The Hox gene Antennapedia is essential for wing development in insects

The Hox genes encode a family of transcriptional regulators that are important in differentiating the bodies of bilaterian animals along their antero-posterior axis (Mallo and Alonso, 2013). Disruptions to individual Hox genes often lead to disruptions of traits that develop in the regions where the Hox gene is expressed (Mallo and Alonso, 2013).

In holometabolous insects, the Hox gene Antennapedia (Antp) is expressed in all thoracic segments, including the forewing and the hindwing, yet no function has been attributed to this gene regarding wing morphogenesis. In Drosophila, wing and haltere primordia can be detected in embryos even in the complete absence of Antp function in homozygous mutants of Antp (Carroll et al., 1995). In addition, a very low level of Antp protein is detected in the growing wing margin that is not the wing primordium region, indicating that forewing formation does not require Antp (Carroll et al., 1995). Similarly, no obvious phenotypes are observed in adult Tribolium elytra (forewing) or hindwing after RNA interference (RNAi) of Antp. These data suggest that wing development takes place without any Antp input (Tomoyasu et al., 2005). In contrast, the Hox gene Ultrabithorax (Ubx), which is expressed exclusively in hindwings, functions to differentiate hindwings from forewings. Forewing development in insects was, therefore, thought to occur without any significant Hox gene input (Lewis, 1978; Struhl, 1982; Carroll et al., 1995; Weatherbee et al., 1998; Weatherbee et al., 1999; Roch and Akam, 2000; Deutsch, 2005; Tomoyasu et al., 2005; Pavlopoulos and Akam, 2011; Tomoyasu, 2017; Liu et al., 2020).

Recently, the Antp dose was shown to affect wing morohology in Drosophila (Paul et al., 2021). However, the potential downstream genes of Antp in wing disc are unclear. We observed that two loss-of-function mutations in the silkworm Bombyx mori Antp gene (BmAntp), Nc and Wes, displayed abnormal wings (Nagata et al., 1996; Chen et al., 2013). These mutations have not been examined beyond the embryonic stage due to lethality, but can be maintained in heterozygous lines. The adults of these lines display reduced and malformed wings.

These two Bombyx Antp mutants share common features with Drosophila Antp mutants that are observable in embryos. The homozygous (Antp^−/−) embryos die late in embryogenesis but display a homeotic transformation of thoracic legs to antenna-like appendages (Denell et al., 1981; Wakimoto and Kaufman, 1981; Nagata et al., 1996; Chen et al., 2013). The novel wing phenotypes in Bombyx heterozygote mutants, however, suggest that Antp is affecting wing development, a role not previously documented for this gene, as embryos die long before the stage at which wings start to develop. In the present study, we have used wild-type and heterozygotes of the Wes strain (Antp^+/−), as the study aims to better understand the role of Antp in wing development.

As defective adult wings were observed in aberrant Antp Wes and Nc mutants (Antp^+/−) (Nagata et al., 1996; Chen et al., 2013), we sought to test when in development Antp input was required. We analyzed the expression profile of BmAntp in the forewing and hindwing of wild-type individuals from the 3rd day of the 5th instar to the adult stage. qRT-PCR revealed that the expression of BmAntp was maintained at a low level in the larval stage, and gradually increased and reached a peak on the 6th day of the pupal stage. Forewings expressed higher levels of Antp relative to hindwings at most times during the pupal stage (Fig. 1A). We then compared the temporal expression profile of BmAntp between mutant Wes (Antp^+/−) and wild-type individuals. BmAntp was expressed at a lower level in the mutants compared with wild-type controls from larva to moth stage (Fig. 1B).

To evaluate the effects of BmAntp expression levels on wing morphology during development, we dissected the wing discs of Wes (Antp^+/−) and wild type from the 3rd day of the last larval instar to the wandering stage larva. Wing disc size increased slowly during the larval stage and was not significantly different between Wes mutants (Antp^+/−) and wild-type individuals. Then, during the wandering stage, the wing morphology changed dramatically, leading to wing discs dysplasia in Wes mutants (Antp^+/−), which presented as folded, curled and unexpanded wing discs. Finally, the wing discs of Wes mutants (Antp^+/−) degenerated to tiny and wrinkled adult wings (Fig. 1C, Fig. S1).

