harris, 1973a). Such cells often become exactly square or triangular, taking on the precise size and shape of the island they happen to be adhering to (see Figure 17). As with the “fried egg” morphology, it is as if fibroblasts deprived of their microtubules lack autonomy in controlling their shape. There are several different kinds of mechanisms by which cytoplasmicmicrotubules might contribute to fibroblast polarity, three of which will now be discussed. One such category of mechanism would simply be that the physical strength and resistance to bending of microtubules would provide enough mechanical support to favor elongation and polarity. This class of explanation is frequently encounteredin implicit form, but seems not to have been developed explicitly. A second class of possible mechanism is that polarization results from the transport functions known to be provided by microtubules,using kinesins or dyneins. The best evidence for the second of these possibilitiescomes from recent work by Rodionov et al. (1993) who injected antibodies specific for the motor domains of the protein kinesin. This pro-

Figure 77. Vinblastine-treated fibroblasts cultured on a series of “haptotactic islands” consisting of squares of vacuum-evaporated palladium metal overlying a nonwettable (and relatively nonadhesive) polystyrene culture dish. Notice how the shapes of these cells approximate the artificial shapes of the adhesive islands. This is in contrast to equivalent fibroblasts not treated with microtubule poisons, which also tend to be confined to the adhesive areas, but adopt their usual stellate shapes much more independently of the shapes of the islands themselves.

238

ALBERT K. HARRIS

tein, of course, is one of those that serves to transport materials along microtubules. The injection of these antibodies did not inhibit the development of normal arrays of cytoplasmic microtubules. But it did mimic some of the morphological effects of colchicine and other microtubule poisons, implying cell polarity does not depend SO much on the presence or absence of microtubules, but rather depends on whether materials can be transported along microtubules. For example, unspecified proteins might be transported to some regions of the cell periphery in preference to other regions, thereby causing these areas to behave differently in respect to contractility, adhesiveness or other properties. In previous work, Vasiliev’s group had studied the disruption of the actin cytoskeleton in response to phorbol ester tumor promoters, which act by overstimulating protein kinase C (Brown et al., 1989; Bershadsky et al., 1990) . A principal result of this over-stimulation is that actomyosin stress fibers break down, and what we here are calling the type A margins crawl forward actively, stretching the cell body out into thin cytoplasmic necks (Dugina et al., 1987). Among the other discoveries made by members of this group was that fibroblasts that had been treated with microtubule poisons became essentially immune to this phorbolinduced redistribution of their actin cytoskeleton (Lyass et al., 1988). Quite reasonably, they interpreted this in terms of the microtubules being needed in some way to transport or otherwise redistribute the actin or other cytoskeletal proteins to new locations and arrangements. Parallel studies by Danowski and Harris had shown that phorbol esters cause a rapid weakening of the tractional and contractile forces exerted by fibroblasts, as revealed by reductions in the wrinkling of rubber substrata on which the cells were cultured (Danowski and Harris, 1988). These studies also found that fibroblasts treated with phorbol esters become able to crawl onto relatively nonadhesive and hydrophobic substrata, to which their adhesions had previously been too weak to permit spreading. This apparently results from their weakened contractility rather than from any increase or other change in their adhesiveness per se. When Danowski (1989) studied whether these effects on contractility might likewise be inhibited by microtubule poisons, following up the work of Lyass et al. described above (1988), she found several very unexpected things. First she confirmed that treating fibroblasts with amicrotubule poison along with the phorbol ester will indeed prevent the usual effect of the phorbol: the actin cytoskeleton is not disrupted. But then she discovered that even if cells have already been treated with the phorbol, and their actin cytoskeletons have already been disrupted, if you then add colchicine or one of the other microtubule inhibitors to the medium, this results in a restoration of the actin cytoskeleton! The cells’ contractility, measured by distortions of rubber substrata, is also restored even though the phorbol ester is still present. In other words, microtubule poisons don’t just prevent the cytoskeletal effects of phorbol esters, they actually counteract and reverse these effects. Obviously, this is not at all what one would expect if the role of microtubules in the process were that of transporting cytoskeletal materials to their new locations. If

Fibroblast Polarity

239

that had been the explanation, then microtubule poisons ought to prevent restoration or normal cytoskeletal patterns, not promote it. An even more surprising discovery (Danowski, 1989) was that colchicine, vinblastine and nocodazole all cause large and rapid strengthening of fibroblast contractility, even in cells not treated with phorbol esters. Within a few minutes of treatment with any one of these three antimicrotubule drugs, fibroblasts increase their contractility to levels two or even three times as strong as they had been exerting just prior to treatment. Danowski discovered this using the (relatively nonquantitative) silicone rubber substratum technique; but by using electronic strain gauges Kolodney and Wysolmerski (1992) were able to confirm her conclusions quite precisely for the case of large numbers of fibroblasts pulling on collagen gels. Danowski also showed that these anti-microtubule drugs promote formation of actomyosin stress fibers, and that not only do these drugs prevent and reverse the disruption of stress fibers by phorbol esters, they also block and reverse the disruption of the cytoskeleton by quite unrelated drugs, including ones acting by altering cytoplasmic concentrations of cyclic AMP. As if all this weren’t surprising enough, when Danowski tried a wide range of concentrations of the anti-microtubule drugs, she discovered that their contractility and stress fiber-promoting effect occurs even at extremely low drug concentrations - so low that many microtubules are visible when the fibroblasts were stained with antibodies against microtubules, and so low that mitosis can still occur (note, the same effects also occur at higher concentrations of these drugs). What all this seems to imply is that what matters is not so much the continued existence of cytoplasmic microtubules, but something more subtle, like their state of dynamic equilibrium. Merely disturbing the microtubules, or their process of assembly, seems to be enough. It somehow favors assembly of actin into stress fibers, thereby magnifying the contractile forces that fibroblasts exert on their surroundings, and preventing (and reversing) the disruption of these stress fibers with TPA and other treatments. To explain the increased contractile force itself, the most obvious interpretation would be that microtubules normally exert a substantial pushing force, which is more than counterbalanced by the pull of the acto-myosin stress fibers. With the removal of these microtubules, the full strength of the actin-based contractility would be revealed. This type of explanation would fit in well with Donald Ingber’s (1993) recent attempts to explain cell mechanics in terms of “tensegrity.” Indeed, given that fibroblast contractile force doubles or even triples in response to colchicine or equivalent drugs, this pushing interpretation would imply that microtubules exert very strong forces indeed. They would have to push with forces of the order of at least 2 or 3 hundreths of a dyne per cell, that being the order of magnitude of the forces that fibroblasts ordinarily apply to rubber substrata (Harris et al., 1980). I doubt if any such direct pushing can be the correct interpretation, however. For one thing, it would not explain why disruption of microtubules is able to reverse the actin-disrupting effects of phorbol ester tumor promoters and poisons of cyclic

