Krüppel-like factor 2 cooperates with the ETS family protein ERG to activate Flk1 expression during vascular development

Vascular endothelial growth factor (VEGF) and its high-affinity receptor,fetal liver kinase 1 (FLK1, also called VEGFR2 or KDR), are essential for formation of the vasculature. Mice lacking a single Vegf allele die at E10.5 owing to defects in the differentiation of endothelial cells and severe disorganization of the vasculature(Carmeliet et al., 1996; Ferrara et al., 1996). Similarly, homozygous Flk1-null embryos die at E8.5-9.5 with severe reduction in endothelial cell number and complete disruption of subsequent vascular development (Shalaby et al.,1995). Since FLK1 is amongst the earliest markers of the endothelial lineage, characterization of transcription mechanisms regulating Flk1 gene expression will help to advance our understanding of the regulatory pathways controlling vascular development.

Mouse transgenic studies have shown that endothelial-specific expression of Flk1 is regulated by an enhancer in the first intron, which contains binding elements for the transcription factors TAL1 (SCL) and members of the GATA and ETS families (Kappel et al.,1999). Mutation of the GATA or ETS motifs abolishes reporter expression in endothelial cells of transgenic mice, whereas mutation of the TAL1 site results in reduced expression levels(Kappel et al., 2000). Additional ETS motifs are located in the promoter region of the mouse Flk1 gene and these have been shown to function together with HIF-2α (EPAS1 - Mouse Genome Informatics) to regulate Flk1transcription (Elvert et al.,2003). The requirement for ETS binding sites in Flk1regulatory regions is consistent with the established function of ETS transcription factors in the regulation of vascular development(Dejana et al., 2007). Approximately 30 ETS factors are known in mammals and all members share a conserved DNA-binding domain that recognizes the core recognition sequence GGA(A/T) (Lelievre et al.,2001; Sharrocks,2001). ETS proteins frequently interact with partners to influence tissue-specific gene regulation (Lelievre et al., 2001; Sharrocks,2001; Oikawa and Yamada,2003), but very little is known about possible ETS partners in endothelial cells. Transgenic analysis has demonstrated the importance of ETS motifs for expression of several vascular genes in addition to Flk1,including Tie2 (Schlaeger et al.,1997) and VE-cadherin (cadherin 5)(Gory et al., 1999). Gain-of-function experiments have shown that ETS factors can upregulate endothelial gene expression in cultured cells(Birdsey et al., 2008; Hasegawa et al., 2004; Schwachtgen et al., 1997; Wakiya et al., 1996). Overexpression of the ETS factor ERG in Xenopus embryos is sufficient to activate ectopic transcription of the vascular marker X-msr(Baltzinger et al., 1999).

At least four ETS genes, Ets1, Erg, Fli1 and Er71(Etv2), are expressed in mouse embryonic endothelial cells(Lelievre et al., 2001; Lee et al., 2008). Owing to functional redundancy between family members, most loss-of-function studies of individual ETS factors have not revealed early vascular phenotypes. A striking exception is the knockout of the mouse Er71 gene, which shows greatly reduced angioblast cell numbers and severe disruption of vascular development(Lee et al., 2008). Zebrafish studies have shown that knockdown of four vascular ETS genes results in a near complete loss of endothelial cells, whereas single knockdowns of individual genes exhibit less severe phenotypes (Pham et al., 2007).

