Microtubules and mitotic cycle phase modulate spatiotemporal distributions of F-actin and myosin II in Drosophila syncytial blastoderm embryos.

Many processes critical for development suggest interactions (of unknown cause) between microtubules, actin and myosin. These include (1) directed lamellipodial protrusion in crawling cells (e.g. Waterman-Storer and Salmon, 1999; Waterman-Storer et al., 1999), (2) sperm-entry-activated, colcemid-sensitive, rotation of the cortex relative to the deep cytoplasm in amphibian eggs, establishing the dorsoventral body axis of the embryo (Larabell et al., 1996; Vincent and Gerhart, 1987), and (3) during cytokinesis, cortical flow away from spindle poles (Cao and Wang, 1990b; Dan, 1954; Fishkind et al., 1996; Wang et al., 1994) and cytoplasmic myosin movement towards the spindle mid-zone (Yumura and Uyeda, 1997a), movements that are oriented by mitotic apparatus position.

In early Drosophila embryogenesis, three large-scale morphogenetic events also suggest interactions between the actomyosin cytoskeleton and microtubule arrays: (1) the stepwise migration of nuclei through the syncytial egg’s interior towards the cortex in phase with cycles of microtubule and F-actin reorganization (Baker et al., 1993; von Dassow and Schubiger, 1994; Zalokar and Erk, 1976), (2) cycles of ‘bud’ and ‘pseudocleavage furrow’ formation after migrating nuclei reach the cortex of the still syncytial embryo (Foe and Alberts, 1983), which we focus on in this study, and (3) cellularization, the actin, myosin and microtubule-requiring process which partitions the syncytial cytoplasm into mononucleate cells (reviewed in Foe et al., 1993; Schejter and Wieschaus, 1993). Several proteins essential for contractile ring formation during cytokinesis – the septin protein Peanut (Neufeld and Rubin, 1994), Diaphanous (Castrillon and Wasserman, 1994) and Anillin (C. M. F., unpublished data) – are also implicated in pseudocleavage furrow formation in Drosophila (Afshar et al., 1999; Field and Alberts, 1995; Miller and Kiehart, 1995; Rothwell et al., 1998). This suggests that studies of bud/pseudocleavage furrow formation will provide insights relevant to cytokinesis.

We observe temporal and spatial changes in the organization of filamentous actin (F-actin hereafter), conventional non-muscle myosin (myosin II hereafter), centrosomes and microtubules in the Drosophila syncytial blastoderm correlated with mitotic cycle phase, and report drug experiments to probe for interactions between the cytoskeletal elements. Our evidence implicates oriented arrays of microtubules in causing rapid spatial reorganizations of myosin II and F-actin in synchrony with the mitotic cycle. These spatial reorganizations occur against a background of embryo-wide, large-amplitude, temporal oscillations in the fraction of actin polymerized and concentrated at the cortex, and in the fraction of myosin II concentrated at the embryo’s cortex (as detected by two different polyclonal antibodies). These embryo-wide oscillations, interesting in their own right, complicate understanding how microtubules may affect F-actin and myosin II spatial patterning. Fortunately, in early Drosophila development, cell cycle-related changes in the cortex that depend on microtubules can be distinguished from those that do not. Before their 10^th division cycle, nuclei with associated microtubule arrays undergo replication and movement cycles in the embryo’s interior without contacting the cortex (Foe and Alberts, 1983; Zalokar and Erk, 1976). Before cycle 10, we can therefore observe microtubule-independent kinematics of cortical F-actin and myosin II. Migrating nuclei reach the cortex during early interphase of cycle 10 (Foe and Alberts, 1983); thereafter their associated microtubule arrays do affect cortical F-actin (Karr and Alberts, 1986; Kellogg et al., 1988; Warn et al., 1984)and myosin II reorganizations . Rapid cycles of cytoskeletal reorganization near the cortex correlate with the formation during interphase of ‘buds’ – cytoplasmic protrusions each containing one nucleus – and bud collapse during anaphase (Foe and Alberts, 1983). A time-lapse video of bud kinematics is available at http://raven.zoology.washington.edu/downTheTubes/. Bud formation cycles require the presence at the cortex of centrosomes and the microtubules they nucleate, but not necessarily of nuclei (Raff and Glover, 1988; Yasuda et al., 1991). The ‘trenches’ between buds, called pseudocleavage furrows (or metaphase furrows) are shallow during cycle 10, but invaginate deep enough in later cycles to isolate adjacent nuclei. Functionally, this isolation of syncytial nuclei prevents chromosome capture by microtubules emanating from centrosomes of adjacent nuclei (Postner et al., 1992; Sullivan et al., 1990). We focus on the cytoskeletal dynamics cyclically producing cytoplasmic buds and pseudocleavage furrows between buds, two features of the same underlying phenomenon.

We prepared miniature egg collection plates by filling 40 mm Falcon Petri plates with 3% Bacto agar (Difco Laboratories), 20% grape juice concentrate (Welch) and 0.15% methyl paraben (Sigma) dissolved in boiling water. We placed these plates, flavored with yeast and acetic acid, as lids on empty inverted bottles containing 3-to 10-day-old adult flies (Sevelin strain). We collected eggs for 20 minutes during the daily maximum oviposition period (5-8 hours after lights-on in a 14 hours/10 hours light/dark cycle), after discarding a 45 minute pre-collection. We used Fig. 1 of Foe et al. (1993) to grossly stage eggs. We dechorionated eggs by 1.5 minute immersion in 50% bleach (Clorox), then rinsed them repeatedly in 0.8% NaCl, 0.1% Triton X-100. Since crowding dechorionated embryos results in anoxia (Foe and Alberts, 1985), embryos, aged on the collection plates, were either not dechorionated much in advance of usage or were kept well dispersed in the rinse solution. We distinguished cycle 8, 9 and 10 dechorionated embryos by the presence of pole buds, pole cells and anterior buds (Foe and Alberts, 1983) with a dissection microscope using transmitted light. We staged embryos mounted for microinjection on slides under Halocarbon oil by the position, number and size of nuclei, using a compound microscope (Foe and Alberts, 1983, 1985).

Bleach-dechorionated embryos in 0.8% NaCl/0.1% Triton X-100 were transferred by pipette to parafilm (American Can Company), rinsed twice with distilled water, dried and transferred individually by forceps tip, onto fine lines of glue dried onto glass slides. The glue is double-stick Scotch tape adhesive dissolved in heptane. Rows of embryos glued to slides, after appropriate dehydration at ambient humidity, were covered with high viscosity (series 700) Halocarbon oil (Halocarbon Products Corp). Injection protocol was as described in Foe and Alberts (1983). We released embryos from the glue with several washes of n-heptane, then transferred them by pipette from heptane to fixative.

Intracellular drug concentrations cited assume a 50-fold dilution of the injected solution (Foe and Alberts, 1983) and diffusion throughout the egg. The injection buffer is phosphate-buffered saline (PBS) diluted 1/1 with distilled water. A stock cycloheximide (Sigma) solution of 10 mg/ml in water was diluted 1/10 into injection buffer. A stock cytochalasin B (Sigma) solution of 40 mg/ml in DMSO was diluted 1/100 into injection buffer. A stock colcemid solution of 4.6 mg/ml in distilled water, protected from light, was diluted 1/25 into injection buffer. Stocks were stored at −20°C.

We used bright-field or DIC optics on Zeiss WL standard microscopes to image injected embryos under Halocarbon without coverslips with Zeiss dry Plan 16/0.35, multi-immersion, Plan NeoFluor 63/1.2 W Korr and 25/0.8 Plan Neofluor objectives. To make time-lapse movies, an Apple G3 computer using NIH-Image software recorded output of a Hammamatsu C2400 CCD camera and control box. Apple’s QuickTime software (Sorensen) compressed the frames.

We use different fixation protocols to preserve different filament systems. Protocol 1, optimized to preserve myosin II and F-actin and to not deform buds, does not preserve microtubules except those stabilized by taxol or bundled into a mitotic spindle. We suspend bleach-devitellinized embryos in 4 ml of PBS and gradually add 4 ml of fix solution made of 3 parts PBS + 1 part 40% methanol-free formaldehyde (Electron Microscope Sciences Inc), while vortexing for 45 seconds, then add 4 ml heptane and agitate vigorously for 23 minutes. We stop this reaction by replacing the fix with several changes of PBS. The rational for exposing the embryos to fix prior to immersion in the permeabilization solvent heptane is to permit fixative, entering through the micropyle, to fix protruding buds at the embryo anterior. Anterior buds collapse when embryos are exposed to heptane before being fixed. The low concentration of formaldehyde (5%) and absence of methanol (a component of formalin) in the fix are important to retain myosin II stainability. We manually devitellinize fixed embryos individually with tungsten needles rather than devitellinizing en masse via methanol immersion because methanol immersion ruins phalloidin stainability of F-actin.

