Co-option of a coordinate system defined by the EGFr and Dpp pathways in the evolution of a morphological novelty

Background Morphological innovation is an elusive and fascinating concept in evolutionary biology. A novel structure may open up an array of possibilities for adaptation, and thus is fundamental to the evolution of complex multicellular life. We use the respiratory appendages on the dorsal-anterior side of the Drosophila eggshell as a model system for morphological novelty. To study the co-option of genetic pathways in the evolution of this novelty we have compared oogenesis and eggshell patterning in Drosophila melanogaster with Ceratitis capitata, a dipteran whose eggs do not bear dorsal appendages. Results During the final stages of oogenesis, the appendages are formed by specific groups of cells in the follicular epithelium of the egg chamber. These cells are defined via signaling activity of the Dpp and EGFr pathways, and we find that both pathways are active in C. capitata oogenesis. The transcription factor gene mirror is expressed downstream of EGFr activation in a dorsolateral domain in the D. melanogaster egg chamber, but could not be detected during C. capitata oogenesis. In D. melanogaster, mirror regulates the expression of two important genes: broad, which defines the appendage primordia, and pipe, involved in embryonic dorsoventral polarity. In C. capitata, broad remains expressed ubiquitously throughout the follicular epithelium, and is not restricted to the appendage primordia. Interestingly pipe expression did not differ between the two species. Conclusions Our analysis identifies both broad and mirror as important nodes that have been redeployed in the Drosophila egg chamber patterning network in the evolution of a morphologically novel feature. Further, our results show how pre-existing signals can provide an epithelium with a spatial coordinate system, which can be co-opted for novel patterns.


Background
Classically, the concept of evolutionary novelty is that of a new trait, usually an anatomical or morphological one, that opens up the possibility of a wide adaptive radiation into new niches [1]. This definition places an emphasis on adaptation and is thus illustrative of the central role novel traits may have on shaping life on earth. Yet, it is a restrictive definition in that it implies *Correspondence: b.vreede@gmail.com; esucena@igc.gulbenkian.pt 1 Instituto Gulbenkian de Ciência, Rua da Quinta Grande 6, Oeiras, Portugal 2 Universidade de Lisboa, Faculdade de Ciências, Departamento da Biologia Animal, Lisbon, Portugal Full list of author information is available at the end of the article knowledge of the adaptive value of the trait, eliminating traits that have been phylogenetically validated as novelties but lack ecological context. Moreover, this definition disregards the ontogenic aspects of the new trait, particularly of novel morphologies, the most prevalent type of novelty reported. An alternative definition has been proposed by Müller and Wagner [2] to a great extent circumventing the limitations described above. They define a morphological novelty as 'a structure that is neither homologous to any structure in the ancestral species nor homonomous to any other structure of the same organism' [2]. Still, this definition is not without problems, as it dislocates the problem of defining novelty to the definition http://www.evodevojournal.com/content/4/1/7 of homology, which is another particularly elusive concept in biology [3][4][5]. Or, as phrased by Moczek [6]: 'our definition of novelty now only becomes as strong as our definition of homology' . Nonetheless, and as new perspectives and conceptual contributions to this debate arise [7], in the confined context of this paper we will adopt this latter, more operational definition of a morphological novelty.
At the mechanistic level, one of the most important contributions of evo-devo to our understanding of the evolutionary process has been the refinement and experimental validation of the gene recruitment concept (cooption). These are key innovations at the genetic level that may underlie differences in cellular growth and morphogenetic processes between related organisms, which have diverged morphologically [8]. In recent years many examples have demonstrated that evolution largely relies on recycling old genes and pathways to generate novel patterns and morphologies [9,10].
The model Drosophila melanogaster has often been criticized for being extremely derived, and therefore a poor reference in understanding the prototypical insect. Here, we turn this argument around and use D. melanogaster as a source of novelty by identifying a novel morphological feature acquired in the evolution of the Drosophilidae family: the egg dorsal appendages ( Figure 1B). The formation of these dorsal-anterior chorionic filaments during Drosophila oogenesis has already been used as a model system for the study of many developmental mechanisms, such as epithelial patterning [11,12], and tube formation [13][14][15].
Most (though not all) eggs of Drosophilidae bear dorsal appendages, which are thought to have a single origin in their last common ancestor [16]. The appendages are hollow tubes protruding from the dorsal-anterior end of the chorion, and provide an oxygen supply to the immersed egg [16,17]. They portray a striking diversity within the Drosophilidae family [18][19][20], which makes were identifiable in C. capitata using criteria described in D. melanogaster. At stage 8 of oogenesis, the oocyte nucleus (on) is localized asymmetrically in the oocyte (oo), which at this stage is of roughly equal size to the nurse cells (nc). At stage 9 the follicle cells (fc) start their migration to posterior: anterior follicle cells stretching over the nurse cells, and posterior follicle cells forming a layer of columnar cells over the oocyte. At the same time, a cluster of border cells (bc) migrates between the nurse cells to the anterior end of the oocyte. Late stage 10 sees the columnar follicle cells migrating centripetally (cmfc), in between nurse cells and oocyte. Stage 11 shows a difference between D. melanogaster and C. capitata egg chambers in the relative thickness of dorsal and ventral follicle cell layers. (B) Eggs of D. melanogaster and C. capitata, the former bearing obvious structures: dorsal appendages (da), operculum (op), and an outward micropyle (mp). (C) Fas-II staining of stage 8 C. capitata egg chamber, identifying the polar cells, part of the border cell cluster. (D) In situ hybridization with a probe against slbo confirms the identity of the border cell cluster in the C. capitata egg chamber. (E) A small pore is visible in the newly formed eggshell of C. capitata, likely a structure homologous to the micropyle (mp). http://www.evodevojournal.com/content/4/1/7 the appendages an interesting subject from an evolutionary perspective. The adaptive advantage of respiratory appendages is emphasized by Hinton [16]: they allow the egg to increase its oxygen-absorbing surface without risking desiccation. Indeed, similar eggshell structures have evolved independently at least 11 more times within Diptera, and at seven more instances in other insects [16,17]. Nonetheless, and despite their assumed evolutionary advantage, they are not so prevalent that a single origin of these structures in all Diptera seems likely.
In addition to the dorsal appendages, the Drosophila egg carries an operculum and a micropyle: structures relevant for hatching and fertilization, respectively ( Figure 1B). These structures are formed during the last stage of oogenesis by designated cells in the follicular epithelium that change shape prior to the deposition of chorionic proteins [13,21]. Specification of the appendage primordia occurs chiefly through the activity of two main signaling pathways: EGFr and Dpp [13,22].

