Introduction

Arthropods are the most speciose animal group on Earth and they have successfully adapted to virtually every ecological habitat. One of the cornerstones of this evolutionary success is the diversity of their appendages, the morphology of which has been adapted to various functions and lifestyles. For instance, sensory antennae are present in many species and are used for odor and taste perception. Other appendage types include for example tooth-like mouthparts like the mandible or locomotory legs used for walking, jumping, and swimming.

Because the appendages are so important for the adaptive evolution and diversification of the arthropods, there is quite some interest in the genetic mechanisms that control the development of the appendage types. In the fly Drosophila melanogaster (Meigen 1830), a number of genes required for walking leg development have been identified. Initially, the cells that will form the legs are specified by the mechanisms that also define the dorsal–ventral polarity of the embryo. The expressions of decapentaplegic (dpp) on the dorsal side and Epidermal growth factor receptor (Egfr) on the ventral side restrict the leg primordium to the lateral side (Goto and Hayashi 1997). The wingless (wg) gene is then required to activate the formation of the proximal–distal axis of the legs, which involves genes like Distal-less (Dll) and dachshund (dac) (Kubota et al. 2003; Estella and Mann 2008; Estella et al. 2008; McKay et al. 2009). The proximal portion of the legs does not require wg or the expression of Dll and dac, but is patterned by the concerted action of the TALE homebox genes extradenticle (exd) and homothorax (hth) (Rieckhof et al. 1997; González-Crespo et al. 1998; Abu-Shaar and Mann 1998; Jaw et al. 2000). After this initial subdivision of the leg into discrete proximal–distal portions, further subdivision is achieved by distal signaling through the Egfr pathway (Campbell 2002; Galindo et al. 2002), and the final number of leg segments is set down by segmentally repeated sources of Notch signaling (de Celis et al. 1998; Rauskolb and Irvine 1999; Bishop et al. 1999; Rauskolb 2001).

These leg patterning mechanisms are generally similar in the other appendage types, but modifications have been described that are responsible for appendage type-specific morphology. For example, in the antenna, the regulation of exd and hth is different from that in the legs (Casares and Mann 1998). The two genes are expressed thoughout the developing antenna and this leads to the activation of various antenna-specific genes (Dong et al. 2000, 2002). Another example is the fly proboscis, a highly modified appendage type comprising all mouthparts fused into one feeding structure. Genetic experiments have shown that in the proboscis, the two Hox genes proboscipedia and Sex-combs-reduced change the regulation of most appendage development genes to a state different from their regulation in the legs (Abzhanov et al. 2001). All available data suggest that a common developmental gene network underlies the development of all appendage types in D. melanogaster, but the exact interplay and regulation of all components differ between the appendage types.

A set of appendage patterning genes has also been studied in a number of other arthropods from all four major arthropods groups, i.e., chelicerates, crustaceans, myriapods, and insects (e.g., Abzhanov and Kaufman 2000; Prpic and Tautz 2003). In terms of species numbers, the insects have been studied most intensively (reviewed in Angelini and Kaufman 2005), but the analysis of gene expression was in most species restricted to the expression of Dll for which a highly cross-reactive antibody was available (e.g., Scholtz et al. 1998; Popadic et al. 1998). In addition, most studies involved higher insects from the Pterygota (winged insects), and only few data are available for more basally branching hexapods from the apterygote orders (e.g., Palopoli and Patel 1998; Mittmann and Scholtz 2001).

Primitively wingless groups (formerly placed in a group called “Apterygota”) are a paraphyletic assemblage at the base of the phylogenetic tree of the hexapods. They comprise the orders Collembola (springtails), Protura (coneheads), Diplura (two-pronged bristletails), Archaeognatha (jumping bristletails), and Zygentoma (silverfish and allies). Traditionally, they belong to the “small insect orders” and have attracted much less attention than their winged relatives. We have therefore studied appendage patterning in two species of apterygote hexapods, namely the white springtail Folsomia candida (Willem 1902) and the firebrat Thermobia domestica (Packard 1873).