Protocol 2, used for Figs 4, 5, 8, 9 and early interphase panels of Fig. 3, preserves microtubules, F-actin, gamma tubulin staining (centrosomes) and nuclei. Embryos are treated as above, except that protocol 2 fix is 1 part PBS + 1 part formalin (37% formaldehyde with 13% methanol). We add heptane simultaneously with fixative and fix for 13 minutes. Gross actin staining is identical in protocols 1 and 2, but detailed F-actin structure is better preserved by protocol 1. Rapid influx of heptane through the micropyle partially collapses anterior buds before fixation stabilizes them. We will publish the 3-D rendering techniques that we devised for 3-D reconstruction elsewhere.

We added taxol (Molecular Probes) from a stock solution of 2.5 mM in DMSO to embryos in 5 ml PBS to a final concentration of 10 μM. Embryos were overlaid with an equal volume of heptane (to permeabilize vitelline membrane) and shaken for 90 seconds. Fixation by protocol 1 (for Figs 6, 7) or protocol 2 (for Fig. 8) was commenced 90 seconds after taxol application. Permeabilization is not instantaneous so we estimate that the live embryo exposure to taxol is between 1 and 1.5 minutes.

Extract of 0-to 4-hour-old Drosophila embryos was passed over filamentous actin or albumin control columns. Column preparation, embryo extracts and chromatography conditions were as described previously (Miller et al., 1989, 1991). Proteins, eluted with 1 M KCl, 1 mM ATP, 1 mM MgCl₂, were further separated by SDS-PAGE using preparative gradient gels. Individual bands were excised and used as antigens for antibody generation in mice. Myosin II migrated at approximately 200 kDa.

Mouse antibody #4 from Miller et al. (1989) was used to screen a λZap expression library constructed from Drosophila ovary poly(A)⁺ RNA (Hay et al., 1988). The library was screened using minor modifications of procedures described by Huynh et al. (1985). Clones from the screen were sequenced using fluorescently labeled chain terminator nucleotides with an Applied Biosystems 373 automated sequencer. The screen produced two unrelated cDNA clones, one of which is approximately 1.3 kb in length and encodes amino acids 959-1496 of the non-muscle myosin heavy chain tail (Ketchum et al., 1990; Kiehart and Feghali, 1986).

A glutathione S-transferase (GST) fusion with the myosin II tail fragment was generated using a pGEX expression vector (Smith and Johnson, 1988). We used this fusion protein to generate polyclonal antibodies in rabbits. Antibodies were affinity purified by passage over columns of immobilized fusion proteins. The serum was first depleted of anti-GST antibodies by repeated passage over a column of GST protein and then passed over a column containing the myosin fragment, as detailed in Field and Alberts (1995).

Micrographs in Figs 1 and 2 were made on a BioRad Radiance Plus laser scanning confocal microscope (LSCM), with three photodetectors. All others were made on a two-photodetector BioRad MRC-600 LSCM, and were all made using a Nikon plan apochromatic 60/1.4 oil immersion objective. Our nuclear staining protocol enables the BioRad MRC-600 to distinguish chromatin staining from other red and green channel probes, thus differentiating three probes using just two photodetectors.

We block fixed embryos for 30 minutes with normal goat serum (50 μl/ml in PBT; PBT is PBS with 0.1% Triton X-100), react with 1/750 polyclonal rabbit anti-myosin and 1/1500 anti-gamma tubulin (to visualize centrosomes) for 4-6 hours at room temperature, wash with PBT for 3× 5 minutes plus 3× 30 minutes, then stain the rabbit antibodies with 1/750 Alexa 568-conjugated goat anti-rabbit IgG (Molecular Probes Inc) or, for three-photodetector imaging, with 1/750 Cy5 goat anti-rabbit (Jackson Immuno Research Labs). We incubate embryos already stained with antibodies in DNAase-free RNAase at 1 mg/ml for 1.5 hours at 37°C, then rinse 3× in PBT and add between 1.0 and 3.0 μl/ml of a 15 mg/ml stock solution of propidium iodide in DMSO to the phalloidin reaction solution (see below) and incubate at room temperature for 1-3 hours. The goal is to obtain a sufficiently low level of chromatin staining by propidium iodide to be undetectable by the LSCM when excited by 1% of the laser intensity, yet staining sufficient to be seen at 10% excitation after the other probes have been bleached away (see details in section on confocal microscopy below). For imaging with three photodetectors (Fig. 2), nuclei are stained with mouse monoclonal anti-histone (Chemicon) at 1/1500 and visualized with 1/1000 rhodamine Red-X goat anti-mouse (Molecular Probes).

We stain F-actin with BODIPY-Fl phallicidin (Molecular Probes), 300 units of which dissolved in 1.5 ml methanol and stored at −20°C constitute our stock solution. Five 10 μl aliquots of this stock are vacuum dried at room temperature, redissolved in PBT, pooled to give 1 ml of solution and embryos already stained with antibodies are suspended in this solution. If nuclei are to be visualized, these embryos were previously treated with RNAase and propidium iodine is added to the phallicidin solution. Embryos were incubated in the phallicidin/propidium iodide solution for 1 to 3 hours at room temperature and then rinsed for 1 hour in PBT. Variation in the staining duration does not affect staining. Prolonged rinsing washes out propidium iodide and diminishes the phallicidin staining, especially that of delicate cortical actin fibers (because phallotoxins can exchange off of F-actin). Therefore phallotoxin or propidiumstained embryos were cleared and mounted immediately after rinsing or, if stored, were kept in small volumes of PBS.

To visualize deep internal cytoskeletal structures, we clear and mount embryos in Murray’s mounting medium (2 parts benzyl benzoate: 1 part benzyl alcohol). This matches the refractive index of Drosophila yolk. Unless all salt is removed before transferring embryos into Murray’s medium, propidium iodide-stained nuclei will not fluoresce. We transfer embryos from PBT to 1/40 PBS in three dilution steps each of 5 minute duration. Because methanol and ethanol damage the phallicidin-stainability of F-actin, we dehydrate embryos in isopropanol, with a minimum exposure. While gently vortexing, we pipette 0.3 ml of 100% isopropanol into an Eppendorf tube containing the embryos in 0.15 ml of 1/40 PBS. After allowing the embryos to settle for 1.5 minutes, we pipette them onto a polylysine-coated slide, and then transfer the slide through a dehydration series of 85%, 95% and 2× 100% isopropanol (allowing 20-30 seconds per step), and then through three 5 minute changes of Murray’s medium. The preparation is covered with a number 1.5 cover glass and sealed with 5 minute Epoxy glue (Devcon) to give a semipermanent preparation. Epoxy adheres to glass slides sufficiently to make a seal only after removing excess Murray’s medium (with xylene-soaked swabs).

We coat slides with polylysine by dipping them in a solution of 100 ml water, 60 mg poly-L-lysine hydrobromide (Sigma), 200 μl Photoflo (Kodak). Dipped slides are air-dried, rinsed in distilled water and redried. We make slides fresh every two weeks because poly-lysine slides slowly lose their adhesivity. Shims, made from Scotch Brand 425 aluminum tape, placed beside embryos support cover glasses to prevent embryo deformation.

How do we differentiate nuclear staining from the staining in the red (myosin) and green (actin) channels using only the MRC 600’s two photodetectors? The F-actin and myosin II images are obtained by exciting with 1% of the laser intensity at the 488 nm and 568 nm laser lines, respectively, and collecting, into the two photodetectors, light emitted at 522 nm and 585 nm, respectively. Then we bleach away the F-actin and myosin II probes by exciting the sample with all laser lines at 10% laser strength. The dim, but more photostable, propidium iodide probe survives this. We imaged it at 10% laser intensity by exciting at 488 nm and collecting the emission light at 585 nm. Images were collected using a dual channel (K1/K2) filter block but excited with one laser line at a time to minimize bleed-through. Nuclei in Figs 1, 3, 4, 5 and 6 were all imaged in this way. The micrographs in Figs 1 and 2 were obtained with a BioRad Radiance Plus confocal microscope, which allows viewing triple probes in the usual way. Where indicated in the captions, images represent stacks of successive optical sections.