EGFr and Dpp signaling define appendage primordia
Pattern formation on the follicular epithelium occurs through the activation of a genetic network by two main input pathways: EGFr and Dpp signaling ( Figure 2). Around stage 8 of Drosophila oogenesis, dorsal patterning is initiated when the TGF-α-like ligand Gurken (Grk) localizes to the dorsal-anterior corner of the oocyte ( Figure 3A). Grk associates with the oocyte nucleus, which is pushed by microtubules to a dorsal-anterior position Figure 2 A simplified representation of the genetic network underlying dorsoventral polarity (pip) and DA-formation (br) during D. melanogaster oogenesis. Input comes from two main signaling pathways, EGFr and Dpp, providing dorsoventral and anteroposterior information, respectively, and results in the specification of domains on the epithelium expressing pip and br. [23], breaking dorsoventral symmetry in the egg chamber [24]. The Grk signal then activates EGFr in the adjacent follicle cells, leading (directly and indirectly) to the expression of several transcriptional targets, among which are mirror (mirr) [25,26], rhomboid (rho) [27], and pointed (pnt) [28] (Figure 2).
Meanwhile, Dpp signaling starts at stage 8 with the expression of dpp in a subset of anterior follicle cells [29] ( Figure 3C). Dpp protein diffuses to more posterior follicle cells, forming a morphogen gradient. It acts via the receptor Thickveins (Tkv) in the follicular epithelium to phosphorylate Mothers Against Dpp (Mad), activating the pathway in a graded manner [30]. Dpp has also been suggested to be required for the expression of mirr [31], which starts at stage 10A in a wide dorsoanterior domain ( Figure 4A). Recent work by Fuchs et al. [32] shows how the transcription factor Mirr, regulated by both Dpp and EGFr activity, and the ETS-domain transcription factor Pnt, expressed in a more narrow stripe along the midline, subsequently establish two groups of cells expressing broad (br) through two rounds of signaling. First, Mirr represses br, which has been expressed in all follicle cells up to this point, in a wide dorsoanterior region through the brE enhancer. Then, br expression is upregulated again by Mirr, but repressed by Pnt, through the brL enhancer ( Figure 2). The two resulting patches of Br-positive cells on either side of the midline are identified as 'roof cells': they will later constrict apically and shape the roof of the appendage tube [33]. Adjacent to the Br-positive patches is a single L-shaped row of cells, bordering the anterior and the central edge of the roof domain. These cells express high levels of rho, and elongate directionally to form the floor of the tube [33]. rho expression is regulated mainly by activation of the EGFr pathway, which is highly dynamic throughout oogenesis, and shows the same L-shaped pattern at the definition of the floor cells [34]. Rho itself is involved in the dynamic EGFr activation as it cleaves the EGFr ligand Spitz (Spi) into its active form, thereby providing a positive feedback loop for EGFr signaling [27,35,36] (Figure 2). Importantly, EGFr signaling also determines the dorsoventral axis of the future embryo [37]. Via Mirr, pipe (pip) expression is restricted to the ventral follicle cells (Figure 2), leaving an asymmetric distribution of Pip protein at the end of oogenesis [32,[38][39][40]. Pip is upstream of a proteolytic cascade in the embryo, leading to the well-known gradient of nuclear Dorsal that regulates the germ layers of the early embryo [41].
Dpp, too, is required for processes other than the specification of the appendage primordia. As the inward movement of the centripetally migrating follicle cells starts ( Figure 1A), dpp is expressed in the leading edge of these cells, and disrupted Dpp signaling has been associated with defects in this migration [29]. Dpp is http://www.evodevojournal.com/content/4/1/7 required furthermore for the formation of the operculum [29,42].
In summary, EGFr and Dpp activity specify dorsoventral and anteroposterior polarity in the epithelium, respectively, and their signaling information is integrated by Br and Rho, which together specify the appendage primordia. In addition, both signaling pathways are crucial for proper egg formation and further embryonic development, linking the formation of secondary (novel) structures to essential (thus presumably ancestral) developmental events.

