Actomyosin contraction of the posterior hemisphere is required for inversion of the Volvox embryo

Folding of a multicellular sheet is a widespread embryological phenomenon observed during gastrulation, neurulation and various other aspects of morphogenesis. Such processes have been analyzed extensively at the cellular level, where it has been repeatedly observed that the cells constituting a folding sheet undergo a predictable series of morphological changes in a spatiotemporally regulated fashion (Ettensohn, 1985; Fristrom, 1988). There have, however, been significantly fewer analyses of the subcellular events that underlie these coordinated changes in cell behavior during folding.

Inversion of Volvox embryo provides a useful model to study sheet folding (Kirk, 1998; Kirk et al., 1982). A Volvox embryo is a hollow spherical monolayer of about 3,000 cells. All cells in the embryo, including the 16 larger cells that are destined to become ‘gonidia’ (reproductive cells), are structurally connected by numerous cytoplasmic bridges. There is, however, a region at the anterior pole where the syncytium is interrupted by an opening called the phialopore (Green et al., 1981). As inversion starts, a portion of the sheet encircling the phialopore curls outward to produce a circular bend region. The bend then moves progressively toward the opposite pole (Kelland, 1977), eventually causing the spherical sheet to turn inside out. Volvox inversion bears some similarity to metazoan gastrulation, except that it progresses reversely and goes to completion.

During bend formation, two cellular events, i.e., specific shape changes and movements of the cytoplasmic bridges, have been reported to take place (Kelland, 1977; Viamontes and Kirk, 1977; Viamontes et al., 1979; Green et al., 1981). Prior to inversion, the cells are spindle shaped and linked by cytoplasmic bridges at their widest points. Cells in the bend form long, narrow projections at their outer ends and the cytoplasmic bridges move so that they become bundled at the tips of these projections. The bend is then propagated progressively from the phialopore toward the opposite pole where these two cellular events are taking place. Propagation of the bend obviously requires precise spatiotemporal regulation of cell shape and movement (Green et al., 1981).

Kirk and his colleagues have implicated microtubules and actin filaments in the two cellular events of inversion (Viamontes et al., 1979). An ultrastructural analysis revealed abundant microtubules aligned along the thin stalks of flask cells (Viamontes and Kirk, 1977), and colchicine or low- temperature treatments, which inhibited inversion, abolished elongated cells, otherwise present in the bend (Viamontes et al., 1979). Inversion was also blocked by inhibitors of actin filaments (Viamontes et al., 1979). Since the cellular location of actin filaments was not determined, the cytological basis for the cytochalasin effect was not fully explained.

In this study, we analyze the spatial distributions of actin filaments, myosin, microtubules and other structures in an inverting embryo by fluorescence and electron microscopy. Combining such studies with the use of actomyosin inhibitors and microsurgery, we provide evidence that localized actomyosin contraction, which is generated so as to make the posterior hemisphere compact, appears to be essential for inversion. This constitutes a previously unrecognized aspect of the inversion process.

Volvox carteri f. nagariensis, strain HK-10, was obtained from the University of Texas Culture Collection of Algae. The axenic culture was maintained in a synchronous and asexual growth with the standard Volvox medium (SVM; Kirk and Kirk, 1983) at 32°C under the light cycle previously reported (Tam and Kirk, 1991). Volvox spheroids 2-4 days after inoculation were harvested by filtration using 90 μm mesh (Cell Strainer: Nippon Bection Dickison Co. Ltd., Tokyo, Japan). They were then ruptured in SVM by five passages through a 25-gauge needle. The released embryos were retrieved on a 40 μm mesh after having been passed through a 90 μm mesh to remove spheroid debris. Embryos were then rinsed in excess SVM. All experiments were done at room temperature.

Stock solutions of cytochalasin B, cytochalasin D and cytochalasin E (Sigma Chemical Co., St Louis, MO, USA) were prepared following the method described previously (Viamontes et al., 1979). They were dissolved at 10 mg/ml in DMSO (Infinity Pure, Wako Chemicals, Osaka, Japan) and diluted 100-fold in SVM before use. A myosin II inhibitor, 2, 3-butandione monoxime (BDM; Sigma) was diluted to 40 mM with SVM from a freshly prepared 0.5 M solution (Cramer and Mitchison, 1995). Sucrose at 50 mM was present in all drug solutions to avoid non-specific osmotic damages (Viamontes et al., 1979). The concentration of sucrose was lowered to 10 mM in the case of BDM to balance the osmolarity.