Fig. 1 comprises laser scanning confocal microscope (hereafter LSCM) sectional views of anterior ends of Drosophila embryos showing anaphase and interphase of cycle 9 (before centrosomes, microtubules and nuclei reach the cortex) and cycle 10 (the first round of bud formation and breakdown). In cycle 9, myosin II staining concentrates in a cortical rim during interphase but leaves the cortex during anaphase (myosin, row 1 versus row 2). Likewise, F-actin concentrates during interphase 9 in a cortical rim, which attenuates greatly during anaphase 9. Throughout interphase 9, with no nuclei/asters near the cortex, cortical myosin II and F-actin co-localize. We observed similar waxing and waning of cortical F-actin and myosin II, co-localized, synchronized with globally synchronous mitotic cycles, in cycle 8 (not shown). Migrating nuclei with microtubule arrays reach the cortex 1 minute after interphase 10 begins (Foe and Alberts, 1983). As telophase 9 ends and interphase 10 begins, F-actin and myosin II re-accumulate co-localized to high levels in a spatially uniform cortical rim (not shown in sectional view, but see Fig. 2, row 2). 2 minutes after nuclei reach the cortex, cortical F-actin and myosin II are no longer co-localized but occur in the complementary patterns shown in Fig. 1, row 3. Myosin II occurs at high levels between buds, but vacates the cortex where buds now protrude (row 3), while F-actin attains high levels precisely on the domes of the buds that myosin II vacated. During anaphase 10, cortical levels of F-actin and myosin II are globally low. Cortical F-actin re-accumulation begins first near centrosomes at anaphase/telophase (Fig. 1, row 4). Regardless of cortical fluctuations, high levels of myosin II staining occur throughout the embryo interior (Fig. 1, rows 1-4); myosin-dark holes are where yolk particles exclude the myosin probe. When myosin dissociates from the cortex, it transiently boosts the concentration of myosin immediately beneath the cortex, but does not significantly boost the concentration of internal myosin globally, presumably because it is dispersing into an ocean of cytoplasmic myosin filling this large cell. Throughout the interior cytoplasm, F-actin occurs diffusely, and additionally in particles, but at lower levels than cortically. von Dassow and Schubiger (1994) have described in detail the cyclic reorganizations of internal F-actin during nuclear migration.

Fig. 2 comprises surface views of Drosophila embryos at about 50% egg length, as overlays of multiple LSCM single sections, showing in finer temporal resolution the rapidly changing organizations of F-actin, myosin II, centrosomes and chromatin just prior to and during the first cycle of bud formation. Relative intensities of the myosin II and F-actin probes in Figs 1 and 2 are quantitatively accurate representations of the intensity changes as the mitotic cycles progress. When migrating nuclei with associated microtubule arrays first reach the periphery early in interphase 10, myosin II staining disappears from a small region immediately above the nuclei (row 3) and, during the next 3 minutes of interphase and prophase, the holes vacated by myosin II staining enlarge into oblong holes (rows 4 and 5). During metaphase and anaphase, the cortical myosin staining dims (rows 6-7), then, during telophase (row 8), myosin staining returns brightly to the cortex except near centrosomes where it remains dim. We interpret holes in the cortical myosin pattern, appearing in the cortex precisely when/where microtubule arrays first contact the cortex, as implying an interaction between myosin II and microtubules. Lajoie-Mazenc and colleagues showed anti-gamma tubulin concentrates on the spindle during metaphase and on the midbody remnant from anaphase through telophase (Lajoie-Mazenc et al., 1994). Therefore, the staining of the spindles in Fig. 2 in the middle columns of metaphase 10 and anaphase 10 may be partly due to anti-gamma tubulin labeling, but only partly, because we find that embryos stained with anti-myosin II in the absence of anti-gamma tubulin also show the spindle and midbody staining pattern displayed in rows 6-7 of Fig. 2. Single section images (not shown) reveal that the cytoplasmic myosin staining at anaphase 10 and early telophase (shown as multisection projections in Fig. 2), in addition to the midbody and cortical staining, includes a low level but discernable accumulation of myosin staining on the forming aster. We obtain the same staining patterns with our polyclonal antibody prepared against a part of the Drosophila myosin II tail and with a polyclonal anti-myosin prepared against whole myosin heavy chain (the generous gift of Daniel Kiehart; data not shown).

As migrating nuclei, centrosomes and microtubule arrays approach the cortex, they affect cortical F-actin differently than myosin II. F-actin accumulates between centrosomes and the plasma membrane, forming ‘caps’ in the regions that myosin II staining vacates (Fig. 2, row 3, compare left red panel – F-actin, with center blue panel – myosin II; also row 3 of Fig. 1). During the next 3 minutes these small caps of actin enlarge in outer diameter and hollow out centrally above each centrosome (rows 3-5 of Fig. 2; also see rows 2 and 5 of Fig. 3 and row 2 of Fig. 4). During metaphase, the cortical F-actin staining attenuates globally with the ghosts of the prophase F-actin rings persisting longest (row 6 of Fig. 2; also row 4 of Fig. 4). Actin staining remains dim globally during anaphase with re-accumulation beginning in the cortex overlying centrosomes before it does so elsewhere (row 7 of Fig. 2; row 4 of Fig. 1; also rows 2 and 4 of Fig. 5). F-actin accumulation into caps accelerates during telophase 10 and early interphase 11 (row 8 of Fig. 2). (For earlier descriptions of F-actin kinematics see Karr and Alberts, 1986; Kellogg et al., 1988; Warn et al., 1984.) In Fig. 2, juxtaposing eight sequential phases of the mitotic cycle clearly exhibits the global waxing and waning of cortical F-actin and myosin II. These changes are spatially homogeneous in cycle 9. During cycle 10, cortical concentrations of myosin II between buds resembles the same-phase, cycle 9, concentrations seen uniformly across the cortex. Both F-actin and myosin II staining attenuate during metaphase, reach a minimum during anaphase, rapidly rebuild during telophase, and maintain a high level during interphase and prophase (see also Fig. 1). These wild-type kinematics suggest the mitotic oscillator modulates F-actin and myosin II polymerization dynamics, and/or the attachment of these molecules to the cell cortex, enabling potentially higher levels of actomyosin contractility from telophase through interphase, and suppressing it during metaphase and anaphase.

Figs 3, 4 and 5 comprise a sequence of stereopair 3-D reconstructions that we made of small volumes of the syncytial blastoderm in fixed and staged embryos. These show how the arrangement of microtubules changes through the mitotic cycle, and how the microtubule kinematics correlate with the spatial rearrangements of F-actin and myosin II in cycle 10 described above. The relative intensities of the myosin II and F-actin in these 3-D reconstructions are not accurate because LSCM gain was set to reveal topological detail in each reconstruction rather than to show intensity changes from one phase of the cycle to the next. Figs 3-5 show microtubules running in all directions. No single focal plane of a 2-D micrograph can portray accurately their 3-D geometry. The conclusions that we glean from Figs 3-5 are found throughout the text, especially in the Discussion section. QuickTime movies and interactive virtual reality objects, from which Figs 3-5 show selected frames, are available at and http://raven.zoology.washington.edu/downTheTubes/. These 100-200 megabyte files merit downloading because the interactive full-rotation vision of the blastoderm cytoskeleton’s geometric complexity that they communicate far surpasses the glimpses that Figs 3-5 provide.

The 3-D reconstructions in Figs 3-5 and the surface views in Fig. 2 are at about 50% egg length. Only at the anterior and posterior poles of the embryo does sufficient space exist between the plasma membrane and the egg shell for buds to protrude into (as in Fig. 1). The flatness of the egg surface in Figs 2-5 is caused by the plasma membrane pressing against the nearly non-deformable vitelline membrane, whose curvature is so slight as to be nearly imperceptible across the 42 μm width of the region imaged. Except at the poles, invaginations of the plasma membrane are possible but protrusions are not. Our 3-D reconstructions of cycle 12 and 13 embryos (not shown here, but see http://raven.zoology.washington.edu/downTheTubes/ for a 3-D reconstruction of metaphase 12) confirm the common understanding that deep invaginations of the cortex, pseudocleavage furrows, form between adjacent nuclei during later syncytial cycles as nuclei become closely packed. Pseudocleavage furrow invaginations along the sides of the embryo are mechanically equivalent to buds protruding out above the egg surface as occurs at the anterior pole of the embryo starting in cycle 10 (Fig. 1). In this paper, we confined our investigation of cytoskeletal kinematics to cycles 10 and earlier precisely because the cortex (except at the poles) remains nearly planar during those cycles. Therefore deformation of the cortex cannot play any important role in causing the dramatic spatial reorganization of myosin II and F-actin shown in Figs 2-5. Rather the spatial rearrangements of myosin II, F-actin and other cytoskeletal proteins into the kind of patterns shown in Figs 2-5 almost certainly cause invagination of pseudocleavage furrows in later cycles (see Discussion).

The F-actin and myosin II reorganizations of Fig. 2, rows 3-5, and Figs 3 and 4, take only 3 minutes, reminiscent of the time it takes a mitotic apparatus to induce a cleavage furrow (reviewed in Rappaport, 1996). Another interesting feature of the cytoskeletal kinematics described above is that they appear always to act to attenuate the potential actomyosin-based contractile strength of the cortex. During interphase, when global amounts of cortical F-actin and myosin II are greatest, the highest concentrations of these two key parts of the actomyosin contractile machinery are kept mostly separate. During metaphase, the F-actin and myosin II that remain at the cortex disperse toward spatially uniform co-localization but, at this phase of the mitotic cycle, cortical concentrations of F-actin and myosin II plummet, attenuating actomyosin contractile potential by different means. In telophase, when myosin II reconcentrates at the cortex, it vacates the regions overlying the centrosomes where F-actin is strongly re-accumulating. Myosin II and F-actin always colocalize at the cortex and throughout the embryo’s cytoplasm. Due to their global disappearance during mitosis and their spatial reorganization into near complementary patterns at other phases, however, the level of myosin II and F-actin colocalization remains always low relative to what it could be. When buds form (telophase through prophase), the cortex between buds (which in later cycles will invaginate as pseudocleavage furrows) are the zones where myosin II and F-actin do co-localize, though F-actin concentrations elsewhere are higher; this point is especially apparent in sectional views (Fig. 1, row 3).

