ETV2/ER71 regulates hematovascular lineage generation and vascularization through an H3K9 demethylase, KDM4A

Summary

ETV2/ER71, an ETS (E-twenty six) transcription factor, is critical for hematopoiesis and vascular development. However, research about the molecular mechanisms behind ETV2-mediated gene transcription is limited. Herein, we demonstrate that ETV2 and KDM4A, an H3K9 demethylase, regulate hematopoietic and endothelial genes. Etv2^−/− mouse embryonic stem cells (mESCs), which fail to generate hematopoietic and endothelial cells, exhibit enhanced H3K9me3 levels in hematopoietic and endothelial genes. ETV2 interacts with KDM4A, and the ETV2-mediated transcriptional activation of hematopoietic and endothelial genes depends on KDM4A histone demethylase activity. The ETV2 and KDM4A complex binds to the transcription regulatory regions of genes directly regulated by ETV2. Mice lacking Kdm4a and Etv2 in endothelial cells (Cdh5Cre:Kdm:Etv2^f/f mice) display a more severe perfusion recovery and neovascularization defect, compared with Cdh5Cre:Kdm4a^f/f mice, Cdh5Cre:Etv2^f/f mice, and controls. Collectively, we demonstrate that ETV2 interacts with KDM4A, and that this interaction is critical for hematovascular lineage generation and vascular regeneration.

Graphical abstract

Highlights

• •Interaction of ETV2 and KDM4A decreases H3K9 trimethylation on hematovascular genes
• •ETV2 and KDM4A cooperatively regulates the expression of hematovascular genes
• •Mice lacking endothelial Etv2 and Kdm4a display severe angiogenic impairment

Molecular biology; Cell biology

Introduction

Generating vascular endothelial cells, hematopoietic cells, and cardiomyocytes in developing embryos is indispensable from embryonic development to adulthood. However, the molecular mechanisms responsible for early stage development of endothelial and hematopoietic cells remain unclear. Previously, we and other groups have demonstrated that mice deficient in Etv2, an ETS transcription factor, succumb to death in utero (embryonic day 9.5–10.5) and have vascular and blood development defects.^1,2,3,4 Furthermore, similar defective phenotypes are conserved among other vertebrate animals such as zebrafish and Xenopus.^5,6,7 In addition to its vital function in embryonic development, endothelial Etv2 is required for vascular regeneration and tissue repair following ischemic insults.⁸ Lentiviral Etv2 promotes perfusion recovery and angiogenesis when delivered to ischemic hindlimbs,⁸ this finding provides further evidence of its potent vascular function. Furthermore, ETV2 overexpression alone or with other factors is able to directly convert nonendothelial cells into cells with endothelial cell (EC) characteristics, phenotypically and functionally, both in vivo and in vitro.^9,10,11,12 Moreover, hematopoietic Etv2 increases the proliferation of hematopoietic stem and progenitor cells in bone marrow.¹³ These studies provide evidence for the important functions of ETV2 in vascular and hematopoietic systems.^14,15

ETV2 studies have mainly focused on its potent transcriptional activity achieved by direct DNA binding on the target genes critical for endothelial and hematopoietic cell development and functions such as VEGFR2/FLK1, CDH5, TAL1/SCL, GATA2, EGFL7, and LMO2.^{3,16,17,18,19} However, recent findings including ours have opened up a better understanding of the functions of ETV2 through protein-protein interactions.^20,21,22 De Val et al. have shown that ETV2 can interact with FOXC2 to cooperatively regulate the expression of endothelial and hematopoietic genes.²⁰ We have also demonstrated that the interaction of ETV2 and OVOL2 synergistically activates the expression of Flk1 and generates FLK1 (VEGFR2)⁺ cells, their downstream lineages, endothelial and blood cells from mouse embryonic stem cells (mESCs).²¹ The onset of endothelial and hematopoietic cell development is marked by the emergence of FLK1⁺ cells in differentiating mESCs and mouse embryos.^23,24 In addition, GATA2 forms a complex with ETV2 in regulating hematovascular lineages,²² suggesting that ETV2 has additional regulatory mechanisms by which ETV2 controls its downstream targets.

KDM4A-D (also known as JMJD2 or JHDM3) is a subfamily of the Jumonji C domain-containing proteins (JMJD) which can demethylase histone proteins, including di- and tri-methylated H3K9 and H3K36.^25,26,27 H3K9me3 (tri-methylated H3K9) generally is a hallmark of heterochromatin domains and is associated with transcriptional silencing.^28,29,30 Despite its role in diverse biological events,^31,32 investigations into KDM4 family function in vascular and hematopoietic lineage development are in the early stages.³³ KDM4 members are essential for early embryonic development demonstrated by gene knockout studies in mice.³⁴ Consistent with their preferential expression in cardiovascular lineages, the knockdown of Kdm4a and Kdm4c leads to a defective generation of ECs from mESCs, further supported by findings with morpholino knockdown in zebrafish.³³ These results imply the importance of KDM4 and thus the impact of H3K9 methylation status on cardiovascular lineage.

In this study, we show that ETV2 binds to an H3K9 demethylase, KDM4A, and that its interaction is important for the expression of hematovascular genes and the generation of these lineages.

Results

ETV2 interacts with KDM4A

Overexpression of ETV2 converts non-ECs into ECs in vivo and in vitro.^9,10,11,12 Furthermore, Etv2 can induce the de novo generation of endothelial and hematopoietic cells from differentiating mESCs under serum-free conditions.^3,17,21 These findings suggest that ETV2 can regulate the epigenetic status of its target genes, leading to the generation of both cell lineages. Since ETV2 does not have known functional domains related to epigenetic regulations, we hypothesized that ETV2 interacts with epigenetic modifiers to activate genes critical for endothelial and hematopoietic cells. Among many regulators, we were particularly interested in the KDM4 family owing to its importance in early embryogenesis.^33,34 Furthermore, EC generation from mESCs and zebrafish was impaired upon Kdm4a and Kdm4c knockdown.³³ To gain a better understanding of the roles of KDM4A and KDM4C in generating hematovascular lineages, we decided to knock out both Kdm4a and Kdm4c (Kdm4a/4c). Since mice deficient in Kdm4a/4c exhibit early embryonic lethality prior to the development of hematopoietic and endothelial lineages,³⁴ we used mESCs derived from Rosa26-CreERT2:Kdm4a/4c^f/f mice³⁴ (Figure S1). After confirming that Kdm4a/4c were efficiently deleted following 4-OH tamoxifen treatment, we differentiated the double knock out (DKO) Kdm4a/4c mESCs and analyzed embryoid bodies (EBs) (i.e., differentiated cell aggregates of mESCs) at day 4 (D4) of differentiation. As shown in Figure S2, our RNA sequencing (RNA-seq) results indicated that a lack of Kdm4a/4c leads to reduced expression of genes associated with mesoderm differentiation and hematovascular development and increased expression of genes related to neurodevelopment. Then, we confirmed the RNA-seq results with RT-qPCR analysis (Figure 1A). Loss of Kdm4a/4c in the differentiating mESCs resulted in reduced expression of early mesodermal lineage markers such as Flk1, Mesp1, and Tbx6, but increased expression of neuroectodermal lineage markers (NeuroD1, Fgf5, and Pax6). Brachyury/T gene expression, a pan mesoderm marker, remained unchanged. Endodermal markers showed altered expression with upregulation of Sox17 and Foxa2 and downregulation of Gata4 and α-Fp. Hnf4a was comparable regardless of Kdm4a/4c expression. Notably, the generation of FLK1⁺ cells was significantly impaired in the absence of Kdm4a/4c (Figures 1B and 1B′). These results suggest that KDM4A/4C could be critical for proper embryonic germ layer formation, potentially favoring the induction of mesodermal lineages and FLK1⁺ cells and their downstream hematovascular lineages.

Figure 1.

ETV2 binds with KDM4A

(A and B) Rosa26-CreERT2;Kdm4a/4c^f/f mESCs ±4′-OH-tamoxifen were differentiated and analyzed on day 4 for RT-qPCR (A, n = 3) and flow cytometry (B, B′, n = 3).

(B) Representative data from three independent experiments.

(B′) Quantification data of flow cytometry. Con, wild-type control; DKO, double knockout; FL1, empty channel of flow cytometry. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, n.s.: not significant.

(C) Overexpression of ETV2 increases genes downregulated in Kdm4a/4c DKO mESCs. Kdm4a/4c DKO mESCs overexpressing Etv2 were differentiated and subjected to RT-qPCR. n = 3. ∗∗∗p < 0.001.

(D) Enhanced H3K9me3 in endothelial and hematopoietic genes in the absence of ETV2. Etv2^−/− mESCs were differentiated, harvested at day 4 and cross-linked. Subsequently, the nuclear genomic DNAs were sonicated, and the fragmented genomic DNAs were immunoprecipitated with rabbit anti-mouse H3K9me3 antibody or rabbit IgG antibody. The immunoprecipitated DNA fragments were qPCR-amplified with primers corresponding to the promoter or enhancer regions of the indicated genes. DNA enrichment is shown as a percentage of input DNA. Data are mean ± SEM of three independent experiments with duplicate assays/experiments. ∗p < 0.05.