To treat embryos with the drugs, isolated embryos were placed on glass coverslips precoated with 0.01% polyethyleneimine (Kirk et al., 1993). Prior to drug application, embryos were incubated in 50 mM sucrose in SVM for about 60 minutes. An appropriate amount of the drug solution was placed onto the embryos and the solution was changed quickly two more times to facilitate solution exchange. Removal of drugs was done similarly with 50 mM sucrose in SVM.

Embryos were observed by differential interference contrast (DIC) on an Olympus inverted microscope (IMT-2, Olympus, Tokyo, Japan) and the images were recorded with a cooled CCD video camera (Image Point; Photometrics, Ltd., Tucson, AZ, USA) onto S-VHS video tapes. Images were enhanced with an image-processor, Σ-III (Nippon Avionics Co., Tokyo, Japan) and captured on a Macintosh 840AV (Apple Computer, Cupertino, CA, USA) with NIH Image 1.59. Edited digital images were printed with a color printer (Pictrography 3000: Fujifilm, Tokyo, Japan). For higher resolution, images were recorded with a high resolution cooled CCD camera (PXL-1400; Photometrics) mounted on a Nikon Optiphoto II (Nikon, Tokyo, Japan) or on a Olympus AX-70.

To acquire images with single-cell resolution, embryos were processed as follows. Embryos on a coverslip were fixed in 2% formaldehyde, 2.5% glutaraldehyde, 2.5% tannic acid in 0.02 M phosphate buffer at pH 8.0 for 12-16 hours and postfixed in 1% OsO₄ in the same buffer for 3 hours (Viamontes and Kirk, 1977). Specimens were then dehydrated in a graded series of ethanol solutions and infused with Technovit 8100 (Kulzer, Wehrhein, Germany). To capture embryos in a Technovit block mountable on a microtome, a fresh Technovit solution contained in a gelatin capsule was placed upside-down on the coverslip and allowed to polymerize in vacuum for 1 hour. After polymerization, 2.5 μm serial sections were cut with a glass knife on an ultramicrotome (Reichert Ultracut E: Leica, Vienna, Austria). Sections were mounted on coverslips with Entellan (Merck, Darmstadt, Germany) and observed with DIC optics.

Embryos attached to PEI-coated coverslips were fixed for 15 minutes in 3.7% formaldehyde, 1 mM DTT, 0.1% Triton X-100, 2 mM MgCl₂, 5 mM EGTA, 150 mM KCl, 10 mM sodium glycerophosphate and 10 mM Hepes at pH 7.1, a solution modified from a fixative originally designed for preservation of labile actin filaments of leukocytes (Redmond and Zigmond, 1993). The coverslips were then washed (with gentle agitation) in a Coplin jar containing 1% NP-40, 1 mM DTT, 1% BSA, 2 mM MgCl₂, 5 mM EGTA, 150 mM KCl and 10 mM sodium glycerophosphate for 1 hour to extract chlorophyll and minimize autofluorescence. After washing three times in TPBS (0.1% Tween 20) supplemented with 0.1% BSA, embryos were stained for 4 hours with 4 μM fluorescein-phalloidin (Molecular Probes, Inc., Eugene, USA) in the same buffer and again washed three times in TPBS. Samples were mounted in 50% glycerol, 1 mg/ml phenylenediamine (an anti-fading agent) in PBS.

For immunostaining, embryos were similarly fixed and incubated for 12 hours either with 10-20 μg/ml anti-tubulin monoclonal antibody (Amersham, Buckingamshire, England), or 10 μg/ml antibodies against Physarum myosin II (Ogihara et al., 1983). Samples were then washed and stained with an FITC-conjugated secondary antibody (Amersham) at 1:200 dilution for 4 hours. To visualize the nucleus, DAPI (4 μg/ml) was included in the secondary antibody solution. The secondary antibodies were preabsorbed following the protocol described by Harlow and Lane (1988). The anti-myosin II from Physarum plasmodium used in this study was affinity purified and is monospecific for the heavy chain. Staining without the primary antibody as a control gives no specific patterns (data not shown).