T. domestica belongs to the Zygentoma. Members of the Zygentoma usually have a flattened body shape (Fig. 1a), and many species have colored or iridescent scales on the upper side of their body. Most species live in moist habitats under stones or leaf litter. Some species are synanthropic and are often found in houses and cellars (e.g., the firebrat and the common silverfish). The head of the Zygentoma is a typical insect head with long, multisegmented antennae and chewing mandibles, maxillae, and a labium. In most species, the maxillae and the labium bear long palps that serve a sensory function in addition to the antennae. The thorax bears three pairs of legs used for fast running on the ground. The abdomen comprises ten segments, plus a long spine at the end. This spine, the filamentum terminale, is often regarded as a modified 11th abdominal segment. In addition to the filamentum terminale, there is a pair of very long cerci growing from the reduced tenth abdominal segment. Apart from these long abdominal outgrowths, there also are shorter styli on at least two abdominal segments. These styli are usually associated with the so-called coxal sacs, small round organs of unclear function (these coxal sacs are lacking, however, in the family Lepismatidae that also includes T. domestica).

Fig. 1
figure 1

Overview of zygentoman and collembolan morphology. a Generalized adult bodyplan of the Zygentoma, lateral view. Note that the mouthparts are external, and the maxilla and labium bear long, segmented palps. After Handschin (1929). b Generalized adult bodyplan of the Collembola, lateral view. After Hopkin (1997). Note that the mouthparts are internalized. The palps of maxilla and labium are short or fully reduced in the adult and not visible from the outside. c Head of an embryo of T. domestica, explanatory drawing in ventral aspect and only one half is shown. d Head of an embryo of F. candida, explanatory drawing in ventral aspect and only one half is shown. Note that the labrum is not a separate structure in this species and forms a unit with the anterior head tissue, thus forming the so-called clypeolabrum. Abbreviations: a1–a10 abdominal segments, an antenna, ce cercus, clr clypeolabrum, fc furca, ft filamentum terminale, ga galea, gl glossa, hl head lobe, la lacinia, lbp labial palp, lr labrum, md mandible, mxp maxillary palp, pg paraglossa, re retinaculum, sl superlingua, st stylus, t1–t3 thoracic segments, vt ventral tubus

Full size image

F. candida belongs to the Collembola. Members of the Collembola usually are small insects that live in the soil or in moss cushions. The defining character of the Collembola is the furca (the “spring”) (Fig. 1b). The furca is reduced in a few species, but in the majority of all springtails the furca is a large and bifurcated appendage on the fourth abdominal segment. It is normally folded on the ventral side of the abdomen and is secured with a small outgrowth on the third abdominal segment, the so-called retinaculum. For jumping, the furca is rapidly released from the retinaculum and folded out, thus catapulting the animal up and forward. On the first abdominal segment, there is a large ventral outgrowth, the ventral tubulus, which is required for osmoregulation (reviewed in Hopkin 1997). The head capsule of the collembolan head not only covers the dorsal side of the head, but is drawn out ventrally and thus completely covers the mouthparts, a condition termed entognathy (in contrast to ectognathy, the condition with free mouthparts which is present in most insects). In the adult, the mouthparts are therefore not visible from the outside, but in the embryo the individual gnathal appendages can be seen (Fig. 1d) (see also Uemiya and Ando 1987), compared to those of insects with free (=ectognathous) mouthparts, like, e.g., T. domestica (Fig. 1c). The mandible is an unbranched appendage in F. candida, but there is a smaller outgrowth next to the base of each mandible. These small buds are the primordia of the superlinguae, lip-like structures that are lacking in most ectognathous insects (including T. domestica). Like in most insects, the maxillae and labial appendages of F. candida and T. domestica have two endites each, termed galea and lacinia in the maxilla, and glossa and paraglossa in the labium. However, the palps of these appendages are strongly reduced in F. candida; in the labium, the palp is just as long as the neighboring endites (Fig. 1d). The labrum is an anterior outgrowth that serves as an upper lip during feeding. While the labrum is a free outgrowth in T. domestica (and many other insect species) (Fig. 1c), in F. candida, the labrum is entirely fused to the clypeus (the “forehead” of the insect head) and is therefore translocated to the anterior end of the head, forming the clypeolabrum (Fig. 1d).

Since the appendages of T. domestica and F. candida show conserved as well as unique features, we were interested whether the expression of some of the appendage patterning genes known from D. melanogaster could be correlated with variations in these morphological features. In this respect, we show here expression data of the genes Dll, dac, exd, and hth during appendage formation and patterning in T. domestica and F. candida.

