Endothelial Cell Direct Reprogramming: past, present, and future

Abstract

Ischemic cardiovascular disease still remains as a leading cause of morbidity and mortality despite various medical, surgical, and interventional therapy. As such, cell therapy has emerged as an attractive option because it tackles underlying problem of the diseases by inducing neovascularization in ischemic tissue. After overall failure of adult stem or progenitor cells, studies attempted to generate endothelial cells (ECs) from pluripotent stem cells (PSCs). While endothelial cells (ECs) differentiated from PSCs successfully induced vascular regeneration, differentiating volatility and tumorigenic potential is a concern for their clinical applications. Alternatively, direct reprogramming strategies employ lineage-specific factors to change cell fate without achieving pluripotency. ECs have been successfully reprogrammed via ectopic expression of transcription factors (TFs) from endothelial lineage. The reprogrammed ECs induced neovascularization in vitro and in vivo and thus demonstrated their therapeutic value in animal models of vascular insufficiency. Methods of delivering reprogramming factors include lentiviral or retroviral vectors and more clinically relevant, non-integrative adenoviral and episomal vectors. Most studies made use of fibroblast as a source cell for reprogramming, but reprogrammability of other clinically relevant source cell types has to be evaluated. Specific mechanisms and small molecules that are involved in the aforementioned processes tackles challenges associated with direct reprogramming efficiency and maintenance of reprogrammed EC characteristics. After all, this review provides summary of past and contemporary methods of direct endothelial reprogramming and discusses the future direction to overcome these challenges to acquire clinically applicable reprogrammed ECs.

1. Introduction

Ischemic cardiovascular diseases are the leading causes of morbidity and mortality [1]. Treatment options for chronic, severe cases of ischemia such as myocardial infarction and critical limb ischemia are still limited. Despite significant improvement in surgical, interventional, and medical therapeutics, patients with advanced cases have poor prognoses even after exhaustive treatment with conventional therapeutics leaving them no other option but cardiac transplantation or lower extremity amputation [2–8]. Therefore, novel approaches have been demanded for restoring proper blood perfusion and tissue repair.

Underlying burden of the coronary and peripheral artery diseases are loss of blood vessels and inability to restore vessels with endogenous mechanisms. Cell therapy has been considered an attractive strategy as it supplies vascular components, especially endothelial cells (ECs), for generation of new functional blood vessels (neovascularization). Early attempts of cell therapy was done with various types of adult stem or progenitor cells. Mesenchymal stem cells, mononuclear cells or endothelial progenitor cells derived from bone marrow or peripheral blood have shown to secrete proangiogenic and cytoprotective factors but are modestly effective, lacking evidence of major transdifferentiation of these cell types into ECs [9–12]. Therefore, researchers have sought to generate ECs which can induce neovascularization with different cell sources.

Pluripotent stem cells (PSCs) including embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) emerged as a method of generating ECs. Particularly, iPSCs have been considered as a more practical candidate over ESCs as they are free from ethical concerns and can be generated from autologous sources. Ever since Yamanaka and colleagues first reported generation of iPSCs with four transcription factors (TFs), differentiation techniques for generating ECs have significantly improved [13–16]. Multiple studies showed differentiation of human iPSCs (hiPSCs) into ECs and these hiPSC-ECs demonstrated excellent therapeutic potential and neovascularization capacity in ischemic murine models [17, 18]. However, complicated differentiation steps and difficulties in phenotype maintenance have delayed clinical translation of iPSC-derived cells [19]. More importantly, residual, undifferentiated or unwantedly differentiated iPSCs which may exist within the differentiated cell cultures may have tumorigenic potential and aberrant tissue formation [20].

To avoid issues associated with PSCs, novel methods of generating target cell types using lineage-specific TFs or small molecules have been developed. These methods directly convert various source cells to target cell types while bypassing the pluripotent states [21]. This novel approach referred to as “direct reprogramming” or “transdifferentiation” provides significant advantages as it provides time- and cost-effective ways of generating target cells while reducing potential risks and inefficiency associated with pluripotent stem cells.

In this review, we dissect methods of directly reprogramming ECs and characteristics of the reprogrammed or induced ECs. We further discuss therapeutic potential, challenges, and future directions of EC reprogramming.