Embryos were imaged on a cooled CCD camera. Serial images in z-axis were precisely incorporated by the use of a computer-assisted focusing devise (Micro Z-mover; Photometrics). Digitally deconvoluted images were obtained using IPLab and Confocal extension (Signal Analytics Corporation, Vienna, VA, USA), following the method by Agard et al. (1989).

Specimens were prepared for electron microscopy as described by Viamontes et al. (1979). Embryos were fixed at 20°C for 15 seconds in 2 ml of freshly prepared 2.5% glutaraldehyde in SVM (pH 7.2), and then 0.5 ml of 2% OsO₄ in H₂O was added. Fixation was continued in the resulting mixture for 10-15 minutes. After washing with SVM, the specimens were transferred to 2.5% tannic acid in H₂O for 15 minutes, washed again in SVM and postfixed in 2% OsO₄ in H₂O for 15 minutes. After another SVM wash, specimens were transferred to 2.5% uranyl acetate in H₂O for 15 minutes, washed in H₂O, dehydrated, embedded and sectioned.

Prior to dissection, the ‘gonidial vesicle’ surrounding an embryo was removed following the previously described method (Ransick, 1991) with modifications as follows. Gonidia were isolated by the method described above for embryos and then were incubated in SVM containing 0.1 mg/ml trypsin (Sigma) for 15-30 seconds and washed with an excess volume of SVM containing 1% BSA. The softened vesicles were then removed from the embryos with micromanipulation using a fine tungsten needle. Vesicle-free embryos were transferred to SVM contained in a plastic ELISA plate and allowed to develop for about 8 hours. After the final cleavage, vesicle-free embryos were transferred to a ∼200 μl drop of SVM placed on a silicon-coated coverslip. Most of the anterior half was then removed from the embryo by pressing a fine glass needle along the equatorial plane.

To measure the width of the posterior hemisphere of inverting embryos with high geometrical accuracy, the digitized images were rotated so that the anteroposterior axis became the y-axis of the analysis coordinate. The posterior hemisphere diameter was then obtained as the maximal surface-to-surface dimension parallel to the x-axis. Quantification and calculation were done using NIH Image 1.59.

Cells changed shape conspicuously during inversion, as seen in the DIC images (Fig. 1). The sequential changes of the cell shape are consistent with the previous report by Viamontes et al. (1979), who named the characteristic shapes, pear, spindle, flask and column, according to the progression of inversion. The pear-shaped preinversion cells (Fig. 1A,F) were nearly round with slightly conical anterior ends, the apices of which were directed toward the interior of the embryo. The nucleus was also conical, with its apex aligned with that of the cell. The phialopore opened, as cells assumed an elongated spindle- shape with a short protuberance pointing outward (Fig. 1B,F). In the characteristic bend region near the phialopore, there were flask-shaped cells with the long stalks (Fig. 1C,F). Cells remained elongated throughout inversion (Fig. 1C,D,F), but then assumed a less-elongate column shape again after inversion had been completed (Fig. 1E,F). Concomitant with these cell shape changes, the distribution of microtubules changed dynamically. In preinversion embryos, the microtubule signal outlined each cell in fairly uniform manner (Fig. 1A,F) but, once cells became elongated, they exhibited an much more intense microtubule signal in regions distal to the nuclei, at the chloroplast outer ends of the cells (Fig. 1B,C,F). Although individual microtubules were not resolved highly, the microtubules appeared to be aligned parallel to the long axis of each cell. Nuclei were located close to the inner surface in uninverted portions of the embryo. Flagella became visible on the newly exposed outer surface, in close proximity to the nucleus, in embryos approaching the end of inversion (Fig. 1D-F).

We next examined the cellular localization of actin filaments by use of fluorescein-phalloidin. When non-fluorescent- phalloidin was added in varying proportions to the fluorescein- phalloidin staining solution, it caused a substantial, concentration-dependent decrease in the fluorescent signal (Fig. 2I-K) leading us to conclude that the fluorescent signal obtained with fluorescein-phalloidin is phalloidin-specific and, hence, presumably actin-specific.

