Etv6 activates vegfa expression through positive and negative transcriptional regulatory networks in Xenopus embryos

VEGFA signaling controls physiological and pathological angiogenesis and hematopoiesis. Although many context-dependent signaling pathways downstream of VEGFA have been uncovered, vegfa transcriptional regulation in vivo remains unclear. Here, we show that the ETS transcription factor, Etv6, positively regulates vegfa expression during Xenopus blood stem cell development through multiple transcriptional inputs. In agreement with its established repressive functions, Etv6 directly inhibits expression of the repressor foxo3, to prevent Foxo3 from binding to and repressing the vegfa promoter. Etv6 also directly activates expression of the activator klf4; reflecting a genome-wide paucity in ETS-binding motifs in Etv6 genomic targets, Klf4 then recruits Etv6 to the vegfa promoter to activate its expression. These two mechanisms (double negative gate and feed-forward loop) are classic features of gene regulatory networks specifying cell fates. Thus, Etv6’s dual function, as a transcriptional repressor and activator, controls a major signaling pathway involved in endothelial and blood development in vivo.

V ascular endothelial growth factor A (VEGFA) signaling is critical for both physiological and pathological processes in the adult, including hematopoiesis, angiogenesis and solid tumour progression. In hematopoiesis, VEGFA regulates hematopoietic stem cell (HSC) survival and function in the bone marrow through both cell intrinsic and extrinsic mechanisms 1,2 . In angiogenesis, VEGFA is required for endothelial cell proliferation, migration and organization in three dimensions to form new vessels during vascular remodelling 3 . Finally, in cancer, VEGFA signaling plays essential roles in the regulation of tumour angiogenesis and metastasis 4 , and has emerged as a key antiangiogenic target in a wide variety of cancer therapies 5 . However, inhibitors of VEGFA-mediated signaling often trigger limited responses in both human tumours and mouse models of cancer, due to the development of evasive and intrinsic resistance mechanisms 5 .
VEGFA signaling is also essential during embryogenesis where it plays pivotal roles in the development of the endothelial and hematopoietic systems. Knockout of VegfA or VegfA receptor (Vegfr1, 2) genes in mice results in early lethality owing to severe defects in vascular development 6 . Mice deficient in Vegfr2 (also known as Flk1 or Kdr) show an absence of yolk sac blood islands and reduction of CD34 + hematopoietic progenitors 7 . In vitro, VEGFA is required for the specification of FLK1 + mesoderm into hematopoietic and cardiovascular lineages in mouse embryonic stem cell differentiation models 8,9 . In embryos, vegfa is essential for HSCs emergence and has been shown to be required at several stages of their programming [10][11][12] . In Xenopus, acting in a paracrine manner from the somites, Vegfa is required for the establishment of definitive hemangioblasts (DHs, precursors of the dorsal aorta (DA) and HSCs) in the dorsal lateral plate (DLP) mesoderm 11,12 . Later, Vegfa secreted by the hypochord guides the migration of DA precursors from the DLP to the midline 11,12 .
Although critical to understand the development of the vascular and hematopoietic systems, very little is known about the regulation of the complex spatio-temporal expression pattern of vegfa during embryogenesis. Because developmental and tumour biology and, in particular, angiogenic and metastatic processes share signaling and transcriptional pathways 13 , a better understanding of the regulatory networks lying upstream of VEGFA during specification of the hematovascular lineage may help identify new druggable targets relevant to VEGFA-dependent diseases.
ETV6, a transcriptional repressor belonging to the ETS family of transcription factors (TFs) [14][15][16] , is a key regulator of hematopoietic, angiogenic and tumorigenic processes, and has been linked to vegfa regulation. It is essential for bone marrow hematopoiesis 17,18 and is involved in many chromosomal rearrangements which lead to childhood leukaemias 19 . Additionally, germline and somatic mutations in ETV6 have recently been shown to be involved in other malignancies such as skin, colon, mammary and salivary gland cancers [20][21][22] . In mouse embryos, Etv6 homozygous deletion leads to early embryonic death because of defective yolk sac angiogenesis 23 .
Previously, we showed that Etv6 specifies DHs and HSCs in developing Xenopus embryos through positive regulation of vegfa expression in the somites at stage 22 of development 12 . Because Etv6 is considered a transcriptional repressor 14,15,24 , this suggested an indirect regulatory mechanism of action whereby Etv6 directly represses expression of a vegfa transcriptional repressor. To test this hypothesis, we set out to investigate the mechanism by which Etv6 activates vegfa expression in the somites. Using genome-wide and functional analyses, we identify Etv6 direct target genes and show that Etv6 employs multiple regulatory mechanisms. To allow vegfa expression, Etv6 directly represses foxo3, a known transcriptional repressor of VegfA in cancer cells 25,26 , thus participating in a double negative gate. Unexpectedly, Etv6 also directly activates expression of klf4, a known activator of VegfA in endothelial cells and a regulator of metastatic processes 27,28 . Consistent with a feed-forward loop mechanism, Etv6 also binds the vegfa promoter through a Klf4dependent mechanism. Thus, our work demonstrates that vegfa expression in the somites is regulated by a complex gene regulatory network (GRN) with Etv6 acting both as a repressor and, unexpectedly, an activator. These findings further our understanding of the in vivo regulation of vegfa and provide a platform for the study of the processes underlying VEGFA-dependent tumorigenesis, such as those involving FOXO3 and KLF4, and for the establishment of future therapeutic strategies.