Meiotic sisters of the female pronucleus, the polar body nuclei, persist until cycle 10 near the dorsal cortex of Drosophila embryos. These lack centrosomes and arrest in permanent metaphase-like configurations without nuclear envelopes, apparently after completing S-phase of their first mitotic cycle (see Foe et al., 1993). The condensed chromosomes from one, two or all three polar bodies form large starbursts with chromosomes joined near their kinetochores (Fig. 6C,E). In wild-type embryos, sparse arrays of short microtubules form around these polar bodies, organized presumably by chromosomes, as no centrosomes are present. The polarity of microtubules in these arrays has not been established. However, chromatids segregate normally during anaphase on centrosome-less spindles (Bonaccorsi et al., 1998; Matthies et al., 1996; Megraw et al., 1999), so it is likely that microtubules nucleated from chromosomes are oriented normally, i.e. with plus ends at kinetochores. If so, microtubule arrays nucleated from polar body chromosomes would form ‘asters’ with polarity opposite that of a centrosome-nucleated asters. In the reversed aster, the minus ends of microtubules would form the outer perimeter of the aster while the plus ends cluster at the aster’s center, near kinetochores, as cartooned in Fig. 11B.

By observing the effect this peculiar aster has on nearby F-actin and myosin, we proposed to answer two questions. (1) Can microtubule asters with no centrosome nucleating them influence spatial reorganization of nearby F-actin and myosin? The answer is, yes. When polar bodies lie close to the cortex, the immediately adjacent cortex is often depleted of both F-actin and myosin. Fig. 6D shows a surface view of myosin II depletion above the polar body of an embryo fixed in late interphase of cycle 8. Myosin depletion above sparse polar body microtubule arrays tends to be less pronounced than myosin-depletion over the denser microtubule arrays nucleated from centrosomes. Taxol treatment, by blocking microtubule disassembly, inhibits microtubule dynamic instability, promotes microtubule growth and induces non-centrosomal microtubule nucleation (Horwitz et al., 1986). Fig. 6A shows, in sectional view, F-actin depletion from the cortex above the polar body in an interphase 8 embryo briefly treated with taxol (see Materials and Methods). Though taxol rapidly increases the number of microtubules associated with polar body nuclei (Fig. 6B), cortical F-actin depletion above polar bodies in embryos fixed without prior taxol treatment (not shown) and after a 1-1.5 minute exposure to taxol are similar. The depletion of cortical F-actin above polar body microtubule arrays contrasts with F-actin accumulation into caps or rings near centrosome-nucleated microtubule arrays of zygotic nuclei.

(2) Do microtubule asters clear F-actin from their centers, concentrating it at their perimeter (as do normal cycle 10 asters shown in Figs 3-5), regardless of the polarity of the microtubules in the aster? To answer this question, we increased the number of microtubules in the polar body asters by brief taxol treatment (Fig. 6B). Notwithstanding the very high density of microtubules in the taxol-stabilized polar body aster, this aster concentrates F-actin at its center (Fig. 6A) rather than re-locating cortical actin towards the perimeter (as in Fig. 2 rows 3-5). Indeed, even without taxol, low concentrations of F-actin are discernable at the internal core of the normal sparse polar body asters by the end of interphase (not shown). We thus tentatively conclude that it is the polarity of the microtubules in an aster that determines where F-actin concentrations re-locate to. To explain the effect microtubule arrays have upon F-actin reorganization in all of Figs 1-6, it appears that F-actin, or a signal to polymerize it, moves toward the plus ends of microtubules in oriented arrays.

We find that treating pre-cycle10 embryos with taxol for 1-1.5 minutes causes a dense disorganized mat of short (2-10 μm) microtubules to form everywhere within about 2 μm of the cell cortex regardless of mitotic phase (shown in sectional view in Fig. 6B and in surface view in Fig. 8A), and causes diminution of centrosome-nucleated microtubules deep inside embryos (presumably by polymerizing at the cortex, where taxol first impinges tubulin dimers that would otherwise make interior microtubules). Earlier studies (Karr and Alberts, 1986) construed this to be the normal microtubule organization in preblastoderm embryos, but our improved fixation techniques show that this mat of cortical microtubules is an artifact of exposing the permeabilized embryo to taxol. Taxol treatment before cycle 10 lets us assay how the rapid genesis of myriad randomly oriented microtubules affects F-actin. Fig. 7 shows single-section high-magnification LSCM micrographs of phalloidin-stained embryos taken in the plane of the embryo surface, with untreated embryos (left) and embryos after a 1-1.5 minute exposure to taxol (right). Taxol administered during interphase 9 induces redistribution of cortical F-actin into long fibers (Fig. 7A,B) over the entire embryo surface (except above polar bodies where cortical F-actin is depleted). Taxol treatment during anaphase 9, (when cortical F-actin levels drop; Fig. 7A-C), induces dramatically fewer and thinner cortical actin fibers than does treatment during interphase (Fig. 7B versus D).

Taxol immersion during cycle 10 (with nuclei and associated microtubule arrays at the cortex) causes redistribution from fewer long, to more numerous but shorter, microtubules in centrosome-nucleated cortical microtubule arrays. It also induces randomly oriented short microtubules in the cortex between buds. Taxol immersion during cycle 10 interphase induces actin fibers over the entire surface but most densely at the actin caps that normally form above centrosomes (Fig. 7E,F). During prophase, taxol induces cortical actin fibers from the F-actin rings that encircle buds and in the cortex between buds, but very few in actin-depleted bud crowns (Fig. 7G,H). Taxol treatment during anaphase 10 resembles anaphase 9 in inducing few if any cortical actin fibers (anaphase 10 not shown). Taxol treatment at telophase 10 induces many F-actin fibers to grow from the forming actin caps, while few grow in the actin-poor zones between caps (Fig. 7J). In summary, in response to taxol stabilization of microtubules, F-actin fibers form wherever the cortical F-actin concentration is high; in cortical regions normally depleted of F-actin because of microtubule or mitotic phase effects, taxol does not promote actin fiber formation. The effect de novo polymerization of microtubules has upon cortical F-actin organization, and the rapidity of that effect, implies an interaction between these two principal filament systems.

We detect no obvious co-alignment between the actin fibers and the short microtubules that taxol induces in the cortex of pre-cycle 10 embryos (Fig. 8A-C). Once nuclei and microtubule arrays approach the cortex, taxol amplifies existing microtubule arrays, presumably by reducing microtubule depolymerization. In the presence of these dense stable arrays, the usual interphase 10 actin disks and rings often have ‘tufts’ extending from their outer edges (Fig. 8E).

Some of these tufts overlie microtubules (see Fig. 8D-F), as would be expected if F-actin were being pulled out along microtubule tracks. Elsewhere, however, actin fibers exist separate from underlying microtubules. Further study on F-actin/microtubule spatial relationships should be done in live embryos where the dynamic natures of actin fibers and microtubules do not complicate interpretation. Just permeabilizing membranes with heptane – essential for our taxol treatment – has little effect on F-actin organization, but it abolishes most of the cortical myosin II staining. This precluded determining what effect taxol-induced ectopic microtubules have on myosin distributions with our experiment.

Early mitotic cycles proceed so rapidly in Drosophila that cytoskeletal geometry is always changing rapidly. No phase lasts long enough for us to glimpse what equilibrium the cytoskeleton might approach during a long-lasting phase. In syncytial Drosophila embryos, protein synthesis inhibitors cause interphase arrest, presumably by inhibiting cyclin synthesis. In early cycle 13, we microinjected cycloheximide (intracellular concentration about 20 μg/ml) to inhibit protein synthesis, thereby causing permanent arrest in interphase 13. Time-lapse videorecordings show that, beginning about 1 hour after cycloheximide injection, nuclei begin leaving the egg periphery. Fixing and staining injected embryos just as nuclei disconnect from the cortex reveals that F-actin that would normally coat the cell’s cortex densely during interphase (Fig. 9A, top panel), has left the cortex (Fig. 9B, top panel) and has aggregated deep inside the embryo around the distal ends of interphase microtubules (Fig. 9B). In regions where patches of nuclei/asters fall away from the cortex, cortical F-actin reappears precisely in the locations that microtubule arrays have vacated (cf. upper and lower panels of Fig. 9B). Injecting colcemid alone into syncytial Drosophila embryos disrupts microtubules and arrests the mitotic cycle in metaphase with chromatin permanently condensed (Zalokar and Erk, 1976). But, simultaneous microinjection of cycloheximide (to 20 μg/ml) plus colcemid (to 100 μM) early in interphase, causes interphase arrest (the chromatin remains decondensed and the nuclei spherical) and microtubule depolymerization (Fig. 9C), i.e., while cycloheximide causes interphase arrest, colcemid eliminates microtubules. During this permanent interphase 13 with microtubules absent F-actin remains concentrated at the cortex even 3 hours after the injection (Fig. 9C top panel), though nuclei have become displaced from the cortex (Fig. 9C bottom panel).

