Heart tube patterning in Drosophila requires integration of axial and segmental information provided by the Bithorax Complex genes and hedgehog signaling

The Drosophila heart is a pulsatile organ, composed of a simple linear cardiac tube subdivided into an ‘aorta’ in the anterior region and a ‘heart’ in the posterior region (Rizki, 1978). The heart beat ensures the flow of hemolymph – insect blood – in an open circulatory system. In the adult, hemolymph enters the cardiac cavity through four pairs of valve-like ostiae situated in the heart and is propelled in a caudal to rostral direction by the contractions of the heart muscle (Rizki, 1978; Curtis et al., 1999; Molina and Cripps, 2001).

The general organization and structure of the larval cardiac tube are similar to that of the adult but with some differences (Curtis et al., 1999; Molina and Cripps, 2001). In particular, there are three pairs of ostiae, which are genuine openings, as opposed to the four differentiated adult valve-like ostiae. In addition, the alary muscles, which support the heart, are less developed in the adult than in the larva and an additional layer of longitudinal muscle cells has been added in the ventral region of the adult cardiac tube (Rizki, 1978; Molina and Cripps, 2001).

The Drosophila cardiac tube is composed of two cell types: cardiac cells, arising from cardioblast precursors, and pericardiac cells. These two cell types differentiate from segmentally arranged clusters of progenitor cells which are located in the dorsal-most strip of mesoderm. The mesenchymal cardiac progenitor cells undergo a mesenchymal-epithelial transition, during germband shortening, to generate two bilateral rows of epithelial cardioblasts that are flanked by pericardiac cells (Dunin-Borkowski et al., 1995; Frémion et al., 1999). Through the process of dorsal closure, the two rows join at the dorsal midline to form the lumen of the dorsal vessel, which is surrounded by irregularly arranged pericardiac cells (Ruggendorff et al., 1994; Zaffran et al., 1995). From that stage onwards, the cardiac tube is composed of repeated units, or segments, each containing six pairs of cardiomyocytes, with the exception of the most posterior segment that contains only two pairs of cells (Bodmer and Frasch, 1999). The segments of the cardiac tube are in register with the segments of the epidermis and are accordingly annotated T3, A1-A8. Each segment boundary is defined by the position of the alary muscles that are attached, on one side, to the segment borders and, on the other side, to the two most posterior cardiomyocytes that express svp (Gajewski et al., 2000; Lo and Frasch, 2001; Molina and Cripps, 2001).

The expression of several genes is metamerically repeated in cardioblasts suggesting an intrasegmental specification of cell identity and potential functional diversification. Within each segment, the four anterior-most pairs of cardioblasts express tinman (tin), β3-tubulin and D-sulfonyl-urea receptor (D-sur) (Bodmer and Frasch, 1999; Kremser et al., 1999; Nasonkin et al., 1999) whereas the two posterior-most pairs express seven up (svp) and Tb66F2 (also known as Dorsocrss1) D-sur (Gajewski et al., 2000; Lo and Frasch, 2001). The two anterior-most pairs of tin-expressing cardioblasts also express ladybird (Jagla et al., 1997). All cardioblasts undergo myogenic differentiation and express D-Mef2 and myosin heavy chain. Based on this combination of molecular markers, the cardioblasts may be subdivided into at least three different subpopulations that could respond to specific developmental programs and differentiate into cells with specific physiological functions. Such a concept is illustrated by the differentiation of ostiae from svp-expressing cardioblasts (Molina and Cripps, 2001).

Inductive extrinsic and intrinsic signals dictate and co-ordinate myocardial patterning. It is generally acknowledged that, in the segmented mesoderm, mesodermal precursor cells differentiate into specific tissues depending on their position within the segment (Azpiazu and Frasch, 1993; Riechmann et al., 1997). According to this view, the anterior region of the dorsal mesoderm in each segment (A domain) gives rise to the cardiogenic region within which cardioblasts respond to two inductive signals secreted by the dorsal ectoderm, namely Decapentaplegic (Dpp) and Wingless (Wg) (Frasch, 1999). The cardioblasts that express svp, which are integrated into the cardiac tube, have been suggested, however, to originate from the posterior domain (P domain) of each segment (Frémion et al., 1999; Gajewski et al., 2000). The information provided by these extrinsic signals is complemented by intrinsic effectors, including the transcription factors sloppy paired, even skipped and muscle specific homeobox gene (msh/Dr; Drop) (d’Alessio and Frasch, 1996; Halfon et al., 2000; Lee and Frasch, 2000). Concomitantly, a further level of diversity is achieved, at least in part, by asymmetric cell divisions (Park et al., 1998; Gajewski et al., 2000; Halfon et al., 2000; Ward and Skeath, 2000).