In preinversion embryos, most actin filaments were restricted to the perinuclear cytoplasm; none were detected in the chloroplast region (Fig. 2A,E). As inversion began, a dramatic change in the organization of actin filaments occurred; the intense perinuclear signals were replaced by a more-or-less continuous, subnuclear actin filament array that traversed the entire embryo from one pole to the other (Fig. 2B,F). Indeed, this filament array appeared to be continuous through the cytoplasmic bridges known to be linking adjacent cells at this stage (Fig. 2G). This shift in actin-filament location appears to be much more extreme in the anterior (phialopore- proximal) hemisphere than in the posterior hemisphere (Fig. 2B). As inversion proceeded (Fig. 2C), the actin filaments were displaced somewhat toward the chloroplast ends of the cells of the uninverted cells, where they formed an intensely staining band that was continuous throughout the posterior portion of the embryo. In marked contrast, actin filaments virtually disappeared from cells that had entered or passed through the bend region of the embryo (Fig. 2C, arrowheads). After inversion was complete, diffuse actin signals returned to somatic cells (Fig. 2D) and more intense signals were observed in the gonidial cells.

Because the images obtained with whole embryos made it appear that the actin array in the uninverted portions of an embryo was continuous from cell to cell (Fig. 2C,F), we next examined optically sectioned embryos from a different angle, in an attempt to obtain higher resolution and to determine if this actually was the case. Fig. 2G shows a cross section of the posterior hemisphere of an inverting embryo in which some cells appear to have protrusions linking them to adjacent cells (which presumably is a reflection of the cytoplasmic bridge system known to link cells at this level at this time; Green et al., 1981). Actin staining is present in such protrusions.

Antibodies against myosin II of Physarum plasmodia (Ogihara et al., 1983) cross-reacted with Volvox myosin. Before inversion, myosin II was localized in the perinuclear region in a pattern similar to that of the actin filaments (Fig. 3A). In addition, however, a discrete spot of intense myosin staining was observed at the inner end of each cell, in a region known to be occupied by the basal bodies and contractile vacuoles (Fig. 3D).

During inversion, myosin II, like actin, became localized to a subnuclear cytoplasmic region, formed arrays that appeared to be continuous from cell to cell in the posterior portion of the embryo, but became diffusely distributed in cells within and beyond the bend region (Fig. 3B). Concomitant with these changes, the myosin signal in the vicinity of the basal body disappeared (Fig. 3E). Postinversion embryos had very faint myosin signals in the cytoplasm; but a bright spot near the basal bodies reappeared (Fig. 3C,F).

In short, the dynamic changes that occur in myosin II localization during inversion are parallel to those observed for actin filaments and are as tightly coupled to the morphogenetic events of inversion.

To determine what role the observed changes in co-localized actin and myosin might be playing in the inversion process, we examined the effects of four actomyosin inhibitors: cytochalasin B, cytochalasin D, cytochalasin E and BDM (Cramer and Mitchison, 1995). Even when added 2 hours prior to the beginning of inversion, all three kinds of cytochalasin inhibited inversion partially (Fig. 4F-J), despite the fact that such treatments had led to almost total disappearance of actin filaments in the subnuclear region within the first an hour (Fig. 4P,Q). BDM had the same effect (Fig. 4K-O). In all four cases, the phialopore opened (Fig. 4F,K) and, subsequently, the lips of the phialopore began to turn outward as in a control embryo (Fig. 4G,L). The propagation of the bend, however, stopped in ∼20 minutes, leaving the anterior hemisphere inverted incompletely (Fig. 4H-J,M-O). The posterior hemisphere was never inverted.

In embryos that underwent such an inversion arrest, we noticed that the posterior hemisphere appeared wider than in control embryos. To quantify this, we measured the diameter of the posterior hemisphere of the treated and untreated embryos, with the results shown in Fig. 4R. In the first 15 minutes after the onset of phialopore opening, the diameter of the posterior hemisphere decreased slowly in both treated and control embryos. But then, whereas the posterior hemisphere of the control embryos decreased in diameter by nearly half during the next 20 minutes, no such rapid decrease in diameter occurred in the treated embryos; they simply continued the initial slow contraction. The time at which abrupt contraction occurred in the control embryos corresponded to the time at which the posterior hemisphere began to pass through the phialopore.

