Pair-rule expression of a cell surface molecule during gastrulation of the moth embryo

In Drosophila, the metameric division of the body along the anterior-posterior axis is effected by a gene control set called the segmentation genes (Nüsslein-Volhard and Weischaus, 1980). Genetic and molecular definition of segmentation gene action suggest that the number, identity and polarity of the different body segments result from the sequential action of different segmentation gene classes (recent reviews by: Scott & O’Farrell, 1986; Akam, 1987; Ingham, 1988). These classes provide such information by subdividing increasingly smaller domains of the body into regions that display increasingly more specific gene expression. Thus various segmentation gene products appear to act as transcriptional regulators that control the expression of segmentation genes within their class, and of genes lower down in an apparent hierarchy (Nüsslein-Volhard et al. 1984; Harding et al. 1986). These interactions are responsible for conferring unique identities to blasto-derm cells by establishing unique combinations of segmentation (and dorsal group) gene expression (Chan & Gehring, 1971; Struhl, 1985; Anderson, 1987; Schneuwly et al. 1987).

Many of these interactions occur within the syncytial blastoderm of Drosophila and prior to the cellularization of the embryo. However, genetic and molecular evidence has more recently suggested that, in Drosophila, cellular interactions are also important for the overall generation of embryonic pattern by the segmentation gene network (DiNardo et al. 1988; Martinez-Arias et al. 1988). For example, the genes wingless and engrailed are expressed in different sets of blastoderm cells, yet in the absence of wingless gene expression, the cellular pattern of engrailed expression is altered. Furthermore, the wingless mutant phenotype is non-autonomous (Morata & Lawrence, 1977) and the deduced structure of the wingless gene product predicts a secreted protein (Rijsewijk, F. et al. 1987; Cabrerra et al. 1987; These findings are consistent with the hypothesis that blastoderm cells communicate with each other to refine the overall pattern by inducing and/or repressing the expression of other segmentation genes within neighboring cells. Likewise, cell surface molecules that were initially discovered due to their presence during later developmental stages, such as the Fasciclin III/ DENS antigen, are first expressed in segmentally repeated fashion immediately following Drosophila gastrulation (Patel et al. 1987; Gauger et al. 1987): such patterns again suggest that the developmental processes that are initiated by segmentation genes may be mediated, in part, by cellular interactions.

We have been using a monoclonal antibody (MAb TN1) to a cell surface epitope in the moth Manduca to study aspects of neurogenesis during embryonic development. In previous reports, the TN-1 MAb was used as a histological tool to visualize initial events in formation of the central, peripheral (Carr & Taghert, 1988a, b) and enteric nervous systems (Copenhaver & Taghert, 1989). By its transient and cell-specific pattern of labelling portions of bundled axons, TNI resembles antibodies generated to grasshopper and Drosophila glycoproteins named fasciclins (Bastiani et al. 1987; Harrelson & Goodman, 1988). We were interested to ask when the TN-1 antigen(s) was first expressed during Manduca embryogenesis and here report that it is initially detected on mesodermal cell surfaces, during gastrulation and in a pair-rule pattern.

A laboratory stock of Manduca was raised on an artificial diet (Bell and Joachim, 1976) and kept at 27°C on a 17L:7D photoperiod. Eggs were collected daily in short (∼1/2 hour) collection periods and placed in an incubator at 25 °C; at this temperature, embryogenesis is typically complete in ∼100h. A timetable of embryonic development has been published by Dorn et al. 1987; more detailed, though less comprehensive timetables are also provided by Carr and Taghert, 1988a,b and Copenhaver and Taghert, 1989.

The methods used were as described in a previous report (Carr and Taghert, 1988a). Briefly, embryos were dissected out from their egg shells and embryonic membranes and fixed in a 2 % paraformaldehyde fixative at room temperature for 1h. Antibody staining was performed at 4°C, with 0·2% Saponin and 0·01 % NaN₃ in phosphate-buffered saline (PBS). Primary antibody incubation (a 1/40 000 dilution of a TNI ascites fluid) was continued for 48 h with agitation and antibodies were localized with a commercially available HRP-based antibody kit (Vector Labs, Burlingame, CA). Embryos were cleared in glycerin, viewed using Nomarski optics, and photographed at up to 650×magnification.