(E–G) (E) Interaction of ETV2 and KDM4A. HEK/293T cells transfected with the indicated constructs were subjected to immunoprecipitation (IP) and western blot analysis. Anti-TUBULIN antibody was used for the internal loading control.

(F) In vitro translated FLAG-ETV2 and HA-KDM4A proteins were subjected to immunoprecipitation with an anti-FLAG antibody and western blot analysis with the indicated antibodies.

(G) The differentiated wild type mESCs were subjected to immunoprecipitation with an anti-KDM4A antibody, followed by a WB blot analysis with an anti-ETV2 antibody.

(H) The cooperative function of ETV2-KDM4A interaction in inducing Flk1 promoter/enhancer (p/e) activity. Each construct expressing ETV2 or KDM4A (wt or H188A, a defective mutant of demethylation activity of KDM4A) was introduced into HEK/293T cells together with the Firefly luciferase reporter construct, pGL3-Flk1(p/e). The Firefly luciferase activity was normalized with Renilla luciferase activity. Data are mean ± SEM of three independent experiments with triplicate assays/experiments. ∗∗p < 0.01, ∗∗∗p < 0.001.

To investigate whether ETV2 has a functional relationship with the KDM4 protein family in the generation of hematopoietic and endothelial cells, we examined Etv2 expression and Etv2-regulated hematovascular genes with respect to Kdm4a/4c. As shown in Figures 1A and S2, Etv2 expression was reduced in differentiated Kdm4a/4c deficient mESCs (i.e., EBs). The Etv2-regulated hematovascular genes such as Flk1 and Cdh5 were downregulated as well. However, the reduced expression of the Etv2-regulated hematovascular genes in Kdm4a/4c DKO differentiated mESCs was rescued by Etv2 overexpression (Figure 1C). Furthermore, we sought to examine the status of H3K9me3 on Etv2-regulated hematovascular genes in Etv2-deficient mESCs.^16,17 Genomic DNA fragments isolated from D4 EBs of Etv2^−/− mESCs were immunoprecipitated with anti-H3K9me3 antibody (Ab), followed by qPCR with primers that amplify regulatory regions such as promoter and enhancer of the ETV2 target genes. The degree of H3K9me3 on the regulatory regions of Flk1, Cdh5, Scl/Tal1, Lmo2, and Gata2 in the absence of Etv2 was significantly increased, compared to wild-type (wt) control (Figure 1D). However, non-ETV2 target genes such as NeuroD and Gata4 had comparable H3K9me3 levels between wt control and Etv2^−/− mESCs. KDM4 family expression was unchanged in the absence of Etv2 (Figure S3). These results suggest a potential role of H3K9 methylation, and thus the role of the KDM4 family in regulating ETV2-mediated transcriptional activation of endothelial and hematopoietic genes.

Since ETV2 and KDM4 proteins function through binding on specific regulatory regions of genomic DNA, initially we examined whether ETV2 interacts with KDM4A. A co-immunoprecipitation (coIP)/western blot analysis revealed that ETV2 binds with KDM4A and KDM4C, and not with KDM4B (Figure 1E). Interestingly, KDM4A binds with ETV2 more strongly than KDM4C. Unlike the previous report’s finding,³⁵ we could not detect the interaction between ETV2 and KDM3A, another member of the H3K9 demethylase family. We were unable to detect the interaction between ETV2 and KDM6B, a H3K27 demethylase,³⁶ suggesting the specificity of the interaction between ETV2 and KDM4A/4C. Our coIP/western blot analysis with in vitro translated ETV2 and KDM4A proteins reveals their direct interaction (Figure 1F). Furthermore, we demonstrated the endogenous interaction between ETV2 and KDM4A in differentiating mESCs (Figure 1G). ETV2 can interact with KDM4A in prostate cancer cells,³⁷ which also supports our findings. However, the report failed to provide detailed molecular mechanisms. Next, to examine whether the interaction has functional significance in regulating endothelial and hematopoietic genes, we conducted luciferase-based promoter assays with a Flk1 promoter/enhancer(p/e)³⁸ luciferase plasmid as a reporter. As shown in Figure 1H, ETV2 significantly induced Flk1 p/e activity; however, KDM4A showed a marginal effect on it. Importantly, while co-transfection of ETV2 and KDM4A exhibited a cooperative effect, the introduction of KDM4A (H188A),²⁷ a catalytic (demethylase) inactive mutant of KDM4A which can function as a dominant negative form, inhibited the ETV2-induced luciferase activity. Furthermore, we found that KDM4C failed to activate Flk1p/e, and that ETV2 and KDM4C interaction did not show a cooperative effect (Figure S4). These results suggest that modulation of H3K9 methylation by KDM4A may contribute to ETV2-regulated FLK1⁺ cell and its downstream lineage generation. Considering the relatively weak KDM4C effects, we focused on KDM4A in the subsequent experiments.

ETV2-induced hematopoietic and endothelial generation is regulated by KDM4A

To determine the biological consequence of ETV2-KDM4A interaction, we employed two lines of mESCs that (1) overexpress FLAG-ETV2 [iFLAG-ETV2]²¹ and (2) overexpress both FLAG-ETV2 and HA-KDM4A (H188A) [iFLAG-ETV2-HA-KDM4A (H188A)] upon doxycycline (Dox) treatment (Figure S5). The mESCs were cultured under serum-free conditions, treated with ± Dox at D2 and analyzed at D4 of differentiation.^3,19,21,39 While FLK1⁺ cells were significantly induced from iFLAG-ETV2 mESCs following Dox treatment, this increase was significantly reduced in iFLAG-ETV2-HA-KDM4A (H188A) mESCs (43.1 ± 5.6% vs. 18.7 ± 2.0%, p = 0.015) (Figures 2A and 2A′). Consistently, the reduced expression of endothelial and hematopoietic ETV2 target genes was also evident in differentiated iFLAG-ETV2-HA-KDM4A (H188A) mESCs, compared to differentiated iFLAG-ETV2 mESCs (Figure 2B). Overall, NeuroD1 and Sox17 expression levels were comparable, regardless of the overexpression of ETV2 and/or KDM4A (H188A) (data not shown). Furthermore, we performed RNA-seq to further demonstrate that ETV2 can regulate endothelial and hematopoietic genes partly through KDM4A demethylase activity. As we previously reported,^16,17 overexpression of Etv2 induced the expression of genes critical for hematovascular development, thereby promoting cardiovascular development and neovascularization (Figure S6). However, the upregulated expression of the Etv2-regulated hematovascular genes was reduced upon the co-expression of ETV2 and KDM4A (H188A). These results suggest that the demethylase activity of KDM4A has an important function in ETV2-regulated hematopoietic and endothelial gene expression.

Figure 2.

The ETV2-mediated generation of FLK1⁺ cells, hematopoietic and endothelial lineages is regulated by the demethylase activity of KDM4A

(A and B) iFLAG-ETV2 and iFLAG-ETV2-HA-KDM4A (H188A) mESCs differentiated for 4 days were subjected to flow cytometry (A, A′, n = 3) and gene expression analysis (B, n = 3). Dox (1 μg/mL) was treated at day 2.

(A) Representative data from three independent experiments is shown. Numbers in the plots denote the percentages of FLK1⁺ cells. FL1: empty channel.

(A′) The quantification data of the flow cytometry.

(B) Expression of each gene was normalized against Gapdh, and the fold change of its expression level (+Dox/-Dox) was calculated. Data are mean ± SEM of three independent experiments with duplicate assays/experiments. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, n.s.: not significant.

(C and D) Flow cytometry analysis for FLK1 and hCD4/SCL (C, C′) and CDH5 (D, D′) in D6 EBs of indicated mESCs. Results are means ± SEM from three independent experiments. FL1: empty channel. ∗p < 0.05, ∗∗p < 0.01.

(E) Hematopoietic colony assay. iFLAG-ETV2 and iFLAG-ETV2-HA-KDM4A (H188A) mESCs differentiated under serum-free conditions were treated with ± Dox at D3 and subjected to a hematopoietic replating assay at day 6. Colonies were counted 4 days later. Results are means ± SEM from three independent experiments. ∗∗p < 0.01.

(F) EB sprouting assay. iFLAG-ETV2 and iFLAG-ETV2-HA-KDM4A (H188A) mESCs differentiated under serum-free conditions were treated with Dox at D3 and subjected to the sprouting assay in a 3D-collagen matrix. The mean sprouting area was measured using ImageJ software 8 days later. 3 EBs/group, n = 3, ∗p < 0.05.

