ladybird, a tandem of homeobox genes that maintain late wingless expression in terminal and dorsal epidermis of the Drosophila embryo

Cell identity in the segmented epidermis of Drosophila embryos is specified by a network of segment polarity genes (for review see Peifer and Bejsovec, 1992; Perrimon, 1994). Most of them are highly conserved in evolution, providing a model system for cellular interactions. Initially activated by the pair rule genes, expression of the segment polarity genes becomes interdependent. One of the best examples of such a regulation is the mutual dependence between wingless (wg) (Baker, 1987) and engrailed (en) (Poole et al., 1985). These genes are expressed on either side of the parasegmental border and define two signaling centres that play a key role in epidermal patterning. wg itself encodes a secreted protein that acts on neighbouring cells and is required to generate naked cuticle in a restricted part of each segment (Heemskerk et al., 1991; Ingham and Hidalgo, 1993). In addition to its epidermal function, wg plays a role in the specification of neuroblast identity (Chu-LaGraff and Doe, 1993) in embryonic mesoderm formation (Wu et al., 1995) and imaginal development (Wilder and Perrimon, 1995). A putative Wg receptor (Bhanot et al., 1996) and the products of other genes, armadillo (arm), dishevelled (dsh) (Noordermeer et al., 1994) and shaggy/GSK-3 (Bourouis et al., 1990) are thought to be required for the transduction of the Wg signal.

After 6 hours of development maintenance of wg activity becomes dependent on two functionally redundant gooseberry genes (gsb and gsbn), which encode transcription factors containing highly related paired domains and prd-type homeodomains (Gutjahr et al., 1993; Li and Noll, 1993). Interactions between wg and gsb are restricted to the ventral epidermis and the loss of gsb function results in defects in cuticle differentiation leading to a lawn of denticles (Li and Noll, 1993). At the same time, cells of the dorsal epidermis and the terminal regions of the embryo undergo distinct patterning processes, and maintenance of wg activity in these cells requires other regulatory mechanisms.

We have isolated two Drosophila homeobox genes called ladybird (Jagla et al., 1993, 1994). Like gsb and gsbn (Li and Noll, 1993), they are clustered in tandem and are expressed in the epidermis, mesoderm and central nervous system (CNS) of embryos. The ladybird genes are located in the 93E homeobox gene cluster, just distally to bagpipe (bap) (Azpiazu and Frasch, 1993) and proximally to S59 (Dohrmann et al., 1990). The most 5′ located gene, ladybird early (lbe), is activated during germ band elongation slightly earlier than its relative ladybird late (lbl). lbl follows the expression pattern of lbe and both lb genes encode transcription factors bearing a specific Ladybird-type homeodomain (Jagla et al., 1994). In addition to the Drosophila genes, orthologous genes have been found in mouse (Lbx1) and human (LBX1) (Jagla et al., 1995), suggesting that the ladybird genes could have an evolutionarily conserved role in development.

Here we report the cloning of full length cDNA sequences of both lbe and lbl genes, analyse their specific epidermal expression patterns and discuss regulatory interactions with wg. We show that activity of lb genes in the epidermis is regulated by the segment polarity gene network and depends on Wg signaling. Analysis of embryos homozygous for a deficiency uncovering the lb locus, as well as embryos with ubiquitous expression of both lb genes, indicates a requirement for lb in dorsal epidermis and anal plate development.

A bidirectional chromosomal walk between the bap and S59 genes was carried out using PCR-generated probes corresponding to the 3′ region of bap and 5′ region of S59. Genomic clones were restricted and aligned with previously obtained lbe and lbl λ clones (Jagla et al., 1993, 1994). The distal breakpoints of Df(3R)GC14, Df(3R)eF1 and Df(3R)eBS2 were analysed by PCR amplification on genomic DNA prepared from single homozygous embryos selected by their bloated gut appearance (Bodmer, 1993). A set of primers targeting tin, bap, lbl, lbe and S59 genes were designed using sequences available in the EMBL/GenBank Database Library.