Finally, more general positional information along the anteroposterior axis is superimposed on this segmental information to complete patterning and differentiation of the cardiac tube. The mature larval cardiac tube broadens in its posterior region to generate the heart with a large cavity or lumen. The heart beat originates from a pacemaker activity in the most posterior region of the tube (Rizki, 1978). The heart is the only part of the larval tube where ostiae differentiate (Molina and Cripps, 2001). A cardiovascular valve develops at the junction between the aorta and heart, in segment A5, the role of which is probably to prevent a backflow of hemolymph (Rizki, 1978; Molina and Cripps, 2001).

Fig. 1 is a schematic synthesis of our current knowledge of the morphology of a fully mature larval cardiac tube.

In the absence of well-characterized molecular markers with which to investigate the anteroposterior subdivision of the cardiac tube, the aim of this work has been to develop tools and methods to study the differentiation and acquisition of cardioblast diversity. Such knowledge is essential for an understanding of the genetic programs underlying functional organogenesis. We show here that the homeotic genes of the Bithorax Complex (BXC), Ubx and abd-A, are responsible for the aorta and heart cardioblast identities, respectively, thus providing instructional genetic information along the anteroposterior axis. In addition, part of the segmental information results from secretion of the Hedgehog morphogen, which induces svp expression in the two most posterior pairs of cardioblasts in each segment. The superimposition of such segmental and axial information is likely to trigger the differentiation of ostiae from the svp-expressing cells exclusively in the heart region.

The Drosophila mutant lines used in this study were: hh^9K (temperature-sensitive), hh^AC (Heemskerk and DiNardo, 1994), svp^e22, svp^AE127 (Kerber et al., 1998), abd-A^m4 (Karch et al., 1990), Ubx^9.22 (Kerridge and Morata, 1982). Df(3)Ubx¹⁰⁹/Dp(3;3)P5,Sb(1) (Ubx, abd-A double mutant) was obtained from the Bloomington Drosophila Stock Center. Homozygous mutant embryos were distinguished from heterozygous embryos in different genotype backgrounds either by the lack of Act-GFP or Dfd-lacZ markers carried by the TM3 balancer, or by the lack of 2^nd and (or) 3^rd midgut constrictions, or by the absence of oenocytes labeled by anti-Prc.

The following UAS and GAL4 lines have been used: UAS-Ci^rep (generous gift from Dr T. Kornberg), UAS-Hh (generous gift from Dr P. Ingham), UAS-HhGPI (Burke et al., 1999), UAS-abd-A (second chromosome), UAS-Armadillo S10 (Pai et al., 1997), UAS-Ubx and UAS-mCD8GFP (second chromosome) (Bloomington Drosophila Stock Center), twist-GAL4 (Greig and Akam, 1993), 24B-GAL4 (Brand and Perrimon, 1993), en-GAL4 (Bloomington Drosophila Stock Center).

The strains w;wg^cx4UAS-Armadillo^S10/Cyo, hh^AC,UAS-Hh/TM2 Ubx and hh^AC,UAS-HhGPI/TM3, Sb,Ser have been constructed by recombination and verified. Several recombinants were obtained in each class and behave in the same manner.

In situ hybridization was performed as described previously (Frémion et al., 1999), using digoxigenin (DIG)-labeled antisense RNA probes specific for svpI cDNA (Mlodzik et al., 1990). Fixed, staged embryos were stained as described by Frémion et al. (Frémion et al., 1999) with the following primary antibodies: mouse or preadsorbed rabbit anti-β-galactosidase (Promega and Cappel, respectively) 1:1000; rabbit anti-Tinman (Azpiazu and Frasch, 1993) 1:800, preadsorbed; mouse anti-Pericardin (EC11) (Zaffran et al., 1995) 1:2; rabbit anti-D-Mef2 (Nguyen et al., 1994) 1:1000, preadsorbed; rabbit anti-α-Spectrin (Lee et al., 1993) 1:500, preadsorbed; mouse anti-En (4D9, Developmental Studies Hybridoma Bank) 1:4; mouse anti-Abd-A (Karch et al., 1990) 1:1000; mouse anti-Ubx (White and Wilcox, 1985) undiluted; mouse anti-Abd-B (Celniker et al., 1989) 1:100. Affinity purified secondary antibodies were either coupled to alkaline phosphatase or to biotin (Jackson Immuno Research Laboratories) and used at a 1:1000 dilution or were either Alexa-488 or Alexa-594 conjugated (Molecular Probes, Inc.) and used at a 1:500 dilution. In some cases, the signal was amplified with the aid of a Tyramide Signal Amplification kit (NEN Life Sciences). The stained embryos were mounted in Geltol Medium (Immunotech, France) or, when fluorescent, in Vectashield (Vector Laboratories) for further observation under an Axiophot Zeiss microscope or a LSM 410 Zeiss confocal microscope.