2. Pathway to direct endothelial reprogramming

2.1. Partial activation of pluripotency

To overcome the problems of PSCs, alternative methods of generating ECs while avoiding or being minimally affected by the pluripotent state have been developed. In 2012, Margariti et al. transiently induced the expression of iPSC-inducing factors OCT4, SOX2, KLF4, and c-MYC (OSKM) to convert human fibroblasts into a dedifferentiated state referred to as partial-iPSCs (PiPSCs) [22]. Four-days of induction of OSKM was sufficient to alter plasticity of the fibroblasts, and subsequent cell culture under endothelial cell growth media-2 (EGM-2) differentiated PiPSCs into ECs (PiPSC-ECs). Unlike iPSCs, tumors were undetected 2 months after PiPSCs were injected into mice with Matrigel. Similarly, Kurian et al. generated a plastic intermediate state from human fetal and adult fibroblasts by introducing miR302 and miR307 with either 4 factors (OSKM) or 6 factors (OCT4, SOX2, KLF4, LMYC, LIN28, p53 short hairpin RNA) for 8 days [23]. These plastic cells were cultured under mesodermal induction medium for 8 days to generate CD34⁺ angioblast-like progenitor cells. Another 8-day culture under EGM-2 culture medium differentiated the progenitor cells into ECs. Notably, pluripotency markers TRA-1–60 and TRA-1–81 were undetectable even after plastic state induction.

Interestingly, Li et al. only used two pluripotent TFs, OCT4 and KLF4, and chemically defined media to generate induced ECs (iECs) [24]. After lentiviral transduction of these two TFs, the cells were cultured with bone morphogenetic protein 4 (BMP4) for first 7 days, cAMP-dependent protein kinase A (8-Br-cAMP) for the next 7 days, and TGF-beta inhibitor (SB43152) for another 14 days. They noted the pluripotency network was minimally affected by the gradual transdifferentiation steps as suggested from the analysis of transcriptomic and epigenetic changes. Taken together, partial or transient activation of pluripotency reduces risks of tumorigenesis as it bypasses the fully pluripotent state while expediting reprogramming procedure for EC generation. However, achieving dedifferentiated or a partial pluripotent state still requires overexpression of pluripotency reprogramming factors, and therefore, is not completely free from the risks posed by PSCs.

2.2. Direct endothelial reprogramming

Inducing cellular reprogramming toward PSCs by overexpression of pluripotency TFs hinted the possibility of direct cellular reprogramming or fate conversion toward ECs by lineage-specific factors. Thus far, ~15 studies were published in endothelial cell generation via direct reprogramming, and we summarized them in Figure 1 and Table 1.

Figure 1.

Methods of direct endothelial reprogramming with various source cell types. Fibroblasts, human adipose-derived stem cells (hADSCs), mesenchymal stem cells (hMSCs), and human amniotic fluid-derived cells (hACs) were reprogrammed into endothelial cells (ECs). A. The diagram showing origin of the source cells. B. The graphic showing source cell types, reprogramming methods, and target cell. Alphabets from A-N indicate studies as referred from Table 1.

Table 1.