To confirm the function of BmAntp in wing development, we performed RNAi injections into wild-type larvae of B. mori. We synthesized dsRNA targeting BmAntp and injected it into larvae on the 1st day of the wandering stage. qRT-PCR showed that BmAntp dsRNA efficiently reduced BmAntp transcript levels compared with controls injected with EGFP dsRNA (Fig. 1D). Nineteen out of 22 (86%) BmAntp dsRNA treated individuals had small wings, similar to the Wes mutant (Antp^+/−), whereas control silkworm adults grew their wings normally (Table S2, Fig. 1E).

To further confirm the function of BmAntp, we performed CRISPR-Cas9 injections into wild-type embryos of B. mori. We generated a genomic disruption of the BmAntp gene by targeting its first exon using four specific single-guide RNAs (sgRNAs) and the Cas9/gRNA ribonucleoprotein (RNP) delivery system (Fig. 1F). After injection, 20 eggs hatched and 18 larvae developed to the adult stage. We found that 61% of the moths (11 individuals) displayed malformed adult wings (Fig. 1G), and confirmed that various insertions and deletions were present at the location targeted by the four sgRNAs (Fig. 1H). Abnormal wings were not observed in controls injected with BmBLOS2 sgRNA, which only led to translucent larval skin. These data indicate that BmAntp is a crucial transcription factor that regulates wing development in B. mori.

We next tested whether the production of abnormal wings in BmAntp mutants was related to deficits in levels of the molting hormone 20-hydroxyecdysone (20E). We tested this hypothesis because: (1) significant differences in the size of wing discs were observed between BmAntp mutants and controls, starting from the onset of the larva-to-pupa transition (Fig. 1C); (2) a pulse of this steroid hormone normally regulates the larva-to-pupa transition; and (3) 20E is a major regulator of wing growth and development (Denell et al., 1981; Wakimoto and Kaufman, 1981).

We first examined the expression level of genes involved in the ecdysteroid biosynthesis pathway in a variety of tissues. We found that spookier, phantom, disembodied and shadow were expressed in the prothoracic gland (PG), as expected, as this is the main source of ecdysteroid synthesis in insect larvae (Struhl, 1982). In addition, the shade gene, which encodes a P450 monooxygenase that converts ecdysone into the active 20E in targeted peripheral tissues (Mizoguchi et al., 2001), was primarily expressed in the wing discs compared with the PG and hemolymph (Fig. 2A–E).

We next explored whether Wes mutants (Antp^+/−) expressed shade at different levels relative to wild-type wings, and whether this impacted levels of 20E in the wing tissue. The shade transcripts were present at higher levels in wild-type than in Wes mutant (Antp^+/−) wings, and levels reached a peak on the 4th day of the pupal stage (Fig. 2F). Titers of ecdysone measured from wing discs on that day (P4), were similar between Wes mutants (Antp^+/−) and wild-type individuals. Titers of 20E, however, were significantly lower in the Wes mutant (Antp^+/−) relative to wild-type wings at P4 and P6 (Fig. 2G,G′).

We next investigated whether the expression levels of Ecdysone Receptor (EcR) and ultraspiracle (usp) (Fujiwara and Hojyo, 1997), the receptors that bind 20E to transduce ecdysone signaling to the nucleus, were also different between Wes and wild-type individuals. This is because 20E signaling is known to upregulate expression of EcR and usp in the wings of Drosophila (D'Avino and Thummel, 2000; Schubiger and Truman, 2000). Significantly lower levels of usp, and of the two isoforms of EcR (EcRA and EcRB mRNA) were detected in the mutant compared with wild-type wings on P4 (Fig. S2). These results suggest that Antp is also regulating the expression of these genes, either directly or indirectly. The latter mechanism could involve Antp upregulating shade, which increases 20E titers in the wing cells which, in turn, upregulates EcR and usp transcription in wings.

We next sought to test whether shade was a direct target of BmAntp. We examined a 2 kb region of DNA immediately 5′ of the start site of shade for potential Antp-binding sites and found a total of five such sites (Fig. 2H,H′). To evaluate the regulatory function of DNA containing one or more of these sites on reporter gene expression, we cloned different sized fragments, containing a different number of Antp-binding sites, upstream of the reporter gene luciferase. We transfected this plasmid into BmN cells and also co-transfected BmAntp in these cells (Fig. S3). The largest fragment (−1985 to −300), containing all five Antp-binding sites, led to significantly increased luciferase activity compared with the other four fragments (Fig. 2I). These data suggest either that a regulatory region −1985 to −1470 containing a key Antp-binding site or, more likely, that all Antp sites together are required for the transcriptional regulation of shade, and that shade is likely a direct target of BmAntp.