Results
Etv6 genome-wide occupancy in the somites of Xenopus embryos. To investigate the mechanism by which Etv6 regulates vegfa expression in the somites, we first set out to identify Etv6 genomic targets through ChIP-seq (Fig. 1a). To this end, we generated ChIP-grade anti-Xenopus-Etv6 polyclonal antibodies in rabbits. Three short peptides were designed (Supplementary Figure 1a) and used to generate six polyclonal antibodies (two/ peptide). Four of these antibodies detected Etv6 protein on western blots (Supplementary Figure 1b) and were tested for their ability to immunoprecipitate endogenous Etv6 protein from protein extracts generated from the somites of stage 22 embryos. Two antibodies (Etv6-2a and Etv6-2b) immunoprecipitated endogenous Etv6 protein (Supplementary Figure 1c). Etv6-2a showed greater affinity to Etv6 and, therefore, was selected for ChIP-seq experiments. Chromatin immunoprecipitation was performed on three independent biological samples obtained from the somites of wild-type (WT) stage 22 embryos (Fig. 1a). Indexed libraries were generated from immunoprecipitated DNA and control input, samples were pooled and sequenced, and reads mapped to the X. laevis genome.
A total of 9128 Etv6 consistent peaks across the three biological replicates were identified using DiffBind 29 (Supplementary Data 1). The genomic distribution of the peaks revealed Etv6 occupancy over distinct features including transcription start sites (TSS), transcription termination sites (TTS), exons, introns and intergenic regions (Fig. 1b). Additionally, 22.6% of the peaks mapped to genomic regions not yet annotated (NA). For all subsequent analyses, and as a first step to identify Etv6 direct transcriptional targets, we focused on the 2440 peaks (26.7%) located in TSS regions that could confidently be associated with 2416 genes. In Fig. 1c, to illustrate the quality of the Etv6 ChIPseq data, we show the profiles of two genes with the highest enrichment of Etv6 in their TSS region (rimklb and Xelaev18005411m). Tal1, a hematopoietic gene not expressed in the somites at stage 22 11 , is an example of a gene devoid of Etv6 peaks at its TSS region.
Next, we performed de novo motif analysis on the sequences underlying all Etv6 peaks associated with TSS regions using Homer software 30 . Etv6 peaks showed significant enrichment of consensus binding sequences corresponding to known TFs (Fig. 1d). Surprisingly, ETS family binding motifs were not represented (Supplementary Table 1). Instead, the most overrepresented motif belonged to the Klf/Sp family, followed by consensus motifs for Nfy, Creb, Meis and Fox TFs. A similar result was obtained when the de novo motif analysis was performed on all 9128 Etv6 peaks across the genome, as the most overrepresented binding motif corresponds to the Klf/Sp family (28.29%, Supplementary Data 2). Only a small fraction of the peaks (3.31%) was found to be enriched with ETS-binding motifs (TACTTCTT, Supplementary Data 2). This suggested that, globally, Etv6 is preferentially recruited to chromatin indirectly, through interaction with other DNA-binding TFs such as members of the Klf/Sp family.
Establishing the somitic transcriptome regulated by Etv6. In Xenopus embryos, vegfa expression in the somites at stage 22 of development is completely dependent on Etv6 12 . To determine the transcriptional targets of Etv6 in this tissue and stage of development, we compared the transcriptome of WT and Etv6deficient somites, the latter generated with a previously validated etv6 antisense morpholino oligonucleotide (MO) 12 (Fig. 2a). Spearman correlation analysis and principal component analysis (PCA) revealed that etv6 deficiency causes significant changes in the transcriptome of the somites (Fig. 2b, Supplementary Figure 2). Indeed, differential expression analysis identified a total of 5186 differentially expressed genes (DEGs, FDR <0.05) with 2236 genes normally repressed by Etv6 (upregulated in Etv6-deficient somites) and 2950 genes normally activated by Etv6 (downregulated in Etv6-deficient somites) (Fig. 2c To validate the RNA-seq data, we next confirmed the differential expression of selected genes in WT and Etv6deficient embryos by Whole-mount in situ hybridization (WISH) (Fig. 2d). During vertebrate development, the homeobox genes meox1 and meox2 regulate somitogenesis and myogenesis 31 . In zebrafish, meox1 also represses the expansion of somite-derived endothelial cells thereby limiting HSC development in the DA 32  meox2.S) and WISH confirmed higher expression in Etv6deficient somites (Fig. 2d). Crim1, a regulator of VEGFA autocrine signaling in endothelial cells 33 , is positively regulated by Etv6 (WT/etv6 MO log 2 FC = 0.68). WISH analysis confirmed downregulation of crim1 expression in the somites of Etv6-deficient embryos (Fig. 2d). Sox18, a TF required for the development of blood vessels 34 , is positively regulated by Etv6 in the somites, as revealed by RNA-seq and confirmed by WISH (WT/etv6 MO log 2 FC = 1.01, and decreased expression in the somites of Etv6deficient embryos, Fig. 2d). Finally, we have previously reported that vegfa expression in the somites is activated by Etv6 12 . Intriguingly, vegfa was not identified as a differentially expressed gene by RNA-seq. WISH analysis did, however, confirm that vegfa is absent in the somites of Etv6-deficient embryos. It also revealed strong vegfa expression in the hypochord of both WT and Etv6deficient embryos (Fig. 2d), thus providing an explanation for the absence of vegfa from the DEG list, as the explants used for RNA extraction contained the hypochord, the unperturbed expression of vegfa in the Etv6-deficient hypochord masked its downregulation in the somites. More importantly, this demonstrates that Etv6 specifically controls the expression of vegfa in the somites. In conclusion, our RNA-seq data reveals a robust transcriptional response to etv6 knock-down.
Identification of Etv6 direct target genes. To identify Etv6 direct transcriptional targets, we compared the list of genes harbouring Etv6 peaks in their TSS regions (2416 genes) to the list of DEGs (5186 genes). This revealed 540 putative direct target genes with 545 peaks in their TSS regions (Fig. 2e, Supplementary Data 5). Surprisingly, around two thirds of Etv6 target genes (339/540, 63%) were downregulated in the somites of Etv6-deficient embryos, strongly indicating that Etv6, typically considered as a transcriptional repressor 19 , can also act as an activator of gene expression in the somites at stage 22. De novo motif analysis on the sequences underlying the peaks linked to genes normally activated by Etv6 identified Klf/Sp, Nfy and Creb motifs. Notably, only 9.94% of these peaks contained ETS motifs (CTTCCGCCCTTT and GGCCGGCAGTGT, Supplementary Table 2). Regarding the genes normally repressed by Etv6, their Etv6 peaks contained Klf/Sp, Nfy, Creb and Hox motifs, and were not enriched in ETS-binding cis-elements (Supplementary Table 2). This is similar to what was observed in de novo motif analysis on all Etv6 ChIP-seq peaks and peaks at TSSs: ETSbinding motifs are present in only a small fraction of the peaks (see above and Supplementary Table 1 and Supplementary Data 2). Therefore, Etv6 recruitment to all its genomic targets, including the genes it normally regulates transcriptionally, largely depends on additional DNA-binding TFs.
