Fli1⁺ cells transcriptional analysis reveals an Lmo2–Prdm16 axis in angiogenesis

Significance

Understanding the molecular signals driving the development of vascular endothelium is important in seeking therapies for human disease. This work highlights the role of an epigenetic modifier and oncogene in the development of the vascular endothelium and angiogenesis in zebrafish. In particular, we show that the expression of the histone methyltransferase prdm16 is high in endothelial cells and that prdm16 is necessary for the endothelial differentiation and migration in vivo and in vitro in iPSC-derived endothelial cells. Moreover, prdm16 expression is mediated by Lmo2, a well-known oncogene implicated in angiogenesis and leukemogenesis. We hence unveil a role of prdm16 in endothelial development and angiogenesis and therefore propose that PRDM16 could be a target for therapeutic modulation of angiogenesis.

Abstract

A network of molecular factors drives the development, differentiation, and maintenance of endothelial cells. Friend leukemia integration 1 transcription factor (FLI1) is a bona fide marker of endothelial cells during early development. In zebrafish Tg(fli1:EGFP)^y1, we identified two endothelial cell populations, high-fli1⁺ and low-fli1⁺, by the intensity of green fluorescent protein signal. By comparing RNA-sequencing analysis of non-fli1 expressing cells (fli1⁻) with these two (fli1⁺) cell populations, we identified several up-regulated genes, not previously recognized as important, during endothelial development. Compared with fli1⁻ and low-fli1⁺ cells, high-fli1⁺ cells showed up-regulated expression of the zinc finger transcription factor PRDI-BF1 and RIZ homology domain containing 16 (prdm16). Prdm16 knockdown (KD) by morpholino in the zebrafish larva was associated with impaired angiogenesis and increased number of low-fli1⁺ cells at the expense of high-fli1⁺ cells. In addition, PRDM16 KD in endothelial cells derived from human-induced pluripotent stem cells impaired their differentiation and migration in vitro. Moreover, zebrafish mutants (mut) with loss of function for the oncogene LIM domain only 2 (lmo2) also showed reduced prdm16 gene expression combined with impaired angiogenesis. Prdm16 expression was reduced further in endothelial (CD31⁺) cells compared with CD31⁻ cells isolated from lmo2-mutants (lmo2-mut) embryos. Chromatin immunoprecipitation–PCR demonstrated that Lmo2 binds to the promoter and directly regulates the transcription of prdm16. This work unveils a mechanism by which prdm16 expression is activated in endothelial cells by Lmo2 and highlights a possible therapeutic pathway by which to modulate endothelial cell growth and repair.

The endothelium is the inner lining of blood and lymphatic vessels made of a continuous monolayer of endothelial cells (ECs) interconnected by specialized junctions. It is present in all vertebrates (1) and is highly specialized to adapt to the need of the different tissues and organs (2). It regulates not only the transit of nutrients and oxygen from the blood into tissues and the removal of metabolic waste products, but also blood fluidity, platelet aggregation, vascular tone, inflammation, and angiogenesis. In fact, endothelial dysfunction is a fundamental pathophysiology associated with many human diseases and conditions, including air pollution (3), smoking (4), heart failure (5), hypertension (6), atherosclerosis (7), and cancer (8). Therefore, a detailed understanding of the underlying mechanisms of EC proliferation and differentiation are crucial to the aim of seeking therapeutic tools to promote restoration of dysfunctional or damaged endothelium.

Zebrafish have a closed circulatory system, and the anatomical processes and the molecular mechanisms underlying vascular development are very similar to humans (9). During embryogenesis, endothelial and blood progenitor cells share a common ancestor, the hemangioblast, derived from the lateral plate of the mesoderm (10) that gives rise to hematopoietic stem cells and endothelial precursors. Early blood vessels form de novo from these endothelial progenitors, the angioblasts, in a process named vasculogenesis (11). This primary vasculature expands and remodels through sprouting into a mature vascular network, also known as developmental angiogenesis (12). The differentiation of ECs requires an orchestrated network of different transcription factors yet to be fully defined (13). One of these, Friend leukemia integration 1 transcription factor (FLI1), a member of the E26 transformation–specific family of transcription factors, can be detected at low levels as early as the three-somite stage within the posterior lateral mesoderm in zebrafish (14). Fli1 is required for hemangioblast formation, as shown in Xenopus and zebrafish (Danio rerio) (10). Hence, Fli1 is one of the earliest known endothelial markers, acting downstream to Npas4l (cloche) (15, 16) and upstream of transcription factors such as Scl/Tal1, Lmo2, Gata2, Etsrp, and Flk1. As such, Fli1 helps to initiate and maintain EC fate (10).

The zebrafish transgenic line Tg(fli1:EGFP)^y1 has enhanced green fluorescent protein signal (EGFP) driven by the fli1 promoter. This model has been extensively used to elucidate endothelial and blood vessel development (14). Fli1 is specific for EC, although low levels of green fluorescent protein (GFP) signal can also be detected in some other tissue, including mesenchyme and developing cartilage of the jaw (14). In the current study, we observed that fli1 gene expression increased rapidly during zebrafish embryogenesis between 6 h and 72 h postfertilization (hpf) (Fig. 1A). Accordingly, we fluorescence-activated cell sorting (FACS)-purified three populations of cells from Tg(fli1:EGFP)^y1 embryos on the basis of their differential expression of fli1 (Fig. 1B), termed fli1⁻, low-fli1⁺, and high-fli1⁺cells. We compared these subpopulations to discover genes involved in early endothelial development. We reasoned that if a gene is specific for the differentiation of EC, it will be expressed more in the high-fli1⁺ population compared with the low-fli1⁺. Hence, separating the high-fli1⁺ from the low-fli1⁺ will achieve two key aims: 1) reveal candidate genes that are highly expressed only in high-fli1⁺ cells, the level of which would be diluted or masked if observed in a population containing both high-fli1⁺ and low-fli1⁺ cells; and 2) discriminate those genes whose expression is significantly higher in low-fli1⁺ compared with high-fli1⁺. For this study, we used embryos at 72 hpf since we observed that fli1 gene expression and the percentage of GFP⁺ cells were close to maximal at this time point (Fig. 1 C–D) and a high-fli1⁺ population was clearly distinguishable at this early developmental stage (Fig. 1B). A GFP⁻ (fli1⁻) population was defined using zebrafish control Wik line, whereas the gate between the two GFP⁺ populations (low-fli1⁺ and high-fli1⁺) was placed at the onset/edge of the high-fli1⁺ population (Fig. 1B). To confirm that cells expressing the highest GFP signal were in fact bona fide EC, low- and high-fli1⁺ cells were sorted and cultured on a fibronectin-coated Petri dish with EC growth medium (Fig. 1E). Fibronectin is an extracellular matrix protein that serves as adhesive support and regulates the spreading, migration, and contractility of EC during vascular development (17). At 24 h after seeding, only 29% of low-fli1⁺ cells were attached to the fibronectin compared with the 89% of high-fli1⁺ cells (Fig. 1 E–F). In addition, from 24 to 48 h after seeding, the increase in high-fli1⁺ cells was significantly greater than low-fli1⁺ cells (45% versus 5%) (Fig. 1 E–F).