Because loss of cortical F-actin brought about by injection of cycloheximide is associated with detachment of nuclei from the cortex, we tested whether loss of cortical F-actin by other means has the same effect. We microinjected cytochalasin B into cycle 8-9 Drosophila embryos to an intracellular concentration of 5μg/ml. By inhibiting actin polymerization (Mabuchi, 1986), cytochalasin leads eventually to F-actin depolymerization. We monitored subsequent development by time-lapse movies, then fixed the embryos in late cycle 13 and stained to reveal F-actin. Fixed preparations show that, by cycle 13, very little filamentous actin remains and nothing resembling normal actin caps or rings exist (not shown). Our time-lapse movies show that, in cytochalasin-injected embryos, nuclei reach the cortex in a timely fashion and continue to undergo cyclic rounds of division (though internuclear spacing during anaphase and telophase is abnormal; see also Zalokar and Erk, 1976). However, between cycles 10 and 13 many peripheral nuclei move away from the cortex and into the interior of the embryo (not shown), providing independent evidence that F-actin plays a role in holding microtubule arrays and nuclei at the cortex.

By cycle 12, the interphase distribution of cortical myosin II is restricted to the narrow base of the pseudocleavage furrows that separate closely packed buds. (see Fig. 6 in Field and Alberts, 1995). Because the heptane treatment that we use to remove oil and detach injected embryos glued to slides degrades the myosin-stainability of the cortex, we were unable to observe directly the effect mitotic cycle arrest or microtubule disruption has on cortical myosin II. However, time-lapse movies reveal that, in embryos stalled in a permanent interphase by cycloheximide, but in which colcemid has disrupted microtubules, disorganized, fast-moving, large-amplitude, convoluted, contractile waves develop and churn the entire cortex for hours (Fig. 10). Assuming these interphase contractions are actomyosin based, this rampant cortical contractility implies that, in the absence of microtubules, abundant myosin II must return to the cortex and co-localize with the F-actin that we detect there (Fig. 9C, upper panel). Nuclei dissociate from the writhing cortex and are displaced to a few microns beneath it (Fig. 9C, lower panel). This result differs strikingly from what is observed in embryos arrested in metaphase by injection with just colcemid; our time-lapse recordings demonstrate that the cortex of cycle 12 and 13 embryos arrested in metaphase by an internal concentration of 100 μM colcemid remains mechanically quiescent for at least the two hours during which we normally film them (videos not shown).

We conclude from these drug injection experiments that: (1) the mechanism that removes F-actin from the cortex and concentrates it far beneath the cortex during interphase arrest must be microtubule-dependent (it fails when colcemid disrupts microtubules), (2) two completely different ways of eliminating F-actin from the cortex (by cytochalasin or by prolonging interphase and letting the microtubule-dependent mechanism strip off F-actin) cause nuclei to fall away from the cortex, but (3) the presence of F-actin alone at the cortex (without microtubules) is not sufficient to hold nuclei at the cortex, and (4) mitotic cycle phase affects cortex contractility, the cortex being actively contractile in interphase-arrested embryos, but inert in metaphase-arrested embryos.

We propose four high-level parsimonious hypotheses about mitotic-cycle-modulated interactions between the cell cortex (the approximately 3 μm deep zone immediately underlying the plasma membrane), F-actin, myosin II, centrosomes and microtubules, which, working together, can explain our, and others’, experimental results. These hypotheses imply no opinion about detailed molecular causes; interactions could be direct or caused by multipart complexes of auxiliary proteins. We state each hypothesis, connect it to our data, then cite literature providing supporting biochemical, genetic, or other data. We discuss cytokinesis studies because we propose an analogy between the cyclic cytoskeletal reorganizations that we observe in syncytial fly embryos, in which karyokinesis proceeds rapidly without cell division, and the way the mitotic apparatus locates the contractile ring during cytokinesis in typical mononucleate animal cells. We end by discussing how the mechanistic interactions that we hypothesize could generate the forces to protrude buds, to invaginate pseudocleavage furrows and to position cleavage furrows for standard cytokinesis.

mitotic phase modulates the amount of myosin II associated with the cortex and the amount of actin polymerized into filaments at the cortex; independent of the proximity of centrosomes and/or microtubules, both cortical myosin II and cortical F-actin increase from telophase through interphase then fall through metaphase and anaphase

H1 simply summarizes the cyclic fluctuations in the cortical rim of high F-actin and myosin II concentration that we see in Drosophila embryos (Figs 1-5), which occur in both cycle 9 (microtubules absent) and cycle 10 (microtubules present). The surface views in Fig. 2 imply a dramatic temporal fluctuation in the total amount of cortical F-actin and myosin II staining, while Figs 1 and 3-5 show that F-actin and myosin II cyclically concentrate within ∼3 μm of the cell surface.

H1 says the cortex is a diffusion trap that accumulates, and possibly stabilizes, myosin II filaments during one part of the mitotic cycle, from telophase to prophase. We lack a molecular understanding of how myosin II associates with the plasma membrane and how this localization is cyclically modulated. One component may be encoded by the gene Drosophila lethal(2)giant larvae, part of a multiprotein complex that associates tightly with plasma membrane and binds myosin II directly (Kalmes et al., 1996; Strand et al., 1994). The association between myosin and the l(2)gl protein is inhibited by phosphorylation (Kalmes et al., 1996), suggesting possible regulation via the p34^cdc2 kinase cascade. Alternately myosin’s association with the actin-rich cortex may be controlled via phosphorylation of myosin itself. Studies on Dictyostelium discoideum show that phosphorylation of myosin heavy chain simultaneously determines myosin’s assembly into bipolar filaments (Liang et al., 1999; Pasternak et al., 1989), the distribution of myosin between cortex and cytoplasm (Egelhoff et al., 1993; Yumura and Uyeda, 1997b), and in vivo actomyosin cortical contractility (Egelhoff et al., 1993).

H1 likewise says the cortex binds filamentous actin. We assume actin filaments are continually depolymerizing, so ongoing polymerization is needed to maintain equilibrium levels of F-actin; an increase in the rate of polymerization at one mitotic phase would yield higher levels of F-actin at that phase (assuming a constant depolymerization rate). We propose that, with no microtubules nearby (e.g. in cycle 9), the actin polymerization rate near the cortex is higher between telophase and prophase than during metaphase and anaphase. (Obviously an inverse variation of filament stability would accomplish the same result.) This same temporal modulation of actin polymerization rates must occur also in cycle 10 (and 11, etc), when centrosomes and microtubules immediately underlie the cortex, because it is obvious from inspection of rows 1-3 of Fig. 2 that there is no way to produce the caps of F-actin observed in early interphase 10 simply by redistributing the cortical F-actin pre-existing in anaphase 9. Whereas proteins that attach actin to plasma membranes are known (see for example Tsukita and Yonemura, 1997), cell cycle effects on these interactions are not yet known. From cycle 10 onward, cortical F-actin distribution is additionally influenced by nearby microtubule arrays, leading us to propose H2 and H3.

We hypothesize the elevation of actin polymerization rates near centrosomes to account for the genesis of the F-actin caps. In the Drosophila syncytium, cortical F-actin concentrations appear near centrosomes as the miniscule metaphase asters (Fig. 4) become extensive anaphase asters (Fig. 5). This occurs both when centrosomes are near the cortex (as shown here), or deep inside the embryo, as reported by von Dassow and Schubiger (1994).