To identify the transcriptional regulators that could mediate Etv6 control of vegfa expression, we next focused on the 18 TFs and chromatin modifiers detected amongst the Etv6 putative direct targets with fold-change in gene expression >1.5 (Fig. 2f).
Many of these transcriptional regulators are involved in the control of vegfa expression in higher vertebrates [35][36][37][38][39] . Amongst those, foxo3 (Forkhead box o3), a transcriptional repressor of vegfa in cancer cells 25,26 , and klf4 (Krüppel-like factor 4), a transcriptional activator of vegfa in endothelial cells and a regulator of metastatic processes 27,28 , are known to exert their functions through direct binding to the VegfA promoter 25,27 . In Xenopus somites, their expression is respectively repressed and activated by Etv6 (Fig. 2f). For these reasons, we hypothesized that Foxo3 and Klf4 may directly contribute to the regulation of vegfa expression in the somites downstream of Etv6.
Etv6 prevents Foxo3-mediated repression of vegfa. To investigate whether Etv6 could be regulating vegfa expression by repressing foxo3 expression, we first confirmed the binding of Etv6 to both foxo3 genes in the X. laevis genome. X. laevis is an allotetraploid organism with a genome organized into two homologous sub-genomes, referred to as L and S, and most genes (>56%) are represented by two distinct homologous copies which can be regulated differently 40 . ChIP-seq tracks in Fig. 3a show that Etv6 peaks are present in the TSS region of both foxo3.L and foxo3.S and Etv6 binding in these regions was confirmed by ChIP-qPCR (Fig. 3b). These peaks contained one Ets (5'-G(A/T) GGAAG(G/T)-3') and several Klf/Sp binding motifs. Furthermore, upregulation of foxo3 in the somites of Etv6-deficient embryos was confirmed by WISH with probes designed to target both foxo3.L and foxo3.S (Fig. 3c) and RT-qPCR on RNA extracted from stage 22 somites with primers designed to detect both foxo3 genes (Fig. 3d). To functionally link the repressive function of Etv6 on foxo3 expression to Etv6-occupied DNA regions, we next tested the transcriptional activity of the sequences underlying the Etv6 peaks in foxo3.L and foxo3.S promoters in luciferase assays. When compared to the empty vector (control-luc), both sequences had the capacity to induce luciferase activity in wild type embryos, with the foxo3.S sequence exhibiting a stronger transcriptional activity (Supplementary Figure 3). Critically, luciferase activity was significantly enhanced in Etv6-deficient embryos, confirming that these sequences are involved in Etv6-mediated repression of foxo3. In conclusion, both foxo3.L and foxo3.S are directly repressed by Etv6 in the somites.
Next, we tested the hypothesis that Foxo3 may repress the expression of vegfa in the somites. First, we determined whether upregulation of foxo3 was sufficient to repress vegfa in the somites by injecting foxo3 mRNA into 2-cell stage embryos and testing vegfa expression by WISH at stage 22. Indeed, vegfa expression was dramatically downregulated in the somites of embryos injected with exogenous foxo3 mRNA (Fig. 3e) whereas expression in the hypochord was unaffected. Importantly, etv6 expression was unaffected in the somites of foxo3-overexpressed embryos (Supplementary Figure 4). This indicated that foxo3, like Etv6, specifically regulates the expression of vegfa in the somites, Fig. 2 Identification of Etv6 direct transcriptional targets in the somites. a Experimental design. RNA-seq was performed on wild type (WT) and Etv6deficient (etv6 MO-injected embryos) somite explants dissected from stage 22 embryos. Differentially expressed genes (DEGs) were identified by comparing the transcriptome of these tissues. b Spearman correlation analysis on triple biological RNA-seq replicates. c MA plot (magnitude of the difference versus amplitude of the signal) showing the fold-change in gene expression of all genes in WT versus Etv6-deficient somites (log 2 FC) compared to their expression levels (log 2 CPM). Black dots, non-significant change; red dots, differentially expressed genes (DEGs). The number of positively and negatively regulated DEGs is indicated. d WISH showing the expression of DEGs in stage 22 WT and Etv6-deficient embryos. Meox2 expression is upregulated in the somites (arrows) in Etv6-deficient embryos whereas expression of crim1, sox18 and vegfa is downregulated. Note that vegfa expression in the hypochord is unaffected (arrowheads). Embryos are shown in lateral view with anterior to the left and dorsal to the top. Numbers in top right corner indicate the number of embryos exhibiting the phenotype pictured (scale bars: 0.5 mm). e The intersection between DEGs and genes harbouring Etv6 ChIP-seq peaks in their TSS region reveals 540 putative direct target genes. In cancer cells, FOXO3 has been reported to repress VegfA expression by directly binding to the VegfA promoter region 25,26 . Using Jaspar 41 , we have identified a putative Foxo3 binding site (5'-GTAAACA-3') 42 in the promoter region of Xenopus vegfa distantly localized upstream the Etv6 ChIP peak (Supplementary Data 6). Foxo3 ChIP-qPCR experiments showed that Foxo3 does indeed bind this sequence and, as expected, occupancy is significantly increased in Etv6-deficient embryos or when foxo3 is overexpressed in embryos (Fig. 3g). Furthermore, in luciferase assays, a vegfa promoter fragment containing the Foxo3 binding motif can indeed repress transcriptional output and, critically, its repressive activity is lost when the Foxo3 binding motif is deleted (Supplementary Figure 7). Taken together, these results indicate that foxo3 directly represses the transcriptional activity of the vegfa promoter (Fig. 3h). gene is a candidate direct target of Etv6 in the somites (Fig. 2f). Interestingly, KLF4 has been shown in human retinal microvascular endothelial cells (HRMECs) 27 and human umbilical vein endothelial cells (HUVECs) 28 to activate VegfA expression through direct binding to its promoter. We, therefore, hypothesized that Etv6 could activate vegfa expression in the somites through Klf4.