We discovered that BDM acts very rapidly on inverting Volvox embryos, and took advantage of this to study the time course of onset and reversal of inhibition more closely (Fig. 5). When BDM was supplied to an embryo in which the phialopore had already opened, the posterior hemisphere including the equator expanded significantly within 3 minutes (Fig. 5A). Our measurements (not shown) indicated that the diameter of the embryo at the equator increased by 10-15%. As expected, removal of BDM resulted in a quick contraction (Fig. 5B). In this quick expansion and contraction, we could not detect any specific cells or regions that expanded and contracted more than the rest of the posterior hemisphere; the posterior hemisphere appeared to contract almost uniformly. The embryos contracted about 10 μm in diameter, which corresponds to a 15 μm change in the circumference. Since an optical section of the posterior hemisphere (Figs 4, 5) contained approximately 30 cells, individual cells are expected to contract much less than 1 μm. The action of BDM could be fully reversed, even in embryos exposed to the drug for more than 3 hours before phialopore opening, if – but only if – the BDM was washed out before inversion had been underway for 20 minutes. When the drug was washed out after this time, the treated embryos never completed inversion. This critical time point corresponded roughly with the time at which the abrupt contraction of the posterior hemisphere began in control embryos, as shown in Fig. 4.

In contrast, the effects of cytochalasins were irreversible in all cases (data not shown). Removal of cytochalasins either before or after the critical time point was unsuccessful. This was probably because of the slow permeation of these chemicals into and out of the embryos.

Results presented in the previous section indicated that a critical point for drug action is passed about 20 minutes after the beginning of inversion, just about the time when the posterior hemisphere of control embryos abruptly becomes smaller and begins to pass through the phialopore opening (Fig. 4). We speculated that beyond this time BDM and cytochalasins deprive the embryos of the ability to undergo the changes of individual cell shapes that are required to complete the inversion process. However, this speculation was not supported by further examination of cytochalasin B-arrested embryos (Fig. 6).

In control embryos, as previously reported (Viamontes et al., 1977), a small bundle of extremely elongated ‘flask’ cells (usually about seven cells in a cross section) is present in each bend region (Fig. 6A). This continues to be true as the bend is propagated toward the posterior pole (Fig. 6B,C). In cytochalasin B-treated embryos, the bend is not propagated beyond the equator to the posterior pole (Fig. 6D,E). However, a cluster of elongated cells is not only present in the bend region at the time that inversion is arrested, but formation of such clusters of elongated cells is propagated beyond the bend within the next 10 minutes (Fig. 6D), eventually reaching the vicinity of the posterior pole about 30 minutes after inversion had been arrested (Fig. 6E). As in a normal bend region, the thin ends of such cells were bundled together. In the uninverted portion of the inhibited embryo (as in the inverted portion), microtubule staining revealed presence of flagella pointing toward inside the embryo sphere (Fig. 6F).

The presence in the posterior hemisphere of arrested embryos of clustered cells that are bundled at their thin, chloroplast ends implied that cytoplasmic bridges might be capable of becoming relocated to the chloroplast ends of the cells even in the presence of actomyosin inhibitors. To determine whether this might be the case, the location of cytoplasmic bridges was examined ultrastructurally (Fig. 7). In BDM-treated embryos (Fig. 7A), cells in the posterior hemisphere clearly had cytoplasmic bridges at their chloroplast ends (Fig. 7B,C). This embryo had been fixed at a time when inversion would have been completed in the absence of BDM. For comparison, we chose a control embryo with curvature of the posterior hemisphere approximately the same as that of the BDM- arrested embryo, meaning that cells in the posterior hemisphere had not yet experienced inversion at the time of fixation. As previously reported, cytoplasmic bridges were found at the nuclear level (Fig. 7D), indicating that bridges in these cells had not yet migrated (Green et al., 1981). In contrast, the cytoplasmic bridges in postinversion regions (Fig. 7E) were located at the chloroplast ends of the cells, like those near the uninverted posterior pole of the BDM-inhibited embryo. Cytoplasmic bridges migrated similarly in cytochalasin B- treated embryos (data not shown).

If actomyosin inhibitors do not block the propagation of cell shape changes or the migration of cytoplasmic bridges into the posterior hemisphere, why do such inhibitors block inversion of the posterior hemisphere? We postulated that this might be because, in the absence of actomyosin-mediated contraction, the diameter of the embryo in the region of the equator is larger than (and hence unable to pass through) the opening that is present in the already-inverted anterior hemisphere.

To test this hypothesis, we performed two kinds of experiment. First, we added BDM to the embryo just after the equatorial region had passed through the bend region. Our hypothesis predicted that, at this stage, BDM would have no effect on inversion at all, because the remaining uninverted portion of the posterior hemisphere would be smaller than the equator. This turned out to be the case (data not shown). Second, we used microsurgery to isolate subhemispherical fragments of BDM-inhibited embryos (Fig. 8A) in which the cut edge formed an opening as wide as the fragment itself (Fig. 8). Again, our hypothesis predicted that BDM should have no inhibitory effect on inversion under these conditions. Fig. 8 shows that this was the case.