Wandering third instar larvae were removed from the culture vessel and initially pinned through the head to a Petri dish containing Sylgard 184 (Dow Corning Corp., Midland, MI, USA) with the dorsal side down. A longitudinal incision was made along the ventral midline using fine scissors, and the body wall was stretched and pinned out to generate a flat fillet. Viscera were removed, taking care not to dislodge the heart in its location at the dorsal midline. The nervous system and other anterior structures including the imaginal discs were left intact because the anterior end of the heart is not attached to the dorsal body wall and the heart would otherwise move too freely.

All dissections were carried out in phosphate-buffered saline (PBS). Samples were fixed in 3.7% formaldehyde in PBS for 20 minutes on ice, then washed twice for 5 minutes in PBS and further processed for antibody labeling as indicated above.

The embryos were dechorionated, glued onto glass slides with sticky tape, covered with a drop of oil (Voltalef 10S) and observed under an inverted Olympus CK40 microscope, with a ×60 objective. Films were recorded, using Adobe Premiere, every 15 minutes between 15 and 23 hours of embryonic development at 25°C.

Whereas the morphology and physiology of the mature larval heart have been extensively studied, little is known about functional cardiogenesis that takes place during embryogenesis. We have developed suitable tools to observe the morphology of cardiomyocytes along the anteroposterior axis and describe their features with precision. We have chosen morphological criteria that are more appropriate than markers of expression to investigate the organogenesis and differentiation of the cardiac tube. Whole-mount embryos labeled with antibodies were observed under a confocal microscope. The antibodies used include anti-α-Spectrin (Lee et al., 1993) that marks the basal-lateral membrane of epithelial cardiomyocytes (Zaffran et al., 1995; Frémion et al., 1999), anti-Pericardin (Chartier et al., 2002) that labels the basal membranes of the cardiomyocytes, anti-β-galactosidase that detects a lacZ reporter gene product inserted in the svp gene, which reflects svp expression (Gajewski et al., 2000; Lo and Frasch, 2001) and anti-DMef2 (Bour et al., 1995) that labels the nuclei of all myogenic cells including cardioblasts.

Cardioblasts undergo a mesenchymal-epithelial transition during embryonic stages 12 and 13, which results in the formation of two bilateral monolayers of cardioblasts situated beneath the most dorsal ectodermal cells, on either side of the dorsal aperture (Zaffran et al., 1995; Frémion et al., 1999). This and subsequent morphological steps occur without further cell division. At these stages, cardioblasts display no evident signs of heterogeneity with respect to their size or morphology along the whole length of the cardiac tube. At stage 14, during dorsal closure, when the two rows of cardioblasts migrate towards the dorsal midline in a movement coordinated with that of the ectoderm (Chartier et al., 2002), clear morphological differences become apparent (Fig. 2). Cells in the presumptive heart region increase in size, elongate along the transverse axis and adopt a columnar shape, while the cells in the aorta region remain smaller and more cuboidal (Fig. 2D,E). Likewise, the round nuclei of the cardioblasts become larger in the heart than in the aorta (Fig. 2K,L). Moreover, differences among the cardioblasts within the heart region begin to emerge. Three pairs of metamerically repeated cells in both rows of cardioblasts develop more even cell contours than the other cardioblasts and become thinner. Their nuclei elongate, adopting an egg-like shape.

The cardiac tube of stage 16 embryos displays the same general morphology as the larval cardiac tube and presents evidence of differentiation along the anteroposterior axis. The cardiac cells are two to three times larger in the heart than in the aorta and this leads to a broadening of the cardiac tube in the heart region (Fig. 2F,G). At the same time, seven segmentally repeated pairs of cardioblasts per half embryo adopt a morphology that distinguishes them from the other cardioblasts, both in the heart and in the aorta. Their nuclei are more elongated (Fig. 2M,N) and the cells are thinner and have less sharp edges than the other cells of the cardiac tube. In addition, these cells are unambiguously larger in the heart than in the aorta.

These seven pairs of cardioblasts express svp (Fig. 2O,P) and lie just beneath the alary muscles, which can be precisely localized by their labeling with anti-Prc (Fig. 2G,L). The svp-expressing cells differentiate in the larva into two distinct structures depending on their position along the anteroposterior axis (Molina and Cripps, 2001). In the aorta, the svp-expressing cells form a butterfly-like structure that evenly interrupts the epithelial sheet constituting the cardiac tube (Fig. 3A), without differentiating into functional ostiae since there are no openings between the two cells (Fig. 3C). In contrast, these same cells give rise to larval ostiae in the heart region (segments A5-A7) (Fig. 3B,D) (Molina and Cripps, 2001).