We first validated binding of Etv6 in the promoter regions of the two Xenopus klf4 genes, klf4.L and klf4.S (Fig. 4a, b), and confirmed by WISH (Fig. 4c) and RT-qPCR (Fig. 4d) on stage 22 Etv6deficient embryos that expression of klf4 in the somites was dependent on Etv6. Additionally, we demonstrated by western blot that Klf4 protein levels were dramatically reduced in the somites of Etv6-deficient embryos (Fig. 4e, etv6 MO). Finally, to further confirm that the transcriptional activity of the klf4 promoter is regulated by Etv6 in the somites, the DNA fragment corresponding to the Etv6 peak in the klf4 promoter was functionally tested in in vivo luciferase assays in both wild-type and Etv6-deficient backgrounds. We observed a significant decrease of the transcriptional activity of this sequence in Etv6-deficient embryos (Supplementary Figure 3). Taken together, these data indicate that klf4 expression in the somites is directly activated by Etv6.
Analysis of expression of klf4 during early embryogenesis shows a remarkably transient expression in the somites where it is only detected at stage 22 (Supplementary Figure 8). Importantly, vegfa expression in the somites is absolutely dependent on Etv6 at this stage 11 . This is also the stage when vegfa secreted from the somites is required for the programming of definitive hemangioblasts in the lateral plate mesoderm 12 . In order to investigate the role of Klf4 in vegfa expression and HSC programming, we designed a MO which efficiently blocks the in vivo translation of Klf4 (Fig. 4e, Klf4 MO) as well as TALENs (transcription activator-like effector nucleases) which efficiently generate null mutations in the second exon of klf4 (Fig. 4f). Klf4-deficient embryos generated either by MO or TALEN injection showed a dramatic downregulation of vegfa expression in the somites at stage 22 (Fig. 4g). Thus, Klf4 is required for vegfa expression in the somites. Furthermore, definitive hemangioblast specification in the lateral plate mesoderm, as indicated by tal1 expression at stage 26-28, and hemogenic endothelium emergence in the dorsal aorta, as indicated by runx1 expression at stage 39, failed in Klf4deficient embryos (Supplementary Figure 9). Therefore, Klf4 is essential for vegfa expression in the somites and the programming of HSCs during embryogenesis, thus recapitulating the Etv6 functions.
Collectively, our studies show that Etv6 positively regulates the expression of vegfa in the somites through direct transcriptional activation of klf4, a transcriptional activator of vegfa (Fig. 4h).
Klf4 is required for Etv6 binding to the vegfa promoter. Analysis of Etv6 ChIP-seq peaks associated with the vegfa genes revealed binding of Etv6 in the vegfa.L (Fig. 5a), but not vegfa.S (Supplementary Figure 13), promoter region (−552 to −244 bp upstream of vegfa.L TSS). Analysis of the genomic structure of vegfa.S indicated that it is a gene undergoing degeneration and it is not transcribed during embryogenesis (Xenbase expression data after Sessions et al. 40 ). Thus, only one vegfa gene, vegfa.L, is functional in X. laevis (referred to as vegfa hereafter) and bound by Etv6 at its TSS. To test whether vegfa could be directly bound by Etv6, we performed Etv6 ChIP-qPCR on the vegfa promoter and confirmed the enrichment of Etv6 in this region (Fig. 5b, c). However, we did not find ETS-binding motif under this peak, suggesting that Etv6 cannot directly bind to the vegfa promoter. Instead, two binding motifs for the Klf/Sp family with high matrix scores were identified (Supplementary Data 6). This is consistent with our de novo motif analysis showing that Klf/Sp binding motifs are the most overrepresented sequences.
As Klf4 is required for vegfa expression in the somites, we wondered whether it was involved in Etv6 binding to the vegfa promoter. Therefore, we performed Klf4 ChIP-qPCR on the vegfa promoter using the same primers as for Etv6 ChIP-qPCR (Fig. 5b,  c). This confirmed that Klf4 binding is indeed enriched in the promoter of vegfa and co-localises with Etv6 binding (Fig. 5d), supporting the hypothesis that Klf4 is involved in Etv6 recruitment to the vegfa promoter. To further confirm this notion, Etv6 ChIP-qPCR was carried out on Klf4-deficient somites (Fig. 5e). Etv6 enrichment in the vegfa promoter was severely reduced in Klf4-deficient embryos, a reduction similar to that observed in Etv6-deficient embryos (Fig. 5e). Importantly, etv6 expression was unaffected in the somites of Klf4-depleted embryos (Supplementary Figure 10). Therefore, Klf4 is required for Etv6 recruitment to the vegfa promoter.
To functionally test the role of the two Klf4 binding motifs identified under the Etv6 peak in the vegfa promoter, we designed TALENs that efficiently deleted them (Fig. 5f, Supplementary  Figure 11). In the somites of stage 22 embryos with Klf4 binding motifs deleted, the enrichment of Klf4 and Etv6 in the vegfa promoter region was dramatically decreased (Fig. 5g). Importantly, vegfa expression in the somites was downregulated in embryos deleted for Klf4 binding motifs, whereas expression in the hypochord was not affected (Fig. 5h), strongly indicating that these Klf4 binding sites are specifically required for vegfa expression in the somites of stage 22 embryos.
