Roles of wingless in patterning the larval epidermis of Drosophila

The larval epidermis of insects secretes a highly patterned array of cuticular structures, revealing an underlying pattern of differences in the epidermal cells within each segment (Wigglesworth, 1940; 1973). For example, the ventral surface of the Drosophila larva has a stereotyped pattern of hooks, or denticles, which occupy the anterior portion of each abdominal segment (see Fig. 1A). Each row of denticles has a slightly different character; this suggests a precise array of positional values in the cells that secrete these structures. Positional values along the anteroposterior axis of the Drosophila embryo are assigned through a cascade of gene activities that subdivide the embryo into increasingly smaller units (reviewed in Akam, 1987; Ingham, 1988). Maternally introduced gene products control expression of the gap genes, which then direct the periodic expression pattern of the pairrule genes, dividing the embryo into segmental units. Pair-rule genes activate the segment polarity genes, wingless (wg) and engrailed (en), in adjacent stripes of cells and these cells interact with each other and with other cells within the segment in order to generate the wild-type epidermal pattern. These two genes are thought to be of central importance because their mutant phenotypes represent the most severe of the segment polarity class (Nüsslein-Volhard and Wieschaus, 1980; and see Fig. 1B and 1C) and because the mutant phenotypes of other segment polarity genes can be explained by inappropriate activity of wg and/or en (Martinez Arias et al. 1988; DiNardo et al. 1988; Hooper and Scott, 1989; Hidalgo and Ingham, 1990). Thus, transcriptional states as defined by pair-rule gene expression are translated into cellular identities from which all the positional values within the segment will be generated.

How these positional values are generated has been a matter of some debate. Two main ways of thinking about this process have been reviewed in Martinez Arias (1989). The gradient model proposes that cells respond to a graded distribution of a morphogenic substance and acquire positional identity based on concentration values at specific points along the gradient (Lawrence, 1966b; Lawrence et al. 1972). The cell interactions model proposes that positional values are generated sequentially by local cell– cell interactions starting from an initial specification event (Martinez Arias et al. 1988; Ingham, 1988). Here we present evidence that both mechanisms are important in specifying intrasegmental pattern. Early in development, wg affects patterning decisions by local interactions with en-expressing cells. Later in development, wg plays a different role: some elements of the wild-type pattern appear to be specified in response to a graded distribution of wg protein.

Segment polarity is thought to involve cell–cell interactions because in embryos mutant for certain segment polarity genes, expression of other segment polarity genes in adjacent cells is affected. For instance, in the absence of functional wg product, en expression is lost early during embryogenesis (DiNardo et al. 1988; this work). Conversely, wg expression is lost in embryos that are deficient for en product (see below). This indicates that some interaction between these stripes of cells leads to stable expression of the two gene products early in embryogenesis. The wg gene encodes a secreted protein (van den Heuvel et al. 1989; Gonzalez et al. 1991) that is expressed in a stripe of cells in each segment. It is transported to adjacent cells within the segment, including cells in a stripe posterior to the wg stripe, where the gene en is expressed (van den Heuvel et al. 1989; Gonzalez et al. 1991). This suggests that wg may have a direct effect on adjacent cells, en is a DNA-binding protein (Despean et al. 1985) that is localized to the nucleus (DiNardo et al, 1985) and thought to be a transcription factor. Consequently, its influence on other cells in the segment is probably mediated indirectly through the genes that it activates.

Here we explore the nature of the interaction between wg and en and assess the relative contributions of the two gene products in the patterning of the larval epidermis. We describe four distinct phases of wg function in epidermal cells; some dependent on en activity but some independent of it.

wg^IL114 is a temperature-sensitive lethal allele generated by EMS mutagenesis (Nüsslein-Voihard et al. 1984). At 25°C the mutant phenotype of wg^IL114 homozygotesis indistinguishable from that of wg null mutants, such as wg^CX4 (Baker, 1987). Balanced stocks were outcrossed to Canton-S wild-type stocks to remove the CyO balancer chromosome and homozygotes were generated by backcrossing. As no difference in homozygous phenotypes were observed, balanced stocks were used routinely in the experiments presented here. Embryos mutant for engrailed were generated by crossing en^IO/CyO flies to Df(2R) en^B/CyO (Eberlein and Russell, 1983). En^IO1Df(2R)en^B embryos show no labelling with a monoclonal anti-en antibody directed against the homeo-domain epitope.