Materials and methods

Animal culture and embryo fixation

Cultures of T. domestica and F. candida were kept as described before (Schaeper et al. 2010). Embryos of F. candida were washed several times with deionized water, boiled for 1 min, and quickly cooled on ice. The embryos were then washed in hypochlorite solution (1:1 mix of deionized water and bleach solution (DanKlorix, Colgate-Palmolive)) for 6 min, washed with water several times, and transferred stepwise to methanol. Then the embryos were sonicated in methanol for 45 s using a Branson tip sonifier. Optimal sonication effect was tested empirically and was found to result from sonication intensities that destroyed over 50 % of the embryos completely. The remaining embryos were then washed with methanol several times, vortexed in several intervals for at least 10 min at maximum speed, washed again with methanol, and then stored in methanol at −20 °C.

The proper devitellinization of the embryos of F. candida is difficult because of the very tough vitelline membrane. Conventional methods for devitellinization (e.g., shearing, osmotic shock, mechanical removal by rolling over a sticky surface, or peeling by hand) were not successful. Therefore, the embryos were sonicated. This destroyed most of the embryos completely, but in a sufficient number of embryos, the destruction was incomplete and larger portions of the germ band that had been released from the destroyed vitelline membrane could be recovered and used for in situ hybridization. The younger germ band stages are quite delicate even after proper fixation, and it was not possible to obtain a sufficient number of fragments after sonication. We therefore focused on older embryonic stages after germ band retraction, of which larger fragments (e.g., complete heads or portions of the abdomen) could be recovered. However, this fragmentation of the embryos also limited the analysis of expression patterns and not all appendages of all the segments could be studied in detail in this study, especially the appendages in the anterior part of the abdomen. This has to await the development of better devitellinization techniques for F. candida that allow for the recovery of complete and undamaged germ bands.

Embryos of T. domestica were washed with deionized water, boiled for 1 min, and cooled on ice. Then the embryos were washed once with heptane and three times with methanol and were then transferred stepwise to phosphate-buffered saline supplemented with 0.1 % Tween-20 and 4 % formaldehyde and postfixed for 30 min. After several washes with phosphate-buffered saline, the embryos were dissected from the vitelline membranes by hand with watchmaker forceps (Dumont No. 5) and sharpened tungsten needles. Dissected embryos were transferred to methanol and stored at −20 °C.

Gene cloning

Total RNA was extracted from embryos with Trizol reagent (Invitrogen) according to the manufacturer's instructions. From the total RNA, adaptor-ligated cDNA (RACE-cDNA) was produced using the SMART PCR cDNA Synthesis kit and SMART RACE cDNA Amplification kit (Clontech). Fragments of exd, hth, and dac from both species were amplified by PCR with the previously described primer sets (Prpic et al. 2001, 2003). An initial short homeobox fragment of Dll from both species was amplified by PCR with the published degenerate primer set (Prpic and Tautz 2003). These PCR fragments were subcloned and sequenced. Based on this sequence information, species-specific primers were designed and used for RACE-PCR to amplify larger parts of the Dll sequence of each species. All gene fragments were cloned into the pCR-II vector (Invitrogen) and sequenced. The sequences were deposited in GenBank under the following accession numbers: T. domestica—Td Dll (HG326265), Td dac (HG326264), Td exd (HG326262), and Td hth (HG326263); F. candida—Fc Dll (HG326261), Fc dac (HG326260), Fc exd (HG326258), and Fc hth (HG326259). The fragments were used to transcribe digoxigenin-labeled RNA probes that were used in whole mount in situ hybridization.

