Reassessing embryogenesis in the Ctenophora: the inductive role of e₁ micromeres in organizing ctene row formation in the ‘mosaic’ embryo, Mnemiopsis leidyi

Ctenophores are a phylum of marine organisms of uncertain phylogenetic relationship. The body plan of the vast majority of ctenophores is seen in Fig. 1. The major body axis is the oral-aboral axis with the mouth situated at one pole and the apical sense organ located at the opposite pole. Most ctenophores possess two bilaterally situated tentacles that are used to capture prey. They locomote through the water column by the coordinated action of eight ctene rows. Ctene rows run along the oral-aboral axis and are composed of multiple ctene plates. Each adult ctene plate consists of a linear array of thousands of cilia bound together by compartmental lamellae that are generated by epithelial polster cells (Tamm and Tamm, 1981). Coordination of the movement of the eight ctene rows is controlled in large part by the apical sense organ, which consists of a gravity sensor (statolith) connected to the ctene rows by ciliated grooves (see Tamm, 1982 for review). If viewed from the oral or aboral pole (Fig. 1), ctenophores appear to be composed of four nearly identical quadrants separated by two orthogonal planes. One plane runs through the two tentacles (the tentacular plane) and the other through the undistended gut (the esophageal or sagittal plane). These axes define planes of rotational (not mirror) symmetry, because adjacent quadrants are not identical to one another. In fact, diagonally opposed quadrants share a more similar organization (Martindale and Henry, 1995).

Ctenophore development is unique and does not bear obvious similarities to embryogenesis in any other extant organisms. First cleavage is unipolar, passes through the future sagittal plane and defines the oral pole (Freeman, 1977). Second cleavage passes through the future tentacular plane and gives rise to four equal-sized cells. Each of the first four blastomeres divides to give rise to E (end) and M (middle) blastomeres, which occupy distinct positions in the embryo (Fig. 2). Each E blastomere gives rise to a series of three smaller micromeres at the aboral pole (called e₁, e₂ and e₃), while the M blastomeres generate two aboral micromeres (m₁ and m₂). Following a period of rapid cell division and gastrulation, differentiated cell types begin to appear. Embryogenesis in most ctenophores is rapid, and direct, with a juvenile Mnemiopsis leidyi developing in less than 24 hours (18-20°C).

The regularity of the early cleavage program and ability to identify individual blastomeres in ctenophore embryos led to the construction of an embryonic fate map. Previous workers placed chalk particles on identified blastomeres and followed them through embryogenesis (Ortolani, 1964). This analysis indicated that ctene rows are generated exclusively from the E lineage while the ciliated groves are derived from the M (m₁) lineage. One ctene row is supposedly generated from each of the two daughter cells of each of the e₁ micromeres, the e₁₁ and e₁₂ blastomeres (Reverberi and Ortolani, 1963; Ortolani, 1964, reviewed by Reverberi, 1971; Ortolani, 1989).

The results of both cell isolation and deletion experiments supported the view that ciliated comb plates are derived from e₁ micromeres. Blastomeres isolated at the 4-cell stage, E cells at the 8-cell stage and e₁ cells at the 16-cell stage all make comb plates. None of the other cells in the embryo differentiate ctene rows when isolated (reviewed by Reverberi, 1971; Ortolani, 1989). Cell deletion experiments also support the notion that e₁ micromeres are the sole progenitors of ctene rows. The deletion of all four E cells or e₁ micromeres resulted in the complete absence of comb plate formation at the end of embryogenesis and two comb plates are missing for each E or e₁ cell that was destroyed (Farfaglio, 1963; Martindale, 1986). The destruction of other identified cells up through the 60-cell stage had no effect on the appearance of ctene rows. These sorts of experiments have led workers to conclude that the developmental potential of individual blastomeres in ctenophores is equal to their developmental fate, the classic definition of a mosaic embryo (Chun, 1880; Driesch and Morgan, 1895).

