The Heartless (Htl) FGF receptor is required for the differentiation of a variety of mesodermal tissues in the Drosophila embryo, yet its ligand is not known. Two FGF genes, thisbe (ths; FGF8-like1) and pyramus (pyr; FGF8-like2), have been identified that probably encode the elusive ligands for this receptor. The two genes were named for the 'heartbroken' lovers described in Ovid's Metamorphoses because the genes are linked and the mutant phenotype exhibits a lack of heart. The genes exhibit dynamic patterns of expression in epithelial tissues adjacent to Htl-expressing mesoderm derivatives, including the neurogenic ectoderm, stomadeum, and hindgut. Embryos that lack ths⁺ and pyr⁺ exhibit defects related to those seen in htl mutants, including delayed mesodermal migration during gastrulation and a loss of cardiac tissues and hindgut musculature. The misexpression of Ths in wild-type and mutant embryos suggests that FGF signaling is required for both cell migration and the transcriptional induction of cardiac gene expression. The characterization of htl and ths regulatory DNAs indicates that high levels of the maternal Dorsal gradient directly activates htl expression, whereas low levels activate ths. It is therefore possible to describe FGF signaling and other aspects of gastrulation as a direct manifestation of discrete threshold readouts of the Dorsal gradient (Stathopoulos, 2004; Gryzik, 2004).

In Drosophila two known FGF receptors, Breathless (Btl) and Htl, are required for cell migratory events, tracheal migration, and mesodermal cell migration, respectively. Despite the annotation of the Drosophila genome sequence, only Bnl, the ligand of Btl, has thus far been identified. Bnl is not expressed in the right temporal-spatial pattern to serve as a ligand for Htl, and bnl mutant embryos do not display defects in early mesoderm morphogenesis. In addition, expression of dominant-negative forms of Btl does not produce mesoderm migration defects, and upon ectopic expression of Bnl within the ectoderm, only Btl-expressing cells show ectopic activation of MAP kinase. Together, these data suggest either that the ligand of Htl was missed in the genome annotation or that Htl might not be activated by FGF-like ligands. The identification is reported of two genes encoding for novel fly FGF homologs, which exhibit features consistent with being ligands for Htl (Gryzik, 2004).

The early expression of FGF8-like1 and FGF8-like2 is restricted to the neuroectoderm and thus corresponds to the predicted source of ligands required for Htl activation. The knockdown of the function of these genes by RNAi leads to a mesoderm cell phenotype that is very similar to that produced by mutations in the Htl receptor. Thus, FGF8-like1 and FGF8-like2 act non-cell-autonomously in the early embryo and are required for mesodermal cell shape changes during gastrulation. In addition, deletion of the two genes blocks activation of MAP kinase in early mesoderm cells. Genetic mapping using isogenic deletions shows that an interval containing 14 genes is responsible for the observed phenotypes. Because RNAi experiments are able to reproduce the migration phenotype of the deletion embryos, it is proposed that FGF8-like1 and FGF8-like2 together are responsible for the mesoderm defects observed in the deletion. In conclusion, these results strongly suggest that FGF8-like1 and FGF8-like2 represent ligands for Htl (Gryzik, 2004).

FGF signaling molecules have been implicated in both the movement and specification of different mesoderm lineages in vertebrate embryos. For example, FGF4 and FGF8 are required for the invagination of epiblasts into the primitive streak of mouse embryos, whereas FGF8 and FGF24 appear to specify the posterior mesoderm cells in zebrafish embryos. Given the large number of FGF signaling molecules in vertebrates, it is often difficult to distinguish between disruptions in movement or specification. It is conceivable that FGF signaling controls both processes (Stathopoulos, 2004).

Invertebrate embryos offer an opportunity to investigate this issue because they contain relatively few FGFs. The nematode worm Caenorhabditis elegans contains two FGF ligands, Egl-17 and Let-756. Mutants lacking egl-17⁺ activity exhibit defects in the migration of the sex myoblasts to the gonad and in the differentiation of the muscle cells that surround the developing vulva (Burdine, 1998). Thus, it would appear that FGF signaling is required for both movement and specification. Mutants lacking Let-756 (Roubin, 1999) display developmental arrest early in larval stages, but the exact cause of this defect has not been determined (Stathopoulos, 2004).

Previously, the only FGF gene known in Drosophila was branchless, which is essential for the morphogenesis of the trachea, air sacs, and male genital imaginal disc. During embryogenesis, branchless is expressed in a highly dynamic fashion in discrete epithelial cells of developing embryos. The Branchless ligand triggers the activation of the Breathless FGF receptor and thereby controls the movement (branching) of the trachea. It is conceivable that this Branchless-Breathless interaction is also important for the specification of at least a subset of the tracheal cell types. Branchless does not appear to influence the specification or movement of mesoderm lineages during earlier periods of Drosophila embryogenesis (Stathopoulos, 2004).