The Krüppel-like factor (KLF) family of transcription regulators is also involved in the regulation of vascular gene expression(Atkins and Jain, 2007). KLFs bind a consensus recognition sequence of CACCC(Bieker, 2001; Dang et al.,2001), and three of the 17 family members, KLF2, KLF4 and KLF6, are expressed in the mouse embryonic vasculature (Kuo et al., 1997; Yet et al.,1998; Kojima et al.,2000; Botella et al.,2002; Lee et al.,2006). KLF proteins can act as either transcriptional activators or repressors and domain mapping of KLF2 has identified transactivating and transrepression domains within the protein(Conkright et al., 2001). Numerous endothelial genes have KLF binding sites in their promoter regions and cell culture studies have shown that KLF2 activates the expression of vascular genes including thrombomodulin(Lin et al., 2005) and eNOS (Nos3) (Parmar et al., 2006; Dekker et al.,2005), but inhibits expression of other vascular genes including endothelin and adrenomedullin (Dekker et al., 2006). Mice lacking either KLF2 or KLF4 activity are not viable; however, early vascular development in both knockouts is normal(Kuo et al., 1997; Lee et al., 2006; Segre et al., 1999). Mice lacking KLF6 function exhibit vascular assembly defects in the yolk sac, but endothelial gene expression and development of blood vessels in the embryo itself are apparently normal (Matsumoto et al., 2006). It is interesting to note that embryonic stem (ES)cells lacking KLF6 activity show reduced levels of FLK1 protein after differentiation into embryoid bodies(Matsumoto et al., 2006),suggesting that KLF6 might function as an activator of Flk1expression. Like the ETS proteins, it is possible that redundant expression of different KLF family members can exert a rescuing function and, indeed,redundancy has been demonstrated in studies using ES cells in which any one of three specific KLFs (KLF2, KLF4 or KLF5) could substitute for the other two in maintaining ES cells in an undifferentiated state(Jiang et al., 2008). Of particular relevance to the study of Flk1 transcriptional regulation,both cell culture and microarray studies using adult endothelial cells have suggested that KLF2 functions as a repressor of Flk1 expression(Bhattacharya et al., 2005; Dekker et al., 2006).

Our investigation into the transcriptional regulation of Flk1strongly suggests a positive role for KLF proteins during embryonic vascular development. We show that KLF2 is sufficient to activate Flk1expression in the Xenopus embryo and that inhibition of KLF2 function results in the disruption of normal vascular development. Furthermore, we show that ETS and KLF proteins physically interact and synergistically activate embryonic expression of the Flk1 gene.

The insert from a full-length Xenopus laevis Klf2 clone (BC043732)was isolated using NotI and SalI and inserted into pBluescript SK(+). For in situ probe synthesis, Klf2-pBluescript was linearized with SalI and transcribed with T3 RNA polymerase(Megascript Kit, Ambion). Xenopus tropicalis VE-cadherin sequences were inserted into pGEM T-easy, linearized with SalI and transcribed with T7 RNA polymerase. Synthesis of Flk1, Erg and X-msr in situ hybridization probes has been described previously(Cleaver et al., 1997; Baltzinger et al., 1999; Devic et al., 1996). The KLF2 coding region was PCR amplified from BC043732 with Pfu polymerase, subcloned into pT7TS and the sequence verified. For synthesis of Klf2 mRNA, Klf2-pT7TS was linearized with XbaI and transcribed with T7 RNA polymerase (Message Machine Kit, Ambion). A dominant-repressor form of KLF2 (DR-KLF2) lacking the first 87 amino acids that make up the transactivation domain was generated by inverse PCR from Klf2-pT7TS template using Pfu polymerase and the sequence verified. The ERG coding sequence (AJ224126) was PCR amplified using Pfu polymerase, cloned into pT7TS,sequence verified, linearized with XbaI and mRNA synthesized using T7 RNA polymerase.

For embryonic expression experiments, mRNAs encoding ERG, KLF2, DR-KLF2 and EGFP were injected into a single vegetal blastomere of a four-cell stage embryo in 0.4× MMR containing 6% Ficoll and cultured thereafter in 0.2× MMR until assay (stages 34-36). Embryos were staged according to Nieuwkoop and Faber (Nieuwkoop and Faber,1994). Whole-mount in situ hybridization was carried out using digoxigenin-labeled probes and standard conditions(Harland, 1991). Preparation and microinjection of morpholino oligomers (MOs) was performed as described previously (Garriock et al.,2005). Klf2 antisense MO (Klf2 MO,5′-ATCCGAATCAGATTGTCAGCAAAAC-3′) was targeted to the 5′untranslated region (UTR) of Klf2 transcripts. Klf2 MO effectively blocked translation of Klf2 test transcripts containing the 5′ UTR plus a portion of the coding region sequences of Klf2 fused to the coding region of EGFP (see Fig. 3D,E). For in vivo experiments, 12.5, 25 or 50 ng of Klf2 MO or control antisense MO(5′-GGTAGTAATAGATGCTGTGATCTAT-3′) was microinjected into the mediolateral region of one cell of two-cell staged embryos and later assayed at stage 34 for Flk1 transcripts by whole-mount in situ hybridization. For measuring Flk1 transcript levels, Klf2 or control MO was injected at the one-cell stage.