In a recent study, we discovered that, during the process of adult regeneration, ctene plates are generated by m₁ micromere derivatives (Martindale and Henry, 1996). We have therefore reinvestigated the embryological origins of ctene rows in the ctenophore Mnemiopsis leidyi using intracellular lineage tracer DiI. We show that the use of DiI dissolved in soybean oil is a reliable cell lineage tracer in ctenophore embryos that is more accurate than previous labeling methods. Injections of specific micromeres reveal that both e₁ and m₁ micromere lineages normally generate the adult ctene rows. Thus, the fact that no ctene rows form during embryogenesis following removal of e₁ micromeres indicates that m₁ micromeres normally require inductive interactions of e₁ micromere and/or their descendants. We have also found that while individual e₁ and m₁ micromeres generate stereotypical fates and regions of the embryo, there is some variability in the exact complement of cells generated from embryo to embryo. This suggests that the fates of e₁ and m₁ micromeres are not rigorously defined as a consequence of the early cleavage program and that field properties may play a role in establishing cell fates and patterning events in ctenophore embryos.

Adult specimens of the lobate ctenophore M. leidyi were collected at Quissett Harbor or off the rock jetty at NOAA in Woods Hole, MA during the months of July and August. One or two self-fertile adults were placed in small bowls of sea water in the dark. After approximately 7-8 hours, fertilized embryos were collected and demembranated with jewelers forceps in gelatin-coated plastic Petri dishes (Zalokar and Sardet, 1984). Individual e₁ and m₁micromeres were injected with glass microelectrodes by pressure with 5 mg/ml DiI (catalog no. D-282, Molecular Probes Inc., Eugene OR) dissolved in soybean oil. A stock solution of 100 mg/ml DiI dissolved in ethanol was diluted 20-fold with soybean oil. Control injections with 5% ethanol in Wesson oil had no effect on the development of injected cells up through the cydippid stage. One or more small DiI oil droplets were placed in each identified cell (see Fig. 3). The diameter of each droplet was controlled by the pressure of nitrogen used to drive the injector and measured 1/5-1/10 the diameter of the injected cell (General Valve, Inc., Fairfield, NJ). Injected embryos were cultured in 0.22 μm filtered sea water (FSW) for 25-48 hours. They were examined live in 6.5% MgCl₂, with slight compression, under a coverslip using a Zeiss Axioplan microscope equipped with DIC and fluorescence optics. Photographs were taken on 400 ISO Kodak Ektachrome film.

There has been some confusion in the literature about the naming of e₁ and m₁ micromere descendants. For the purposes of this paper, the daughter cell that is closest to the sagittal (esophageal) plane is designated as the e₁₁ or m₁₁ cell, and the daughter furthest away from the sagittal plane the e₁₂ or m₁₂ blastomere (Fig. 2).

In order to perform the various cell deletion experiments, uncleaved embryos were demembranated with fine forceps in gelatin-coated plastic Petri dishes. Hand-pulled needles were dipped in molten 2% agarose to keep the embryos from sticking to the needles during orientation (Martindale, 1986). Individual, or pairs, of e₁ or m₁ micromeres were killed by stabbing them with fine glass needles at the 16-cell stage. Operated embryos were carefully transferred into individual wells of gelatin-coated 24-well plates in FSW and examined 22-26 hours after the deletions were performed. Only embryos that remained intact and did not lose additional cells were scored.

Intracellular cell lineage analyses of the embryological origins of ctene rows have not been performed previously in ctenophores. In order to determine the fidelity of the lipophilic dye DiI as a lineage tracer, we injected identified blastomeres and watched the domains of fluorescent labeling through subsequent development. Injected droplets of DiI quickly labeled the internal and plasma membranes of the injected, but never the adjacent, blastomeres (Fig. 3A,B). Labeled clones stay together in a contiguous domain through gastrulation (Fig. 3C). Development proceeded normally in both injected and uninjected cells through cydippid formation. These results indicate that the intracellular injection of DiI dissolved in soybean oil is an effective and reliable cell lineage tracer in the Mnemiopsis embryo.

