Contribution of maternal factors and cellular interaction to determination of archenteron in the starfish embryo

Analysis of the determination mechanism in echinoderms has been mainly based on sea urchin embryos. The results of classic studies have suggested that the area of the archenteron is generally specified by localization of maternal factor(s) around the vegetal pole (VP) of the unfertilized egg and determined finally by cellular interaction during and/or after the cleavage stage (reviewed in Hörstadius, 1973). This idea has been reaffirmed and partially modified by recent experiments using new axis markers, altered tracers and differentiation markers (Maruyama et al., 1985; Henry et al., 1989; Khaner and Wilt, 1990, 1991; Ransick and Davidson, 1993).

As a material for analyzing the mechanism of determination of the archenteron, the eggs and embryos of starfish have several advantages compared with those of sea urchins. (1) They have good markers of the animal-vegetal axis even before cleavage (Shirai and Kanatani, 1980; Schroeder, 1985). In ordinary sea urchins, in contrast, the animal and vegetal pole is not evident until the micromeres appear at the vegetal pole after the fourth cleavage. In some species, the pigment band surrounding the subequatorial zone can be used as the marker for the animal-vegetal axis (Schroeder, 1980b), though such species are not cosmopolitan. Position of the microcanal marks the animal pole in ordinary sea urchins (Schroeder, 1980a; Maruyama et al., 1985). However, the position of the canal cannot be recognized unless the eggs are immersed in suspension of small particles such as sumi ink, and gradually becomes indistinguishable as the jelly coat disperses away in the sea water. (2) They are larger than those of sea urchins, making it easier to prepare oocyte fragments for deletion and transplantation experiments designed to clarify the presence and distribution of the maternal cytoplasmic factor(s). (3) Both oocyte fragments and blastomeres derived from the animal hemisphere have little capacity for archenteron formation (Maruyama and Shinoda, 1990; Zhang et al., 1990; Kiyomoto and Shirai, 1993a), unlike those of sea urchins (Henry et al., 1989; Khaner and Wilt, 1990, 1991), and appear to be good recipients for implantation experiments.

Accordingly, Kiyomoto and Shirai (1993b) fused the VP fragment obtained from immature oocytes with the fertilized and denuded egg fragment derived from the animal half by electric pulse, and showed that the VP fragment contains maternal cytoplasmic factors that are both necessary and sufficient for archenteron formation. After the first five cleavages, embryos are composed of four tiers each of which has eight cells. All the blastomeres are almost equal in diameter. Kiyomoto and Shirai (1993a) also showed that some vegetal blastomeres obtained from the vegetal half, probably from the most vegetal tier (veg2), have capacity to respecify ectoderm cells to endoderm, by co-culturing animal half embryos with the vegetal blastomeres from the 32-cell-stage embryos.

In the present study, the contribution of the maternal factor(s) and the cellular interaction to determination of each part of the archenteron of starfish larvae was assessed by means of the following approaches. (1) Temporal and positional pattern of expression of archenteron-specific enzyme, alkaline phosphatase (AP), was examined histochemically. (2) VP fragment was deleted from immature oocytes and the archenteron-forming capacity of the partial embryos derived from the remaining animal fragment was examined. (3) The inductive capacity of the VP region was tested by making the vegetal polar region of one of the 2-cell-stage blastomeres face the animal polar region of the other using the blastomere-rotation technique.

Adults of Asterina pectinifera were collected in the subtidal zone at Akiya, Kanagawa prefecture, during their breeding seasons in May, 1991-1993, and kept in running seawater. Follicle-cell-free oocytes and spermatozoa were prepared as described elsewhere (Kuraishi and Osanai, 1988). The oocytes were inseminated by adding diluted suspension of spermatozoa about 45 minutes after start of treatment with 1 μM 1-methyladenine (1MA). In this text, developmental time is described as time after start of 1MA treatment. The fertilized eggs were allowed to develop at 19°C.

Positional pattern of expression of AP was examined histochemically. When the larvae reached the given stages, they were fixed in cold ethanol (4°C) for 30 seconds as described in Nishida (1992). The fixed specimens were stained histochemically for AP according to Whittaker and Meedel (1989). The stained specimens were transferred into a wedge-shaped egg holder (Kishimoto, 1986), oriented using a glass needle attached to a micromanipulator, and observed and photographed by bright-field microscopy as a whole mount.