Although only one FGF ligand has been identified, Drosophila contains two FGF receptors, Breathless and Heartless (Htl). The Htl receptor is essential for the development of various mesoderm lineages, including cardiac tissues, hindgut visceral musculature, and the body wall muscles. Htl is initially expressed throughout the mesoderm of early embryos, and its activation is thought to trigger the spreading of the mesoderm across the internal surface of the neurogenic ectoderm. The mesoderm cells that come into contact with the dorsal ectoderm receive an inductive signal, Dpp, which triggers the expression of genes such as tinman (tin) and even-skipped (eve) that are required for the differentiation of cardiac and pericardial tissues, respectively. However, the mechanism of Htl activation is uncertain (for review, see Wilson, 2000). It has been suggested that localized FGFs emanating from the neurogenic ectoderm might be responsible for Htl activation and provide an instructive cue that guides the migration of the mesoderm. An alternative view is that Htl plays a permissive role in migration by rendering the mesoderm competent to respond to an unknown localized signal (Stathopoulos, 2004).

Htl may be required both for the spreading of the mesoderm and the subsequent specification of cardiac tissues. The misexpression of Dpp throughout the ectoderm, in both dorsal and ventral regions, causes widespread activation of tin expression within the mesoderm. However, eve expression is not expanded, and it has been suggested that its activation depends on both Dpp signaling (normally achieved through spreading) and a second dorsally localized signal, possibly FGF (Frasch 1995; Carmena, 1998; Michelson, 1998b; Halfon, 2000). The analysis of the hindgut visceral musculature provides evidence for this dual role of FGF signaling in movement and specification (San Martin, 2001). The activation of Htl is required for the initial spreading of the visceral mesoderm around the hindgut, as well as the subsequent differentiation of the hindgut musculature (Stathopoulos, 2004).

To investigate the function of FGF signaling in the early embryo, Htl ligands, which have eluded intensive genetic screens, have been identified. This study identified two closely linked genes, thisbe (ths) and pyramus (pyr), which encode FGF signaling molecules that appear to function in a partially redundant fashion to activate Htl. Ths and Pyr are most closely related to the FGF8/17/18 subfamily, which controls gastrulation as well as heart and limb development in vertebrates (e.g., Maruoka, 1998; Sun, 1999; Reifers, 2000). Both ths and pyr are expressed in the neurogenic ectoderm during the spreading of the internal mesoderm in gastrulating embryos. These two genes also exhibit dynamic expression in the stomadeum, hindgut, and muscle attachment sites of older embryos. These sites of expression closely match the genetic function of htl described in previous studies. Moreover, a small deletion that removes both ths and pyr causes a variety of patterning defects, including delayed spreading of the mesoderm during gastrulation, the loss of cardiac tissues and hindgut visceral musculature, and abnormal patterning of the body wall muscles. These defects are similar to those seen for htl mutants. The ectopic expression of Ths in the early mesoderm of gastrulating embryos causes an expansion in the domain of Htl activation and a corresponding expansion in the eve expression pattern. These observations suggest that Htl controls both the spreading of the mesoderm and (along with Dpp and Wingless) the specification of pericardial cells. Computational methods were used to identify a mesoderm-specific enhancer for htl that is directly activated by peak levels of the maternal Dorsal gradient. Because ths is directly activated by low levels of the gradient, it is possible to describe gastrulation as a direct manifestation of discrete threshold readouts of the Dorsal gradient (Stathopoulos, 2004).

Several lines of evidence suggest that Ths and Pyr correspond to ligands for the Htl FGF receptor. (1) The two genes exhibit dynamic patterns of expression in tissues that influence the development of different mesoderm lineages, including the neurogenic ectoderm (early mesoderm spreading), muscle precursors (dorsal muscles, visceral muscles, and heart), hindgut (visceral musculature), and neuroblasts. (2) Mutant embryos that lack both ths⁺ and pyr⁺ gene activity exhibit defects that are quite similar to those seen in htl mutants, including a delay in mesoderm spreading during gastrulation, a reduction in dorsal mesoderm lineages, the loss of pericardial and cardial cells, the absence of hindgut musculature, and disruptions in the ventral oblique muscles. Misexpression of Ths throughout the early mesoderm causes an expansion in the Eve expression pattern, consistent with expanded induction of pericardial and/or dorsal muscle founder cells. (3) Expression of activated Htl or Ths rescues the loss of dorsal mesoderm lineages in mutant embryos. Pyr and Ths might also activate the Htl receptor at later stages of the life cycle. For example, a recent microarray screen (Butler, 2003) identified CG13194 (pyr) and CG12443 (ths) transcripts in the body wall muscle of wing imaginal disks, where htl is also expressed (Stathopoulos, 2004).