Eggs were collected from agar/apple juice plates (Wieschaus and Nüsslein-Volhard, 1986) and dechorionated with bleach. In all cases, embryos were staged from gastrulation, 3h after egg laying (AEL).

Anti-engrailed staining was as described by DiNardo et al. (1985). Anti-wingless staining was performed using a rat polyclonal antiserum (Gonzalez et al. 1991). Embryos were fixed in 4% formaldehyde in BBS (Leiss et al. 1988) for 50 min and processed following the protocol of Leiss et al. (1988).

Embryos were aged at the appropriate temperature for the equivalent of 24 h at 25°C, dechorionated in bleach and dissected free of the vitelline membrane in Drosophila Ringer’s solution (Roberts, 1986). They were then mounted in Hoyer’s medium mixed 1:1 with lactic acid as described by Wieschaus and Nüsslein-Voihard (1986). Cuticles were observed with phase-contrast optics on a Zeiss Axiophot.

Embryos collected from a wg^IL114/CyO stock were staged by selecting gastrulae within a 10min interval. The stage chosen is visible for less than 5 min at 25°C. Therefore embryos are synchronized to within 15 min of development. Shifts were performed by transferring synchronized embryos to Drosophila Ringer’s solution at the appropriate temperature. AH low temperature manipulations were performed in a constant temperature room that fluctuates no more than 1°C. Ages of embryos cultured at low temperature were adjusted to correspond to developmental stage at 25°C. This was accomplished by timing the appearance of characteristic developmental stages in embryos cultured at 17.5°C and using those times to calculate a conversion factor for adjustment to correspond to times at 25°C. Mutant embryos were identified on the basis of consistent phenotypes observed in one-quarter of the experimental embryos. At least 20 mutant embryos were examined for each time point described.

Response of wg^1L114 mutant protein was measured directly by staining with anti-wg antiserum. At 25°C the mutant protein is retained in the secretory apparatus (see Gonzalez et al. 1991, for a more complete description). Shifting down to 17.5°C restores a normal distribution of vesicles over several cell diameters within 30min. We do not know if the genetic lesion only involves a temperature–sensitive defect in secretion, or if the mutant protein is defective in activity as well. In either case, the response of en expression to removal or restoration of wg via temperature shifts can be observed within 1 h following a shift.

The expression of both wg and en is activated in the cellular blastoderm just before gastrulation (3h of development), in adjacent stripes of cells in each segment. As the germ band extends, en protein is detected in the nuclei of a discontinuous and jagged line of cells that slowly resolves into a coherent stripe (see Fig. 2A,C). In embryos mutant for wg function (Fig. 2B, D), en expression is activated in a segmentaily repeated pattern at the appropriate time but it does not stabilize and the en-expressing cells never become coherent stripes as in wild-type embryos (Fig. 2A,C). The en staining remains spotty and discontinuous and by 4h of development, most en expression in the epidermis of thoracic and abdominal segments has disappeared (Fig. 2D). Therefore functional wg product is not only required for the maintenance of en expression but also appears essential for the proper establishment of en stripes.

en function is required for maintenance of wg expression, but this reciprocal maintenance requirement has a very different timing, en mutants express wg properly until about 5h of development (Fig. 3A,B). At this time there is a transition in the wild-type pattern of wg expression. The stripes of protein fade and are replaced by a second pattern of wg expression with staining in a ventral wedge and in a dorsal patch (Fig. 3A,C). In en mutants, wg expression begins to decay just before the transition occurs and the second phase of wg expression is altered, with weaker staining in the ventral portion of alternate segments (Fig. 3D). This pair-rule pattern of decay is rapidly superseded by loss of wg in all cells of the ventral epidermis, leaving dorsal patches of wg expression that eventually decay after germ band retraction is complete. Notice that en does not appear to influence the initial expression of wg, as wg does en.