Analysis of later differentiation of cardioblasts by immunohistochemistry is considerably hampered by the concomitant differentiation of an insulating cuticle, which restricts the penetration of the antibodies into the embryos. We have therefore studied the late differentiation of cardiac cells by direct observation of the development of the cardiac tube in living embryos. Moreover, in vivo observation offers the additional advantage of being able to monitor the acquisition of cardiac function, namely heart beating and ostiae function.

Fig. 4 shows a series of photographs extracted from a video movie that are representative of the main features of embryonic heart development (movie available at http://dev.biologists.org/supplemental/ ). Immediately after the tube has closed, the cardiomyocytes became columnar and discernible size differences are already observed between aorta and heart. Rapidly, around 17 hours after fertilization at 25°C, the first signs of muscular activity are observed in the posterior region of the heart. Simultaneously, the lumen in the heart region dramatically increases in size and the cell wall of the cardiac cells become thinner, the cells change from columnar to a more corrugated form that results from the protrusion of the nuclei of tin-positive cells into the lumen. In contrast, the aorta does not exhibit many signs of differentiation. A lumen forms but remains small and no autonomous muscular activity is perceptible. Fig. 4E-V depicts the cardiac tube of embryos shortly before hatching.

Observation of living embryos led to the conclusion that ostiae were functional very soon after the first signs of heart muscle activity. They open and close, allowing the entry of rare and large hemocytes that, once in the cardiac tube, are readily propelled towards the aorta (Fig. 4 and movie). The two most posterior pairs of ostiae are the first to form and to become functional. The aperture of the ostiae is regulated by the movement of the two component cells as illustrated in Fig. 4M-V. A filamentous network links together the ostiae cells on either side of the cardiac tube. This filamentous material might serve to isolate the different compartments or chambers of the heart. The hemocytes stay for some time in a chamber, moving backwards and forwards like a ping-pong ball, before eventually being expelled into the aorta.

The recording of heart activity in living embryos has also provided the opportunity to measure another important aspect, namely the sizes of end-diastolic and end-systolic diameters. We calculated these values at eclosion (22 hours at 25°C) to be 22.6 μm and 15.1 μm, respectively.

These results establish morphological and functional criteria that have been used to subsequently characterize the state of differentiation of subpopulations of cardioblasts in various genetic backgrounds.

We hypothesized that BXC homeotic genes are likely candidates to control patterning of the cardiac tube along the anteroposterior axis, primarily because these genes are responsible for axial patterning of the whole organism. Secondly, Ubx and abd-A have been reported to be expressed in the cardiac tube in which they can be expected to display a regionalized expression profile (Karch et al., 1990; Bate, 1993). We have analyzed in detail the sites of expression of BXC proteins in the embryonic cardiac tube. The sites of Abd-A and Abd-B expression were analyzed with respect to the svp-expressing cells, which were identified by monitoring lacZ reporter gene activity in the svp^AE127 line. To investigate Ubx expression, cardioblasts were double-labeled with anti-Ubx and anti-Tin in wild-type embryos instead of svp^AE127 embryos to circumvent the fact that the TM3 balancer carries a mutation in Ubx that might affect the wild-type level of Ubx expression.

The level of expression of Ubx in the cardiac tube was low compared to that in the ectoderm or gut. Strongest expression was observed in the cardiac and pericardiac cells of the aorta, from segment A2 to the middle of segment A5 (Fig. 5A). Weaker expression was detected in the more anterior segments A1 and T3 and only extremely low expression was observed more posteriorly, in the heart region.

In contrast, Abd-A expression was observed in all the cardioblasts in the heart region, except in the most posterior cardioblasts in segment A8 and in the two pairs of most anterior cardioblasts in segment A5, in which the expression of Abd-A was considerably lower (Fig. 5D). The region of strong Abd-A expression encompasses the two pairs of cells expressing svp in segment A5 anteriorly and, the two last pairs of svp-expressing cells that belong to segment A7 posteriorly. A subpopulation of pericardiac cells present in segments A5 to A7 also strongly expresses Abd-A.

Abd-B expression in the cardiac tube is restricted to the 4 most posterior cardioblasts in segment A8 (Fig. 5G). The two pairs of svp-expressing cells in segment A7 also express Abd-B, although to a lesser extent.

