Behaviour of microtubules and actin filaments in living Drosophila embryos

The early events in Drosophila embryogenesis have been extensively studied (Rabinowitz, 1941; Sonnenblick, 1950; Turner & Mahowald, 1976; Zalokar & Erk, 1976; Foe & Alberts, 1983). The first thirteen nuclear divisions occur rapidly in a syncytial cytoplasm. All of the nuclei are initially located in the interior of the embryo, but during nuclear cycles 8 and 9 most nuclei migrate to the cortex of the embryo. By interphase of nuclear cycle 10, these nuclei form a continuous monolayer just beneath the surface of the plasma membrane. The surface nuclei in this ‘syncytial blastoderm’ embryo divide four more times in a nearly synchronous fashion and then become synchronously cellularized during interphase of nuclear cycle 14; subsequent local infoldings of the newly formed cell sheet mark the beginning of gastrulation.

A number of recent studies suggest that the cytoskeleton plays an important role in the localization of developmental information in early embryos (Edgar et al. 1988; Strome & Wood, 1983; Ubbels et al. 1983). One therefore suspects that a detailed knowledge of cytoskeletal organization and function will be integral to our understanding of the pattern formation process during embryogenesis. Several recent studies have initiated a characterization of the cytoskeleton in early Drosophila embryos (Warn et al. 1984, 1985, 1987; Walter & Alberts, 1984; Karr & Alberts, 1986; Warn & Warn, 1986). These studies reveal that two regions of the early embryo have a high density of cytoskeletal filaments. These same regions can be detected in the light microscope as cleared zones that exclude yolk particles and other large organelles (Scriba, 1964). One such region is the cortical layer of cytoplasm that lies just beneath the plasma membrane before the nuclei migrate to the periphery of the embryo. Actin filaments, microtubules and intermediate filaments form a dense meshwork in this cortical cytoplasm to a depth of about 3pm. The other region that is enriched in cytoskeletal elements constitutes a special ‘cytoplasmic domain’ that surrounds each nucleus. During interphase in the syncytial blastoderm, microtubules are organized around each nucleus by a pair of centrosomes located on the apical side of the nucleus and actin filaments form a cap-like layer above these centrosomes, just beneath the plasma membrane. By the time that mitosis begins, the centrosomes have migrated to either side of the nucleus, where they assemble a mitotic spindle. The actin cap enlarges as the nuclear cycle progresses, spreading into transient furrows surrounding each spindle during mitosis and dividing into two smaller caps at telophase. The cytoskeletal rearrangements that take place in the syncytial embryo are remarkable in their rapidity, since the successive nuclear cycles require only 8 (cycle 10) to 21 (cycle 13) min to complete (Foe & Alberts, 1983).

Earlier studies have employed a variety of fixation and staining techniques to visualize the cytoskeleton of the early Drosophila embryo. These methods have several disadvantages. One can never be sure that a fixation procedure accurately preserves the in vivo distribution of a protein, and the view that one obtains from a fixed specimen is necessarily a static one. In order to circumvent these problems, we have been developing techniques for visualizing fluorescently labelled cytoskeletal proteins that have been injected into living Drosophila embryos, similar to the techniques that have been used successfully in other biological systems (for reviews see Taylor & Wang, 1980; Taylor et al. 1986). By using novel fluorescent derivatives of actin and tubulin to visualize the behaviour of actin filaments and microtubules during the early syncytial nuclear divisions in the Drosophila embryo, the initial studies reported here have allowed us to gain a better understanding of cytoskeletal organization and dynamics in the early Drosophila embryo; they have also provided a standard for comparison that has allowed us to develop improved fixation procedures.

5-(4,6-dichlorotriazinyl) amino fluorescein (DTAF), lissamine rhodamine B sulphonyl chloride (LRSC), and 2V-hydroxy succinimidyl 5-carboxytetramethyl rhodamine (NHSR) were obtained from Molecular Probes. Dimethyl formamide and dimethyl sulphoxide were obtained from Aldrich (sure seal grade).