Our injections confirm previous cell lineage experiments of ctenophore embryogenesis that each e₁ micromere generates a pair of ctene rows (Reverberi and Ortolani, 1963; Ortolani, 1964). Labeled e₁ clones (n=32) always include both adesophageal and adtentacular ctene rows and their respective ciliated grooves (Fig. 4A,B), as well as contributions to the floor of the apical organ, tentacle apparatus and a small amount of epidermis. Lithocytes and dome cilia, both components of the apical organ, were not labeled by e₁ injections. Fluorescence was seen in the base of the polster cells as well as all the way to the tips of the ctene plates. Interestingly, there was some variation in the pattern of labeling from clone to clone. For example, in some clones, a strip of epidermis was labeled from the apical region down to the mouth while, in others, this epidermal staining was present from the tentacle sheath to the oral opening. These variations were detected despite efforts to vary the intensity of fluorescent labeling by adjusting the size and number of injected dye droplets. Labeling in the ctene rows was not always homogeneous, with the edges of some ctene rows appearing to be unlabeled. There were several cases in which the adtentacular halves of the row were brighter than the adesophageal region of the row, suggesting that there is some variability in the contributions of cells to these regions. The presence of unlabeled portions of the ctene rows following the injection of e₁ micromeres suggested that other cells contribute to these structures.

Previous reports indicated that e₁₁ micromeres generate the adesophageal ctene row while e₁₂ micromeres generate the adtentacular row (Reverberi and Ortolani, 1963; Ortolani, 1964). Our injections indicate that this is not entirely correct. We injected each of the two daughter cells of the e₁ micromeres, the e₁₁ (n=14) and e₁₂ (n=11) cells in order to see how their fates contributed to each ctene row. Fig. 4C,D shows an example in which an e₁₁ micromere clone has generated the adesophageal ctene row and ciliated groove, as well as the oralmost ctene plate of the adtentacular row. Fig. 4E,F illustrates another embryo in which the sister cell of the e₁₁ blastomere, called e₁₂, has generated the adtentacular row and ciliated groove plus a large portion of the adtentacular region of the adesophageal row and groove. 12 of the 14 e₁₁ injections labeled both adesophageal and adtentacular comb rows. The remaining two only labeled adesophageal rows. 10 of 11 e₁₂ fills labeled both comb rows with the remaining case labeling only the adtentacular row. Both e₁₁ and e₁₂ cells generated cells in the apical organ, tentacle sheath and epidermis as well, but did not generate consistently unique cell-specific domains of labeled tissue. Thus, these two e₁ subclones are not precisely identical from embryo to embryo and indicate that there is normally variation in the partitioning of adult descendants in injected labeled e₁-derived clones. Only ectodermal derivatives were labeled by injecting e₁ micromeres and/or their descendants. No contribution to the endodermal canal system was observed.

The labeling patterns derived from m₁ fills were distinctly different than those observed from e1 fills. M ₁clones give rise to large epidermal domains that span from the aboral pole to the mouth (Fig. 5A,B). Labeled epidermal clones are interspersed with unlabeled cells derived from other cell lineages, resulting in a ‘leopard skin’ pattern. Epidermal patterns varied slightly from embryo to embryo. m1 micromeres also contribute to the dome cilia, portions of the tentacular apparatus, part of the floor of the apical organ and cells underneath the comb rows, probably components of the adult gonad. Unexpectedly, we also observed that 17/26 cases examined also generated one or more ctene plates in the injected quadrants. In a few cases (5/26), part of the ciliated groove was also generated by m1 descendants. Based on the position of the m1 micromeres during embryogenesis (i.e. closer to the sagittal plane than e1 micromeres), one would expect the m1 lineage to contribute to just the adesophageal components of the ctene rows (Fig. 2); however, we observed contributions to either or both ctene rows in the quadrant of the injected blastomere. In two of the 26 cases, m1 derivatives contributed to two adesophageal comb rows in adjacent quadrants.

Although we performed an extensive cell lineage analysis, no other blastomeres in the M. leidyi embryo were found to contribute to the formation of comb plates Martindale and Henry, unpublished data).

In contrast to what is seen in the e₁ lineage, the daughters of m₁ micromeres generally give rise to distinctly different subclones (Fig. 5C-F). Both m₁ descendants contributed to epidermis, the polar fields, the floor of the apical organ and sub-ctene row material (possibly including the somatic gonad). m₁₁ descendants (n=13) gave rise predominantly to epidermal and sub-ctene row material associated with the somatic gonad while m₁₂ cells make ctene plates (n=18/22), sub-ctene tissue, dome cilia (n=11/15) and components of the tentacular system (n=12/22). Epidermal domains were highly variable and unlike contributions to other structures, cells often crossed the boundaries separating individual quadrant territories.