Several embryonic cDNA libraries were first tested by PCR for the presence of both lbe and lbl clones. Selected libraries were screened using homeobox-containing genomic fragments of lbe and lbl as previously reported (Jagla et al., 1993, 1994). One lbe and three lbl full length cDNA clones were obtained from the embryonic λ Zap libraries kindly provided by K. Zinn (Caltech, Pasadena) and C. S. Thummel (University of Utah, Salt Lake City) and sequenced using the TaqDyeDeoxy Terminator Cycle kit and an automated DNA sequencer.

lbe and lbl coding sequences were amplified by PCR from cDNA templates, using DeepVent DNA polymerase (Biolabs), and cloned downstream to the 6His-encoding motif of a His-pET expression vector. Chimeric proteins were produced in the E. coli BL21 pLysS strain and purified on Ni-agarose affinity column as previously described (Jagla et al., 1994). Polyclonal and monoclonal antibodies against these proteins were produced and tested by ELISA, western blots and by whole-mount embryo immunocytochemical staining. In addition, the monoclonal antibodies were selected using immunostaining of COS cells transfected by the pSG5-lbe or lbl expression vectors.

Embryos were collected from apple-juice agar plates, dechorionated, fixed and processed according to the method of Tautz and Pfeifie (1989). The digoxigenin (Dig)-labeled DNA probes targeting lb genes were prepared by PCR (Jagla et al., 1994). The wg 3.0 kb cDNA probe (Baker, 1987) was labeled by random priming. After overnight hybridization with the Dig probes (final concentration of 2-4 ng/ml), embryos were incubated (1 hour) with preadsorbed anti-Dig antibody coupled with alkaline phosphatase (Boehringer) (1:2000). Colour reaction was performed using NBT and X-phosphate as substrate. For antibody staining whole-mount embryo preparations were blocked (1 hour) in 5% normal goat serum and incubated overnight at 4°C with first antibody. The secondary antibodies were biotinylated horse anti-rabbit or anti-mouse IgG, detected using an ABC-AP or Elite-ABChorseradish peroxidase (HRP) kit (Vector Laboratories). To identify cell positions or to distinguish heterozygous embryos, some preparations were double stained with anti-β -gal, anti-En or anti-Eve antibodies. Stained embryos were dehydrated, mounted in Canada balsam and photographed using a Nomarski optics.

Cuticles were prepared essentially as described by Li and Noll (1993), mounted in Hoyer’s medium and photographed under phase contrast optics.

The heat-shock lb constructs (hs-lbe and hs-lbl) were made by inserting the full-length lbe and lbl cDNAs into the EcoRI site of the P-element vector pCaSpeR-hs and injected into w¹¹¹⁸ embryos according to standard procedure (Rubin and Spradling, 1982). From several transformants, the hs-lbe4A and hs-lbl5/1 insertions on chromosome X were used in most experiments. Double hs-lbe/lbl transgenic flies were generated by recombination of hs-lbl transgenes on hs-lbe containing X chromosome. Single hs-lb transgenes were combined with Df(3R)eF1 for the rescue experiments. Embryos homozygous for the Df(3R)eF1 deletion were identified by a lack of pericardial cells when immunostained with anti-Eve antibody.

hs-gsb, hs-lbe, hs-lbl, hs-lbe;hs-lbl and (wild-type control) embryos were collected, aged on agar plates at 25°C and heat-shocked (15 minutes 37°C) in water. After incubation in a humidified chamber at 25°C the embryos for in situ hybridization and immunostaining were removed at 7 to 8 hours AEL, those for cuticle preparation were left up to 24 hours AEL. wg^IL114 embryos were aged at 18°C and shifted to the non-permissive temperature (29°C) at 6 hours AEL.

Flies were raised on standard Drosophila medium at 25°C. The null alleles ftz^W20, en^IIB86 and hypomorphic eve^ID19 were kindly provided by W. Gehring (Biozentrum, Basel). hh^IJ35, ptc^IN108, wg^CX4, wg^IL114, nkd^7H16, Df(2R)gsb^IIX62, and y arm^XK22mutants were obtained from the Tübingen stock center. Df(2R)gsb^IIX62 (Bopp et al.,1986) carries a deletion encompassing both gsb and gsbn genes. The hs-gsb transgenic strain was from M. Noll (Institute of Molecular Biology, Zurich) and deficiencies of the 93 region (Df(3R)eGp4, Df(3R)GC14, Df(3R)eF1, Df(3R)eBS2) were from the Bloomingthon stock center. bap²⁰⁸ and tin^EC40 EMS alleles were provided by M. Frasch. The Oregon R strain was used as a wild-type control.