Tubulin has been labelled with N-hydroxy succinimidyl 5-carboxytetramethyl rhodamine (NHSR) or lissamine rhodamine B sulphonyl chloride (LRSC). For preparation of NHSR-labelled tubulin, 3 –4 mg of phosphocellulose purified bovine brain tubulin (Mitchison & Kirschner, 1984) at l –5 μg ml⁻¹ in 80mm-potassium Pipes (pH6 ·8), 1mm-Na₃EGTA, 1 mm-MgCl₂ (buffer designated BRB80) was polymerized into microtubules by adjusting the solution to 10% dimethylsulphoxide, 1mm-GTP, and 4mm-MgCl₂, followed by incubation at 37 °C for 25 min. The microtubules were then collected by centrifugation at 35 °C through a 4-5 ml cushion consisting of BRB80 plus 50 % sucrose (60 min at 50 000 revs min⁻¹ in a Beckman 50Ti rotor). The pellet was resuspended in enough BRB80 to give a protein concentration of 5 –8 mg ml⁻¹, and incubated on ice for 30 min to depolymerize microtubules. The tubulin solution was then clarified for 5 min in a Beckman microfuge and the pellet discarded. This initial cycle of polymerization and depolymerization is necessary to remove the residual β-mercaptoethanol (left over from the tubulin purification procedure) that interferes with the NHSR reaction. The tubulin was repolymerized into microtubules as described above, and then adjusted to 45 % sucrose by addition of prewarmed BRB80 containing 80 % sucrose (the addition of sucrose doubled the subsequent yield). A 50 mg ml⁻¹ stock of NHSR was made up in dimethyl formamide and stored at −70 °C. From this stock, NHSR was added to a final concentration of 3 mg ml⁻¹ and the reaction was allowed to proceed for 2 h at 37 °C. At the end of the labelling period, the microtubules were collected by centrifugation at 35°C through a 4 ·5 ml cushion consisting of BRB80, 50% sucrose, 10 mm-potassium glutamate, and 5 mm-dithiothreitol (DTT) (60 min at 50 000 revs min⁻¹ in a Beckman 50Ti rotor). The pellet was resuspended in 0 ·4ml of BRB80 containing 0 ·1 mm-GTP, 10 mm-potassium glutamate, 5mm-DTT and incubated on ice for 30 min to depolymerize the microtubules (the potassium glutamate and DTT were included in this step and the preceding one in order to block unreacted NHSR). The tubulin solution was then clarified by centrifugation at 4 °C for 20 min at 50 000 revs min’¹ (Beckman 50Ti rotor), followed by a brief spin in a microfuge.

In order to remove tubulin damaged by the coupling of rhodamine, the modified tubulin was subjected to two cycles of polymerization and depolymerization. Thus, the supernatant was adjusted to 33% glycerol, 1 mm-GTP, 4 mm-MgCl₂ and the tubulin repolymerized by incubation at 37 °C for 30min. The microtubules were then pelleted through a 50 % sucrose cushion in BRB80 as previously described. The pellet was resuspended in 0 ·4 ml of BRB80 containing 0 ·1 mm-GTP and depolymerized by incubation on ice for 30 min. Insoluble material was removed and the resulting tubulin solution was subjected to an additional round of glycerol-induced assembly and disassembly, in which the microtubule pellet was resuspended in 150 of 50 mm-potassium glutamate (pH 6 ·8), 0 ·5mm-MgCl₂, before depolymerization on ice for 30 min. This final tubulin solution was then centrifuged at 4 °C for 7 min at 301bfin⁻¹ (1 11bfin⁻¹ ≈6 ·9 kPa) in a Beckman airfuge and the supernatant frozen on liquid nitrogen in small aliquots. The procedure described produces an overall protein yield of 30 % and a stoichiometry that is typically about 0 ·9 rhodamines per tubulin dimer.