Follicle-cell-free oocytes were transferred to a wedge-shaped egg holder and held loosely between the upper and lower glasses. Each oocyte was bisected perpendicular to the animal-vegetal axis using a micro glass needle attached to a micromanipulator. The animal and vegetal poles were identified as the point of contact of the germinal vesicle with the oocyte surface and its opposite end, respectively (Shirai and Kanatani, 1980; Schroeder, 1985).

The degree of deletion of the VP fragment was evaluated in terms of the ratio of the volume of the VP fragment relative to the sum of those of the animal and vegetal fragments. Both fragments maintained a spindle shape for a while after bisection. The lengths of their long and short axes were measured using a micrometer attached to the eye piece of a microscope. Then the volume of the fragments was calculated assuming that they were spheroids.

The animal fragments were inseminated 45 minutes after the start of treatment with 1MA and allowed to develop individually in filtered seawater at 19°C in 24-hole microtest culture plates. In some cases, 100 units/ml penicillin G and 50 μg/ml streptomycin sulfate were added to the filtered seawater.

The swimming larvae were observed at every third hour from 18 to 36 hours and at every 6th or 12th hour after that, using a brightfield stereomicroscope (25 –80×), and checked for the presence of an archenteron.

One of the blastomeres was rotated around the axis of apicobasal polarity at the 2-cell stage within the fertilization envelope (according to Kuraishi and Osanai, 1988), to make its animal and vegetal pole face the vegetal and animal pole of the other. The rotated blastomere was injected by pressure with 5 –10% horseradish peroxidase (HRP; Type VI, Sigma Inc.) in distilled water or 5 –10% lysyl rhodamine dextran (M_r 20,000, Molecular Probes) in distilled water soon after the rotation. In some cases, two granddaughter cells of the rotated blastomere derived only from its animal half were injected with 5% fluorescein isothiocyanate dextran (M_r 70,000, Research Organics, Inc.) in distilled water at the 8-cell stage.

It is difficult to maintain the orientation of blastomeres in recombined embryos of starfish, since the adhesiveness of the blastomeres is very weak, especially in the early stage of cleavage. The intraenvelope system applied in this study made it possible to preserve blastomere orientation after the operation and specify the origin of the blastomeres (Kuraishi and Osanai, 1988, 1989).

When the larvae reached given stages, they were fixed in 1% glutaraldehyde in filtered seawater for several hours and stained histochemically as described in Kuraishi and Osanai (1988). The stained specimens were transferred to an egg holder and photographed, controlling the orientation using the fine glass needle attached to the micromanipulator. In some cases, the specimens were dried in a graded series of ethanol, cleared in xylol and mounted in xylol-balsam to obtain higher contrast between the labeled and unlabeled regions.

When the labeled larvae reached the given stages, their ciliary movement was inhibited by treatment with 100 mM sodium azide in 80% filtered seawater, followed by transfer to an egg holder (Kuraishi and Osanai, 1992). The larvae were then observed and photographed using a fluorescence microscope. The observed larvae were washed twice with filtered seawater and allowed to develop in filtered seawater at 19°C until the next observation.

Expression of AP is used as a marker for differentiation of archenteron or endoderm in many marine invertebrates including sea urchins (Henry et al., 1989; Khaner and Wilt, 1990), ascidians (Bates and Jeffery, 1987; Whittaker and Meedel, 1987; Nishida, 1992) and brachiopods (Freeman, 1993). To clarify whether it could also be used as a marker in starfish larvae, the temporal and positional pattern of AP expression was examined histochemically in normal development.

