The functions of pannier during Drosophila embryogenesis

The morphological diversity in Drosophila is primarily established along the anteroposterior (AP) and the dorsoventral (DV) body axes. A great deal is known about the genetic factors that generate the diversity along the AP axis (Lawrence, 1992). Maternal products such as Bicoid and Caudal form functional gradients, which are resolved in the activation of a cascade of zygotic (gap, pair-rule, polarity) genes (Nüsslein-Volhard and Wieschaus, 1980). The end product of this process is the formation of a chain of 14 metameric units (parasegments) (Martinez-Arias and Lawrence, 1985), each composed of two stripes of cells, one expressing the gene engrailed (en) and the other not. The morphological diversity is then generated by the various Hox genes, which become active in different sets of parasegments (Lawrence and Morata, 1994; Mann and Morata, 2000).

By contrast, relatively little is known about genetic subdivisions of the body in the DV axis. A crucial event is the formation in early embryogenesis of a gradient of nuclear expression of the Dorsal protein, whose nuclear translocation requires the activity of the Toll receptor. Spatial restriction of Toll activity is dependent on the accumulation in the ventral region of the active form of the Toll ligand Spatzle, the result of a proteolytic processing catalysed by the serine protease encoded by the gene easter. In turn, the restriction of Easter activity to the ventral region depends on the localised activity of the heparan sulfate transferase encoded by pipe. The Pipe protein is thought to modify the proteoglycans of the matrix to allow interaction with the Easter protease in order to cleave the Spatzle protein (Anderson, 1998).

The Dorsal gradient is a principal element establishing local differences along the DV axis. Different levels of nuclear Dorsal regulate the activity of zygotic target genes such as snail, rhomboid and decapentaplegic (dpp), which are involved in the specification of different cell types. In early embryos, the Dpp product is localised in the dorsal half and its activity determines the formation of dorsal embryonic structures: in absence of dpp activity embryos become ventralised (Irish and Gelbart, 1987), whereas derepressed activity of the Dpp pathway results in dorsalised embryos (Nellen et al., 1996).

The subdivision of the dorsal ectoderm into distinct parts is achieved through the establishment of a complex Dpp activity gradient in the early embryo. This involves the function of another TGFβ molecule, encoded by Screw (Scw), which potentiates Dpp signalling (Arora et al., 1994), and that of the secreted protein Short gastrulation (Sog). The activity of Dpp/Scw is modulated by the Sog gradient (Neul and Ferguson, 1998): high Sog levels in the lateral region block Dpp/Scw and allow the formation of neuroectoderm; intermediate levels attenuate Dpp/Scw function to specify dorsal epidermis; low Sog levels enhance Dpp/Scw activity to form the most dorsal tissue, amnioserosa (Ashe and Levine, 1999). The response to Dpp/Scw is further complicated by the activity of brinker (brk), which encodes a transcriptional repressor (Zhang et al., 2001) and is expressed in lateral stripes in the neuroectoderm (Jazwinska et al., 1999b). brk suppresses the response to Dpp signalling, but its activity is repressed by high levels of Dpp in the dorsal ectoderm (Jazwinska et al., 1999b). The nature of the interactions between Dpp and Brk draws a border of the patterning influence of Dpp/Scw.

In addition to its early role specifying dorsal ectoderm, dpp has other embryonic functions that are independent of the polarity of the DV axis; it plays a role in dorsal closure, midgut development and tracheal formation (Affolter et al., 1994). The existence of these other functions is reflected in the dynamics of its expression. Although in early development dpp transcripts cover the half dorsal region of the embryo, after stage 11 (germ band elongation) they disappear from much of the dorsal embryos and become restricted to two longitudinal stripes: a dorsal one at the border of the epidermis with the amnioserosa, and the other in the lateral region (St Johnston and Gelbart, 1987). The expression of dpp in the dorsal stripe is under the control of the JNK pathway (Glise and Noselli, 1997) and is involved in embryonic dorsal closure.

The Dpp activity gradient probably achieves its different roles by activating various target genes. Several of these have been identified that respond to different thresholds of the Dpp/Scw gradient (Ashe et al., 2000). They include Race, hindsight, tailup, u-shaped (ush) and pannier (pnr) (Ramain et al., 1993; Frank and Rushlow, 1996; Rush and Levine, 1997), which define distinct dorsal domains and are probably instrumental in subdividing the dorsal ectoderm into different parts. pnr is a gene encoding a zinc-finger protein containing a GATA motif (Ramain et al., 1993; Winick et al., 1993), which has several embryonic functions connected with dorsal closure and heart development (Heitzler et al., 1996; Gajewski et al., 1999).