We also find that wg is itself required for the second phase of wg expression. In wg mutants that make protein which is detectable with our antibody, such as the temperature-sensitive wg mutant, wg^IL114, we observe that the nonfunctional protein is expressed correctly until 5.5 h when it diminishes on schedule but the second pattern of expression never appears. This is not simply due to the loss of en expression since mutants lacking en continue to express wg after 5.5 h of development (Fig. 3D).

We conclude that the second phase of wg expression depends on the initial activity of functional wg protein and requires en activity for its maintenance. The delay in wg response to the absence of en expression allows us to assess the effects of wg on en expression before this point. Manipulating wg expression during this time will not be confounded by a secondary effect on wg expression from the early loss of en.

We have used a temperature–sensitive allele of wg to assess the requirements for wg at different times during development. The temperature–sensitive wg mutant, wg^IL114, makes protein that is defective in secretion (Gonzalez et al. 1991) and function at 25 °C, but shows a wild-type protein distribution and is indistinguishable from wild-type with respect to en expression (this work) and cuticle phenotype (Baker, 1988) when raised at 17.5°C. We have made use of this allele to remove or restore functional wg product at various times during development.

Shifts in temperature up to 25°C at or before gastrulation result in a wg null mutant phenotype. Shifts down to 17.5°C at these times result in a wild-type phenotype. Therefore, no requirement for wg in cuticle phenotype (Baker, 1988) or en expression is detected before 3h of development. Shifts after 3h alter the length of time that an embryo has been exposed to functional wg protein. For instance, a wg^IL114 mutant embryo shifted up at 4h has had wild-type wg expression for one hour during the time that wg is required, wg function was then removed by shifting to the restrictive temperature and maintaining that temperature until the embryo had reached the desired age. The effect of these shifts on en expression was assessed by examining embryos just before and just after germ band shortening.

Such temperature shifts with wg^IL114 embryos have allowed us to define the time during which wg is acting to maintain en expression and have revealed an unexpected property of this maintenance – that different populations of cells become competent to maintain en expression at different times. Upshifts at times between 4 and 7h of development result in progressively more cells maintaining en expression (Fig. 4). The location and arrangement of these cells changes over time. Because there are reproducible positional differences in the pattern of cells that maintain en expression at any given time, counting the number of en-expressing cells is not an accurate representation of the results and we prefer to present the data in a descriptive manner.

Upshifts removing wg function at 4 h of development result in the decay of en within every stripe except for a lateral cluster of cells (Fig. 4A,B), roughly adjacent to that portion of the wg stripe which is absent in the second phase of wg expression (Fig. 3C). Thus these lateral cells are ‘specified’ to maintain their en expression earlier than are other cells in the epidermal stripe. Note that similar numbers of en-expressing cells are present in embryos fixed at 7 h and at 10 h (compare Fig. 4A and B). Cell death in wg mutant embryos occurs within this interval (unpublished observations) and therefore does not contribute to the pattern that we observe.

Upshifts removing wg function at 5h result in a greater expanse of the dorsolateral en stripe being maintained (Fig. 4D), while en expression decays predominantly in a discrete region of the ventral midline (Fig. 4C). Upshifts at 6h result in a wild-type pattern of en expression throughout the ventral epidermis, with some gaps in the dorsal-most portion of the stripes (compare Fig. 4E and F). In upshifts later than this, the dorsal patterning of en is virtually wildtype. This indicates that the requirement for wg activity to maintain en expression ends at about 6h of development.

We conclude that longer exposure to wg expression causes more cells to maintain their en expression, and that discrete groups of cells along the dorsoventral axis are specified to maintain en at different times, with lateral cells being specified first and ventral cells last. We believe that this also reflects the timing of cell identity specification because pattern elements in the larval cuticles secreted by these shifted embryos are acquired in this same order.