LRSC-tubulin and DTAF-tubulin were prepared in essentially the same manner as NHSR-tubulin, with the following modifications. LRSC was added to the tubulin solution in two equal-sized samples spaced by 5 min. The LRSC was added to a concentration of 0 ·8 mg ml⁻¹ (final) from a 50 mg ml⁻¹ stock made up in dimethyl formamide moments before using and the total reaction time was 10 min. Except for terminating the reaction by addition of 10 mm-potassium glutamate, the remainder of the procedure was the same as for NHSR-tubulin and it produced a stoichiometry of about 1. For DTAF-tubulin, DTAF was added to the reaction in 5 aliquots at 10 min intervals to a final concentration of 2-6 mg ml⁻¹. The DTAF was added from a 50 mg ml⁻¹ stock made up in dimethyl formamide and the total reaction time was 2 ·5 h, producing a stoichiometry of about 0 ·5.

Drosophila tubulin was prepared by a modification of procedures used by Detrich & Wilson (1983) to purify sea urchin embryo tubulin (D. R. Kellogg, C. M. Field and B. M. Alberts, unpublished data). Drosophila tubulin was labelled with LRSC as described for bovine tubulin.

Rabbit muscle actin was prepared according to Pardee & Spudich (1982) and stored at 4°C as filaments at 4-7 mg ml⁻¹ in a buffer containing 50mm-Tris-HCl, pH8 ·1, 100mm-KCl, 0-2mm-ATP, and 0 ·02% sodium azide. For preparation of fluorescently labelled actin, 2 –5 mg of this filamentous actin was dialysed for 3 h against 200ml of 50mm-potassium Pipes (pH6 ·8), 50mm-KCl, 0 ·2mm-CaC12, and 0 ·2 mm-ATP. In order to break up actin aggregates, the dialysed actin was stirred vigorously with a Teflon dounce homogenizer (several strokes) and brought to 2 ·3 mg ml⁻¹ by addition of the dialysis buffer. ATP was added to 0-5 HIM and NHSR was added to a concentration of 3 ·2 mg ml⁻¹ (from a 50 mg ml⁻¹ stock in dimethyl formamide, as described above). The reaction was allowed to proceed for 1 h at room temperature and actin filaments were then pelleted by spinning for 60 min at 50 000 rev min⁻¹ in a Beckman 50Ti rotor (20 °C). The pellet was suspended in 0-35ml of 20mm-Tris-HCl (pH8-l), 5mm-DTT, 5mm-potassium glutamate, 0-20tnM-CaC12 and 1 mm-ATP. An equal volume of a solution containing 1 ·8 M-KC1, 50 mm-Tris-HCl (pH 8-7), 5 mm-potassium glutamate, 3mm-CaCl₂, and 15 mm-ATP was then added. After swirling for 30 min at 0°C, the solution of depolymerized actin was spun for 5 min in a Beckman microfuge. The supernatant from this spin was layered onto a 1 ×25cm Biorad P10 column previously equilibrated with G buffer (5 mm-Tris-HCl, pH 8 ·1, 0 ·2mm-CaCl₂, 0 ·2 mm-ATP, and0 ·2niM-DTT). The protein pool in the void volume was adjusted to F buffer (50mm-Tris-HCl, pH8-l, 50mm-KCl, 1mm-ATP, 0 ·2mm-CaCl₂, 0 ·2mm-DTT) and left at room temperature for 45 min. Actin filaments were pelleted at 50 000 rev min⁻¹ for 60 min at 4°C in a Beckman 50Ti rotor, resuspended in 300 μl of G buffer, and dialysed for 36 h at 4 °C against G buffer to depolymerize actin filaments. The dialysed actin was then centrifuged for 7 min at 301bfin⁻¹ in a Beckman airfuge. A final cycle of assembly and disassembly was carried out by adjusting the supernatant to F buffer, incubating at room temperature and repelleting the actin filaments. The pellet was resuspended in 150 μl of G buffer, dialysed for 36 h against G buffer containing 0 ·1mm-DTT, centrifuged for 7 min at 301bfin⁻¹ in a Beckman airfuge at 4 °C and frozen on liquid nitrogen in small aliquots. The final protein yield was approximately 50 % and the stoichiometry was about 0 ·9 rhodamine molecules per actin monomer.