During embryogenesis, pnr is expressed in a complex pattern (Winick et al., 1993; Heitzler et al., 1996); in early embryos it is expressed in a broad dorsal domain extending from 20% to 60% of the egg length (Winick et al., 1993), a region including the presumptive amnioserosa and dorsal epidermis. This pattern is later refined, and by germ band retraction pnr is expressed in a longitudinal dorsal domain extending along the thoracic and abdominal segments (Calleja et al., 2000). This late embryonic pattern resembles that described for imaginal development, where it has been shown that pnr has an instructive, selector-like function, determining the identity of the medial dorsal structures of thoracic and abdominal segments (Calleja et al., 2000).

pnr embryonic expression and its role in adult development suggest that it may be involved in subdividing the dorsal part of the body into distinct genetic domains, but to date this possibility has not been examined. It has been reported that in pnr mutant embryos dorsal closure is defective and as a consequence the embryos present ‘holes’ in the dorsal cuticle (Heitzler et al., 1996). We investigate the embryonic function of pnr by studying the effects of alterations of pnr activity on the larval patterns and on the expression of genes involved in larval patterning. We show that it has an instructive role in specifying the dorsomedial pattern of all thoracic and abdominal segments. Our results indicate that pnr is the gene responsible for a major subdivision along the DV axis in the Drosophila body. We also show that pnr is involved in embryonic dorsal closure by activating dpp in the cells in the leading edge.

The pnr^VX6 allele has been described previously (Heitzler et al., 1996), and can be regarded as a null allele, as most of the coding sequence (except that coding for nine amino acids) is lacking. In addition we used the following mutants all which are considered null alleles: Abd-M^M1 (Casanova et al., 1986); grain^7L12 (Brown and Castelli-Gair Hombria, 2000); Df(2L)5 (deficient for sal and sal-r) (de Celis et al., 1996); ems^9Q64 (Dalton et al., 1989); brk^M68 (Jazwinska et al., 1999a); Df(3)iro² (Leyns et al., 1996); and lin^G1 (Bokor and DiNardo, 1996).

To distinguish hemizygous or homozygous mutant embryos from their heterozygous siblings, we made use of balancer chromosomes carrying lacZ transgenes: FM7c ftz-lacZ (Klambt et al., 1991), CyO wg-lacZ (Ingham et al., 1991) and TM3 hb lacZ (Hyduk and Percival-Smith, 1996). Other lacZ lines used were en-lacZ (Simcox et al., 1991), brk^M12-lacZ (Jazwinska et al., 1999a) and ush-lacZ.

The UAS-pnr chromosome was a gift from Mariann Bienz and has been described elsewhere (Heitzler et al., 1996). We also used the UAS-tkv^DN (Haerry et al., 1998). The Gal4 lines used were en-Gal4 (Tabata et al., 1995), Ubx-Gal4 (M. Calleja and G. M., unpublished), arm-Gal4 (Sanson et al., 1996), ptc-Gal4 (Wilder and Perrimon, 1995) and wg-Gal4 (M. Calleja and G. M., unpublished). LP1-Gal4 drives expression in the amnioserosa (G. M., unpublished).

Embryos were collected overnight and aged an additional 12 hours, then first instar larvae were dechorionated in commercial bleach for 3 minutes and the vitelline membrane removed using heptano-methanol 1:1. Then, after washing with methanol and 0.1% Triton X-100, larvae were mounted in Hoyer’s lactic acid 1:1 and allow to clear at 65°C for at least 24 hours.

Embryos were stained using standard procedures for confocal microscopy (Gonzalez-Crespo et al., 1998); secondary antibodies were coupled to Red-X and FITC fluorochroms (Jackson Immunoresearch) and embryos were analysed under a laser-scan Zeiss microscope.

In situ hybridisation and antibody/in situ hybridisation double labelling were performed as described previously (Azpiazu and Frasch, 1993), and embryos were mounted in Permount (Fisher Scientific). Digoxigenin-labelled RNA probes were synthesised as described (Tautz and Pfeifle, 1989). Those used were pnr full-length antisense RNA probe synthesised from a plasmid provided by Pat Simpson and dpp antisense RNA probe synthesised from a plasmid provided by Ana Macías.