This specific anteroposterior distribution of the three BXC gene products in the cardiac tube was analyzed in different mutant backgrounds. In embryos homozygous for a null abd-A mutation (abd-A^m4), Ubx expression extended into the heart region (Fig. 5B). In contrast, Abd-A expression, in embryos homozygous for a null mutation in Ubx (Ubx^9.22) was similar to that in wild-type embryos (Fig. 5E). Likewise, Abd-B expression was not affected in abd-A mutants or in Ubx,abd-A double mutants (Fig. 5H,I).

Overexpression of Abd-A in the whole mesoderm, including cardiac cells, using twistGAL4 as driver, led to a significant decrease of Ubx expression in the aorta (Fig. 5C). In contrast, ectopic overexpression of Ubx did not result in any significant change in Abd-A expression (Fig. 5F).

In homozygous abd-A^m4 mutant embryos, in contrast to the situation encountered in wild-type embryos, the general morphology of the cardioblasts was uniform along the whole length of the cardiac tube even at late stage 16. All the cells had the configuration of aorta cardioblasts, including the svp-expressing cells which displayed the features of svp-aorta cells (Fig. 6A,B). This observation was further supported by analysis of cardiac tube development in living embryos. In abd-A mutants, the heart region never differentiated and the entire cardiac tube remained as an aorta (Fig. 6C,D). Cardiac cells in the heart region did not modify their shape nor did the heart beat. No sign of ostiae function was observed.

In contrast, in Ubx loss-of-function mutant embryos, heart cardioblasts were properly differentiated and the function of the heart was similar to that in wild-type embryos (not shown). The most anterior part of the aorta showed, however, some perturbations (compare Fig. 6E and 6F). Polarization and thus differentiation of the cardioblasts are not clearly apparent in segments T3 and A1 in Ubx mutant embryos as is the case in wild-type embryos (Fig. 6E,F). These features seem, however, to extend more posteriorly to segment A2 in mutant embryos (Fig. 6E). In addition, pericardiac cells were disorganized in several positions, particularly along these anterior segments.

In Ubx-abd-A double mutants, the early stages of cardiac tube morphogenesis proceeded as in wild-type embryos but later differentiation of cardiomyocytes was impaired. The number of cardioblasts in the cardiac tube as well as in each segment was not modified and the segmental expression of svp was normal (Fig. 6I,J). No heterogeneity in the morphology of the cardioblasts could be observed along the anteroposterior axis nor did the tube differentiate into heart and aorta (Fig. 6G,H). The cells were not conspicuously polarized along the whole length of the cardiac tube and had a configuration similar to that of the cells in the anterior-most segments (T3 and A1) of wild-type and Ubx mutant embryos. Prc-expressing cells were disorganized as in Ubx single mutations. Other cells that did not express Prc and that were hypothesized to be lymph gland cells were found in ectopic locations along the tube.

All these observations suggest that in the absence of Ubx function, aorta cardiomyocytes could be transformed in cells and structures normally present in segments more anterior than A2 in wild-type embryos.

Overexpression of abd-A in the whole cardiac tube resulting in a gain-of-function of abd-A in the aorta led to a partial transformation of aorta cardioblasts into heart cardioblasts (Fig. 6K). The most posterior segments of the aorta (A4 and A5) were invariably transformed into heart, including the svp-expressing cells that give rise to functional ostiae in these segments (Fig. 6L,M). Although varying from embryo to embryo, the transformation of more anterior segments into a heart phenotype was more or less complete. Also, overexpression of Ubx significantly impaired the general morphology of the cardioblasts particularly in the heart region (Fig. 6N). In living embryos, the heart lumen and the timetable for heart beat appeared normal. An important delay was observed in cardioblasts differentiation (Fig. 6O,P) and the development of ostiae was perturbed and they never became functional.

Complete loss-of-function of svp (svp^e22 or svp^AE127) resulted in embryonic lethality. However, the development of the embryos proceeded as in wild-type embryos although muscular movements were impaired at the end of embryogenesis and the embryos never hatched and eventually died. The cardiac tube (Fig. 7A-D) differentiated properly into aorta and heart and cardiac muscular activity was observed, although with some delay. Similarly, the ostiae cells exhibited a delay in their shape remodeling and appeared more like tin-positive cardiomyocytes (Fig. 7A-D) in stage 16 embryos. In 22-hour-old svp mutant embryos, the ostiae cells regained part of their characteristic morphological features, but were still incompletely differentiated (Fig. 7F,G). They were not functional as evident from a lack of opening and closing and no hemocytes were ever observed entering the apertures. Finally, the loss of svp function did not affect the limits of the expression domains of the BXC genes, including abd-A (Fig. 7E). These observations suggest that the function of svp is required for functional ostiae to differentiate properly but that svp does not participate in the process that leads to differential patterning of the aorta and heart.