Our previous data (Jagla et al., 1994) showed that lbe and lbl are tandemly located in the 93E homeobox gene cluster. In order to define the position of lb inside the cluster, the distal breakpoints of four deficiencies were analysed (Fig. 1A-C) and a genomic walk was carried out (Fig. 1B). Initially, by PCR (see Materials and methods), we determined that lbl maps between the distal breakpoints of Df(3R)GC14 and Df(3R)eF1 while the S59 gene is located outside from Df(3R)eF1 (Fig. 1B,C). Since, bap alleles complement Df(3R)GC14 whereas alleles of its immediate neighbour tin do not (Azpiazu and Frasch, 1993), the PCR results indicate that lb genes are located distally to bap and proximally to S59. This gene order was confirmed by our genomic walk (Fig. 1B) which also revealed that lbe maps distally to lbl and that both genes are transcribed from the opposite DNA strand compared to tin, bap and S59. In addition, as determined by Southern blot analysis (not shown), the λ clones encompassing the genomic region between lbe and S59 (Fig. 1B) contained a Drosophila homologue of the insulin receptor (inr) gene (Fernandez et al., 1995). The position of the distal breakpoint of Df(3R)eF1 was analysed by PCR using the lbe and two different lbl primers targeting 3′ (lbl1) and 5′ (lbl2) coding sequences (Fig. 2B,C). The lbl gene was found to be inside and lbe outside the region deleted by Df(3R)eF1. However, the lack of lbe activity in the Df(3R)eF1 homozygous embryos (Fig. 8H) suggests that Df(3R)eF1 deficiency carries an additional mutation within the lbe gene, or deletes some lbe regulatory sequences located downstream of the gene.

As found previously, Lbe and Lbl share highly homologous regions extending downstream from the homeodomain (Jagla et al., 1994). The sequence of full length lb cDNA clones revealed conservation of additional regions located immediately upstream from the homeodomain (Fig. 2A,B). The 2045 bp lbe cDNA clone (not shown, see GenBank accession number) contains a leader sequence of 227 bp, an ORF of 1440 bp, and a trailer of 378 bp. Analysis of the coding sequence, predicts a protein of 479 amino acids with a Prd-like repeat (Bopp et al., 1986) in the amino-terminal region, a homeodomain in the carboxy-terminal part and two poly-Ala stretches on either side of the protein (Fig. 2A,B). The lbl cDNA clones lbl20 (1477 bp), lbl5 (1867 bp) and lbl6 (1982 bp) (not shown) correspond to transcripts differentially spliced by retention of the short introns (dark grey boxes in Fig. 2A,B), a phenomenon previously reported in human (Cooke et al., 1988). These cDNAs predict three proteins of 346, 372 and 411 amino acids (Fig. 2B). The longest (Lbl6) differs from the others at the carboxy-terminal region (Fig. 2B). The common features of lbl clones are (i) a short 5′ leader sequence (about 150 bp), (ii) the same initiating methionine codon, and (iii) several polyadenylation signals (AATAAA) preceding the poly(A) stretches. In addition to the short alternative introns, comparison between lbl genomic and cDNA sequences, revealed a large 20 kb intron upstream of the homeobox, and a 571 bp intron located inside the homeobox (Figs.1B, 2B) (Jagla et al., 1993). The predicted Lbe and Lbl proteins contain highly conserved homeodomains (97%) and regions of conservation upstream (6181%) and downstream (77%) of the homeodomain (Fig. 2C). To further characterize these proteins, we raised antisera to His-tagged-Ladybird fusion proteins in both mice and rabbits. The anti-Lbe and anti-Lbl antibodies specifically recognize the corresponding recombinant protein (Fig. 3A) and detect nuclear proteins in COS cells transfected by the corresponding expression vector (Fig. 3B and data not shown) and in whole-mount embryos (Fig. 8C).