The antibodies used were anti-Cad (Macdonald and Struhl, 1986), anti BP102 (hybridoma bank), anti-Eve (Frasch et al., 1986), anti-Ftz (Krause et al., 1988), anti-Kr (provided by Jordi Casanova), anti-Pnr and rabbit anti-β-galactosidase (Capel).

We have studied (with the help of Juan Pablo Albar of the Centro Nacional de Biotecnologia) the amino-acid sequence of the Pnr protein using the ‘PeptieStructure’ program, which makes secondary structure predictions for a peptide sequence. The predictions include measures for antigenicity index, chain flexibility, hydrophobicity and surface probability. In accordance with these data, we have chosen two peptides: a first peptide spanning amino acids 7 to 26 (DGDSTSDQQSTRDYPHFSGDYC) and a second from amino acid 272 to 284 (TRKRKPKKTGSGSC). The peptides were prepared as a fusion with KLH to increase the antigenicity of each peptide. Antiserum against these peptides was raised in rabbits. We performed the first injection with 250 μg of a mix of the two peptides and the next five injections with a mix of 125 μg each injection. The second injection was 21 days after the first, and the other boosters were given also with a 21 days interval. Antiserum from the rabbit was tested against fixed Drosophila embryos.

The embryonic expression of pnr has already been described (Winick et al., 1993; Heitzler et al., 1996; Calleja et al., 2000), therefore it will only be considered briefly here. We have assayed the distribution of Pnr products by in situ hybridisation using an RNA probe and also using an anti-Pnr antibody made in our laboratory (see Materials and Methods). As expected for a protein containing DNA-binding motifs, the Pnr product appears localised to the cell nuclei (Fig. 1). We found a good correlation between the patterns of RNA and protein distribution after embryonic stage 7, when the Pnr protein is first detected with the antibody.

In early embryos, pnr is expressed in a broad region on the dorsal side, which may occupy as much as 40% of the circumference of the embryo. It does not extend to the entire length of the embryo. The anterior and posterior borders of expression can be delimited by double staining of pnr with even-skipped (eve), fuzhi tarazu (ftz), caudal (cad) and engrailed (en) (Fig. 1). The anterior border is slightly anterior to the second eve stripe, which corresponds to parasegment 3 (Labp-T1a), whereas the posterior border abuts the 7th ftz stripe, which marks the anterior limit of parasegment 14 (Lawrence, 1992). Thus, the pnr embryonic domain extends from the labial to the ninth abdominal (A9) segment: the presumptive region of part of the head and the entire thorax and abdomen of larvae and adult flies.

As development proceeds the overall extent of the pnr domain in the AP axis does not change; the only significant modification is that between embryonic stages 10 and 11, pnr transcription is repressed in much of the A8 segment (Fig. 1B,D), thus leaving a gap of expression that has already been noted (Calleja et al., 2000). The small posterior portion in the posterior region in late embryos that does not contains pnr activity exhibits caudal (cad) activity (Fig. 1E). It corresponds to the presumptive A10 segment (Moreno and Morata, 1999), which gives rise to analia structures.

The extent of the Pnr domain in the DV axis is also modified during embryogenesis; by stage 10 there is no detectable pnr activity in the amnioserosa cells (Fig. 1B), even though in earlier stages it is expressed in the amnioserosa presumptive region (Fig. 1A,C). We have mapped pnr expression (Fig. 1F) with respect to that of dpp: a determinant of dorsal development in embryos and a positive regulator of pnr (Winick et al., 1993). In early embryonic stages, dpp is expressed in the dorsal half of the embryo (Ferguson and Anderson, 1992), but by stage 10 the Dpp product lacks in the most dorsal tissue (amnioserosa) and occupies about half of the epidermis, from the border of the amnioserosa to a mid-lateral region (St Johnston and Gelbart, 1987). This is later resolved in two stripes by subsequent loss of expression in the mid-dorsal region. The expression of pnr is confined within the domain defined by the two dpp stripes. dpp and pnr overlap in the dorsal region, and share a common border with the amnioserosa (Fig. 1F).

We have studied some aspects of the regulation of pnr activity. The loss of expression in the A8 segment depends on Abdominal-B (Abd-B) activity: in Abd-B mutants pnr is expressed in the A8 site (Fig. 2A). However, none of the known Abd-B target genes expressed in the A8 segment, spalt (sal), empty spiracles (ems) or grain (gnr) (Castelli-Gair, 1998), mediates this regulation, because pnr expression is not altered after mutation of any of these genes (not shown). Finally, we found that in mutant embryos for lines (lin), a co-factor of Abd-B function (Castelli-Gair, 1998), pnr is not downregulated in A8 (Fig. 2B).