A culture system for Manduca embryos has been previously described that will support as much as 40% of continued development in vitro (Copenhaver and Taghert, 1989). Briefly, eggs were surface-sterilized and dissected to free the embryo from egg shell, embryonic membranes and yolk. The culture medium contained 47·5 % Schneider’s Drosophila medium (Gibco, without glutamate) and 37·5 % Eagle’s basic salts (after Chen and Levi-Montalcini, 1969), ∼9·9% fetal calf serum, 0·1% penicillin-streptomycin (culture grade, Sigma), 0·01 % insulin (Sigma, after Seecoff et al. 1971), and 5% Manduca haemolymph (heat-inactivated, from wandering, 5th instar larvae; see Dominick & Truman, 1983, for details of Manduca larval stages). Cultures were maintained on glass slides within a bed of Sylgard (Corning) at 27 °C in normal atmosphere.

Cell surface expression of the TNI antigen(s) was demon-strated by immunofluorescent staining of living moth embryos. Embryos were dissected from their egg shells and associated membranes as described above, and placed into embryo culture (Copenhaver and Taghert, 1989). TNI ascites fluid was added to the culture medium at a dilution of 1:1000. Embryos were allowed to incubate for 30–60 min in primary antibody, and then washed with fresh culture media for 30 min. Culture medium containing preabsorbed goat antimouse IgG (Sigma Chemical; at a concentration of 2·5 %) was used to visualize the TNI staining; preabsorption was accomplished by exposing the culture media-fluorescent antibody mixture to homogenized moth embryos for ∼6h, followed by centrifugation at 10000revs min⁻¹ for 10min.

Preparations were viewed and photographed with one of two techniques. Embryos that had reached 35 % of development displayed a fluorescent signal bright enough to be photographed directly with standard fluorescent microscopy techniques. Gastrulating embryos ∼15–17% of development) could be viewed in similar fashion, though high-resolution photography required the use of an image-processing system. The image from a Dage/MTI 66 silicon-intensified target video camera was digitized with 8 bits of resolution (256 grey levels) and analyzed with a Trapix image processor (RSI) connected to a Microvax computer. Final photographs represented the average of 16 fluorescent images from which an out-of-focus image was subtracted to correct for background staining. For both imaging techniques, and in order to ensure the viability of the embryos at the time of antibody staining, some embryos were subsequently maintained in culture for a further 3 to 6 h to ascertain that development would proceed according to a variety of morphological markers (N.J. Tublitz and C.M. Bate, per. comm.; Carr and Taghert, 1988a,b).

Manduca embryos complete development in ∼100 h at 25 °C. A general schedule of tissue and organ development has been published (Dorn et al. 1987). We have supplemented this with our own observations, and those of M. Bate and N. Tublitz (unpublished results). A brief listing of major developmental landmarks is provided in Table 1.

The monoclonal antibody TNI was generated using the transverse nerves of the pharate adult moth, Manduca sexta, as the immunogen (Taghert et al. 1983). Immuno-cytochemical staining with this antibody revealed cell-specific patterns of expression in most embryonic tissues (Fig. 1). Within the developing nervous system (Fig. 1A, C & D), there are three salient features that describe this pattern. First, staining was observed in a subset of neurons and, for the greater part of embryo-genesis, was found predominantly along axons that run longitudinally in the rostral-caudal aspect. The second feature is that this axonal staining was regional: individual neurons typically did not display any antigen on their cell bodies or proximal neurites; however, many axons were specifically stained once within longitudinal tracts (Fig. 1D) and once within peripheral nerves (Fig. 1A). Third, within segmental ganglia, some commissural staining of axons was seen, but it was limited to just three of the commissural tracts and was a transient aspect, no longer seen after ∼45% of development. Within the brain, subsets of commissural axons were also labelled to varying extents (Fig. 1C). Certain enteric neurons (Copenhaver & Taghert, 1989) and peripheral neuroendocrine neurons were also labelled by this antibody, but sensory neurons were not (Carr & Taghert, 1988a). This pattern of labelling in the developing nervous system resembles that found with anti-bodies to the fasciclin II glycoprotein in grasshopper embryos (Bastiani et al. 1987; Harrelson & Goodman, 1988) .

Specific staining was not limited to the nervous system, but was seen in all tissue types. Patterned (though transient) labelling was observed of glial cells associated with peripheral nerves (Fig. 1A), of epidermal cells (Fig. IB), of mesodermally-derived cells associated with the gonad (Fig. IE), of mesodermally and ectodermally-derived cells of the gut (Fig. IF), as well as of tracheal cells and glandular cells (not shown). We believe that most of the labeling was due to recognition of cell-surface epitopes as described below (Fig. 6). In all cases, the intensity and pattern of staining changed markedly as a function of development such that, by late embryonic periods, most CNS and epidermal expression was greatly diminished. These patterns of TNI staining were found in each segment from 20 % through 100 % of embryonic development, with only slight segment-specific variation.