We have shown that Klf4 is required for the recruitment of Etv6 to the vegfa promoter (Fig. 5e). To test whether this occurs through direct physical interaction between these TFs, we performed co-immunoprecipitation (co-IP) assays (Fig. 6). First, Fig. 3 Etv6 prevents Foxo3-mediated repression of vegfa in the somites. a IGV showing Etv6 peaks (red boxes) in the TSS region of foxo3.L and foxo3.S. Input is shown as control. For the peak tracks of individual replicates, see Supplementary Figure 13. b ChIP-qPCR analysis confirming that Etv6 binding is enriched in the foxo3 promoter region. The diagram above the histogram depicts the foxo3 locus and the location of the primers used to amplify the region of the Etv6 peak in the foxo3 TSS region (peak primers) and a negative control region in exon2 (exon2 primers). Primers were designed to target both foxo3. L and foxo3.S. IgG ChIP was used as negative control. Error bars represent SEM of three biological replicates. *P = 0.014, two-tailed Student's t-test. c WISH showing that foxo3 expression is upregulated in the somites (arrows) of Etv6-deficient (etv6 MO) embryos (scale bars: 0.5 mm). d RT-qPCR confirming that foxo3 is upregulated in stage 22 Etv6-deficient somites. Expression was normalized to odc1. Error bars represent SEM of three biological replicates. *P = 0.020, two-tailed Student's test. e WISH showing that overexpression of foxo3 (0.25 ng foxo3.L mRNA + 0.25 ng foxo3.S mRNA) blocks vegfa expression in the somites (arrows) whereas expression in the hypochord (arrowheads) is unaffected (scale bars: 0.5 mm). f WISH showing that blocking foxo3 translation with MOs rescues vegfa expression in the somites (arrows) of Etv6-deficient (etv6 MO) embryos. Arrowheads indicate expression in the hypochord (scale bars: 0.5 mm). g Foxo3 ChIP-qPCR on the vegfa promoter region. The assay was performed on somitic material isolated from wild-type, Etv6-deficient (etv6 MO) and foxo3 mRNA overexpressing embryos at stage 22. IgG ChIP was used as negative control. Error bars represent SEM of three biological replicates. **P = 0.0014, ***P = 0.00018, two-tailed Student's t-test. h Diagram illustrating that vegfa expression in the somites requires the repression of foxo3 by Etv6, i.e., Etv6 represses a repressor of vegfa. Images in c, e, f show stage 22 embryos in lateral view with anterior to the left and dorsal to the top. Numbers in top right corner indicate the number of embryos exhibiting the phenotype pictured NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09050-y ARTICLE Etv6 was overexpressed in developing embryos and its ability to co-immunoprecipitate endogenous Klf4 was assessed. Western blot analysis of Etv6 immunoprecipitates showed co-purification of Klf4 (Fig. 6a). To ask whether the interaction between Etv6 and Klf4 could be direct, mRNAs encoding these proteins were translated in vitro using the reticulocyte lysate system and the translated proteins were used directly in co-IP assays. In vitro translated Etv6 and Klf4 did not co-immunoprecipitate ( Fig. 6b), in contrast to two TFs known to directly interact, Lmo2 and Ldb1 43 , used as positive control (Fig. 6c).
The expression of vegfa in the somites and the establishment of definitive hematopoiesis are impaired in both Etv6-and Klf4deficient embryos. As expression of etv6 in Klf4-deficient embryos is unaffected (Supplementary Figure 10), we hypothesized that Klf4 could be the main regulator of these processes. To test this, we attempted to rescue the haematopoietic phenotype of   Figure 12), thus strongly supporting a model in which cooperation between Etv6 and Klf4 is required in vivo for the development of the HSC lineage. Taken together, our results show that klf4 is not only a direct target of Etv6, but also acts as its recruiting factor to the vegfa promoter to activate its expression.

Discussion
In this study, we show that expression of the important signaling molecule vegfa is under tight transcriptional control by the ETS TF Etv6 during the early stages of hematovascular development. Previously, we showed that Etv6 is essential for expression of vegfa in the somites 12 . We now demonstrate that it works through both repressive and activating transcriptional mechanisms. As discussed below, this reveals unexpected molecular mechanisms engaged by Etv6 and a complex GRN upstream of vegfa in vivo.
To establish how Etv6 controls vegfa expression in the somites, we identified Etv6 direct target genes by focusing on genes that are both abnormally expressed in Etv6-deficient somites and bound by Etv6 in their TSS. This stringent approach revealed numerous biologically relevant targets. Indeed, many of the TFs and chromatin modifiers identified amongst Etv6's 540 direct target genes have previously been directly or indirectly implicated in the regulation of VegfA in endothelial and cancer cells. In addition to Foxo3 and Klf4 (the two targets examined in this study), these are Ets2, shown to down-regulate VegfA expression in HUVECs treated with arsenic trioxide (ATO) 39 , the tumour suppressor Rbl2 (also known as RB2/p130), that inhibits tumour formation in mice by inhibiting VegfA-mediated angiogenesis 37 , the transcriptional repressor Mecp2, that represses VegfA expression by binding to methylated CpG islands in the VegfA promoter 35 , and the transcriptional regulator Creg1, that regulates human endothelial homoeostasis in vivo and promotes vasculogenesis in mouse ES cells through the activation of VegfA 36 . These findings validate our experimental design and the robustness of the list of Etv6's direct target genes.
ETV6 and its Drosophila orthologue, Yan, are well-established transcriptional repressors 14,15,24 . Surprisingly, in Xenopus somites, around two thirds (339/540, 63%) of Etv6 direct target genes are normally transcriptionally activated. Therefore, during the early stages of specification of the HSC lineage, Etv6 acts both as a transcriptional activator and a repressor. As detailed below, molecular and functional examination of Etv6 target genes revealed that Etv6's dual function is required for the regulation of vegfa expression.