As mentioned above, pnr expression is switched off in the amnioserosa region before germ band extension. The dorsal limit of pnr expression coincides with the morphological boundary between dorsal epidermis and the amnioserosa (Fig. 1C) and abuts the expression domain of Kruppel (Kr), which at that time is expressed in all amnioserosa cells. We do not know the identity of the factor(s) that suppress pnr transcription in the amnioserosa, although we have observed that there is no alteration of pnr expression in Kr mutants (not shown). Several amnioserosa-specific genes [Race, zen, tail-up, hindsight and serpent (Frank and Rushlow, 1996)] are candidates for this regulation.

On the ventral side, the pnr domain abuts that of iro (Calleja et al., 2000), raising the possibility that iro might be a negative regulator of pnr. However homozygous Df(3L)iro2 embryos, totally deficient for the Iroquois complex (Leyns et al., 1996), show normal pnr expression (not shown).

Because pnr is activated by dpp in early development (Winick et al., 1993; Ashe et al., 2000), we have checked whether its late expression is negatively regulated by brk, an antagonist of the Dpp pathway (Campbell and Tomlinson, 1999; Jazwinska, et al., 1999a; Minami et al., 1999). brk is expressed in a longitudinal domain in the lateral region of the embryo (Jazwinska et al., 1999b), close to the Pnr domain; thus, it might regulate a possible activating role of dpp on pnr. Besides, there is evidence (Jazwinska et al., 1999b; Ashe et al., 2000) that alterations in brk activity modify the extent of the early Pnr domain. Double label experiments show that in wild-type late embryos (from stage 10), pnr and brk define parallel longitudinal domains. brk is expressed in a more ventral position but there is an ample zone of overlap between the two domains (Fig. 2C). In brkM68 embryos, pnr expression from stage 10 onwards is like the wild type (Marty et al., 2000). As there is compelling evidence that early pnr activity is regulated by brk levels (Jazwinska et al., 1999b; Ashe et al., 2000), it suggests that pnr is under different control in late embryonic development. This is supported by the observation that in embryos lacking tkv zygotic function, the extent of the pnr domain is normal, although expression levels are weaker than in wild-type embryos (Affolter et al., 1994). It also supported by our finding that driving a dominant negative form of thick veins (UAS-tkv-DN) (Haerry et al., 1998) with Ubx-Gal4 does not alter normal pnr activity (not shown).

We have studied the effects on the larval cuticle patterns of alterations in pnr activity. The principal morphological features of the dorsal and ventral epidermis of the wild-type first instar larva are illustrated in Fig. 3. There are various types of cuticle differentiations on the dorsal side, which are easily discernible from the thick denticles present in the ventral side. The arrangement of cell types is not uniform in the dorsal cuticle. The dorsomedial region differentiates all the dorsal pattern elements, described by Heemskerk and DiNardo (Heemskerk and DiNardo, 1994), but the dorsolateral region lacks some of these elements. Especially relevant is the lack of dorsal triangles [cell type 1 in Heemskerk and DiNardo (Heemskerk and DiNardo, 1994)] in the dorsolateral region (Fig. 4), which differentiates only spinules (cell type 4). These dorsal triangles are especially clear in the abdominal segments. As they do not extend to the lateral region, the distinction between the medial and lateral region of the dorsal epidermis can be assayed by the presence or absence of dorsal triangles.

For the description of the null phenotype of pnr, we have used the pnr^VX6 mutation, which has been characterised genetically and molecularly (Ramain et al., 1993; Heitzler et al., 1996). It is a small deletion that removes all but nine amino acids of the Pnr protein (Ramain et al., 1993); it can therefore be considered as a null mutation. Homozygous pnr^VX6 embryos show no staining with anti-Pnr antibody.

There are two principal phenotypic alterations in pnr^VX6 embryos. The first is that dorsal closure is defective, as has already been reported (Heitzler et al., 1996). The left and right sides do not fuse properly, often leaving ‘holes’ in the dorsal cuticle, which gives the embryos a characteristic basket shape. This indicates an involvement on pnr in dorsal closure that we examine below. Although there are holes in the dorsal epidermis, dorsal cuticular elements are present in pnr^VX6 larvae (Fig. 3).