240

ALBERT K. HARRIS

AMP metabolism. Furthermore, Danowski also tried the effects of the antimicrotubule drug taxol, which acts by promoting microtubule assembly rather than inhibiting it. This drug has inconsistent effects on the exertion of contractile forces, but even though it caused all the microtubules to form a dense radial ball at one end of the cells, nevertheless when the taxol was washed out and replaced by medium containing nocodazole or equivalent drugs, the result was still a large increase in contractile force relative to what the cells had been exerting in the presence of the taxol. The point here is that the microtubules in the taxol-treated cells couldn’t have been doing any effective pushing along the cell’s lengths, since they were all concentrated into a ball at one end. These and other results strongly support Danowski’s conclusion that there must be some actual stimulation of actin fiber assembly produced by the disruption of microtubules. It is a puzzle why this should occur, either in terms of mechanism or functional significance. One class of possibilities is that one or more of the MAPS (microtubule associated proteins), being released into the cytoplasm by the disrupted microtubules, could promote assembly either of actin itself, or of one of the other components of stress fibers. Functionally, it could be part of the special mechanism for controlling locations of actin polymerization, with the dynamic instability of microtubules serving to control actin organization.To the extent that this control were inhibitory in nature; then it would make sense that any generalized depolymerization of cytoplasmic microtubules would, as a side effect, promote an equally generalized polymerization of cytoplasmic actin. One of the consequences would be the loss of directional polarity, while another would be the strengthening of contractility.

VII. QUESTIONS ABOUT THE AUTONOMY OF FIBROBLAST POLARITY Some of the more important questions one can ask about fibroblast polarity concern the continuity of differences in polarity over time, and the mechanisms by which variations in polarization can be passed by a cell to its mitotic progeny. To see dramatic evidence of such inheritance, one need only take fibroblasts of an established line, trypsinize them and plate them out on glass or plastic at low enough population densities to produce discrete, clonal colonies, and then wait long enough for 3 or 4 cell cycles to have been completed. When doing this in the course of other experiments, I have frequently been strongly impressed by “family resemblances” between the 8 to 16 or so cells making up each clone. You see a bunch of long straight ones here, a colony of circular ones there, and above it a clone with nearly all triangular cells. Any observant tissue culturist will have noticed such recurring patterns, although the effect is of course much less striking in more homogeneous cultures, such as primary explants and secondary cultures, since the cell morphologies are so much more uniform within the population as a whole.

Fibroblast Polarity

241

The systematic study of similarities in sibling cells, especially their locomotory behaviors, has somehow become the province of a single researcher, AlbrechtBuehler, whose work I will now attempt to summarize. He began with reports that mitotic sister cells tend to crawl in paths that are mirror images of one another (Albrecht-Buehler, 1977), including some degree of mirror image similarity between the actin cytoskeletons of sister cells; he then proposed that cell directionality is controlled by the 9+0 structure of the centriole, with this control being reflected in a tendency of fibroblasts to turn at angles preferentially close to some multiple of 40 degrees (Albrecht-Buehler, 1979,1981). In case the reason for 40 degrees is not obvious, it is 360 degrees divided by the 9 outer doublets of the typical centriole or basal body structure. Not only does he regard the centriole as “the brain of the cell”, but also proposes that it serves as a visual organ, by means of which fibroblasts are able to extend lamellae directionally toward sites of emission of near infrared light (Albrecht-Buehler, 1991, 1992), especially if this light flashes on and off a hundred or so times per second. This is suggested to be a mechanism of communication between cells, even on opposite sides of sheets of glass, by which they control one another’s directions of elongation. This same author has also proposed that the computational processes of the cellular “brain,” equivalent to synapses and transistors, are accomplished at the molecular level by assembly and disassembly of protein monomers, such as tubulin and actin, into cytoplasmic fibers (Albrecht-Buehler, 1985).The ideais that monomer assembly encodes information in a manner analogous to that of a digital computer. We can regard it as a credit to the openness of science to new and imaginative ideas that all these proposals have been published in the best journals. We also must ask whether the boundary has been crossed from the imaginative to the imaginary.

Figure 78. A pair of mitotic sister 3T3 fibroblasts, just after completing division. Notice how nearly they seem to be mirror images of one another. In fact, until seconds before the picture was taken, there had also been nearly a mirror image pattern of compression wrinkles beneath the cells in the silicone rubber membrane on which they are spreading.