How could centrosomes promote nearby actin polymerization during anaphase and telophase? Waterman-Storer and colleagues (1999) show that, in 3T3 fibroblasts, polymerization of microtubules activates Rac 1 GTPase, which in turn boosts actin polymerization. Neither the mere presence of stable microtubules, nor their depolymerization, has this effect. In the fly syncytial blastoderm at metaphase, few microtubules and only tiny asters at the spindle poles exist (Fig. 4). But during anaphase (Fig. 5), centrosomes nucleate large asters, which enlarge in telophase. If the mechanism that Waterman-Storer et al. (1999) discovered in 3T3 cells also operates in the fly embryos, then the large anaphase/telophase burst of centrosome-nucleated microtubule polymerization should cause a proportional increment in F-actin polymerization. Our 3-D reconstructions confirm that this burst of actin polymerization concentrates in the asters, near centrosomes (Fig. 5 rows 2 and 4). When centrosomes closely underlie the cortex (see Figs 1, 3-5), the Rac 1 mechanism of concentrating actin polymerization near centrosomes (H2) together with F-actin’s increasing affinity for the cortex upon exit from mitosis (H1), could account for the formation of cortical F-actin caps shown in Figs 2-5. H2 could additionally account for the response to taxol that we observe. Brief treatment of permeabilized embryos with taxol causes a dramatic polymerization of haphazard microtubules just underlying the cortex and a concomitant change in cortical F-actin cortical organization (Fig. 7). Waterman-Storer et al. (1999) find that taxol treatment of 3T3 cells also promotes haphazard microtubule polymerization, which in turn causes a transient burst of actin polymerization.

The maternal-effect gene mat(2)syn^HK21 (Schupbach and Wieschaus, 1989), renamed centrosomin by Kaufman and coworkers, encodes a protein localized at centrosomes throughout syncytial blastoderm stages (Heuer et al., 1995; Li and Kaufman, 1996; Vaizel-Ohayon and Schejter, 1999). Homozygous cnn-defective mothers produce embryos with dramatic early defects in spindle organization, nuclear migration and pseudocleavage furrow formation. These embryos die without cellularizing (Li and Kaufman, 1996; Megraw et al., 1999; Vaizel-Ohayon and Schejter, 1999). Cnn protein, which is essential for normal centrosome assembly, also affects actin polymerization dynamics (Vaizel-Ohayon and Schejter, 1999). Although cortical F-actin in cnn mutant embryos is reportedly normal prior to nuclear arrival, when migrating nuclei reach the cortex, F-actin caps fail to form above them and pseudocleavage furrows fail to invaginate (Vaizel-Ohayon and Schejter, 1999); astral microtubule arrays form poorly or not at all, though functional spindles do form (Megraw et al., 1999; Vaizel-Ohayon and Schejter, 1999), apparently nucleated from chromosomes and assembled into spindles by motor proteins. These observations, like those of Waterman-Storer et al. (1999) implicate centrosome-nucleated microtubule dynamics in actin polymerization dynamics.

This hypothesis accounts for effects microtubules have on reorganizing cytoplasmic F-actin three-dimensionally in the embryo interior and reorganizing cortical F-actin in the plane of the cortex. When microtubules lie close to, and tangent to, the cortex, this hypothesis implies plus end transport of cortex-bound filaments in the plane of the cortex, until continued transport along a microtubule would tear an actin filament away from the cortex. Then either the transport mechanism drags both the actin filament and the cortex that it is attached to towards the embryo’s interior (as in pseudocleavage furrow invagination), or, if the membrane stays put, the kinesin motor stalls and the actin filament remains attached until depolymerization melts it.

Plus ends of elongating microtubules growing out from centrosomes, leaving a wake of activated Rac 1, leaving in turn a wake of actin polymerization (Waterman-Storer et al., 1999), could account for the formation of F-actin caps during anaphase through early interphase (Fig. 5, rows 2 and 4; Fig. 2, rows 3, 7 and 8). It will not explain, however, the subsequent hollowing out of F-actin caps into rings (Fig. 2, rows 3-5; Fig. 4, row 2) or account for the removal of F-actin from the cortex during a drug-prolonged interphase and its concentration near the plus ends of microtubules (Fig. 9B). The conversion of F-actin caps into hollow rings could result from nuclei pressing up into, and thereby excluding F-actin from the apices of buds. However, the equally common ‘figure-of-eight’ F-actin pattern observed overlying many prophase nuclei (with centrosomes at the center of each ring and a ‘cross bar’ of F-actin between the apical crown of the nuclear membrane and the plasma membrane – see right-most nuclei in row 5 of Fig. 2 and row 2 of Fig. 4) excludes this possible explanation. The gradual conversion of F-actin caps into hollow rings might arise simply because dense microtubule asters exclude F-actin. But the concentration of F-actin at the center of dense, assumed reversed-polarity microtubule asters (Fig. 6A), which polar body chromosomes nucleate and taxol amplifies, argues against that. However, movement of F-actin along centrosome-nucleated microtubules, towards microtubule plus ends, can simultaneously explain the prophase figure-of-eight F-actin patterns around separated individual centrosomes and F-actin rings encircling both centrosomes (eg. row 5 of Fig. 2). Cortical F-actin moving away from centrosomes is also consistent the accumulation of F-actin in the central core where the plus ends of microtubules of the reversed polarity aster converge (Fig. 6A), and the outcomes of our drug studies, discussed below. Fig. 11A,B cartoon the reorganizations of F-actin that hypothesis H3 would produce in centrosome-nucleated and reverse-polarity microtubule arrays, respectively.

The F-actin kinematics of Fig. 2, and everything just described in the previous paragraph, would result if something that promotes actin polymerization, rather than F-actin itself, associated with microtubules and continually moved along them toward microtubule plus ends. We favor the hypothesis that it is actin filaments that attach to and move along microtubules by some complex involving a kinesin because: (1) studies of cytokinesis in other model organisms have shown that, by some means, cortical flows of stabilized actin filaments do occur in vivo (Cao and Wang, 1990a,b); and (2) a beautiful in vitro study using cell-free lysates of Xenopus eggs showed that stabilized F-actin binds along the length of microtubules nucleated from demembranated sperm (Sider et al., 1999). Sider et al. (1999) find that F-actin associates with microtubules in the presence of cell lysate, but not in its absence, implying the involvement of intermediary molecules. Moreover, the distribution of F-actin along microtubules changes with time: 5-minute-old asters contain more F-actin in the aster center than do 15-minute-old asters, which have relatively more F-actin at their distal (plus) ends, suggesting microtubule-based plus-end-directed transport at work. The co-localization of cortical actin with taxol-stabilized microtubules at the outer ragged edge of actin rings (Fig. 8D-F) is likewise consistent with this interpretation.

Another reason our data favors attachment of actin filaments to microtubules and transport along them is that the proposed mechanical attachment of F-actin to both cortex and microtubules has the side effect that F-actin would act as a ‘glue’ holding microtubule asters to the cortex, consistent with our drug studies. H3 implies that dissociating F-actin from the cortex would permit microtubule asters (and associated nuclei) to fall away from the cortex, as we observe after cytochalasin injections eliminate F-actin or after cycloheximide injection produces an interphase of such long duration that all F-actin is stripped from the cortex. After microinjection of both cycloheximide (to cause interphase arrest) and colcemid, the F-actin remains concentrated at the cortex, but cortical contractions nevertheless displace nuclei from the cortex (Fig. 9C-bottom panel), presumably because, when colcemid eliminates microtubules, it breaks one link in the cortex–F-actin–microtubule chain.

H3 also explains the result that most F-actin is removed from the cortex during the cycloheximide-induced hour-long arrest in interphase 13, and ends up concentrated basally where the plus ends of microtubules terminate (Fig. 9B). Continuous operation of H3 would establish a conveyor belt of F-actin motoring towards the plus ends of long microtubules. But why wouldn’t cortical attachments prevent removal of F-actin from the cortex? Because, we assume, individual actin filaments persist no longer than a couple of minutes before depolymerizing (Theriot and Mitchison, 1992). Actin repolymerizing near the cortex, but not close enough to attach to that cortex, can be carried into the embryo’s interior along microtubules. Actin transported in filamentous form to the basal ends of long microtubules can, upon depolymerization, diffuse back towards the cortex. We assume it takes 60-80 minutes of a drug-prolonged interphase for the H3 transport mechanism to win the race against diffusion of actin monomers back to the cortex and for all F-actin to be concentrated in the embryo interior. Thus H3 would not be expected to clear F-actin from the cortex in the normal short-duration syncytial interphases. In the cycloheximide-prolonged interphase experiment, immediately F-actin is depleted from the cortex and so eliminates the ‘actin glue effect’, microtubule asters and their nuclei start falling away from the cortex. Immediately microtubule arrays leave the F-actin-depleted cortex, F-actin begins reappearing in the patches that microtubules have vacated (Fig. 9B), demonstrating the cortex’s normal interphase ability to accumulate F-actin (H1) wherever F-actin is not being convected away by microtubules. In the dual cycloheximide/colcemid injection experiment, depolymerization of microtubules defeats the transport process, so F-actin remains at the cortex permanently (Fig. 9C).

von Dassow and Schubiger (1994) studying nuclear migration in early Drosophila embryos (cycles 4-6), before nuclei and microtubules make contact with the cortex, observed cyclic F-actin reorganization patterns consistent with H2-H3, but in the embryo interior. Particles of cytoplasmic F-actin accumulate near centrosomes in anaphase and telophase. As interphase progresses to prophase, F-actin particles accumulate in an enlarging halo around centrosomes with the center of the halo increasingly depleted of F-actin. Prolonging interphase by cycloheximide injection (an experiment analogous to that illustrated here in Fig. 9B) massively concentrates F-actin granules where the distal plus ends of microtubules from adjacent asters converge (presumably in antiparallel orientation; see Fig. 12C in von Dassow and Schubiger, 1994) and depletes the area around centrosomes/ nuclei of F-actin. In cycloheximide-injected embryos, cytoplasmic F-actin re-localizes more rapidly than F-actin attached to the cortex (as the logic of the previous paragraph predicts), but both reorganizations are consistent with plus-end transport of F-actin along microtubules. Is there genetic evidence for a motor that (possibly indirectly) moves actin along astral microtubules towards their plus ends? We speculate that the essential kinesin-like protein PAV-KLP (homologue of mammalian MKLP-1), perhaps in association with Polo Kinase (Adams et al., 1998; Carmena et al., 1998; Nigg, 1998), may be such a motor.