Fig. 1.

Characterization of low- and high-fli1⁺ cells. (A) qPCR analysis of fli1 gene expression during zebrafish development (from 1 to 120 hpf). (B) Following enzymatic disaggregation of Tg(fli1:EGFP)^y1 zebrafish embryos at 72 hpf, two populations of low- and high-fli1⁺ cells were identified. Wik zebrafish were used to set the gate for fli1⁻ cells. (C and D) Line graphs showing the percentage of low- and high-fli1⁺ cells at different developmental stages (C) and relative fli1 gene expression (D). (E) Fluorescent images of low- and high-fli1⁺ cells 24 h after seeded on fibronectin-coated dishes with endothelial growth medium. (F) Table of cell characterization (percent attached cells and cell number increase). n = 3 experiments; ANOVA test followed by Bonferroni post hoc was used to compare means. ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05.

To investigate the gene expression profile, we performed RNA-sequencing (RNA-Seq) for fli1⁻, low-fli1⁺, and high-fli1⁺ cell populations. We compared the level of genes in each pair of cell populations and defined differential genes by requiring a fold change larger than 2 and a false discovery rate (FDR) <0.05 (Fig. 2). We were particularly interested in genes encoding epigenetic factors that were previously unrecognized in endothelial differentiation.

Fig. 2.

Bioinformatic analysis. (A–C) Heat maps displaying the expression level of genes up- or down-regulated in the negative-, low-, and high-fli1⁺ cells, with each of the two populations compared separately. Genes for which the fold change was bigger than two and FDR was less than 0.05 were included in the heat maps. (D and E) Bar plots showing enrichment Q- values of functional terms in genes up-regulated (D) or down-regulated (E) in high-fli1⁺ compared with low-fli1⁺. Up-regulated or down-regulated genes were defined based on EdgeR FDR cutoff 1e-5.

RNA-Seq analysis revealed a set of 441 genes up-regulated and 1,024 genes down-regulated in low-fli1⁺ compared with fli1⁻ cells (Fig. 2A), a set of 660 genes up-regulated and 1,333 genes down-regulated in high-fli1⁺ compared with fli1⁻ cells (Fig. 2B), and a set of 751 genes up-regulated and 793 genes down-regulated in high-fli1⁺ compared with low-fli1⁺ cells (Fig. 2C). Mesodermal markers showed an increased level of fragments per kilobase of transcript per million mapped reads from fli1⁻ to low-fli1⁺ to high-fli1⁺ cells (SI Appendix, Fig. S1A), whereas opposite patterns were observed for ectodermal (SI Appendix, Fig. S1B) and endodermal markers (SI Appendix, Fig. S1C). Pathway enrichment analysis showed clear differences between the groups (Fig. 2 D–E and SI Appendix, Fig. S2 and Dataset S1). Compared with fli1⁻, and as expected, low-fli1⁺ and high-fli1⁺ groups showed enrichment of blood vessel development pathways, including vascular development, angiogenesis, lymphangiogenesis, and tgf-β signaling pathway (SI Appendix, Fig. S2A), with the highest enrichment of these pathways observed in the high-fli1⁺ cells (Fig. 2D). This was also true for other endothelial markers, such as chd5, kdr, pecam, tie1, and fli1 (SI Appendix, Fig. S3 A and B), as expected and as confirmed by qPCR (SI Appendix, Fig. S3C). These data suggest that a high-fli1⁺ population is composed of ECs, whereas the low-fli1⁺ population could be composed of a variety of cell types, including ECs at an early maturation stage (Fig. 2D and SI Appendix, Figs. S2A and S3).

While several transcription factors are known to play a role in endothelial development and maturation (13), the role of other epigenetic factors in these processes is not fully understood. These are enzymes (methyltransferases, demethylases, deacetylases, acetyltransferases, and chromatin remodelers) (18) that modify the epigenome directly through DNA methylation or modifications of histones. Accordingly, we identified 510 epigenetic factors from our RNA-Seq dataset that were differentially expressed in our three defined cell populations (SI Appendix, Fig. S4). We were particularly interested to screen epigenetic factors that had low or null expression in fli1⁻ and high expression in the other two cell groups.

In a shortlist of 15 candidate genes, we found three members of the PRDI-BF1 and RIZ homology domain containing (prdm16) gene family: prdm5, 11, and 16 (Fig. 3A). Of these, prdm16 showed the highest level of expression in the high-fli1⁺ cells (Fig. 3B), confirmed by qPCR (Fig. 3C), and so this was selected for further study. Prdm16 is a 140-kDa zinc finger protein, the chromatin modifying activity of which is structurally defined by the presence of a conserved N-terminal histone methyltransferase (19). Prdm16 is quite ubiquitously expressed and is known to be involved in hematopoiesis (20), palatogenesis (21), brown fat determination and differentiation (22), and neurovascular network formation during brain development (23). In zebrafish, prdm16 plays a role in craniofacial development (24). It has also been implicated in several human conditions, including cardiomyopathy (25) and leukemogenesis (26).

Fig. 3.