242

ALBERT K. HARRIS

With regard to whether mitotic daughter cells are sometimes mirror images of each other, or whether they follow mirror image paths, the question is not whether such things sometimes occur, but whether mirror images occur significantly more frequently than would be expected to occur at random. We also need to ask what specific features of the cells’ movements need to be non-randomly correlated in order to produce a given type and frequency of apparent mirror images. Figure 18 shows a pair of mitotic sibling fibroblasts which not only were noticeably near to being mirror images of one another, and (up until just before the photograph was taken!) had even been producing wrinkling patterns in their rubber substrata that were also mirror images! Even the most negative critic should admit such things do occur, sometimes. But how often? There can be an awful lot of fibroblasts in one Petri dish. The proposal of mirror-image paths arose out of studies of the ability of fibroblasts to clear away small particles in their paths by means of the retrograde transport process discussed previously. Albrecht-Buehler (1977) showed that this phenomenon can be used to track cell paths, calling the cleared areas “phagokinetic tracks” (although phagocytosis plays only a minor role, if any; and the accumulation of too many particles on a cell’s surface was already known to alter its behavior). By first coating a substratum with closely spaced particles of colloidal gold, and then observing by dark field microscopy after the fibroblasts have had a chance to crawl about, one finds that the cells’ paths become dramatically visible by their relative darkness, compared with the brightness of the gold that has been left in place. One can easily observe thousands of cell paths, including the branched paths of cells that had undergone one or more mitotic divisions during the time since they were plated onto the gold covered surface, The illustrations in the paper show over a dozen such cases in which the paths of mitotic sibling cells seemed to be approximate mirror images of each other; and when the sibling cells were fluorescently stained for actin stress fibers, considerable degrees of similarity were found in their cytoskeletons as well. Albrecht-Buehler reported that approximately 60% of mitotic sister cells moved in paths that he classified either as approximate mirror images, or as “identical”. At this point, we need to ask ourselves how frequently such paths should occur randomly, i.e. if there were no greater correlation between the directions and speeds of sister cells than there is between unrelated cells. Intuitively, it probably doesn’t seem as if this random frequency could possibly be anywhere near as large as 60%. Figure 19 shows a series of computer-generated pairs of paths meant to be analogous to those of mitotic sistercells. How many of these pairs would you count as being “mirror images”? Many people might count only a, b and c. On the other hand, Albrecht-Buehler also counted pairs in which the turning directions were the same, rather than opposite; these are the pairs he classifies as “identical”; this means counting d and e, as well as a, b, and c. His reasoning was that one of the cells might have gotten turned upside down during mitosis, so that its path was really a mirror image. Thus all but the last (f) above, would be counted. In other words, if neither

Fibroblast Polarity

243

Five possible pairs of pathsof mitotic sister cells. In a, neither undergoes any major turns; in b, c, d, and e, each undergoes a single turn; and in f one turns and the other doesn’t. By Albrecht-Buehler’s criteria, all but f would be classified as mirror images.

Figure 19.

cell turns during the observation period, or if both turn only once (as is quite often the case, to judge from Albrecht-Buehler’s published illustrations, and also based on my own experience), then all possible paths would be counted. The only way to escape that fate is to make a different number of turns, or to turn more than once; and for the case of two turns each, no fewer than half the possible directional combinations would be counted as “mirror images”. The point is that daughter cells need only resemble one another in respect to how frequently they turn in order to produce quite a high percentage of “mirror images.” Perhaps what should surprise us is that Albrecht-Buehler did not find even more than the reported 60%, using such elastic criteria. The eye can often “see” patterns in what is actually randomness, for example in the case of the Rorschach (inkblot) test in psychology. Figure 20 shows a series of computer-generated random path-pairs, about half of which would be counted as approximate “mirror images” by the criteria used by Albrecht-Buehler. These are the first 10 paths generated by a program, written in Turbo-Pascal on a Macintosh computer, with “directions” and “speeds” being assigned by sequences of random numbers. For each of the pairs, two circles were started out at the same location; displacements in the X and Y directions were separately determined for each, based on the compiler’s own random number sequence. For each of the 40 recalculations per cell, its new speed in the X direction equals the old “X speed” plus a random number between -9 and +9; and its new speed in the Y direction likewise equals its old Y speed plus another random number in this same range. There is no connection at all between the calculations for one cell and its sibling. In addition, for each cell at each of the 40 cycles, there is a random one-tenth chance that its speeds in both the X and Y directions will be reset to zero, so as to produce the effect of a random turn. The percentage of apparent “mirror-images” can be increased further by such variations as making the turns occur every so many cycles. My conclusion is that there need not be any real correlations in the directions of cell turns to produce the types and frequencies of “mirror images” reported by Albrecht-Buehler. The eye can find pat-

244

ALBERT K. HARRIS

figure 20. Ten path-pairs that were generated by a computer program. Although all these paths are actually entirely random, the majority could easily be classified as ”mirror images.“ terns even in randomness; and the author will be glad to provide free copies of this Pascal program to anyone interested in trying it out. The reports of positive phototaxis to certain frequencies of infrared light (Albrecht-Buehler, 1991,1992) also deserve comment. There is certainly no reason to doubt the possibility of some such taxis; there are plenty of other physical stimuli to which fibroblasts are known to orient themselves, including several reviewed above. On the other hand, the reported need for the light’s intensity to oscillate at certain frequencies would certainly be a puzzle, and is also part of what makes independent repetition of this work very much more difficult than it seems at first. The use of a special sapphire lens is also part of this difficulty. As in earlier the “mirror image” work, these reports lack explicit criteria for comparing frequencies of extension of cellular processes toward the light source, as compared with what should occur randomly. Anyone knows, who has observed the particular type of fibroblasts used in these studies, that they are constantly and continually extending and retracting leading lamellae and lamellipodia of all sorts of sizes and durations of existences. This means you have to decide which protrusions are going to “count”. For example, you might decide to count only those protrusions of greater than, say, 5 micrometers in width and length, and of these only those that maintain at least these dimensions for longer than 10 minutes. Equally essential are criteria for the directionality of these protrusions. For example, you might draw an imaginary line from the center of the cell’s nucleus outward to the furthest tip of each cytoplasmic protrusion, or perhaps a line bisecting the two sides of the protrusion. With such criteria in hand, one could then count the frequency with which a fibroblast extends processes to within plus or minus 20 degrees of any given

Fibroblast Polarity

245

direction. Alternatively, you might subdivide the cell periphery into quadrants and count the frequency of protrusions into each one. What you can’t do is just look at the video recordings and say “that extension counts as being toward the light source; that other one doesn’t.” If the cells in question were like Amoebaproteus, with an actual front and arear, or even if they had the degree of directionality of leucocytes, than subjectivejudgments could perhaps be sufficient, especially if the directional response were marked. Nor is the situation aided by doing statistical analyses on one’s previous subjective estimates of which protrusions should “count,” especially when that observer is strongly committed to the hypothesis. As regards the demonstrable tendency of fibroblasts on opposite sides of very thin sheets of glass to orient perpendicular to one another, I would suggest that this is an example of contact guidance and reflects the slight curvature of the glass sheets. On the convex surface, fibroblasts tend to orient in the direction of minimum curvature; while on a slightly concave surface, their direction tends to be across the curvature, with the cells arranged like bow strings relative to the curve of the substratum. One might try flexing such a thin sheet slowly, so that one side is convex one day, with the other side becoming convex the next day. The point is that such phenomena often have fairly mundane explanations, without the need to postulate infrared Morse code being sent from one cell to another.