The following experiments were performed to determine the function of the ETV2-KDM4A interaction in blood and endothelial lineage generation from mESCs. To this end, we subjected the mESCs to differentiation under serum free condition, treated the cells with Dox and analyzed the cells at D6 as previously reported.^3,39 Since SCL, a hematopoietic marker in early-stage embryos, is a transcription factor, a non-functional human CD4 consisting of extracellular and transmembrane domain was knocked into the endogenous regulatory region of one allele of Scl, allowing detection of SCL⁺ cells with human CD4 Ab using flow cytometry.⁴⁰ The generation of SCL/CD4⁺ hematopoietic cells⁴⁰ was increased upon ETV2 overexpression (iFLAG-ETV2 mESCs + Dox) but decreased when the expression of ETV2 and KDM4A (H188A) was induced (i.e., iFLAG-ETV2-HA-KDM4A (H188A) mESCs + Dox) (Figures 2C and 2C′). In agreement with the results, the number of hematopoietic colonies from iFLAG-ETV2-HA-KDM4A (H188A) mESCs was substantially decreased, compared to iFLAG-ETV2 mESCs (Figure 2E). To examine the role of the interaction in the generation of endothelial cells, we performed flow cytometry and found a decreased number of CDH5⁺ cells from iFLAG-ETV2-HA-KDM4A (H188A) mESCs, compared to iFLAG-ETV2 mESCs (Figures 2D and 2D′). Furthermore, vascular sprouting was significantly impaired in EBs derived from iFLAG-ETV2-HA-KDM4A (H188A) mESCs, compared to the EBs from iFLAG-ETV2 mESCs. (Figures 2F and S7). Individual EBs of iFLAG-ETV2 mESCs and iFLAG-ETV2-HA-KDM4A (H188A) mESCs at D3 of differentiation were mixed with collagen matrix and cultured for an additional 8 days.⁸ Taken together, these results suggest that KDM4A plays an important role in ETV2-regulated hematovascular lineage generation from mESCs.

H3K9 methylation of ETV2 target genes is regulated by KDM4A

To address if ETV2 can regulate the observed functions by modulating H3K9 methylation through KDM4A, we differentiated iETV2 mESCs, treated with ± Dox at D2, and performed a chromatin immunoprecipitation sequencing (ChIP-seq) with anti-H3K9me3 Ab or IgG at D4 to determine the degree of H3K9me3 on ETV2 target genes (Figure S8 and Table S2). H3K9me3 is considered a hallmark of heterochromatin regions associated with transcriptional silencing.²⁶ In peak calling, we tried to find consensus peaks to find peak regions showing differences. We identified 4,306 Dox-treated group (overexpression of ETV2) specific peaks, 3,247 no Dox control group specific peaks, and 19,212 common peaks (observable in both groups). We then focused on the peaks preferentially identified in the no Dox control group and compared them with genes known for hematovascular development and functions. As shown in Figure S8, we found H3K9me3 ChIP peaks specific to the no Dox control group on hematovascular genes including Cdh5, Tek, and Nrp2, compared to the Dox-treated group. Intriguingly, we failed to detect differential H3K9me3 peaks on Flk1, a well-known hematovascular and Etv2-regulated gene.³ Nonetheless, these results suggest a potential role of ETV2 in regulating hematovascular gene expression via H3K9 methylation. To validate the ChIP-seq results and further determine the H3K9me3 status in genes associated with ETV2, we conducted ChIP-PCR. Unlike the ChIP-seq analysis, our ChIP-PCR results clearly showed that the degree of H3K9me3 in the Flk1p/e region was significantly decreased in differentiating iFLAG-ETV2 mESCs following Dox treatment (overexpression of ETV2), compared to the no Dox group (control) (Figure 3). Importantly, the H3K9me3 level in the Flk1p/e region reduced by ETV2 was partially restored when ETV2 and KDM4A (H188A) were co-overexpressed. A similar H3K9me3 status was observed on the other ETV2 direct targets, including Cdh5, Scl, and Gata2. However, the degree of H3K9me3 in the promoter regions of NeuroD1 and Gata4, the non-ETV2 target genes, was similar regardless of the induction of ETV2 or ETV2-KDM4A (H188A). Altogether, these results suggest that ETV2 modulates H3K9 methylation of hematovascular genes through KDM4A demethylase activity.

Figure 3.

KDM4A regulates H3K9me3 methylation on the target genes of ETV2

iFLAG-ETV2 and iFLAG-ETV2-HA-KDM4A (H188A) mESCs were differentiated ± Dox and EBs at day 4 were cross-linked, sonicated, and the fragmented genomic DNAs were subsequently immunoprecipitated with rabbit anti-mouse H3K9me3 antibody or rabbit IgG antibody. The immunoprecipitated DNA fragments were qPCR-amplified with primers corresponding to the promoter regions of the indicated genes. DNA enrichment is shown as a percentage of input DNA. Data are mean ± SEM of three independent experiments with duplicate assays/experiments. ∗p < 0.05.

ETV2-KDM4A complex binds to the regulatory regions of hematovascular genes

To determine if the ETV2-KDM4A complex binds to the promoters or enhancers of ETV2’s target genes, we differentiated iFLAG-ETV2 mESCs and iFLAG-ETV2-HA-KDM4A mESCs, treated with Dox at D2 and did a ChIP assay at D4 with anti-FLAG Ab to immunoprecipitate FLAG-ETV2 or with anti-HA Ab to immunoprecipitate HA-KDM4A, followed by qPCR with primers that amplify the promoter/enhancer regions. As shown in Figure 4, enriched ChIP peaks in the Flk1p/e region were found when EBs of iFLAG-ETV2 mESCs were subjected to a ChIP assay with anti-FLAG Ab, not by anti-HA antibody. However, enhanced binding of FLAG-ETV2 and HA-KDM4A in the Flk1p/e region was evident when EBs of iFLAG-ETV2-HA-KDM4A mESCs were subjected to a ChIP assay with anti-FLAG Ab and anti-HA Ab, respectively, suggesting the occupancy of the ETV2 and KDM4A complex in the Flk1p/e region. We found consistent results when analyzing other ETV2 target genes, including Cdh5, Scl/Tal1, and Tek. These findings were further validated using Re-ChIP PCR experiments (Figures S9 and S10). These results suggest that the ETV2-KDM4A complex indeed binds to the promoter/enhancer regions of the genes whose H3K9 methylation is regulated by ETV2 and KDM4A in differentiating mESCs. While ETV2 failed to bind the non-ETV2 target genes, Gata4 and NeuroD, KDM4A was able to bind their regulatory regions, confirming the endothelial-specific functions of ETV2. Collectively, our findings suggest that ETV2 can regulate the status of H3K9 methylation of its target genes through interaction with KDM4A.

Figure 4.

ETV2 and KDM4A complex binds to hematopoietic and endothelial genes

The genomic DNAs prepared for ChIP-PCR from differentiated iFLAG-ETV2 and iFLAG-ETV2-HA-KDM4A ESCs were incubated with rabbit anti-FLAG (for ETV2), rabbit anti-HA (for KDM4A), and rabbit IgG antibody, followed by qPCR with primers corresponding to the promoter or enhancer regions of the indicated genes. The DNA enrichment is shown as a percentage of input DNA. Data are mean ± SEM of three independent experiments with duplicate assays/experiments. ∗p < 0.05.

ETV2 increases DNA demethylation of hematopoietic and endothelial genes

Because of the intimate relationship between DNA methylation and H3K9 trimethylation,⁴¹ we questioned whether DNA methylation in CpG islands which has implications for the status of gene expression^41,42 can be regulated by ETV2. To this end, iFLAG-ETV2 mESCs were differentiated for 4 days in the presence or absence of Dox, and the resulting cells were subjected to the enzyme-based DNA methylation analysis.⁴³ The degree of DNA methylation in CpG islands on Flk1 and other ETV2 targets, including Tal1/Scl, Lmo2, and Nrp2, was substantially reduced in Dox-treated cells, compared to untreated controls (Figure S11A). The DNA methylation on Notch1 was comparable regardless of the overexpression of ETV2, suggesting the DNA methylation-independent regulation of Notch1 by ETV2. Neither NeuroD1 nor Gata4 responded to ETV2. These results suggest that ETV2 can also regulate the status of DNA methylation on CpG islands on its target genes. Interestingly, KDM4A (H188A) overexpression partly abrogated the ETV2-induced demethylation on Flk1 CpG islands (Figure S11B).