A Flk1 genomic fragment comprising ∼2.5 kb of sequence from upstream of the transcriptional start site to within exon 2 was isolated. For transgenic analysis, the EGFP coding region was inserted into exon 1 of a Flk1 construction containing 2.5 kb of 5′ flanking sequences plus 1.5 kb of first intron sequences. Transgenic mutant constructs for the ETS and KLF sites were generated by inverse PCR using Pfu polymerase and the sequence verified. The ETS site 5′-GGAT-3′ was mutated to 5′-GTAT-3′ and the KLF site 5′-CACCCT-3′ was mutated to 5′-CGGTCG-3′. Xenopus transgenic embryos were generated as described (Kroll and Amaya,1996; Sparrow et al.,2000).

For luciferase assays, mouse bEnd.3 endothelial cells (ATCC# CRL-2299) were transfected with 0.5 μg of reporter plasmid using Fugene 6 (Roche). Flk1 reporter constructions (wild-type, ETS mutant and KLF mutant)were identical to transgenic constructions except the luciferase coding region was substituted in place of EGFP. For assay of the ETS/KLF module alone, three copies of the ETS/KLF sequence were inserted in tandem upstream of the minimal SV40 promoter in the luciferase reporter construction pGL3-Promoter vector(Promega). For both sets of experiments, cells were co-transfected with 0.1μg of plasmid containing a CMV promoter driving β-galactosidase for normalization of transfection efficiency. Transfected cells were incubated in a 12-well plate for 20 hours and luciferase and β-galactosidase activity was measured using a luminometer. Experimental transfections were performed in quadruplicate.

Nuclear protein extracts from mouse EOMA (ATCC# CRL-2586) and L cell fibroblast (CCL-1) cell lines were isolated according to standard procedures. HA-tagged mouse KLF2 was generated by transfection of a HA-KLF2 expression plasmid into COS-7 cells using Fugene 6, and protein extracts were isolated in 15 mM Tris-HCl (pH 7.5) containing 1% Triton X-100, 120 mM NaCl, 25 mM KCl and protease inhibitors. Radiolabeled probes included: WT used in Fig. 2A(5′-GTACTCTCCACCCTGGTGC-3′) and WT used in Fig. 2B(5′-TAAGACTCCACCCTGGCC-3′). Competition oligonucleotides included a mutated KLF site (Mut, 5′-GTACTTTAGATGCAGGTGC-3′) and a serum-response element (SRE, 5′-CTAGGTTTCAGGGTCCTGCCATAAAAG-3′). EMSA was performed using standard techniques.

Quantitative RT-PCR analysis was performed using a Rotor-gene 6000(Corbett). Ventral posterior tissue from 15 embryos (stage 34) injected with 150 pg of Klf2 mRNA, 25 pg of Erg mRNA and 500 pg GFP mRNA alone or in combination were processed for RNA isolation,the RNA converted into cDNA and used for real-time PCR analysis. The following primers were used (F, forward; R, reverse): Flk1 F,5′-ACATTCCTGTAGAGCCTGTGGT-3′ and R,5′-GGACTGGTAGTCGCTAGTTTGG-3′; Tie2 F,5′-TGAGAAGCCTCTGAACTGTG-3′ and R,5′-TTCAGCAGAACAGTCAATCC-3′; PECAM F,5′-ATGAGTGTGACGTGTTCTGTG-3′ and R,5′-GACAGCCATGCAATGTCTATG-3′; ornithine decarboxylase(ODC) was used to normalize samples(Bouwmeester et al., 1996).

Co-immunoprecipitation experiments were conducted as described(Meadows et al., 2008). Briefly, HA-tagged mouse KLF2 was generated by transfection into COS-7 cells and verified by protein blotting with detection using chemiluminescent solution (Supersignal, West Dura). ERG and EGFP proteins were radiolabeled with ³⁵S methionine by translation in the Wheat Germ Cell-Free Translation System (Promega). COS-7 cell extracts containing HA-KLF2 were mixed with in vitro translated products and immunoprecipitation was carried out using standard protocols and anti-HA antibody (Roche). Bound proteins were fractionated on a 10% SDS-PAGE gel and visualized by autoradiography. Human ERG was modified with a FLAG epitope, inserted into the expression vector pcDNA3.1+ (Promega) and co-transfected into COS-7 cells with HA-KLF2. Immunoprecipitation was carried out using anti-FLAG monoclonal antibody(Sigma) and proteins detected as described above.