Larvae developing normally were fixed at 18, 24, 27, 30, 36, 39, 42, 48 and 72 hours, stained histochemically for AP and observed as a whole mounts. About 18 hours after 1MA treatment, the VP area of larvae starts to invaginate. No staining was recognized from the start of gastrulation (18 hours, Fig. 1A) to the middle of mesenchyme-differentiation stage (24 hours, Fig. 1B). At the end of the mesenchyme differentiation stage, the apical cortex of the whole archenteron except the tip of the archenteron became AP positive (27 hours, Fig. 1C). The staining became stronger and spread to the basal part of the positive area during the intermission of archenteron elongation up to the start of mes-enchyme migration (30 hours, Fig. 1D). The stained area did not spread in a longitudinal direction. Thus, the posterior part of the archenteron became AP negative, because of translocation of negative cells from the outer layer (Fig. 1E). The apical cortex of the posterior part of the archenteron became AP positive after the early mouth-formation stage (39 hours, Fig. 1F). Then the staining spread over the entire cell (Fig. 1G-J).

During the mouth-formation stage (40–48 hours), the rudiment of the left posterior coelom projected into the blastocoel from the dorsal side of the boundary zone between the first and second area of AP expression (Fig. 1H). The rudiment was AP negative throughout the mouth formation stage (Fig. 1H,J).

At the bipinnaria stage, the entire area of the endoderm (esophagus, stomach and intestine) was AP positive and no staining was recognized in the rest of the larval body (Fig. 1K,L). This two-step expression of AP suggests that at least two mechanisms participate in commitment of archenteron.

To clarify the roles of the maternal factors around the VP in determination of archenteron, we deleted an anucleate fragment from the VP of the immature oocytes and monitored its effects on archenteron formation. When the volume of the deleted fragment was less than 6% of the whole oocyte, most of the larvae formed an archenteron. In contrast, VP deletion over 8% strongly suppressed archenteron formation (Fig. 2A). As a control to assess the effect of the cutting operation itself, a fragment was cut off from the equatorial region. Equatorial fragments less than 7%, ranging 7 to 15% and ranging 15 to 25% were removed from two, four and thirteen oocytes, respectively. In all cases examined, the operated larvae formed an archenteron. This result suggests that some factors responsible for archenteron formation are concentrated in the 7% region around the VP of immature oocytes.

In normal development, cells from the outer layer are continuously added to the posterior end of the archenteron as a result of involution throughout the gastrula stage (Kuraishi and Osanai, 1992). If the time of involution of each part of the archenteron is specified to some extent by maternal cytoplasmic factor(s) around the VP of immature oocytes, VP deletion may cause delay in start of gastrulation. Thus, relationship between the time after 1MA and the percentage of gastrulated larvae that had an archenteron at 72 hours in the above experiment (45 vegetally and 19 equatorially deleted) was plotted in Fig. 2B. The deleted volume ranged from 1.3 to 16% and 4.2 to 24.3 % in vegetally and equatorially deleted groups, respectively. In the former group, most of the larvae started gastrulation between 18 and 30 hours, while most of the larvae in the latter group started gastrulation at 18 hours, just as the intact controls. Though involution continues even after the early mesenchyme-migration stage (30 hours) in normal development (Kuraishi and Osanai, 1992), few larvae started archenteron formation after that. This result suggests that the area that involutes before 30 hours is committed to some extent by maternal cytoplasmic factors, while the area that is added to the archenteron after 30 hours is not determined at the prophase of meiosis.

To get detailed information about the effect of VP deletion on embryogenesis, the morphology and the expression of AP of VP-deleted specimens was examined. For morphological observations, the same samples as the above experiment were used.

When the volume deleted was less than 3% of the total, the larvae started gastrulation at almost the same time as the intact ones and formed apparently normal gastrulae and bipinnariae (Fig. 3A). When the volume deleted was 4-7% of the total, the larvae started gastrulation by 30 hours as described above. The archenteron appeared smaller than that of the intact larvae, especially in the inflated area at the tip of the archenteron (Fig. 3B2). In these larvae, differentiation of the digestive tract was suppressed or delayed to some extent. The larvae shown in Fig. 3B had only reached the very early bipinnaria stage at 72 hours when the intact controls formed early bipinnaria. The anterior coelomic pouches and esophagus, which are formed from the inflated area in normal development (Kuraishi and Osanai, 1992), were less well developed than those of the intact larvae at the very early bipinnaria stage (Fig. 3B3).