Previous genetic screens failed to identify ths and pyr, possibly because of overlap in the activities of the encoded proteins, which are closely related members of the FGF8/FGF17/FGF18 subfamily of FGF signaling molecules. Mutations in either gene alone might be insufficient to produce robust dorsal-ventral patterning defects, as seen for htl mutants. Indeed, two related FGF genes, FGF8 and FGF24, are required for the patterning of the posterior mesoderm in zebrafish embryos (Draper, 2003). A mutation in the FGF8 gene alone causes a relatively mild phenotype, but a severe loss of the posterior mesoderm is observed when FGF24 activity is also diminished. Similarly, a small chromosome deficiency that removes both ths and pyr produces severe embryonic patterning defects (Stathopoulos, 2004).

It is conceivable that the spreading of the mesoderm across the internal surface of the neurogenic ectoderm is a simple manifestation of cell-cell contact. FGF signaling might cause each mesoderm cell to make maximal contact with the neurogenic ectoderm. According to this view, the Ths and Pyr ligands are permissive, and simply promote cell adhesion. An alternative view is that Ths and Pyr are spatially activated in a manner that promotes a temporal gradient of information that guides the movement of the mesoderm toward the dorsal ectoderm. The expression of dpERK is consistent with an early requirement of FGF signaling acting permissively to activate Htl and allow the mesoderm to start spreading. Staining is first seen throughout the mesoderm that is in contact with the ectoderm during early phases of gastrulation when these individual mesoderm cells come into contact with the neurogenic ectoderm. Later, dpERK staining is restricted to the leading edge of the mesoderm as it spreads into the dorsal ectoderm. These data support a model in which the FGF ligands, Ths and/or Pyr, activate Htl in an instructive manner that guides the mesoderm during the later stage of spreading. The expression of ths and pyr is consistent with this model of early permissive and late instructive roles of the ligands in Htl activation. Early, ths is expressed in a broad staining pattern, which might reflect a role in the promotion of initial contact between the mesoderm and neurogenic ectoderm. The restricted staining of both ths and pyr seen later may reflect a Ths/Pyr activity gradient emanating from increasingly more dorsal regions of the neurogenic ectoderm (Stathopoulos, 2004).

The combined ths and pyr expression profiles might produce a dynamic FGF activity gradient within the neurogenic ectoderm that guides the spreading of the mesoderm into the dorsal ectoderm. pyr expression is particularly dynamic, and rapidly lost in the neurogenic ectoderm, whereas ths expression is progressively lost first in ventral regions and then in more dorsal regions of the neurogenic ectoderm. In principle, this putative FGF gradient could provide a precise guidance cue for the coordinated spreading of the mesoderm into the dorsal ectoderm. However, it is also conceivable that the production of an FGF signaling gradient depends on post-transcriptional regulation, such as the translational regulation of mRNA expression or differential processing of FGF precursor proteins. For instance, the early embryonic enhancer isolated for ths does not support expression during germ-band elongation even though ths mRNA can be detected by in situ hybridization at this same stage. One interpretation of these results is that the ths mRNA is not synthesized during mesoderm migration. Differential degradation might help shape an FGF ligand activity gradient, as observed for FGF8. In addition, negative regulators of signaling downstream of the Htl receptor could contribute to the production and sharpening of an FGF signaling activity gradient (Stathopoulos, 2004).

Ths and Pyr are related to FGF signaling molecules that control both cell movement and differentiation. For example, EGL-17 directs the movement of the sex myo-blasts in the gonad and FGF8 is required for the migration of the mesoderm into the primitive streak of vertebrate embryos and later for heart development (Burdine, 1998; Sun, 1999; Reifers, 2000). Ths and Pyr are required both for the spreading of the mesoderm along the internal surface of the neurogenic ectoderm during gastrulation, as well as the subsequent induction of the dorsal mesoderm to form pericardial tissues (Stathopoulos, 2004).

Evidence that neurogenic expression of Ths and Pyr is important for the orderly spreading of the mesoderm was obtained by misexpressing Ths. Embryos misrepresenting Ths in the mesoderm exhibit a variety of defects including mild twisting of the germ band, abnormal patterning of the body wall muscles, and an expansion of cardiac tissues. The latter phenotype can be explained on the basis of expanded induction of dorsal mesoderm (there is at least a threefold increase in the number of Eve-expressing cells). A more uniform rescue phenotype was obtained when exogenous Ths products were expressed in the ectoderm using the 69B-gal4 transgene (Stathopoulos, 2004).