Members of the ETS family of TFs are defined by a highly conserved ETS domain that recognizes a core sequence 5'-GGA (A/T)-3' motif within the context of a 9-to 10-bp DNA sequence 44 . As all ETS TFs bind to the same motif, additional mechanisms regulating the selection of specific transcriptional targets within biological contexts are required. It has been suggested that high-affinity ETS motifs found in the promoter of housekeeping genes can be bound by any ETS TF whereas loweraffinity ETS-binding sites found in tissue-specific promoters and only bound by a subset of ETS TFs are flanked by binding sites of other TFs 45,46 . Cooperative binding with other TFs in sequences with composite binding sites results in a higher affinity and stable binding to DNA, and in synergistic repression or activation of specific target genes 47,48 .
Here, de novo motif analysis of the 545 peaks associated with the promoters of genes normally activated or repressed by Etv6 showed (i) high prevalence of Klf/Sp binding sequences in both sets of targets; (ii) absence of ETS-binding motif enrichment in the peaks associated with genes repressed by Etv6; and (iii) ETS-binding motif enrichment in the peaks associated with a small fraction of genes activated by Etv6 (9.94%). This suggested that members of the Klf/Sp family are required for both the activating and repressive transcriptional activities of Etv6 and that Etv6 could be recruited to DNA through its interaction with other DNA-binding TFs. What determines Etv6 to act either as a transcriptional repressor or an activator remains to be fully explored.
Strikingly, klf4 was one of the target genes activated by Etv6 in somites. KLF4 is a multifunctional TF which is critical for pluripotency, tissue homoeostasis, stemness, maintenance of cancer stem cells and normal hematopoiesis 49 , and has previously been implicated in the expression of VegfA in human and mouse cells both positively and negatively. KLF4 binds the VegfA promoter and represses Vegfa expression during EMT in mouse mammary gland cells 50 , whereas, in HRMECs and HUVECs, it promotes angiogenesis by transcriptionally activating VegfA 27,28 . Here, we show that Klf4 not only binds to conserved Klf/Sp motifs but also recruits Etv6 to the vegfa promoter region to activate vegfa expression. Our in vivo and in vitro co-immunoprecipitation (co-IP) assays indicate that these two proteins form part of the same transcriptional complex controlling gene expression in the Xenopus somites but are likely to interact indirectly. In support of this, it has been shown, in overexpression experiments with tagged proteins, that the ETS protein ERG coimmunoprecipitates with KLF2 51 . Similarly, in situ proximity ligation assays have demonstrated that KLF4 and the phosphorylated form of the ETS TF, ELK1, physically interact in human coronary artery cells 52 . Moreover, analysis of EHF (an ETS TF structurally related to ETV6) ChIP-seq peaks indicates that a significant number of peaks contained no ETS-binding motifs and many of them harboured Klf4/5 binding motifs only 53 . Interestingly, EHF is involved in the maintenance of  Figure 13. b ChIP-qPCR analysis confirming that Etv6 is enriched in the klf4 promoter region. The diagram above the histogram depicts the klf4 locus and the location of the primers used to amplify the region of the Etv6 peak in the klf4 TSS region (peak primers) and a negative control region in intron2 (intron2 primers). Primers were designed to target both klf4.L and klf4.S. IgG ChIP was used as negative control. Error bars represent SEM of three biological replicates. **P = 0.0036, two-tailed Student's t-test. (c) WISH showing that klf4 expression in the somites (arrows) is downregulated in Etv6-deficient embryos (scale bars: 0.5 mm). d RT-qPCR confirming that klf4 is downregulated in stage 22 Etv6-deficient somites. Expression was normalized to odc1. Error bars represent SEM of three biological replicates. *P = 0.020, two-tailed Student's t-test. e Western blot showing that Klf4 protein is depleted in Etv6-deficient embryos and that klf4 MO blocks efficiently the translation of Klf4. Histone H3 was used as a loading control.  53 . In conclusion, the mechanistic relationship between Etv6 and Klf4 in regulating vegfa expression is compatible with a coherent feed-forward loop, a regulatory circuit where two pathways downstream of a TF control expression of the same gene: Etv6 activates klf4, Klf4 binds and activates vegfa and Etv6 binds vegfa 54,55 (Fig. 7). How Etv6 recruitment to vegfa promoter is functionally integrated to Klf4 activity in the control of vegfa expression remains to be fully explored as they cannot interact directly in our study.
In addition to activating an activator of vegfa, Etv6 achieves vegfa activation by repressing a transcriptional repressor of vegfa, Foxo3. In humans, FOXO3 harbours tumour suppressor activity as it represses processes such as VEGFA-driven tumour angiogenesis 56 . As an example, in breast cancer, FOXO3 activation correlates with VEGFA downregulation 25  human cells, FOXO3 represses VegfA expression by binding to a Forkhead response element in the VegfA promoter 25 . We have confirmed that Foxo3 can bind to the conserved Forkhead binding motif (5'-GTAAACA-3') 42 in the Xenopus vegfa promoter region, and negatively regulates its expression in the somites as (i) overexpression of this TF results in robust vegfa promoter occupancy and repression of vegfa expression, and (ii) downregulation of foxo3 in Etv6-deficient embryos rescues vegfa expression. We propose that this double negative gate 57 (Etv6 repressing Foxo3 that represses vegfa), together with the feedforward loop described above, unlocks the endothelial cell fate in the lateral plate mesoderm through tight control of vegfa expression (Fig. 7). Gene repression by ETV6 and Yan is mediated by the pointed (SAM) and linker domains. These domains can repress transcription independently through different mechanisms. The linker domain represses transcription by complexing with co-repressors which recruit histone deacetylases (HDACs) 14,24,58 . The pointed domain represses transcription by a mechanism that does not involve co-repressors recruiting HDACs 14 . Additionally, the pointed domain has the capacity to form homo-and hetero-typic oligomers and it has been proposed that polymerization of ETV6 could facilitate the spreading of transcriptional repression complexes along long stretches of DNA 16 . However, the formation of ETV6 polymers has not yet been demonstrated in vivo. We found a total of 9128 Etv6 ChIP-seq peaks in our analysis but only 102 of them were > 2 kb in size (Supplementary Data 1) and none were associated with DEGs. Therefore, although long stretches of DNA occupancy are present in the genome, Etv6 polymerization does not mediate gene repression in Xenopus somites.