Fig. 11 diagrams the effect centrosome-nucleated microtubule arrays and reversed-polarity polar body asters would have on myosin II, via H4. H4 accounts for why myosin II normally vacates regions of the cortex centered at centrosomes from which microtubules radiate (Figs 1-3) and likewise vacates regions where microtubule arrays nucleated from polar bodies contact the cortex (Fig. 6D). An alternative would be radiation, toward microtubule plus ends, of a signal that causes myosin II to release its attachment to the cortex, after which it just diffuses away. We prefer directed movement of myosin II along microtubules because that hypothesis, but not the alternate, can also explain myosin II relocalizations in Dictyostelium discoideum mutants (Yumura and Uyeda, 1997a; Zang and Spudich, 1998).

In Dictyostelium myosin mutants whose myosin heavy chains cannot be phosphorylated (and hence cannot be depolymerized), filaments of mutant myosin still localize normally to the contractile furrow during mitosis and to the tails of migrating interphase cells (Yumura and Uyeda, 1997b). Thus, in this lower eukaryote, myosin II thick filaments actually move rather than depolymerize at one site and re-polymerize elsewhere. It is also relevant that, in Dictyostelium cells null for the normal myosin II gene, three different mutant myosin II constructs move to the equator of dividing cells even though none of these three mutant constructs can hydrolyze ATP, or cause furrow constrictions (Yumura and Uyeda, 1997a). Likewise myosin II tails shorn of their actin-binding heads still aggregate to the cell equator in mitotic Dictyostelium cells null for normal myosin II, though the myosin tail polypeptide remains cytoplasmic and does not localize to the cleavage furrow cortex (Zang and Spudich, 1998). While neither research team tested whether myosin II localization requires microtubules, their data provide strong evidence that actomyosin contractility is not required to transport at least cytoplasmic myosin II to the cell equator, overturning a common assumption, and one consistent with H4. If myosin II translocation to the equator is occurring via microtubule transport, the finding that ‘triple–ala’ mutants (which mimic constitutive dephosphorylation) overaccumulate myosin at the contractile ring (Sabry et al., 1997) would then mean that, in Dictyostelium, it is probably only dephosphorylated myosin II heavy chain assembled into filaments that moves in this fashion.

If myosin II translocation occurs via microtubules, how might this occur? Recent work on unconventional (non-filament forming) myosins suggests a possible answer. Huang and coworkers have shown that a region in the tail of mammalian myosin VA, a two-headed myosin with actin-activated ATPase activity (Cheney et al., 1993), interacts with a roughly 100-amino-acid portion of the tail of conventional kinesin heavy chain (Huang et al., 1999). This interaction, identified using a yeast two-hybrid screen, is highly specific. Perhaps myosin II heavy chain likewise interacts via its tail directly or indirectly with a kinesin.

We find that the cortex of Drosophila syncytial blastoderm embryos arrested in interphase (by cycloheximide injection) with microtubules depolymerized (by colcemid) is actively contractile (Fig. 10), whereas the cortex in embryos in metaphase arrest (by colcemid injection) is noncontractile. This matches studies of cortical contraction in frog blastomeres in which relaxation occurs in response to activated p34^cdc2, while contraction occurs upon p34^cdc2 inactivation (Rankin and Kirschner, 1997). In cycle 9 and 10 syncytial Drosophila embryos, we observe a decrease in myosin II (and F-actin) at the cortex during each metaphase and anaphase, then a rapid rise during telophase to a high level, which persists through interphase and prophase. Cyclic changes that simply determine whether F-actin and myosin co-localize at the cortex are potentially sufficient to explain our observed cyclic changes in cortical contractility. We do not know whether in higher eukaryotes myosin II is independently attached to the cortex (like actin) with this attachment subject to mitotic control, or whether myosin accumulates at the cortex via its interactions with actin, which interactions are mitotically controlled, or whether both types of regulation occur. Jordan and Karess (1997) find that, in Drosophila, oocyctes lacking myosin regulatory light chain (RLC) accumulate cytoplasmic aggregates of myosin heavy chain (among other abnormalities) hinting that RLC may affect myosin’s ability to remain cortically bound. Perhaps RLC inhibitory sites ser-1 and ser-2, which p34^cdc2can phosphorylate (Pollard et al., 1990; Satterwhite et al., 1992; Yamakita et al., 1994), regulate cyclic changes in myosin’s association with the cortex.

If true and phylogenetically general, the four processes that we hypothesize would have multiple and far-reaching mechanical implications for cell morphogenesis, of which we now focus on two examples. The left side of Fig. 12 illustrates how the cytoskeletal interactions that we propose, acting in concert, might generate forces that would cause bud protrusion, and pseudocleavage furrow invagination in the syncytial Drosophila embryo. The righthand diagrams show how precisely the same interactions could position and time F-actin and myosin II assembly between bipolar microtubule arrays in preparation for cytokinesis.

Why do buds protrude from the embryo surface? At the Drosophila egg’s anterior end, where space between the plasma membrane and vitelline shell accommodates protrusions, we believe three forces, all acting in the same direction, cause buds to protrude. The region of the membrane that must be stretched to form each bud overlies centrosomes, which is where F-actin caps form (H2) beginning during telophase (H1), latching the microtubule array to the cortex (H3). This is also the region of the cortex that myosin II vacates (H4) immediately the telophase nucleus contacts the cortex. Removal of myosin II presumably weakens the very region of the cortex that will erupt in the fashion of an aneurysm, and this weakening, an instance of polar relaxation, sets the stage for a combination of three different forces to generate the aneurysm. First, microtubules, growing into the interior of the egg, collide with yolk particles and push upon them, either via kinesin-based transport of yolk particles, or simply by a ‘polymerization ram mechanism’ (see Peskin et al., 1993). This, iterated in interphase of cycles 10-14, progressively packs all the yolk particles into the center region of the embryo. Inward forces by microtubules on yolk produce opposite and equal outward reaction forces on microtubules, which they transmit to the centrosomes. This push of the entire nucleus-centrosome-microtubule aster assembly outward is focused precisely on that area of the cortex weakened by microtubule-dependent depletion of myosin II. The second force is a subtle consequence of H3; a kinesin-like motor can drag an actin filament attached to the cortex toward the microtubule’s plus end only by overcoming viscous drag forces resisting the sliding of the actin filament’s attachment to the cortex. The resulting myriad equal-and-opposite reaction forces required by Newton’s laws push the microtubules that contact cortical F-actin back toward the centrosomes. Before the membrane overlying the nucleus erupts outward, the astral microtubules radiating out from the centrosome, tangent to and just underlying the membrane, are like the spokes of an erect umbrella in which all the spokes are in compression pushing on the umbrella’s central hub (the centrosome). This arrangement is prone to buckling outward (or inward). Once any other action initiates a slight outward protrusion of a bud, this mechanism amplifies it; the larger the diameter of the hollow actin rings surrounding centrosomes become (see rows 4 and 5 of Fig. 2), the more effective are the compressive forces on the microtubules underlying the membrane at causing outward buckling of the bud. The third force may result from an actomyosin contraction of the bands between buds where the F-actin and myosin II coexist. Seen at low magnification, these bands constitute a roughly hexagonal network of interconnected actomyosin strands, presumably capable of contracting during interphase, surrounding the egg’s entire surface. Contraction of this net of strands would cause all of them to indent the cortex. But the constraint that the plasma membrane always encloses a constant volume means any indentations made in the membrane cause equal volume outward bulges.

Note that the depolymerization of all but spindle microtubules in metaphase eliminates the scaffolding through which the first two forces are transmitted. The third (contractile net) force is maximal in interphase but vanishes at anaphase due to the observed dissipation of the cortical F-actin and myosin II at anaphase (H1). Therefore, buds start collapsing at metaphase and become flat by anaphase, as observed (live sequences of bud formation and breakdown are shown in Fig. 4 of Foe and Alberts (1983) and at http://raven.zoology.washington.edu/downTheTubes/.