Once mesoderm spreading is complete, the leading edge of the mesoderm comes into contact with Dpp-expressing cells in the dorsal ectoderm. Dpp signaling might be sufficient for the activation of some of the target genes required for the patterning of the visceral mesoderm, such as tin and bap during stage 10. However, Dpp is insufficient for other inductive events such as the activation of tin and eve in different heart precursors. The loss of eve expression in ths;pyr and htl mutants does not appear to be due to a breakdown in mesoderm spreading. Although this spreading is delayed in the mutants, it does ultimately occur. The late activation of the Htl receptor may be essential for the induction of eve expression and the specification of pericardial tissues. Previous studies suggest that Dpp works together with another signal that may be localized in the dorsal ectoderm. This second signal appears to trigger Ras signaling because the expression of a constitutively activated form of Ras causes expanded expression of eve. Evidence that the second signal might be FGF stems from the analysis of a dominant-negative Htl receptor, which blocks the full expression of cardiac and pericardial gene markers after the mesoderm has spread. The present study considerably strengthens the case that FGF is the second signal that patterns the dorsal mesoderm. The misexpression of Ths in the mesoderm causes a substantial expansion in the dorsal mesoderm and the number of Eve-expressing cells. Moreover, ths and pyr are expressed in specific 'spots' within the dorsal ectoderm that are adjacent to the internal mesoderm where eve is activated. Thus, the simplest interpretation of the results is that FGF signaling controls both the spreading and patterning of the dorsal mesoderm (Stathopoulos, 2004).

The spreading and subsequent subdivision of the mesoderm into distinct dorsal and ventral lineages can be viewed as direct readouts of the Dorsal gradient. The identification of mesoderm enhancers for htl and dof/hbr/smsf based on clustering of Dorsal-binding sites (and associated sequence motifs) suggests that these genes are directly activated by high levels of the Dorsal gradient. Htl-dependent signaling is triggered by Ths and Pyr, which are selectively expressed in the neurogenic ectoderm in response to low levels of the Dorsal gradient. After spreading, dorsal mesoderm cells comes into contact with Dpp-expressing cells in the dorsal ectoderm, and are thereby induced to form dorsal lineages such as cardiac tissues. The same low levels of the Dorsal gradient that activate ths and pyr also activate sog expression and repress dpp. The Sog inhibitor ensures that Dpp signaling is restricted to the dorsal ectoderm. Thus, the differential regulation of Htl and its ligands determines the precise limits of mesoderm-ectoderm germ-layer interactions during gastrulation (Stathopoulos, 2004).

GENE STRUCTURE

A recent microarray screen identified a gene called Neu4 (CG12443) that is directly activated by low levels of the maternal Dorsal gradient in broad lateral stripes that encompass the entire presumptive neurogenic ectoderm (Stathopoulos, 2002). Computational methods were used to identify a tissue-specific enhancer that directs this Neu4 expression pattern. The enhancer contains three tightly linked, high-affinity Dorsal-binding sites. It was found to map quite far from the predicted Neu4 promoter, ~16 kb 5' of the coding region. To confirm this remote linkage of the Neu4 enhancer, rapid amplification of cDNA ends (RACE) was done using embryonic RNA to identify 5' Neu4 coding sequence. These assays identified three short exons that map far 5' of the previously identified coding sequences. The newly identified exons place the Neu4 enhancer within intron 2 of the transcription unit and not in the 5' regulatory region. The new exons were found to share homology with another predicted open reading frame (ORF), CG13194, that maps ~80 kb 5' of the Neu4 gene. RACE assays were done to identify the 5' CG13194 coding sequences. Neu4/CG12443 is referred to as thisbe (ths) and the second and related gene, CG13194, as pyramus (pyr (Stathopoulos, 2004).

PROTEIN STRUCTURE

Amino Acids - 699 for pyramus and 702 for thisbe

Structural Domains

CG12443 and CG13194 were predicted to consist of two exons each, and a BLASTp analysis exhibited a 36% amino acid identity for the primary sequences of the two genes within a stretch of 47 amino acids in the amino terminus. Furthermore, this amino-terminal sequence of CG13194 exhibited 36% amino acid identity with FGF8 from the African clawed frog Xenopus laevis and FGF8 from the axolotl (Ambystoma mexicanum). The sequence homology of CG13194 is within the conserved FGF core domain of FGF8 and includes highly conserved amino acids such as Cys-101, Phe-103 and Glu-105, which are conserved in all FGFs that can be found in the databases. Because of the similarity of the predicted gene products to FGFs and because no other gene within this interval exhibited an early zygotic expression profile, focus of further analysis was placed on these two candidates (Gryzik, 2004).