Ceratitis capitata
Considering the relatively novel acquisition of eggshell appendages in the family Drosophilidae, it is interesting to examine the underlying patterning network in the context of a fly species that does not possess these specialized structures. Tephritidae are estimated to be separated by about 65 million years of evolution from Drosophilidae [43]. For our comparison we chose a Tephritid fly that has been established as a laboratory organism: the Mediterranean fruit fly Ceratitis capitata. C. capitata is an agricultural pest, which has motivated widespread international research, including a genome project and the development of genetic tools [44][45][46].
In this study, we have examined both EGFr and Dpp signaling as well as their downstream targets in C. capitata oogenesis, in order to understand the genetic network patterning the follicular epithelium prior to the evolution of dorsal appendages. Determining which genes behave differently in the formation of appendage-bearing (D. melanogaster) and appendage-less (C. capitata) eggshells http://www.evodevojournal.com/content/4/1/7 can help us understand the co-option of genes and the genetic network in the evolution of this novel feature. Our analysis points to a key role for the transcription factor Mirr, both in its regulation and in its transcriptional targets. Furthermore, the activity of both the EGFr and the Dpp pathway in C. capitata oogenesis leads us to hypothesize that these pathways provided positional information to the ancestral follicular epithelium, which could have facilitated further downstream patterning required for the development of the dorsal appendages.

Fly maintenance
Our initial Ceratitis capitata culture was kindly (and repeatedly) provided by Andrew Jessup (IAEA Seibersdorf, Austria), originating from flies captured in Argentina. Adult flies were maintained on a diet of sugar and hydrolyzed yeast protein, and larvae were reared on a mixture of bran, sugar and yeast. All stages were maintained at room temperature. Drosophila melanogaster Oregon R. was maintained on regular fly food at room temperature.

Cloning
Gene-specific sequences were isolated from C. capitata cDNA by PCR using degenerate primers (for dpp, mirr, rho and tkv), as well as C. capitata specific primers (for Cc-br, Cc-cic (capicua), Cc-grk, Cc-pnt, Cc-pip and Ccslbo). Specific primers were designed using contigs from For Cc-pip two primer combinations were used, generating two separate probes for in situ hybridization. These probes were (1) against the common part of all pip isoforms, and (2) against Cc-pip-ST2, the homologue of Dmpip-ST2 (isoform A). Corresponding probes were made for the positive controls in D. melanogaster.

Immunohistochemistry
Ovaries were dissected in cold PBS and fixed for 20 minutes at room temperature in 4% formaldehyde in PBTx (0.1% Triton-x100 in PBS). After fixation they were washed several times in PBTx-B (1% BSA in PBTx) at http://www.evodevojournal.com/content/4/1/7 room temperature for 1 hour. Antibody incubation was done overnight at 4°C. The rabbit anti-pMad antibody was kindly provided by the laboratory of Ginés Morata (Centro de Biología Molecular Severo Ochoa, Autonomous University of Madrid, Spain), and was used at a concentration of 1 : 100 in PBTx-B. Anti-Fasciclin II (1D4) was obtained from the Developmental Studies Hybridoma Bank (maintained at the University of Iowa, Department of Biology, Iowa), and was used at a concentration of 1 : 50. Secondary antibodies (Alexa fluor 488/546 goatanti-rabbit or anti-mouse IgG (H+L), Molecular Probes) were used at a concentration of 1 : 2000, overnight at 4°C. Nuclear staining was done with Dapi.