To determine whether BmAntp protein could directly bind to the in silico identified Antp-binding sites of the shade promoter, we designed a specific biotinylated probe covering the −1985 to −1470 genomic region of shade and conducted electrophoretic mobility shift assay (EMSA) (Fig. 2J). We further validated the direct regulation of BmAntp on shade transcription with ChIP-PCR on chromatin from BmN cells expressing recombinant BmAntp protein with a FLAG-tag (Fig. 2K, Fig. S4). This experiment showed that BmAntp binds the tested genomic region in BmN cells, providing extra support for a direct activation of shade transcription by BmAntp.

In order to explore potential additional targets of Antp, besides shade, that might have contributed to the small wings of adult Wes mutants, we investigated the expression of four cuticular protein genes with a known expression profile, which matched that of Antp, in both wild type and Wes mutants. In particular, expression levels of CPH28, CPG24 and CPG9 peaked at P5, as did expression of Antp (Fig. 1B) (D'Avino and Thummel, 2000). CPG11, by contrast, was expressed primarily during the early 5th instar, and was used as a control gene (Futahashi et al., 2008). Previous work has shown that cuticular proteins are major components of insect wings, and that both EcR-mediated signaling and other transcription factors regulate their very dynamic and specific expression profiles (Riddiford et al., 2000; Petryk et al., 2003; Wang et al., 2010). qRT-PCR analysis showed that the expression levels of CPH28, CPG24, CPG9 and CPG11 in Wes (Antp^+/−) were remarkably lower than those of wild-type P5 wings (Fig. 3A). We explored the direct regulation of these four cuticular proteins by Antp by conducting Luciferase reporter assays in BmN cells with candidate genomic regions (3 kb upstream of each gene) containing putative Antp-binding sites (Fig. 3B). Increasing BmAntp levels in these cells significantly upregulated the transcription of CPH28, CPG24 and CPG9 (Fig. 3C-F), but not CPG11. A dual-Luciferase assay with CPH28 further showed that BmAntp can directly elevate the expression of CPH28 (Fig. 3G). Moreover, an EMSA and ChIP-PCR assay showed that BmAntp was able to directly bind the in silico identified Antp-binding sites in the CPH28 promoter (Fig. 3H,I, Fig. S5). These results indicate that BmAntp can upregulate the transcription of these three wing cuticular protein genes, and CPH28 is likely upregulated by a direct interaction of Antp with the promoter of this gene.

To determine whether CPH28 is essential for wing development, we knocked it down using RNAi. CPH28-siRNA was injected into 18 pupae, and the same quantity of scrambled siRNA sequence was injected in control animals. Levels of CPH28 decreased significantly in the wing discs 48 h after CPH28-siRNA injections relative to control injections (Fig. 3J). The ratio of malformed wings reached 80% after eclosion (Fig. 3K, Table S3). In contrast, all moths in the control group had normal wings (Fig. 3K). These results indicate that CPH28 is required for the generation of normal wings in silkworms.

To evaluate whether the function of Antp in wing development is conserved across other insect orders, we examined the wings of adult flies and beetles after Antp downregulation. In Drosophila, we drove expression of Antp RNAi hairpins in larval and pupal wing discs under the control of the nubbin-gal4 (nub-gal4) driver. All individuals in which Antp was knocked down had rudimentary wings that were reduced in size compared with controls (Fig. 4A-D, Fig. S6A-D). In Tribolium, we injected Antp/ptl dsRNA during the last larval stage, immediately before the onset of rapid wing growth (Tomoyasu et al., 2005). These injections led to lower mRNA levels of Antp/ptl (Fig. S7A) and to wrinkled and shortened forewings (elytra) and hindwings (Fig. 4E-J, Figs S6E-H, S7B,C). Additionally, the uniform mesonotum phenotype observed in the Antp/ptl RNAi adults was consistent with that reported by Tomoyasu and colleagues (Figs 4K,L, Fig. S7D,E) (Tomoyasu et al., 2005). These observations indicate that Antp plays a crucial role in the development of wings in Drosophila and Tribolium. Taken together, these results demonstrate that Antp participates in insect wing development in a conserved manner.