The second phenotypic trait is the disappearance in the abdominal region of the most dorsal pattern elements, the dorsal triangles, which appear to be replaced by dorsolateral spinules (Fig. 3). Our interpretation is that in the absence of pnr the dorsomedial pattern cannot be formed and the dorsolateral pattern extends dorsally. We have measured the width of the dorsal domain (as indicated by the distance from the border of the amnioserosa to the middle Dpp stripe) and found that there is a normal number of cells. This suggests that in the absence of pnr function, there is no cell loss but that the dorsomedial domain is transformed into the dorsolateral one. This is in good agreement with the previous observation (Calleja et al., 2000) that in pnr^VX6 mutant embryos the iro domain extends dorsally. As expected, no effect is seen in the ventral body region.

We expected an effect on the amnioserosa, because pnr is expressed in early embryos in the entire dorsal half, which includes the presumptive amnioserosa region. Moreover, Heitzler et al. (Heizler et al., 1996) report that in pnr^VX6 mutants, amnioserosa cells die prematurely. However, we fail to see any alteration in pnr^VX6 embryos; the amnioserosa cells appear morphologically normal until the end of embryogenesis. They also express molecular markers such as ush (Fig. 5B). Besides, a characteristic phenotypic trait of the genes required for amnioserosa development is that the mutant embryos adopt an u-shaped morphology (Frank and Rushlow, 1996), owing to their inability to retract the germ band. In pnr^VX6 embryos, germ band retraction is normal, suggesting that amnioserosa development is not affected.

We have used the Gal4/UAS method (Brand and Perrimon, 1993) to study the developmental potential of the Pnr protein during embryogenesis. Some Gal4 lines drive generalised expression (arm-Gal4) and others are restricted to different body parts (Ubx-Gal4, ptc-Gal4, wg-Gal4, en-Gal4, LP1-Gal4). We first observed that increased levels of the Pnr product (as in pnr-Gal4/UAS-pnr) do not have a detectable effect on larval patterns. This was expected because flies of pnr-Gal4/UAS-pnr genotype survive and show virtually wild-type phenotype (M. Calleja and G. M., unpublished).

The principal conclusion from the ectopic expression experiments can be summarised by saying that pnr is able to induce a transformation of the ventral and dorsolateral patterns into the mediodorsal ones. In arm-Gal4/UAS-pnr larvae, the entire epidermis becomes dorsalised (Fig. 4F,G). Close inspection of these larvae shows the presence in the abdominal segments of a continuous belt of dorsal triangles, indicating that the transformation is towards the mediodorsal pattern. A similar observation is made using a Ubx-Gal4 driver (Fig. 4B,E). This line mimics the expression of the wild-type Ultrabithorax (Ubx) gene in embryos and shows expression from the posterior compartment of the second thoracic segment (T2) down to the abdominal segment A7, although it is weaker in the more posterior abdominal segments. The presence of the Pnr protein in the entire Ubx domain can be demonstrated with the anti-Pnr antibody (Fig. 4C).

The transformation of ventral to dorsal epidermis can also be seen in lines driving expression in restricted regions of segments. en-Gal4/UAS-pnr embryos show the transformation in the P compartments; a thin stripe of dorsal epidermis can be seen in the ventral region of each segment. In wg-Gal4/UAS-pnr embryos the transformation is restricted to a portion of the anterior compartment (Fig. 4H) that corresponds to the embryonic expression of wg, just anterior to the en stripe (Bejsovec and Martinez Arias, 1991). These results suggest that transformation induced by Pnr is cell autonomous, restricted only to the cells containing the product.

We have not tested whether the effect of ectopic pnr expression extends to the mesoderm, but it clearly affects the central nervous system (CNS). In Ubx-Gal4/UAS-pnr embryos, the ventral cord is clearly altered, precisely in the Ubx domain (not shown), suggesting that the transformation induced by pnr affects all the ectodermal derivatives.

In contrast to the observed in the epidermis and the CNS, ectopic pnr expression does not seem to affect in the amnioserosa, the most dorsal ectodermal derivative. In Ubx-Gal4/UAS-pnr embryos the amnioserosa develops normally even though it contains high levels of Pnr protein (Fig. 4C). We have used a amnioserosa specific driver LP1 (see Fig. 7) to express pnr only in this tissue and do not observe any defect. In Ubx-Gal4/UAS-pnr or LP1/UAS-pnr embryos, germ band retraction is normal.