ETV2 and KDM4A cooperatively promote blood perfusion recovery and enhance vessel formation

To further determine the biological consequence of the ETV2-KDM4A interaction in postnatal life, we generated Cdh5-Cre:Etv2^f/f, Cdh5-Cre:Kdm4a^f/f, and Cdh5-Cre:Etv2^f/f:Kdm4a^f/f double conditional knockout (CKO) mice, and found that these mice were born alive and fertile (data not shown). Then, we subjected the mice to a mouse model of hindlimb ischemia, instrumental in evaluating vascular regeneration ability.^44,45 As shown in Figure 5A, 5A′, Cdh5-Cre;Etv2^f/f and Cdh5-Cre:Kdm4a^f/f single CKO mice had significantly impaired blood perfusion recovery compared to control mice. Importantly, the lack of both endothelial Etv2 and Kdm4a (Cdh5-Cre:Etv2^f/f:Kdm4a^f/f DCKO mice) led to a further decreased blood perfusion recovery compared to the single CKO mice. When analyzed with FITC-conjugated ISOLECTINB4, Cdh5-Cre:Etv2^f/f:Kdm4a^f/f DCKO mice had the most defective vasculature among the mice examined as demonstrated by vascular density in the hindlimbs (Figures 5B and 5B′). A similar result was found when necrosis and fibrosis in the hindlimbs were examined. (Figures 5C and 5D). Taken together, these results suggest that ETV2 and KDM4A function together to regulate vascular regeneration, consistent with the findings from the mESC differentiation approach.

Figure 5.

ETV2 and KDM4A cooperatively promote neovascularization in a mouse model of hindlimb ischemia

Mice deficient in endothelial Kdm4a, Etv2, Etv2 and Kdm4a, wild type were subjected to a mouse model of hindlimb ischemia.

(A) Representative images showing physiological status and blood perfusion measured by laser speckle contrast analyzer (LASCA) on day 28 post-injury. n = 5/group.

(A′) The blood perfusion ratio of ischemic limbs was measured by LASCA on days 0, 7, 21, and 28 after injury. n = 5/group, ∗∗∗p < 0.001 (littermate wt control vs. Kdm4a^f/fCKO (Cdh5-Cre;Kdm4a^f/f) and Etv2^f/fCKO (Cdh5-Cre;Etv2^f/f)), ^###p < 0.001 (littermate wt control vs. Kdm4a^f/f;Etv2^f/fCKO (Cdh5-Cre;Kdm4a^f/f;Etv2^f/f)), ^$$$p < 0.001 (Kdm4a^f/f;Etv2^f/f CKO (Cdh5-Cre;Kdm4a^f/f;Etv2^f/f) vs. Kdm4a^f/f CKO (Cdh5-Cre;Kdm4a^f/f) and Etv2^f/f CKO (Cdh5-Cre;Etv2^f/f)). Error bars indicate ± SDs.

(B) Representative images of ISOLECTIN B4⁺ vessels in the ischemic hindlimbs. Adductor muscles from the mice were harvested 28 days after injury and subjected to immunohistochemistry with FITC-conjugated ISOLECTIN B4 to detect vascular structure (green). DAPI (blue) was used for nuclear staining. Scale bars: 100 μm.

(B′) Vessel density as represented by the proportion of ISOLECTIN B4 stained area over the total area in the ischemic hindlimbs on day 28 post-injury. (n = 18 images; 2 images/section × 3 sections/mouse × 3 mice/group). ∗∗∗p < 0.001. Error bars indicate ± SDs.

(C and D) (C) Representative images of the ischemic hindlimb showing muscle necrotic lesions (H&E stain; white arrowhead) and (D) fibrotic lesions stained in blue (Masson’s trichrome stain; black arrowhead). Scale bars: 100 μm. Insets in the photographs in c and d show a magnified image of the area marked with an asterisk.

Discussion

Since the discovery of ETV2 as a master regulator of hematopoietic and endothelial cell development, studies have extensively investigated detailed regulatory mechanisms used by ETV2 in this process. The early wave of investigations focused mainly on the DNA binding ability of ETV2 through the ETS domain and identified a myriad of direct downstream targets of ETV2.^1,3,17 Subsequently, studies including our own have revealed that ETV2 can interact with other nuclear proteins, such as OVOL2 and GATA2, and that such interactions cooperatively promote the transcriptional activity of ETV2.^21,22 In the case of ETV2-OVOL2 interaction, one of the mechanisms ascribed to this cooperative effect is the enhanced protein stability of ETV2 upon overexpression of OVOL2.²¹ While these studies provide valuable insights about ETV2, how ETV2 governs the expression of hematopoietic and endothelial genes at the level of epigenetics remains largely unexplored. The significance of our study of ETV2 in the generation of hematopoietic and endothelial cell lineages is several-fold. First, we found augmented H3K9me3 on hematovascular genes in the absence of Etv2. Second, our study demonstrated the interaction of ETV2 and KDM4A, and its functional significance in regulating hematopoietic and endothelial cell generation from differentiating mESCs. Third, we demonstrated the regulation of H3K9 trimethylation on ETV2 target genes in a KDM4A demethylase-dependent manner. Fourth, we found that the ETV2-KDM4A complex occupies the regulatory regions of hematovascular genes. Finally, we showed that ETV2 in conjunction with KDM4A regulates neovascularization in response to ischemic injury.

Numerous studies reveal the biological consequences of histone modifications, such as methylation and acetylation. The importance of H3K9 methylation in mouse embryogenesis comes from studies showing that mouse embryos deficient in Kdm4a/4c die before establishing embryonic germ layers (E6.5).³⁴ In this study, taking advantage of in vitro mESC differentiation, we further revealed that Kdm4a/4c is required to properly form the mesoderm and its derivatives including hematopoietic and endothelial cell lineages as well as non-hematovascular mesoderm lineages such as bones, and smooth muscle cells. By performing a series of biochemical assays, we also uncovered a mechanism by which KDM4A/4C and thus H3K9 methylation regulates blood and endothelial cell generation. Our results indicated that ETV2 can interact with KDM4A in differentiating mESCs. Interestingly, while ETV2 showed a strong ability to bind to KDM4A and a weak ability to bind to KDM4C, it could not bind to KDM4B. Considering a high degree of homology among the members, no plausible explanation of the different binding specificity exists. Still, the binding results are compatible with findings from knockout studies that KDM4A and KDM4C have functional redundancy in early embryogenesis, unlike KDM4B.³⁴ A previous study reported that KDM4B and KDM6B are associated with osteogenic differentiation in mesenchymal stem cells.⁴⁶ We did not detect interactions between ETV2 and KDM6B or KDM3A, further suggesting the specificity of ETV2-KDM4A interaction. A recent study showed that ETV2 can interact with KDM4A to regulate matrix metalloproteinases 1 and 7 in prostate cancer cells,³⁷ supporting our findings. However, the report did not provide detailed molecular mechanisms. Since Kdm4a/4c deficiency leads to decreased expression of genes associated with non-hematovascular mesoderm lineages such as bone and smooth muscle cells, studying the potential role of KDM4A/4C in the generation of these lineages would be an interesting topic.

A previous study demonstrated the highly enriched ChIP peaks of KDM4A/4C in H3K4me3-positive transcription start regions.^34,47 Wu et al. reported that KDM4A and KDM4C play an important role in endothelial gene regulation.³³ Although the study suggests the transcriptional regulatory roles of KDM4A and KDM4C in regulating EC genes, it is unclear how these proteins regulate the expression of the target genes. In this study, we provide a molecular mechanism via which KDM4A and possibly KDM4C act as interacting proteins of ETV2, activating ETV2 target genes. We showed that ETV2-induced Flk1 promoter/enhancer activation was impaired by the demethylase mutant form of KDM4A. Furthermore, while overexpression of ETV2 significantly reduced the level of H3K9me3 on the promoter and enhancer region of Flk1 and other key hematopoietic and endothelial genes, the reduced level of H3K9me3 was restored upon overexpression of demethylase defective KDM4A. Consistent with the H3K9me3 alterations in response to the ETV2 + KDM4A mutant, the induced expression of ETV2-controlled genes was decreased when the ETV2 and KDM4A mutant were co-overexpressed. Co-overexpression of the ETV2 and KDM4A mutant led to defective hematopoietic and endothelial cell generation from mESCs. Therefore, we conclude that ETV2 induces the expression of genes essential for hematopoietic and endothelial cells by regulating H3K9 methylation through interaction with KDM4A. Therefore, our study warrants genome-wide profiling studies of histone modification in response to ETV2.

DNA methylation is one of the major epigenetic modifications intimately associated with transcriptional repression.^48,49 Guo et al. reported that DNA methylation is closely correlated with chromatin accessibility during fetal germ cell development in humans and mice.⁴² In addition, deficiency of DNA methyltransferases in FLK1⁺ mesoderm cells derived from mESCs leads to the conversion of mesoderm cells to endoderm by the drug-inducible Gata4 system, which induces endoderm differentiation.⁵⁰ Furthermore, CpG methylation on or near the transcription start site of a gene is functionally related to the inactive status of gene expression.⁴⁸ In this study, we examined whether ETV2 can induce hypomethylation on the CpG islands present in the target genes of ETV2. Although we investigated limited numbers of key downstream targets, overexpression of ETV2 reduces DNA methylation, which agrees with the results of the transcriptome analysis.^16,51,52 Intriguingly, a recent study showed that ETV2 regulates the expression of Robo4 through the interaction of TET1, a DNA demethylase, supporting the role of ETV2-regulated DNA demethylation.⁵¹ Whether or not the interaction of ETV2 and TET1 is conserved in regulating the expression of ETV2 target genes during early embryogenesis remains to be determined. Both the CpG methylation near transcription start sites and the patterns of DNA methylation throughout the genome are also important markers to evaluate gene expression such as differentially methylated regions (DMRs).⁵³ Therefore, our results provide an in-depth insight into the function of ETV2 in altering DNA methylation.