In situ hybridization (ISH)
Ovaries were dissected in cold PBT (0.1% Tween-20 in PBS) and transferred to 4% paraformaldehyde in PBS, where they were fixed overnight. They were subsequently washed in PBS, dehydrated and stored in 100% MeOH at −20 • C. The protocol for ISH was taken from Tautz and Pfeifle [47] and modified for oogenesis. The main change concerned the adjustment of the proteinase K digestion to 10 minutes 50 μg/mL at room temperature.
To ensure identical conditions during the experiment, the positive controls with embryos were done in the same well as the ovaries, starting at the pre-hybridization incubation in hybridization buffer at hybridization temperature. This was done because the proteinase K treatment for ovaries is much harsher than the one we used for embryos (10 minutes 50 μg/mL vs. no proteinase K at all).

C. capitata oogenesis is a suitable system for comparison
The C. capitata eggshell carries no structures that can be identified as homologues of the operculum, outward micropyle and dorsal appendages ( Figure 1B). Still, it is not entirely symmetrical, both over the anteroposterior axis and the dorsoventral axis. The anterior end of the chorion shows markedly stronger imprints of (previously present) follicle cells when compared to the posterior end. While we cannot say with certainty which side is dorsal and which is ventral, it is clear that one is more convex than the other. As both late stage egg chambers ( Figure 1A,D) and early embryos (data not shown) are clearly more convex at the ventral side, it is a reasonable assumption that the convex side of the egg is ventral.
From an initial observation of C. capitata oogenesis we can conclude first and foremost that it is a suitable system for comparison with D. melanogaster. C. capitata ovaries, like those of Drosophilidae, are meroistic polytrophic ovaries. While the egg chamber of C. capitata is usually larger than the corresponding stage in D. melanogaster, there is no notable difference in the number of cells that make up the follicular epithelium. Instead, the size of C. capitata follicle cells is increased with respect to those of D. melanogaster, thus contributing to a larger egg chamber as a whole ( Figure 1A).
The structure of the egg chambers as well as the progression of stages is nearly identical to that of Drosophila, providing a good basis for comparison ( Figure 1A). Starting at mid-oogenesis, we can observe the asymmetric localization of the oocyte nucleus (stage 8), as well as follicle cell migration (stage 9), and centripetal migration (stage 10B). Also visible is the dumping of nurse-cell content into the oocyte, as evidenced by the increasing size of the oocyte relative to the nurse cells, which disappear eventually. All these are important and stagedefining steps in Drosophila oogenesis. We will therefore refer to the stages defined in D. melanogaster [48] when describing C. capitata oogenesis.
In addition to the migration of the main body follicle cells, a cluster of anterior follicle cells can be seen to migrate between the nurse cells at stage 9. Their migration ends at the posterior edge of the nurse cells, adjacent to the oocyte, where they are shortly joined by the centripetally migrating follicle cells. In D. melanogaster these cells are known as border cells, and can be identified by the expression of slbo [49], as well as with the polar-cell-specific label Fasciclin II. Both markers confirmed the identity of the border cell cluster in C. capitata ( Figure 1C,D). Interestingly, as the border cells have been associated in D. melanogaster with the formation of the micropyle, no obvious external micropyle can be seen on the C. capitata egg ( Figure 1B). However, upon closer examination of the newly formed eggshell we found a pore-like structure on the anterior side of the eggshell, likely homologous to the micropyle pore ( Figure 1E). This is consistent with the observed border cell localization in C. capitata, as these cells are known to form the pore of the micropyle, but not the outwardly visible structure [48].