Limited experiments in previous Drosophila studies, focusing on embryonic and larval stages, likely prevented the identification of a role for Antp in later stages of wing development. Fly embryos homozygous for Antp^W10, a mutation in the Antp sequence, led to normal wing primordia, whereas ectopic expression of Antp in third instar larval wing discs had no effect on larval wing disc morphology (Carroll et al., 1995). In the present study, the nub-Gal4 driver was used to drive UAS-Antp^RNAi expression in fly wing discs. We chose this driver as its expression was first detected in late 2nd instar wing discs and persisted through late pupal wings (Cifuentes and García-Bellido, 1997). This led to a prolonged silencing of Antp expression and to malformed adult wings in Drosophila. Recently, an improved immunofluorescence assay in Drosophila wing discs using a more recent 8C11 anti-Antp antibody showed that Antp is dynamically expressed in the wing pouch from L1 to L3 larval stages (Paul et al., 2021). In addition, knockdown of Antp expression in the whole pouch of wing discs caused reduced wing size and weak margin defects, indicating that Antp was required for adult wing blade formation (Paul et al., 2021). Thus, we speculate that there is little requirement for Antp function during the embryo stage, but Antp is important for wing development in the later larval and pupal stages.

Our RNAi experiment in Tribolium castaneum also identified strong wing defects that were not previously identified in a similar RNAi experiment (Tomoyasu et al., 2005). This previous study only reported variation of mesonotum morphology (Tomoyasu et al., 2005), which was also found in our experiments. We performed the Antp RNAi experiment twice (>250 individuals) and obtained consistent defective wing morphologies that were not observed in control animals injected with dsRNA against EGFP. We speculate that the different outcomes of the two experiments might be due to the different dsAntp fragments used. We used two fragments covering a larger region of the Antp gene (922 bp) compared with the 535 bp fragment used by Tomoyasu et al. (2005). Based on the present results, we propose that Antp is necessary for wing development in Bombyx, Drosophila and Tribolium.

Recently, Antp input was found to be required for the development of two novel traits in the wings of the nymphalid butterfly Bicyclus anynana: silver scales and eyespot patterns, in both forewings and hindwings, but only minor wing growth deformities were reported (Matsuoka and Monteiro, 2021). It is possible that the role of Antp has shifted from a general wing growth role to a more specialized role in color pattern formation. This might be the case in this species and in other nymphalids where Antp expression has been visualized in the eyespots (Cifuentes and García-Bellido, 1997; Hombría, 2011). Alternatively, the mosaic disruptions obtained with this CRISPR-Cas9 experiment were insufficient to uncover a more general role of Antp in wing growth and development. Most interestingly, the effect of Antp on shade expression should be investigated in connection with 20E-mediated eyespot size plasticity in this species (Cifuentes and García-Bellido, 1997; Hombría, 2011).

We showed that Antp directly binds to the promoters of shade, a gene encoding the last step in the production of the active ecdysteroid 20E, and that 20E was produced inside wing tissues from the precursor ecdysone produced in the prothoracic gland (Petryk et al., 2003; Rewitz et al., 2006). The biosynthesis of 20E, the main hormonal regulator of molting and methamorphosis in insects (Wakimoto and Kaufman, 1981; Shahin et al., 2018), is mediated by the Halloween genes, such as spookier, shroud, disembodied, shadow and shade (Gilbert and Warren, 2005). shade is known to convert ecdysone into 20E in peripheral organs such as the fat body, midgut and Malpighian tubules (Petryk et al., 2003; Rewitz et al., 2006). As expected, the mRNA encoding shade was present at an extremely low levels in the prothoracic gland and also in the hemolymph, but at a higher levels in wing discs. Given that the mRNA expression of shade in wing discs of Wes (Antp^+/−) mutants was significantly lower than that in normal wing discs, this explains the observed lower levels of 20E, but not of ecdysone, in the wing tissue of these mutants, and associated wing disc growth disruptions.

Cuticular proteins are major components of insect wings and previous studies had already implicated the regulation of these proteins by other Hox genes (Futahashi et al., 2008; Shahin et al., 2018). A total of 52 cuticular protein genes were detected in silkworm wing discs by expressed sequence tags (Futahashi et al., 2008). One of those detected proteins, BmWCP4, has previously been shown to be regulated by the co-binding of the Hox gene BmAbd-A with BmPOUM2, in the promoter region of the gene (Ou et al., 2014). In the present study, we focused on investigating wing cuticular protein genes whose expression patterns were largely congruent with that of Antp (Shahin et al., 2018). We showed that they were remarkably downregulated in mutant (Antp^+/−) individuals, and that disruptions to one of these proteins impaired wing development. It is possible that many more additional Antp targets remain to be described.