Several studies have suggested the therapeutic potential of ETV2 for blood and vascular-related diseases. Previously, we reported that a single delivery of lentiviral Etv2 to ischemic hindlimbs can promote limb salvage and enhance perfusion recovery from ischemic damage.⁸ Furthermore, studies including ours have demonstrated that ETV2 alone or with other transcription factors such as GATA2, ERG, and FLI1, can directly generate hematopoietic and endothelial cells from pluripotent stem cells or non-ECs.^9,10,11,54 Expansion and regeneration of hematopoietic stem cells require the axis of the ETV2 and c-KIT, as demonstrated by bone marrow (BM) transplantation and hematopoietic injury.¹³ These findings suggest that ETV2 is critical for blood and endothelial cell development, but also plays an important role in postnatal hematopoiesis and angiogenesis. However, the detailed molecular mechanism by which ETV2 controls those processes remains unexplored. Extrapolating from our current study, it is tempting to speculate that KDM4A strengthens ETV2’s ability to activate its downstream genes through demethylation of H3K9me3 in the promoter or enhancer regions. Indeed, EB sprouting and hematopoietic colony formation induced by ETV2 were impaired when the demethylase activity of KDM4A was defective. KDM4A morpholino revealed impaired vasculature during zebrafish embryo development.³³ KDM4 activity is also critical for bone marrow hematopoiesis through regulating hematopoietic stem cell maintenance, demonstrated by CKO of KDM4A, 4B, and 4C.⁵⁵ Thus, investigating the functional significance of the ETV2-KDM4A axis in hematopoietic and vascular regeneration under pathophysiological conditions would be an interesting area for future studies, particularly as our study revealed the important functions of ETV2-KDM4A in the regulation of ischemia-induced angiogenesis.

In conclusion, our findings demonstrate that ETV2 regulates H3K9 methylation in promoter or enhancer regions of its downstream genes through KDM4A, activating their transcription and promoting the generation of hematopoietic and endothelial lineages. Thus, investigation of histone methylation and acetylation on H3K4, H3K27, and H3K9 (acetylation), as well as chromatin accessibility with respect to ETV2-KDM4A interaction would be warranted.

Limitations of the study

The mouse embryo is a useful model to investigate early events in development. However, there are some limitations such as fast kinetics and limited numbers of target cells, which make in vivo study challenging. Furthermore, early embryonic lethality of Kdm4a/4c^−/− mice and Etv2^−/− mice prompted us to find an alternative method to examine the questions that we raised. The mESC differentiation system has been instrumental as a surrogate system. Indeed, developmental events are well conserved between these two systems especially during early hematopoietic and endothelial cell generation. Therefore, we employed in vitro mESC differentiation to test our hypothesis and demonstrated the mechanism by which ETV2 regulates hematovascular lineage generation. In this study, we found KDM4A as an ETV2 interacting protein through a candidate-based approach and revealed its functional significance in regulating hematovascular gene regulation. Therefore, more comprehensive profiling ETV2 binding proteins (e.g., LC-MS/MS or yeast two hybrid screen) with special emphasis on epigenetic and chromatin remodeling proteins would be warranted. We performed a ChIP-seq for H3K9me3, but we do not show the comprehensive genome-wide information on H3K9 methylation including H3K9me1 and H3K9me2, as well as other histone marks. While it is informative, the comprehensive epigenetic profiling would require lengthy procedures (e.g., cell preparation time [mESC differentiation followed by viable cell sorting with flow cytometry], sequencing and bioinformatics analysis) and numerous cell numbers (for the H3K9me3 ChIP-seq, we prepared ∼180 plates of 100 mm dishes [triplicates × 30 plates/IP × 2 groups (control vs. ETV2)]). Furthermore, it is beyond the scope of the current study since the major focus of this manuscript is on the ETV2-regulated gene transcription. Our study reveals the functional significance of the interaction between ETV2 and KDM4A in postnatal angiogenesis. However, we are not able to demonstrate the H3K9me3 status of ETV2-regulated genes in this setting due to the limited number of ECs from the ischemic hindlimbs. Nonetheless, this is an important study to demonstrate the functional consequence of the interaction between ETV2 and KDMA4A in ischemia-induced angiogenesis. Thus, the potential therapeutic effect of the interaction in mediating neovascularization would be warranted.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Changwon Park (changwon.park@lsuhs.edu).

Materials availability

Material generated in this study will be made available upon reasonable request.

Data and code availability

• •RNA-seq data and ChIP-seq data have been deposited at NCBI (PRJNA1051681, https://www.ncbi.nlm.nih.gov/search/all/?term=PRJNA1051681%20). The link is also listed in the key resourcestable.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This work was supported by the Basic Science Research Program through the National Research Foundation of Korea funded from the Ministry of Science, ICT and future Planning 2016R1A2B4011905 (to H.S.); NIH R01HL133053 (to B.-Y.K); the Danish Cancer Society R167-A10877 (to K.H.) through a center grant from the NNF to the NNF Center for Stem Cell Biology NNF17CC0027852 (to K.H.) through the Memorial Sloan Kettering Cancer Center Support Grant NIH P30 CA008748 (to K.H.); post-doctoral fellowship, Center for Cardiovascular Disease and Science at LSU health science (to M.S.K.); Children’s Heart Research and Outcomes Center and Children’s Healthcare of Atlanta 00060337 (to C.P.); the Children's Miracle Network, 660085-1116 (to C.P.); and NIH R01HL119291 (to C.P.). The graphical abstract was created in BioRender.com.

Author contributions

M.S.K., R.L., and D.H.L. conceptualized and designed this study, collected, assembled and analyzed the data, and wrote the manuscript with the help of H.S., T.H., J.K.K., B.-Y.K., D.S., and Y.K. performed RNA-seq and ChIP-seq analysis, helped interpretation of data and manuscript writing. K.A. and H.L. provided study material, helped in data analysis, interpretation and in manuscript writing. C.P. conceptualized and designed this study, collected, assembled, analyzed and interpreted data, wrote the manuscript and provided financial support.