Summary of direct endothelial reprogramming methods

#	Reprogrammed cell name	Source cell type	Key factors		Delivery method	Culture condition	Culture duration	Reprogramming efficiency	In vivo animal model (follow up)	Reference
			Genes	Small molecules
A	rAC-VEC	Human amniotic fluid-derived cell	ETV2, FLI1, and ERG1	VEGF and SB431542	Lentiviral infection	EM: SB431542, EC supplement, and Heparin	28 days	D14: 39.1% CDH5+, 8.0% KDR+ D21: 86.5% CDH5+, 48.5% PECAM1+	Matrigel Plug (2 weeks) Liver engraftment assay (3 months)	[26]
B	iEC	Mouse skin and tail-tip fibroblast	Etv2, Foxo1, Klf2, Tal1 and Lmo2		Lentiviral infection	EBM-2	12 days	D12: 4% Tie2-GFP+, 2.3% PECAM1+, 4.2% CDH5+, 2.6% KDR+, 7.3% vWF+	HLI model LDPI & IHC (2 weeks)	[25]
C	iEC*	Human neonatal dermal fibroblast		(Poly I:C) bFGF, VEGF, BMP4, 8-Br- cAMP, SB43152	Chemical stimulation	Activation of innate immunity: Poly I:C Transdifferentiation medium I: bFGF, VEGF, and BMP4 Transdifferentiation medium II: EGM-2, bFGF, VEGF, BMP4, and 8-Br-cAMP After sorting: EGM-2 and SB431542	7 days + 7 days + 14 days and more	D14: 2% PECAM1+	Matrigel Plug (2 Weeks) HLI Model 1 st injection immediately after HLI surgery 2nd injection 10 days after HLI surgery LDPI (14 days) & IHC (18 days)	[28]
D	ETVEC	Human adult fibroblast	ETV2	VEGF and bFGF	Lentiviral infection	EGM-2, VEGF, and bFGF	25 days (beyond 50 days)	D14: 40% KDR+ D15: 3.4% CDH5+ & PECAM1+ D25: 95.3% PECAM1+	Matrigel Plug (2 weeks and 6 weeks) HLI model (2 weeks)	[27]
E	iEC	Human neonatal fibroblast	ETV2, FLI1, GATA2, and KLF4	BMP4, VEGF, bFGF, and SB431542	Lentiviral infection	Differentiation medium: BMP4, VEGF, and bFGF EC growth medium: EGM-2 MV and SB431542	3 days + 25 days	D14: 16% PECAM1+	N/A	[29]
F	rEC (early vs. late)	Human dermal fibroblast	ETV2	VEGFA and VPA	Lentiviral infection	Early rEC: EGM-2 and VEGFA Late rEC: EGM-2, temporal treatment of VPA, and VEGFA	7 days for early rEC and 3 months for late rEC	D7: 50% CDH5+, 39% KDR+ D93: 60% CDH5+, 83% PECAM1 +	HLI model LDPI (4 weeks) IHC: Early rEC (4 weeks) and Late rEC (3 months)	[30]
G	EiEC	Human adipose- derived stem cell and human umbilical mesenchymal stem cells	ETV2	SB431542, CHIR99021, VEGF, bFGF, EGF, and BMP4	Lentiviral infection	EIM: Insulin, ascorbic acid, Heparin, VEGF, bFGF, EGF, SB431542, CHIR99021 and BMP4 EMM: SB431542, VEGF, bFGF, and EGF	10 days(up to 2 months)	D10: 47.2% KDR+ D30: 73.8% PECAM1+, 60.6% KDR+, 88.1% CDH5+	HLI Model (2 weeks) Injected with 30% Matrigel	[35]
H	iEC	Human embryonic lung fibroblast	DKK3	VEGF	Adenoviral infection	EGM-2 and VEGF	10 days or more	N/A	Matrigel plug (1 week)	[31]
I	Fsk-iEC	Human fibroblast and UCB-MSC	ETV2	forskolin	Lentiviral or Retroviral infection	EGM-2 and Forskolin	14 days	D14: 55.9% CDH5+, 43.9% PECAM1+	Matrigel plug (1 week) HLI model (2 weeks)	[33]
J	iVEC	Human dermal fibroblast	ETV2		Retroviral infection	EGM-2 MV	N/A	N/A	N/A	[32]
K	rCVT	Mouse tail-tip fibroblast	miR-208b- 3p	ascorbic acid and BMP4	Lipid-based transfection	DMEM/F-12, 10% FBS, ascorbic acid, and BMP4	10 days	D10: 28% PECAM1+	MI model Masson’s trichrome staining and Echocardiography (12 weeks) IHC (16 weeks)	[46]
L	iEC	Human adult dermal fibroblast	ETV2, KLF2 and T AL1 with siTWIST 1	Rosiglitazone	Lentiviral infection	1st stage: EGM-2 MV and rosiglitazone 2nd stage: EGM-2 MV	4 weeks + 2 weeks	D28: 63% CDH5+, 9.5% PECAM1 + D42: 19.6% PECAM1+ and CDH5+	N/A	[34]
M	SCAP-EC*	Human MSC from apical papilla		VPA, CHIR99021, Repsox, Forskolin, Y- 27632	Chemical stimulation	1 st Induction Medium: EGM-2, VEGF and BMP4 2nd Induction Medium: EGM-2, VEGF, and 8- Br-3,5-cAMP	4 days + 4 days	D8: 42.88% KDR+, 39.20% TEK+, 14.10% PECAM1 +	Matrigel plug (2 weeks)	[37]
N	ETV2 overexpressing DPSC	Dental pulp stem cells	ETV2		Lentiviral infection	Differentiation media: EGM	14 days	D7: 69.9% CDH5 +	Matrigel plug (1 week)	[36]

DPSC = dental pulp stem cell; EBM-2 = Endothelial Cell Growth Basal Medium-2; EC = endothelial cell; EGM-2 MV = Microvascular Endothelial Cell Growth Medium-2; EiEC = ETV2-induced endothelial cell; EIM = endothelial induction medium; EM = endothelial growth media; EMM = endothelial maintenance medium; Fsk = forskolin; HLI = hindlimb ischemia model; iEC = induced endothelial cell; iEnd cell = induced endothelial cell; IHC=immunohistochemistry; iVEC = induced vascular endothelial cell; LDPI = laser doppler perfusion imaging; N/A = not applicable; rAC-VEC = reprogrammed amniotic fluid-derived cell-vascular endothelial cell; rCVT = reprogrammed cardiovascular tissue; rEC = reprogrammed endothelial cell; SCAP-EC = stem cell from apical papilia-derived chemical-induced endothelial cells; siTWIST = small interfering RNA of TWIST1; UCB-MSC = umbilical cord blood-derived mesenchymal stem cell; VEGF = vascular endothelial growth factor; VPA = valproic acid.