Fig. 5A– C shows cuticles secreted by embryos shifted up early in development. Upshifts at 3 h result in a typical wg null mutant phenotype, with no pattern elements on the lateral surface, reduced patterning of the dorsal surface and disrupted ventral patterning (Fig. 5A). Between 3 and 4h of development, pattern elements on the lateral surface are specified (Fig. 5B). This embryo still looks abnormal because wg has a later function in dorsal closure. In rare mutant embryos where dorsal closure has proceeded normally, we find that dorsal and lateral pattern elements are specified by4.5h(Fig. 5C).

The dorsoventral difference in timing of pattern acquisition is shown even more dramatically in downshifts (Fig. 5D– F). In downshifts, embryos are mutant for wg until the time of the shift, when wg function is restored by culturing at the permissive temperature. All aspects of the larval pattern can still be rescued when embryos are shifted down at or before 4 h of development (Fig. 5D,E). However, embryos shifted down at 4.5 h show defective dorsal and lateral patterning, while the ventral pattern is not greatly affected. Thus ventral pattern elements can still be rescued by restoring wg function at this time, but dorsal and lateral pattern elements cannot.

These cuticle patterns (Fig. 5) correlate with the effects on en expression in downshifted embryos. Downshifts at 4 h of development restore en expression in a pattern resembling the wild type. Such embryos fixed at 7h display en stripes of the correct shape, although the dorsal portion of the stripe is thin and discontinuous (Fig. 6A,B). This indicates that the activity of wg prior to 4h of development is not generally required for maintenance of en expression in the epidermis. In 4h old wg mutant embryos, most epidermal en expression has decayed (Fig. 2D). Therefore, en expression in downshifted embryos largely represents reactivated expression. This suggests a direct role for wingless at 4 h in initiating engrailed expression or in promoting it front levels below those that we can detect with our anti-en antibody.

In downshifts at 5h, some en expression can still be reactivated, but the dorsolateral part of the stripe is not maintained. Restoring wg function at this time promotes a transient expression of en in the dorsal portion of the stripe, while persistent en expression is established only in cells of the ventral wedge (compare Fig. 6C and D). Downshifts at or later than 6h of development result in virtually no restoration of en expression (Fig. 6E,F). Restoration of en expression appears to occur in a pattern roughly complementary to the pattern of stabilized expression acquired in upshifts; however, the correlation does not seem to be exact. For instance, a patch of cells just dorsal to the ventral wedge is stabilized in both upshifts and downshifts at 4.5 h (not shown).

An earlier study reported a continuous requirement for wg throughout embryogenesis in the patterning of the ventral epidermis (Baker, 1988). While our observations confirm such a requirement, we find that there are at least three qualitatively different functions of wingless for epidermal pattern formation. The first two are contingent on engrailed activity, the third is not.

Embryos homozygous for wg null afieles primarily secrete denticles typical of the fifth row in a wild-type abdominal denticle belt. These row 5-type denticles are arranged in a metamerically repeating pattern of anterior– posterior polarity reversal (Fig. 7A). In the absence of both wg and en, this metamerism is lost and the epidermis secretes denticles in a uniform field pointing toward the ventral midline (Fig. 7B); the only trace of metamerism left in these embryos is a repeating pattern of rosettes of denticles pointing outwards along the ventral midline. This suggests that a brief exposure of the epidermal precursors to en activity, in the wg single mutant, provides some degree of information reflected in the final cuticular pattern. Patterning of the ventral epidermis beyond this default state hinges on the establishment and maintenance of proper en expression. This requires wg expression between 3.0 and 6h of development.

During the time that wg is acting to maintain en expression, patterning events can be detected in the ventral denticle belts secreted by shifted embryos. Downshifts indicate that, although the ventral pattern of en expression can be rescued (Fig. 6D), there is a point when the ventral pattern of the cuticle cannot (Fig. 8F– J). Shifts down to the permissive temperature between 5 and 5.5 h restore en expression in the ventral portion of the stripe, but cuticles secreted by embryos shifted at these times resemble those secreted by embryos mutant for en (compare Fig. 8H and I with Fig. 7D). This suggests that after 5h, restored en expression is irrelevant to patterning decisions in the epidermis.