VIII. CONCLUSIONS AND PROSPECTS To summarize the current situation in this field: much information has been gathered and still is accumulating about the spatial distributions within crawling fibroblasts of the various cytoskeletal and membrane proteins apparently responsible for their locomotion (Beshadsky and Vasiliev, 1988).The force-exertion mechanism of this locomotion seems at last to have been identified, and consists of a continuous assembly of cytoplasmic actin localized along the cell’s leading margins, with actin fibers flowing centripetally from there just beneath the plasma membrane, pulling certain integral membrane proteins along with them, with these “tine proteins” serving to transmit quite strong shearing forces (traction) tangentially through the plasma membrane, from the inside to the outside (reviewed in Heath and Hollifield, 1991;Harris, 1990).On the other hand, this information is just beginning to be applied to questions of cell shape and polarity. Presumably, a fibroblast’s polarity reflects the localization of such processes as actin assembly, the formation of new cell-substratum adhesions, the aggregation of transmembrane adhesion proteins into focal plaques, the disruption of old adhesions, and so on, so that they occur in the correct spatial locations relative to one another. Each molecular and physical event of locomotion must somehow be concentrated at the right places, rather than occurring either randomly or homogeneously. We also have to presume that this localization is self-organizing as well as

246

ALBERT K. HARRIS

self- perpetuating; otherwise, trypsinized cells would not be able to re-establish polarity and maintain it for periods of a few hours or more. Here are some specific questions. By what means is actin assembly focused so precisely right along the outer edges of what we have been calling the “type A” margins? Perhaps the mechanisms are comparable to those found in leucocytes (Cassimeris et al., 1990). Why and how are the type I myosins concentrated at these edges? Do they somehow promote or localize actin assembly, in the sense that if they were relocated to what had been a type B margin, would actin assembly then begin there? Conversely, which of the special protein constituents of these type A margins would you have to take away in order to convert it to either a type B margin or a type C margin? Another sort of question would involve forces. What does it mean that the herniation of the plasma membrane into blebs is often localized right along the edge of type A margins? It could be that the resistance to outward pressure is either weakest or least stable there; or it could be that the outward forces themselves are concentrated there. Based on subjective impressions of watching too many time lapse films, I would suggest that an important controlling factor may be the state of tension in the cell body. Certainly, an generalized loss of tension will almost always lead to a reorganization of cell polarity. Of course, we already know that tension is created by the outward crawling and traction of the type A margins; but perhaps this tension itself promotes the subdivision of the cell margin into type A versus type C behaviors; in particular, the type C behavior would be promoted where the direction of stretching is approximately parallel to the edge, whereas type A behavior would be promoted where the stretching is directed perpendicular to the edge. If so, then one might predict that rotating the directions of maximum tension, either using flexible substrata or micromanipulator needles, would allow you to control which parts of a cell’s margins behaved in the “A,” the “B,” or the “C” patterns. Another promising direction for research in this field would seem to be that of reinterpreting directional control phenomena, such as contact inhibition and contact guidance, in terms of what we now know about “actin treadmilling” and “membrane raking.” Perhaps we will now also be able to explain how these forcegenerating processes get inactivated near sites of cell-cell contact, why some cells are more susceptible to this inactivation than others, and to what extent reductions in this susceptibility contribute to the greater invasiveness of such things as macrophages and cancer cells. Likewise, in the case of contact guidance or galvanotaxis, we need to ask which molecular processes are promoted or inhibited, and how external cues produce these effects. Also important and timely is the control of actin organization and contraction by microtubules, and specifically by their disassembly. Some sort of controlling function has long been suspected for microtubules in relation to cell polarity and actin organization; but no one expected the discovery by Danowski that disassembling microtubules causes increased assembly of actin, and even reverses disruption of actin by other treatments. Small and Rinnerthaler (1985) have noted that the forrna-

Fibroblast Polarity

247

tion of new adhesion plaques tends to occur near the tips of microtubules, no mechanism for this has been known (see also Geiger et al., 1984; also see Rinnerthaler et al., 1988). Possibly it is related to the formation of the actin stress fibers extending from the adhesion plaque. Certainly, in the case of cytokinesis, it has long been suspected that the tips of the aster microtubules may serve to signal to cortical actin where to form the contractile ring, and thereby the cleavage furrow. In that case, it has been debated whether the aster microtubules should inhibit cortical contraction or promote it; but no one seems to have asked if it is merely the presence of microtubules that cause the stimulation or inhibition, or whether these effects might actually be caused by the disassembly of the microtubule tips. Recent discoveries about the dynamic equilibrium state of cytoplasmic microtubules seem to raise this latter possibility. These are the types of questions we now need to be asking about the control of fibroblast polarity.

