By a complex and little understood mechanism, segment polarity genes control patterning in each segment of the Drosophila embryo. During this process, cell to cell communication plays a pivotal role and is under direct control of the products of segment polarity genes. Many of the cloned segment polarity genes have been found to be highly conserved in evolution, providing a model system for cellular interactions in other organisms. In Drosophila, two of these genes, engrailed and wingless, are expressed on either side of the parasegment border. wingless encodes a secreted molecule and engrailed a nuclear protein with a homeobox. Maintenance of engrailed expression is dependent on wingless and vice versa. To investigate the role of other segment polarity genes in the mutual control between these two genes, we have examined wingless and engrailed protein distribution in embryos mutant for each of the segment polarity genes.
In embryos mutant for armadillo, dishevelled and porcupine, the changes in engrailed expression are identical to those in wingless mutant embryos, suggesting that their gene products act in the wingless pathway. In embryos mutant for hedgehog, fused, cubitus inter- ruptus Dominant and gooseberry, expression of engrailed is affected to varying degrees. However wingless expression in the latter group decays in a similar way earlier than engrailed expression, indicating that these gene products might function in the maintenance of wingless expression. Using double mutant embryos, epistatic relationships between some segment polarity genes have been established. We present a model showing a current view of segment polarity gene interactions.
The process of segmentation in Drosophila melanogaster embryos is coordinated by a cascade of genes dividing the embryo into 15 segments. Phenotypically three classes of zygotic segmentation genes can be defined (Nüsslein- Volhard and Wieschaus, 1980): the gap, pair rule and segment polarity genes. Of these three groups, the segment polarity genes are the last to act and are thought to define internal organization within each segment. The onset of expression of the zygotic segment polarity genes coincides temporally with cellularization of the embryo (for reviews see Ingham, 1988; Hooper and Scott, 1992). Cell-cell interactions and intracellular signal transduction are presumably important for the coordination of gene expression. Some of the cloned segment polarity genes encode molecules that would appear to be involved in signalling pathways. The patched (ptc) and hedgehog (hh) genes encode putative transmembrane proteins (Nakano et al., 1989; Hooper and Scott, 1989; Lee et al., 1992; Mohler and Vani, 1992;Tabata et al., 1992). Serine-threonine kinases are encoded by the genes for zeste white-3 (zw-3) (Siegfried et al., 1990; Bourouis et al., 1990) and fused (fu) (Preat et al., 1990), while wingless (wg) encodes a secreted molecule (Rijsewijk et al., 1987). Since many of the cloned segment polarity genes have been shown to be highly conserved in evolution, the mechanism by which they control pattern may have implications for patterning in other animals. It is of particular importance to understand the way segment polarity genes interact with each other in Drosophila because of its unique accessibility for gene interaction studies.
After the initial activation of some segment polarity genes by the pair rule genes (Howard and Ingham, 1986; DiNardo and O’Farrell, 1987; Ingham et al., 1988), expression of the segment polarity genes becomes interdependent. Loss of function of one gene causes misexpression or loss of expression of others. The best known example of such regulation is the mutual dependence between wg and engrailed (en;DiNardo et al., 1988; Martinez-Arias et al., 1988). wg is expressed in the cells just anterior to the parasegment border (Baker, 1987; van den Heuvel et al., and en is found expressed immediately next to the cells expressing wg, in all cells of the posterior compartment (Ingham et al., 1985; Kornberg et al., 1985; DiNardo et al., 1985). wg protein can be found outside the cells producing it and occasionally in neighbouring cells, including those expressing en (van den Heuvel et al., 1989; González et al., 1991). The maintenance of en by wg therefore might be a direct effect of the wg protein travelling between these cells. The expression of wg in the cells just anterior to the parasegment border is in turn dependent on en function (Martinez-Arias et al., 1988; Bejsovec and Martinez Arias, 1991). en encodes a nuclear homeobox protein that acts as a transcription factor (Jaynes and O’Farrell, 1988). Maintenance of wg expression in adjacent cells is therefore likely to depend on an extracellular signalling pathway originating from the en cell. It has been postulated that an interaction between hh and ptc is responsible for this regulation (Ingham et al., 1991).