The cuticle patterns of mutant embryos shifted up during this interval reveal that a progressive transition in denticle polarity is taking place, changing from symmetry to asymmetry within the segment (Fig. 8A– E). The wg null mutant phenotype shows symmetrical groupings of a single type of denticle (Fig. 8A), this gradually changes into an asymmetric pattern of different denticle types (Fig. 8E). Upshifts at 5h result in denticle types being specified, other than the default null phenotype denticle, but the arrangement is still fairly symmetrical. Asymmetry becomes apparent in upshifts at 5.5h (Fig. 8D), and is complete by 6h (Fig. 8E). At this point, all denticles of the wild-type denticle belt are present and arranged correctly in the anterior portion of the segment, but the posterior portion is occupied by denticles of indeterminate type. We find it striking that this transition from symmetry to asymmetry coincides precisely with the time during which wg is acting to maintain en expression in the ventral epidermis. We propose that the emergence of polarity between 5 and 6 h results from the interaction of wg- and en-expressing cells. At this stage the two gene functions appear tightly linked; wg expression begins to decay in the absence of engrailed at about this time.

After 6h of development, wg no longer affects en expression. The main role of wg now is to specify naked cuticle in the posterior portion of the segment, replacing denticles of indeterminate type (see Fig. 8E, 9A). Upshifts after 6h show progressively more naked cuticle replacing the rows of indeterminate denticles in a posterior-to-anterior direction (Fig. 9A– C). Thus, after 6h of development, wg activity can be measured directly by how much naked cuticle is specified in the ventral epidermis. This direct relationship is maintained until the end of germ band retraction. Throughout this period of the extended germ band stage, we also detect an apparently graded distribution of the wg protein extending anterior to those cells expressing the wg gene (Fig. 9E– G). Whereas wg protein is evenly distributed on either side of the cells expressing it at earlier times (Fig. 3A), at later times we observe a sharp cut-off just posterior to the line of cells expressing it and we detect wg staining several cell diameters anterior to this line of cells. We suspect that specification of naked cuticle may be a direct response to wg protein transported to these cells that lie at a distance from the wg-expressing cells.

The progressive phase of wg function ends at about9.5 h. After this point only cells of the ventral midline exhibit a requirement for wg. Upshifts between 9.5 and 15 h result in a virtually wild-type cuticle pattern except for persistence of indeterminate denticles in the naked cuticle zone at the ventral midline (Fig. 9C).

In interpreting the cuticle phenotypes of upshifted homozygous wg^IL114 embryos, we cannot distinguish between dose-dependence due to concentration of wg protein or to length of time exposed to functional wg. At least some aspects of the wg mutant phenotype probably result from the actual concentration of wg protein, because absence of en product reveals an underlying dosage effect of wg. While heterozygotes for wg null alleles are perfectly viable and show no detectable difference from wild-type flies, heterozygosity for wg in embryos that are completely deficient for engrailed (Fig. 7C) renders the phenotype far more severe than that of homozygous en embryos (Fig. 7D). The ventral cuticle pattern appears to be intermediate between that of embryos mutant for wg and that of wg en mutants, while the dorsal surface is not so severely affected. Temperature shifts of embryos heterozygous for vvg^ILI14 and homozygous for en show that the time during which the full dose of wg is required coincides with the time of the wg–en interactive event (data not shown). The mutant phenotype of en also shows some temperature sensitivity (compare Fig. 8H, which is indistinguishable from en^IO/Df(2R)en^B cultured at 17.5°C, and Fig. 1C, which is en^IO/Df(2R)en^B cultured at 25°C), but at each temperature the phenotype of the doubly mutant strain is more severe than the en single mutant.