In conclusion, our investigation of the functions of Foxo3 and Klf4 in vegfa expression in the somites unveils the foundations of a complex Etv6 GRN. It indicates that Etv6 has the capacity to regulate vegfa expression through the deployment of pathways involving both positive and negative regulators of gene expression and begins to unveil the complexity of vegfa regulation during hematovascular development. This work also further strengthens the parallel between oncogenic and developmental processes 13 . We propose that this Etv6-vegfa GRN is conserved through vertebrate evolution, from Xenopus to human, and that Etv6 uses, depending on tissue and context, particular components and/or branches of this network to ensure the correct levels of vegfa expression. Further investigation will establish whether Klf4, or additional members of the Klf/Sp family of TFs, are required for the regulation of some of the other Etv6 targets identified in our study and known to regulate Vegfa in the tumorigenic processes. If so, studying these molecular mechanisms could lead to new strategies for the treatment of pathological processes driven by the Etv6-vegfa GRN.   Prepare a 10 x stock, and adjust pH to 7.5) supplemented with 3% Ficoll. After 6 h, embryos were washed and cultured in 0.1×MMR at 17-19°C 59 until they were collected for analysis. Embryos were staged according to Nieuwkoop and Faber 60 . The somites of stage 22 embryos were manually dissected in 0.35 × MMR using forceps, these explants also contained the hypochord, neural tube, notochord, and some dorsal endoderm.
MO and mRNA for injection. MOs (morpholino antisense oligonucleotide) were obtained from GeneTools LLC (Corvallis, OR). Etv6 MO targeting both etv6.L and etv6.S was previously published 12 . A klf4 MO targeting both klf4.L and klf4.S was designed (Fig. 4e). As single MO blocking both foxo3.L and foxo3.S could not be designed (Supplementary Figure 5a), two MOs, one targeting foxo3.L (foxo3.L MO) and the other targeting foxo3.S (foxo3.S MO), were co-injected in a 1:1 ratio in order to generate Foxo3-deficient embryos (Supplementary Figure 5c). Every MO was titrated in order to determine the optimal concentration for embryo injection. MO sequences are as indicated in Supplementary Table 4. A two-step Golden Gate assembly method using the Golden Gate TALEN and TAL effector kit 2.0 (Addgene) was used to construct the TALEN plasmids containing the homodimer-type FokI nuclease domain 61 . TALEN sequences were designed using the online design tool, Mojo Hand (http://www.talendesign.org/). TALEN mRNA for injection was generated by linearizing the plasmids with NotI and transcribing with SP6 RNA polymerase using the mMESSAGE mMACHINE Kit (Ambion). TALEN mRNAs were injected at the 1-cell stage of development. To determine the mutagenesis caused by the TALENs, DNA was extracted from the somites of stage 22 TALEN-injected embryos and the genomic DNA fragment containing the TALEN binding sites was amplified and Sanger sequenced (Fig. 4f,  Supplementary Figure 11).
Antibodies and western blot analysis. Polyclonal anti X. laevis Etv6 antibodies were generated in collaboration with NovoPro Bioscience Inc. (Shanghai, China). In brief, the amino acid sequence of X. laevis Etv6 (GeneBank number EU760352) was analyzed using a proprietary algorithm, NovoFocus TM antigen design, to identify epitopes with good hydrophilicity, surface probability and high antigenic index. Based on this information, three epitopes with little or no conservation with other proteins, including ETS TFs, were selected (Supplementary Figure 1a). Short peptides corresponding to these epitopes were synthesized, conjugated to keyhole limpet hemocyanin (KLH) and used in immunizations at a purity ≥90%. Two rabbits were immunized with each peptide and six polyclonal antibodies (Etv6-1a, -1b, -2a, -2b, -3a, -3b) were affinity-purified after 5-6 rounds of immunization. Each antibody had an ELISA titre ≥1:50,000 against the peptide antigen. The specificity of the antibodies was verified by Western Blot analyses.
ChIP-seq. Etv6 ChIP-seq was performed using a standard procotol 62,63 with some modifications. Briefly, the somites of 40 stage 22 embryos were dissected, homogenized, using a pestle, in 1 ml ice-cold PBS containing protease inhibitors and 20 mM sodium butyrate, and centrifuged at 1300 rpm for 5 min at 4°C. Pellet was resuspended and cross-linked in 2 mM EGS (ethylene glycol bis(succinimidyl succinate)) for 20 min at RT, followed by fixation in 1% formaldehyde for 20 min at RT and gentle rocking before treating with glycine (final concentration: 125 mM) for 5 min. Samples were centrifuged (1,300 rpm, 5 min, 4°C) and pellets rinsed twice with cold PBS, lysed and chromatin sonicated to an average size of~200 bp using Covaris S220 (Duty factor: 10%; Cycle per burst: 200; Intensity: 5; Duration time: 300 s; Cycle: 3). Samples were centrifuged for 10 min at 14,000 × g and supernatant recovered. After saving 1% of each sample as input, Triton X-100 (final concentration: 1%) was added to the supernatant and incubated overnight at 4°C with slow shaking in the presence of anti-Etv6 antibody already bound to Dynabeads protein A. Beads were washed in RIPA buffer (4°C, 10 × 5 min) followed by 2 washes (5 min) in TEN buffer (10 mM Tris-HCl, 1 mM EDTA, 150 mM NaCl; 4°C ). Complexes were eluted in 1% SDS elution buffer (50 mM Tris-HCl, 1 mM EDTA and 1% SDS) by shaking the Dynabeads for 15 min at 65°C in a thermomixer set to 1000 rpm. Eluted samples were centrifuged at full speed for 1 min at RT. The supernatants and the reserved input samples were de-crosslinked by adding 1/20 volume of 5 M NaCl and incubating at 65°C overnight. After treating with RNase A (final concentration: 0.2 mg/ml, 1 h at 37°C) and Proteinase K (final concentration: 0.2 mg/ml, 3 h at 55°C), DNA was extracted using phenol:chloroform:isoamyl alcohol purification followed by ethanol precipitation. Immunoprecipitated DNA and input DNA were quantified using Qubit Fluorometer and sequencing libraries were constructed with 1 ng of DNA using NEBNext Ultra DNA Library Prep Kit (NEB#E7370) for Illumina sequencing. DNA was sequenced (∼15-38 million paired reads/library, 2 × 40 bp read length) using NextSeq 500. Sequencing was performed on three independent biological replicates with corresponding inputs as control.