Assessment of prdm16 function in Tg(fli1:EGFP)^y1 zebrafish. (A) Prdm16 was selected from a list of epigenetic modifiers. (B) RNA-Seq read mapping of prdm16 for all three groups of fli1⁻, low-, and high-fli1⁺, in duplicate. (C) qPCR analysis confirmed RNA-Seq data showing the higher expression of prdm16 in fli1⁺ cells. (D) Kaplan-Meyer survival curve of zebrafish embryos following prdm16-targeted Mo injection and control (mismatch). (E) Brightfield images (Upper) of whole embryos showing the gross phenotype following injection of prdm16 Mo. Fluorescent images of the trunk regions in Tg(fli1:EGFP) (Middle) and Tg(kdrl:mCherry) (Lower) embryos, showing changes in the ISVs following prdm16 KD. These effects were rescued when the prdm16 Mo was coinjected with prdm16 mRNA. (F) The average lengths of the ISVs. (G and H) FACS graph and data analysis showing changes in the percentage of low- and high-fli1⁺ population following prdm16 KD. (I–K) Analysis of total fli1⁺ cells isolated from prdm16 KD and control embryos and cultured on fibronectin-coated dishes. Merged brightfield and fluorescent images (I), fluorescence images showing the uptake of acetylated low-density lipoprotein (J), and nitrate/nitrite assay as measure of nitric oxide (K). n = 3 experiments; ANOVA test followed by Bonferroni post hoc was used to compared means. NS, not significant. ***P ≤ 0.001, *P ≤ 0.05.

To assess the role of prdm16 in zebrafish vascular development, we conducted prdm16 knockdown (KD) studies by injection of two different, nonoverlapping prdm16 translation-blocking morpholinos (Mo) in Tg(fli1:EGFP)^y1 one-cell stage embryos, following previous guidelines (27). We injected each Mo separately, at the same dose (0.5 ng per egg), and then we coinjected each Mo at 1/2 dose of each (0.25 ng per egg). While the phenotype was just slightly apparent with 1/2 dose of each Mo when injected alone, coinjection of two different Mo produced a stronger effect, similar to that produced by a single Mo at the dose of 0.5 ng per egg. This additive action strongly suggests that the Mo effects are specific for prdm16. The effective KD was verified by Western blot (SI Appendix, Fig. S5A). The survival rate of prdm16-KD embryos at 120 hpf was ∼70% compared with 92% in controls (Fig. 3D). The main features of the phenotype in prdm16-KD embryos assessed by brightfield microscopy included body bending and reduced skin pigmentation (SI Appendix, Fig. S5 B–C). We did not observe an increase in tp53 gene expression, that could cause off-target effects at the dose of Mo used in this study and up to 4 ng; therefore, we decided not to coinject tp53 Mo. Using fluorescence microscopy, we observed that the development of intersegmental vessels (ISV), which sprout from dorsal aorta via angiogenesis (28), was significantly reduced in prdm16-KD embryos compared with control (Fig. 3 E and F), with a phenotype penetrance of >70%. Indeed, prdm16 Mo–injected embryos with the ISV phenotype showed a significant reduction of prdm16 compared with control embryos or those embryos not showing the ISV phenotype, as shown by Western blot analysis (SI Appendix, Fig. S5D). This suggests that the ISV deficiency is a specific feature of the prdm16 depletion. In fact, at 30 hpf, ISVs bridged fully ventro-dorsally in control embryos, whereas they sprouted only halfway across their dorsal trajectory in prdm16 morphants (data are reported graphically as average ISVs length, Fig. 3 E and F). These observations were confirmed in the Tg(kdrl:mCherry)^y171 zebrafish, where the red fluorochrome mCherry is expressed under control of kdrl gene, an endothelial marker. The phenotype of prdm16 morphants was significantly rescued by coinjecting prdm16 Mo and prdm16 messenger RNA (mRNA), indicating that the observed effects were specific for prdm16 KD (Fig. 3 E and F and SI Appendix, Fig. S5A). As a negative control of the rescue experiment (27), coinjection of prdm16 Mo with a mutant RNA that does not encode proteins failed to rescue the Mo-induced defects. In contrast, vasculogenesis was not affected in prdm16-KD embryos, as shown by the normal development of dorsal aorta and cardinal vein (Fig. 3E). We performed cell proliferation and apoptosis assays in Tg(fli1:EGFP)^y1 zebrafish whole embryos at 22 hpf (end of segmentation stage) and at 30, 38, and 46 hpf (pharyngula period) to assess if these processes were involved in the reduced angiogenesis observed in prdm16-KD embryos (SI Appendix, Fig. S6). Ethynyl-2′-deoxyuridine (EdU) proliferation assay (SI Appendix, Fig. S6A) showed that the number of GFP⁺-EdU⁺ cells, marking proliferating ECs, were not different in prdm16-KD embryos compared with controls. Interestingly, GFP⁺-EdU⁺ cells in prdm16-KD embryos are more localized in the dorsal aorta compared with control where GFP⁺-EdU⁺ cells are localized also in the ISV. Similarly, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay (SI Appendix, Fig. S6B) showed that the number of GFP⁺-TUNEL⁺ cells, labeling apoptotic ECs, were not different in prdm16-KD embryos compared with controls. Overall, these data suggest that prdm16 is required for developmental angiogenesis. FACS analysis of cells isolated from prdm16-KD Tg(fli1:EGFP)^y1 embryos revealed that the percentage of high-fli1⁺ cells was significantly reduced compared with control (1.5% versus 7%, respectively), whereas the low-fli1⁺ population was significantly increased (7.9% versus 3.2% in control) (Fig. 3 G and H). The expression of endothelial markers was significantly lower in fli1⁺ from prdm16-KD compared with control (SI Appendix, Fig. S7). In vitro, total fli1⁺ cells plated on fibronectin-coated dishes and isolated from prdm16-KD exhibited a lower GFP signal compared with controls. In addition, they showed a nonhomogeneous morphology and a reduced low-density lipoprotein (LDL) uptake and nitric oxide production compared with controls (Fig. 3 I–K). Taken together, these data suggest that prdm16 is essential for angiogenesis in vivo and that prdm16 regulates the functionality of EC.