Activation of lbe slightly precedes that of lbl and appears during germ band elongation (3 hours 30 minutes AEL) in the primordium of the anal plate (Fig. 4A) and subsequently during neuroblast segregation (about 4 hours AEL) in the epiderm of gnathal segments and in some neuroblasts (Fig. 4B). Later, after completion of germ band elongation (4 hours 20 minutes AEL), epidermal domains of the terminal regions of the body contain only low levels of lbl transcripts (not shown) and Lbl protein (Fig. 4G,H). This is consistent with a spatial distribution of lbl gene products that evolves rapidly during embryogenesis into a pattern corresponding to that of lbe. The only difference concerns the trunk epidermis, where lbe transcripts are much more abundant (compare Fig. 4D and J).

Dynamic appearance of new epidermal and mesodermal domains of lbe and lbl gene expression takes place between 5 and 7 hours AEL, although the most prominent region of lb activity remains the anal plate (Fig. 4C,E,I,K). At about 5 hours AEL, both lbe and lbl start to be expressed in a cluster of mesodermal cells corresponding to the heart precursors (Jagla, K. and Frasch, M., unpublished data; see Fig. 4C,J). At the onset of segmental groove formation just posterior to these mesodermal cells in the thoracic and abdominal segments (A1-A7) a one-cell-wide epidermal lb stripe appears (Fig. 4C,J). This dorsal stripe broadens anteriorly up to 4-5 cells at 7 hours AEL (Fig. 4D) and then up to 6-7 cells after germ band retraction (Fig. 4F,L). Surprisingly, no expression of either lb genes is detected in the dorsal epidermal cells of the most posterior abdominal (A8 and A9) segments (Fig. 4C,F,I,L). At the late extended germ band stage lb transcripts also appear in the ventral, but not the lateral epidermis (Fig. 4D,E). These ventral lb stripes are weaker than dorsal patches (Fig. 4D) and in the case of lbl become visible during germ band retraction (not shown). After germ band retraction, the anti-Lbe and anti-Lbl antibodies clearly label broad epidermal stripes of dorsal cells (Fig. 4F,L) which migrate towards dorsal closure. Subsequently, during head involution, lbe and lbl expression decreases throughout the ventral but not the dorsal epidermis (Fig. 4L) and still persists in the terminal regions of the body corresponding to the head segments and the anal plate (not shown). At later stages of embryogenesis, lb gene expression progressively disappears from the epidermis and becomes restricted to the segmental border muscles and clusters of cells in the central and peripheral nervous system (not shown).

The distribution of lb genes products in the epidermis of the trunk is particularly reminiscent of the late wg expression pattern. To localise more precisely the lb expressing epidermal cells along the anteroposterior axis of individual segments, we used, as a positional marker the En-positive cells of the posterior compartment (Poole et al., 1985) specifically stained with anti-En antibody. Double immunostaining for Wg/En (Fig 5A), Gsb/En (Fig. 5B) and En-immunostaining of embryos previously hybridized to the lbe probe (Fig. 5C,D) revealed a major domain of lb expression that overlaps the cells expressing wg and gsb, located just anteriorly to the cells expressing en. The posterior-most row of lb-expressing cells crosses the parasegmental groove and overlaps the en domain, as is observed for gsb (Li and Noll, 1993). In the dorsal epidermis, however, gsb stripes disappear at 6 hours AEL (Fig. 5B) whereas lbe (Fig. 5D) and lbl (not shown) remain co-expressed with wg. In addition to the expression in the trunk, we also observed common epidermal domains of lb and wg activity in the terminal regions of the body, corresponding to the labrum (Fig. 5E,F) and to the anal plate (Fig. 5E,G). These regions represent the earliest domains of activity for both lb genes.