Other segment polarity genes are likely to be required to mediate the maintenance of wg and en. To investigate possible functions of these genes we have surveyed the expression patterns of the wg and en proteins in all known segment polarity mutants and in some double mutant combinations. We present these data here, in the context of other studies on gene expression in segment polarity mutants.
RESULTS AND CONCLUSIONS
To examine the cross-regulation between the segment polarity genes, we chose to investigate the protein expression patterns of the wingless (wg) and engrailed (en) proteins. The maintenance of wg and en is of crucial importance for subsequent development, which is evident from the strong pattern aberrations in mutants of either gene. Based on our results, we divide the known segment polarity genes in three groups. (A) Genes that seem to be involved in wg signalling. In mutant embryos, en expression disappears before wg expression. (B) Genes that seem to be involved in wg regulation. In mutant embryos, wg expression disappears before en expression, and (C) genes that when mutant result in misexpression of wg and en. Some genes in this latter group have been shown to be involved in en or wg suppression (Ingham et al., 1991; Siegfried et al., 1992).
In most cases, we analyzed several mutant alleles, including the strongest available (see Table 1). We used mutant strains with balancer chromosomes carrying a LadZ. fusion gene to mark non-homozygous mutant embryos. In the case of maternally acting genes (fused, armadillo, dishevelled, porcupine, zeste-white 3), we have generated germ line clones using the dominant female sterile technique, to remove both the maternal and the zygotic gene products; here the paternally contributed wild-type X chromosome was also marked by a LacZ fusion gene.
Cell death, prominent in some of these mutants at later stages, is a confounding factor in interpreting the results (Klingensmith et al., 1989 and N. P., unpublished observations). However, in most of the mutants the initial aberrations from wild-type staining patterns occur earlier in development than cell death is detectable.
At embryonic stage 10, wg protein is detected in a discontinuous stripe one to two cells wide, along the parasegment border. For a description of wg and en protein patterns in wild-type embryos, see van den Heuvel et al. (1989); Gonzalez et al. (1991), and DiNardo et al. (1985). The en stripe is continuous, two cells wide just posterior to the parasegment border. In all mutants, the early segmental patterns of wg and en protein expression are identical to what is seen in wild-type embryos, arguing that the initial expression of these genes is independent of the other segment polarity genes and presumably totally regulated by earlier acting segmentation genes (Howard and Ingham, 1986; DiNardo and O’Farrell, 1987; Ingham et al., 1988). wg and en protein localizations are presented in Fig. 1 and the findings are summarized in Table 2.
Class A, wingless (wg), dishevelled (dsh), armadillo (arm) and porcupine (pore)
In embryos mutant for the wg allele used here, no wg protein is found at any time during embryogenesis, wg transcription is initialed normally in most wg alleles and disappears from the germband during stage 10 (Fig. 2; for a full description of wg patterns in wg alleles, see van den Heuvel et al., 1993). wg protein fades from the epidermis (dorsal fust) during stage 10 in arm and dsh embryos; no more wg staining is observed by the end of stage II. In contrast, in embryos lacking pore, the wg protein is found throughout most of embryogenesis but the subccllular localization appears altered, pore embryos show a retention of the wg protein in producing cells (van den Heuvel et al., 1993). Interestingly. whereas wg protein in pore embryos is present throughout stage 13. its mRNA can no longer be delected at the end of stage II. as in arm. dsh and wg embryos (not shown). en protein in wg embryos disappears from the epidermal cell layer al late stage 9 (sec DiNardo et al., 1988; Martinez- Arias et al., 1988; Bejsovcc and Martinez Arias. 1991). Only neuroblast cells on the ventral side are then positive for the en antigen and a distinct pattern in the gnathal and first thoracic segments in the epidermal cell layer persists. Identical effects on en expression are seen in arm. dsh and pore embryos and represent the earliest sign of segment polarity inlerregulation (for arm sec also Peifer et al., 1991). In arm. dsh and pore embryos, it is thus possible to detect wg protein at certain stages while the epidermal en pattern is already completely disrupted. Apparently, the wg signalling pathway lo en is impaired in these mutants. How might these gene products interact in a wg signalling pathway? Both arm and dsh function autonomously (Wieschaus and Rigglcman, 1987; .1. K. and N. P., unpublished data), consistent with a role in reception of the wg signal. Since arm is homologous to the intracellular vertebrate proteins. Plakoglobin /β-calenin (Peifer and Wieschaus. 1990; McCrea et al., 1991). a proposed function for (win as a receptor for wg (Peifer cl al., 1991) seems unlikely. An observation by Rigglcman cl al. (1990) indícales that the intracellular location of (he arm protein is dependent on wg but also on dsh. Perhaps, arm protein becomes associated with different proteins upon activation by the wg signal and this reassociation is required for wg function, dsh is necessary for both aspects of wg activity: maintenance of en and the relocalization of arm protein, dsh seems therefore a good candidate for a protein involved in the reception of the wg signal. However the molecular cloning of dsh does not clarify what its function is (J. K. and N. P., unpublished data).