To assess whether the role of PRDM16 is conserved during the differentiation of human EC, we differentiated human-induced pluripotent stem cells (iPSC). iPSCs at passage between 21 to 23 were differentiated to endothelial lineage using our standardized protocol (29) (Fig. 4A). Cells were treated with small interfering RNA (siRNA), control, or targeting PRDM16 for 6 h until the end of the mesodermal differentiation. At the end of the protocol, ECs showed the typical EC cobblestone–like shape in the two groups (Fig. 4B). qPCR showed a reduction of PRDM16 expression of ≥70% in siRNA PRDM16–treated cells compared with control, as also evidenced by Western blot (Fig. 4C). Furthermore, qPCR showed a significant reduction in endothelial markers following PRDM16 KD (Fig. 4D). Accordingly, at this stage, we sorted double-positive cells for the endothelial markers CD31 (PECAM1 gene) and CD144 (CDH5 gene). We observed a reduced percentage of CD31⁺-CD144⁺ cells generated from iPSCs treated with siRNA PRDM16 compared with control (51.7% versus 6.4%; P ≤ 0.01; Fig. 4 E and F). We purified cells that were double positive for CD31 and CD144 and confirmed the reduced expression of endothelial markers in EC after PRDM16 KD (Fig. 4G). We observed no difference in the proliferation rate of iPSC-derived EC in the two groups, as shown by EdU assay (SI Appendix, Fig. S8A). TUNEL assay in cells from 9 to 15 d of the differentiation protocol, which encompasses differentiation and maturation stages of EC, shows that cell death is not increased following PRDM16 KD (SI Appendix, Fig. S8B). Furthermore, in the EC migration assay, we observed that iPSC-derived EC treated with siRNA for PRDM16 migrated significantly less, as shown by a larger residual gap between the edge of the wound monolayer, compared with cells treated with siRNA control (Fig. 4 H and I).

Fig. 4.

Assessment of PRDM16 function in iPSC-EC. (A) Differentiation protocol for the generation of iPSC-derived ECs (see details in Methods). At day 7, cells were incubated with siRNA control (Ctrl) or targeting PRDM16. (B) Brightfield images of an iPSC colony and of the iPSC-derived ECs, showing a different phenotype following treatment with SiRNA for PRDM16. (C) qPCR bar graph and Western blot showing the reduced expression of PRDM16 mRNA and protein in cells following PRDM16 siRNA treatment. (D) qPCR data showing the reduced expression of endothelial markers in iPSC-derived EC following PRDM16 KD. (E) FACS analysis for the endothelial markers CD31 and CD144 in iPSC-EC showing a significantly reduced population of CD31⁺-CD144⁺ in PRDM16 KD cells compared with controls, as shown in the bar graph in F. In negative samples, no first antibodies were added. (G) Immunofluorescence staining for endothelial markers CD31 and CD144 (vascular endothelial (VE)-Cadherin). iPSC-EC with no first antibodies added were used as negative control. (H) Migration assay in iPSC-EC showed that PRDM16 KD reduces cell migration, as shown by a larger residual gap between the edges of the wounded monolayer, quantified in I. n = 3 experiments; Student’s t test, ***P ≤ 0.001, **P ≤ 0.01, compared with controls.

Horn et al. (30) showed that PRDM16 is expressed in mouse embryos by E14.5 in a broad range of tissues, including brain, lung, kidney, and gastrointestinal tract. The mechanisms by which a ubiquitous factor such as PRDM16 is modified by tissue or organ context to provide such a variety of functions is mostly unknown. There are several possible mechanisms acting in isolation or in combination, such as protein modifications, specific epigenetic landscapes, or interactions with various cell- or tissue-specific factors, that make PRDM16 both multifunctional and tissue specific at the same time.

LIM domain only 2 (LMO2) is a transcription factor essential to both hematopoietic (31) and endothelial pathways (32), and it is normally expressed in mature vascular endothelium (33). LMO2 is a protooncogene implicated in leukemogenesis (34, 35) and has been suggested as a potential therapeutic target in various clinical indications (36). By virtue of its LIM-domain zinc-finger–like structures, LMO2’s canonical function is to act as a bridging molecule (37) to assemble a DNA-binding complex, which includes the TAL1, E47, GATA1, and Ldb1 proteins in erythroid cells (38) and in EC (39). We have previously shown that Lmo2 is necessary for EC proliferation during tissue regeneration in adult zebrafish (40) and for the regulation of EC migration mediated by Sphk1 in zebrafish embryos (41). In the present study, we have knocked out lmo2 in zebrafish by CRISPR/Cas9 (Fig. 5) to further investigate the role of this transcription factor in embryogenesis, and we have explored its possible interaction with prdm16. The guide RNA (gRNA) targeted the sequence between intron/exon two of the lmo2 gene and produced an insertion of four nucleotides (AGAT) resulting in a frameshift and a stop codon a few nucleotides downstream (Fig. 5A). Western blot analysis confirmed lmo2 knockout (Fig. 5B). The survival of lmo2-mut embryos was ∼75% at 120 hpf compared with 94% in controls (Fig. 5C). The gross phenotype of lmo2-mut embryos under brightfield microscopy appeared normal and exhibits mild features of body bending and reduced skin pigmentation (SI Appendix, Fig. S9 A and B). These features are similar to those shown previously. As previously reported, we also found that lmo2 mutants had fewer circulating red blood cells, consistent with reduced Gata1 gene expression (42) and impaired hematopoiesis (43), mild pericardial edema, (43) and a mild cephalomegaly (42). Immunostaining for Fli1 and Kdr showed a significantly reduced length of ISV in lmo2-mut embryos compared with control (Fig. 5 D and E), confirming the role of lmo2 in developmental angiogenesis. In contrast, vasculogenesis was only slightly affected, consistent with a previous study in mice (32).

Fig. 5.

Lmo2 impacts zebrafish angiogenesis by regulating prdm16. (A) lmo2-mut in zebrafish by CRISPR/Cas9. gRNAs produced a 4-nucleotide insertion in the exon 2 and a stop codon downstream. (B) Western blot using an anti-Lmo2 antibody confirmed the absence of Lmo2. β-Tubulin was used as loading control. (C) Kaplan-Meyer survival curve of lmo2-mut zebrafish. (D) Brightfield images (Upper) showing the lmo2-mut embryo gross phenotype. Fluorescent images of the trunk region following immunostaining for Fli1 (Middle) and Kdrl (Lower), showing the reduced average length of ISVs in lmo2-mut and the rescue of vascular defects when prdm16 mRNA was injected. These data are graphically shown in E. (F) qPCR analysis showing the expression of PRDM16 in CD31⁺ and CD31⁻ cell isolated from lmo2-mut embryos. (G) ChIP–PCR showing the association of Lmo2 to the prdm16 gene; gel electrophoresis of PCR products. The prdm16 promoter region, but not a control region ∼1-kb downstream, was precipitated by the anti-Lmo2 antibody. Sheared chromatin before immunoprecipitation and immunoprecipitation using spermine served as positive and negative controls, respectively. (H) Graphical abstract showing the relationship between Lmo2–prdm16 and the role of prdm16 on angiogenesis. n = 3 experiments; ANOVA test followed by Bonferroni post hoc was used to compare means. NS, not significant. ***P ≤ 0.001, *P ≤ 0.05.