The similarity between lb and wg gene expression patterns (Fig. 6A) prompted us to analyse lbe transcript distribution in some of the segmentation mutants. lbe expression is altered by mutations of the pair rule genes even skipped (eve) and fushi tarazu (ftz) (data not shown) which are both required for proper expression of en and wg (Frasch et al., 1987, Ingham et al., 1988). However, since lbe gene activity appears late, during the extended germ band stage, it is unlikely to be regulated directly by the early-acting pair rule genes, but may be a target or even a component of the Wg signaling pathway. To investigate this possibility we have analysed lbe activity in the absence of the major elements of the en-wg regulatory loop (Fig. 6B-F). Using the double staining procedure with anti-En antibodies and a lbe specific probe we found that in en (Fig. 6B) and wg (data not shown) null mutants lbe transcription is down-regulated. Since en is not affected and lbe activity decays in wg^IL114 embryos shifted to the non-permissive temperature at 6 hours AEL (Fig. 6C) we postulate that wg rather than en is directly required for lbe transcription. This observation is consistent with the wg-like distribution of lbe transcripts in nkd⁻ (Fig. 6D) and ptc⁻ embryos (Fig. 6E). In particular, in ptc mutants, the expanded wg domain is limited anteriorly by a narrow ectopic en stripe (Ingham et al., 1991) and the lbe expression territory broadens to the same limit (Fig. 6E). A similar but unlimited expansion of the lbe domain was observed in nkd mutants (Fig. 6D). nkd is known to repress en autoactivation (Heemskerk et al., 1991). Since lbe expression is lost in wg mutants, we tested embryos mutant for arm that is thought to abolish transduction of Wg signalling in target cells. lbe expression decays in the majority of epidermal cells of arm⁻ embryos (Fig. 6F) suggesting that the Wg signalling pathway is necessary for lb activity.

To further confirm the dependence of lb genes upon wg activity, we analysed hs-gsb and hs-gsb;wg⁻ embryos (Fig. 6G–I). Previous studies of Li and Noll (1993) have shown that ubiquitous expression of gsb ectopically activates the endogenous gsb gene in cells located anteriorly to the wild-type stripe (Fig. 6G). This ectopic induction, however, was not observed in a wg⁻ background (Li and Noll, 1993). As shown in Fig. 6I, hs-gsb is also able to activate ectopic lbe stripe formation and as for gsb, this phenomenon is wg-dependent and cannot be detected in hs gsb/wg⁻ embryos (Fig. 6J). Therefore, it is likely that wg function is required for both activation and maintenance of lb expression.

The maintenance of wg expression becomes gsb-dependent during segmental groove formation (about 6 hours AEL), (Li and Noll, 1993). However since at this time the dorsal gsb stripes decay, the wg-gsb autoregulatory loop may be restricted to the ventral region. Indeed, our analysis of wg (Fig. 7A,B) and lbe (Fig. 7D,E) expression patterns in gsb⁻ embryos clearly shows that in the dorsal epidermis (Fig. 7A,D), both wg and lbe are gsbindependent. Furthermore, wg and lbe expression in gsb mutants, also persists in the epidermal domains of the head and tail, in particular, the labrum and anal plate (Fig. 7D,E). In contrast, in wg⁻ embryos, lbe (Fig. 7C) and lbl (Fig. 7F) activity completely disappears from the epidermis, including the head segments and the anal plate. The central nervous system seems to be the only domain of lb expression which is only partially or not affected by wg loss of function (Fig. 7C,F). This observation suggests that, from about 6 hours AEL, maintenance of wg expression in the dorsal epidermis, does not involve gsb but may require other transcriptional regulators such as the lb gene products.

In order to test the influence of lb activity on wg expression we used embryos homozygous for Df(3R)eF1 (Fig. 8F-J) that lacks both lbe and lbl gene products (Fig. 8H). These embryos were found to lack Wg protein in the labrum and anal plate (Fig. 8F) and to have reduced levels in the dorsal epidermis from 8 hours AEL (Fig. 8G). These domains correspond to the region of gsb independent expression of wg and lb (Figs 5, 7). Consequently, Df(3R)eF1 embryos do not develop the anal plate (Fig. 8J) and display defects in the dorsal cuticle (Fig. 8I). The most affected region of the dorsal cuticle corresponds to the wg-dependent (Bokor and DiNardo, 1996) type 4° denticles (reduced number and abnormal pigmentation) but modifications appear also in type 3° cells. This cuticular phenotype is not detected in embryos carrying Df(3R)eGp4 (Fig. 8A-E) or Df(3R)GC14 (Mohler and Pardue, 1984, see Fig. 1A) which delete together the same genomic region but retain bap and lb loci. This suggests that the phenotype observed in Df(3R)eF1 embryos is due to a loss of either bap or lb function. A loss of bap function does not seem to be responsible for the cuticle phenotype of Df(3R)eF1, since a wild-type cuticle pattern was observed in bap²⁰⁸ embryos (not shown). That it is due to lb was shown by rescue experiments in which continuous expression between 4 and 9 hours AEL of lbl gene is sufficient to restore terminal wg expression (not shown) and anal plate formation (Fig. 8K). In the dorsal epidermis wg expression is restored partially (not shown), suggesting that in this region wg activity requires two lb gene products and cannot be fully rescued by one of them. Together this data indicates that lb genes maintain late wg expression in dorsal and terminal epidermis and that a wg-lb regulatory feedback loop is required for anal plate formation. The function of wg, late, in the dorsal epidermis is not fully understood, but whatever its role, lb genes seem to be required for the broadening of the wg expression domain in this region during germ band retraction (Fig. 8B).