Whole mount in situ hybridization

For in situ hybridization, the embryos of both species were transferred stepwise to ethanol, rehydrated stepwise in deoxycholate buffer (150 mM sodium chloride, 0.5 % sodium deoxycholate, 0.1 % sodium dodecyl sulfate, 1 mM ethylenediaminetetraacetic acid, 50 mM tris(hydroxymethyl)aminomethane-HCl), washed in pure deoxycholate buffer for 1 h, and washed twice in phosphate-buffered saline supplemented with 0.1 % Tween-20 (PBST). Then the embryos were incubated in 8 μg proteinase K in 1 ml PBST for 7 min, washed several times with PBST, postfixed in 4 % paraformaldehyde in PBST, washed several times with PBST, and incubated at 55 °C for 4 h in hybridization buffer (50 % formamide, 5× saline sodium citrate buffer (pH = 4.5), 100 μg/ml heparine, 100 μg/ml DNA from calf thymus, 0.1 % Tween-20) with pH adjusted to 5.0. After the addition of the digoxigenin-labeled RNA probe, the embryos were incubated in hybridization buffer at 55 °C overnight and the entire following day with occasional swirling. After this incubation period, the embryos were washed for a complete day with several changes of fresh hybridization buffer at 55 °C to remove all probe that did not hybridize. On the following day, the embryos were washed several times at room temperature with hybridization buffer, transferred stepwise to PBST, and then incubated in blocking solution (PBST supplemented with 2 % heat inactivated sheep serum and 10 mg/ml bovine serum albumine) for 1 h. After the addition of alkaline phosphatase conjugated antidigoxigenin antibody (concentration 1:2,000), the embryos were incubated overnight at 4 °C. Then the embryos were washed several times in blocking solution, washed several times in PBST supplemented with 1 % Tween-20, and then transferred to staining buffer (0.1 M tris(hydroxymethyl)aminomethane-HCl (pH = 9.5), 0.05 M magnesium chloride, 0.1 M sodium chloride, 0.1 % Tween-20 in water). After the addition of substrate (NBT/BCIP), the staining reaction was stopped by several washes with PBST.

Results

Expression of Distal-less in T. domestica

The pattern of Td Dll mRNA expression is similar to the protein expression pattern described previously (Rogers et al. 2002; Ohde et al. 2009). Td Dll is expressed in all developing appendages, except for the mandibles, and there is additional expression in the head lobes (Fig. 2a, b) and at the posterior end of the germ band. During germ band elongation, this posterior expression domain appears to be correlated with the developing hindgut (Fig. 2a, asterisk), but in stages after germ band elongation, the posterior expression is obviously located in the primordium of the filamentum terminale (Fig. 2b, asterisk). It is therefore likely that the earlier posterior expression domain is not correlated with the later posterior expression in the filamentum terminale, and instead, these domains represent separate functions of Td Dll. Indeed, RNAi with Td Dll leads to the shortening of the filamentum terminale, suggesting a role in filamentum growth (Ohde et al. 2009). In the antenna, maxilla, labium, and the thoracic legs, Td Dll expression initially is restricted to distal cells (Fig. 2a, b). In the antenna, this expression resolves into weaker distal expression and a stronger medial ring after full germ band extension (Fig. 2c). As soon as the endites start developing on the maxilla and on the labium, the proximal endites (i.e., the lacinia and the glossa) also show distinct expression of Td Dll (Fig. 2b (arrow), c). Td Dll is also expressed in the labrum and the pleuropodia. These limbs express Td Dll even before there are morphologically visible buds (Fig. 2a). The buds of the cerci also express Td Dll (Fig. 2b).

Fig. 2
figure 2

Expression of Distal-less in T. domestica embryos. a Embryo at early germ band elongation. The asterisk denotes posterior expression of Td Dll in the region of the developing hindgut. b Embryo at full germ band extension. The arrow points to the expression domain in the lacinia. The asterisk denotes posterior expression of Td Dll in the primordium of the filamentum terminale. c Gnathal region of an embryo at germ band retraction. Anterior is to the left in all panels. Abbreviations: an antenna, ce cerci, lb labium, lr labrum, md mandible, mx maxilla, oc ocular region, pl pleuropodium, t1–3 thoracic legs

Full size image

Expression of dachshund in T. domestica

Throughout germ band elongation, the Td dac gene is expressed in the head lobes (Fig. 3c, arrows) and in a dotted pattern in the developing ventral nerve cord, which presumably corresponds to neuroblasts or neuroblast groups. There is also expression in the area of the hindgut. In the labrum, there is a very weak lateral expression domain (Fig. 4a, arrows) and stronger expression is seen at the base of the labrum (Fig. 4a, asterisks). Expression of Td dac in the antenna is very weak and, during germ band retraction, resolves into two thin rings at a proximal and medial level, respectively (Fig. 4b). Very strong expression of Td dac is seen in the outer half of the mandible (Fig. 4c). Strikingly similar expression is present in the proximal endite of the maxilla (lacinia) (Fig. 4d). Further Td dac expression in the maxilla is present in a broad ring in the palp and at the palp base and, from the palp base, appears to extend very weakly into the base of the distal endite (galea). Expression in the labium is very similar to the expression in the maxilla (Fig. 4e). However, the expression in the proximal endite (glossa) is not visibly stronger than the remaining expression in this appendage. In the thoracic legs, there is a broad medial ring and an additional dorsal domain in the proximal part of the legs (Fig. 4f).