For both actin and tubulin, the stoichiometry of dye labelling did not change from one cycle of assembly to another, indicating that labelled and unlabelled proteins assemble with about equal efficiency. To estimate the stoichiometry of dye coupling, the concentration of dye linked to protein was determined by reading absorbance at 555 nm in 6 M-guanidinium-HCl (NHSR), or 0 ·1m-Tris-HCl pH9 ·5 (DTAF), with an extinction coefficient of 10⁵ used for both tetramethyl rhodamine and fluorescein (see Kodak laboratory chemicals catalogue). Protein concentration was estimated by the Bradford procedure, using bovine serum albumin as a standard.

Embryos were manually dechorionated and injected at 50 % egg length according to standard procedures (Santamaria, 1986). All solutions were spun for several minutes in a microfuge before loading into needles. The injected actin and tubulin tended to diffuse throughout the entire embryo within 10 min, although the concentration of the injected protein remained highest near the site of injection. In order to obtain solutions containing a mixture of NHSR-actin and DTAF-tubulin (Figs 4, 5) or NHSR-actin and NHSR-tubulin (Fig. 6), the appropriate protein solutions were mixed just before injection from the following stocks: 14 mg ml⁻¹ DTAF-tubulin (stoichiometry = 0 ·45), 5 ·8 mg ml⁻¹ NHSR-tubulin (stoichiometry = 0 ·6) and 5 ·6 mg ml⁻¹ NHSR-actin (stoichiometry = 1 ·3).

Injected embryos were photographed using a Nikon Dia-phot-inverted microscope fitted with a 35 mm camera back. All micrographs were taken with a Zeiss ×100 neofluor objective; and a mercury arc lamp (HBO 100W) was used for illumination. Kodak Technical Pan 2415 film was hypersensitized by exposure to 41bf in⁻¹ hydrogen at 30 °C for 5 days (apparatus and information available from Lumicon, 2111 Research Drive, Livermore, CA 94550). This produces an ASA of approximately 1400 and allows a 2 –4 s exposure, which greatly reduces photobleaching and blurring of the image due to cytoplasmic movements within the embryo. Video recordings were made using a Zeiss IM35 microscope and a Zeiss X100 neofluor objective. A Zeiss halogen illuminator turned to full power was used as a light source for low-light-level video recordings, although satisfactory images could be obtained with lower light levels. The microscope was coupled to a Cohu SIT Camera and an image processor from Interactive Video Systems.

We have used two procedures for fixation of embryos for immunofluorescence. One is a slight modification of the procedure of Warn & Warn (1986). Dechorionated embryos are immersed in a mixture of 1 part heptane and 1 part 97% methanol: 3% 0 ·5M-Na₃EGTA (EGTA stock adjusted to pH 7 ·5) at room temperature. After shaking vigorously for 1min, embryos that have lost their vitelline layer sink to the bottom of the tube. These are recovered and stored in the methanol/EGTA mixture for several hours at room temperature, or overnight at 4 °C. The embryos are rehydrated by passage through a series of 20, 40,60 and 80 % PBS in methanol. The fixation procedure of Karr & Alberts (1986) was carried out as described, except that twice as much formaldehyde was added.

We have used three fluorescent probes to label bovine tubulin: dichlorotriazinyl amino fluorescein (DTAF), lissamine rhodamine B sulphonyl chloride (LRSC), and A-hydroxy succinimidyl tetramethyl rhodamine (NHSR). For all three tubulins, reaction conditions yielding a labelling stoichiometry of 0 ·4 –1 ·0 were selected, so that most of the labelled molecules carry only a single fluorescein (DTAF-tubulin) or rhodamine (LRSC-tubulin and NHSR-tubulin).

DTAF-tubulin is a well-characterized derivative that has been used in many other laboratories; it appears to function normally in vitro and in vivo, as shown by a variety of criteria (Keith et al. 1981 ; Leslie et al. 1984; and Salmon et al. 1984). However, the fluorescein chromophore is relatively light sensitive and is thus rapidly photobleached at the light levels needed for observation. To avoid this problem, we have preferred to use rhodamine reagents to label tubulin. As expected, tubulin labelled with LRSC is significantly more resistant to photobleaching. When LRSC-tubulin is injected into early Drosophila embryos or tissue culture cells, it becomes rapidly incorporated into endogenous microtubule networks. However, this incorporation is short-lived; within 5 min one observes a decrease in the labelling of tubulin networks, accompanied by the appearance of rhodamine fluorescence in vacuole-like structures (data not shown). This phenomenon occurs even in the dark and is presumably due to the selective proteolytic degradation of LRSC-tubulin.