As Lmo2 and Prdm16 similarly affect angiogenesis, we investigated the relation between the two genes. To test the hypothesis that Lmo2 regulates prdm16, we first assessed prdm16 gene expression in lmo2-mut embryos. We observed that prdm16 expression was significantly reduced from 48 to 120 h postfertilization compared with control (SI Appendix, Fig. S10). Interestingly, there is a gradual increase of prdm16 gene expression in LMO2-mut embryos, suggesting that other factors could be implicated in the regulation of prdm16. Therefore, we tested the ability of Prdm16 to rescue the vascular defects of lmo2-mut. We showed that injection of prdm16 mRNA in lmo2-mut resulted in ISV of similar length to controls (Fig. 5 D and E). Embryos control injected with prdm16 mRNA alone did not show any difference in ISV phenotype compared with control, showing that the rescue of the ISV phenotype in lmo2-mut embryos was a clean rescue experiment (Fig. 5 D and E). To test whether the reduced expression of prdm16 in lmo2-mut was global or more specific to ECs, we performed qPCR for prdm16 in CD31⁺ and CD31⁻ cells FACS purified from lmo2-mut and control. We found that the reduction of prdm16 gene expression was more significant in CD31⁺ cells compared with the reduction observed in CD31⁻ cells, demonstrating EC-specific effects of Lmo2 for prdm16 (Fig. 5F). To further investigate whether Lmo2 associates with the promoter region of the prdm16 gene, a chromatin immunoprecipitation (ChIP) assay followed by PCR was performed in zebrafish embryos using an Lmo2-specific antibody. PCR identified a DNA fragment from the prdm16 promoter region, which was coprecipitated by the Lmo2 antibody but not by preimmune serum (Fig. 5G), indicating an association of Lmo2 with the prdm16 promoter. As a control, a region located 1 kb downstream of the Lmo2-binding site could not be efficiently coprecipitated. Taken together, these data demonstrate that Lmo2 associates with and activates transcription of prdm16.

In this report, we took advantage of the Tg(fli1:EGFP)^y1 zebrafish to discover epigenetic factors implicated in endothelial development and differentiation (Fig. 5H). Among many candidates, we focused on PRDM16. We demonstrated that PRDM16 is essential for angiogenesis and that its expression is mediated by LMO2 specifically in ECs by LMO2 association with the PRDM16 promoter region. We showed that PRDM16 is involved in the process of differentiation and maturation of EC as shown in vivo in the zebrafish and in vitro during differentiation of iPSC-derived EC. This work highlights a mechanism by which PRDM16 could promote endothelial lineage by LMO2-mediated regulation during endothelial development and points toward a potentially therapeutic target for endothelial dysfunction in a wide range of vascular disease settings.

Methods

Ethics Statement.

All experiments with zebrafish were performed in accordance with the recommendations of the Institutional Animal Care and Use Committee at the Houston Methodist Research Institute and with the United Kingdom Animals (Scientific Procedures) Act 1986 at the Queens Medical Research Institute (44).

Zebrafish Aquaculture and Husbandry.

Zebrafish Wik, Tg(fli1:EGFP)^y1, and Tg(kdrl:mCherry)^y171 strains were maintained according to standard procedures. Fish were kept at 28 °C under a 14/10-h light/dark cycle and fed with dry meal (Gemma Micro) twice per day. Embryos were obtained by natural mating and kept in E3 embryo medium at 28.5 °C. All the experimental procedures were performed under anesthesia with Tricaine 0.02 mg/mL.

Maintenance of Human iPSC.

The human iPSC lines were obtained from Coriell Cell Repositories and were maintained on Matrigel (BD Biosciences, catalog number 354277) coated plates (Corning) in mTeSR1 medium (STEMCELL Technologies catalog number 85850) according to their protocol. The iPSCs were passaged approximately every 4 d, and RELSR (STEMCELL Technologies, catalog number 05873) dissociation reagent was used to detach colonies. Cells were maintained in humidified incubators at 37 °C and 5% CO₂. Pluripotency of iPSC was periodically characterized by morphology and immunostaining of pluripotency markers.

Differentiation of iPSC to ECs.

EC differentiation was carried out following an established protocol (45), with some modifications. In brief, the iPSCs at passage between 21 to 23 were grown to 80% confluence and placed in differentiation medium Advanced Dulbecco's modified Eagle's medium (DMEM)/F12 (ThermoFisher, catalog number 11320033), supplemented with Wnt agonist CHIR 99021 5 µM (Selleck, catalog number S2924), bone morphogenetic protein-4 (25 ng/m) (Peprotech, catalog number 120-05), B27 supplement (ThermoFisher, catalog number 17504044), and N2 supplement (ThermoFisher, catalog number 17502048) (mesodermal differentiation). After 3 d, cells were dissociated with HyQtase (GE Healthcare, catalog number SV30030.01) and plated in StemPro media (ThermoFisher, catalog number 10639011), supplemented with forskolin 5 µM (LC Laboratories catalog number F-9929), vascular endothelial growth factor (VEGF) 50 ng/mL (Peprotech, catalog number 100-20), and polyvinyl alcohol 2 mg/mL (Sigma-Aldrich, catalog number 360627) (endothelial differentiation). After 4 d, cells were washed twice with phosphate-buffered saline (PBS) and cultured in endothelial growth media (EGM-2MV, Lonza, catalog number CC-3202), supplemented with additional VEGF (100 ng/mL) for 4 more days (endothelial maturation). Cells were passaged once they reached 80 to 90% confluence. During the differentiation protocol, cells were maintained at 37 °C and 5% CO₂ in a humidified incubator.

PRDM16 KD in Differentiating ECs.