Fig. 3
figure 3

Expression of extradenticle (a), homothorax (b), and dachshund (c) in T. domestica embryos. a Embryo at mid-germ band extension. b Embryo at full germ band extension. c Embryo at germ band retraction. The arrows point to expression in the ocular region. The asterisk denotes the cerci in all panels. Anterior is to the left in all panels. For abbreviations, see Fig. 2

Full size image
Fig. 4
figure 4

Expression of dachshund in the appendages of T. domestica. a Labrum. The arrows point to weak expression in a group of cells on each side of the labrum and the asterisks denote expression at the base of the labrum. b Antenna. The asterisk denotes expression of Td dac in the primordial antennal ganglion. The arrows point to two thin expression rings in the medial antenna. c Mandible. d Maxilla. Note the similarity between expression in the maxillary lacinia and the mandible. e Labium. f Thoracic leg

Full size image

Expression of extradenticle and homothorax in T. domestica

During germ band formation, both genes are expressed in most cells of the embryo. The Td exd gene shows a ubiquitous and homogeneous expression throughout the germband (Fig. 3a). Ubiquitous expression is also present in all appendages, including the labrum, antenna, mandible, maxilla, labium, and legs (Fig. 5a–f). The Td hth gene is expressed in a pattern of higher and lower expression levels (Fig. 3b). Expression is weak in most parts of the body, whereas stronger expression is detected in most of the appendages. No expression is present in the posterior structures (cerci and primordium of the filamentum terminale). In the labrum, two lateral expression domains are present (Fig. 5g). In the antenna (Fig. 5h) and the legs (Fig. 5l), Td hth expression is mainly in the base, plus a ring of expression still in the proximal portion of these appendages. A similar ring is present near the root of the palp of the maxilla and the labium (Fig. 5j, k). In these appendages, Td hth is also expressed in the base and the endites. Interestingly, the expression levels in parts of the endites are higher than in the remaining tissue. In the labium, both endites and, in the maxilla, only the distal endite strongly express Td hth. No expression of Td hth was detected in the mandible (Fig. 5i).

Fig. 5
figure 5

Expression of extradenticle and homothorax in the appendages of T. domestica. a–f Expression of Td exd in the labrum (a), antenna (b), mandible (c), maxilla (d), labium (e), and leg (f). g–l Expression of Td hth in the labrum (g), antenna (h), mandible (i), maxilla (j), labium (k), and leg (l). The asterisk in h, j, k, and l denotes the expression ring at the border between the proximal and medial appendage parts

Full size image

Expression of appendage patterning genes in F. candida

In the head of F. candida embryos at germ band retraction, Fc Dll is expressed in the inner endites of the maxilla (lacinia) and the labium (glossa) (Fig. 6a). There is no expression in the palp of either appendage, in the mandibles or in the superlinguae (Fig. 6a). Strong expression of Fc Dll is detected in the distal half of the thoracic legs (Fig. 6b) and in the distalmost portion of the furca (Fig. 6c). There is also a posterior expression domain in the last body segment (Fig. 6c, asterisk). The Fc dac gene is expressed in the lacinia of the maxilla and the glossa of the labium and in the tip of the mandibles (Fig. 6e). In the legs, Fc dac is expressed in a broad medial ring (Fig. 6f). A similar but thinner ring is present in the furca (Fig. 6g). Both Fc exd and Fc hth are expressed throughout the embryo. In the legs, however, Fc hth expression is restricted to proximal leg parts (Fig. 6h), whereas Fc exd is expressed throughout the legs (Fig. 6d).