In order to explore regulatory pathways that may be involved in development of the vascular endothelial lineage, we investigated the transcriptional control of the gene encoding FLK1. Alignment of sequences within the first intron of the Flk1 gene from different species identified a ∼120 bp segment that is highly conserved from human to frog. This conserved element is contained within a region of the mouse Flk1 intron that has previously been demonstrated to possess enhancer activity(Kappel et al., 1999; Kappel et al., 2000). As shown in Fig. 1A, a module within the larger conserved region contains consensus binding sites for ETS and KLF proteins. This ETS binding site in the mouse Flk1 gene is known to be essential for expression in transgenic animals(Kappel et al., 2000), but the KLF element has not previously been reported. The conservation of the putative KLF binding site in evolutionarily distant organisms suggests that it might also be important for regulation of Flk1 expression.

To determine the possible functional relevance of the KLF site, we used the electrophoretic mobility shift assay (EMSA) to determine whether nuclear proteins from endothelial cells associate with the KLF element. Nuclear extracts were prepared from the mouse EOMA endothelial cell line, which expresses the Flk1 gene (data not shown). A single EMSA band was observed when protein extracts were incubated with an oligonucleotide containing the KLF binding sequence found in the enhancer(Fig. 1B, lane 2). This shift could be competed with wild-type KLF sequences(Fig. 1B, lane 3), but not with a mutated KLF sequence or with other binding site sequences(Fig. 1B, lanes 4 and 5). Nuclear proteins from a fibroblast cell line (CCL-1) failed to form a complex with the KLF binding sequence (data not shown). These experiments indicate that a binding activity specific for the KLF site in the Flk1enhancer is present in endothelial cell nuclear extracts.

To determine whether KLF proteins interact with the conserved element, gel shift assays were performed using protein extracts from COS-7 cells expressing an epitope-tagged KLF2 protein, as this family member is known to be expressed in endothelial cells (Kuo et al.,1997). Extracts from non-transfected COS-7 cells yielded a shifted band (Fig. 1C, lane 1). Since KLF proteins are expressed in a wide range of cell types(Kaczynski et al., 2003), we cannot exclude the possibility that this band results from the binding of one or more KLF proteins to the probe sequence. However, an additional, prominent complex was detected when nuclear extracts expressing HA-tagged KLF2 were incubated with the probe (Fig. 1C, lane 2). This band could be supershifted using an antibody directed against the HA-tag on the KLF2 protein(Fig. 1C, lane 3). These experiments demonstrate that KLF2, a representative endothelially expressed member of the KLF protein family, specifically binds to the KLF motif located in the Flk1 enhancer.

To address the function of the ETS and KLF sites for transcriptional control of Flk1, we utilized the Xenopus transgenesis system(Kroll and Amaya, 1996; Sparrow et al., 2000). A construction containing 2.5 kb of Xenopus Flk1 5′ flanking sequences plus first intron sequences is sufficient to drive GFP expression throughout the blood vessels of transgenic embryos(Fig. 2A-C)(Warkman et al., 2004; Doherty et al., 2007). Expression is evident in major vessels including the posterior cardinal veins,intersomitic vessels and the vascular plexus, and is equivalent to endogenous Flk1 expression at a comparable stage (compare Fig. 2A with Fig. 3A). Reporter expression persists during subsequent development, when GFP fluorescence is observed throughout the smallest vessels of the trunk and tail(Fig. 2B,C)(Doherty et al., 2007).

To determine whether the conserved ETS and KLF binding sites were required for embryonic expression of the Flk1 gene, the individual sites were mutated in the context of the GFP reporter construction. In these experiments,the γ-crystallin promoter driving GFP expression in the eye was used as a marker for transgenesis. As expected from previous mouse studies(Kappel et al., 2000),mutation of the core ETS sequence reduced transgene expression to almost undetectable levels (n=28) (Fig. 2D,E). To assess the role of the KLF binding site, the consensus sequence in the Flk1-GFP reporter construction was altered from CACCC to CGGTC and transgenic embryos were generated. Transgenic embryos showed a dramatic reduction in GFP reporter expression compared with the wild-type construction (n=31) (compare Fig. 2F,G with 2B), but any detectable expression remained endothelial-specific. We conclude from these experiments that the KLF site is required for efficient expression of the Flk1gene during embryonic development and that the normal function of the site is to activate transcription.