In most larval/adult ctenophores (Hyman, 1940), two different diagonally opposed pairs of quadrants can be identified by morphological criteria. One pair that contains anal canals, the so-called ‘slash’ or / pair (Martindale and Henry, 1995) and another pair, the ‘backslash’ or \ pair, that do not (Fig. 2). These two pairs of quadrants can be distinguished following third cleavage due to the asymmetric position of the E and M macromeres. We made every effort to detect differences in the fates of /e₁ (n=11 cases) and \e₁ (n=14 cases) micromere descendants and /m₁ (n=7 cases) and \m₁ (n=9 cases) micromere descendants. Each micromere injection gave rise to progeny in the expected quadrant, but no consistent differences in the compliments of labeled progeny were detected between\ and / quadrants in these lineages.

In order to determine whether inductive interactions are required for the formation of ctene rows in M. leidyi, we deleted either two adjacent e₁ micromeres or two adjacent m₁ micromeres and then raised these embryos for 24 hours. Two adjacent micromeres were killed in order to assess any contributions that micromere descendants might have made to structures straddling quadrant boundaries (e.g. the tentacle apparatus). Furthermore, the results of single micromere deletions have been previously reported (Farfaglio, 1963; Ortolani, 1963; Martindale, 1986). In all 59 cases in which adjacent e₁ micromeres were deleted, there was a complete absence of four ctene rows on the operated side (Fig. 6A). In addition, the endodermal outpocketings (derived from E and M macromeres), which normally underlie each ctene row were missing and the tentacular apparatus was very reduced (57/59 cases) or missing (2/59 cases). This is in keeping with previous reports of the role of the e₁ lineage in endodermal canal formation (Farfaglio, 1963; Martindale, 1986). Our lineage analyses reported above have ruled out the direct contribution of e₁ descendants in the formation of the endodermal canals, therefore, e₁ micromeres must play an inductive role in controlling endodermal, as well as m₁ derivatives during ctenophore development.

In 22/24 cases in which two adjacent m₁ micromeres were deleted, all eight ctene rows including their ciliated grooves, and endodermal canals appeared, although, in some cases, the ctene plates in the four ctene rows on the deleted side appeared to be narrower (Fig. 6). In the remaining two cases, two ctene rows were missing. These results indicate that, although e₁ descendants have profound effects on m₁ derivatives, the reciprocal relationship does not appear to occur.

We have performed intracellular lineage analysis on embryos of the lobate ctenophore, M. leidyi, to examine the contributions and roles of specific cells in the formation of adult comb rows. We have shown that previous fate maps describing the development of identified blastomeres in ctenophores are incomplete. In particular, the two longitudinal ctene rows in each body quadrant are generated from two different micromere lineages, the e₁ micromeres, as previously described (Reverberi and Ortolani, 1963; Ortolani, 1964) as well as the m₁ micromeres. Furthermore, other differences in the previously accepted ctenophore fate map (Reverberi, 1971) were also detected in the course of this investigation. For example, the lithocytes present in the apical organ are not generated by either e₁ or m₁ micromeres. In addition, m₁ micromeres (not e₁) consistently generated dome cilia covering the apical organ. Finally, the e₁ micromeres normally make the ciliated grooves that provide continuity between the apical sense organ and the individual comb rows. Our labeling experiments also revealed a subtle degree of variability in the contributions of labeled clones derived from identified blastomeres. A detailed description of the fates of clones derived from identified ctenophore blastomeres not described in this paper will be reported in a subsequent publication (Martindale and Henry, unpublished data).

