Regulation of the BMP Signaling-Responsive Transcriptional Network in the Drosophila Embryo

Abstract

The BMP signaling pathway has a conserved role in dorsal-ventral axis patterning during embryonic development. In Drosophila, graded BMP signaling is transduced by the Mad transcription factor and opposed by the Brinker repressor. In this study, using the Drosophila embryo as a model, we combine RNA-seq with Mad and Brinker ChIP-seq to decipher the BMP-responsive transcriptional network underpinning differentiation of the dorsal ectoderm during dorsal-ventral axis patterning. We identify multiple new BMP target genes, including positive and negative regulators of EGF signaling. Manipulation of EGF signaling levels by loss- and gain-of-function studies reveals that EGF signaling negatively regulates embryonic BMP-responsive transcription. Therefore, the BMP gene network has a self-regulating property in that it establishes a balance between its activity and that of the antagonistic EGF signaling pathway to facilitate correct patterning. In terms of BMP-dependent transcription, we identify key roles for the Zelda and Zerknüllt transcription factors in establishing the resulting expression domain, and find widespread binding of insulator proteins to the Mad and Brinker-bound genomic regions. Analysis of embryos lacking the BEAF-32 insulator protein shows reduced transcription of a peak BMP target gene and a reduction in the number of amnioserosa cells, the fate specified by peak BMP signaling. We incorporate our findings into a model for Mad-dependent activation, and discuss its relevance to BMP signal interpretation in vertebrates.

Fig 1. Identification of Dpp target genes.

(A) Race expression in dorsal and lateral views of wildtype and Tkv^QD over-expressing embryos at the onset of gastrulation. (B) RNA in situ hybridization staining of wildtype embryos with antisense probes to detect transcripts identified as being positively regulated by Dpp signaling, based on the RNA-seq analysis. Dorsal and lateral views of stage 6 and 9 embryos, respectively, are shown. (C) As in (B), except stage 6 and 11 embryos are shown. (D) Expression of selected transcripts in embryos lacking dpp or carrying 4 copies of dpp, compared to the wildtype pattern.

Fig 2. pMad and Brk ChIP-seq validation.

(A) Browser images of ChIP-seq reads normalized to the PI control for pMad (dark green, 2–2.5 h, light green, 3–3.5 h AEL) and Brk (dark blue, 2–2.5 h, light blue, 3–3.5 h AEL) for Race and zen. Called peaks are shown as black bars below the Brk and pMad tracks (top row 2–2.5 h, bottom row 3–3.5 h). The red scale bar at the top right of each browser view represents 1kb. Known enhancers are shaded in orange. (B) Venn diagrams showing the overlap between the data sets for each factor between the two time points. (C) Browser views as in (A) for putative enhancers based on pMad/Brk binding. In situ hybridizations below the browser views show the endogenous expression pattern for the indicated gene, and the lacZ reporter gene expression pattern driven by the test ChIP region. Embryos are dorsal views with the exception of the net and Ama embryos, which are shown as lateral views to highlight that the test region only drives the dorsal ectoderm part of the expression pattern. Grey shaded areas indicate the tested enhancer fragments, orange shading indicates a known enhancer. Where two are tested, the regions are labelled 1 and 2 as seen from left to right. The enhancer for zen2 is located within the pb gene. The absence of a gene structure below the tracks reflects distal enhancer positioning.

Fig 3. Dpp-EGF signaling cross-talk in the embryo.

(A) RNA in situ hybridization showing expression of the peak Dpp target genes Race, hnt and ush in stage 6 wildtype and kek1 mutant embryos (dorsal views). (B) Visualization of Race mRNA by fluorescent in situ hybridization (red) and dpERK protein (green) in wildtype and kek1 mutant embryos. Top row shows the merge image with DAPI (blue), lower panels show the separate channels for Race and dpERK. (C) As in (B) except that kek1 and wildtype embryos are stained for Race mRNA and pMad protein at stages 5 and 6, as labelled. (D) RNA in situ hybridizations showing egr mRNA in stage 5 and 6 wildtype and kek1 mutant embryos. (E) Graph showing the expression widths of the Dpp target genes in wildtype and kek1 mutant embryos. Error bars are SEM, n≥10 across different biological repeats, *P<0.0001, two-way ANOVA.

Fig 4. Negative regulation of the Dpp pathway by EGF signaling.

(A) Visualization of Race and hnt mRNA in stage 6 wildtype and egfr mutant embryos by RNA in situ hybridization. (B) Detection of Race mRNA by fluorescent RNA in situ hybridization (red) and dpERK protein (green) by immunostaining in wildtype and egfr mutant embryos (late stage 6, dorsolateral views). (C) As in (B) except that rho mutant embryos are compared to wildtype, embryos are at early stage 6 and shown as dorsal views. (D) Graph showing the expression widths of the Dpp target genes in wildtype embryos and the mutants tested in (A)-(C). Error bars are SEM, n≥10 across different biological repeats, *P<0.0001, ordinary one-way ANOVA.