Fig. 6
figure 6

Appendage patterning gene expression in F. candida embryos. a Expression of Fc Dll. Arrows point to expression in the proximal endites of the maxilla and labium. Ventral view of the head, anterior is up. b Expression of Fc Dll in the thoracic leg. c Expression of Fc Dll in the distal tip of the furca and at the posterior end of the embryo (asterisk). Lateral view, ventral is up and anterior is to the left. d Expression of Fc exd in the three thoracic legs. e Expression of Fc dac. Ventral view of the head, anterior is up. f Expression of Fc dac in the thoracic legs. Note that the legs t1 and t3 have been damaged during the process of in situ hybridization and are lacking the distal parts. g Expression of Fc dac in the medial portion of the furca. Lateral view, ventral is up and anterior is to the left. h Expression of Fc hth in the thoracic leg. The arrowhead points to the boundary between Fc hth expressing and nonexpressing leg portions. The asterisk in a and e denotes the superlinguae. For abbreviations, see Fig. 2

Full size image

Discussion

In general, the expression patterns of the studied genes in T. domestica and F. candida are very similar to the patterns known from other insects. However, significant diversity exists for example with respect to dac, which is usually expressed in a medial ring in the appendages, but, e.g., in the legs of the cricket Gryllus bimaculatus, there are two medial rings (Inoue et al. 2002), and in the beetle Tribolium castaneum, a number of separate proximal domains exist (Prpic et al. 2001). Moreover, dac expression in the antenna is diverse, probably corresponding with the morphological diversity of this particular appendage. Another example for divergent expression patterns is Dll, that in the leg of pterygote insects is expressed in a “ring-and-sock” pattern: this is a bipartite pattern comprising a proximal ring and a distal “sock” and that develops gradually from an earlier uniform distal domain (see, e.g., Abzhanov and Kaufman 2000; Prpic et al. 2001). In the stages of F. candida and T. domestica available to us, we did not observe a ring-and-sock pattern of Dll in the legs of either species. This is also consistent with previous findings in collembolans and zygentomans: F. candida (Palopoli and Patel 1998), Tetrodontophora bielanensis and Lepisma saccharina (Scholtz et al. 1998), and T. domestica (Rogers et al. 2002; Ohde et al. 2009). This potentially apterygote insect typical expression pattern of Dll is similar to the likewise undivided Dll domain in the legs of spiders (e.g., Abzhanov and Kaufman 2000; Prpic and Damen 2004).

The use of gene expression data for serial homology inference

The segmental appendages are generally regarded as serially homologous organs, and thus, all the different appendage types trace from a common ancestral appendage and should share a common ground plan. The common elements of this ground plan are usually identified on the basis of morphological similarities of form, function, and the relative position of structures. However, in some cases, these characters are not sufficient, and therefore, the homology of several appendage parts is debated (see, e.g., Coulcher and Telford 2013). Gene expression patterns are now often used as characters to infer homology of problematic appendage parts: Due to its important role in distal appendage development (reviewed in Panganiban 2000), Dll expression has been used generally as an indicator of distal identity (e.g., Panganiban et al. 1995, 1997). Indeed, this notion also fits well with the classical interpretation of the insect mandible as an appendage lacking all distal elements: Dll expression studies in various insect species showed that the mandible is indeed lacking Dll expression (Popadic et al. 1996). However, Dll is also expressed in structures that, on morphological grounds, are believed to belong to the proximal leg. For example, some of the endites on the maxilla and the labium express Dll (e.g., Giorgianni and Patel 2004; Jockusch et al. 2004), and separate Dll expression is also present in endites in spiders and crustaceans (e.g., Williams et al. 2002; Prpic and Damen 2004). These observations contradict the notion that only distal elements express Dll. The endite expression of Dll is presumably not functionally related to Dll expression in the distal elements and rather indicates a second, endite-specific function of Dll. Thus, information about gene function is crucial when comparing gene expression patterns across appendage types and across species. A recent study in the red flour beetle T. castaneum, for example, used gene expression patterns to infer homology of the mandible to the other mouthparts (Coulcher and Telford 2013). Traditionally, the two teeth in the insect mandible are homologized with the two endites on the maxilla or the labium (see Machida 2000). Using the expression patterns of the genes dac and paired (prd), Coulcher and Telford (2013) propose that the two teeth of the mandible are in fact homologous only with the proximal endite, i.e., the lacinia in the maxilla and the glossa in the labium. Indeed, this notion is also strongly supported by the Td dac expression data in T. domestica: the expression of Td dac in the mandible (Fig. 4c) and the lacinia of the maxilla (Fig. 4d) is strikingly similar. In addition, the expression levels of Td hth are increased in the galea (Fig. 5j), but not in the lacinia or the mandible (Fig. 5i). The expression of Fc dac in the mouthparts of F. candida also supports the homology between the mandible, lacinia, and glossa (Fig. 6e). However, a different homology is supported by the expression of Dll. This gene is not expressed in the mandible and the distal endites in T. domestica and F. candida (at least in the stages available to us) and thus suggests homology between the mandible, galea, and paraglossa. Thus, the result of serial homology inferences on the basis of gene expression patterns critically depends on the gene that is used for study and might be confounded if the gene has multiple functions. And the matter is further complicated by the fact that in other insect species Dll is expressed in both the distal and the proximal endites (e.g. Giorgianni and Patel 2004). Therefore, as already discussed by Bolker and Raff (1996), gene expression data can serve as additional characters, but their value for homology inferences increases, when they are combined with morphological and gene function data and when inferences are not based on only a small number of genes.