^*Chemically driven direct reprogramming towards ECs.

2.2.1. Source cells for endothelial reprogramming.

Therapeutic application of direct reprogramming starts with the selection of appropriate source cells. An adequate number of source cells should be acquired in non-invasive and cost-effective manners. In addition, the success of direct reprogramming greatly depends on the source cells. It is important to distinguish the reprogrammability of each source as reprogramming efficiency may differ for many aspects such as species, cell types, and tissue of origin. For example, Han et al. successfully reprogrammed ECs from murine fibroblasts using Foxo1, Etv2, Klf2, Tal1, and Lmo2, but they failed to generate ECs from human bone marrow mononuclear cells with the five factors [25]. Ginsberg et al. were the first to demonstrate direct EC reprogramming from human cells [26]. They suggested human ACs as an ideal source amenable for reprogramming. The cells are highly proliferative, can be human leukocyte antigens (HLA)-typed, and are cryopreservable for clinical use. However, human ACs are not suitable for autologous cell therapy, and their reprogramming methods were not applicable for human adult somatic cells. They acknowledged that ACs express ETV2 cofactors FOXC2, which drives endothelial gene expression with FOX:ETS elements. Though they showed ACs did not express pluripotency markers like OCT4, SOX2, and NANOG, midgestation ACs possibly have more plastic chromatin state permissible for reprogramming. To date, the majority of studies gave attention to fibroblasts as the possible source cells applicable for autologous cell therapy [27–34]. Cheng et al. suggested hADSC, an easily accessible cell type with multipotentiality, as a candidate somatic cell source for autologous cell therapy [35]. They successfully generated ECs from hADSCs via ETV2 overexpression. Interestingly, Li et al. and Yi et al. used a unique type of stem or progenitor cells derived from dental tissue in their reprogramming studies [36, 37]. Li et al. overexpressed ETV2 to generate ECs from dental pulp stem cells (DPSCs) [36]. Yi et al. used chemicals and small molecules to convert mesenchymal stem cells from apical papilla (SCAPs) into ECs [37].

2.2.2. Reprogramming strategy using transcription factors.

In early 2010, direct conversion of somatic cells to other cell types without activating pluripotency was demonstrated in generating neuronal cells from mouse fibroblasts [38]. Since then, numerous studies attempted to directly reprogram various cell types with unique sets of TFs and culture conditions [39–42]. In 2012, Ginsberg et al. pioneered a direct EC reprogramming strategy [26]. They reprogrammed endothelial cells from human amniotic fluid-derived cells (ACs) by overexpressing ETV2, FLI1, and ERG1. They suggested temporal regulation of ETV2 expression and TGFβ inhibition, increasing reprogramming efficiency and maturity of induced ECs. ETV2 was induced up to day 14 while FLI1 and ERG1 were continually over expressed. They treated SB431542, small molecule inhibitor for TGFβ, up to day 21. The reprogrammed cells, called rAC-VECs, displayed mature EC characteristics and formed tubular structures in vitro and in vivo Matrigel assays. Two years later, Han and colleagues used Foxo1, Etv2, Klf2, Tal1, and Lmo2 to directly reprogram mouse skin fibroblasts from Tie2-GFP reporter mice into induced ECs (iECs) [25]. Tie2-GFP positive iECs had similar in vivo characteristics as primary ECs isolated from the mouse lung. In another study, Wong and Cooke screened transcription factors and directly reprogrammed ECs from human dermal fibroblasts (HDFs) with ETV2, FLI1, GATA2, and KLF4 [29]. They noted ETV2 is the most essential factor for inducing PECAM1⁺ cell generation.