VP deletion ranging from 8 to 15% strongly suppressed gastrulation. However, the larvae became flattened parallel to the animal-vegetal axis and a depression, apparently homologous to the stomodaeum, formed in the center of the flattened surface (Fig. 4A). This suggests that the larvae had an axis of dorsoventral polarity. When the volume deleted exceeded 15%, such flattening was hardly evident and the depression formed at the anterior end of the larva (Fig. 4B). Thus it is probable that some factor(s) responsible for the establishment of dorsoventral polarity resides in the vegetal polar 15% region.

As mentioned above, deletion from the equatorial region did not suppress gastrulation. When the volume deleted was relatively large, the profile of the larvae appeared smaller than that of intact ones. However, no critical difference was observed in the size and shape of the archenteron between the deleted and intact larvae (Fig. 4C).

Expression of AP was examined using fixed specimens obtained from another set of VP-deleted larvae. The larvae were fixed about 48-55 hours, when the intact controls started the secondary expression of AP and stained histochemically for AP. In this set of samples, deletion of more than 7% also suppressed archenteron formation strongly (Table 1). Those larvae that had not gastrulated did not express AP, with only one exception, while all the larvae that had gastrulated expressed AP (Table 1; Fig. 5). In cases after 1-7% deletion, the larvae often expressed AP even in the tip of the archenteron or in the rudiment of the anterior coelomic pouch (Table 1; Fig. 5A). In normal larvae, these parts of the larvae never expressed AP (Fig. 1C-H).

The results of VP deletion suggests that the area that enters the archenteron before 27 hours is determined to some extent by the maternal cytoplasmic factor(s) localized around the VP. In contrast, commitment of the area that is added to the posterior end of the archenteron after that seems to require cellular interaction from the VP area. In order to examine if the VP area has capacity to induce the posterior part of the archenteron, the vegetal pole of one of the 2-cell-stage blastomeres was made to contact the animal pole (the presumptive area of ectoderm) of the other by blastomere rotation.

As shown previously (Kuraishi and Osanai, 1988), when one of the blastomeres had been rotated 180° around the axis of apicobasal polarity at the 2-cell stage, the embryos first became spherical blastulae and then formed two independent archentera at the two vegetal poles (Fig. 6A). The archentera gradually curved toward the hemisphere that had been formed from the other blastomere of the 2-cell-stage embryo until the early mesenchyme-migration stage (Fig. 6B). The boundary between the labeled and unlabeled region, which could be detected even by bright-field microscopy (Fig. 6A1, A2), stayed on the inner edge of the blastopore lip until that stage (Fig. 6B2). This bend of the archentera may have resulted from onesided supply of cells into the archenteron from the area composed of the same half as the archenteron. After the start of mesenchyme migration at 30 hours, the boundary gradually moved into the archenteron and became undetectable by bright-field microscopy (Fig. 6C).

At the bipinnaria stage, the larvae developed duplicitas posterior, with shared frontal plane and anterior organs (Fig. 7B) as the authors reported elsewhere (Kuraishi and Osanai, 1988). An intermittent band of the labeled cells was detected in the middle-to-posterior part of the stomach and the intestine of the unlabeled digestive tract. Since the cells in the embryonic wall of early starfish larvae are not considered to be intermingled to a great degree (Kominami, 1983), the cells in this band may be derived from the animal half.

To ensure that the cells really originated from the animal half, the two animal-half descendants of the rotated blastomere were labeled at the 8-cell stage. At the bipinnaria stage, the labeled cells were detected not only in the ectoderm but also in the middle-to-posterior part of the stomach and the intestine in one of the digestive tracts (Fig. 7C). Some of the blastomererotated larvae were examined histochemically for expression of AP at 72 hours to make it clear whether the involuted population of the animal-half progeny had differentiated into endoderm. In all cases examined, the animal-half progeny in the digestive tract expressed AP (Fig. 7A).