Previous studies have suggested that KLF2 acts as an inhibitor of Flk1 expression (Bhattacharya et al., 2005; Dekker et al.,2006). Therefore, we wished to confirm that the KLF site in the enhancer indeed functioned as a positive regulatory element. First, the coding region of luciferase was substituted for GFP in the wild-type and ETS and KLF mutant Flk1 reporter constructions. These constructions were transfected into a mouse endothelial cell line (bEnd.3) and luciferase activity was measured. The results indicated that mutation of either the ETS or KLF elements results in a significant reduction in transcriptional activity compared with the wild-type construction(Fig. 2H). Second, in order to investigate the activity of the ETS and KLF binding sites independent of other potential regulatory sequences in the Flk1 gene, we inserted three tandem copies of the ETS/KLF module upstream of a minimal promoter driving a luciferase reporter. Reporter expression was then assayed in transfected bEnd.3 cells. As shown in Fig. 2I, mutation of either the ETS or the KLF binding site resulted in a reduction in reporter expression in endothelial cells. Taken together, the transgenesis and cell culture studies strongly suggest that the ETS and KLF sites in the Flk1 enhancer function as positive regulators of Flk1 transcription.

Several ETS family genes, including Ets1, Fli and Erg,are known to be expressed in Xenopus embryonic endothelial cells(Stiegler et al., 1990; Stiegler et al., 1993; Meyer et al., 1993; Meyer et al., 1995; Meyer et al., 1997; Baltzinger et al., 1999). For example, whole-mount in situ hybridization for Flk1 and Ergtranscripts (Fig. 3A,B) show very similar expression patterns, with transcripts detected in endothelial cells of the posterior cardinal veins, intersomitic vessels, aortic arches and the ventral vascular plexus. Previous reports have shown that Klf2 is expressed in the developing vasculature of mouse and zebrafish embryos(Kuo et al., 1997; Oates et al., 2001; Lee et al., 2006), but expression of Klf2 in Xenopus has not been reported. As shown in Fig. 3C, Klf2is expressed in all major developing blood vessels of the Xenopusembryo, in a pattern very similar to Flk1 and Erg. The expression of Erg and Klf2 in early endothelial structures indicates that these factors are present in the correct place at the correct time to regulate embryonic transcription of Flk1.

To address the function of KLF2 in Flk1 expression during development, we used an antisense morpholino oligomer (MO) to specifically inhibit the translation of KLF2 in the embryo. The MO effectively blocked translation of a test mRNA transcript in control experiments(Fig. 3D,E). The MO was injected into one cell of the two-cell embryo so that the uninjected side functioned as a time-matched control. Klf2 MO-treated Xenopus embryos were assayed for Flk1 expression at the tailbud stage by whole-mount in situ hybridization. MO treatment resulted in a major reduction of Flk1 expression in the vascular plexus, aortic arches, posterior cardinal vein and intersomitic vessels when compared with embryos injected with control MO (compare Fig. 3G,H with 3F). The effects on Flk1 expression were dose-dependent, with 64% of embryos showing a significant inhibition of transcript levels when 50 ng of Klf2 MO was injected (Table 1). Quantitation of the reduction by PCR indicated that Flk1 transcript levels were significantly reduced (Fig. 3L), but were not eliminated, consistent with the results of the transgenic studies and in situ detection of Flk1 expression(Fig. 2F,G and Fig. 3G,H, respectively).

To complement and extend the MO studies, we used a dominant-repressor form of the KLF2 protein (DR-KLF2) that lacks the transactivation domain but retains the repressor domain (Conkright et al., 2001). Since at least three KLF proteins are expressed in mouse embryonic endothelial cells, we reasoned that a dominant-repressor form of KLF2 might inhibit the function of possibly as yet uncharacterized KLF proteins that are also expressed in the frog vasculature. Xenopusembryos were injected with mRNA encoding DR-KLF2 and assayed at the tailbud stage for Flk1 expression. Similar to the results obtained with Klf2 MO, DR-KLF2-expressing embryos displayed a significant reduction in Flk1 transcripts throughout the embryonic vasculature (compare Fig. 3J,K with 3G,H). Again, the effect was dose-dependent, and at the highest levels of DR-KLF2 transcript injected (250 pg), 87% of embryos exhibited a reduction in Flk1 expression(Table 1). The observation that general inhibition of KLF function using DR-KLF2 was slightly more inhibitory than with Klf2 MO, raises the possibility that additional KLF proteins might also be involved in the regulation of Flk1 expression in the frog embryo. Finally, we examined whether knockdown of KLF2 activity inhibited formation of vascular structures in the embryo, as would be expected when FLK1 receptor function is reduced. In histological sections, vascular tubes were absent on the injected side of a Klf2 MO-treated embryo(Fig. 3M-M″). Taken together, these inhibition studies support a role for KLF proteins in the activation of Flk1 expression during embryonic development.