Human postnatal cell reprogramming toward ECs were first demonstrated by two studies (Lee et al. and Morita et al.), which showed that ETV2 alone is sufficient for reprogramming ECs from HDFs [27, 30]. Morita et al. demonstrated that ETV2 coordinates with endogenous FOXC2 to promote the FLI1 and ERG1 expression essential for downstream EC generation [27]. They noted that basic fibroblast growth factor (bFGF) and vascular endothelial growth factor (VEGF) supplements improve reprogramming efficiency. Though they suggested that transient expression of ETV2 is sufficient to maintain PECAM1⁺ population, the reprogrammed cells retained high ETV2 expression even after 50 days in culture. This phenotype could be a transient or transitional cell because ETV2 is minimally expressed in postnatal ECs, and sustained expression of ETV2 during embryogenesis leads to vascular abnormalities. Extreme overexpression of ETV2 forced expression of CDH5 and PECAM1, which are direct targets of ETV2. Thus, the selected population with ectopic expression of EC markers were named as ETVECs, which were failed to show the maturation process of ECs. Lee et al., on the other hand, emphasized that temporal expression of ETV2 is not only necessary for directly reprogramming toward ECs but is sufficient for maturation of reprogrammed EC by applying appropriate culture strategies [30]. They presented the reprogrammed cells in two different stages. After initial activation of ETV2, KDR⁺ cells were sorted at day 7. These sorted cells, referred to as early reprogrammed ECs (rECs), showed EC phenotypes while lacking some mature characteristics. However, these early rECs lost their EC phenotypes 30 days after continued culture. At day 30, transient re-introduction of ETV2 with angiogenic culture conditions with or without valproic acid (VPA) for another 2 months generated more mature ECs and was called late rECs.

A pivotal role of ETV2 for human endothelial reprogramming was demonstrated by many follow-up studies. Cheng et al. reported direct conversion of ECs from human adipose-derived stem cells (hADSCs) by transiently expressing ETV2 and inhibiting TGFβ signaling pathway [35]. More recently, Kim et al. explored molecular mechanisms underlying ETV2-mediated endothelial transdifferentiation via RNA-seq and ChIP-seq analyses. They noted cyclic AMP (cAMP), exchange protein directly activated by cAMP (EPAC) and RAP1 signaling pathway played essential roles in EC reprogramming, and treating forskolin, a cAMP signaling activator, effectively increased reprogramming efficiency [33]. Bersini et al. converted human fibroblasts from donors of different age groups and patients with Hutchinson-Gilford Progeria Syndrome (HGPS) into induced smooth muscle cells (iSMC) and induced vascular ECs (iVECs) by over expressing MYOCD or ETV2, respectively [32]. Han et al. used three TFs (ETV2, KLF2, TAL1) to generate iECs from HDFs [34]. They noted CDH5 single positive cells lacked EC functionality and retained source cell signatures. On the other hand, CDH5 and PECAM1 double positive cells formed capillary tubes in Matrigel, secreted nitrogen oxide, and clustered closer to human umbilical vein EC (HUVEC) based on the sequencing analysis. From the RNA-seq analysis, they identified TWIST1, an epithelial-mesenchymal transition (EMT) activator as a factor blocking reprogramming. Supplementing Rosiglitazone, an MET inhibitor, and knocking out TWIST1 with siRNA system improved double positive cell generation. They also pointed out an additional two- week culture after CDH5 sorting helps increase double positive iEC generation. On the other hand, another study demonstrated endothelial reprogramming without ETV2. Chen et al. used DKK3, a secreted glycoprotein, a member of the dickkopf family of Wnt inhibitors, to convert human embryonic lung fibroblasts into functional ECs. DKK3 gene expression level picked up 2 days after infection and returned back to baseline after day 7. Such transient activation of DKK3 was sufficient to induce KDR gene expression. They also demonstrated that DKK3 contributes to the mesenchymal-to-epithelial transition (MET) and promotes KDR⁺ EC generation by modulating the miR-125a-5p/Stat3 axis pathway [31].

2.2.3. Delivery of reprogramming factors.

2.2.3.1. Viral delivery method.

For therapeutic translation of direct reprogramming, surveying effective and clinically compatible modes of gene delivery is required. To date, the majority of reprogramming studies employ lentiviral or retroviral vectors, which are integrated into the host genome and induce ectopic expression of the target gene expression [25–27, 29, 30, 32–35]. Ginsberg et al. and Lee et al. employed a doxycycline-inducible expression system to control ETV2 gene expression [26, 30]. After doxycycline (DOX) treatment, ETV2 gene and protein levels were elevated and induced downstream EC gene expression. After withdrawing DOX, directly reprogrammed ECs displayed phenotypes similar to mature fetal and adult ECs as they had low ETV2 levels while maintaining EC features. Even though the lentiviral or retroviral method is very efficient in activating target gene expression, insertional mutations induced by these viral methods is the major drawback for clinical application. Thus, non-integrating methods of gene delivery using DNA viruses such as adenovirus and adeno-associated virus (AAV) is more desirable. To date, one study reported the application of an adenoviral vector to overexpress a non-TF gene, DKK3, to reprogram human embryonic lung fibroblasts [31]. We have been using adenoviruses to transduce ETV2 to reprogram human adult cells into rECs (unpublished data). There is currently no study using AAV to directly induce EC reprogramming, but feasibility was demonstrated in in-vivo hepatic reprogramming study in which hepatic TFs were delivered via AAV6 vectors [43].