In Manduca embryogenesis, the blastoderm (5 –10% of development) is an oblong disc and the pattern of segmentation conforms most to that of long germ band insects. Only a few most-caudal body segments remain to be added in sequential fashion following blastoderm formation. No specific TNI staining was detected at this stage (Fig. 2A). Gastrulation began at ∼11% in the presumptive thoracic regions and proceeded as a wave that ran both anteriorly and posteriorly; the establishment of distinct germ layers was complete by ∼18%. The onset of TNI immunostaining occurred at ∼11 ·5 % of embryogenesis and was roughly coincident with the onset of gastrulation: typically, the TNI antigen(s) appeared earliest in the segment primordium that corresponded to subesophageal segment 3 (S3), where gastrulation was largely complete, and in the segment primordia that corresponded to abdominal segment 1 (Al), where gastrulation was just beginning (Fig. 2B). As the gastral groove moved posteriorly, TNI immunoreactivity followed to eventually create a pair-rule pattern of antigen expression that was most prominent at ∼18% (Fig. 2C); at this stage of development, odd-numbered segmental primordia (e.g. S3, T2, Al, A3, etc.) expressed the TNI antigen while even-numbered segments did not. As development proceeded and organogenesis began, pair-rule expression of the TNI antigen(s) rapidly converted such that, by ~20%, a pattern of similar expression in all segments was evident (Fig. 2D).

The creation of this pair-rule pattern was characterized by two features. The first was a four-segment periodicity that often anticipated the two-segment pattern. For example, cells in segments S3, Al, A5 and A9 often reached their final, high level of TNI staining in advance of the other odd segments (T2, A3 and A7). In some cases (2 of 10 embryos stained at ∼13%), we found that segments Al, A5 and A9 had begun TNI antigen expression prior to any of the intervening segments; in these preparations segments A5 and A9 began their expression prior to gastrulation. The second salient feature of the pair-rule pattern creation was that many even-numbered segments demonstrated slight and very brief levels of expression as adjacent odd-numbered segments first showed staining. Immuno-reactivity in even-numbered segments was briefly seen as the gastral groove first reached their respective levels, though the intensity of immunoreactivity was less than in their odd-numbered neighbors. In this same regard, one particular pattern was seen in approximately one third of the preparations studied: staining in segments A5, A6 & A7 would emerge at ∼ 14 –15 % as a uniform block. As mentioned, this was not a consistent feature and, by 18 %, little if any expression in A6 could be detected. Thus, when TNI immunoreactivity was first detected in an odd-numbered segment, some expression could also be detected in the adjacent, evennumbered segment. As the odd segments increased their levels of immunoreactivity, staining decreased in the even segments, creating the pair-rule pattern by 18% (Fig. 2C).

At all times of pair-rule TNI expression, the posterior boundary of immunoreactivity was clearly delimited, whereas the anterior border was ragged (Fig. 3A). We could not determine whether the ragged anterior boundaries were due simply to fewer mesodermal cells in these regions, or due to intermingled cells that displayed varying degrees of immunoreactivity. The alternating blocks of TN1-positive cells could represent either segmental or parasegmental units (Martinez-Arias & Lawrence, 1985): in this regard, we could not identify tracheal pits in Manduca embryos before ∼20% of development. However, the posterior borders of TNI immunoreactivity were always in register with slight deformations in the ectoderm that indicated future segment boundaries. Hence, pair-rule TNI expression in Manduca observed segmental, and not parasegmental, boundaries.

The pair-rule expression of the TNI antigen(s) during gastrulation was exclusively associated with mesodermal cells (Fig. 3B.). From the earliest period of detection, specific labelling was only observed among the medial-positioned groups of cells that were destined to gastrulate or that had already begun to do so. Following gastrulation and throughout the period during which mesodermal cells migrated laterally, no TNI staining of the underlying ectoderm was ever seen. Even-numbered segments (those that showed little if any TNI expression during gastrulation) did not appear to be delayed in the initiation of gastrulation relative to adjacent odd-numbered segmental primordia. Rather the front of the gastral groove proceeded through prospective segmental primordia irrespective of whether or not TNI epitopes were present.