The archenteron of the larvae of A. pectinifera is divided into three major regions on the basis of the temporal and positional pattern of expression of AP. The first region is the AP-negative area at the tip of the archenteron. This region never expressed AP throughout the early development and formed mesodermal tissues such as anterior coeloms and mesenchymes (anterior mesodermal area: AMA), as reported in Kuraishi and Osanai (1992). The second region is AP-positive area 1 (APA1), which entered the archenteron and started to express AP at the end of the mesenchyme-differentiation stage (27 hours). The third region was added to the posterior end of the archenteron after 27 hours and became AP positive after 39 hours (AP-positive area 2: APA2). Besides the three major regions, the rudiment of the left posterior coelom forms on the dorsal side of the boundary zone between APA1 and APA2. The rudiment and the differentiated left posterior coelom never expressed AP. The different properties on expression of AP suggest that the mechanisms of commitment of these regions differ. According to the results of the trace of movements of archenteron cells (Kuraishi and Osanai, 1992), APA1 is considered to form the esophagus and the anterior end of the stomach of bipinnaria larvae, whereas APA2 is considered to form the middle-to-posterior part of the stomach and the intestine (Fig. 8A).

The results of VP deletion indicate that the vegetal 7% area is necessary for archenteron formation. Embryos derived from blastomeres in the animal half (Maruyama and Shinoda, 1990) and those from animal half fragment of oocytes (Zhang et al., 1990; Kiyomoto and Shirai, 1993a) do not gastrulate. These findings suggested that maternal cytoplasmic factor(s) responsible for archenteron formation is concentrated in the vegetal hemisphere. Our results narrow the supposed distribution of the factor(s) to a much more limited area, namely the vegetal 7% area.

VP deletion less than 7% caused a delay in the start of gastrulation. Even in these cases, gastrulation started before 30 hours. The operated gastrulae had a poorly developed anterior coelomic pouch and esophagus. These facts suggest that the fate of the area that enters the archenteron before 30 hours is determined by the maternal factor(s) to some extent. Since increase in archenteron volume ceases between 27 and 30 hours (Kuraishi and Osanai, 1992), this area corresponds to the combined area of AMA and APA1 (Figs 1C,D, 8A).

When the deleted volume exceeds 7%, no more delay in gastrulation was observed and the archenteron formation was strongly suppressed. In normal development, the founder area of APA2 is added to the posterior end of the archenteron after 30 hours. Provided the 7% deletion had not removed the presumptive area of APA2, this result indicates that APA2 is not determined in the immature oocyte. The result of blastomere rotation shows that the cells around the VP have capacity to induce the area corresponding to APA2. It is probable that APA2 is specified during development by inductive influence from the vegetal polar region.

To inspect this possibility, we analyzed the movement of the 7% area during early development, examining if it gives rise to AMA and APA1. The factors that are responsible for archenteron formation are considered to be concentrated in the oocyte cortex (Shirai and Kanatani, 1980; Kiyomoto and Shirai, 1993b). Thus we analyzed the delivery of the cortex of the 7% region during the early development.

When a sphere is cut into two across a plane, the relationship between the relative volume (V (%)) and cortical area (C (%)) inherited by one of the two fragments is described by the simple formula: C3-150C2+5000V=0. Using this formula, the surface area belonging to the 7% area is calculated to be about 16% of the total.

In normal development of A. pectinifera, the oocyte cortex becomes the apical cortex of blastomere. Each cleavage plane divides both the apical cortex and the cytoplasm almost equally, at least up to the eighth cleavage (Kuraishi and Osanai, 1989). Thus the ratio of cytoplasm allotted to each part of an embryo is the same as that of the cortex. As a result, the volume of the embryonic wall, which possesses the cortex of the 7% area is estimated to be about 16% of the whole. According to the results of measurement of archenteron volume (Kuraishi and Osanai, 1992), the proportion of the whole larval wall occupied by the combined AMA and APA1 is considered to be a little bit larger than 15% (by cell volume). Thus it is suggested that the vegetal 7% area gives rise to the combined area of AMA and APA1 (Fig. 8A).