The segmental diversity of the cardioblasts is unlikely to result from the action of the BXC genes, since the pattern of segmental expression of tin and svp were not altered in Ubx and abd-A mutants (see Fig. 5 and Fig. 6). A more likely hypothesis is that segmental diversity relies on a segmental information, which could be delivered by inductive morphogens secreted by the segmented ectoderm. svp expression in the cardiac tube is restricted to the two most posterior pairs of cardioblasts in all segments from A1 to A7 (Gajewski et al., 2000; Lo and Frasch, 2001) (Fig. 8A-C), a position that could be influenced by Hedgehog signaling, which is secreted by cells in the posterior domain of ectoderm parasegments.

In hh mutant embryos, the expression of svp RNA was dramatically reduced or even abolished (Fig. 8E). A temperature-sensitive mutant allele of hh, hh^9k, has been used in our experiments and the temperature shift was performed at a time where the activities of the hh and wg signaling were no longer dependent on each other (Bejsovec and Martinez-Arias, 1991) such that the observed effects are due to specific inhibition of hh signaling activity. In parallel, Tin expression was ectopically extended (Fig. 8F) due at least in part, to the repression of tin expression by svp (Gajewski et al., 2000; Lo and Frasch, 2001). It has been shown that tin and svp expression in cardioblasts are exclusive (Gajewski et al., 2000; Lo and Frasch, 2001). In a svp mutant embryo, tin is ectopically expressed in the svp-expressing cardioblasts, and reciprocally, ectopic expression of svp represses tin expression. The expression of Engrailed (En) was taken as a reference for the position of the segments.

The decrease in svp expression in hh mutants was confirmed by expressing a repressor form of Cubitus interruptus (Ci^rep) (Aza-Blanc et al., 1997) in the whole mesoderm, including cardiac precursor cells, using twist-GAL4/24B as driver. In this situation, in which Ci^rep inhibits hh signaling in the cells that receive it, svp expression was significantly reduced in the cardioblasts (Fig. 8G), although not completely abolished and the reduction was less effective than that observed in a hh^9k mutant embryo. One explanation could be that part of the hh signal was transmitted by a Ci-independent pathway (Gallet et al., 2000; Apidianakis et al., 2001) or, more likely, low level expression of Ci^rep in the svp-expressing cardioblasts was insufficient to fully repress hh signaling. Nonetheless, the marked reduction of svp expression in repressed cardioblasts suggests that hh signaling is required for the activation of svp expression in cardioblasts.

In support of this conclusion, the lack of svp could be rescued by expressing hh in the posterior compartment of the ectodermal segments, using the enGAL4/hhUAS system in a hh^AC null mutant background (Fig. 8H). It should be noted that no rescue of svp expression was obtained by using, in the same genetic context, an Hh form that cannot be secreted and that remains bound to the membrane (Burke et al., 1999) (Fig. 8I). These results suggest that svp expression in cardioblasts is under the control of Hh produced and secreted by ectodermal cells.

Contribution of Wg signaling to svp expression was ruled out by maintaining hh expression in a wg null mutant background. A constitutively active form of Armadillo (Arm^S10) was expressed under the control of En in wg^cx4 embryos, which can be identified by the widened En positive stripes (Pai et al., 1997). In such embryos, svp expression was similar to that in wild-type embryos (Fig. 8J), confirming that Hh secretion by the ectoderm is required for svp expression in the mesoderm.

In conclusion these results indicate that the segmental diversity of the cardioblasts is provided in part by the secreted morphogen Hh.

Generation of cell diversity within an organ is critical to its physiological function. In Drosophila, the functional cardiac tube contains only 104 cardiomyocytes that diversify to generate distinct subtypes of cells with specific developmental and cellular properties. Owing to its relative simplicity, Drosophila cardiogenesis constitutes a unique model with which to investigate the genetic programs that provide cell diversity and specific cell differentiation within an organ.

Two types of seemingly independent positional information, axial information along the anteroposterior axis and segmental information, are required to pattern the cardiac tube. In regard to the first type of information, the results presented herein indicate that the homeotic genes, Ubx and abd-A, are responsible for the identity of aorta and heart cardiomyocytes, respectively. Concerning segmental information, the cardiac tube is indeed segmented and the hh signaling pathway induces (by turning on svp expression) the diversification of a subpopulation of cardiomyocytes within each segment in the tube. It is the intersection of these two types of positional information that triggers the differentiation of heart cardiomyocytes expressing both abd-A and svp, into three pairs of ostiae. The scheme in Fig. 9 summarizes these observations (Fig. 9).