The immunofluorescence studies performed here indicate that microtubules are present abundantly during the entire course of Volvox inversion and that changes in their organization correlate with changes in cell shape and embryo morphology. As the clearest example, the long stalks of the flask-shaped cells that are characteristic of the bend region are shown to be richly supplied with aligned microtubules. These observations support the conclusion, drawn from ultrastructural observations made by others, that microtubule-mediated changes in cell shape play an important role in the inversion process (Green et al., 1981; Viamontes et al., 1979; Viamontes and Kirk, 1977).

The present study also extends the analysis of Volvox inversion in a new direction by examining for the first time the distribution of myosin and actin filaments in inverting embryos and by reexamining the effects of actomyosin inhibitors on the inversion process. These studies have uncovered an important aspect of the inversion mechanism that had not previously been detected.

As inversion begins, actin and myosin move in concert from the perinuclear region to the subnuclear region of each cell and then, as cells approach the bend region, their actomyosin filaments move progressively further from the nuclei, toward the chloroplast ends of the cells. This is reminiscent of the progressive change that has been shown to occur in the locations of the cytoplasmic bridges during the inversion process (Green et al., 1981, figure 10). Moreover, some actin filaments appear to be continuous from cell to cell (Fig. 2G). This leads to the obvious suggestion that actomyosin filaments may actually traverse the cytoplasmic bridges that link neighboring cells (Green et al., 1981). This possibility should be examined at the ultrastructural level but, unfortunately, we have not yet succeeded in preserving the actin filaments of Volvox embryos in specimens that are suitable for EM examination (unpublished observation). A similar lack of success has been reported by others (Green et al., 1981; D. Kirk, personal communication).

It is particularly noteworthy that the actin-filament network disappears, and myosin becomes diffusely distributed, as soon cells enter the bend region of the inverting embryo (Figs 2C, 3B). This clearly indicates that, whatever role(s) the actomyosin complex plays in inversion, that role must be completed before the cells participate in formation of a bend region. This assumption was reinforced by our observation on the effects of actomyosin inhibitors, to be discussed next.

All four of the actomyosin inhibitors used here (cytochalasin B, cytochalasin D and cytochalasin E, plus BDM) had a similar effect when applied to Volvox embryos continuously from 2 hours before the beginning of inversion through the time at which control embryos had completed inversion. They permitted inversion to be initiated relatively normally and to proceed to the halfway of the anterior hemisphere, but they blocked inversion of the posterior hemisphere irreversibly (Fig. 4).

It was previously reported that cytochalasin D acted on the Volvox embryo quickly and reversibly to block both the initial generation of, and the propagation of, the bend regions (Viamontes et al., 1979). In contrast, under our conditions of observation, all three cytochalasins acted much more slowly, and irreversibly, and did not prevent either the initial generation of bend regions or their progression to the halfway point of the treated embryos (Fig. 4G,P,Q). These slow effects of cytochalasins were taken into consideration in the design of our experiments. We observed with immunofluorescence that no effects on the actin filament network could be observed in the first 5 minutes of cytochalasin treatment (data not shown) and it took 30-60 minutes to fully depolymerize actin filaments in the subnuclear region (Fig. 4P). Thus, cytochalasin treatment for 2 hours before inversion, which we performed here, is thought to be long enough to induce the drug effects in the embryos, and hence the incomplete progression of the bend was never due to insufficient permeation of the drugs. It is unclear whether the different results obtained in this study and the earlier one (Viamontes et al., 1979), are due to subtle difference in strains, culture conditions, purity of the drugs, method of drug administration or some other unknown factor. In any case, the present study and the previous one indicate that the sequence of cell-shape transformations that accompany inversion are largely independent of actomyosin activity and, although they are undoubtedly necessary, they are not sufficient to allow inversion to be completed.