REFERENCES Abercrombie, M. (1967). Contact inhibition: the phenomenon and its biological implications. Natl. Cancer Inst. Monogr. 26,249-277. Abercrombie, M. (1970). Contact inhibition in tissue culture. In Vitro 6, 128-142. Abercrombie, M. & Ambrose, E. J. (1958). Interference reflectionmicroscopystudies ofcell contacts in tissue culture. Exp. Cell Res. 15, 332-345. Abercrombie, M. & Dunn, G. A. (1975). Adhesions of fibroblasts to substratum during contact inhibition observed by interference reflection microscopy. Exp. Cell Res. 92, 57-62. Abercrombie, M. & Heaysman, J. E. M. (1953). Observations on the social behaviour of cells in tissue culture. I. Speed of movement of chick heart fibroblasts in relation to their mutual contacts. Exp. Cell Res. 5 , 111-131. Abercrombie, M. & Heaysman, J. E. M. (1954). Observations on the social behaviour of cells in tissue culture. 11. Monolayering of fibroblasts. Exp. Cell Res. 6,293-306. Abercrombie, M., Heaysman, J. E. M. & Karthauser, H. M. (1957). Social behaviour of cells in tissue culture. 111. Mutual influence of sarcoma cells and fibroblasts. Exp. Cell Res. 13, 276-291. Abercrombie, M., Heaysman, J. E. M. & Pegrum, S. M. (1970a). The locomotion of fibroblasts in culture I. Movement of the leading edge. Exp. Cell Res. 59,393-398 Abercrombie, M., Heaysman, J.E.M. & Pegrum, S. M. (1970b). The locomotionoffibroblasts in culture 11. ‘Ruffling’ Exp. Cell Res. 60,437444. Abercrombie, M., Heaysman, J.E.M. & Pegrum, S. M. (1970~).The locomotionof fibroblasts in culture Ill. Movements of particles on the dorsal surface of the leading lamella. Exp. Cell Res. 62, 389-398. Abercrombie, M., Heaysman, J. E. M. & Pegrum, S. M. (1971). The locomotionoffibroblasts in culture IV. Electron microscopy of the leading lamella. Exp. Cell Res. 67,359-367. Abercrombie,M.,Heaysman, J. E. M. &Pegrum,S. M. (1972). Thelocomotionoffibroblastsinculture V. Surface marking with concanavalin A. Exp. Cell Res. 73,536-539 Albrecht-Buehler, G. (1977). Daughter 3T3 cells, are they mirror images ofeach other? J. Cell Biol. 72, 595-603 Albrecht-Buehler, G. (1979). The angular distribution of directional changes in guided 3T3 cells. J. Cell Biol. 80, 53-60.

248

ALBERT K. HARRIS

Albrecht-Buehler, G. (1981). Does the geometric design of centrioles imply their function? Cell Motility 1,237-245 Albrecht-Buehler,G. (1985). Iscytoplasmintelligenttoo?InCellandMuscleMotility. J. W. Shay(ed.), VOI. 6. pp. 1-21. Albrecht-Buehler, G. (1991). Surface extensions of 3T3 cells towards distant infrared light sources. J. Cell Biol. 114,493-502. Albrecht-Buehler, G. (1992). Rudimentary form of cellular “vision.” Proc. Natl. Acad. Sci. USA. 89, 8288-8292. Bell, P. B. Jr. (1977). Locomotory behavior, contact inhibition, and pattern formation of 3T3 and polyoma virus-transformed cells in culture. J. Cell Biol. 74, 963-982. Bershadsky, A. D. & Vasiliev, J.M. (1988). Cytoskeleton. Plenum Press, New York. Bershadsky, A. D., Ivanova, 0. Y., Lyass, L. A., Pletyushkina, 0. Y., Vasiliev, J. M., & Gelfand, I. M. (1990). Cytoskeletal reorganizations responsible for the phorbol ester-induced formation of cytoplasmic processes: Possible involvement of intermediate filaments. Proc. Nat. Acad. Sci. USA 87, 1884-1888. Blodgett, K. B. (1935). Films built by depositing successive monomolecular layers on a solid surface. J. Am Chem. SOC.57, 1007-1022. Britland, S. Clark, P. Connolly , P. & Moores, G.(1992). Micropatterned substratum adhesiveness, a model for morphogenetic cues controlling cell behavior. Exp. Cell Res. 198, 124-129 Brown, A. F. Dugina, V.; Dunn, G. A. & Vasiliev, J. M. (1989). A quantitative analysis ofalterations in the shape of cultured fibroblasts induced by tumour-promoting phorbol ester. Cell Biol. Int. Rep. 13,357-366. Brunette, D. M. (1986). Spreading and orientation of epithelial cells on grooved substrata. Exp. Cell Res. 167,203-217. Brunette, D. M. (1988). The effect of surface topography on cell migration and adhesion. In, Surface Charaterization of Biomaterials. Ratner, BD (ed.), pp.203-217. Elsevier Science Publishers, Amsterdam. Brunette, D. M., Kenner, G. S. & Gould, T. R. L. (1983). Grooved titanium surfaces orient growth and migration of cells from human gingival explants. J. Dent. Res. 62, 1045-1048. Boyde, A,, Grainger, F. & James, D. W. (1969). Scanning electron microscopic observations of chick embryo fibroblasts in vitro, with particular reference to the movement of cells under others. Z. Zellforsch. mikrosk. Anat. 94,4&55. Buck, R. C. (1980). Reorientation response of cells to repeated stretch and recoil ofthe substratum Exp. Cell Res. 127,47&474. Burridge, K. & Connell, L. (1983). A new protein of adhesion plaques and ruMing membranes. J. Cell Biol. 97,35%367. Burridge, K. & Fath, K. (1989). Focal contacts: transmembrane links between the extracellular matrix and the cytoskeleton. Bioessays 10, 104-108. Campbell, R. D. & Marcum, B. A. (1980). Nematocyte migration in Hydra: evidence for contact guidance in vivo. J. Cell Sci. 41,33-51. Cassimeris, L., McNeill, H. & Zigmond, S. H. (1990). Chemotactant-stimulated polymorphonuclear leucocytes contain two populations of actin filaments that differ in their spatial distribution and relative stabilities. J. Cell Biol. 110, 1067-1075. Chen, W-T. (1981). Mechanism of retraction of the trailing edge during fibroblast movement. J. Cell Biol. 90, 187-200. Clark, P. Connolly, P., Curtis, A. S. G., Dow, J. A. T. & Wilkinson, C. D. W. (1987). Topographical control of cell behavior: I . Simple step cues. Development 99,439448. Clark, P. Connolly, P., Curtis, A. S. C., Dow, J. A. T. & Wilkinson, C. D. W. (1990). Topographical control of cell behavior: 11. Multiple grooved substrata. Development 108,635464. Clark, P.Connolly, P., Curtis, A. S. G., Dow, J. A. T. & Wilkinson, C. D. W. (1991). Cell guidance by ultrafine topography in vitro. J. Cell Sci. 99,73-79.