Ubiquitous lb expression was induced by heat-shock treatment of transgenic embryos carrying lbe, lbl or both coding sequences under an hsp70 promoter (see Materials and methods). Since lb genes are required for late wg expression in dorsal epidermis (Fig. 8) we focused our analysis on this region. Triple heat-shock treatments (15 minutes each) administered between 4 and 9 hours AEL on control wild-type embryos had no effect on lbe (not shown) or wg expression (Fig. 9A) and dorsal cuticle pattern (Fig. 9B). The same treatment of hs-lbe embryos led to uniform lbe expression and induced ectopic activation of wg transcripts in dorsal epidermal cells (Fig. 9C, arrows). Moreover, hs-lbe embryos displayed abnormalities in the dorsal denticle pattern (Fig. 9D). Similar, although stronger, ectopic expression of wg (Fig. 9E) and cuticle alterations (Fig. 9F) were observed after simultaneous heat-shock induction of both hs-lbl and hs-lbe transgenes. The ectopic dorsal expression of wg is reminiscent of that induced by hs-gsb in the ventral region (Li and Noll, 1993). However, unlike the ventral cells after hs-gsb treatment, the dorsal cells do not form ectopic wg stripes. They are arranged anteriorly to the most lateral wg expressing cells from the dorsal stripes (Fig. 9C,E). The hs-lb dorsal cuticle is affected in the region overlapping the segmental border row of cells (row 1°) and cells just posterior to the row 1° (for a definition of cell types see Bokor and DiNardo, 1996). We observe that smooth cuticule adjacent to the row 1° is absent and fate of type 3° cells altered and difficult to distinguish from that of type 1° (Fig. 9D,F). This cuticle phenotype of hs-lb embryos may result from ectopic expression of wg and in consequence from alterations in Hh signaling.

In this report we show that lbe and lbl, members of the 93E homeobox gene cluster, code for highly related nuclear proteins. They play an important role in tail development and maintain late wg expression in dorsal epidermis. The high similarity of the predicted DNA-binding domains strongly suggest that lbe and lbl may recognize common target sequences and collaborate in the regulation of downstream genes.

Although lbe and lbl genes have a common epidermal expression domain, the lbe transcripts appear earlier and are more abundant. Like the majority of segment polarity genes, lb genes display specific expression in the underlying CNS. In contrast to the segment polarity genes which are activated at the blastoderm stage (for review see Perrimon, 1994), epidermal expression of lb genes appears later at the extended germ band stage. We have found a striking similarity between the expression patterns of wg (Baker, 1987; Ingham and Hidalgo, 1993; van den Heuvel et al., 1993) and lbe gene. The expression domains coincide in the terminal regions of the body (labrum and anal plate), and are both restricted to laterally interrupted stripes in the trunk. The wg and lb epidermal expression domains differ in the most posterior abdominal segments (A8 and A9) where lb activity is absent in the dorsal cells. In an attempt to understand the regulatory events that direct lbe gene expression we have analysed a set of embryos carrying mutations affecting wg function. We found, that embryos lacking functional products of wg, en and the other components of Wg and En/Hh signaling pathways (Perrimon, 1994) show misexpression of the lb genes that follows changes in wg pattern. As determined by analysis of hs-gsb and hs-gsb/wg⁻ embryos, the induction of ectopic lbe gene expression is wg dependent and cannot be detected in wg⁻ context. Since Arm, a component of Wg signaling (Perrimon, 1994), is required for lb activity, we conclude that the Wg signal activates and maintains the epidermal expression of lb genes.