Because the DTAF-tubulin is stable in the dark after its microinjection, the instability of the LRSC-tubulin is presumably due either to the rhodamine chromophore or to the different linkage of this chromophore to tubulin. Another rhodamine derivative was therefore prepared and tested by microinjection. Tubulin labelled with NHSR is both resistant to photobleaching and stable in vivo. When injected into embryos, the NHSR-tubulin is incorporated into endogenous microtubule networks in less than 30 s, and it provides strikingly clear images of microtubule arrays in the living embryo.

The major Drosophila embryo tubulins are 96% (alphaj tubulin) and 95 % (beta_x tubulin) identical to their vertebrate analogues in porcine brain and chicken brain, respectively (Theurkauf et al. 1986; Rudolph et al. 1987). As a check on the suitability of bovine tubulin as a marker for Drosophila microtubules, tubulin was prepared from Drosophila embryos and labelled with LRSC (see Methods). When this tubulin was injected into embryos, it produced labelling patterns that were indistinguishable from those obtained with bovine tubulin (data not shown). Most experiments were therefore performed with bovine tubulin because of its ready availability from brain. We also saw no differences in the labelling patterns generated by the different fluorescent derivatives of tubulin; indicating that our tubulin derivatives are reliable in vivo probes.

Fig. 1 presents a series of fluorescence micrographs taken at the indicated time intervals of an embryo injected with a 7 mg ml⁻¹ solution of NHSR-tubulin. We estimate that this amount of injected tubulin represents 2 ·5 % of the total tubulin pool in the Drosophila embryo, assuming that tubulin represents 3 % of the total protein (Loyd et al. 1981) and that the injected volume and the embryo volume are approximately 2×10⁻⁴μl and l ·5×10⁻²μl, respectively (Foe & Alberts, 1983). In order to reduce the background due to out-of-focus fluorescence and light scattering in these thick specimens, a ×100 times objective was used and the field diaphragm was closed down to include only a single nucleus. The nucleus shown in Fig. 1A is in late interphase; microtubules are seen to radiate from asters on either side of a dark region that corresponds to the nucleus. Fig 1B-L are identical views that show the changes in microtubule distribution that take place as this nucleus proceeds from interphase of cycle 10 through metaphase of cycle 11 in the syncytial blastoderm embryo. In cycle 10, the Drosophila embryo contains about 300 nuclei that are regularly arranged in a monolayer just beneath the egg plasma membrane (Zalokar & Erk, 1976; Foe & Alberts, 1983). All of these nuclei behave indistinguishably when viewed in this way (see also Karr & Alberts, 1986).

Video intensification techniques allow substantially lower light levels than the film procedures used for Fig. 1, and they thereby allow the continuous recording of embryos injected with fluorescently labelled proteins (for reviews see Bright & Taylor, 1986 and Inoue, 1986). Fig. 2 presents a series of photographs taken from a video monitor showing the changes in microtubule distribution that take place during the transition from nuclear cycle 10 to 11. Similar recordings have allowed single embryos to be monitored from cycles 10 through 14, a period of 2h, without damage to the embryo -as judged by its subsequent hatching to a normal larva (data not shown).

Data such as those shown in Figs 1 and 2 reveal a regular pattern of changes in microtubule distribution during each nuclear cycle. Data for the timing of events in cycles 10 and 11 are compiled in Table 1. In telophase, a pair of centrosomes is located near the surface of the nucleus that is closest to the plasma membrane (Figs 1H and 2E). These centrosomes require a few minutes of interphase to migrate to opposite sides of the nuclear envelope, where they remain for a few more minutes before the next mitosis begins.