Since wg protein is present and accumulates in pore embryos (sec also van den Heuvel cl al., 199.3), pore may be involved in processing of the wg protein. Consistent with a role in processing the wg protein is (he non-aulonomous function of pore (.1. K. and N. P” unpublished observations). Al what stage of (he processing of the wg protein pore might act is not known, but it luis been reported that the wg- dependent relocalization of the arm protein in pore embryos is restricted to the cells that express wg (Riggleman et al., 1990) This suggests that the wg protein can still function intracellularly.
In arm, dsh and pore mutants, transcription of wg is lost (in pore embryos the wg mRNA is lost) in a pattern very similar to that seen in wg mutants, indicating that the wg signalling pathway might also regulate wg expression. This will be discussed later.
Class B, engrailed (en), lines (lin), Cell, cubitus interruptus Dominant (cjD), fused (fu), gooseberry (gsb), hedgehog (hh) and smooth (smo)
The wg protein in en embryos disappears from the odd parasegmental stripes during stage 10. Staining in the even stripes persists but at stage 12 no more wg protein is detected in the germband (see also Martinez-Arias et al., 1988; Bejsovec and Martinez Arias, 1991). The transient seven stripe pattern of remaining wg expression is noteworthy since the cuticle phenotype of en also shows a paired- segment pattern. Indeed the most aberrant segments in the cuticle (Kornberg, 1981) correspond to the weak wg bands in the embryo.
In lin embryos, a paired-segment pattern for wg protein is seen, similar to the pattern in en mutants, although it arises later (stage 12) and is never as well defined. The expression of en is hardly affected in lin mutants; only some cells lose expression.
As argued by Orenic (Orenic et al., 1987; Orenic et al., 1990), Cell and ciD could be allelic. In both mutations, dorsal wg expression is lost during stage 10, while ventrally wg protein persists longer. In gsb embryos, wg expression is lost from the ventral epidermis during stage 10 (see also Hidalgo and Ingham, 1990; Hidalgo, 1991), while dorsally, protein is present longer. In both ciD/Cell and gsb, small gaps in the en domains are formed by stage 11. The patchy nature of the expression domain of en becomes clearer later in development.
In fu and in hh embryos, the wg protein fades from the dorsal epidermis by stage 10. By the end of stage 11, all staining has disappeared from the segmented region (see also Limbourg-Bouchon et al., 1991; Hidalgo and Ingham, 1990). In fu and hh embryos, gaps appear in the en stripes during stage 11 (see also DiNardo et al., 1988; Limbourg- Bouchon et al., 1991). The discontinuity of the en stripes becomes more obvious in later stages.
Embryos mutant for smo display normal patterns of expression of wg until stage 10. Most, but not all wg protein disappears from the ventral epidermis during subsequent development. During stage 11, small gaps appear in the en stripes which become clearer during subsequent development.