fibroblast Polarity

249

Clark, P., Connolly, P. & Moores, G. R. (1992). Cell guidance by micropatterned adhesiveness invitro. J. Cell Sci. 103,287-292. Cooper, M. S. & Keller, R. E. (1984). Perpendicular orientation and directional migration of amphibian neural crest cells in DC electrical fields. Proc. Natl. Acad. Sci. USA 81, 16G164. Curtis, A. S. G. &Clark, P. (1990). The effects oftopographic and mechanical properties of materials on cell behavior. Crit. Rev. Biocompat. 5,343-362. Danowski, B. A. (1989). Fibroblast contractility and actin organization are stimulated by microtubule inhibitors. J. Cell Sci. 93,255-266. Danowski, B. A. & Harris, A. K. (1988). Changes in fibroblast contractility, morphology and adhesion in responses to a phorbol ester tumor promoter. Exp. Cell Res. 177,47-59 DePasquale, J. A. & lzzard, C. S. (1987). Evidence for an actin-containingcytoplasmicprecursor ofthe focal contact and the timing of incorporation of vinculin at the focal contact. J.Cell Biol. 105, 2803-2809. DePasquale, J. A. & Izzard, C. S. (1991). Accumulation of talin in nodes at the edge of the lamellipodium and separate incorporation into adhesion plaques at focal contacts in fibroblasts. J. Cell Biol. 113, 1351-1359. DiPasquale, A. & Bell, P. B. (1974). The upper cell surface: its inability to support active cell movement in culture. J. Cell Biol. 62, 198-214. Dow, J. A. T., Clark, P., Connolly, P., Curtis, A. S. G. & Wilkinson, C. D. W. (1987). Novel methods for the guidance and monitoring of single cells and simple networks in culture. In, Cell Behavior: Shape, Adhesion and Motility. Heaysman, J. E. M., Middletion, C. A,, & Watt, F. M. (ed.) J. Cell Sci. Suppl. 8, 55-79. Dugina, V. B., Svitkina, T. M., Vasiliev, J. M. & Gelfand, I. M. (1987). Special type ofmorphological reorganization induced by phorbol ester: reversible partition of cell into motile and stable domains. Proc. Natl. Acad. Sci. USA 84,41224125. Dunn, G. A (1991). How do cells respond to ultrafine surface contours? Bioessays 13,541-543. Dunn, G. A. &Brown, A. F. (1986). Alignmentoffibroblastson groovedsurfaces described by asimple geometric transformation. J. Cell Sci. 83,313-340. Dunn, G. A. & Heath, J. P. (1976). A new hypothesis ofcontact guidance in tissue cells. Exp. Cell Res. 101, 1-14. Elsdale, T. R. (1968). Parallel orientation of fibroblasts in vitro. Exp. Cell Res. 51,439450. Erickson, C. A. (1978a). Analysis of the formation of parallel arrays by BHK cells in vitro. Exp. Cell Res. 115,303-315. Erickson, C. A. (1978b). Contact behaviour and pattern formation of BHK and polyoma virus-transformed BHK fibroblasts in culture. J. Cell Sci. 33, 53-84. be influenced Erickson, C. A. &Nuccitelli, R. (1984). Embryonicfibroblastmotilityandorientationcan by physiological electric fields. J. Cell Biol. 98,296-307. Euteneuer, U. & Schliwa, M. (1992). Mechanism ofcentrosome positioningduring the wound response in BSC-I cells. J. Cell Biol. 116, 1157-1 166. Ferrier, J., Ross, S. M., Kanehisa, J. & Aubin, J. E. (1986). Osteoclasts and osteoblasts migrate in opposite directions in response to a constant electrical field. J. Cell Physiol. 129,283-288. Fisher, G. W., Conrad, P. A,, DeBiasio, R. L. & Taylor, D. L. (1988). Centripetal transport of cytoplasm, actin, and the cell surface in lamellipodia of fibroblasts. Cell Motil. & Cytoskel. 11, 235-247. Forscher, P. & Smith, S. J. (1988). Actions of cytochalasins on the organization of actin filaments and microtubules in a neuronal growth cone. J. Cell Biol. 107, 1505-1516. Garber, B. (1952). Quantitative studies on the dependenceofcell morphology andmotility uponthe fine structure of the medium in tissue culture. Exp. Cell Res. 5, 132-146 Geiger, B., Avnur, Z., Rinnerthaler, G., Hinssen, H. & Small, J. V. (1984). Microfilament organizing centers in areas of cell contact: Cytoskeletal interactions during cell attachment and locomotion. J. Cell Biol. 99, 83-91,