Throughout this interphase period, prominent astral microtubules emanate from each centrosome and extend away from the nucleus (Fig. 1A,I). After centrosome migration, some microtubules traverse the surface of the nucleus between the two centrosomes (not shown) and a fine ring of fluorescence appears that surrounds the nuclear envelope as mitosis begins (Figs 1J,B and 2B). A partial nuclear envelope breakdown follows (see Stafstrom & Staehlin, 1984), allowing the metaphase spindle to be assembled. This spindle has an unusual morphology, inasmuch as the poles appear to be separated by a small gap from the remainder of the spindle (Figs 1D,L and 2C); a similar morphology has been observed by electron microscopy in isolated spindles of the sea urchin embryo (Salmon & Segall, 1980). During metaphase, the spindles of adjacent nuclei often make short oscillatory movements relative to each other, typically making excursions of 3gm or so. Early anaphase is marked by the appearance of prominent astral microtubules and the elongation of the spindle (Fig. 1E). The asters grow rapidly and by late anaphase the spindle is largely disassembled, leaving only a bundle of interzonal microtubules and large asters associated with each of the reforming daughter nuclei (Figs 1F and 2D). The interzonal microtubules persist well into telophase (Fig. 1G). Telophase ends with the paired centrosomes at their interphase position on top of each nucleus and the disappearance of the interzonal microtubules (Fig. 1H).

In addition to filamentous fluorescence., each nucleus is surrounded by a domain of weaker, diffuse fluorescence that persists throughout the nuclear cycle and is assumed to represent unpolymerized tubulin. This diffuse domain of fluorescence becomes clearest in interphase/prophase of nuclear cycle 11 (Fig. 3A). Similar domains of diffuse fluorescence are seen to surround each nucleus in embryos that have been injected with fluorescently labelled dextrans of 35 000 M_r (Fig. 3B). When focusing up and down through each nuclear domain, the separation between the adjacent domains in Fig. 3 is detected only in the half of each domain closest to the plasma membrane. It is therefore possible that the surface protrusions seen as ‘somatic buds’ above interphase and prophase nuclei by Foe & Alberts (1983) and Stafstrom & Staehlin (1984) produce the apparent separation in the fluorescent domains around each nucleus.

The viability of embryos is unaffected by injection of fluorescently labelled tubulin. Typically, 80-90% of embryos injected with control buffer or labelled tubulin develop into normal hatching larvae. The amount of light required to visualize the labelled tubulin also does not seem to affect viability, since the majority of the embryos photographed develop into normal hatching larvae.

Our images of microtubule networks in vivo are consistent with those of Warn et al. (1987), who have used microinjection of a fluorescently labelled antibody against tyrosinated α-tubulin to visualize microtubule networks in living Drosophila embryos.

We have used rabbit muscle actin labelled with NHSR as a probe for actin filament networks in the Drosophila embryo. In our initial experiments, we labelled rabbit muscle actin in a pH 7 ·6 buffer -a standard condition used for preparing actin (Pardee & Spudich, 1982). This labelled actin behaved in a manner indistinguishable from unlabelled actin through two cycles of in vitro assembly and disassembly, and it rapidly became incorporated into the endogenous actin networks when injected into embryos. However, as with LRSC-tubulin, the actin labelled in this way was degraded within 5 –10min, producing a vacuolar staining pattern. This was unexpected, since tubulin labelled with NHSR is not degraded in vivo. Because the NHSR-tubulin had been labelled at pH 6-8, we repeated the NHSR labelling of actin at a lower pH. As with the pH 7 ·6 labelling, the actin labelled to a good stoichiometry at pH 6-8 and behaved normally through two cycles of in vitro polymerization and depolymerization. However, this actin showed no significant degradation in vivo. The absorbance spectra of the different NHSR-actins suggest that different functional groups are labelled at pH 7 ·6 and pH 6 ·8. The absorbance spectrum of actin labelled at pH 6 ·8 shows a largely monophasic peak in the rhodamine region (E_max at 555 nm), while actin labelled at pH7 ·6 shows a biphasic spectrum (peaks at 520 nm and 555 nm). We believe that the splitting of the rhodamine peak seen for actin labelled at pH 7 ·6 is due to an interaction of the rhodamine ring structure with the native actin molecule. This hypothesis is supported by the observation that the absorbance spectrum of the pH 7·6 NHSR-actin after denaturation with 6M-guanidinium-HCl closely resembles the monophasic peak seen for the pH 6·8 labelled actin.