Previous studies have assumed that the forewing is a Hox-free wing (Struhl, 1982; Weatherbee et al., 1998; Tomoyasu et al., 2005). Our data indicate that Antp is crucial for wing development in insects (Fig. 5). It does this by directly enhancing transcription of the steroidogenic enzyme gene shade in wings and, thus, controlling the synthesis of an essential growth hormone, 20E, directly in the wing tissue.

The mutant strains Wes were homeotic mutations caused by a single gene BmAntp mutation (Chen et al., 2013). The homozygous (Antp^−/−) embryos are all lethal and display a transformation of thoracic legs to antenna-like appendages. Whereas the Antp^+/− heterozygote can be completely viable, we observed that all of the heterozygote adults exhibit reduced and malformed wings. We therefore used the heterozygote Wes mutation to explore the role of Antp on wing development in this study. The wild-type strain DaZao and mutant strain Wes (Antp^+/−) were obtained from the Silkworm Gene Bank of Southwest University, China. Silkworms were reared on mulberry leaves at 25°C in ∼75% relative humidity with a 12:12 h (light:dark) photoperiod during their entire life.

The following fly stocks were used in this study: the wild-type yw and nub-gal4 enhancer trap lines (BCF391#) were obtained from Core Facility of Drosophila Resource and Technology. The UAS-Antp^RNAi (THU2760) was supplied by the Tisng Hua Fly Center. The wild-type yw were used as control flies. All individuals were incubated at 25°C.

The Tribolium castaneum GA-1 strain was used in this study. Insects were reared in whole-wheat flour containing 5% brewer's yeast at 30°C under standard conditions.

The Bombyx mori ovary-derived cell line BmN was cultured at 27°C in TC-100 medium (United States Biological) supplemented with 10% fetal bovine serum (Gibco) and 2% penicillin/streptomycin (Gibco).

Total RNA samples were isolated from wing discs, prothoracic glands, hemolymph, BmN cells and the whole beetles at different time points or under different conditions, using the MicroElute Total RNA kit (Omega) in accordance with manufacturer instructions. The cDNA was synthesized with 1 μg total RNA using the PrimeScript RT Reagent Kit with gDNA Eraser (TaKaRa). qRT-PCR was performed using a qTOWER³G system (Analytikjena) and a qPCR SYBR Green Master Mix (Yeasen). The eukaryotic translation initiation factor 4A (BmMDB probe ID sw22934) was used as an internal reference in Bombyx, and ribosomal protein S3 (rps3) was used as an internal reference in Tribolium castaneum. All experiments were independently performed with three biological replicates and the results were calculated using the 2−ΔΔCT method. Primers are listed in Table S1.

The double-strand RNA (dsRNA) of Antp, CPH28, ptl1, ptl2 and EGFP were synthesized using the RiboMAX Large Scale RNA Production System T7 kit (Promega). Approximately 100 μg of synthesized dsAntp was injected into the second chest spiracle at the first day of Bombyx larval wandering stage. We injected 0.4-0.5 μg of dsptl at the ratio of 1:1 mix ptl1and ptl2 final instar larvae of Tribolium castaneum. To knock down CPH28 expression in the silkworm pupal stage, the siRNA sites 5′-GCAGCAAUUGUUCGCACAATT-3′ and 5′-GGAAGCUUUACAUUCGGUUTT-3′ (GenePharma) for CPH28 were designed. Ten μl of siRNA (1 μg/μl) was injected from the breathing-valve into the wing disc on the 4th day of the pupal stage. In addition, after injection, all insects were reared in a suitable living environment until analysis.

We used the Gal4/UAS system to knockdown Antp gene expression in Drosophila wings. We crossed the UAS-Antp^RNAi males with nub-gal4 virgin females and then incubated them at 25°C on a yeast/saccharose medium. The wing phenotypes of F1 adults were observed.

The sgRNAs for knocking out Antp was designed by http://crispr.dbcls.jp/ and synthesized using the RiboMAXTM Large Scale RNA Production System T7 kit (Promega). Cas9 protein was purchased from Invitrogen (Thermo). The four sgRNAs and the Cas9 protein were mixed at a dose of 500 ng/μl. The mixture was incubated for 15 min at 37°C to produce a ribonuclearprotein complex (RNP) and micro-injected into the silkworm embryos within 2 h post oviposition. The injected embryos were incubated at 25°C and >90% relative humidity until they hatched. Genomic DNA of adult wings was extracted using the DNAzol (Takara) according to the manufacturer protocol. The target region was amplified using site-specific primers (Table S1). PCR products were checked by PAGE gel and sequencing approach. Related promoters are listed in Table S1. These sgRNAs synthesized in vitro were mixed with Cas9 protein and micro-injected into preblastoderm embryos of the DaZao strain.