Both EGFr and Dpp pathways are active in C. capitata oogenesis
In C. capitata ovaries, the initial activation of the dorsoventral patterning cascade by the ligand Gurken occurs similarly to D. melanogaster. In the early stages, the Cc-grk transcript is visible in the oocyte at the anterior cortex (data not shown), and around stage 8 the pattern becomes restricted to the putative dorsoanterior side of the oocyte ( Figure 3A,B). The transcript disappears around stage 11.
While we were unable to obtain patterns of EGFr activation because of practical difficulties, the fact that TGFα-EGFr signaling is conserved in insects as distant as Tribolium and Gryllus [50], functioning upstream of embryonic dorsoventral patterning even in drastically different systems of oogenesis, makes it unlikely that this would be any different in C. capitata. Indeed, we observed http://www.evodevojournal.com/content/4/1/7 the dorsal repression of a known target of EGFr signaling in D. melanogaster: the gene pip ( Figure 4D,D ).
In contrast with oogenesis in D. melanogaster, Cc-dpp is not expressed in the somatic follicle cells, but instead in the germ line. Expression of Cc-dpp is first visible as early as the germarium. Once the egg chamber is formed, the dpp transcript localizes to the oocyte. When the oocyte increases in size, the mRNA seems to accumulate at the putative anterior end of the oocyte, in a ring around the edge, adjacent to the follicle cells ( Figure 3D,D ,I). Interestingly, this ring is reminiscent of the D. melanogaster pattern, where dpp is expressed in the stretched follicle cells as well as a few anterior rows of columnar follicle cells, resulting in a similar ring of dpp expression around the anterior end of the oocyte ( Figure 3C ). The main difference, of course, is that the transcript is located in different cell types.
One exception to the exclusive germ line expression of Cc-dpp is the border cell cluster. This migrating group of anterior follicle cells is not known to express dpp in D. melanogaster, but is the only group of somatic cells during oogenesis to express Cc-dpp. Expression is visible around stage 8, when the cell cluster is defined ( Figure 3D, empty arrowhead), and persists through migration until the edge of the nurse cells is reached.
A possible second group of Cc-dpp expressing follicle cells was identified using fluorescent in situ hybridization (FISH). This group of cells is centrally located between the nurse cells and the oocyte in late stage 11 ( Figure 3I ). Due to the very small sample size we cannot say with certainty whether these cells are the border cells or part of the follicle cells that have centripetally migrated inwards. As the signal of Cc-dpp expression does not persist in the border cell cluster after migration is completed, the observation could either indicate a new round of Cc-dpp expression in this cluster should these cells indeed be border cells, or it could point to conservation of dpp expression in the leading edge of centripetally migrating follicle cells.
While expression of the ligand may differ somewhat between the two species, downstream signaling is remarkably similar. The expression of the homologue of the Dpp pathway type I receptor tkv is not visibly different in C. capitata from D. melanogaster: Cc-tkv is expressed in the follicular epithelium ( Figure 3E,F), and disappears around stage 11 or 12. More importantly, the activity of the pathway, shown through immunohistochemistry for the phosphorylated form of Mad (pMad), is initially not different between the two species, despite the altered localization of the dpp transcript ( Figure 3G,H).
Differences in Dpp pathway activation between C. capitata and D. melanogaster start around stage 10B, when expression of Dm-tkv becomes restricted to the Br-positive cells of the appendage primordia, naturally affecting pMad patterns [51,52]. These dynamics were not observed in C. capitata, where no Br-positive domains are formed ( Figure 4F ).

Patterning of the follicular epithelium downstream of EGFr and Dpp
The dynamics of EGFr and Dpp signaling and subsequent epithelial patterning in D. melanogaster egg chambers are key in defining the appendage primordia. Identifying the point in the genetic network where C. capitata no longer resembles D. melanogaster is therefore an important step in understanding the evolution of the dorsal appendages, as it could indicate the point where the network was co-opted.
Our first candidate for co-option was found when we saw that no expression of mirr could be detected in C. capitata egg chambers ( Figure 4B). The probe against Ccmirr did reveal clear expression in the C. capitata embryo, in a pattern familiar from expression in D. melanogaster ( Figure 4B ) [53].
Mirr regulates the transcription of br in those cells that will give rise to the dorsal appendages ( Figure 4E ). Unsurprisingly, the br-positive domains do not appear on the C. capitata stage 10B follicular epithelium ( Figure 4F ), nor during any other stage of oogenesis. Early expression of br could be seen uniformly in the follicular epithelium, as in D. melanogaster, but the late expression dynamics, both the dorsal-anterior repression and the appearance of the two domains, were not observed; instead, expression diminished around stage 11 and had disappeared entirely by stage 12.
Preliminary results indicate that two other genes relevant for D. melanogaster epithelial patterning do not play a role in the C. capitata dorsal-anterior epithelium: expression of pnt, encoding the transcription factor responsible for the midline repression of br, could not be detected in the dorsal-anterior follicular epithelium of C. capitata. A second known expression domain of pnt at the posterior pole of the egg chamber was clearly visible from an early stage (stage 8), providing a positive control for the in situ hybridization and the pnt probe (Additional file 1). Transcription of the gene rho was also not detected in either the early broad dorsoanterior domain, or in the late hinge-shaped patterns adjacent to the br expressing domains [22] (Additional file 1). However, as both early rho expression and the dorsoanterior domain of pnt can be difficult to detect in D. melanogaster egg chambers as well, we cannot be completely certain of the absence of pnt and rho transcripts in the dorsoanterior follicular epithelium of C. capitata.