When the NHSR-actin prepared at pH 6-8 is injected into Drosophila embryos, it is incorporated into endogenous actin networks within less than a minute. As with fluorescently labelled tubulin, the viability of embryos is unaffected by injection of NHSR-actin. Fig. 4 presents a series of fluorescence micrographs showing the changes in actin distribution that take place from interphase of cycle 11 to prophase of cycle 12. We were able to assign exact stages in the cell cycle by coinjecting fluorescein-labelled tubulin with the rhodamine-labelled actin. Fig. 5 presents several examples of fluorescence micrographs taken from the Fig. 4 embryo, with a switch to the fluorescein filter cassette used to visualize microtubules.

Our work with fluorescently labelled actin confirms the results of previous studies that used fixed embryos and a specific actin stain (Warn et al. 1984; Karr & Alberts, 1986). In interphase of nuclear cycle 11, each nucleus is centred beneath a flat cap-like layer of actin, which has a coarse fibrous appearance (Fig. 4A). As metaphase approaches, the cap enlarges by spreading laterally until it comes into contact with adjacent actin caps (Fig. 4B,C). In micrographs taken at a focal plane just below the surface actin layer, the actin at this stage can be seen to extend below the surface of the actin cap at the regions of contact between adjacent actin domains. By metaphase, each spindle is therefore surrounded by a shell of actin that extends over its surface and around its sides. This distribution of actin is particularly clear in embryos that have been coinjected with a mixture of NHSR-actin and NHSR-tubulin, so that both actin filaments and microtubules are labelled with rhodamine. Fig. 6 shows micrographs taken from such an embryo in metaphase of nuclear cycle 12. The micrograph at the left has been taken at a focal plane above the spindle and shows the surface of the actin cap (Fig. 6A); the other micrograph was taken at a deeper focal plane and shows the metaphase spindle surrounded by a belt of actin (Fig. 6B). During metaphase, a slight concentration of actin fluorescence can be seen in the region of the spindle (Fig. 4D,E). However, a similar concentration of fluorescence in the spindle region is seen in control embryos injected with equivalent amounts of fluorescently labelled dextran of 35 000 M_T (not shown). At anaphase, the actin domains elongate and the belts of actin surrounding each domain begin to disappear. By late anaphase or early telophase, the actin appears to form a continuous layer over the surface of the embryo (Fig. 4F), and it is difficult to detect evidence of individual domains corresponding to each nucleus. The actin domains reappear during late telophase (Fig. 4G), and by interphase each nucleus is again overlaid by its own unique actin cap just below the plasma membrane (Fig. 4H).

The availability of the labelled embryos just described has allowed us to evaluate fixation procedures to determine which ones give immunofluorescent staining that most accurately mimics the images that we have obtained in vivo. By this criterion, microtubules were not perfectly preserved by any tested procedure, although the best results were obtained by using a variation of a methanol-fixation procedure (Warn & Warn, 1986). Fig. 7 compares the distribution of microtubules in vivo (Fig. 7A-C) and in methanol-fixed embryos (Fig. 7D-F) at three stages of the mitotic cycle. The figure shows that although the bulk distribution of microtubules is preserved in the methanol fixation, much of the fine structure is lost. Thus far we have been unable to find a better fixation procedure for microtubules. In particular, inclusion of glutaraldehyde in the methanol fixation procedure (0 ·1 –1%) did not seem to improve significantly the preservation of microtubules. The formaldehyde/taxol procedure used by Karr and Alberts gives striking images, but it causes an explosion of the centrosomal region and the appearance of excess astral microtubules during interphase and prophase (Fig. 7G-I).

Similar studies show that actin is reasonably well preserved by the methods of Warn et al. (1984) and Karr & Alberts (1986), and that the methanol fixation that best preserves microtubules poorly preserves actin networks (not shown).

We would like to acknowledge David M. States for his expert preparation of the manuscript. This research was supported by NTH grant no. GM23928 to B.S.A. and by NTH institutional training grant no. GM08120 to D.R.K.