Embryos carrying a null mutation of both lb genes, do not show wg-like cuticle phenotype in the ventral epidermis, probably because wg expression in this region is maintained by gsb (Li and Noll, 1993) and does not require lb function. However, the anal plate does not develop and dorsal cuticle shows alterations of wg-dependent hairs (Bokor and DiNardo, 1996). In these regions, Wg protein decays suggesting a mutual requirement for lb and wg genes. The wg-lb regulatory feedback loop in the anal plate (Fig. 10) appears at the same time (about 6 hours AEL) as the wg-gsb loop in the ventral epidermis (Li and Noll, 1993) and seems to supply the information required to specify the anal plate cells from the non-differentiated epidermal tail cells. lbe and lbl gene activity in the anal plate is likely to be activated by the homeotic gene fork head (fkh) (Weigel et al., 1989) which governs terminal development. Alternatively, the homeobox gene caudal (cad) required for the anal pads, tuft and anal sense organ formation (Macdonald and Struhl, 1986), could be part of the genetic circuitry that switches on the wglb autoregulatory loop in the terminal region.

Mutual activation of lb and wg genes was observed at distinct times in the anal plate and dorsal epidermis (Fig. 10). In the dorsal epidermis, wg becomes dependent on lb genes during germ band retraction (at about 8 hours AEL). Temporal asymmetry between the appearance of the wg-gsb autoregulatory loop in the ventral epidermis (Li and Noll, 1993) and that of wg-lb in the dorsal epidermis suggests that other factors, as yet unknown, may maintain wg expression in dorsal epidermis between 6 and 8 hours AEL. Since the late wg function (after 9 hours AEL) in the dorsal region is unknown, we can only speculate about a role for the lb-wg autoregulatory loop in this region. The most attractive possibility is that the late Wg signal in the dorsal epidermis, like the early one (Wu et al., 1995), is required for differentiation of the underlying heart mesoderm. This hypothesis seems to be supported by the observation that lb-wg interactions are restricted only to the seven abdominal segments in which heart develops. Following the wg-lb regulatory interactions, the dorsal lb-dependent wg expression domain broadens anteriorly and posteriorly, suggesting self-propagation of the lb-wg autoregulatory loop. In this case, the secreted Wg protein (van den Heuvel et al., 1993) might activate lb in neighbouring cells where lb gene products, in turn, may switch on wg expression. In contrast, the ventral wg expression domain, maintained by the wg-gsb autoregulatory loop, is restricted anteriorly by the repressing activity of ptc, and posteriorly the activity of en (Li and Noll, 1993). As a consequence of this dorsal wg expansion, only cells at the segment boundaries do not express both wg and lb genes. The ectopic expression of Lb proteins in these segment boundary cells, after heat-shock induction, results in severe defects of the dorsal cuticle pattern.

The tandem organisation of lb genes appears to be specific for Drosophila and was not evolutionarily conserved, since in mouse and human we detected only one orthologous lb gene per locus (Jagla et al., 1995). This could indicate a particular protection for developmental decisions involving these two Lb proteins in Drosophila. Here we show that lbe and lbl are related by their structure and almost identical expression patterns. Our analysis of phenotypes generated by the ubiquitous expression of both lb genes (unpublished observations) and rescue experiments in which lbl gene product was sufficient to replace two lacking Lb proteins have led to the conclusion that lbe and lbl are functionally redundant. As a consequence of this redundancy, we failed to identify null lbe and lbl alleles among lethal 93E EMS (Azpiazu and Frasch, 1993) and insertional P-element mutants (Spradling et al., 1995). In this report, we focused our analysis on the epidermal lb functions and showed that they are required for late Wg signaling in the anal plate and dorsal epidermis. Although tail development requires lb function, we presume that lb genes are also involved in differentiation of mesodermal and CNS lineages in which they are specifically expressed. This aspect remains to be investigated.

We are grateful to Pierre Chambon for the support, Manfred Frasch for the 93E EMS alleles, Joël Acker for expression vectors, Yves Lutz for the generation of anti-Lb antibodies and Marie-Louise Nullans for excellent technical assistance. We thank also Marc Bourouis, Pascal Dollé and Pat Simpson for critical reading of manuscript. This work was supported by grants from the Ministère de la Recherche, the CNRS, the INSERM, the Fondation pour la Recherche Médicale and the Association pour la Recherche sur le Cancer.