As discussed above, 7% deletion would have removed only the founder area of AMA and APA1 (Fig. 8C). However, the relative volume of the mesendoderm in the whole embryo finally exceeds 25% in starfish larvae (Kuraishi and Osanai, 1992). Taking into account the discussion above, it is calculated that at least 15% deletion would be necessary to remove all the presumptive area of mesendoderm (AMA, APA1 and APA2). The fact that the VP deletion ranging from 7 to 15% suppressed archenteron formation strongly suggests that APA2 is not determined in the immature oocyte and that it requires some effects from the more vegetal region. As discussed below, VP deletion of about 1% would be sufficient to remove the presumptive area of AMA. Since archenteron elongation continues after 30 hours in larvae after VP deletion ranging from 1 to 7%, the presumptive area of AMA does not seem to be required for specification of APA2. Thus it is suggested that the influence required for specification of APA2 is a homoiogenetic induction from APA1. In blastomere-rotated embryos, in contrast, cells derived from the animal hemisphere entered the archenteron after 30 hours. At the bipinnaria stage, these cells formed a part of the middle-to-posterior part of the stomach and the intestine, and expressed AP. This result indicates that cells around VP does have sufficient capacity to induce the area corresponding to APA2.

Kiyomoto and Shirai (1993a) cultured embryos derived from the animal-half fragment of oocytes with a blastomere obtained from the vegetal hemisphere of 32-cell-stage embryos and obtained gastrulae. In about a half of the cases, cells derived from the animal-half fragment were observed in the archenteron. Although inductive capacity of the VP area is shown by their result, its contribution in normal archenteron formation has not been clarified. Our results postulate the indispensability of the inductive effect from APA1 in specification of APA2. The boundary between the induced area (APA2) and the inductor (APA1) comes at the anterior part of the stomach at the bipinnaria stage (Fig. 8A). The position coincides with the most anterior end of the animal-half-derived area observed in the recombinants reported by Kiyomoto and Shirai (1993a). It is interesting that the boundary between the inductor and the induced area does not coincide with that of the organs of bipinnariae (the esophagus, stomach and intestine). The capacity to form APA2 is practically the same between the presumptive area of APA2 and that of the ectoderm.

In the blastomere-rotated embryos, cells from the animal half stayed out of the blastopore lip until 30 hours. The bend of the archentera and periblastopore region in these embryos suggests that the animal cells stayed out of the blastopore lip against the tractoring force from the archenteron. In normal development, increase of archenteron volume ceases once from 27 to 30 hours, suggesting that the cells in the presumptive area of APA2 pause by the blastopore lip for several hours before they enter the archenteron. The cells that contribute to APA2 seem to require respecification out of the blastopore lip before they are added to the archenteron.

The volume of the presumptive area of anterior coelomic pouch (identical with AMA) of mesenchyme-differentiation stage larvae is calculated to be less than 5% of the total volume of the larva, adopting the method of Kuraishi and Osanai (1992). Thus, deletion of 1% fragment from the VP would be sufficient to remove the founder area of the anterior coelomic pouch completely. In the present study, VP deletion ranging from 1 to 7% did not suppress formation of anterior coelomic pouch completely. In these cases, AP-positive cells often appeared in the presumptive area or the rudiment of the anterior coelomic pouch. This result suggests that a proportion of cells in the presumptive area of APA1 changed their fate from endoderm to mesoderm in these embryos. In other words, the default fate of cells in at least a part of the presumptive area of the APA1 is mesoderm. Cells in AMA may exert suppressive influence to those in the presumptive area of APA1 and change their fate to endoderm in normal development. It is interesting that the transformed cells in VP-deleted larvae still expressed AP in the rudiment of the anterior coelomic pouch. Expression of AP may have been determined in these cells before the decision as to whether it becomes endoderm or mesoderm.

The sequence of determination of mesendoderm in starfish is schematically illustrated in Fig. 8. In an immature oocyte, the founder areas of the mesoderm (AMA), endoderm (APA1 and APA2) and ectoderm are piled up sequentially from the vegetal pole to the animal pole (Fig. 8A). As discussed above, at least three mechanisms participates in determination of archenteron: (1) determination of the combined area of AMA and APA1 by maternal cytoplasmic factor(s) localized in the founder area of it, (2) specification of APA2 by homoiogenetic induction from APA1, and (3) specification of APA1 by suppressive interaction from AMA.

Kiyomoto and Shirai (1993b) fused an oocyte fragment obtained from VP with a fertilized animal-half oocyte fragment, which hardly forms gastrula by itself, and proved that the VP fragment contained factors necessary and sufficient for archenteron formation. We reaffirmed that the maternal cytoplasmic factor(s) around VP is indispensable for archenteron formation. Additionally, our results suggest that the factors only participate in determination of the anterior part of the archenteron, namely AMA and APA1 (Fig. 8A,C).