The observation that the segmental diversity of cardiomyocytes was not influenced by the function of homeotic genes suggests that the two programs operate independently, in parallel pathways. The segmental restriction of svp expression to two cardioblasts per segment was unchanged in a Ubx,abd-A double mutant cardiac tube (Fig. 6). Similarly, the pattern of expression of Abd-A was maintained (Fig. 7). How information from these two independent pathways is interpreted by specific transcriptional targets will be a major goal for future studies. A likely target in the heart cells forming the ostiae could be wingless, which is expressed exclusively in the mature cardiac tube in svp-positive cardiomyocytes of the heart (our own observations) (Gajewski et al., 2001). It can also be anticipated that tin-positive cells in the heart will express specific genes responsible for the distinct physiological activity of the heart versus the aorta cardiomyocytes. One can expect that the transcription of these candidate genes will be activated by tin and abd-A but not by tin and Ubx.

The morphological and functional criteria that we have defined in this study have allowed us to subdivide cardiomyocytes into two distinct populations that acquire different identities and differentiate according to their positions along the anteroposterior axis. Ubx is expressed in almost all cardiomyocytes of the aorta whereas abd-A is expressed in almost all cardiomyocytes of the heart. The lack (or a very low level) of Ubx expression in the T3 and A1 segments of the aorta suggests that cardiomyocytes in these segments may be exposed to a distinct mode of differentiation. In support of this hypothesis, our morphological analysis has revealed distinct features in the most anterior region (segments T3,A1) of aorta. These particular traits were nonetheless difficult to unmask owing to a hindering of the aorta inside the embryo and to the presence of the ring and lymph glands surrounding the cardiac tube. Similarly, the lack of abd-A expression (and the strong Abd-B expression) in the four most posterior cardioblasts of the heart implies that these cells respond to specific genetic and differentiation programs that do not operate in more anterior heart cardiomyocytes. While no obvious morphological features permit these most posterior cardioblasts to be distinguished, their position in the caudal most region of the cardiac tube suggests that they are likely candidates to form the pacemaker center of the organ (Rizki, 1978).

In contrast to the situation in the ectoderm (Akam, 1998), the domains of expression of the BXC homeotic genes in the cardiac tube do not overlap, are contiguous and mutually exclusive. The same type of regionalized expression is also encountered in the visceral mesoderm in which, for example, Ubx is expressed in PS7 while abd-A expression encompasses the segments PS8 to PS12 (Bienz, 1994). Nevertheless, whatever the tissue, ectoderm or visceral mesoderm, the more posteriorly expressed gene represses (or dominates over) more anteriorly expressed genes, conforming to the phenotypic suppression (Gonzalez-Reyes and Morata, 1990) or posterior prevalence (Duboule, 1995) rules. Accordingly, loss-of-function of abd-A leads to a posteriorization of Ubx expression and a concomitant transformation of the heart into aorta. Similarly, the anterior boundary of the expression domains of abd-A and Abd-B were not modified in Ubx and abd-A mutants, respectively. Reciprocally, overexpression of abd-A in the whole cardiac tube repressed Ubx expression and transformed the most posterior aorta cardiomyocytes into heart cardiomyocytes. However, ectopic expression of Ubx also impaired the differentiation of cardioblasts, although to a lesser extent than when abd-A was overexpressed and it did not significantly repress Abd-A expression. This latter result suggests that Ubx and Abd-A may be in competition for common downstream targets.

In Ubx embryos, or in Ubx, abd-A double mutants, differentiation of the most anterior region of the aorta was affected. This observation suggests that, as in the visceral mesoderm, Antennapedia (Antp) might be expressed in the anterior domain (segments T3, A1) of the aorta, in which the lymph glands and the ring gland are located and that Antp transcription is repressed by Ubx in segment A2 and more posterior segments. In the absence of Ubx function, the Antp expression domain could be extended posteriorly and lead to the formation of ectopic lymph and/or ring gland cells. Finally, the fact that an additional effect on cardioblast differentiation was observed in double mutant embryos when compared with each single mutation, suggests that Ubx and abd-A participate in cardiomyocyte differentiation independently of their role in axial patterning.