At day 3 of the mesodermal differentiation stage, 6 h before replacing the medium, cells were transfected with 5 nM siRNA targeting PRDM16 or control (Silencer Select siRNA, ThermoFisher) using Lipofectamine RNAiMax (Life Technologies, catalog number 13778-075) in Opti-MEM (Life Technologies, catalog number 31985-062). After 6 h, mesodermal medium was aspirated and replaced with endothelial differentiation medium. Effective KD was analyzed by qPCR and Western blot (anti-PRDM16 antibody, human polyclonal, catalog number PA5-20872, Thermo Fisher Scientific).

Flow Cytometry Characterization and Purification of iPSC-Derived EC.

At the end of the protocol (see Fig. 4A), cells were dissociated into single cells with HyQtase for 5 min at 37 °C, washed for 5 min with PBS containing 5% bovine serum albumin, and then passed through a 40-μm cell strainer. Cells were then incubated with either CD31 mAb (PE mouse anti-human, BD Pharmingen, catalog number 555446) or CD144 mAb (FITC mouse anti-human, BD Pharmingen, catalog number 560411) for 30 min. Isotype-matched antibody served as negative control. Cells were sorted using FACSAria (BD) flow cytometer, and data were analyzed by Flowjo software.

Proliferation Assay.

Click-IT EdU kit (Thermo Fisher Scientific) was used according to the manufacturer’s instructions to assess proliferation in zebrafish larvae and in iPSC-derived ECs.

Apoptosis Assay.

Apoptosis was detected by TUNEL assay using the ApopTag rhodamine In Situ Apoptosis Detection kit (Chemicon) following the manufacturer’s instructions.

For experiments in whole-mount zebrafish, embryos were fixed in 4% paraformaldehyde (PFA) at 4 °C, washed in PBS twice for 5 min, and permeabilized with proteinase K (10 μg/mL) for 20 min at room temperature, followed by two washes in PBS. Then, they were placed in prechilled ethanol:acetic acid (2:1) at −20 °C for 10 min and washed in PBS 1×–0.1% Tween-20 twice before incubation in equilibration buffer and further steps as described in the manufacturer’s protocol. TUNEL assay staining was quantified by counting positive staining puncta (TUNEL⁺ nuclei) in the vessel of Tg(fli1:EGFP)^y1 in z-stack confocal images using ImageJ. For in vitro experiments, cultured iPSC cells during the endothelial differentiation and maturation stages were treated as above for TUNEL and then immunostained for CD144.

Cell Migration Assay.

Confluent ECs were cultured on 12-well dishes. Cells were wounded by scratching with a micropipette tip, rinsed with PBS, and then incubated for 16 h. Wound closure was monitored through the use of digital photography and measured using the ImageJ program. Cell migration was expressed as the migrated distance.

Enzymatic Isolation of Cells from Zebrafish Embryos.

Cells were isolated according to Shestopalov et al. (46). In brief, Tg(fli1:EGFP)^y1 embryos at the appropriate developmental stage were dechorionated, euthanized, and washed three times with sterile PBS. Embryos were placed in a 1.5-mL tube, washed with calcium-free Ringer solution (200 μL for 30 embryos; 116 mM NaCl, 2.6 mM KCl, and 5 mM Hepes, pH 7.0), and replaced with 1 mL solution of 1× PBS containing trypsin (0.25%, Gibco), 50 μg collagenase P (Roche), and 1 mM EDTA. Embryos were disaggregated using a 200-μL pipette tip and incubated for 30 min at 28.5 °C, with further pipetting every 5 min. Digestion was quenched with stop solution (200 μL; 1× PBS containing 30% [volume/volume] calf serum and 10 mM CaCl2), and samples were centrifuged at 400 × g for 5 min at 4 °C. Supernatants were discarded, and cell pellets were washed twice with chilled solution of DMEM containing 1% (volume/volume) calf serum, 0.8 mM CaCl2, and 50 U/mL⁻¹ penicillin/streptomycin, centrifuged, and resuspended in the same medium. Cell suspensions were filtered through a 40-μm cell strainer (BD Biosciences) into FACS tubes.

Flow Cytometry Characterization and Purification of fli1⁺ Cells from Zebrafish.

Cell suspensions were analyzed using a BD FACS Aria (BD Biosciences). 4′,6-diamidino-2-phenylindole (DAPI) was used to identify viable single cells, whereas FSC-H and FSC-A were used to select cell singlets. Wild-type (Wik) zebrafish were used to set the gate between GFP⁻ (i.e., fli1⁻) and GFP⁺ (low-fli1⁺) cell populations, whereas the gate between low- and high-fli1⁺ populations was placed at the onset of the high-fli1⁺ cells. At least 50,000 of fli1⁻, low-, and high-fli1⁺ cells (excitation [Ex]: 488 nm; emission [Em]: 530 nm) were sorted from groups of n = 50 Tg(fli1:EGFP)^y1 embryos into a 15-mL falcon tube containing chilled PBS and 10% fetal bovine serum. The number of viable cells was confirmed under fluorescence stereomicroscope (Leica M205) by using a Neubauer chamber.

Culture of Zebrafish Cells.

Low- and high-fli1⁺cells were cultured on Petri dishes coated with fibronectin at a concentration of 2 µg/cm² in endothelial basal medium supplemented with EGM2 bullet kits (Lonza). The medium was replaced every 2 d.

Dil-Ac-LDL Uptake Assay.

Uptake of Ac-LDL was assessed by incubating cells with ac-LDL-594 (Thermo Fisher Scientific) at 1:200 dilutions for 3 h. Then, cells were washed with PBS, and the fluorescence was measured in at least n = 5 high-power fields using ImageJ and plotted as mean of the fluorescent signal (integrated density) in each cell.

Nitric Oxide Assay.

The nitric oxide (NO) metabolites nitrite and nitrate were detected in the culture medium as an indirect measurement of cellular NO (Cayman Chemical, Nitrate/Nitrite Assay Kit). In brief, cell culture medium was first added with nitrate reductase that converts nitrate to nitrite. Then, Griess reagents added to the sample convert nitrite into a deep purple azo compound. Photometric measurement of the absorbance (540-nm wavelength) derived from this azo-chromophore accurately determines nitrite concentration. Cellular nitrate/nitrite production is quantitated by subtracting the level of nitrate/nitrite present in the media alone from the total nitrate/nitrite level present in the medium during cell growth.