Kominami (1984) showed that treatment with Li+-rich sea water during the period from 7 to 10 hours (around 256-cell stage) increased the relative volume of the archenteron of early-to-middle gastrula (early mesenchyme-migration stage gastrula in this text) at 20°C. This fact indicates that the combined area of AMA and APA1 is finally determined after the sensitive period for Li+. One of the most reasonable explanations of this phenomena is that the concentration of the factor(s) responsible for determination of AMA and APA1 gradually decreases in the periphery of the presumptive area, and that the area that contains the factor at more than a threshold level around the 256-cell stage give rise to APA1 and AMA. Li+ treatment may expand the combined area by lowering the threshold level.

The cellular interaction that specifies the boundary between the ectoderm and the endoderm in starfish embryo shown in present study is slightly different from that known in sea urchins. In both animals, the archenteron is roughly determined by the cytoplasmic factor(s) concentrated around the VP (Hörstadius, 1973; Maruyama et al., 1985; Zhang et al., 1990; Kiyomoto and Shirai, 1993a,b). However, the cellular interaction that finally specifies the archenteron differs between the two animals. In starfish, only part of the presumptive area of the archenteron (AMA and APA1) acquires archenteron-forming capacity. Then it changes the fate of the surrounding area (APA2 in normal development) to the middle-to-posterior part of stomach and intestine by inductive influence (Fig. 8). In sea urchins, it is well known that the most-vegetal cells in a 16-cell-stage embryo, micromeres, have extended capacity to induce archenteron (Hörstadius, 1973; Ransick and Davidson, 1993). However, recombinant larvae consisting of mesomeres isolated at 16-cell stage and/or veg1 blastomeres at 64-cell stage, which give rise to ectoderm in normal development, also forms archenteron (Khaner and Wilt, 1990, 1991). It is considered that suppressive interaction finally limits the archenteron to the veg2 descendant. In starfish, involution takes place even in the later stages of gastrulation (Kuraishi and Osanai, 1992), whereas it lasts only for the first few hours of gastrulation in sea urchins (Burke et al., 1991). The difference in duration of involution may be related to the difference in the type of cellular interaction.

Significance of cytoplasmic factors and cellular interaction on determination of mesendoderm is well known in amphibians and ascidians. However, the arrangement of the dermis and the sequence of determination is quite different from those reported here on starfish. In amphibians, the endoderm area, which occupies the most vegetal part of a fertilized egg, is determined first and then it induces the mesoderm in the surrounding ectoderm area (reviewed in Gilbert, 1988). In ascidians embryos, the presumptive area of the endoderm resides at VP, but many organs can differentiate autonomously without VP area (reviewed in Satoh, 1994). The most evident difference from starfish is that the localization of the factors is not established until the cytoplasmic segregation takes place after fertilization in these animals. In ascidian embryos, for instance, the cytoplasmic segregation takes place in two phases after fertilization. The factors responsible for determination of muscle and endoderm accumulates around the VP during the first phase of the segregation and then moves to the founder area of each tissue during the second phase of the segregation (Nishida, 1992, 1993). The factor responsible for gastrulation is temporarily concentrated at the VP between the two phases of segregation (Bates and Jeffery, 1987; Jeffery, 1990a,b). In fertilized eggs of starfish, in contrast, cytoplasmic segregation is not evident. Localization of factors responsible for archenteron formation and differentiation of endoderm is already completed during oogenesis.

The authors thank Dr H. Nishida (Tokyo Institute of Technology), Dr H. Shirai (Ushimado Marine Laboratory, Okayama University) and Dr M. Kiyomoto (Ochanomizu University) for useful discussions and technical advice. Thanks are also due to Dr K. Sano (Akkeshi Marine Biological Station, Hokkaido University) and Dr T. Kishimoto (Tokyo Institute of Technology) for managing cooperative collection of starfish. We also thank Messrs S. Tamura and M. Washio of Asamushi Marine Biological Station for their assistance in rearing the animals.