250

ALBERT K. HARRIS

Gelfand, V. I., Glushankova,N. A,, Ivanova,O. Yu., Mittelman, L. A,, Pletyushkina,O. Yu. Vasiliev. J. M. & Gelfand. I . M. (1985). Polarization of cytoplasmic fragments microsurgically detached from mouse fibroblasts. Cell Biol. Int. Rep. 9,883-892. Guelstein, V.I., Ivanova, 0. Y., Margolis, L. B., Vasiliev, J. M. & Gelfand, I.M. (1973). Contact inhibition of movement in the cultures of transformed cells. Proc. Natl. Acad. Sci. USA 70, 201 1-2014. Gundersen, G . G. & Bulinski, J. C. (1988). Selective stabilization of microtubules oriented toward the direction of cell migration. Proc. Natl. Acad. Sci. USA 85, 59465950. Harris, A. K. (1973a). Behavior of cultured cells on substrata ofvariable adhesiveness. Exp. Cell Res. 77,285-297. Harris, A. K. (1973b). Location of cellular adhesions to solid substrata. Dev. Biol. 35, 97-1 14. Harris, A. K. (1974). Contact inhibition of cell locomotion. In Cell communication. R.P. Cox (ed.) John Wiley & Sons, New York, pp. 147-185. Harris, A. K. (1982). Traction and its relations to contraction in tissue cell locomotion. In Cell Behavior: a tribute to Michael Abercrombie. R. Bellairs, A. Curtis & Dunn, G. (eds.) Cambridge U. Press, Cambridge, pp. 109-134. Harris, A. K. (1990). Protrusive activity of the cell surface and the movements of tissue cells. In Biomechanics of Active Movements and Deformation of Cells. Akkas, N. (ed.), NATO AS1 Series H: Cell Biology, vol. 42. pp. 249-294. Springer-Verlag, New York. Harris, A. K. & Dunn G. A. (1972). Centripetal transport of attached particles on both surfaces of moving fibroblasts. Exp. Cell Res. 73, 519-523. Harris, A. K., Wild, P. & Stopak, D. (1980). Silicone rubber substrata: anew wrinkle in the study ofcell locomotion. Science 208, 177-179. Harris, A. K., Pryer, N. K. & Paydarfar, D. (1990). Effects of electric fields on fibroblast contractility and the cytoskeleton. J. Exp. Zool. 253, 163-176. Harrison,R. G. (1914). Thereactionofembryoniccells tosolidstructures. J. Exp. Zool. 17,521-544. Heaysman, J.E.M. (1978). Contact inhibitionoflocomotion:areappraisal. Int. Rev. Cytol. 55,4946. Heath, J. P. & Dunn, G. A. (1978). Cell to substratum contacts of chick fibroblasts and their relation to the microfilament system. A correlated interference-reflexion and high voltage electron-microscope study. J. Cell Sci. 29, 197-212. Heath, J. P. & Holifield, B. F. (1991). Cell locomotion: new research tests old ideas on membrane and cytoskeletal flow. Cell Motility and the Cytoskeleton 18, 245-257. Hinckle, L. C. D., Mc Caig, C. D. & Robinson, K. R. (1981). The direction of growth of differentiating neurons and myoblasts from frog embryos in an appiied electric field. J. Physiol. (Lond.) 314, 121-135. Holifield, B. F., Ishihara, A. & Jacobson, K. (1990). Comparative behavior of membrane protein-antibody complexes on motile fibroblasts: Implications for a mechanism of capping. J. Cell Biol. 111,2499-2512. Ingber, D. E. (1993). Cellular tensegrity: defining new rules of biological design that govern the cytoskeleton. J. Cell Sci. 104, 613-627. Ingvar, S. (1920). Reactions of cells to the galvanic current in tissue cultures Proc. Am. SOC.Exp. & Med. 17, 198 Ishihara, A,, Holifield, B. & Jacobson, K. (1988). Analysis of lateral redistribution of a monoclonal antibody complex plasma membrane glycoprotein which occurs during cell locomotion. J. Cell Biol. 106,329-343. Izzard, C. S. (1988). A precursor ofthe focal contact in cultured fibroblasts. Cell Motil. Cytoskeleton. 10, 137-142. Izzard, S. L. & Izzard, C. S. (1987). Actin-associated proteins related to focal and close cell-substrate contacts in murine fibroblasts. Exp. Cell. Res. 170,214-227. Izzard, C. S. & Lochner, L. R. (1976). Cell to substrate contacts in living fibroblasts: an interference relexion study with an evaluation of the technique. J. Cell Sci. 21, 129-159.

Fibroblast Polarity

251

h a r d , C. S. & Lochner, L. R. (1980). Formationofcell-to-substratecontacts duringfibroblastmotility: an interference-reflexion study. J. Cell Sci. 42,81-116. Kolodney, M. S. & Wysolmerski, R. B. (1992). Isometric contraction by fibroblasts and endothelid cells in tissue culture; a quantitative study. J. Cell Biol. 117, 73-82. Luther, P.W., Peng, H. B. & Lin, J. J-C. (1983). Changes in cell shape and actin distribution induced by constant electric fields. Nature 303, 61-64. Lyass, L. A., Bershadsky, A. D., Vasiliev, J. M. & Gelfand, I. M. (1988). Microtubule-dependenteffect of phorbol ester on the contractility of cytoskeleton of cultured fibroblasts. Proc. Natl. Acad. Sci. USA. 85,9538-9541. Middleton, C. A., Brown, A. F., Brown, R. M., Karavanova, 1. D., Roberts, D. J. & Vasiliev, J. M. (1989). The polarization of fibroblasts in early primary cultures is independent of microtubule integrity. J. Cell Sci. 94,25-32. Nuccitelli, R. & Erickson, C. A. (1983). Embryonic cell motility can be guided by physiological electric fields. Cell Motil. 2,243-255. Oakley, C. & Brunette, D. M. (1993). The sequence of alignment of microtubules, focal contacts and actin filaments in fibroblasts spreadingonsmooth and grooved titanium substrata. J. Cell Sci. 106, 343-354. O’Hara, P. T. & Buck, R. C. (1979). Contact guidance in vitro. A light, transmission and scanning electron microscopic study. Exp. Cell Res. 121,235-249. Orida, N (1980). Directed lamellipodial protrusive activity in macrophages induced by extracellular electric fields J. Cell Biol. 87, 92a. Paddock, S. W. & Dunn, G. A. (1986). Analysingcollisionsbetween fibroblasts and fibrosarcomacells: fibrosarcomacells show an active invasionary response. J. Cell Sci. 81, 163-187. Rich, A. M. & Harris, A. K. (198 1). Anomalous preferences of cultured macrophages for hydrophobic and roughened surfaces. J. Cell Sci. 50, 1-7. Rinnerthaler, G., Geiger, B. & Small, J. V. (1988). Contact formation during fibroblast locomotion: Involvement of membrane ruffles and microtubules. J. Cell Biol. 106, 747-760. Robinson, K. R. (1985). The responses of cells to electrical fields: A review. J. Cell Biol. 101, 2023-2027. Rodionov, V. I., Gyoeva, F. K., Tanaka, E., Bershadsky, A. D., Vasiliev, J. M. & Gelfand, V. I. (1993). Microtubule-dependent control of cell shape and pseudopodial activity is inhibited by the antibody to kinesin motor domain. J. Cell Biol. 123, 181 1-1820. Rosenberg, M. ( I 962). Long-range interactions between cell and substratum. Proc Natl. Acad. Sci. USA 48, 1342-1349. Rosenberg, M. (1963). Cell guidance by alterations in monomolecular films. Science 139, 41 1-412. Astudy Rovensky, Y. A. & Slavnaja, 1. L (1974). Spreadingoffibroblast-likecellsongroovedsurfaces. by scanning electron microscopy. Exp. Cell Res. 84, 199-206. Schutze, K., Maniotis, A. & Schliwa, M. (1991). The position of the microtubule-organizing center in directionally migrating fibroblasts depends on the nature of the substratum. Proc. Natl. Acad. Sci. USA 88,8367-8371. Singer, S. J. & Kupfer, A. (1986). The directed migration of eukaryotic cells. Ann. Rev. Cell Biol. 2, 337-365. Small, J. V. (1981). Organization of actin in the leading edge of cultured cells: Influence of osmium tetroxide and dehydration on the ultrastructure of actin networks. J. Cell Biol. 91,695-705. Small, J. V. & Rinnerthaler, G. (1985). Cytostructural dynamics of contact formation during fibroblast locomotion in vitro. Exp. Bio. Med. 10, 54-68. Soong, H. K., Parkinson, W. C., SuIik, G. L. & Bafna, S. (1990). Effects of electric fields on cytoskeleton of corneal stromal fibroblasts. Cum. Eye Res. 9, 893-901. Stopak, D. & Harris, A. K. (1982). Connective tissue morphogenesis by fibroblast traction I . Tissue culture observations. Dev. Biol. 90, 385-398.

