Maternal bicoid transcripts are localized at the anterior pole of the embryo, but BICOID protein is distributed in an anterior-to-posterior concentration gradient across the early embryonic syncytium (20–22). The gene is required before gastrulation for the spatially restricted activation of segmentation genes such as hunchback, Krüppel, and knirps and for translational repression to prevent maternal caudal activity in the anterior portion of the embryo (23–28). bicoid is not expressed at other stages of the life cycle. The maternal expression of bicoid, as well as a role in anterior patterning, is conserved throughout Cyclorrhaphan flies (Fig. 1A, B) (8, 14, 29, 30). The zygotic zerknüllt transcripts and the ZERKNÜLLT protein are expressed on the dorsal side of the early embryo (11, 17). During precellular nuclear division cycles 11–13, zerknüllt is activated in a broad dorsal-on/ventral-off pattern by an unknown factor. Expression is subsequently maintained by DECAPENTAPLEGIC, but during cellularization in nuclear division cycle 14 expression is restricted to a narrow dorsal strip, where it is required for the specification of the extraembryonic anlage (31, 32). In addition, zerknüllt is expressed in a subset of pole cells (germ line), where its function remains obscure (11). zerknüllt is not expressed at other stages of the life cycle. In the lower Cyclorrhaphan fly Megaselia abdita (Phoridae), zerknüllt transcripts are restricted to the extraembryonic anlage/tissue, and neither the early broad expression nor the expression in the pole cells known from Drosophila is observed (Fig. 1 C and D) (15). Zygotic expression of Hox3/zerknüllt in the extraembryonic anlage/tissue is conserved throughout insects, although the topology of the anlage varies (10, 33). In addition, maternal expression of Hox3/zerknüllt has been reported in the locust Schistocerca gregaria (Orthoptera), a rather primitive winged insect (34). This finding raises the question of whether bicoid and zerknüllt evolved from a progenitor that had combined functions of the two genes in higher Diptera. However, in the beetle Tribolium castaneum (Coleoptera), Hox3/zerknüllt apparently is not expressed maternally (10, 34), raising a question as to the likely ancestral state of expression. Here, we describe Hox3/zerknüllt homologues from three basal Dipteran species, the dancefly Empis livida (Empididae), which belongs to the sister taxon of Cyclorrhaphan flies, the horsefly Haematopota pluvialis (Tabanidae), a very basal species in the suborder Brachycera, and the mothfly Clogmia albipunctata (Psychodidae). The results suggest that maternal zerknüllt expression was lost in the stem lineage of Cyclorrhaphan flies concomitantly with a reorganization of extraembryonic tissue.

Figure 1 Maternal expression of bicoid genes versus zygotic expression of zerknüllt genes in Cyclorrhaphan flies. (A–D) Transcript distribution of Megaselia-bicoid (Ma-bcd) and Megaselia-zerknüllt (Ma-zen) in egg chambers (A and C (more ...)

Species. Clogmia (syn. Telmatoscopus, Psychoda) albipunctata Williston 1893 (Diptera; Psychodidae) and Megaselia. abdita Schmitz 1959 (Diptera: Phoridae) were reared in the laboratory. Our Clogmia and Megaselia strains were derived from samples collected by Klaus Sander (Albert-Ludwigs-Universität, Freiburg, Germany) in Turkey near Seljuk (Ephesus) and in Freiburg, respectively. Clogmia was kept on moist paper sprinkled with stinging-nettle powder, and synchronized Clogmia embryos were obtained after experimental egg activation in water (35). Megaselia was kept on wet paper towels sprinkled with aquarium fish food TetraRubin (Tetra, Melle, Germany). For a description of embryonic development of both species see refs. 36 and 37. Haematopota pluvialis L. 1758 (Diptera, Brachycera, Tabanidae) (38) and Empis (subgenus Kritempis Collin) livida L. 1758 (Diptera, Brachycera, Empididae) (39) were caught in the rural environment of Göttingen (Germany) and fed on honey water. To obtain developing ovarian egg chambers from Haematopota, females were fed on human blood and kept for 2–4 days at 21°C before the ovaries were dissected.

Cloning of Homologues. Homeobox fragments of Hox3/zerknüllt homologues were amplified from genomic DNA by PCR with use of the degenerate primer pairs AARMGIWSIMGIACNGCNTWYACNAGT/TTYTTRWAYTTCATICKICKRTTYTG for Clogmia, and CARCTBGTDGARCTIGARAAYGARTT/TTYTTRWAYTTCATICKICKRTTYTG for Haematopota and Empis. (PCR protocols are available on request.) cDNAs were obtained by 5′ and 3′ rapid amplification of cDNA ends (RACE) by using for Clogmia the SMART RACE, and for Haematopota and Empis the Marathon cDNA Amplification Kits (CLONTECH). cDNA was prepared from ovarian poly(A)⁺ RNA. Sequence alignments were performed by using clustal method alignment parameters in the MEGALIGN/DNASTAR software package.