The homeotic genes abd-A and Ubx are transcription factors which probably induce differential activation of particular gene networks which, in turn, could confer specific physiological function on distinct subsets of cardiomyocytes. For example, studies performed on the cardiac tube of another insect, Samia caecropia (McCann, 1970), provide good evidence that the electrophysiological properties of the cardiomyocytes are different in the aorta and in the heart. abd-A function may be necessary to activate genes responsible for heart activity or genes that participate in cardiomyocyte growth. Aorta and heart cardiomyocytes respond to a differential control of cell growth since, at the end of embryogenesis, the heart cardiomyocytes are at least two to three times larger than the aorta cardiomyocytes. Alternatively, Ubx could repress the growth of the aorta cardiomyocytes analogous to its role in haltere cells (Roch and Akam, 2000). Growth control of cardiomyocytes is probably not the unique function exerted by Ubx and abd-A in the cardiac tube, since in the absence of both gene activities the cells did not differentiate properly.

The segmentally repeated expression of svp is regulated by a positive inductive effect of Hh secreted by cells from the overlying ectoderm. A role for hh signaling in Drosophila cardiogenesis has not previously been acknowledged. It has been observed that in hh mutant embryos heart progenitors were lacking (Park et al., 1996), however this has been interpreted to be an indirect influence of hh upon wg signaling. The results reported herein strongly favour the idea that hh has a direct and positive effect on the determination and specification of the sub-population of cardioblasts that expresses svp. It has been proposed that each segment of the trunk is sub-divided into two domains, A and P (Azpiazu et al., 1996; Riechmann et al., 1997). The cells from the anterior domain (A domain) of the dorsal mesoderm would be directed towards a cardiogenic fate while cells from the posterior domain (P domain) would adopt a visceral mesoderm fate. The wg and hh signals released, respectively, from the anterior and posterior compartments of the ectodermal parasegments have been proposed to be the determinants in specification of the two domains. Our observations, however, provide strong evidence that a subtype of cardiac cells could originate from the mesodermal P domain. The P domain origin of some cardioblast progenitors had previously been suggested by Frémion et al. (Frémion et al., 1999) who reported the presence at stage 11-12, within the P domains, of bkh-expressing cells, which will contribute to the cardiac epithelium later in development. It seems, therefore, that there is not a perfect superimposition between A domains within mesodermal segments and the capacity of the cardiac cells to be integrated into the cardiac tube.

The hh signal secreted by cells belonging to the posterior compartments of the segmented ectoderm is sufficient to promote svp expression, as illustrated in Fig. 8. The Hh morphogen needs to be secreted and to freely diffuse from the ectoderm to the underlying mesoderm, as judged from the loss of svp expression in the cardioblasts when a membrane-bound form of Hh is expressed in the same genetic background in place of endogenous Hh. The existence of a specific mechanism to constrain diffusion of the secreted morphogen to the cardioblasts of the P-domain can thus be postulated. Further investigation of this mechanism will provide insight into how specificity of morphogen signaling is achieved across embryonic germ layers.

Based on gene expression patterns, Hh signaling is likely to be instrumental in the specification of tin- and svp-cardioblasts by inducing the expression of svp in cardioblasts which, in turn, leads to the repression of tin. Such a repressive action of svp has already been reported (Gajewski et al., 2000; Lo and Frasch, 2001), although a direct interaction between tin regulatory sequences and svp has not been demonstrated (Lo and Frasch, 2001).

Similar relationships between homologs of hh, svp and tin have been described in vertebrate cardiogenesis. An homolog of svp, COUP-TFII is expressed in the posterior region of the mouse primitive heart tube where it is required for heart development (Pereira et al., 1999); furthermore the expression of COUP-TFII is induced by Sonic hedgehog (Krishnan et al., 1997). Shh (and Indian hedgehog) participates in mouse cardiac morphogenesis but, in contrast to the situation in Drosophila, induces rather than represses the expression of the tin homolog, NKx2.5 (Zhang et al., 2001). It must be concluded, from these remarks, that the genetic networks can be differently interpreted and utilized in invertebrates and vertebrates. Further studies should give better insights into the conservation of the genetic programs at work in heart development.

Our thanks are due to the Bloomington Stock Center for fly stocks. We thank Drs R. Dubreuil, H. Nguyen, D. Gratecos, M. Frasch, M. Hoch, V. Brodu, S. Merabet, Y. Graba, S. Kerridge and M. Akam for their generous gifts of some of the antibodies and strains used in this work. We thank M. Mokrane, C. Moretti, and P. Weber for their helpful assistance in the production of photographs and of the movie. We acknowledge S. Long and F. Graziani for their technical assistance in food preparation and the maintenance of the stock flies. We are deeply indebted to Dr D. Gratecos for her fruitful comments and her help in the preparation of the manuscript. We thank Drs R. Kelly and S. Zaffran for their careful reading of the manuscript. This work was supported by the Centre National de la Recherche Scientifique and by grants from Association Française contre les Myopathies (AFM), Ligue Régionale contre le Cancer (LRCC) and Association pour la Recherche contre le Cancer (ARC).