Raw sequence reads were checked for base quality, trimmed and filtered to exclude adapters using Trimmomatic (Version 0.32) 64 , and then mapped to the X. laevis V9.1 40 with BWA version 0.7.12. Peaks were analyzed using MACS2 peak calling software 65 at default thresholds, with the input samples as control for each replicate. The package DiffBind (Bioconductor package) 29 was employed to identify consistent peaks between replicates using the function dba.peakset with the parameter minOverlap = 3. The consensus peak output of DiffBind was generated by adding and merging the three replicate peak sets. Peaks located in TSS regions were identified using Homer software and used to perform TFs binding site analysis with Homer software (findMotifsGenome.pl) 30 . Examples of the Etv6 peaks for individual replicates are shown in Supplementary Figure 13. For TF binding site prediction in the vegfa promoter region (−2 kb to +1 bp) and the Etv6 peak region (−552 to −244 bp upstream of vegfa.L TSS), DNA sequences were analyzed using Jaspar 41 . ChIP-seq data were visualized on the Integrative Genome Viewer (IGV).
Sequenced reads were checked for base qualities, trimmed and filtered to exclude adapters using Trimmomatic (Version 0.32) 64 and then mapped to the X. laevis V9.1 40 using STAR 66 with default parameters. Aligned read features were counted using Subread tool: featureCounts method (version 1.4.5-p1). Differential gene expression analysis was carried out using EdgR (Bioconductor Package) 67 . For this analysis, samples were CPM (counts per million)-normalized in order to filter out genes with low counts and, then, TPM (transcript per kilobase million)normalized for differential expression analysis. CPM normalization accounts for library size differences between samples and produces normalized values that can be compared on an absolute scale.
An ANOVA-like test analysis 68 was performed between WT and MO samples, using the generalized linear model 69 followed by Estimates Dispersion, fitted with negative binomial model and estimates Generalized linear model likelihood ratio. Genes with a false discovery rate (FDR) above 0.05 were filtered out. The functional annotation (gene ontology analysis) of Etv6-regulated transcriptome was performed using DAVID Bioinformatics Resources 70 .
qPCR. qPCR (quantitative PCR) was performed using Fast SYBR Green Master Mix (Thermo Fisher Scientific, Catalog number:4385612) and StepOne Real-Time PCR system. For RT-qPCR, cDNA was made from 1 µg of total RNA and relative expression levels of each gene were calculated and then normalized to odc1 gene. Details of primer sequences for RT-qPCR and ChIP-qPCR are indicated in Supplementary Tables 6 and 7, respectively.
Luciferase assay. The vegfa promoter region containing a conserved Foxo3 binding site as well as the klf4 and foxo3 promoter regions containing Etv6 peaks were amplified from X. laevis genomic DNA by PCR using primers described in Supplementary  Table 5. Two-step PCR-mediated mutagenesis was used to delete the Foxo3 binding site in the vegfa promoter fragment. All the produced sequences were subcloned into the pGL4.10 vector (Promega). 100 pg reporter plasmid DNA together with 5 pg Renilla luciferase construct pRL1-TK (Promega) and etv6 MO or foxo3 mRNA, were coinjected into 2-cell stage embryos. Somites were dissected at stage 22, and assayed for luciferase activity using the Dual Luciferase Reporter Assay Kit (Promega).
In vitro protein synthesis. Xenopus Etv6-HA, Klf4-HA, Ldb1-Myc, mouse Lmo2, and GFP proteins were synthesized from mRNAs (Supplementary Table 8) using the in vitro rabbit reticulocyte lysate translation system (Promega) according to the manufacturer's instructions.
Co-immunoprecipitation assay. Embryos were injected with etv6-HA mRNA at the 2-cell stage and grown to stage 22, when proteins were extracted by lysing 30 somites in 300 µl of lysis buffer (10 mM Tris-HCl pH 7.5, 100 mM NaCl, 0.5% Triton X-100, and 2.5 mM MgCl 2 , protease inhibitors added). After sonication for 20 min at 4°C using Bioruptor Plus (30 s on and 30 s off for each round of sonication), the extracts were centrifuged at 13,000 rpm at 4°C for 10 min. After saving 1% of each sample as input, the supernatants were incubated overnight at 4°C with slow shaking in the presence of anti-Etv6 or IgG antibody already bound to Dynabeads protein A. After washing three times in lysis buffer containing protease inhibitors, protein complexes were eluted in 1% SDS elution buffer by shaking the Dynabeads for 15 min at 65°C in a thermomixer set to 1000 rpm. Eluted proteins were separated by SDS-PAGE and transferred onto a 0.2 μm nitrocellulose membrane for western blotting. For in vitro co-IP, synthesized proteins were first incubated in lysis buffer for 1 h at 4°C. After saving 10% of each sample as input, the lysates were used for immunoprecipitation as described above.
WISH images and numbers shown in Figures are from one experiment and are representative of three biological replicates.
No statistical methods were used to predetermine the sample size.
No data or animal have been excluded from this study. Xenopus laevis embryos were allocated to experimental groups on the basis of different treatments and randomized within the given group. Investigators were not blinded to the group allocation during experiments and outcome assessment.
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All ChIP-seq and RNA-seq datasets have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE115225.