Previous studies have shown that expression of the ETS factor ERG in the Xenopus embryo is sufficient to induce ectopic expression of the endothelial marker X-msr(Baltzinger et al., 1999). This activation of an endothelial marker is a remarkable observation and has not been replicated with any other class of transcription factor. In order to determine whether expression of ERG is also sufficient to activate ectopic expression of Flk1, we injected Erg mRNA into the frog embryo. Injections were targeted to ventral blastomeres that contribute to posterior ventral regions of the embryo, which are largely free of endothelial cells. In all mRNA injection experiments, the test mRNA was co-injected with EGFP mRNA, which served as a tracer. Injection of Erg mRNA resulted in strong, ectopic expression of Flk1 in a dose-dependent manner (Fig. 4D-F, Table 2), whereas embryos injected with EGFP mRNA alone showed no ectopic expression(Fig. 4A-C, Table 2). Significantly,injection of Klf2 mRNA was also sufficient to activate ectopic expression of the Flk1 gene (Fig. 4G-I). At the highest doses of Klf2 examined (1 ng of mRNA), 79% of embryos showed the presence of ectopic Flk1 transcripts(Table 2). Ectopic expression was observed in both mesodermal and endodermal tissue but apparently not in the ectoderm. These experiments demonstrate that ERG and KLF2 are independently sufficient to activate ectopic Flk1 transcription in the frog embryo.

ETS proteins are known to interact with many other transcription factors to modulate expression of target genes(Sharrocks, 2001; Oikawa and Yamada, 2003). However, much less is known concerning possible regulatory partners for the KLF family proteins (Atkins and Jain,2007; Turner and Crossley,1999). Since our results (Fig. 4) show that ERG and KLF2 are independently capable of activating Flk1 transcription, and because the ETS and KLF sites are adjacent in the Flk1 enhancer region (Fig. 1A), we considered the possibility that ETS and KLF proteins might cooperate to activate transcription of Flk1. To test this, we first titrated the individual doses of Erg and Kfl2 mRNA to amounts that resulted in little to no ectopic activation of Flk1(Fig. 5B,C, Table 3). However, when these limiting amounts of Klf2 and Erg mRNA were co-expressed,robust ectopic expression of Flk1 was observed in a large proportion of embryos (Fig. 5D, Table 3). The large increase in expression observed when both Klf2 and Erg sequences were present suggests possible synergistic activation of Flk1transcription.

To better characterize the effect of Klf2 and Ergco-expression on Flk1 transcript levels, we carried out quantitative(q) RT-PCR analysis. Posterior ventral regions were dissected from Xenopus embryos expressing either GFP control, low doses of KLF2 or ERG alone, or a combination of KLF2 and ERG. These samples were then assayed for Flk1 transcripts by qRT-PCR(Fig. 5F). Samples expressing GFP mRNA showed very low levels of Flk1 transcripts,probably owing to the presence of a few endogenous endothelial cells in the dissected tissue. As expected, samples expressing KLF2 or ERG alone also showed extremely low levels of Flk1 transcripts, essentially equivalent to background levels. However, co-expression of KLF2 and ERG resulted in strong, synergistic activation of Flk1 expression to∼25-fold above background levels. Strong activation was also observed for the vascular-specific angiopoietin receptor gene Tie2, but not for the PECAM gene. These experiments demonstrate that KLF2 and ERG synergize to activate embryonic expression of Flk1. Although cooperation was also observed for Tie2 expression, failure to activate PECAM indicates that not all endothelial genes respond to KLF/ERG co-regulation.

The functional cooperation between KLF2 and ERG in the regulation of Flk1 raised the possibility that the two proteins might physically associate. To test this, we carried out co-immunoprecipitation (Co-IP)experiments using HA-tagged KLF2 protein (HA-KLF2) produced in COS-7 cells,and radiolabeled in vitro translated ERG protein(Fig. 6A, lanes 1 and 2). ERG co-precipitated with HA-KLF2 (Fig. 6A, lanes 4 and 5), whereas control GFP protein(Fig. 6A, lane 3) failed to associate with HA-KLF2 (Fig. 6A, lane 5). In reciprocal experiments, FLAG-tagged ERG and HA-tagged KLF2 proteins were co-expressed in COS-7 cells. Using anti-FLAG antibody, KLF2 co-precipitated with ERG protein, but no significant precipitation of GFP control protein was observed under the same conditions(Fig. 6B, lanes 4 and 5). These studies indicate that ERG and KLF2 are components of a single physical complex and very likely associate through direct protein-protein interactions.