2.2.3.2. Non-viral delivery method.

Alternative to the viral vectors, non-viral delivery systems have been developed to avoid issues associated with cytotoxicity. Multiple studies demonstrated non-integrating episomal vectors to transfer genes and induced cellular reprogramming. In 2008, the Yamanaka group repeatedly transfected two genetically engineered plasmids and successfully generated human iPSCs [44]. Yu et al. also generated iPSCs with a single transfection of episomal vectors and confirmed complete removal of vector sequences from the reprogrammed cells [45]. Similarly, Kurien et al. episomally delivered OCT4, SOX2, KLF4, LMYC, LIN28, p53 short hairpin RNA, miR302, and miR367 to establish intermediate partial pluripotent state [23]. Most recently, Cho et al. simultaneously reprogrammed cardiomyocytes (rCMs), smooth muscle cells (rSMCs), and rECs from mouse fibroblasts by transfecting synthetic miRNA mimics, miR-208b-3p, with Lipofectamine [46]. Altogether, studies clearly demonstrate the feasibility of non-integrative delivery methods in inducing cellular reprograming.

2.2.4. Reprogramming mechanisms and small molecules for endothelial reprogramming

Most studies identified ETV2 as an essential TF for direct endothelial reprogramming. ETV2, a member of the ETS TF family, has an indispensable role in vascular development during embryogenesis [47]. Between embryonic days 7.0 and 9.0 in mice, hemato-endothelial progenitors in the mesoderm exhibit transient expression of ETV2 [48, 49]. ETV2-deficient mice embryos are unable to form vasculature and die at an early gestational age due to complete absence of hemato-endothelial lineage development [49, 50]. Etv2 has shown to activate key transcriptional regulators like Ets1 and Fli1 for endothelial development, and Tal1 and Gata1 for erythropoietic development [51–56]. Sinha et al. uncovered that different threshold Etv2 expression is required for initiating these different lineage development [57]. Etv2 functions in a feedforward manner, directly inducing intermediate transcription factors such as Ets1 and Fli1. Additionally, it binds to the promoter of EC related genes like Flk1, Cdh5, and Pecam1 and regulates their gene expressions as previously known [49, 50, 58]. The authors noted that Fli1 and Ets1 are not dose sensitive, and decrease in Etv2 expression dose does not disrupt vascular development. On the other hand, Tal1 is dose sensitive, and changes in Etv2 expression interrupts erythropoietic development. Furthermore, ETV2 collaborates with various co-factors, such as FOXC2, GATA2, and OVOL2 to specify endothelial lineage transitions [58–60]. Recently, Gong et al. addressed mechanisms of direct endothelial reprogramming by ETV2 induction employing scRNA-seq analysis [61]. Using mouse embryonic fibroblasts (MEFs), they elucidated that, at the initial step, ETV2 binds and recruits BRG1, an ATPase component of the SWI/SNF chromatin remodeling complex, onto closed chromatin domains to open chromatin of endothelial genes, as a pioneer factor. They also showed that BRG1 is required to maintain open states of chromatin by ETV2 binding. In another study, Kim et al. revealed that ETV2 targets RAP1 signaling. RAP1, a key regulator for integrin- and cadherin-based cell adhesion system, has an essential role in angiogenesis and vessel stabilization [62]. RAP1 is a downstream effector of cAMP/EPAC1 signaling, and supplementation of 8-Br-cAMP and forskolin, a cAMP analog and a cAMP inducer respectively, has shown to foster ETV2-mediated reprogramming with cAMP/EPAC/RAP1 signaling [33].

There are several studies, reporting the engagement of MET during the EC direct reprogramming process. Chen et al. reported that direct reprogramming of human embryonic lung fibroblasts into ECs by overexpression of DKK3 and found that a conversion process involves MET and miR-125a-5p/Stat3 signaling axis [31]. Han et al. showed inhibition of EMT with siTWIST and Rosiglitazone enhances the conversion efficiency [34]. Likewise, inhibition of TGFβ signaling cascade with SB431542 prevents EMT and helps maintain endothelial features [26].