Whole-Mount in Situ Hybridization. Egg chambers were dissected in PBS (140 mM NaCl/3 mM KCl/7 mM Na₂HPO₄/3 mM NaH₂PO₄, pH 7.8) and fixed in 178 μl of PBS/22 μl of formaldehyde solution (37%), 20 μl of dimethyl sulfoxide, and 500 μl of n-heptane for 25 min. Clogmia embryos were dechorionated with 50% commercial bleach for 2 min and fixed in 350 μl of PEMS (100 mM Pipes/2 mM MgSO₄/100 mM malic acid, pH 6.9), 45 μl of formaldehyde solution (37%), and 500 μl of n-heptane for 40 min. To remove the vitelline layer, the fixing solution was replaced by 500 μl of n-heptane, to which 500 μl of ice-cold methanol (−80°C) was added. After vigorous shaking, the embryos were washed with methanol (−80°C) and heated to +70°C. The temperature shock was repeated twice, and in situ hybridization was done with digoxygenin-labeled RNA antisense probes covering the ORFs of the respective genes following standard procedures as described (40).

Northern Hybridization. Total RNA was isolated from dissected ovaries by using TRI REAGENT solution (WAK-Chemie Medical, Bad Soden, Germany) following the producer's instructions. Poly(A)⁺ RNA was prepared from total RNA according to the instructions of the Oligotex mRNA Kit (Qiagen, Hilden, Germany). One microgram of Megaselia mRNA, 1 μg of Drosophila mRNA, 350 ng of Clogmia mRNA, 500 ng of Haematopota mRNA, and 750 ng of Empis mRNA were separated in a denaturing 1% agarose gel together with the 0.24- to 9.5-kb RNA Ladder (Life Technologies, Rockville, MD). Separated RNA was transferred to nylon membrane (Hybond-N+, Amersham Pharmacia) (41) and hybridized in Rapid-hyb buffer (Amersham Pharmacia) at 65°C for several hours to radio-labeled DNA probes. Nonspecifically bound probe molecules were removed by washing in 0.2× SSC (3 mM sodium citrate/30 mM NaCl, pH 7) and 0.1% SDS at 65°C for 30 min.

Isolation of Hox3 Genes from Lower Diptera. A survey of the literature on bicoid and zerknüllt gene expression suggested that in Cyclorrhaphan flies bicoid homologues are expressed only maternally, whereas zerknüllt homologues are expressed only zygotically (see Introduction). We confirmed lack of maternal zerknüllt expression in Megaselia (Fig. 1 C–E) and Drosophila (Fig. 1E; data not shown) by whole-mount in situ staining of ovarian egg chambers and Northern blotting experiments. Next we asked how Hox3 genes are expressed in lower (non-Cyclorrhaphan) flies. To identify Hox3 genes in lower Diptera, we took a PCR-based approach by using a variety of degenerate primers to amplify both bicoid and zerknüllt homologous DNA fragments from the mothfly Clogmia (Psychodidae), the horsefly Haematopota (Tabanidae), and the dancefly Empis (Empididae) covering a wide range of non-Cyclorrhaphan Diptera (Fig. 2A). In each species, we identified a single Hox3 homeobox fragment corresponding to a zerknüllt-type gene. cDNAs covering all the ORFs of the genes were obtained by 5′ and 3′ rapid amplification of cDNA ends on templates prepared from ovaries. The cDNAs were 1,243 bp (Clogmia), 2,458 bp (Haematopota), and 1,681 bp (Empis) in length and are available under GenBank accession numbers AJ419659, AJ419660, and AJ419661, respectively. The predicted amino acid sequences of the full-length proteins were used in protein database searches with BLAST software (42) and produced highest scores with vertebrate HoxA3 genes (Haematopota) or insect zerknüllt genes (Empis, Clogmia). Alignment of the predicted amino acid sequences revealed two conserved sequence stretches C-terminal to the homeodomain (Fig. 2B). These results establish that the newly identified sequences are direct zerknüllt homologues referred to as Ca-zen for Clogmia, Hp-zen for Haematopota, and El-zen for Empis. PCR fragments of the bicoid type were not found, although 11 degenerate primers against conserved sequence stretches were tried in various combinations on genomic and cDNA. The findings suggest there are no direct bicoid homologous genes in any of these three species. Thus, the newly identified Hox3 sequences could be orthologous to both zerknüllt and bicoid if both genes originated from a gene duplication in the stem lineage of Cyclorrhaphan flies, or they could be orthologous to zerknüllt only if the duplication occurred before the radiation of Diptera. Gene trees with the homeodomains of zerknüllt and bicoid homologues from Cyclorrhapha and other Hox genes were already known to support a bicoid/zerknüllt kinship (ref. 15; data not shown). When comparing Hox3/zerknüllt from a wider selection of insect species, the relationship of labial (Hox1) and proboscipedia (Hox2), bicoid homologues, and Hox3/zerknüllt homologues is not resolved because of the diverged Hox3/zerknüllt sequences. Therefore, we assumed that Hox3 was duplicated after the basal radiation of Diptera and asked whether expression of Ca-zen in Clogmia differs from zerknüllt expression in Cyclorrhaphan flies.