There exists no space between the plasma membrane and the vitelline shell across most of the egg’s perimeter. In these regions, ‘buds’ still form – we believe by exactly the same mechanisms – but their formation involves invagination of ‘pseudocleavage furrows’ between nuclei rather than protrusion of buds containing nuclei. Mechanically, viewed from a coordinate system attached to the centrosomes, the two kinds of deformation of the membrane are congruent. The reason why these interactions would also cause cortical nuclei to rotate 90° with respect to the cortex between anaphase and telophase (Foe and Alberts, 1983), as diagrammed in Fig. 12, is explained in Foe et al. (1993). There we reviewed evidence for, and first explicitly hypothesized, F-actin-microtubule interactions via a kinesin intermediary. Note also that H1, in concert with H3 and H4, provide a mechanism for moving actomyosin rings along microtubules from their initial apical locations down to the basal ends of elongated nuclei during the ‘slow phase’ of cellularization in interphase 14. Fig. 8 in Foe et al. (1993) diagrams this.

H1 explains the cyclic appearance and disappearance of myosin II staining at the syncytial cortex, and H4 accounts for myosin II leaving the cortex where microtubule arrays contact the cortex. When cycloheximide causes an everlasting interphase 13 arrest and colcemid eliminates all microtubules during that interphase 13 arrest, rampant contractility is unleashed (Fig. 10). We interpret this to mean that continuing operation of just H1, in the absence of microtubules, fails to segregate F-actin and myosin II into the complementary localizations shown in row 3 of Fig. 1 and rows 3, 4, 5 and 8 of Fig. 2. Myosin returning to the cortex, and encountering there the spatially heterogeneous interphase distribution of F-actin (in caps or rings), will initiate strong cortical contraction instabilities that dramatically churn the embryo during the artificially prolonged interphase. Embryos arrested in metaphase would be expected to remain quiescent, as observed, because the cortex loses its associated myosin II during metaphase/anaphase (H1). Why doesn’t co-localized cortical F-actin and myosin II (as in row 1 of Fig. 1) destructively churn the embryo cortex during the interphase of cycle 9 and earlier, before migrating nuclei arrive? Firstly, before microtubule arrays contact the cortex in cycle 10, the interphase distributions of cortical F-actin and myosin II are globally uniform (cf. Fig. 2 row 2 versus rows 3-5) and, secondly, contractile instabilities take time to develop. Compared to the hours-long interphase that we effect by injecting cycloheximide, pre-cycle 10 interphases are very brief (<5 minutes at 25°C) and are interrupted by mitosis when cortical F-actin and myosin concentrations both plummet.

The same four mechanical hypotheses that we propose to explain pseudocleavage furrow and bud formation in the fly syncytial blastoderm, if operative in dividing mononucleate cells, could time and initiate localization of the actomyosin components of the contractile ring for cytokinesis (cartooned in righthand panels of Fig. 12). H1, by melting down prior interphase F-actin structures during metaphase, and H2 by causing F-actin to re-polymerize near centrosomes beginning in anaphase, in effect force a redeployment of the cell’s actin just prior to beginning the specialized task of cytokinesis. By H1, the cortical concentrations of F-actin and myosin II, having fallen to low concentrations in metaphase-anaphase, rebuild in telophase. But the expanding telophase microtubule asters approaching the cortex at opposite poles of the cell would trigger, via H4, depletion of cortical myosin II filaments near the spindle poles where microtubules impinge, while simultaneously concentrating myosin II filaments by moving them through the cytoplasm towards the cell mid-zone. H4 will thus eventually concentrate myosin II in a three-dimensional disk whose perimeter will become the contractile ring. By H2, actin polymer will form coincidentally with microtubule outgrowth initially most concentrated near centrosomes, followed later in telophase by a migration of F-actin along astral microtubules away from centrosomes (by H3) and toward concentration in a three-dimensional disk-shaped volume centered at the equator where it will co-localize with myosin II. The ‘rings’ of cortical F-actin seen in late interphase in Figs 2 and 3, and sketched in cross-sectional cartoons in the bottom left panel of Fig. 12 as cortical F-actin concentrations flanking the bud, correspond to where these equatorial disk volumes intersect the cortex in the dividing mononucleate cell – diagrammed in righthand panels of Fig. 12. The regions between buds in Figs 1 and 2, where both F-actin and myosin II are present together at the cortex during interphase (though concentrations of F-actin may be higher elsewhere) constitute the so-called pseudocleavage furrows in the fly syncytial blastoderm. In our view, this region would be homologous to the cortex of the cleavage furrow, which constricts during cytokinesis (cf. right and left panels in Fig. 12).

Wolpert (1960) proposed that, in spherical cells, transient ‘polar relaxation’ would convert a global cortical contraction into a self-amplifying equatorial contraction. Computer simulations (White and Borisy, 1983) showed how contracting an initially isotropic actomyosin meshwork into an equatorial belt aligns the filaments parallel to the equator, as in a contractile ring, positioning them to cleave a cell in two by a purse-string contraction. Plus-end transport along astral microtubules of cortical F-actin (H3) and of myosin II filaments (H4) could provide a mechanistic explanation for this oft-hypothesized polar relaxation, while simultaneously causing an increase in equatorial tension (assuming that cortical contractile strength is proportional to co-localized F-actin and myosin II concentrations). The right side of Fig. 12 cartoons these reorganizations. Either polar relaxation, or equatorial strengthening, or both together, can set the stage for the kind of actomyosin contraction-based cytokinesis that Wolpert first proposed and Borisy and White modeled. Schroeder (1981) proposed that the establishment of the contractile furrow was actually a mechanism requiring two independent steps – the release in tension at the poles hypothesized by Wolpert, preceded by (or coincident with) a global increase in cortical tensioning. The global loss of actin and myosin from the cortex during metaphase/anaphase and their return beginning in telophase, which we observe in Drosophila embryos (H1), if phylogenetically general, could underlie the cyclic changes in cortical contractility emphasized by Shroeder (1981). Operating together, H1-H4 are potentially capable of implementing a ‘global contraction-polar relaxation’ mechanism of furrow initiation for cytokinesis. If cortical contractile strength is proportional to co-localized F-actin and myosin II concentrations, then H3 and H4 would bring about equatorial strengthening of the cortical actomyosin meshwork, in principal also implementing ‘equatorial stimulation’ (Rappaport, 1996). Note that, in flattened cells with small asters, the microtubules in the mid-zone between spindle poles are positioned to execute the same actomyosin rearrangements as astral microtubules in spherical cells (Wheatley and Wang, 1996). Note also that H3 plus H4 can cause equatorial concentration of whatever cytoplasmic actomyosin network a cell contains, focusing internal forces on the furrow cortex with the potential to aid furrow invagination in non-spherical cells. Homologies between cleavage and pseudocleavage while attractive come with the caveat that cortical tensioning (Schroeder, 1981) and microtubule outgrowth (see for example Henson et al., 1989) both occur earlier in the mitotic cycle during cytokinesis in echinoderms than during pseudocleavage furrow formation in Drosophila syncytia. We do not know the significance of this timing difference, and have initiated experiments in sand dollar embryos to explore whether the interactions that we detect in fly embryos are phylogenetically general.

In summary, our study has aimed to deduce, from descriptions of wild-type and drug-perturbed cytoskeletal kinematics of microtubules, F-actin and myosin II in syncytial Drosophila embryos, the specific ways that these filament systems must be interacting. The speculative mechanistic hypotheses we deduce, H1-H4, are consistent with a large body of circumstantial and partial evidence reviewed above. This machinery, if phylogenetically general, could unify old ideas about cytokinesis with new molecular findings, reconcile polar relaxation with equatorial stimulation models of furrow formation, and homologize cytoskeletal pseudocleavage furrow formation in syncytia with cleavage furrow formation in mononucleate cells. We anticipate exciting future revelations of the molecular details by which the products of an ensemble of key genes (e.g. anillin, centrosomin, diaphanous, KLP-3A, pavarotti, polo kinase, Rac 1, septins, etc.) collaborate to bring about and regulate the interactions between F-actin, myosin II, centrosomes and microtubules our studies imply.

V. E. Foe gratefully acknowledges fellowship support from the MacArthur Foundation without which this exploration of unorthodox ideas and large time investment in technique development would have been untenable. G. M. Odell and V. E. Foe thank the University of Washington’s Royalty Recovery Fund for seed money grant RRF 65-2885. We thank Vera C. Foe for her interest and the gift of a fast computer at a crucial juncture; to her we dedicate this paper. C. M.Field is supported by NIH GM 23928 awarded to T. J. Mitchison. For the generous gift of antibodies, we thank Michelle Moritz (for anti-gamma tubulin) and Daniel Kiehart (for anti-cytoplasmic myosin II). We thank George von Dassow, Eli Meir, Ed Munro, Jon Alberts, John Sisson and two anonymous reviewers for their many helpful comments on our manuscript.