Pair-rule expression in the mesoderm was in place by the time gastrulation was complete, and persisted through the following period of mesodermal migration. Soon thereafter, the pair-rule pattern of TNI staining changed such that cells within both even and odd segments began to stain at comparable levels (Fig. 2D). When consistent expression finally began in even-numbered segments (at ∼20%), gastrulated cells had migrated to comparable extents in all segments to create anlagen of both the splanchnic and somatic mesoderm. For example, mesodermal cells had invaded the developing limb buds of adjacent segments by the 20% stage, despite having differentially expressed the TNI antigen while migrating into these positions. Like-wise the differentiation of ectodermal derivatives (e.g. formation of limb buds and mouth part appendages) had also proceeded at comparable rates among both even- and odd-numbered segments.

The TNI epitope(s) was localized to embryonic cell surfaces by incubating living embryos (that were dissected free from embryonic membranes and yolk) in culture, first with TNI ascites fluid then with a fluor-escent secondary antibody (Fig. 4). Following observation, the embryos were cultured for a further 3 –6 h, during which time they continued to develop. Such experiments were performed at later stages (∼35 %) when expression was prominent in the developing nervous system and in every segment (Fig. 4A,B), and at early stages (∼16%) when pair-rule expression was effected (Fig. 4C,D). At both developmental stages, we found that the basic elements of the patterns of immu-noreactivity (and as many of their fine details as could be perceived) were also present in the living embryos.

The most striking result of these studies is the immuno-cytochemical demonstration that a molecule associated with the surfaces of embryonic cells in the moth embryo is initially expressed in alternate segments. This represents the first observation of a pair-rule pattern of development in an insect different from Drosophila. DiNardo et al. (1988) have argued that pair-rule mechanisms may represent an evolutionary adaptation for the rapid program of embryonic development that is displayed by Drosophila. The evolution of insect species can be roughly indicated by germ band length of the embryo, with short germ bands indicating relatively older species. Manduca and Drosophila display intermediate and long germ band development, respectively. Hence, the degree to which pattern-forming mechanisms of the pair-rule type can exclusively be equated with late evolutionary adaptations is less certain. Pair-rule patterns of development have more recently been observed in the development of the hindbrain in chick embryos (Lumsden and Keynes,1989) .

The expression of the TNI epitope in alternate segments suggests a corresponding biochemical difference between gastrulating mesodermal cells of adjacent segments. We have considered more trivial explanations for this principal result, namely, (i) that the antibody lacks proper access to cells in ‘off’ segments, or (ii) that the intervening segments are delayed in their development with respect to those segments that are TN1-immunoreactive. The first explanation is unlikely because transient immunoreactivity was routinely observed in many such even-numbered segments just as the gastral groove reached their level: hence, access by the antibody to all mesodermal cells appeared the same at developmental times just preceding the peak of pair-rule expression. The second explanation -that even-numbered segments are uniformly delayed in their developmental rate -also seems implausible because at the stage when TNI immunoreactivity was first seen uniformly in each segment (20%), we could see no difference in the degree to which either splanchnic or somatic mesodermal cells had migrated or coalesced. Likewise, ectodermal derivatives (e.g. the central nervous system, limb buds) did not display any obvious alternation by segment in their developmental rates.

The principle features of TNI expression in the early moth embryo resemble those that describe the expression of certain segmentation genes in Drosophila (Hafen et al. 1984; Akam & Martinez-Arias, 1985; Fjose et al. 1985; Kornberg et al. 1985; Ingham et al. 1985; Carroll & Scott, 1985; DiNardo et al. 1985; Kilcherr et al. 1986; Harding et al. 1986; MacDonald et al. 1986; Martinez-Arias et al. 1987). For example, engrailed RNA expression in Drosophila is first detectable at the conclusion of the blastoderm stage in a four-segment periodicity that anticipates a rapidly developed two-segment periodicity (Weir & Kornberg, 1985; DiNardo et al. 1985). By the onset of gastrulation, en protein expression is nearly uniform in its expression by subsets of cells in each of 14 segments. The transition from a four-segment periodicity to a two-segment periodicity (with slight expression in ‘off’ segments), and finally, to comparable levels of expression in all segments, are features that also describe the initial pattern of TNI immunoreactivity in Manduca.

Two differences between these systems are note-worthy: TNI appears later in Manduca development than does en in Drosophila, typically not before gastrulation has begun. Second, low levels of TNI expression in ‘off’ segmental primordia diminish considerably to create a true pair-rule pattern by the end of gastrulation, before reappearing during organogenesis; in this second detail, TNI expression is more directly comparable to the expression of certain pair-rule genes (Weir and Kornberg, 1985; Ingham et al. 1985). The creation of the pair-rule TNI expression thus resembles the creation of en stripes in that new bands of TNI immunoreactivity are sequentially added. It also re-sembles the creation of fushi tarazu and hairy stripes because there occurs a simultaneous suppression in other (even-numbered segments) regions.