Figure 2 (A) The phylogenetic relationship of all higher level taxa of Diptera including an estimated 125,000 species (43, 44). Relative species abundance in each taxon is indicated by the width of each triangle. Arrows indicate strongly supported monophyletic (more ...)

Maternal and Zygotic Expression of Ca-zen. Ca-zen is strongly expressed in the germ-line cells of ovarian egg chambers. The transcripts are detected in the nurse cells (“n”, Fig. 3A) and the oocyte (“o”, Fig. 3A), and seem to be evenly distributed in the early embryo (Fig. 3B). To corroborate maternal expression of Ca-zen, we performed a Northern blot analysis with mRNA prepared from ovaries. A single band of expected size is obtained after hybridization with a Ca-zen probe (Fig. 3C). Zygotic expression of Ca-zen starts during cellularization of the blastoderm in an anterior and dorsal domain, which corresponds to the anlage of extraembryonic tissue (Fig. 3 D and E). Extraembryonic Ca-zen expression is maintained during gastrulation, but no zygotic Ca-zen expression outside the extraembryonic anlage/tissue is observed. The expression pattern of Ca-zen differs in two important ways from conserved zerknüllt expression in Cyclorrhaphan flies (Fig. 1D). First, Ca-zen is expressed maternally, whereas in Cyclorrhapha zerknüllt genes are expressed strictly zygotically. Second, zygotic Ca-zen expression extends to the anterior tip of the cellular blastoderm, whereas zerknüllt expression in Cyclorrhapha is restricted to a narrow dorsal strip at the same developmental stage. Thus, Ca-zen in Clogmia combines expression characteristics of bicoid and zerknüllt in Cyclorrhapha (Figs. 1 A–D and 3 A–D).

Figure 3 Expression of Hox3/zerknüllt genes in lower Diptera. (A and B) Ca-zen in nurse cells (n) and oocyte (o) of an ovarian egg chamber (A) and an early embryo (B). (C) Northern blot of ovaries from Clogmia (Left), Haematopota (Middle (more ...)

Maternal Expression of Hox3/zerknüllt Was Lost in Cyclorrhaphan Flies. To determine the relevance of maternal Hox3/zerknüllt expression for reconstructing the evolution of bicoid, it is important to know whether maternal expression of Hox3/zerknüllt homologues occurs throughout lower Diptera. Therefore, we analyzed the expression of Hp-zen and El-zen in ovarian egg chambers of each species by whole-mount in situ hybridization and Northern blot analysis. Hp-zen transcripts are strongly detected in the nurse cells and in the oocyte of Haematopota ovarian egg chambers (Fig. 3F), and a single band of expected size is detected in Northern blots by using ovarian mRNA and a labeled Hp-zen probe (Fig. 3C). In ovarian egg chambers of Empis, El-zen transcripts are first detected in distal nurse cells (Fig. 3G). Only toward the end of oogenesis are transcripts also found in the proximal nurse cells and in the oocyte (Fig. 3H). A projection of these results on the phylogenetic tree of Diptera suggests that maternal expression of Hox3 genes of the zerknüllt type is lost in the stem lineage of Cyclorrhaphan flies. This conclusion obviously depends on the resolution of the phylogenetic tree of Diptera. Therefore, it is important to note that the monophyly of the higher ranked taxa Schizophora, Cyclorrhapha, Empidoidea, and Brachycera is strongly supported (43, 44). In addition, this tree is supported by the fossil record of flies, because the appearance of these taxa in the fossil record matches their appearance as inferred from the phylogenetic tree (45).