We have analysed the regulatory interactions of pnr with ush and dpp, whose expression domains overlap with that of pnr. The negative control role of pnr on iro activity has already been reported (Calleja et al., 2000).

The wild-type expression of ush is shown in Fig. 5A; it covers the amnioserosa and also part of the dorsal domain in the epidermis, where it overlaps with pnr. In the dorsal epidermis, the ush domain is similar to that of pnr: the two genes define longitudinal domains and both are downregulated in A8. The difference is that the ush domain is narrower. In absence of pnr activity (pnr^VX6 embryos), ush expression in the epidermis is abolished, whereas that in the amnioserosa it is unaffected (Fig. 5B). Conversely, ectopic pnr activity induces ush expression outside its normal domain (Fig. 5C), suggesting an upstream control by pnr. This control of ush by pnr provides an explanation for the downregulation of ush in the A8 segment: ush expression depends on that of pnr, which is turned off. We note that ush has to have other regulators, because its expression in the amnioserosa does not depend on pnr (Fig. 5B).

The wild-type expression of dpp changes during embryogenesis, suggesting the existence of several regulatory tiers; the original broad dorsal expression is resolved in late embryonic stages into two thin stripes running in the anteroposterior direction (Fig. 5D). A dorsal stripe is located at the junction of the epidermis with the amnioserosa, whereas the other is located more ventrally. The dorsal stripe probably reflects a requirement for activity of the Dpp pathway in dorsal closure, as indicated by the dorsal open phenotype of mutations in Dpp transducers such as think veins and punt (Affolter et al., 1994). It is under the control of the JNK pathway (Glise and Noselli, 1997). It requires the activity of hemipterous (hep) a mitogen-activated protein kinase kinase (MAPKK) related to vertebrate Jun N-terminal kinase kinase (JNKK). hep controls dorsal closure by independently activating dpp and puckered (pc) a gene necessary for the movement of the leading edges during dorsal closure (Martin-Blanco et al., 1998).

We find that, just as in hep mutants, in pnr^VX6 embryos, the dorsal Dpp stripe disappears, although the stripe located more ventrally is not altered (Fig. 5E). Moreover, ectopic pnr activity also induces ectopic dpp expression (Fig. 5F). These results argue that pnr acts as a positive regulator of dpp in late embryogenesis. We note, however, that the dorsal dpp stripe of the wild type is not interrupted in A8, as might be expected if it required continuous pnr activity. As the downregulation of pnr occurs between stages 10 and 11 (Fig. 1B,D) and by that time the dorsal dpp stripe is already formed, we suspect the earlier pnr expression induces dpp activity in A8 and later dpp maintains its own expression.

The loss of the dorsal dpp stripe in the absence of either pnr function or JNK activity (Glise and Noselli, 1997) suggested that pnr might be required for the initiation or functioning of the JNK pathway. Therefore, we checked the activity of the JNK pathway in pnr^VX6 embryos by examining the expression of puc, the final element of the cascade. The result is that puc activity is not altered (Fig. 6), indicating that the formation of the dorsal dpp stripe requires independent inputs from the JNK pathway and from pnr.

Together, the preceding observations indicate that in late embryogenesis pnr acts as a positive regulator of both ush and dpp. These results also show that the regulatory interactions between dpp and pnr are reversed during development: whereas in early development dpp acts upstream pnr (Winick et al., 1993; Ashe et al., 2000), in late embryogenesis pnr upregulates dpp activity. This probably reflects the acquisition of different roles in the course of development.

Despite its overall effect on the epidermis and the CNS, there is no detectable effect of pnr activity in the amnioserosa. For example, in Ubx-Gal4/UAS-pnr embryos the amnioserosa appeared to be unaffected even though it contains Pnr protein (Fig. 4C). Moreover, in those embryos there is an expansion of dpp expression all over the epidermis except in the amnioserosa (not shown), suggesting that pnr is unable to induce dpp activity there. We have explored this phenomenon by using a new Gal4 line, LP1, which drives high expression levels specifically in the amnioserosa (Fig. 7). In LP1/UAS-pnr embryos there is a high level of Pnr protein in the amnioserosa (Fig. 7B) but no sign of ectopic dpp expression (Fig. 7D). In addition, the expression of specific amnioserosa genes such as Kr is not affected (Fig. 7C) and germ band retraction is normal. This result suggests that the developmental function of pnr is inhibited in the amnioserosa at the post-transcriptional level. It resembles the phenomenon of phenotypic suppression/posterior prevalence, described for Hox gene function in the AP axis (Gonzalez-Reyes and Morata, 1990; Duboule, 1991; Duboule and Morata, 1994).