Our studies suggest that KLF2 binds to a site within the Flk1enhancer and directly activates transcription of Flk1. It would be interesting to determine whether KLF2 also regulates the expression of other vascular marker genes. Using overexpression and MO knockdown approaches, we determined that KLF2 is required for normal expression of both the apelin receptor X-msr and the vascular cell-adhesion molecule VE-cadherin (Fig. 7, Table 4). These results raise the possibility that KLF2 is involved in the regulation of a considerable number of vascular genes in the embryo, in addition to Flk1. It remains to be determined, however, whether this activation is direct, through specific binding of KLF2 to regulatory elements in the X-msr and VE-cadherin genes, or whether activation is indirect, occurring via other transcription factors.

Examination of sequences within the first intron of the Flk1 gene revealed the presence of paired ETS and KLF binding sites that are conserved in all species examined, from frog to human(Fig. 1A). Previous studies have shown that the ETS site is essential for expression of a Flk1reporter in transgenic mice (Kappel et al., 2000), but the KLF site has not previously been recognized. Using a Flk1 transgenic reporter construction in Xenopus, we confirmed the importance of the ETS binding site(Fig. 2D,E) and also demonstrated that the KLF site is essential for efficient Flk1expression (Fig. 2F,G). These results demonstrate that both the conserved ETS and KLF motifs in the intronic enhancer play an important role in the developmental regulation of the Flk1 gene.

Numerous studies have demonstrated the importance of ETS proteins for developmental regulation of vascular endothelial gene expression(Dejana et al., 2007; Lelievre et al., 2001; Pham et al., 2007). However,less attention has been directed towards the role of KLF family proteins during early vascular development. At least three KLF genes (Klf2,Klf4 and Klf6) are expressed in the endothelium during mouse development (Kuo et al., 1997; Yet et al., 1998; Kojima et al., 2000; Botella et al., 2002; Lee et al., 2006). Similarly,a Klf2 gene (klf2a) is expressed in endothelial cells of the zebrafish embryo (Lee et al.,2006; Oates et al.,2001). Mouse embryos lacking KLF2 or KLF6 function die as embryos owing to defects in the ability to recruit smooth muscle cells to blood vessels and disorganization of the yolk sac vasculature, respectively(Kuo et al., 1997; Matsumoto et al., 2006), but early formation of the embryonic endothelium is normal in these animals. Knockout of Klf4, which is also endothelially expressed, results in perinatal death owing to loss of skin barrier function(Segre et al., 1999), but vascular development is again normal. The presence of multiple KLF proteins in the endothelium raises the possibility of functional redundancy. An argument for functional equivalence of at least some KLF proteins is supported by studies carried out using ES cells, which suggested that any one of three specific KLF proteins (KLF2, KLF4 or KLF5) could substitute for the other two in maintaining ES cells in an undifferentiated state(Jiang et al., 2008). In agreement with observations from other species, we determined that Klf2 is expressed in early endothelial cells of the Xenopusembryo (Fig. 3C). Inhibition of KLF2 function, using either a sequence-specific MO(Fig. 3F-H, Table 1) or more general interference with KLF function using a dominant-repressor construction(Fig. 3I-K, Table 1), results in significant downregulation of Flk1 expression in the Xenopusembryo. We note, however, that inhibition of KLF function never resulted in complete loss of Flk1 expression. This is consistent with the results of the transgenic studies and cell transfection experiments(Fig. 2F-I), in which mutation of the KLF site resulted in a reduction, but not elimination, of reporter gene expression. Taken together, these results indicate that KLF2 is required for normal regulation of Flk1 expression in the frog embryo. We cannot exclude the possibility that other KLF genes are expressed in frog embryonic endothelial cells, but the KLF2-specific knockdown studies indicate that any such proteins are insufficient to completely rescue the loss of KLF2 function.