Conserved expression of pip
Interestingly, especially in the light of the absence of detectable Cc-mirr expression, Cc-pip is repressed dorsally: the transcript is expressed asymmetrically, and http://www.evodevojournal.com/content/4/1/7 clearly localizes to the ventral follicular epithelium. In a similar dynamic-though not precisely identical-to D. melanogaster, Cc-pip expression starts at stage 8 in follicle cells at the posterior pole of the egg chamber ( Figure 4D). This posterior expression domain during stages 8 and 9 is well known in D. melanogaster [38,40]. During early stage 10, ventral follicle cells start expressing Cc-pip, and by late stage 10 expression in ventral and posterior follicle cells is of equal strength ( Figure 4D ). The pattern at this stage is identical to the expression pattern of Dm-pip ( Figure 4C ), including the sharp on-off boundary between ventral and dorsal cells. These results were obtained using two separate probes: one against the common part of all pip isoforms, and one specific to the homologue of isoform A (or pipe-ST2), confirming that the same isoform is used in C. capitata oogenesis as is known to function in D. melanogaster [54].

Pre-existing functional signals provide positional information
A first conclusion we can draw from the work presented is the fact that the activity during oogenesis of the two main patterning pathways, EGFr and Dpp, preceded the evolution of dorsal appendages and their underlying epithelial patterns [29]. The ancestral role of EGFr signaling lies in determining the dorsoventral axis of the future embryo [50], while Dpp is involved in various cell migrations required for the developmental progression of the egg chamber. Activity from these pathways provides the epithelium with positional information that may constitute an important facilitator for novel patterns to evolve.
A formalism for pattern formation on the Drosophila melanogaster follicular epithelium was developed in 2008 by Yakoby et al. [11]. They propose a combinatorial code of principle patterns from which all expression patterns at the dorsal-anterior follicular epithelium can be derived. As their formalism includes both the EGFr and Dpp input, as well as three additional primary building blocks specific to the dorsal-anterior epithelium, this system constitutes the next step of pattern formation in the evolution of Drosophila eggshell morphology. Interestingly, this specification of patterns from a system of higher order components is an emerging theme in regulatory evolution [55]. Our results therefore fit within the larger research theme of how pre-existing information may bias the future evolution of pattern formation and morphology.

Upstream differences in Dpp signaling between D. melanogaster and C. capitata
Despite the fact that the pattern of Dpp activity is similar between C. capitata and D. melanogaster, the differences in the underlying expression of its ligand dpp are puzzling. Not only are there differences in the expression patterns of Dm-dpp and Cc-dpp, but the transcripts are produced by an altogether different cell type. Cc-dpp is likely expressed by the nurse cells and transported to the oocyte, both of which are germ line, while D. melanogaster requires dpp expression in the somatic follicle cells.
Several functions have been described for Dpp signaling in D. melanogaster [56]. In the context of oogenesis, the need for Dpp signaling in the formation of anterior eggshell structures has been clearly established: dpp is expressed in the cells that will form the operculum, and disruptions of Dpp signaling cause misplaced and deformed appendages [29,[57][58][59]. Additionally, Dpp signaling is needed for the centripetal migration of follicle cells, to maintain structural integrity of the egg chamber, and for dumping of nurse cell content into the oocyte [29]. However, expression of the signaling molecule Dpp is only required in the somatic follicular epithelium: in D. melanogaster germ line dpp is not required during oogenesis [60]. In C. capitata, dpp is clearly expressed in the germ line, and the signal acts through receptors in the soma. While it cannot be ruled out that the Dpp activity in the follicular epithelium is a response to early dpp expression in the border cell cluster, the transcript in the nurse cells as well as the ring of dpp in the oocyte are a likely origin for Dpp signaling in the stretched and centripetally migrating follicle cells, respectively ( Figure 3D ,H,I).
Although it is intriguing to observe such apparent dramatic changes in expression patterns, conservation of phenotype in face of substantial changes in the architecture of developmental programs has been widely reported and discussed under the concept of developmental systems drift [61]. In this particular case, it is important to remember that the functional event, the actual Dpp signal, remains a cooperative act between the ligand and its receptors. Thus, the selective pressure for Dpp function will be on this signaling event, as opposed to the source of the ligand. Interestingly, a similar interaction between germ line and soma has been described regarding Dpp signaling in the honeybee Apis mellifera [62]. In this system, dpp mRNA is localized to a dorsal stripe in the oocyte, and signaling activity is observed in the overlying follicle cells. While the absence of data on dpp expression in other closely related dipteran species precludes a clear evolutionary interpretation of these patterns, it does suggest that dpp expression in the follicle cells is a recent adaptation. A possible reason could be to prevent Dpp from remaining in the perivitelline cleft at the end of oogenesis, which could interfere with future embryonic dorsoventral patterning in which Dpp plays a large role.