Declaration of interests

C.P. has a United States Patent (No. 10,023,842 B2. Title: Endothelial and endothelial-like cells produced from fibroblasts and uses related thereto), but the patent has nothing to do with the current manuscript.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
APC streptavidin-conjugated secondary antibody	Biolegend	Cat# 405207; RRID:AB_2869478
Biotin-conjugated anti-human CD4 antibody	Biolegend	Cat# 300504; RRID:AB_314072
Biotin-conjugated anti-mouse CDH5 antibody	Biolegend	Cat# 138008; RRID:AB_10641138
FITC-conjugated ISOLECTIN B4	Vector Laboratories	Cat# FL-1201; RRID:AB_2314663
Goat polyclonal anti-α-tubulin antibody	Origene	Cat# AB0046-200; RRID: N/A
Mouse monoclonal anti-FLAG antibody	Sigma Aldrich	Cat# F1804; RRID:AB_262044
Mouse monoclonal anti-HA antibody	Abcam	Cat# Ab1424; RRID:AB_301017
Mouse IgG antibody	Diagenode	Cat# C15400001; RRID:AB_2722553
PE-conjugated Rat anti-mouse FLK1/VEGFR2 antibody	Biolegend	Cat# 136403; RRID:AB_1967093
Rabbit monoclonal anti-FLAG antibody	Cell Signaling	Cat# 14793; RRID:AB_2572291
Rabbit polyclonal anti-H3K9me3 antibody	Epigentek	Cat# A-4036; RRID:AB_2920606
Rabbit polyclonal anti-HA antibody	Abcam	Cat# Ab9110; RRID:AB_307019
Rabbit IgG antibody	Diagenode	Cat# C15410206; RRID:AB_2722554
Rabbit recombinant monoclonal anti-human KDM4A antibody	Abcam	Cat# Ab191433; RRID: N/A
Rabbit polyclonal anti-human KDM4A antibody	Bethyl Laboratories	Cat# A300-861A; RRID: N/A
Rabbit recombinant monoclonal anti-ETV2 antibody	Abcam	Cat# ab181847; RRID: N/A
Rat monoclonal anti-mouse ETV2 antibody	R&D Systems	Cat# MAB7740; RRID:AB_2076679
Bacterial and virus strains
Stbl3 (chemically competent E. coli)	Thermofisher Scientific	Cat# C737303
Lentiviral ETV2	Park lab	N/A
Biological samples
C57BL/6J-Cdh5Cre:Etv2f/f mouse tissues	This paper	N/A
C57BL/6J-Cdh5Cre:Kdm4af/f mouse tissues	This paper	N/A
C57BL/6J-Cdh5Cre:Kdm4af/f:Etv2f/f mouse tissues	This paper	N/A
C57BL/6J-littermate control mouse tissues	This paper	N/A
Chemicals, peptides, and recombinant proteins
1-Thioglycerol	Sigma Aldrich	Cat# M6145
4-hydroxytamoxifen	Sigma Aldrich	Cat# H6278
Accutase	Sigma Aldrich	Cat# A6964
Antifade mounting medium with DAPI	Vector Laboratories	Cat# H-2000
Ascorbic acid	Sigma Aldrich	Cat# A4544
Blasticidin S HCl	Thermofisher Scientific	Cat# A1113903
Collagen solution	Sigma Aldrich	Cat# ECM675
Direct-zol RNA miniprep kit	Zymo Research	Cat# R2052
Doxycycline	Sigma Aldrich	Cat# D1822
EndoFectin™ max transfection reagent	GeneCopoeia	Cat# EF014
H&E staining kit (Hematoxylin and Eosin)	Abcam	Cat# ab245880
Hygromycin	Thermofisher Scientific	Cat# ant-hg-1
KnockOut™ serum replacement	Thermo Fisher Scientific	Cat# 10828028
Lipofectamine 2000	Thermo Fisher Scientific	Cat# 11668019
Lipofectamine 3000	Thermo Fisher Scientific	Cat# L3000015
Methylcellulose containing hematopoietic cytokines	Stemcell Technologies	Cat# M3434
Mitomycin C	Tocris	Cat# 3258
PFHM-II Protein-Free Hybridoma Medium	Thermofisher Scientific	Cat# 12040077,
Poly(2-hydroxyethyl methacrylate)	Sigma Aldrich	Cat# P3932
Polybrene	Sigma Aldrich	Cat# TR-1003
Polyethylene glycol (PEG) 6000	Sigma Aldrich	Cat# 81260
Protease inhibitor cocktail (100X)	Thermo Fisher Scientific	Cat# 87785
Protein A/G agarose beads	Santa Cruz	Cat# B01060010
Protein G magnetic beads	Sigma Aldrich	Cat# 16-662
Random primers	Promega	Cat# C1181
Recombinant human VEGF 165 protein	Bio Techne	Cat# 293-VE-010/CF
Recombinant mouse LIF protein	Sigma Aldrich	Cat# ESG1106
RNaseOUT	Thermofisher Scientific	Cat# 10777019
SuperScript IV Reverse Transcriptase	Thermo Fisher Scientific	Cat# 18 090 200
SYBR® green PCR master mix	Thermo Fisher Scientific	Cat# 4367660
Trichrome Stain Kit (Connective Tissue Stain)	Abcam	Cat# ab150686
Trizol reagent	Thermo Fisher Scientific	Cat# 15596018
Water, Cell Culture Grade	VWR	Cat# VWRL0200-0500
Zeocine	Invivogen	Cat# ant-zn
Critical commercial assays
ChIP DNA purification kit	Active Motif	Cat# 58002
DNA methylation enzyme kit	Qiagen	Cat# 335452
Dual-Luciferase reporter assay system	Promega	Cat# E1960
qPCR Lentivirus Titer Kit	Abm	Cat# LV900
Re-ChIP-IT Kit	Active Motif	Cat# 53016
TnT® coupled reticulocyte lysate systems (T7)	Promega	Cat# L4610
Deposited data
ChIP-seq data	NIH	https://www.ncbi.nlm.nih.gov/sra/PRJNA1051681
RNA-seq data	NIH	https://www.ncbi.nlm.nih.gov/sra/PRJNA1051681
Experimental models: Cell lines
HEK293T	ATCC	Cat# CRL-3216; RRID:CVCL_0063
mESCs; Doxycyline inducible (i)	Kimet al.21	N/A
mESCs: iFLAG-ETV2	This paper	N/A
mESCs: iFLAG-ETV2-HA-KDM4A	This paper	N/A
mESCs: iFLAG-ETV2-HA-KDM4A (H188A)	This paper	N/A
mESCs: Kdm4a/4c DKO	Pedersenet al.34	N/A
mESCs: Kdm4a/4c DKO + Etv2	This paper	N/A
PMEF	Sigma Aldrich	Cat# PMEF-NL
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Laboratory	Strain code: 000664
Mouse: Cdh5Cre	Jackson Laboratory	Strain code: 006137
Mouse: Cdh5Cre:Etv2f/f	Parket al.8	N/A
Mouse: Cdh5Cre:Kdm4af/f	This paper	N/A
Mouse: Cdh5Cre:Kdm4af/f:Etv2f/f	This paper	N/A
Mouse: Etv2f/f	Parket al.8	N/A
Mouse: Kdm4af/f	Jackson Laboratory	Strain code: 029424
Oligonucleotides
See Table S1 for all primer sequences	This paper	N/A
Recombinant DNA
CSII-EF1α-ETV2	This paper	N/A
pCMV-HA-GASC1	Addgene	Cat# 24214
pCMV-HA-JMJD2A	Addgene	Cat# 24180
pCMV-HA-JMJD2B	Addgene	Cat# 24181
pCMV-HA-JMJD3	Addgene	Cat# 24167
pcDNA4-FLAG-Jhdm2a	Addgene	Cat# 38136
pcDNA6-3xHA-hETV2	Parket al.8	N/A
pCDNA6.1-HA-KDM4A	This paper	N/A
pcDNA6.1-HA-KDM4A-H188A	This paper	N/A
pCDNA6.1-HA-KDM4C	This paper	N/A
pcDNA6.1-HA-KDM4C-S198M	This paper	N/A
pcDNA-Flag-hKDM4A	Addgene	Cat# 101051
pGL2-mFlk1 promoter/enhancer	Kappelet al.38	N/A
pMD2.G	Addgene	Cat# 12259
pMDLg/pRRE	Addgene	Cat# 12251
pRL-null	Promega	Cat# E2271
pRSV-Rev	Addgene	Cat# 12253
Software and algorithms
Fiji	ImageJ	https://imagej.net/software/fiji/
FlowJo	FlowJo	https://www.flowjo.com/
GraphPad Prism 4	GraphPad Software	N/A
Pimsoft	Perimed	https://www.perimed-instruments.com/
R 4.1.0	R Foundation	https://www.r-project.org/

Experimental model and study participant details

Cell culture and cell line generation

Maintenance and differentiation of mESCs were performed as described previously.^21,39,56 Briefly, mESCs cultured on mitomycin C treated mouse embryonic fibroblasts were cultivated on 0.1% gelatin coated plates. Two days later, dissociated mESCs were induced to differentiate in bacterial plates containing 15% preselected FBS and then subjected to flow cytometry and gene expression. For differentiation under serum-free conditions, cells were cultured in the presence of 15% KnockOut Serum Replacement (Thermofisher Scientific, Waltham, MA) ± Doxycycline (Dox, 1 μg/mL). The conditional deletion of Kdm4a and Kdm4c was achieved by incubating Rosa26-CreERT2;Kdm4a/4c^f/f mESCs in the presence of 500 nM 4-hydroxytamoxifen (Sigma-Aldrich, St. Louis, MO) for 14 days.³⁴ For the generation of iFLAG-ETV2-HA-KDM4A mESCs, the PCR-amplified HA-KDM4A inserts were first ligated into the BamHI and PacI sites of FUW-tetO vector (Addgene plasmid # 84008),⁵⁷ generating FUW-tetO-HA-KDM4A. Subsequently, doxycycline inducible iFLAG-ETV2 mESCs²¹ were transfected with the construct using EndoFectin Max (Genecopoeia, Rockville, MD) and then selected under Zeocine at 8 μg/mL (Invivogen, San Diego, CA). The positive clones were verified using genomic DNA PCR and transgene expression analysis. Primer sequences are listed in Table S1.

Mice

Animal husbandry, generation and handling were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of Louisiana State University Health Shreveport (P20-040) and Emory University (PROTO201700187/PROTO201700061). Cdh5-Cre mice, Kdm4a^f/f mice (The Jackson Laboratory, Bar Harbor, ME) and Etv2^f/f mice⁸ were used to generate endothelial Kdm4a^f/f, Etv2^f/f, Kdm4a^f/f;Etv2^f/f mice. Hindlimb ischemia was induced as previously described.^8,45 Mice (n = 5/group, 10–12 weeks old, male and female, C57BL/6) were anesthetized and subjected to femoral artery ligation, followed by laser speckle perfusion analysis.

Method details

Laser speckle perfusion analysis

Surface blood perfusion in hindlimbs was measured using a Laser Speckle Contrast Analyzer (LASCA, PeriCam PSI System, Järfälla Sweden) on days 0, 7, 14, 21, and 28 after the injury. Digital color-coded images were analyzed to quantify blood perfusion in the region from the hind foot to the toe, followed by calculation of the mean values of perfusion.