These experiments also revealed a temperature sensitivity in the function of wild-type wg protein. In a background sensitized by the loss of en, a greater activity of the wild-type dose of wg can be detected at 17.5°C. This phenomenon was noted in the course of performing control experiments with a null allele of wg, wg^cx2. We wonder if in fact this effect is responsible for the slight temperature sensitivity of some segment polarity mutations, such as en (see above) and are currently investigating this possibility.

The diversity of cell identity in the Drosophila segment is generated by the action of segment polarity genes (Nüsslein-Volhard and Wieschaus, 1980). Mutations in these genes generate deletions and pattern duplications in the cuticle, partly caused by defective expression of other genes of this class, such that the mutant phenotypes for each gene do not accurately reflect the absence of that gene product alone (Martinez Arias et al. 1988; DiNardo et al. 1988). Thus it is difficult to assess the relative contribution of each member in this complex interplay of gene products that culminates in the wild-type cuticle pattern. To attempt to dissect this network of interactions, we have focussed on two gene products critical in the process, those encoded by wingless and engrailed. The correct expression of the segment polarity genes wg and en is an essential element in the postblastoderm patterning process. Their expression in the epidermis is activated in a periodic fashion by pair-rule gene activity (DiNardo and O’Farrell, 1987; Ingham et al. 1988), but their gene products then act to specify cell identities beyond their transcriptional domains. It has been postulated that the wg/en domain is an essential reference for generating other positional values within the segment, either as the source of a gradient (Lawrence et al. 1987) or as a key element in promoting local cell– cell interactions (Ingham, 1988). Our observations suggest that this combined function can account for some cell identities but that there appear to be specification functions for the two gene products individually, separate in space and time. In particular, we show that wg performs different roles at different times in development.

The different requirements for wingless that we report here indicate three important phases of the epidermal patterning process: first, the establishment of cell states; second, interactions between cell states to generate new cell states and pattern; and third, elaborations of the basic pattern.

We have noted an early role for wg not only in maintaining en expression, but also in establishing the pattern of en expression correctly. A role for wg in the initiation of the odd numbered en stripes had been inferred from the behaviour of en-lacZ promoter fusions (Kassis, 1990), but our observations suggest that wg may contribute to establishing correct expression of all thoracic and abdominal en stripes. We propose that this may be an instructive function of wg since restoration of wg activity in downshifts, at times when most en epidermal expression has decayed, will reactivate full expression of en in a fairly normal pattern (Fig. 6A). This observation also demonstrates an underlying polarity, dictating on which side of the wg stripe en expression should appear. This implicit polarity does not require wg function continuously, since engrailed expression does not come on at random with respect to the wingless stripe when wg function is restored.

The maintenance function of wg is progressive, with discrete groups of cells along the dorsoventral axis acquiring stable en expression at different times:

Dorsal 3----- 4.5 h AEL

Lateral 3 —4

Ventral 5---- 6

wg activity gradually stabilizes en expression in these adjacent cells, such that by 6h they no longer require wg function for continued expression of en. This transition in the regulation of en expression, from wg- dependent to wg-independent, has also been described by Heemskerk et al. (1991). The dorsoventral differences that we observe imply that cells switch over to a different mode of en regulation at different times; dorsoventral differences in en promoter activity have been observed by DiNardo et al. (1988).

We also note that the specification of the cuticle pattern reflects a similar schedule; lateral and dorsal pattern elements are specified first and ventral pattern elements later. Thus cell identities appear to be assigned in a distinct temporal order. This difference may relate to the fact that profiferation is completed earlier dorsally than ventrally (Campos Ortega and Hartenstein, 1985) and that the commitment of some ventral ectodermal cells to neurogenesis is an important factor in the different timing of cell specification.