Regulation of mirr
One of the most interesting and salient aspects of this model system is the intimate genetic link between the novel phenotype-the dorsal appendages-and an http://www.evodevojournal.com/content/4/1/7 ancestral and vital feature of embryonic developmentdorsoventral polarity. One element of the network draws specific attention: the transcription factor Mirror (Mirr). Mirr regulates both the expression of pipe (pip), the gene encoding a sulfotransferase that is pivotal in providing dorsoventral polarity to the embryo, and broad (br), the gene that defines the dorsal appendage primordia. Our results show that pip expression is conserved, while its upstream regulator Mirr appears to be part of the novel branch of the network in D. melanogaster. This observation suggests that Mirr, rather than Br, operates as the key node of the network underlying the evolution of dorsal appendages.
Understanding the regulation of mirr in D. melanogaster then is necessary to understand how Mirr could have been co-opted to regulate br, and possibly pip, in a novel manner. The best substantiated link between EGFr activation and mirr expression is the HMG-box transcription factor Capicua (Cic). Cic is a repressor of mirr in ventral and lateral follicle cells, and is downregulated in response to EGFr signaling [63,64]. However, global de-repression of mirr through Cic loss-of-function only results in visible expression of mirr in anterior follicle cells [65]. This observation suggests the additional involvement of Dpp activity in mirr regulation [31].
However, it has recently become clear that detectable mirr may not fully represent Mirr activity throughout the follicular epithelium. Indeed, local de-repression of mirr through follicle cell clones in the posterior part of the epithelium is still sufficient to repress pip [40]. Interestingly, pip expression is unaffected when the Dpp pathway is disrupted [30]. Moreover, computational analyses have shown that the two-dimensional EGFr signaling profile is sufficient to explain the pip expression pattern, without any additional requirements for Dpp or other factors [66].
Thus, the fact that mirr expression is not seen with in situ hybridization does not preclude its activity in the follicular epithelium at a level sufficient to repress pip. In other words, we cannot conclude that mirr expression is absent in C. capitata from our data alone. However, although low (undetectable) mirr expression may be sufficient in D. melanogaster for regulation of pip, high (detectable) expression levels are necessary for activating the late enhancer of br and defining the dorsal appendage fate [31,32,65]. These high levels of mirr expression are clearly absent in C. capitata, and constitute a novel expression pattern related to the formation of a novel trait. Expression of br also depends on input from the Dpp pathway [22,30].
Based on this data and our observations in C. capitata we propose a model that separates the contribution of mirr to dorsoventral polarity from its function in epithelial patterning, using two regulatory modules in mirr to pip repression requires only low levels of mirr, which are provided through the green module, whereas the br(L) enhancer is activated only when mirr levels are sufficiently high, which is achieved through the orange module. (C) A proposal for the evolution of two regulatory modules, using two enhancers. A single enhancer (mirLo) is responsive to EGFr signaling only, and is sufficient to provide the low level of mirr expression required to repress pip as part of an ancestral signaling cassette. A second enhancer (mirHi) has evolved in Drosophila, which now drives mirr expression in response to both EGFr and Dpp signaling, in high levels that are sufficient for the activation of the br(L) enhancer. http://www.evodevojournal.com/content/4/1/7 obtain two distinct levels of expression ( Figure 5A,B). One of these, responding only to EGFr activation-presumably through Cic down-regulation-is sufficient to generate the low expression levels required to repress pip (and likely also act on the early enhancer of br). Conversely, the other module requires Dpp signaling in addition to Cic down-regulation, and is able to regulate mirr expression to the high levels observed with in situ hybridization in the dorsal-anterior follicle cells of wild-type Drosophila egg chambers.
With this model we predict that dorsal de-repression of mirr through Cic is sufficient for pip repression, and constitutes an ancestral signaling cassette linking EGFr activation to embryonic dorsoventral patterning. Due to a non-cross-reactive antibody we were unable to confirm whether the localization of Cic in the C. capitata follicular epithelium fits our model, but we do note that cic mRNA is expressed in the egg chambers (Additional file 1).
Alternatively, mirr expression could be absent in C. capitata altogether, and pip could be regulated by another transcription factor. However, given the important role of Pip in embryonic dorsoventral axis determination, and the dramatic defects that are caused with minimal variation in factors along the anteroposterior axis [67,68], we consider it likely that pip regulation happens through a conserved mechanism involving mirr. Pending the completion of the C. capitata whole genome sequencing project, it will be possible to test this hypothesis with a search for Mirrresponsive elements in a Cc-pip regulatory region.
Presenting both modules as enhancers of mirr provides us with a hypothesis regarding their evolution ( Figure 5C). The predicted 'mirLo' enhancer is expected to be ancestral, as we base its existence on the mirr and pip expression patterns in C. capitata egg chambers. MirLo would drive mirr expression in dorsal follicle cells in a level sufficiently high to repress pip, thus regulating dorsoventral polarity of the future embryo, downstream of EGFr signaling and independent of Dpp. The appearance of the second enhancer 'mirHi' would allow mirr to start responding to information from the Dpp pathway, and open up the evolutionary road to new patterns on the follicular epithelium.