A clue to the mechanism by which actomyosin inhibitors prevent inversion from going to completion came with the observation that, in the presence of such inhibitors, the posterior hemisphere of the embryo appeared to be wider than it was in control embryos at the same stage. Measurements confirmed this impression: by 15-20 minutes after the beginning of inversion, the posterior hemisphere of control embryos began an accelerated contraction that reduced its diameter appreciably, but treated embryos underwent no such rapid contraction. The rapidity and irreversibility with which the myosin II inhibitor, BDM (Cramer and Mitchison, 1995) acted on the Volvox embryos permitted us to study this phase of the inversion process more closely and show that an actomyosin inhibitor exerts its inhibitory influence on the inversion process during this period. If BDM was present at the 20-minute time point, inversion of the posterior hemisphere was blocked. In contrast, even in embryos that had been exposed to BDM for 3 hours, if the drug was washed out before the 20-minute time point, inversion was completed normally.

This led us to the hypothesis that actomyosin inhibitors block the last half of inversion by blocking a contraction of the equatorial region of the embryo that is necessary to make it small enough to pass through the opening that is formed by already inverted anterior hemisphere. Two additional pieces of evidence supported this hypothesis. The first was the observation that, if BDM was added after the equatorial region had passed through this opening, no inhibition of the rest of the inversion process was observed. The second was the observation that if a subposterior hemispherical fragment was isolated by microdissection, BDM failed to prevent it from inverting (Fig. 8). As shown in Figs 6 and 7, cell shape changes and migration of cytoplasmic bridges occurred normally even in the arrested embryos. Therefore, removal of the anterior hemisphere, which would otherwise have imposed a physical constraint on the swollen posterior hemisphere, permitted inversion of the arrested posterior hemisphere. In Fig. 9, we schematize how these mechanics cooperate in inversion with special reference to actomyosin and microtubules. Our hypothesis does not require any cellular differentiation between cells in the anterior and the posterior hemispheres to explain why anti-actomyosin drugs arrest inversion at approximately the equator.

The role of actomyosin in Volvox inversion appears similar to its role in reducing the dimensions of a multicellular epithelial sheet during animal morphogenesis (Owaribe and Masuda, 1982). It is noteworthy, however, that the function of actomyosin contraction may be different in these two cases. Whereas the effect of actomyosin inhibitors in the animal system is attributed to inhibition of the contraction that occurs at one end of each cell and causes them to assume ‘flask’ or ‘bottle’ shape, formation of such cellular shapes in Volvox is not prevented by actomyosin inhibitors and appears to be driven by microtubules (Fig.6; Viamontes et al., 1979). Nevertheless the role proposed here for actomyosin contraction (namely, to relieve the folding cellular sheet from the physical constraints) may be applicable in many other examples of epithelial folding, especially when these processes include the passage of a part of the multicellular sheet through a small opening, such as in gastrulation, evagination of imaginal disk (Fristrom, 1988) etc. Mathematical models of epithelial folding (for example, Odell et al., 1981) might also need to pay more careful attention to resolving such spatial constraints.

We speculate that the disappearance of actin filaments from cells in the already inverted anterior hemisphere (Figs 2C, 3B) may play an essential positive role in inversion. If the cells were to remain in a contracted state as they entered and passed through the bend regions, this would almost certainly counteract the effect of contractions in the posterior hemisphere, resulting in a decrease in the diameter of the opening through which the posterior hemisphere must pass. Thus loss of actomyosin filaments and relaxation of cells that have traversed the bend region may well be as indispensable for the completion of inversion as is the actomyosin contraction in the posterior hemisphere.

In conclusion, we have shown that actomyosin exerts a contractile force that causes the posterior hemisphere to become compact and thus facilitates the propagation of the bending region, which is generated largely by microtubules. It is necessary for these activities of actomyosin to be coordinated in space and time with the activities of microtubules during inversion. Genetic analysis of these dynamic phenomena may be feasible. A variety of Volvox mutants deficient in various phases of inversion has been isolated (Sessoms and Huskey, 1973; Kirk et al., 1982) and techniques suitable for analyzing the molecular basis of such defects have recently been developed (Miller et al., 1993; Schiedlmeier et al., 1994; Hallmann et al., 1997). We are now attempting to isolate transposon-tagged mutants with phenotypes that resemble the drug-arrested embryos of the molecular-genetic program involved in coordinating cytoskeletal activities in space and time to bring about this interesting morphogenetic transformation.

We would like to thank Dr David L. Kirk for useful suggestions and critical reading of this manuscript. I. N. is supported by Research Fellowships of the Japan Society for the Promotion of Science for Japanese Young Scientists.