In contrast to class A embryos, wg expression is lost before en expression in all of the class B mutants. This happens during stage 10 of development, as is seen for the odd stripes of wg expression in en mutants. These gene products might therefore act in a pathway that maintains wg expression downstream of en activity. In gsb and ciD embryos, loss of wg expression is seen, initially more or less confined to the ventral and the dorsal side of the embryo, respectively. ciD is expressed in all cells expressing wg (Orenic et al., 1990) and gsbcl expression becomes restricted to the ventral epidermal cells overlapping the (ventral) wg and en domains (Baumgartner et al., 1987). Since both ciD and gsb encode putative transcription factors, they could directly regulate wg expression. The fu kinase most likely also acts in this pathway. However, it is not clear what the substrate of fu is and the ubiquitous expression does not clarify in which cell fu works, hh has been implicated in the maintenance of wg expression as a possible signalling molecule (Hidalgo and Ingham, 1990; Ingham et al., 1991), consistent with its apparent non-autonomy (Mohler, 1988), its molecular structure and its expression in the cells marked by en (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992). We have investigated and extended this proposed function of hh in several double mutant combinations.
(1) If en strictly regulates hh activity, the double mutant en;hh should display a wg expression pattern as in en mutants. We observe, however, a pattern as in hh mutants, indicating that hh activity is not regulated solely by en. Possibly pair rule genes are involved in the early regulation of hh (see also Lee et al., 1992; Tabata et al., 1992). Such an influence could explain the pair rule pattern of disappearance of wg expression in en mutants and thereby the cuticle phenotype of en mutant embryos. In the even parasegments, expression of hh, and thereby expression of wg, is maintained by pair rule gene activity. In the odd stripes, wg expression would be regulated by en via hh. On the other hand, maintenance of hh in the en cells is thought to be also dependent on wg signalling (Lee et al., 1992; Tabata et al., 1992); in wg mutants hh expression disappears as en expression. This indicates that an unknown gene acts downstream of wg signalling to regulate hh expression.
(2) If hh acts as a signal to maintain wg expression, it might also function to induce the ectopic wg expression seen in nkd embryos (see below). In double mutant nkd.hh embryos, we found no ectopic expression of wg, consistent with a role for hh as a signal from en cells to induce or maintain wg expression in neighbouring cells.
The pattern of disappearance of wg expression in these mutants can be directly correlated to the pattern that is seen in the mutants that are thought to function in the wg signalling pathway and in wg mutants. These results indicate that wg regulates its own transcription in a paracrine fashion (see also Ingham and Hidalgo, 1993). A wg signal is transduced which maintains en and hh activity in the neighbouring cell, hh then might function as a signal to maintain wg expression in the cell originally expressing wg.
Class C, naked (nkd), zeste white-3 (zw-3) and patched (ptc)
In the thoracic and abdominal segments of nkd embryos, wg becomes expressed in a row of cells anterior to the normal expression domain (see also Martinez-Arias et al., 1988; Limbourg-Bouchon et al., 1991), resulting in two stripes of wg per segment by stage 11. The en protein domain expands into the cells posterior to the wildtype expression pattern (see also Martinez-Arias et al., 1988), resulting in a domain twice the normal width at stage 10.
In embryos lacking zw-3 function, the wg protein is observed in the normal and in an ectopic domain, and the en protein domain enlarges, both in exactly the same manner as in nkd embryos. The expansion of the wg/en patterns in zw-3 and nkd embryos indicates that these genes might function in the same pathway; that is repression of wg/en in the anterior compartment. Recently, a model has been proposed in which cw-3 functions as an antagonist of en autoregulalion. The wg signal would repress zw-3 activity, and thereby maintain en expression in its appropriate position. This model is inferred in part from the broadening ol the en domain in the cw-3;wg double mutant combination (Siegfried et al., 1992). If nkd functions in this pathway, a similar pattern for en should be seen in the mutant combination wg;nkd. However. wg:nkd mutants show the loss of en expression as in wg mutants, a result that does not corroborate function of, zw-3 ami nkd in the same pathway. On the other hand, it is not known if the existing nkd alleles arc amorphic and residual activity of nkd might result in the observations we present.