Chromatin immunoprecipitation (ChIP)-qPCR

iFLAG-ETV2 mESCs or iFLAG-ETV2-HA-KDM4A mESCs at day 3.5–4 of differentiation were cross-linked with 0.5% formaldehyde as described previously.³ Fragmentation of chromatin was achieved using a Covaris E210 Ultrasonicator (Covaris, Woburn, MA). The sonicated chromatin was incubated with 4 μg of anti-rabbit H3K9me3 antibody (Epigentek, Farmingdale, NY), anti-FLAG antibody (Sigma-Aldrich, St. Louis, MO), or anti-HA antibody (Abcam, Cambridge, MA). Anti-rabbit IgG antibody was used as a control. Re-chromatin immunoprecipitation (Re-ChIP) was performed using Re-ChIP-IT Kit (Active Motif, Carlsbad, CA). iFLAG-ETV2-HA-KDM4A mESCs ± Dox at D4 of differentiation were cross-linked with 0.5% formaldehyde and sonicated with a Bioruptor Pico sonication device (Diagenode, Denville, NJ). The sonicated DNAs were incubated with 3 μg of anti-FLAG antibody bound to protein G magnetic beads (Sigma-Aldrich, St. Louis, MO) overnight at 4°C. Subsequently, the bound DNAs were eluted and incubated with 3 μg of anti-HA antibody (Abcam, Cambridge, MA) overnight at 4°C. For Re-ChIP with endogenous ETV2 and KDM4A, D4 EBs differentiated from mESCs without Dox treatment were fixed, and sonicated. The resulting chromatin was incubated with 3 μg of rabbit anti-human KDM4A antibody (Bethyl Laboratories, Montgomery, TX) or rabbit IgG antibody (Diagenode, Denville, NJ) overnight at 4°C with protein G magnetic beads to conjugate with each antibody. Subsequently, the bound DNAs were eluted and further incubated with 3 μg of rabbit anti-mouse ETV2 antibody (Abcam, Cambridge, MA) overnight at 4°C. The bound DNAs were recovered and subjected to qPCR using the primers listed in Table S1.

Co-immunoprecipitation assay

HEK/293T cells were transfected with 5 μg of plasmids encoding HA-KDM4A (Addgene plasmid # 24180),⁵⁸ HA-KDM4B (Addgene plasmid # 24181),⁵⁸ HA-KDM4C (Addgene plasmid # 24214),⁵⁸ HA-KDM6B (Addgene plasmid # 24167),³⁶ FLAG-KDM3A (Addgene plasmid # 38136),⁵⁹ FLAG-ETV2 and/or HA-ETV2. After 48 h, harvested cells were lysed with NP40 lysis/IP buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10% glycerol, 0.5% IGEPAL, 0.2 mM DTT, protease inhibitor cocktail, 0.2 mM Na3VO4, 50 mM NaF). The resulting cell lysates were incubated in the buffer with 4 μg rabbit anti-FLAG (Sigma-Aldrich, St. Louis, MO) or 4 μg rabbit anti-HA (Abcam, Cambridge, MA) overnight at 4°C. Subsequently, 30 μL (50% suspension) of protein A/G agarose beads (Santa Cruz Biotechnology, Dallas, TX) were added and incubated for 4 h at 4°C. The antibody-bead complexes were then washed three times with the lysis/IP buffer and subjected to immunoblotting with mouse anti-FALG antibody (Sigma-Aldrich, St. Louis, MO, 1:1000) or mouse anti-HA antibody (Sigma-Aldrich, St. Louis, MO,1:1000). For the loading control, goat anti-α-TUBULIN antibody (Cell Signaling Technology, Danvers, MA, 1:1500) was used. For the direct interaction between ETV2 and KDM4A, the reaction mixture containing in vitro translated HA-ETV2, FLAG-KDM4A proteins or both were pre-cleared with 20 μL of a 50% slurry bead (Santa Cruz Biotechnology, Dallas, TX) in 1 mL of NP-40/IP binding buffer (20 mM Tris pH 8.0, 137 mM NaCl, 10% Glycerol, 1% NP-40, 2 mM EDTA, protease inhibitor cocktail) for 3 h at 4°C. The cleared supernatants were then incubated with 3 μL of anti-FLAG antibody (Sigma-Aldrich, St. Louis, MO) overnight at 4°C. Next day, 30 μL of 50% slurry beads were added to each tube and further incubated for 5 h at 4°C. After centrifugation, the beads were washed with NP40/IP buffer three times and subjected to immunoblotting with mouse anti-FLAG antibody (Sigma-Aldrich, St. Louis, MO, 1:1000) or rabbit anti-HA antibody (Abcam, Cambridge, MA). For the endogenous interaction between ETV2 and KDM4A, the differentiated wild type mESCs were lysed in NP-40 lysis buffer [20 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 1% NP-40, supplemented with 1xPIC] and the lysate (1.5 mg/IP) was pre-cleared with 50 μL of a 50% slurry bead (Santa Cruz Biotechnology, Dallas, TX) in 1 mL of NP-40/IP binding buffer [20 mM Tris pH 8.0, 137 mM NaCl, 10% Glycerol, 1% NP-40, 2 mM EDTA, Protease inhibitor cocktail] for 3 h at 4°C. The cleared supernatant was then incubated with 5 μL of rabbit anti-human KDM4A antibody (Bethyl Laboratories, Montgomery, TX) or rabbit IgG antibody (Diagenode, Denville, NJ) at 4°C overnight, followed by immunoblotting with rat anti-mouse ETV2 antibody (R&D Systems, northeast Minneapolis, MN).

In vitro transcription and translation

Recombinant proteins ETV2 (pcDNA6-3xHA-hETV2) and KDM4A (pcDNA-Flag-hKDM4A, addgene #101051) were produced in vitro using a transcription and translation system (TnT Coupled Reticulocyte Lysate Systems T7, Promega, Madison, WI). Briefly, each plasmid (2 μg) was incubated with the transcription/translation reaction mixture containing 25 μL of TNT Rabbit Reticulocyte Lysate, 2 μL of TNT Reaction Buffer, 1 μL of TNT RNA Polymerase (T7), 0.5 μL of both Amino Acid Mixtures minus Leucine and Methionine (1mM each), 1 μL of RNasin Ribonuclease Inhibitor (40 u/μL), and 1 μL of Transcend Biotin-Lysyl-tRNA for 90 min at 30°C. Subsequently, the resulting reaction mixture containing in vitro translated HA-ETV2 protein and FLAG-KDM4A, respectively, was mixed and subjected to Co-IP/western blot analysis as described above.

Luciferase-based promoter assay

HEK/293T cells (1.3 x 10⁵/well in a 24-well plate) were transfected with 1 μg ETV2 expression plasmid (pIRES-FLAG-ETV2), 1 μg KDM4A or KDM4A-H188A expression plasmid (pCDNA6.1-HA-KDM4A or pcDNA6.1-HA-KDM4A-H188A), 1 μg KDM4C or KDM4C-S198M expression plasmid (pCDNA6.1-HA-KDM4C or pcDNA6.1-HA-KDM4C-S198M), 200 ng pGL2-mFlk1 promoter/enhancer constructs,³⁸ and 30 ng pRL-null using lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA). Forty-eight hours later, cells were harvested and the luciferase activity was measured using the Dual-Luciferase reporter assay system (Promega, Madison, WI) according to the manufacturer’s instructions. Firefly luciferase values were divided by Renilla luciferase values to normalize the transfection efficiency.

Flow cytometry

A single cell suspension was incubated with freshly diluted PE anti-mouse FLK1 antibody (Biolegend, San Diego, CA, 1:200), biotin-conjugated anti-human CD4 antibody, and/or biotin-conjugated anti-mouse CDH5 antibody (Biolegend, San Diego, CA, 1:200) in staining/wash buffer (4% FCS in PBS) for 15 min on ice, washed three times with the staining/wash buffer, and then incubated with APC streptavidin-conjugated secondary antibody (Biolegend, San Diego, CA, 1:300) in staining/wash buffer for 15 min. Cells were then analyzed on an LSRII flow cytometer (Becton-Dickinson, Franklin Lakes, NJ). The data acquired were analyzed with Flow-Jo software (Becton-Dickinson, Franklin Lakes, NJ).