ELISA was used to calibrate the ecdysteroid titer in wing disc of wild type and Antp mutants. Silkworm wing discs were collected from ∼50 pupae, and the pooled sample homogenized in methanol. The homogenate was centrifuged and we evaporated the supernatant at 55°C. The solid matter remaining was redissolved in 1 ml EIA buffer (Cayman Chemical) for 20E measurement and 1 ml sample diluents (BIOHJ) for ecdysone measurement, respectively. Ecdysteroid titers were assayed by an ELISA kit, according to manufacturer instructions (Cayman Chemical or BIOHJ). Absorbance was measured at 414 nm for Cayman kit or 450 nm for BIOHJ kit on a BioTek H1 microplate reader.

For 20E treatment in Bombyx and BmN cells, 20E (Adooq) was dissolved in DMSO and then diluted to the experimental concentrations with deionized distilled water. The final concentration of DMSO was 0.1% (v/v) in water. A total of 4 μg 20E was injected into larvae at the mesothoracic region on the 1st day of the larval wandering stage. An equal volume of DMSO at a final concentration of 0.1% (v/v) was used as the control. After 24 h, the wing discs were dissected in TRK lysis buffer (Omega). 20E (5 μm) was applied to BmN cells for 24 h and then collected. An equal volume of DMSO was used as the control.

The different lengths of shade, CPH28 and Antp promoters were subcloned into the pGL3-basic vector (Promega). The ORF of red fluorescent protein gene (RFP)-fused Antp was inserted into a pIZ/V5-His vector (Invitrogen) driven by the OpIE2 promoter. Different truncated promoters of pGL3-basic vector were co-transfected with pIZ/V5-His-Antp or treated with 20E at a concentration of 5 μM. After ∼24 or 48 h transient transfection, dual-luciferase activities were measured using the Dual-Glo Luciferase Assay Kit (Promega). A pRL-TK vector containing the Renilla luciferase gene was used as an internal control.

The full-length recombinant Antp nuclear proteins were extracted from E. coli strain BL21 (DE3) competent cells (TransGen). The potential Antp-binding sites of the shade and CPH28 promoters were predicted by the GENOMATIX system (http://www.genomatix.de/solutions/index.html) and JASPAR CORE (http://jaspar.genereg.net/). The DNA oligonucleotides containing Antp-binding sites were labeled with biotin at the 5′-end and annealed to generate probes. EMSA experiments were conducted according to manufacturer instructions for the EMSA/Gel-Shift Kit (Beyotime). The binding reactions were performed with 4 μg recombinant Antp protein and different amounts of biotin-labeled probes (10 pmol, 20 pmol and 40 pmol) for 30 min at room temperature. For competition assays, 40 pmol unlabeled competitor probes were added to the reaction mixture. These samples were electrophoresed on 5% polyacrylamide gels in 0.5×TBE at room temperature. The total probes are listed in Table S1.

To further detect the effects of Antp on the activity of the shade and CPH28 promoters, the ChIP assay was performed following kit instructions (GST). BmN cells were transfected with a Flag-Antp expression vector and harvested at 48 h. These cells were fixed with 37% formaldehyde, and then DNA-containing proteins were sonicated to obtain 200-1000 bp DNA fragments. The immunoprecipitation reactions were enriched with 1 μg antibody against Flag or IgG. The precipitated DNA and input were used for PCR analysis. The primers used for amplifying the sequences containing potential Antp-binding sites are listed in Table S1.

Statistical analyses were performed using GraphPad Prism 7 (GraphPad Software). The data are mean±s.e.m. The differences between two sets of data were analyzed with a two-tailed Student’s t-test. A value of P<0.05 was considered statistically significant: *P<0.05, **P<0.01 and ***P<0.001.

We thank members of the Tong lab for initial discussions on the paper and three anonymous reviewers for comments that improved the article. We are grateful to the Core Facility of Drosophila Resource and Technology and to the Tisng Hua Fly Center for fly stocks.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199841.