We have addressed the problem of how morphological diversity is achieved in the DV axis of the embryo. There are two pertinent questions to be answered: (1) how the embryo is subdivided into different parts along the DV axis; and (2) the identities of the genes responsible for making the various parts different from each other. Our results indicate that pnr is involved in the process: it participates in the subdivision of the dorsal region of the embryo into two distinct domains and also specifies the identity of the dorsomedial domain. These results, together with those previously reported on pnr function in adult patterns (Calleja et al., 2000), strongly indicate pnr has a principal role in establishing the Drosophila body plan. We discuss these findings and also other aspects of the function and regulation of pnr during embryogenesis.

In early development, pnr is activated in response to dpp activity (Winick et al., 1993; Ashe et al., 2000) in a broad dorsal domain, which we show extends from parasegments 2/3 to the border between 13/14, although the borders are not strictly parasegmental. The control by dpp is consistent with the effect of brk mutations on early pnr expression (Jazwinska et al., 1999b; Ashe et al., 2000).

The original expression domain is substantially modified during embryogenesis. By germ band extension (stage 10) pnr activity is limited dorsally by the border between the epidermis and the amnioserosa, and laterally by the dorsal border of iro (Calleja et al., 2000). We do not know which factor(s) is responsible for the loss of expression in the amnioserosa, although likely candidates are several genes specifically active in this region, such as Race, zen, hindsight or serpent (Frank and Rushlow, 1996; Rush and Levine, 1997). In addition, we do not know how the late expression is regulated at the lateral border. It is not achieved by iro, as the loss of the entire Iroquois complex does not affect pnr expression.

Another modification occurs between stages 10 and 11, and is the loss of expression in the A8 segment. Expectedly, it is under the control of Abd-B; in Abd-B mutants the gap in A8 does not appear (Fig. 2A). However, none of the known Abd-B target genes sal, ems and grn (Castelli-Gair, 1998) is involved in the regulation, as their mutations do not affect pnr expression. Our finding that lin, which is considered as a co-factor of Abd-B (Castelli-Gair, 1998), is involved (Fig. 2B), suggests that downregulation of pnr in the A8 segment is mediated either by an unknown Abd-B target or directly by interaction between the Abd-B and Lin products. It is not clear why pnr activity has to be eliminated precisely in the A8 segment. We notice that this segment gives rise to the spiracles, protruding structures that are very different from those differentiated by the other abdominal segments where pnr remains active. In fact, there are several Abd-B target genes specifically activated in the spiracles (Castelli-Gair, 1998). It is possible that the formation of these structures demands that the pnr activity, which specifies larval epidermis of very different morphology, be turned off.

Interestingly, whereas early pnr expression is under dpp control, the late expression is not. Late inactivation of the Dpp pathway, using a dominant negative form of thick veins, does not modify pnr expression. In addition, mutations at brk, which allow higher response levels to Dpp signalling (Campbell and Tomlinson, 1999; Jazwinska et al., 1999a; Minami et al., 1999) fail to affect pnr expression in late development (Marty et al., 2000; H. H. and G. M., unpublished), although they affect early expression (Jazwinska et al., 1999b; Ashe et al., 2000). This indicates that pnr expression is controlled independently in early and late development, and by different factors.

There is already evidence that pnr has distinct functions during embryogenesis. Its activity in the dorsal epidermis is required for dorsal closure (Heitzler et al., 1996) and it is also expressed in the dorsal mesoderm where it is involved in the specification of cardiac cells (Gajewski et al., 1999).

We provide evidence for another and more general function of pnr. Our results indicate that it specifies the identity of a dorsomedial body region that spans from the labial segment to the end of the abdomen. This is clearly demonstrated by the effects seen in mutant embryos and after ectopic expression experiments. In pnr^VX6 embryos, the dorsomedial cuticle does not form, and there is an expansion of the dorsolateral epidermis (Fig. 4), suggesting that the cells of the dorsomedial domain acquire a dorsolateral fate. The ectopic expression experiments also point to the same conclusion. In larvae like arm-Gal4/UAS-pnr, the entire larval epidermis acquires dorsomedial features (Fig. 4F,G), whereas using more restricted drivers (Ubx-Gal4, wg-Gal4) the transformation is limited to the region where the Pnr protein is present (Fig. 4D,E.H), suggesting that the effect of pnr is cell autonomous. Thus, the Pnr protein is able by itself to trigger a developmental pathway, a typical property of selector gene products (Mann and Morata, 2000). In addition, it induces a ventral to dorsal transformation as corresponding to each segment, indicating that it acts in combination with Hox genes. These observations indicate that selector genes in the AP and DV axes have to co-operate to determine the different spatial patterns.