In order to make sure that M. leidyi embryos behave like other ctenophore embryos previously examined (Farfaglio, 1963; Martindale, 1986), we deleted adjacent pairs of e₁ or m₁ micromeres at the 16-cell stage. Our results demonstrate that two ctene rows fail to form for every e₁ micromere deleted, but that all eight ctene rows appear regardless of the number of m₁ micromeres destroyed. One pair of ctene rows did not appear in 2/24 cases in which 2 adjacent m₁ micromeres were killed. This latter finding might indicate that m₁ derivatives have weak inductive potential on e₁ micromeres. A more likely explanation, however, is that we inadvertently killed one e₁ micromere during these operations. Our results essentially corroborate those of Farfaglio (1963) with respect to the comb rows, however, we observed ciliated grooves following the destruction of m₁ cells. This is in contrast to previous reports that claimed that the ciliated grooves were derived from m₁ cells (Reverberi and Ortolani, 1963; Ortolani, 1963, 1964; Reverberi, 1971). The fact that m₁ micromere descendants fail to generate ctene plates following the destruction of e₁ micromeres indicates that e₁ micromeres, and/or their descendants, are required to induce m₁ derivatives to contribute to ctene plate formation during normal development. Further more, destruction of e₁ micromeres also results in the failure of the endodermal canal system to form under the ctene row (Farfaglio, 1963; Martindale, 1986). Our cell lineage experiments have now formally shown for the first time that the canal system is not generated by e₁ micromere descendants themselves. Therefore, e₁ micromeres and/or their descendants are also required for the appearance of these structures. It is clear that e₁ micromeres and/or their descendants are important ‘organizers’ for m₁ and other cell lineages during ctenophore development.

How does axial information arise in ctenophore embryos? Cell labeling studies have shown that uncleaved ctenophore embryos do not possess any meaningful axial information, and that the major body axis, the oral-aboral axis, is set up by the site of first cleavage (Freeman, 1977). Cutting experiments have shown that factors required for ctene row formation by e₁ micromeres are distributed uniformly in the uncleaved ctenophore embryo and that ctene row potential is transiently localized to the future oral pole at the beginning of first cleavage (Houliston et al., 1993), and subsequently to the aboral pole along the poles of the tentacular axis over the next two cell cycles (Freeman, 1976a,b). Thus, factors are ultimately segregated into the e₁ lineage, which are responsible for the ability of these cells to generate ctene plates. Likewise, one must also argue that there is a segregation of developmental potential, which imparts the e₁ lineage with the ability to induce ctene and endodermal canal formation in adjacent lineages. This developmental potential cannot be uniformly distributed because the deletion of other cell lineages has no effect on ctene plate production (Reverberi, 1971; Ortolani, 1989; our unpublished data) and its segregation to the e₁ lineage is likely to be coupled with the mechanics of the early ctenophore cleavage program (Freeman, 1976a,b).

Ctenophores develop rapidly and directly to juvenile adult form. During this embryonic period, ctenophore embryos fail to replace (i.e., regulate) missing regions of the embryo. After hatching, however, structures, like comb plates, can be formed. A recent cell lineage analysis of the origins of ‘regenerated’ comb plates in juvenile ctenophores derived from embryos in which e₁ micromeres were destroyed at the 16-cell stage indicated that the new combs were derived from the m_1, but not other, lineages (Martindale and Henry, 1996). This result makes some sense now that we know that m₁ cells normally contribute to ctene rows during embryogenesis. The fact that new ctenes form in juveniles suggests that soon after hatching, a new source of comb plate inducing potential may arise which is required by m₁ derivatives for comb plate production. This new source of inducing material may be endoderm, because in juveniles, newly generated comb rows are associated with underlying endodermal tissue (Martindale, 1986; Martindale and Henry, 1996). The exact role that e₁ micromeres play in organizing ctenophore development is unclear. For example, e₁ micromeres and/or their descendants may directly induce both m₁ and endodermal derivatives simultaneously during embryo genesis. Alternatively, e₁ micromeres could directly induce endodermal cells, which subsequently induce m₁ descendants. This latter scenario might explain how m₁ descendants can serve to regenerate ctene rows in the adult (Martindale and Henry, 1996). This notion of sequential inductive interactions can be tested by culturing e₁ and m₁ micromeres either with or without the presence of endodermal precursors to see if m₁ descendants require the presence of endoderm to form comb plates.