Various combinations of growth factors, cytokines, and small molecules have been investigated to enhance reprogramming efficiency. Especially, EC growth factors like VEGF and bFGF have been commonly used. VEGF plays a vital role in EC differentiation by promoting the expression of EC-specific proteins [63–65]. bFGF signaling plays an important role in controlling endothelial homeostasis and maintaining EC fate [66, 67]. BMP4, a member of a TGFβ super family, was also used frequently to promote mesodermal lineage specification [66]. Lee et al. showed that VPA, a HDAC inhibitor, plays an additive role in enhancing EC reprogramming [30].

While many studies used TFs as primary reprogramming inducers, Sayed et al. used only small molecule compounds to convert HDFs into ECs without overexpressing TFs [28]. They first activated innate immunity with Poly I:C, a toll like receptor 3 (TLR3) agonist, for 7 days. They cultured the cells in a transdifferentiation media containing VEGF, bFGF, and BMP4 for another week. Then, they added 8-Br-cAMP in the media and cultured for 2 weeks. Lastly, they sorted the PECAM1⁺ cells and maintained the cells in the EGM-2 media with SB431542. The study suggests innate immunity is crucial for epigenetic plasticity and provides an environment suitable for EC transdifferentiation. Recently, Yi et al. used VPA, CHIR99021, an inhibitor of GSK3 resulting in activation of the WNT pathway, Repsox, a TGFβ signaling inhibitor suppressing EMT, Forskolin, an activator of cAMP pathway, and Y-27632 to convert SCAP into ECs [37]. Y-27632 is a Rho-associated protein kinase (ROCK) inhibitor shown to improve differentiation and expansion of ECs generated from embryonic-derived Flk1⁺ mesodermal precursor cells [68]. In addition to these 5 factors, they cultured the cells with VEGF, BMP4, and 8-Br-cAMP. The SCAP derived ECs had enriched endothelial gene expression and formed tubular structure both in vitro and in vivo Matrigel assays.

2.2.5. Role of reprogrammed endothelial cells in therapeutic neovascularization

2.2.5.1. Neovascularization capacity of reprogrammed ECs in non-disease model.

Clinical translation of directly reprogrammed ECs requires thorough investigation of their therapeutic potential and neovascularization capacity. A majority of studies used Matrigel to demonstrate functionality of their reprogrammed cells in vitro and in vivo [25–31, 33–37]. Distinctively, Chen et al. and Kim et al. assessed neovascularization capacity of directly reprogrammed ECs utilizing tissue engineered vascular grafts [31, 33]. Chen et al. seeded iECs in a decellularized mouse aortic graft with an ex vivo circulation bioreactor system [31]. The resultant iECs-reconstructed tissue graft displayed iECs lining the inner most layer surrounded by multiple SMC layers mimicking native vasculature. Kim et al. seeded KDR sorted iECs to an acellular rat liver scaffold and cultured the scaffold with a perfusion bioreactor system [33]. DOX and forskolin supplements in the perfusion medium promoted iECs to form confluent endothelium surrounding lumen of the scaffold as confirmed with immunostaining with Griffonia (Bandeiraea) Simplicifolia Lectin I (GSL I or BSL I), CDH5, and CLDN5. Ginsberg et al. demonstrated engraftment of rAV-VECs into the liver sinusoidal vessels of the NSG mice undergone 70% partial hepatectomy [26]. Three months after transplanting rAV-VECs into the intrasplenic route, they identified colocalization of their reprogrammed cells into the isolectin B4 perfused vessels. Though above studies demonstrated the potential of the reprogrammed ECs for various therapeutic applications, these models and studies did not directly address the therapeutic potential of directly reprogrammed ECs for ischemic cardiovascular diseases.