Immunocytochemistry.

Embryos were euthanized in tricaine and enzymatically dissociated as above, fixed in 2% PFA for 10 min, washed in 2× PBS, blocked in PBS-TritonX 0.1% for 30 min, stained with anti-CD31 (PECAM-1) Monoclonal Antibody FITC-conjugated (Thermo Fisher Scientific, catalog number 11-0311-82, rat 1:200) for 1 h, washed in 2× PBS, and FACS analyzed.

Immunohistochemistry.

Zebrafish embryos were euthanized in Tricaine 1 mM and fixed in 4% PFA (Sigma) at 4 °C overnight. Embryos were permeabilized using proteinase K (10 μg/mL), fixed again in PFA 4% for 30 min, washed three times in PBS-TritonX 100 (0.1%), and blocked in bovine serum albumin 5% in PBS for 3 h. Then, embryos were incubated with anti-FLI1 antibody (Sigma, catalog number SAB2100822, rabbit, dilution 1:100) or anti-Kdrl antibody, followed by incubation with anti-rabbit antibody (Alexa Fluor 555, Cell Signaling, 1:500). Then, specimens were washed in PBS and mounted in glycerol 100%. Blood vessel formation and ECs were assessed in whole-mount embryos.

Imaging.

Fluorescence images of zebrafish were acquired using a Leica M205FA stereomicroscope equipped with a mercury lamp with filter sets (Ex: 470 nm, Em: 525 nm for GFP; Ex: 555 nm; Em: 565 nm for Alexa Fluor 555) and with a Leica DFC500 digital camera and a confocal microscope (Leica SP5) to capture high resolution z-stack images. The same microscope was also used to capture brightfield images of embryos. Images of isolated cells were acquired using an EVOS FL imaging systems AMF4300 (Thermo Fisher Scientific).

Preparation of RNA-seq Libraries and Sequencing.

fli1⁻, low-fli1⁺, and hig-fli1⁺ from at least n = 50 Tg(fli1:EGFP)^y1 embryos per group were harvested by FACS at 72 h postfertilization and prepared for analysis by RNA-seq. Total RNA from cells was isolated, fragmented, reverse transcribed to complementary DNA (cDNA), ligated to adapters, and subjected to brief PCR amplification in preparation of the Illumina library. The integrity and quality of RNA and complementary DNA were monitored using an Agilent Bioanalyzer 2100. The samples were run on an Illumina Hi-SEq 2500 system with 100 base paired-end sequencing (50 million reads per sample). Samples were run in duplicate.

Bioinformatic Analysis.

RNA-seq reads were aligned to the zebrafish genome danRer10. We used the full set of known genes downloaded from the University of California Santa Cruz (UCSC) Genome browser as reference genes. RNA-seq read counts for each gene in each sample were calculated using Cuffdiff function in Cufflinks version 2.2.1. The Cuffdiff also calculates fragment per kilobase per million reads for each gene. We further subjected the read counts to EdgeR version 3.12.0 for differential expression analysis and defined differential genes based on FDR cutoff 1e-5. We subjected interesting gene groups to the DAVID website (https://david.ncifcrf.gov) for functional enrichment analysis. Enriched functional terms were defined based on a Benjamini adjusted P value cutoff of 0.05. Genomic tracks were generated by UCSC Genome browser. The EpiFactors gene list was downloaded from https://epifactors.autosome.ru/.

ChIP–PCR Assay.

ChIP assay was performed following the manufacturer’s instructions (Cell Signaling Technology). Briefly, 50 to 70 embryos per group were disaggregated in single cells as described above. DNA and protein were cross-linked by 1% formaldehyde. Chromatin was isolated and digested with Micrococcal Nuclease. Then, the DNA–protein complex was precipitated with control IgG or antibody against LMO2 (rabbit polyclonal, ChIP grade, Abcam, AB72841) overnight at 4 °C and protein A/G conjugated magnetic beads for 1 h. Cross-links were reversed. The extracted DNA was used as a template for PCR amplification of the targeted promoter region. The extracted DNA from unprecipitated DNA–protein complex was used as input. The promoter region of prdm16 (National Center for Biotechnology Information [NCBI] reference sequence: XM_021478491.1) was identified with FINDM software. Primers sequences used were (Forward) 5′-GCA​GAG​TGC​GAC​GGT​AAA-3′ and (Reverse) 5′-CGT​CCA​GAC​AGA​ACT​TCA​CAT-3′ to detect prdm16 promoter and (Forward) 5′-CAC​TTC​TCA​AGA​GCC​CAC​TTA​AT-3′ and (Reverse) 5′-CTG-​CTG​AGA​CTA​CTC​CCT​ATG​T-3′ for control sequence.

Generation of Zebrafish Mutants using CRISPR/Cas9.

Zebrafish mutant lines for lmo2 were generated via CRISPR/Cas9-based mutagenesis as previously described (47). In brief, gRNA specific for target sites on lmo2 gene sequence were identified using CHOPCHOP (https://chopchop.cbu.uib.no). The gRNA sequence was 5′-GCT​GAT​CTG​CAG​GGA​GCC​GG-3′ and was prepared as previously described (48). gRNAs were then coinjected with 600 ng/μL of Cas9 protein (PNA bio) and 200 mM KCl. Cas9/single gRNA complexes were formed by incubating Cas9 protein with gRNA at room temperature for 5 min prior to injecting into the cytoplasm of wild-type AB zebrafish embryos at the one-cell stage. For detecting the zebrafish mutants, genomic DNA was extracted from individual zebrafish larva using the Quick-DNA isolation kit (Zymogen), and a short genomic region (200 to 400 bp) flanking the target site was amplified by PCR. For lmo2, the primers were (Forward) 5′-GCA​CAT​GTT​TGC​CTG​TAT​TTG​T-3′ and (Reverse) 5′-CAG​AGG​TCA​CAG-​CTC​AGA​CAG​T-3′. Purified PCR products (200 ng) were denatured, reannealed, and then digested with EnGen Mutation Detection kit (New England Biolabs), which uses T7 Endonuclease I. The samples were run out on agarose gel 2% to distinguish mutant from wild-type embryos. PCR products of positive mutants were subcloned into pGEM-T vectors (Promega) that were then used to transform competent cells. After overnight culture at 37 °C, a single colony was selected for sequencing.