Lentivirus production and rescue experiments

HEK293FT cells (Thermofisher Scientific, Waltham, MA) seeded at 4 x 10⁶ cells in a 100mm dish were transfected with 10 μg of CSII-EF1α-ETV2, 2.5 μg of pMD2.G (Addgene; #12259), 2.5 μg of pMDLg/pRRE (Addgene; #12251), and 2.5 μg of pRSV-Rev (Addgene; #12253) using 40 μL of lipofectamine 3000 (ThermoFisher Scientific, Waltham, MA). On day 3 post-transfection, the supernatant containing lentiviral particles from the culture media was collected, centrifuged at 500 x g for 10 min at 4°C and filtered through a 0.45 μm cellulose acetate membrane syringe filter (Thermofisher Scientific, Waltham, MA). The lentiviral particles were subsequently concentrated using polyethylene glycol (PEG) 6000 as previously described.⁶⁰ Lentiviral titer was determined using qPCR Lentivirus Titer Kit (Abm, Richmond, BC, Canada) according to the manufacturer’s instruction. To generate KDM4c/4c DKO mESCs overexpressing Etv2, the cells seeded at 1 x 10⁵ cells/well on a 6 well plate for 16–24 h were incubated with the lentiviral particles of Etv2 (MOI, 50) with 8 μL/mL of polybrene (Millipore sigma, Burlington, MA) were added into the wells. Twenty-four hours later, the culture medium containing the lentiviral particles was replaced with a fresh medium and the cells were selected with 30 μg/mL of hygromycin (Thermofisher Scientific, Waltham, MA) for 10 days. The resulting positive clones were validated and subjected to differentiation as previously described.²¹

Hematopoietic colony assay

Hematopoietic colonies were generated as described previously.^3,21 Briefly, 6 x 10⁴ cells obtained from differentiated mESCs (i.e., EBs, three-dimensional aggregates of differentiated ESCs) maintained for 6 days under serum free conditions ± Dox (1 μg/mL) were mixed with 1mL methylcellulose containing hematopoietic cytokines (M3434, STEMCELL Technologies, Canada) and replated on a 35 mm bacterial dish. Colonies were counted 4–6 days later.

Embryoid body sprouting assay

Embryo bodies harvested at D4 of differentiation using the hanging drop method (20,000 ES cells/20 μL drop) were mixed with chilled collagen solution (5–6 EBs in 200 μL collagen solution) (ECM675, MilliporeSigma, Burlington, MA) in a 24-well plate.⁸ After incubating for 30 min at 37°C, the mixtures were cultured in sprouting assay medium (2% FBS, 20 ng/mL VEGF in IMDM) ± Dox for 8 days with the medium changed every other day. A phase contrast sprouting image was taken with ImageJ using the magic wand function.

Immunohistochemistry

At day 28 post injury, muscles harvested from the ischemic hindlimbs were fixed in 10% (v/v) buffered formaldehyde, embedded in paraffin and sectioned (8–10 μm in thickness). For immunohistochemistry, the tissue sections were incubated overnight at 4°C with FITC-conjugated ISOLECTIN B4 (1:75, Vector laboratories, Burlingame, CA). Nuclei were counterstained using Slow Fade Gold with 4′, 6-diaminoindole (DAPI) (Vector laboratories, Burlingame, CA) and examined using a confocal laser scanning microscope (Leica TCS SP5 Confocal, Leica, Wetzlar, Germany). For vessel density analysis, images were randomly selected, and vessel density was analyzed by determining the percentage of ISOLECTIN B4⁺ vessel area over the total measurement area (2 images x 3 sections x 3 mice/group = 18 images/group). H&E and Masson’s trichrome collagen staining was performed to analyze tissue necrosis and fibrosis in the ischemic regions, respectively.

Quantitative real-time RT-PCR

Total RNAs were extracted by the addition of 1 mL of TRIzol reagent (Thermo Fisher Scientific, Waltham, MA) following the manufacturer’s instructions. One μg of total RNA was subjected to cDNA synthesis with SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific, Waltham, MA) and random primers (Promega, Madison, WI). SYBR Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA) was used to detect the accumulation of PCR products during cycling with the ABI Real-time PCR 7500 (Applied Biosystems, Foster City, CA). Information on each primer is provided in Table S1. Fold differences in the expression level of each gene were calculated for each treatment group using CT values normalized to transcript levels of Gapdh.

RNA sequencing and analysis

All bulk RNA-seq and the initial analysis for differential genes were performed with Amera Health, NJ, USA. Briefly, total RNA prepared by the Direct-zol RNA miniprep kit (Zymo Research, Irvine, CA, USA) was subjected to the quality analysis with Agilent Tapestation using RNA ScreenTape (Agilent Technologies Inc., CA, USA) and quantification with Qubit 2.0 RNA HS assay (Thermofisher Scientific, Waltham, MA). After depleting Ribosomal RNA depletion with QIAseq FastSelect rRNA HMR Kit (Qiagen, Hilden, Germany), all libraries were constructed according to the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England BioLabs, Ipswich, MA). Subsequently, the quality and the quantity of the libraries were analyzed by TapeStation HSD1000 ScreenTape (Agilent Technologies Inc., CA, USA) and Qubit 2.0 (Thermofisher Scientific, Waltham, MA), respectively. The final library size was about 430bp with an insert size of about 300bp. Illumina 8-nt dual-indices were used. An Illumina Novaseq platform (Illumina, San Diego, CA) with a read length configuration of 150 PE for 60M PE reads per sample (30M in each direction) was used to sequence the libraries pooled equimolarly. Analysis of the sequencing results is as follows. The quality of raw reads was controlled with astQC (version v0.12.1) and cutting adaptors and trimming low-quality bases were performed using Trimmomatic (version v0.39) with default setting. Alignment of the reads and the quality check of mapping were conducted with STAR Aligner (version 2.7.10b) and the package of Picard tools (version 3.0.0), respectively. Subsequently, while construction of the RNA-Seq alignments into potential transcripts was performed with StringTie (version 2.2.1), counting mapped reads for detailed genomic features including genes, exons and promoters with FeatureCounts (version 2.0.6) or HTSeq (version 2.0.3), followed by differential analysis with DESeq2 (version 1.18.1). Genes with adjusted p Value less than 0.05 were defined as significant. Genes with log fold change which equals or larger than 1 were defined as differentially expressed genes. All downstream analysis including plotting and statistical tests were conducted using R 4.3.2. Gene ontology (GO) analyses were performed using the Panther overrepresentation test and GO database annotation with Fisher test and Bonferroni correction. Gene interactome plot was obtained using GENEMANIA (genemania.org).

ChIP sequencing (ChIP-seq) and analysis

The H3K9me3 ChIP-seq experiment was performed with Diagenode (Denville, NJ). The chromatin was prepared by Diagenode ChIP-seq/ChIP-qPCR Profiling service (Diagenode, Denville, NJ) using the iDeal ChIP-seq kit for Histones (Diagenode, Denville, NJ). Briefly, iETV2-FLAG mESCs differentiated for 4days ± 1 μg/mL Dox (control and Dox treated groups, n = 3/group) were cross-linked with 1% formaldehyde and the cells were subjected to chromatin fragmentation using Bioruptor Pico sonication device (1 x 10⁶ cells in 100 μL) (Diagenode, Denville, NJ). The size of the fragmented genomic DNA was determined by High Sensitivity NGS Fragment Analysis Kit (DNF-474) on a Fragment Analyzer (Agilent Technologies Inc., CA, USA). Then, the fragmented DNAs were immunoprecipitated (ChIP’d) with anti-H3K9me3 antibody or a negative control isotype IgG. Library preparations were performed with the ChIP’d DNA and input DNA using MicroPlex Library Preparation Kit v3/96 rxns (Diagenode, Denville, NJ) with UDI for MicroPlex v3 - Set I and Set II (Diagenode, Denville, NJ). After assessing optimal library amplification by qPCR and by using High Sensitivity NGS Fragment Analysis Kit (DNF-474) on a Fragment Analyzer (Agilent, Santa Clara, CA), the libraries were amplified and purified using Agencourt AMPure XP (Beckman Coulter, Brea, CA). The quantity and fragment size of the library was determined using Qubit dsDNA HS Assay Kit (Thermofisher Scientific, Waltham, MA) and High Sensitivity NGS Fragment Analysis Kit (DNF-474) on a Fragment Analyzer (Agilent, Santa Clara, CA), respectively. The resulting libraries were sequenced using the Illumina HiSeq2000 as paired-end reads. Reads were filtered based on a mean base quality score, and mapped to GRCm38/mm10 genome using Bowtie.⁶¹ To minimize PCR amplification bias, we removed duplicate reads were removed using MarkDuplicates.jar from picard-tools package (GitHub: https://broadinstitute.github.io/picard/). EPIC was used for peak calling (GitHub: https://github.com/biocore-ntnu/epic). We performed a comparison between the two groups (control and Dox treated) on called peaks using Diffbind (R package).⁶² In addition, we estimated consensus peaks in each group using Diffbind (R package) and discovered specific peaks to each group. Afterward, annotation and functional analysis were performed for each specific consensus peak using ChIPseeker (R package).⁶³

Quantification and statistical analyses

The data were analyzed using the GraphPad Prism 4 version 4.03 software (Graph-Pad Software, La Jolla, CA). The differences in all data were assessed using one-way analysis of variance. When the p value in the analysis of variance test indicated statistical significance, the differences were assessed using the Tukey’s test. A value of p < 0.05 was considered statistically significant.

Published: December 6, 2024