The ventral cuticle shows another kind of progressive patterning event during the time that wg activity is stabilizing en expression: the generation of asymmetry within the segment. This process culminates at 6h with the specification of positional values in the anterior portion of the segment, as reflected by the ability of the cells to produce the correct denticle type in the correct orientation. We do not know to what extent these patterning events are performed by wg function alone or by wg in conjunction with en activity. It seems likely, however, that there is a critical point prior to 5h of development when both genes must be expressed together for subsequent generation of the wild-type pattern. Although stable en expression can be restored to the ventral epidermis (along with wild-type wg function) after this point, the embryo can only produce pattern elements typical of an en mutant. The timing of these patterning events coincides with the time that the wg- and en-expressing cells are known to interact: maintenance of en expression requires wg function between 4 and 6h; maintenance of wg expression requires en function sometime before 5.5 h. Thus the process of specifying these denticle identities may be driven by an interactive event between these two lines of cells.

All functions of wg after 6h depend on en only in the sense that they are contingent on these early events involving local cell interactions. After this time, wg acts to specify naked cuticle in the posterior portion of the segment. This function appears to depend on the acquisition of wingless product by cells lying at a distance from the stripe of wg-expressing cells. Longer times at the permissive temperature result in cells more distant from the source being instructed to make naked cuticle. It is likely that either length of exposure to wg or accumulation of the product to higher levels over time promotes naked cuticle identity. This view is supported by the observation that string mutants, which make segments consisting of fewer cells, secrete naked cuticle. Embryos doubly mutant for string and wg, however, secrete row5-type denticles (unpublished observations). Therefore, levels of wg sufficient to specify naked cuticle identity reach all cells within the much smaller segments of string mutants.

It is relevant here to ask precisely how many cells wg protein would have to traverse if it is in fact responsible for specifying all naked cuticle in the wild-type ventral epidermis. Although the denticle belt occupies roughly one-third of the length of abdominal segments 2– 7 in a wild-type larva, this does not reflect the proportion of epidermal cells that contribute to it. Late in embryogenesis, the shape of epidermal cells changes dramatically such that the cell diameters underlying the denticle belt region are shorter in the anteroposterior axis and those underlying the naked region tend to be longer (D. Currie and M. Bate, personal communication). On average, between 5 and 6 cell lengths underlie the naked, interbelt region with a similar number of cell lengths lying beneath the denticle belt portion. In late stage embryos stained with wg antibody, we detect wg protein 3 to 4 cell diameters anterior to the line of wg- expressing cells. Thus the number of cell diameters over which wg may be distributed (including those cells expressing it) is almost sufficient to account for the number of cells expected to secrete naked cuticle.

Furthermore, a role for wg protein in specifying naked cuticle is consistent with the phenotypes of several segment polarity mutants in which wg is incorrectly expressed, naked mutant embryos show an expansion of the en expression domain which induces ectopic stripes of wg expression (Martinez Arias et al. 1988). These additional sources of wg protein at about 6h may directly confer naked cuticle identity to cells in the region that should secrete denticles, thereby giving rise to the observed mutant phenotype. This implies that ectopic wg activity can override earlier specification of cells to form denticles typical of anterior portions of the segment. A direct role in specifying naked cuticle could also account for the cuticle pattern secreted by en mutants. The late expression of wg in alternate segments of the extended germ band (Fig. 3D) may be sufficient to specify naked cuticle identity, resulting in a pair-rule type pattern of naked patches (Fig. 1C and 7D). Although patched mutants show an expanded wg domain with ectopic en stripes (the converse of the nkd phenotype, Martinez Arias et al. 1988) the mutant embryos produce a cuticle with patches of denticles. Examination of the wg protein distribution reveals a sharp cut-off at the anterior limit of the expanded domain similar to the posterior cut-off seen in the wild-type (see Fig. 9D) and in ptc mutants (unpublished observations). We surmise that the ectopic en stripe restricts the spread of wg protein - in this case its anterior movement – as the normal en stripe appears to do for posterior movement in late-stage wild-type embryos. The denticles seen in ptc mutants are localized to the region from which wg is excluded, between the endogenous and the ectopic en stripes.