252

ALBERT K. HARRIS

Svitkina,T. M.,Neyfakh, A. A. Jr. & Bershadsky,A. D. (1986). Actincytoskeletonofspreadfibroblasts appears to assemble at the cell edges. J. Cell Sci. 82,235-248. Symons, M. H. & Mitchison, T. J. (1991). Control of actin permeabilization in live and permeabilized fibroblasts. J. Cell Biol. 114,503-513. Trinkaus, J. P. ,Betchaku, T. & Krulikowski, L. S. (1971). Local inhibition of ruffling during contact inhibition of cell movement. Exp. Cell Res. 64,291-300. Vasiliev, J. M. (1991). Polarization of pseudopodial activities: cytoskeletal mechanisms J. Cell Sci. 98, 1-4. Vasiliev, J. M. &Gelfand, I. M. (1976). Effectofcolcemidon morphogeneticprocessesand locomotion of fibroblasts. In Cell Motility, ed. Goldman, R., Pollard T. & Rosenbaum. J. (eds.), Cold Spring Harbor Conferences on Cell Proliferation, Vol. 3, Book A, pp. 279-304, New York, Cold Spring Harbor Laboratory. Vasiliev, J. M. & Gelfand, I. M. (1977). Mechanisms of morphogenesis in cell cultures. lnt. Rev. Cytol. 50, 159-274. Vasiliev, J. M., Gelfand, I. M.,Domnina,L. V., Ivanova,O. Y. Komm, S.G. & Olshevkaja,L. V. (1970). Effect of colcemid on the locomotory behavior of fibroblasts. J. Embryol. Exp. Morphol. 24, 625-640. Vasiliev, J. M., Gelfand, I. M., Domnina, L. V., Zacharova, 0. S. & Ljubimov, A.V. (1975). Contact inhibition of phagocytosis in epithelial cell sheets: Alterations of cell surface properties induced by cell-cell contact. Proc. Natl. Acad. Sci. USA 72, 719-722. Vasiliev, J. M., Gelfand,l.M., Domnina,L.V., Dorfman,N.A. & Pletyushkina,O.Y. (1976). Activecell edge and movements of concanavalin A receptors of the surface of epithelial and fibroblastic cells. Proc. Natl. Acad. Sci. USA 73,4085-4089. Wagner, M. C., Barylko, B. & Albanesi, J. P. (1992). Tissue distribution and subcellular localization of mammalian myosin I. J. Cell Biol. 119, 163-170. Wang, Y-L. (1985). Exchange ofactin subunits at the leading edge ofliving fibroblasts: Possible role of treadmilling. J. Cell Biol. 101,597-602. Weiss, P. A. (1934). In vitro experiments on the factors determining the course of the outgrowing nerve fiber. J. Exp. Zool. 68,393-448. Weiss, P. A. (1961). Guidingprinciples in cell locomotionand cell aggregation. Exp. Cell Res. Suppl. 8, 260-28 1. Weiss, P. A. & Garber, B. (1952). Shape and movement of mesenchyme cells as functions of the physical structure of the medium. Contributions to a quantitative cell morphology. Proc. Nat. Acad. Sci. USA 38,264-280. Weiss, P. A. & Scott, B. I. H. (1963). Polarizationofcell locomotion in vitro. ProcNatl. Acad. Sci. USA 50,33&336. Witkowski, J. A. & W. D. Broughton (1971). Stages of spreading of human diploid cells on glass surfaces. Exp. Cell Res. 68,372-380. Zand, M. S. & Albrecht-Buehler, G. (1989). What structures, besides adhesions prevent spread cells from rounding up? Cell Motility and the Cytoskeleton. 13,54-67. Zand, M. S. & Albrecht-Buehler, G. (1992). Mechanical perturbationofwebbededges in 3T3 cells. Cell Motility and the Cytoskeleton. 21, 195-21 1.