In pic embryos, a broadened stripe of wg is observed (see Martinez-Arias et al., 1988; DiNardo et al., 1988). consistent with its proposed role as a repressor of wg expression. An ectopic en stripe is observed slightly later than the broad wg stripe is generated (not shown). nkd, pic and zw-3 appear to be involved in repression of en and wg in the anterior part of the segment, because of the ectopic expression found in these mutants. Interestingly, the patterns are established in two temporal stages. In nkd mutants, broadened expression of en is seen lirst and subsequently ectopic wg is detected. In ptc embryos, the stripe of wg is broadened and then an ectopic en stripe is induced. Expression of the second antigen might depend on the functional expression of the lirst. since it is known that wg and en arc dependent on each other for continual expression. This possibility has been investigated in the double mutant eir.nkd. Indeed no induction of ectopic wg is seen in these embryos. In a plc.wg double mutant no ectopic en is induced, as previously observed (DiNardo cl al., 1988).
Once the initial expression domains of some of the segment polarity genes arc established by pair rule gene activity. most of the segment polarity genes appear to function in two regulatory pathways, controlling maintenance and correct localization of ng and en expression on either side of the parasegment border. These pathways can be distinguished in time: lirst ng acts to stabilize en expression and subsequently ng expression is maintained by a signalling pathway originating from the en cell. Both act within a short lime window (stage 9-10/11). although some of these genes have been shown to have later embryonic functions as well (Bejsovec and Martinez Arias. 1991; 1 lecmskerk et al., 1991). Fig. 3 shows a simplified scheme of both pathways. The presentation or secretion of the ng protein is regulated by the pore gene product. The ng protein is secreted and interacts with the neighbouring (and perhaps also the producing) cell possibly via a putative transmembrane receptor. The dsh protein might be associated with or be downstream of the receptor. The interaction between ng and dsh and other putative molecules possibly leads to the inactivation of the protein kinase cw-3. which by itself is a negative regulator of en activity, arm functions upstream of en. Recent genetic epistasis experiments using a heat-shock ng transgene combined with loss of functions mutations in other segment polarity genes have shown that arm and dsh arc both required for ectopic en expression induced by IIS- wg (Noordermeer et al., 1993). In another series of double mutant embryos. Siegfried cl al. (1993) have obtained evidence that cw-3 acts downstream of dsh and upstream of arm. The hh transcript is only expressed in cells that express en and its activity is maintained by both en and possibly other genes, in conjunction with wg. hh activity controls wg expression, perhaps by relieving the negative action of pic. hh protein is seen inside the (‘hh-expressing cells and also in neighbouring cells (Taylor et al., 1993). consistent with a role as a signal. The protein kinase hh and the product of smo are involved in this pathway, either in the presentation or in the interpretation of the hh signal (thus either in the hh- producing or in the hh-receiving cell). The transcription factors ciD and gsh ultimately control wg expression. The precise role of nkd in (he repression of wg/en is not clear for the moment, lin appears to be acting late in development, resulting in cuticle defects but in minor defects in wg and en expression.
Some of the gene products that are thought to function intracellularly in these pathways (e.g. arm, fu, zw-3) are maternally provided. The genes are located on the X chromosome of Drosophila. Techniques to remove all activity, including maternal, for X-chromosome located genes have been used to screen for mutants as the ones discussed here (Perrimon et al., 1986). Similar techniques for the other chromosomes have become available only recently (Chou et al., 1993) and it is likely that more mutants with a segment polarity phenotype will be isolated, hopefully to advance our understanding of patterning within the Drosophila segment.
We thank Claudie Lamour-Isnard, Christiane Niisslein-Volhard, Trish Wilson, Phil Ingham, Thomas Kornberg, Peter Lawrence, Alfonso Martinez-Arias, Jym Mohler, Matthew Scott and Gary Struhl for fly stocks. The members of the Nusse Lab are acknowledged for their continuous interest. Many thanks to Phil Ingham for the advice. J. K. was in part supported by a grant from the National Institutes of Health. R. N. and N. P. are investigators of the Howard Hughes Medical Institute.