Conclusions
In the evolution of dorsal appendages, several genes have been co-opted into a network that originally regulated only dorsoventral polarity, using the input from a second signaling pathway active in the tissue. The activity of these pathways, EGFr and Dpp, defined a coordinate system on the epithelium upon which novel gene expression patterns were built. We have shown here how this coordinate system predates the evolution of dorsal appendages, providing positional information to the follicular epithelium that played a crucial role in future pattern formation.
The main regulators in this novel genetic network are transcription factors Mirr, Pnt and Br. The latter integrates the information from upstream Mirr and Pnt to specify the appendage primordia, and drives morphogenesis of the appendage [32,57]. Interestingly, while all three transcription factors have co-opted novel expression patterns and interactions to provide the main regulatory information for the epithelial positional cues to be translated into a novel morphology, both br, pnt and very likely mirr were already expressed in the ancestral non-appendage-forming epithelium. In this case it is notable that evolution may have taught 'old genes new tricks' [69] within the same broad spatial and developmental context. A novel morphological feature has been achieved by modifying the levels of existing expression patterns allocated to ancient developmental roles (br and likely mirr) with new enhancers responding to new information, or through the addition of an expression domain at another end of the same epithelium (pnt).
Most likely, the transcription factors Br, Mirr and Pnt have not been the only ones to evolve new roles. Future research could go into the EGFr feedback loop, looking at rho and aos, as well as other EGFr targets. However, while the EGFr feedback loop was long thought to be the main patterning component of the eggshell [35], it has meanwhile been shown that eggshell patterning functions normally without several elements of this feedback system [70,71]. We can therefore conclude that changes in the regulation of mirr, pnt and br played important roles in the evolution of this novel morphology, and the detailed dissection of its molecular genetic basis constitutes the most important and immediate research agenda in the comprehension of the evolution of this morphological novelty.
Finally, we would like to stress that, as a derived organism, the model system D. melanogaster provides us with an excellent handle to tackle the question of evolutionary novelties, and for the dissection of the molecular mechanisms that underlie the process of co-option. We propose that many other traits that define Drosophila as derivedbe they morphological or behavioural in nature-can be amenable to study using a comparative approach with other emergent dipteran model systems.

Additional file
Additional file 1: Preliminary results: expression patterns of Cc-rho, Cc-pnt and Cc-cic. Rhomboid: stage 9 and stage 10 ovaries do not show expression of the gene rhomboid in Ceratitis capitata. Positive control for the protocol and the probe is shown in a stage 9 to 10 embryo (ventral view). Pointed: Stage 9 clearly shows the familiar posterior expression of pointed, which persists in stage 10 to 11 ovaries of Ceratitis capitata. No expression http://www.evodevojournal.com/content/4/1/7 at the dorsal-anterior side of egg chambers could be detected. Fibers surrounding the egg chamber sometimes obscure the results, but changing the focal plane can confirm the identity of a signal as either coming from aspecific staining in fibers or staining of the cells in the follicular epithelium. Capicua: Strong expression of capicua is seen in the nurse cells, but weaker expression can also be seen in the follicular epithelium (stage 10 egg chamber, below).