The use of gene expression data for appendage homology inference

The same conclusion applies to the use of gene expression patterns for the inference of the appendicular nature of outgrowths. As already mentioned, the expression of Dll has been interpreted as an indicator of appendicular identity. However, the mandible is also an undisputed appendage, but lacks Dll expression. This has been explained by the interpretation that the mandible lacks a palp and therefore consists only of the coxa and its endites (Popadic et al. 1996). This interpretation, though presumably correct, is problematic per se because (1) in other appendage types, the endites express Dll; and (2) in F. candida, palps are present, but lack Dll expression. Thus, neither is a lack of palps required for the lack of Dll expression, nor is endite identity always coupled with a lack of Dll expression. In fact, the expression of Dll does not simply indicate the presence of serially homologous appendages, but mirrors the various functions of Dll and their species-specific modifications. For example, expression of Dll in the cerci and the pleuropodia of T. domestica and in the furca of F. candida likely indicates a distal function homologous to the function in the legs, the antennae, and the palps of the mouthparts. This is also supported by Dll RNAi phenotypes in T. domestica, because in Td Dll RNAi animals, all these appendages are affected in a similar way, and this also extends to the filamentum terminale, which is of disputed appendicular nature (Ohde et al. 2009). The lack of Dll expression in the palps of F. candida might be a modification of this distal role correlated with the strong reduction of the palps in this species. But the lack of Dll in a morphological outgrowth should not be interpreted automatically as the result of reduction: the superlinguae in F. candida, for example, lack Dll expression probably because they are not serially homologous to appendages.

On the other hand, expression of Dll elsewhere might not have anything to do with distal appendage development, but could indicate other functions. Expression in the labrum might be necessary for labrum formation, but apparently can be lacking in cases where the labrum is fused with the lower face (like in the clypeolabrum of F. candida). The early posterior expression domain in T. domestica might reflect a function in hindgut development. And the expression in the endites of the gnathal appendages might be correlated with the development of sensory organs rather than the endites themselves (see Mittmann and Scholtz 2001; Williams et al. 2002).

Spatial specificity of extradenticle and homothorax

The genes exd and hth are required for the determination of the proximal leg portions in D. melanogaster (González-Crespo et al. 1998; Abu-Shaar and Mann 1998). As a consequence, both genes are expressed in the proximal areas of the leg imaginal discs. However, exd (but not hth) is additionally expressed in all distal portions of the leg discs. This expression does not lead to a proximalization of the distal areas because exd and hth encode cofactors that are only stable and functional in the cell if both are coexpressed (Pai et al. 1998; Abu-Shaar et al. 1999; Berthelsen et al. 1999). Thus, the distal expression of exd in D. melanogaster is not functional because its cofactor hth is not present there. Proximal specificity of the exd–hth gene pair is thus achieved by restricting one partner, i.e., hth, to the proximal leg portion and relaxing the regulation of the other partner, i.e., exd. It has been noted previously that in several pterygote insect species (e.g., Inoue et al. 2002; Prpic and Tautz 2003; Angelini and Kaufman 2004), as well as in the crustacean Parhyale hawaiensis (Prpic and Telford 2008) and in the velvet worm Euperipatoides kanangrensis (Janssen et al. 2010), the proximal specificity of exd–hth is achieved in the same way. In spiders and myriapods, the proximal coexpression of exd and hth is conserved, but the spatial specificity is achieved in a different way (Prpic et al. 2003; Prpic and Tautz 2003). Instead of hth, the expression of exd is restricted to the proximal leg parts and the regulation of hth is relaxed. Our data on exd and hth expression in T. domestica and F. candida reveal that the spatial pattern known from pterygote insects is also present in more basal hexapod groups and thus was likely already present in the last common ancestor of all hexapods.