Prdm16 KD.

KD of prdm16 gene (NCBI: DQ851827.1) in zebrafish embryos was achieved by the injection of two antisense nonoverlapping Mo (Gene Tools) targeting the translational start site: 5′-CCT​CGC​CTT​GGA​TCT​CAT​CTT-​GTC-3′ and 5′-TTG​TAG​ATT​CCT​CGC​GTC​CTC​CTT​G-3′. A mismatch was used as control: 5′-CgT​CcC​CTT​cGA​TCT​CAT​gTT​cTC-3′. Using a standard microinjector (IM300 Microinjector; Narishige), an optimized dose of 0.5 ng (0.5 nL bolus) Mo placed in a pulled glass capillary was injected in each embryo at one- to two-cell stages, just beneath the blastoderm.

In Vitro Transcription of prdm16 mRNA.

Prdm16 mRNA, with 7-methyl guanosine cap structure at the 5′ end and Poli(A) tail at the 3′ end, were transcribed using the HiScribe T7 ARCA mRNA Kit (New England Biolabs, catalog number E2060) following the manufacturer’s instructions.

Rescue Experiments by Injection of prdm16 mRNA.

To determine whether the effects of the prdm16-targeted Mo were specifically due to the loss of the target gene, we performed rescue experiments by coinjecting prdm16-Mo with prdm16 modified (m)mRNA wild type. A bolus of 1 nL solution containing 0.5 ng prdm16-Mo and 0.5 ng prdm16 RNA wild type was injected into each egg. As a negative control for the rescue experiment, 0.5 ng mutant RNA (120 base pair; Ultramer RNA, IDT Technologies) lacking the 5′-UTR region and that does not encode any functional protein was coinjected with the Mo. Furthermore, a bolus of 1 nL solution containing 0.5 ng prdm16 mRNA was also used to rescue the phenotype in lmo2 mutants.

Defining the Zebrafish Embryo Phenotype.

Whole embryo phenotype in lmo2-mut embryos and the following prdm16 Mo and rescue experiments were described on the basis of morphologic features observed under bright-field microscopy: reduced body length, curved body, reduced swimming, chorionated larvae, and edema. The phenotype was assessed using a simple 6-point scoring system according to the severity of that feature and where one point was normal. At least four different clutches of larvae were assessed per experimental group. The percentage of embryos showing the phenotype (penetrance) was recorded.

Kaplan-Meier Analysis of Survival.

Kaplan-Meier analysis was used to measure the survival of adult zebrafish or larvae following each defined treatment, using PRISM 7 software.

RNA Extraction and qPCR.

mRNA was extracted from embryos using column purification (RNeasy Mini Kit, Qiagen, catalog number 74104) according to the manufacturer’s instructions. Working surfaces were cleaned with RNaseZap (Life Technologies) to deactivate environmental RNase. Efficient disruption and homogenization of tissue was done using sterile RNase-free disposable pestles (Fisher Scientific, catalog number 12-141-368) mounted on a cordless motor for 30 s and then passing the lysate 5 to 10 times through the needle (18 to 21 gauge) mounted on a RNase-free syringe. RNA integrity was assessed on the basis of 18S and 28S ribosomal RNA bands. mRNA was reverse transcribed into cDNA using qScript cDNA Synthesis Kit (Quanta Bio, catalog number 95047.) Primers (IDT Technologies) targeting all genes of interest (see SI Appendix, Table S1 for the full list of primers) and an SYBR Green PCR kit (Invitrogen) were used for real-time qPCR, which was performed with the QuantStudio 12 k Flex system (Applied Biosystems) following the manufacturer’s instructions. Gene expression was expressed as relative fold changes using the ΔCt method of analysis and normalized to α-actin.

Extraction of Proteins.

Zebrafish embryos were euthanized with an overdose of tricaine and then washed three times in PBS and homogenized with a pestle (Sigma, catalog number Z359971) in 100 µL radioimmunoprecipitation assay (RIPA) buffer (25 mmol/L Tris⋅HCl pH 7.6, 150 mmol/L NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, and 0.1% SDS) supplemented with protease/phosphatase inhibitor mixture. The lysate was kept on ice for 40 min. Then, the tube was centrifuged at 3,000 × g for 5 min, and the supernatant was transferred into a clean prechilled tube. Bicinchoninic Acid Protein Assay (Thermo Scientific, catalog number 23225) was used to measure protein concentration.

Western Blot.

Samples were loaded on polyacrylamide gel electrophoresis (4 to 15% gradient) for 2 h and transferred on polyvinylidene fluoride membranes (PVDF) for 2 h. Membranes were blocked with nonfat milk 5% in PBS + 0.1% Tween for 1 h at room temperature and probed with primary antibody overnight at 4 °C. The antibodies used were anti-PRDM16 rabbit polyclonal (ProSci, catalog number 5555) anti-LMO2 rabbit polyclonal (Abcam, catalog number AB72841), and anti–β-tubulin (loading control) rabbit polyclonal (Abcam, catalog number Ab6046). Membranes were washed three times (5 min per wash) with PBS and incubated with horseradish peroxidase–conjugated (HRP) goat anti-mouse or abbit antibodies for 1 h at room temperature (RT). Membranes were washed three times with PBS for 5 min. Antigen antibody complexes were then detected by exposure for 5 min to the enhanced chemiluminescence solution (Amersham). Then, the membrane was placed down on a film layer and exposed to photographic film (BioMax XAR Film Kodak, Sigma-Aldrich). The film was developed, and immunoreactivity (band density) was quantified by using densitometry (source: https://rsbweb.nih.gov/ij/docs/user-guide.pdf) with ImageJ.

Statistical Analysis.

Results were expressed as the mean ± SEM. Each experiment was performed three times. The Shapiro-Wilk test was used to confirm the null hypothesis that the data follow a normal distribution. Statistical comparisons between two groups or multiple groups were then performed, respectively, via Student’s t test or ANOVA test using PRISM 7 software followed by Bonferroni post hoc test. P < 0.05 was considered significant.

Data Availability

All RNA-seq data have been deposited in Gene Expression Omnibus (GSE149152) and are freely accessible.