Our studies consistently indicate that KLF proteins function as activators of Flk1 expression. First, we have shown that mutation of the KLF site within the first intron of the Flk1-EGFP reporter construction results in a major reduction of reporter expression in transgenic embryos(Fig. 2). Second, expression of KLF2 in the frog embryo has the remarkable ability to activate ectopic transcription of Flk1 (Fig. 4G-I). Third, inhibition of KLF2 activity using MOs or a dominant-repressor protein reduces embryonic expression of the endogenous Flk1 gene (Fig. 3, Table 1). These observations stand in apparent contradiction to previous studies showing that KLF2 may act as an inhibitor of Flk1 expression(Bhattacharya et al., 2005; Dekker et al., 2006). When KLF2 was introduced into human primary endothelial cells using an adenoviral vector, levels of both FLK1 mRNA and protein were reduced(Bhattacharya et al., 2005). Using a reporter construction containing the human FLK1 promoter, but not the conserved first intron sequences, KLF2 inhibitory activity was found to be mediated through SP1 binding sites close to the transcription start site. Another study suggesting KLF2 inhibition involved microarrays, where FLK1 was included in a population of genes downregulated following expression of KLF2 in HUVECs (Dekker et al., 2006). In these experiments, gene expression was analyzed after 7 days of forced KLF2 expression, raising the possibility that the effect on FLK1 transcript levels was indirect. Our results, however,are in broad agreement with studies of mouse KLF6 which indicated that FLK1 expression was reduced in embryoid bodies derived from Klf6^-/- ES cells(Matsumoto et al., 2006). The most likely explanation for these contrasting observations is that the studies were carried out over different time periods and using different sources of endothelial cells. KLF proteins exhibit both activator and repressor properties (Conkright et al.,2001; Yet et al.,1998) and different regulatory activities are likely to be revealed in distinct cellular contexts, probably through interactions with different partner proteins.

ETS family proteins frequently interact with partner proteins to regulate gene expression (Lelievre et al.,2001; Sharrocks,2001; Oikawa and Yamada,2003). These binding partners act to modulate ETS DNA-binding affinity and transcriptional activity (Li et al., 2000). However, despite many documented interacting proteins, no partner that functions specifically in endothelial cells has been characterized. Our studies suggest that endothelially expressed KLF and ETS family members might function together to regulate transcription of Flk1. First, both ERG and KLF2 are expressed in embryonic angioblasts during early vascular development (Fig. 3A-C) and are therefore in the right place at the right time to activate or maintain Flk1 transcription. Second, we have shown that ERG and KLF2 regulate Flk1 expression synergistically. When low amounts of Klf2 and Erg mRNA were expressed in the frog embryo, we never observed high levels of ectopic Flk1 expression(Fig. 5B,C, Table 3). However, when the same low doses were co-expressed, approximately one-third of embryos showed strong ectopic Flk1 expression(Fig. 5D, Table 3). Quantitation of equivalent experiments using real-time PCR showed that co-expression of ERG and KLF2 resulted in a ∼25-fold increase in Flk1 transcript levels over either ERG or KLF2 alone, strongly suggesting synergistic activation of Flk1 expression. Finally, we demonstrated that ERG and KLF2 are able to physically associate with each other(Fig. 6A,B). Our experiments do not exclude the possibility that other proteins are present in the complex,and do not demonstrate direct physical interaction between ERG and KLF2, but they do place the two proteins in the same complex. Additional studies will be required to determine whether the ability to form a physical complex is specific for ERG and KLF2 or whether this property is shared by additional family members.

Our studies have focused largely on transcriptional regulation of Flk1 because this represents one of the earliest and most important steps in the endothelial differentiation program. However, the observed cooperation between ERG and KLF2 might have implications for regulation of additional vascular endothelial genes. ETS binding sites are important for regulation of many endothelial genes(Dejana et al., 2007; Schlaeger et al., 1997; Gory et al., 1999) and recent experiments have shown that a large number of endothelial genes are potentially responsive to KLF2 regulation(Dekker et al., 2006). Although preliminary, we have obtained results showing that KLF2 is required for the expression of other embryonic vascular markers including X-msr and VE-cadherin and is capable of activating ectopic transcription of these genes (Fig. 7, Table 4). However, until functional studies are carried out, we do not know whether KLF2 is a direct transcriptional regulator of these genes. In summary, our studies suggest that cooperation between ETS and KLF proteins is important for activation of the vascular program in the embryo. Cooperation might also play a role in the maintenance of normal vascular function and in vascular disease.