The transformation of ventral and dorsolateral epidermis towards dorsomedial observed after ectopic pnr expression is also reflected in the activity of marker genes of the distinct regions. Characteristic genes of the ventral neuroectoderm such as BP102 for the CNS (not shown) or buttonhead (C. Estella and G. M., unpublished) are suppressed. In addition, pnr is able to suppress iro activity (Calleja et al., 2000), a property that, as in the adult cells, is important to keep the dorsomedial and dorsolateral domains separate during embryogenesis.

The developmental effects observed after either loss or the gain of pnr function in the larval epidermis resemble those reported for the adult cuticle. In the latter, it has been shown that the activity of pnr maintains the segregation of the dorsal cuticle into medial and lateral domains, and also specifies the identity of a medial one (Calleja et al., 2000). This indicates that pnr has a general function involved in the subdivision of the body along the DV axis. The longitudinal stripe of pnr expression established during embryogenesis is probably a major constituent of the body and represents an zone of common identity.

In addition, Pnr has other more concrete functions connected with the specification of cardiac cells (Gajewski et al., 1999) and embryonic dorsal closure (Heitzler et al., 1996). Our results indicate that the involvement of pnr in dorsal closure is exerted through its activation of dpp in late embryogenesis, which is responsible for the formation of the Dpp stripe at the junction of the epidermis with the amnioserosa. Normal functioning of the Dpp pathway in this region is required for dorsal closure (Affolter et al., 1994; Glise and Noselli, 1997), suggesting that defects in dorsal closure observed in pnr mutant embryos (Heitzler et al., 1996) is the result of the lack of the dorsal dpp stripe.

There is evidence that this dpp expression requires function of the JNK kinase pathway (Glise and Noselli, 1997), and we show that it also requires pnr activity. Our observation that in absence of pnr activity the expression of puc, the end element of the JNK pathway (Martin-Blanco et al., 1998) is normal, indicates that in pnr mutants the JNK pathway is normally active. In turn, it shows that the activation of dpp in the dorsal stripe requires independent inputs from both the JNK pathway and pnr.

One intriguing aspect of pnr function is that it is able to induce a developmental modification in all ectodermal structures along the DV body axis except in the amnioserosa, the most dorsal tissue. Even under conditions in which pnr is transcribed and translated in all the amnioserosa cells (Fig. 7), it does not appear to elicit any developmental effect; none of the amnioserosa marker genes is affected by forcing pnr activity and the retraction of the germ band [a morphological indicator of the function of specific amnioserosa genes (Frank and Rushlow, 1996)] is also normal. Similarly, pnr is able to induce dpp activity all over the body except in the amnioserosa (Fig. 5, Fig. 7), where the presence of the Pnr protein appears to be inconsequential. This situation resembles the phenotypic suppression/posterior prevalence phenomenon discovered in the Hox genes specifying the AP body axis (Gonzalez-Reyes and Morata, 1990; Duboule, 1991; Duboule and Morata, 1994). It consists of a functional inactivation of a Hox protein by the presence of another normally expressed in a more posterior region of the body. It is conceivable that there might be a ‘dorsal prevalence’ in the DV axis, by which dorsal expressing genes are functionally dominant over the ventral expressing ones. It would be expected that genes specifying amnioserosa would be able to transform all structures as they would be ranking highest in the functional hierarchy.

We thank Juan Pablo Albar for his help with the analysis of the amino acid sequence of Pnr, for advise on preparing the anti-Pnr antibody and the generous gift of the peptides. We thank Mariann Bienz for the UAS-pnr flies. We also thank Natalia Azpiazu, Manuel Calleja, Carlos Estella and Ernesto Sanchez-Herrero for discussions and comments on the work, and Rosa Gonzalez and Angélica Cantarero for their technical help. The experimental work has been supported by grants from the Dirección General de Investigación Científica y Técnica, the Comunidad Autónoma de Madrid and an institutional grant from the Fundación Ramón Areces.