Fig 5. Peak and intermediate Dpp targets are disrupted in zen mutants.

RNA in situ hybridization of stage 6 embryos (dorsal views) showing the expression patterns of the Dpp target genes hnt, ush and tup in zen mutant embryos compared to wildtype.

Fig 6. Zld is required for Dpp gradient interpretation.

(A) Graph showing enrichment of two variants of the Zld binding motif in the indicated ChIP data sets relative to that of a set of housekeeping enhancers [20]. Other variants of the Zld binding site exist but are not included here. The line drawn at 1 represents no relative enrichment. The percentage of peaks harboring the motif in each data set is shown above the bar. Enrichment of the motif relative to the control set is significant at **P<0.01 based on Fisher’s exact two-tailed test. (B) Graph showing percentage of the peaks in each of the four data sets that overlap Zld regions identified by ChIP-seq [25]. (C) Cartoon showing the position of the Smad, Zen and Zld binding sites in the 533 bp wildtype and altered Race enhancer variants. Relevant binding site sequences are shown underneath with nucleotide shading as in the cartoon (purple—Smad, orange–Zelda, green–Zen). Only part of the sequences of the Zen binding sites is shown. The Smad binding sites are as described [26], although weak Smad binding has also been reported between the upstream Zen and Zelda binding sites [22]. (D) RNA in situ hybridization with a lacZ probe of embryos carrying a transgene with either a wildtype or altered Race enhancer, as shown in (C), upstream of a lacZ reporter gene. The transgenes are integrated at the same genomic site. (E) As in (D) except that lacZ expression is directed by 3 copies of either the Zld motif, Mad motif or the combined Zld-Mad motifs. Spacing between the three Zld or Mad motifs when tested in isolation is the same as when tested together.

Fig 7. Insulator proteins bind to the pMad/Brk regions.

(A, B) Graphs showing percentage overlaps between the pMad/Brk ChIP peaks and the indicated insulator proteins using ChIP data obtained from 2–4 h or 0–12 h embryos (A) or S2 cells (B), as labeled. In graphs (A-B), regions within the DTS data set that overlap with pMad/Brk peaks were removed prior to calculating the DTS-insulator binding protein overlap, in order to provide a cleaner comparison with the pMad/Brk data. However, this removal typically only lowers the overlap between the DTS data sets and the insulator binding proteins by less than 3%, with the exception of the Chromator-DTS overlap that is reduced by 9%. (C) Graph showing the percentage of ChIP peaks in each data set that bind at least one of the insulator proteins tested in (A, B) with the exception of GAF and using only the embryo data for BEAF-32. (D) Graph showing overlap with NELF-B and NELF-E binding based on ChIP data obtained for these factors in S2 cells. (E) Graph (Ei) showing the number of amnioserosa cells, as determined by Hnt staining, in wildtype and BEAF^AB-KO null embryos (n = 30 across 3 biological repeats, error bars are SEM, *p<0.01, two tailed t-test). Representative embryos are shown in (Eii), amnioserosa staining is indicated by black arrowheads, magnified views of the regions in the dashed squares are shown. (F) RNA in situ hybridizations showing a loss of posterior Race expression in embryos collected from BEAF^AB-KO heterozygous adults, compared to wildtype. (G) Graph showing the percentage lethality of dpp^hr27 progeny from a cross of either wildtype or BEAF32^ABKO/+ females to dpp^hr27/CyO males (n = 3, >200 flies counted for each biological repeat, error bars show SEM, *p<0.05, two tailed t-test).

Fig 8. Models of pMad-dpERK antagonism and pMad activation.

(A) Model showing the equilibrium between the positive and negative effects of Dpp and EGF, respectively, on pMad. The top panels show dpERK-pMad dynamics in a wildtype or kek1 mutant cell at the dorsal midline (labeled Dorsal) compared to one positioned more dorsolaterally (Dorsolateral) at stages 5 and 6, as labelled, with the various proteins indicated in the key. The dpERK-pMad equilibrium is disrupted in kek1 mutant embryos leading to reduced pMad. For each stage and genotype, the resulting dpERK, pMad and Race activation/expression domains within the whole embryo are depicted in the lower cartoons. Cartoons (stages as described above) are also shown for the EGF loss-of-function mutants tested. (B) Stepwise model of pMad-dependent activation, involving Zld, Zen and insulator binding proteins.