The use of intracellular lineage tracers have modified the interpretation of classical cell lineage work in other systems as well. For example, Nishida and Satoh (1983, 1985) performed intracellular lineage analysis in ascidian embryos and found that six of the first eight blastomeres normally give rise to muscle cells, where only two had been thought to generate muscle on the basis of earlier descriptive cell lineage investigations (Conklin, 1905a). This example is particularly pertinent due to the fact that it also forced a reinterpretation of the experimental embryology of the ascidians. In ascidians, two of the six muscle precursors are determined to give rise to (the majority) of tail muscle cells at the time of their birth (Conklin, 1905b), but the other cells require inductive interactions from adjacent cells and are thus conditionally specified. These results indicate that further work utilizing modern cell lineage techniques is required in order to obtain accurate information about how embryos develop.

We observed some variation in the labeling patterns following the injection of e₁ and m₁ micromeres, as well as other identified cells in the ctenophore (Martindale and Henry, unpublished data). This variation could be due to the technique that we used to label cells, however, we performed a number of experiments at various stages of development that indicate that development proceeds normally and that DiI spreads rapidly within the injected cell but not to adjacent cells. It therefore appears that a certain amount of variability in the fates of adult cells generated by identified cells is a natural component of development in M. leidyi. This means that all cell fates are not rigidly established at the time of their formation, as previously thought, and that ctenophore development exhibits field properties found in many other kinds of embryos.

So why exhibit a highly stereotyped cleavage pattern if it is not sufficient to establish lineage-specific cell fates? Based on both morphological and molecular data (Harbison, 1985; Ax, 1989; Brusca and Brusca, 1990; Christen et al., 1991; Eernisse et al., 1992; Morris, 1993; Wainright et al., 1993) the cnidarians appear to be the closest living relative of the ctenophores (see Nielsen, 1995 for an alternative scenario). However, most cnidarian embryos possess a loosely-defined cleavage program (called ‘chaotic’ cleavage), are highly regulative, and demonstrate field properties into late embryonic and larval stages. For example, axial properties can be entrained in dissociated hydrozoan larval cell cultures by grafting experiments (see Freeman, 1990 for a review). In ctenophores, it would appear that the overt regularity in the fates of identified blastomeres is due, in part, to the fact that the cleavage program generates cells in defined and regular positions with respect to other cell signaling centers in the embryo (i.e. the e₁ micromeres and/or their descendants). This role of the cleavage program can also be invoked for other lower metazoans. Recent evidence in the soil nematode, C. elegans, has indicated that the highly stereotyped cleavage program in these embryos is designed to insure that the appropriate arrangement of cells is established for subsequent inductive interactions to occur, a process that starts as early as the four cell stage (Schierenberg, 1987; Priess and Thomson, 1987; Goldstein, 1992, 1995; Mello et al., 1994). In addition, many invertebrate phyla display a highly conserved and regular pattern of development called spiral cleavage. Individual blastomeres are born in distinct, predictable positions, that can be named, and the fates of which are highly conserved, even between individual phyla (Wilson, 1898; Verdonk and van den Biggelaar, 1983). In these forms, it is clear that fates along the dorsoventral axis are causally determined by interactions with the so-called ‘D quadrant’ (see van den Biggelaar and Guerrier, 1983; Henry and Martindale, 1987, 1994). Thus, it can be argued that the primary role of the spiralian cleavage program is to establish a signaling center, the D quadrant, and well defined positions for neighboring cells to interact with this region of the embryo to trigger sequential cascades of molecular events in the proper spatial and temporal order.

It is currently unclear to what extent cell-cell signaling occurs between other cells in the ctenophore embryo (Martindale and Henry, in progress). Taken together, our data argue that ctenophores might be an important group of animals to investigate the relationship of the cleavage program to cell and axial determination in bilaterian metazoans. Recent, findings have demonstrated that there is a surprisingly conserved set of molecular features that link many metazoans (e.g. eye development, Hox genes and dorsoventral polarity) and the reassessment of ctenophore embryogenesis opens up the possibility that the evolution of the cleavage pattern and its role in establishing spatial organization may have deep roots in the Metazoa.

The authors thank the community of the Marine Biological Laboratory for facilitating these studies and Steve Q. Irvine and Matt Kourakis for comments on the manuscript. M. Q. M. was supported by an American Cancer Society Illinois Division Grant no. 92-43, NSF grant no. 9315653, and MBL Spiegel and Davis Fellowships. J. Q. H. (J. J. H.) was supported by an MBL Associates Fellowship, and Lemann Fellowship.