Correlation between the Reorganization of Extraembryonic Tissue and Loss of Maternal Hox3/zerknüllt Expression during Dipteran Evolution. The maternal expression of all three newly identified Hox3/zerknüllt homologues implies selection for this trait in lower Diptera and release from a corresponding specific constraint in Cyclorrhapha. To understand better the changing developmental constraints during the evolution of Diptera, we compared embryonic development throughout this taxon. Embryonic development in Diptera seems rather uniform and resembles that of Drosophila. However, an important difference within Diptera occurs with respect to extraembryonic tissue organization (Fig. 4). The establishment of extraembryonic tissue requires the activity of a Hox3/zerknüllt gene not only in Drosophila (17, 46, 47), but most likely in all winged insects (10, 34). In species of several insect orders and, in particular, in all non-Cyclorrhaphan flies analyzed so far, including the three species of this study, extraembryonic tissue consists of an amnion and a serosa. These two epithelia do not contribute to the embryo proper but transiently wrap the embryo (Fig. 4) (36, 48–50). In contrast, Cyclorrhapha including Megaselia and Drosophila develop without such wrapping, and the extraembryonic tissue is reduced to a transient dorsal epithelium, termed amnioserosa (Fig. 4) (15, 49, 51) and, as recently discovered in Drosophila, some additional cells surrounding the yolk (52). Thus, the transition in extraembryonic tissue organization in the stem lineage of Cyclorrhaphan flies occurred in a period when maternal expression was lost in the zerknüllt-type Hox3 genes.

Figure 4 Schematic cross-sections of Dipteran embryos showing extraembryonic epithelia. Cyclorrhaphan (red) and non-Cyclorrhaphan Diptera (blue) differ with respect to the formation of an amniotic cavity and the serosa. The ventral side of the germ band is (more ...)

Model for the Evolution of bicoid and zerknüllt. On the basis of our findings, we propose that bicoid and zerknüllt evolved in the stem lineage of Cyclorrhaphan flies from a Hox3 gene with maternal and zygotic expression, which is still found in non-Cyclorrhaphan Diptera (Fig. 5). In the common progenitor, zygotic activity was required for extraembryonic development, a feature conserved by the Cyclorrhaphan zerknüllt genes. Maternal activity of Hox3/zerknüllt homologues is not understood currently and will require adopting methods for gene inactivation in non-Cyclorrhaphan Diptera (29, 53). However, because maternal expression of Hox3/zerknüllt homologues is conserved in all non-Cyclorrhaphan Diptera analyzed so far, maternal activities of these genes are important for development. We obviously still lack an understanding of how maternal Hox3/zerknüllt activity turned into maternal bicoid activity, one of the problems being the different DNA- and RNA-binding properties of BICOID compared with all other Hox genes (28, 54–59). In Drosophila, ectopic expression of zerknüllt induces extraembryonic tissue (60) and ectopic expression of bicoid induces anterior embryonic structures (61, 62). Thus, both genes have counteracting effects and cannot convert their respective activities in the same spatial domain of the embryo. Therefore, a separation of the functional expression domains of bicoid and zerknüllt in time and space, as well as selective loss of maternal versus zygotic enhancer elements, seems to be an important prerequisite for subsequent divergent evolution of both genes in Cyclorrhapha. We suggest that anterior localization of bicoid, which is based on specific sequence elements in the 3′ untranslated region of the gene (63), and respecification of anterior blastoderm toward an embryonic fate were important steps toward this goal. In summary, the key features of our model are as follows: a single Hox3 gene with maternal and zygotic activity is present in the stem lineage of Diptera; it was duplicated in the stem lineage of Cyclorrhapha, giving birth to maternal bicoid and zygotic zerknüllt; and the functional evolution of BICOID-specific DNA- and RNA-binding properties became possible after the reduction of the extraembryonic anlage/tissue. It will be challenging to test this model at the levels of genomics, developmental genetics, and morphology.

Figure 5 Hypothetical scenario of the evolution of bicoid and zerknüllt. The common ancestral gene is thought to have been expressed maternally (hatched) providing transcripts throughout the early embryo, and zygotically (filled) in the extraembryonic (more ...)

We thank Herbert Jäckle (Max-Planck-Institut für biophysikalische Chemie, Göttingen) for support, Klaus Sander (Albert-Ludwigs-Universität, Freiburg) for the cultures of Megaselia and Clogmia, Marcel Leclercq (Beyne Heusay), Andreas Stark (Halle) for help with the determinations of H. pluvialis and E. livida, and our colleagues Ralf Pflanz, Martin Zeidler, and Philip Shaw for very helpful comments on the manuscript. We thank Diethard Tautz (Universität zu Köln) for the suggestion to use malic acid in fixation protocols. This work was supported by the Max-Planck-Gesellschaft and the Deutsche Forschungsgemeinschaft (U.S.-O.).