We conclude that there are two distinct mechanisms by which positional information is specified at different times and in different locations within the epidermis of the Drosophila embryo. Both mechanisms have been predicted by models formulated to explain segment polarity mutant phenotypes. Local cell– cell interactions are important early in development for establishment and maintenance of appropriate wg and en expression patterns, for specifying positional values in the dorsal and lateral epidermis, for generating anteroposterior asymmetry in the ventral segment and for specifying initially plastic values in the anterior portion of the ventral segment. Contingent on this event, a graded distribution of wg protein is established in the cells anterior to the wg-expressing cells in the ventral epidermis; this correlates with progressive specification of naked cuticle identity in these cells.

Our observations suggest that cells receiving a threshold level of wg product become specified to secrete naked cuticle. This process is reminiscent of the morphogenetic gradient postulated to specify positional values within the field of an insect segment (Locke, 1959; Lawrence, 1966b). Computer simulation reveals that a gradient of freely diffusing material does not fit with experimental results on pattern changes in larval epidermis following surgical manipulation (Lawrence et al. 1972). Our observations that wg may not diffuse freely through extracellular spaces, but may be actively transported by cells (Gonzalez et al. 1991), thereby creating ‘resistance’ to the flow of putative morphogen, is consistent with the computer model that does fit experimental results in Rhodnius (model 2 of Lawrence et al. 1972).

Postulating two temporally distinct mechanisms for the specification of cell identity allows us to reinterpret some aspects of segment polarity gene action. It has been proposed that wg acts as a signal, based on its homology to the vertebrate oncogene, int-l (Rijsewijk et al. 1987), and that reception of the wg signal is mediated by at least one protein encoded by another segment polarity gene, armadillo (Wieschaus and Riggleman, 1987; Riggleman et al. 1990; Peifer et al. 1991). Earlier work (Wieschaus and Riggleman, 1987) had indicated that arm is cell autonomous; mutant clones detected in the naked cuticle region secrete denticles. Two additional features of this clonal analysis were notable – most of the clones produced denticles of the correct polarity and clones were never detected in the posterior portion of the naked cuticle, overlying the en-expressing cells. Our observations suggest that these features are consistent with an arm mutant effect in clones on the late functioning of wg only. Due to the perdurance of arm product, by the time clones are truly deficient in arm, the wg-en interactive phase has already occurred and (1) polarity has been initiated in the segment, (2) en expression is stably maintained, and (3) wg is no longer found at high levels in the en- expressing cells, but is moving to more anterior cells. One would then predict that arm deficient clones in these anterior cells could not receive wg signal and would produce indeterminate denticles – oriented correctly. In addition, since reception of wg in en cells is probably dispensable at this point, arm-deficient clones in this posterior region may not express any mutant phenotype.

Peifer and Wieschaus (1990) have demonstrated that the armadillo protein, a known component of the wg signal reception pathway, is homologous to plako-globin, identified in vertebrates as a component of desmosomes that may be important in anchoring microfilaments to the cell membrane. This raises the intriguing possibility that wg-induced restructuring of the cytoskeleton might provide a simple and direct mechanism for suppressing denticle formation and thereby specifying that naked cuticle be secreted. Cuticular protrusions of the insect integument are formed by outpocketings of the epidermal cells, produced and stabilized by cytoskeletal elements (Wigglesworth, 1933; Lawrence, 1966a; Overton, 1966). Nübler-Jung (1987) has demonstrated that colchicine treatment alters denticle formation in the larval epidermis of Dysdercus. We observe that the wg mediated response is facilitated at 17.5°C versus 25°C, which may relate to the fact that microtubules tend to be less stable at lower temperatur.es (reviewed in Inoue and Ritter, 1975). Thus a cellular response to the wg signal may involve destabilizing cytoskeletal elements that support denticle outpocketings. We are currently investigating this possibility.

We are grateful to H. Skaer, M. Bate, and M. Peifer for helpful comments on this manuscript. We also wish to thank T. Kornberg and M. Wilcox for the anti-en antibody. This work was supported by the National Institutes of Health (A.B.) and a Wellcome Trust Senior Fellowship in the Biomedical Sciences (A.M.A.).