The alignment of the open reading frame of CG13194 with the core domain of mammalian FGFs, as well as the lack of a signal peptide sequence, suggested that the predicted exons in the genome annotation represented only part of the gene. A Northern blot analysis demonstrated that the transcripts were indeed larger than the suggested annotation of the genes. The CG13194 probe detected a single band of 3.4 kb, and the CG12443 probe detected a prominent 4.9 kb band. The cDNA of CG12443 was cloned by RT PCR and a 2.5 kb product was obtained and sequenced. Comparison of the cDNA sequence to the genomic sequence revealed that the gene, which was annotated as CG12443, contains four exons. The CG12443 cDNA encodes a novel FGF with an N-terminal signal peptide of 21 amino acids, a FGF core domain of 67 amino acids, and a long C-terminal region of 611 amino acids with no significant homologies to other proteins. The FGF core domains of CG12443 and CG13194 exhibit 39% identical amino acid residues. Interestingly, the homology with vertebrate FGF8, FGF17, and FGF18 is similarly high; it ranges from 32% to 35% amino acid identity. Because of the high degree of homology of CG13194 and CG12443 to vertebrate FGF8 and the fact that FGF8 plays an essential role in vertebrate gastrulation, it is proposed that these genes be named FGF8-like1 (CG12443) and FGF8-like2 (CG13194) (Gryzik, 2004).

The newly identified ths and pyr 5'-exons encode peptides related to FGF ligands. FGF molecules are highly divergent and exhibit limited amino acid sequence identity (Szebenyi, 1999; Ornitz 2000). The core FGF domain is composed of 12 ß strands separated by coiled-coil regions that form a trefoil structure. The most highly conserved amino acid residues between the Ths and Pyr FGF domains, and other FGFs including FGF8 and Branchless, tend to map in regions that are essential for the integrity of the structure, particularly the ß strands. Another notable feature of the Ths and Pyr proteins is that they are both predicted to contain N-terminal signal sequences, suggesting that the proteins are present extracellularly. They also lack several of the signature amino acid residues that have been implicated in the interaction of certain FGFs with heparin sulfate proteoglycans (HSPGs; Ornitz 2000). Because FGF-HSPG interactions are thought to mediate sustained activation of FGF receptors, it is conceivable that Ths and Pyr function as transient signals, which is consistent with their dynamic patterns of expression during development (Stathopoulos, 2004).

Phylogenetic analyses were performed using the neighbor-joining method. The principal finding is that among all known FGFs, Ths and Pyr are most closely related to EGL-17 in C. elegans and the FGF8/17/18 subfamily in vertebrates (Burdine, 1997; Maruoka, 1998). In contrast, Branchless (Bnl) and Let-756, the only other nonchordate FGFs, are most closely related to viral FGF-like molecules. Both ths and pyr are conserved in the Drosophila pseudoobscura (D.pse) genome, but there is only a single gene in the mosquito, Anopheles gambiae (A.gam). This single FGF gene is more closely related to ths than pyr. It would appear that ancestral dipterans contained a single copy of an FGF gene that underwent duplication after the divergence of mosquito but prior to the divergence of D. melanogaster and D. pseudoobscura. Two Drosophila genes, CG13195 and CG12444, that are closely linked to pyr and ths are also related to one another, suggesting that an ~50-kb interval containing the ancestral FGF gene, along with its neighbor, underwent a tandem duplication event. Greater sequence conservation exists between thisbe orthologs (D.mel Thisbe and D.pse Thisbe) than pyramus orthologs (D.mel Pyramus and D.pse Pyramus). This result suggests that selection may be acting on ths to maintain some ancestral function, whereas pyramus has been released from constraint and is rapidly evolving in the Drosophilids (Stathopoulos, 2004).

An alignment was generated of the putative protein sequences of D. melanogaster Thisbe and Pyramus with D. melanogaster Branchless, Human FGF8, and with the PFAM Hidden Markov Model (HMM) consensus for the FGF protein family. The HMM consensus represents the residues having the highest probability of occurrence for each position in an alignment of all known FGF proteins. Conserved positions exhibited 80% identity; in other positions 80% of the residues are similar using a PAM 250 substitution matrix. Conserved were 12 ß strands in the FGF trefoil structure (Stathopoulos, 2004).

date revised: 30 May 2004

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.