These differences are not surprising given that the TNI antigen is associated with the cell surface, whereas previously described ‘pair-rule patterns’ have been derived from studies of proteins that localize to the nucleus (e.g. en:Desplan et al. 1985; DiNardo et al. 1985). Differences in suspected functions for such differentially localized molecules may account for variations in the spatial and temporal details of their expression. In addition, as mentioned earlier, this comparison considers two animals whose developmental schedules differ in the manner by which they reach the stage of gastrulation. Drosophila belongs to the category of insect embryos that are termed ‘long germ band’, wherein a rapid series of nuclear divisions generates all prospective segmental primordia by the stage of blastoderm cellularization. In ‘short germ band’ embryos, the more posterior segments are sequentially generated following cellularization, from a growth zone that is situated near the head (Anderson, 1972). Manduca displays an intermediate pattern: we find that gastrulation begins at a time when only a few segmental primordia have yet to be clearly established. Another notable difference is that while many stripes of pair-rule activity have sharp anterior borders in Dros-ophila (Lawrence and Johnston, 1989), the pair-rule function described here in Manduca displays ragged anterior borders but sharp posterior borders.

The qualitative correspondence between the spatial patterns that are observed with TNI staining in Manduca and the expression of certain segmentation genes in Drosophila suggests that TNI expression is closely or directly regulated by homologues of the latter in Manduca. A Manduca gene that shares considerable homology with the homeodomain of the Drosophila abdominal-A gene has been cloned (L.M. Riddiford, per. commun.). It is not unreasonable to assume that many such homologues exist and that they may play similar roles in lepidopteran embryonic development; this assumption is supported by genetic analyses of homeotic mutations in Manduca (Booker and Truman, 1989) and in another lepidopteran insect, the silkworm Bombyx mori (Tazima, 1964).

The pattern of TNI immunoreactivity in the moth embryo resembles that seen with anti-fasciclin II anti-bodies in grasshopper in a number of details (Harrelson and Goodman, 1988). The most salient resemblance is the preponderant labelling of longitudinal axonal tracts; many differences are also apparent, however, such that the two patterns are overlapping in many ways, but are not completely identical. The fasciclin II glycoprotein in grasshopper has a mass of ∼ 95 × 10³; Western blotting experiments of Manduca tissue with the TNI antibody have not revealed specifically stained bands (N. Platt, per. commun.). Hence, we have no molecular information at present by which to further implicate the TNI antigen as a potential fasciclin. Regulation of fasciclin expression by segmentation genes has already been surmised on the basis of patterned fasciclin III/DENS expression by epidermal cells at each segment boundary in Drosophila (Patel et al. 1987; Gauger et al. 1987).

TNI immunoreactivity was first detected among meso-dermal cells that are in the process of invaginating during gastrulation. In Drosophila embryos, homeotic genes like Ubx are expressed in mesoderm as well as ectoderm, and often in different parasegmental domains (Beinz and Tremml, 1988; Akam & Martinez-Arias, 1985). These differences in spatial patterns are thought to reflect the independent governance of gene expression by distinct czs-acting elements (Beinz and Tremml, 1988). Genetic and molecular studies in Dros-ophila have indicated that a number of other gene functions and products are specifically required, or are expressed, by mesodermal cells during embryogenesis. These include the twist gene (Thisse et al. 1988), a gene encoding the PS2 antigen (a Drosophila integrin -Bogaert et al. 1987) and a novel homeobox-containing protein, H2.0 (Barad et al. 1988). The early expression of the TNI antigen just at the beginning of gastrulation suggests that its function (if any) is required during the earliest stages in the formation of the distinct germ layers. Its resemblance by spatial pattern of expression to presumed adhesion molecules like fasciclin II, and its cell surface localization during gastrulation, would suggest a role in mesodermal cell motility. However, its initial restriction to alternate segments at this developmental stage suggests a different (or perhaps additional) role: that of signalling segmental identity among migrating cells, or of preventing the mixing of cells that have different segmental origins during migration.

We are grateful to Philip Copenhaver, Ian Duncan, Martha O’Brien and Joshua Sanes for their comments on an earlier draft of this manuscript. Thanks also to Michael Bate, Nathan Tublitz and Lynn Riddiford for communicating unpublished results, to Megan Morgan and Gerry Fischbach for help with the image processing and to Jim Voyvodic for writing the analysis program. This work was supported by an NSF predoctoral fellowship to J.N.C. and a grant from the NIH (NS21749) to P.H.T.