2.2.5.2. Contribution of reprogrammed ECs in therapeutic neovascularization.

Hindlimb ischemia (HLI) model has been used most commonly to investigate therapeutic effects of directly reprogrammed ECs. Han et al. intramuscularly injected iECs to the ischemic hindlimb and monitored blood flow recovery with laser Doppler perfusion imaging (LDPI) over 2 weeks [25]. iEC-injected group demonstrated blood flow recovery, vascular density, and tissue protective effects similar to the primary EC-injected group. The histological examination of the thigh muscle revealed that iECs were incorporated into the host vasculature labeled with BSL I. Sayed et al. injected their reprogrammed cells intramuscularly on the gastrocnemius muscle at two different time points [28]. First injection was done immediately after HLI induction. Second injection was done at 10 days after surgery when the signaling from bioluminescent labeled iECs became undetectable. They monitored blood flow recovery over 2 weeks with LDPI, and then they assessed degree of tissue damage and quantified blood capillary density on day 18 post-surgery. iEC injected group minimized tissue damage, toe necrosis at most, and improved blood flow in the ischemic hindlimbs similar to the human lung microvascular endothelial cells injected group. iEC-injected group had significantly higher capillary density compared to control group. Morita et al. demonstrated that ETVECs enhanced blood flow recovery and protected ischemic hindlimb from necrosis at day 14 after the cells were intramuscular injected into the adductor muscle immediately after HLI surgery [27]. Though, longer follow-ups for histological analysis were done only in Matrigel plug extracts. At day 28 after Matrigel plug implantation, they found colocalization of ETVECs, co-labeled by Ulexeuropaeus agglutinin I (UEA I) and CD34, with the CDH5⁺ host blood vessels. At day 42, ETVEC-constituting vessels were surrounded by ACTA2⁺ mural cells and expressed endothelial nitric oxide synthase (eNOS) indicating formation of mature and functional vessels. Similarly, Cheng et al. injected ETV2-induced endothelial-like cells (EiECs) suspended in 30% Matrigel intramuscularly on the adductor muscle immediately after HLI surgery [35]. EiECs injected group had improved blood perfusion and minimum tissue damage similar to the HUVEC injected group. In 14-day hindlimb tissue, EiECs constituting vessels were successfully stained with vascular markers like CDH5, PECAM1 and VWF, and had ACTA2⁺ mural cells surrounding them. Kim et al. collaterally injected forskolin-treated and KDR-sorted iECs (Fsk-iECs) from proximal to distal transected sites immediately after femoral artery ligation [33]. The Fsk-iECs treated group displayed blood flow recovery, tissue protective effect, and vessel density similar to HUVEC treated group. Lee et al. most clearly assessed therapeutic effects and neovascularization capacity of human directly reprogrammed EC [30]. They showed that both early and late rECs generated at two different stages of reprogramming have robust neovascularization potential in HLI tissue. Early rECs demonstrated better tissue protective effect, improved blood flow recovery, and higher capillary density compared to HUVECs over 4 weeks post-surgery. They were able to identify incorporation of early and late rEC into the host vasculature at 4 weeks and 3 months post-surgery.

While above studies presented promising therapeutic and neovascularization potential of direct reprogrammed ECs, long-term fate of the reprogrammed cells in vivo disease models are not well investigated. Han et al. showed that around 9% of the transplanted cells were engrafted in the ischemic tissue at 14 days after surgery even though the reprogrammed cells were generated from same murine origin [25]. Lee et al. reported around 7% of rECs were incorporated or anastomosed to the host vessels in the ischemic hindlimb tissue 28 days after surgery [30]. Recently, Cho et al. simultaneously generated cardiovascular tissue (rCVT) composed of rCMs, rSMCs and rECs with extracellular matrix via a direct reprogramming method [46]. They used miR-208b-3p, ascorbic acid, and BMP4 to directly convert mouse tail-tip fibroblasts into the tissue. Reprogrammed GFP labeled rCVT was transplanted onto the infarcted mouse heart. rCVT improved cardiac function, protected host myocardium from fibrosis, and reduced left ventricular wall thinning and compensatory enlargement of ventricles. They reported that rCVT-derived rECs migrated into the host infarcted heart and contributed to new vessel formation. Sixteen weeks after transplantation, 21% of vessels in the infarcted area were composed of or coaptated with GFP⁺ BSL I⁺ rECs, and GFP⁺ ACTA2⁺ reprogrammed SMCs ensheathed over the vessels composed of GFP⁻ PECAM1⁺ host ECs and GFP⁺ PECAM1⁺ rECs. Even though the extent to which rECs alone contribute to therapeutical neovascularization on the infarcted heart still remain unknown, it is clear that administration of the vessel-forming cells embedded within extracellular matrix improved long-term cell retention and contribution to durable vessel formation.

3. Challenges for clinical application of reprogrammed ECs

3.1. Source cells and their origins

Cell source is also a critical aspect to consider when translating rECs for therapeutic purposes. The early studies of EC direct reprogramming used mouse cells and human fetal and neonatal foreskin fibroblasts demonstrating the concept and feasibility of direct reprogramming. Thereafter, fibroblasts collected from various tissues and subjects with different physiological conditions have been tested to systematically generate reprogrammed ECs. However, sourcing fibroblasts requires performing an invasive procedure and can potentially cause complications limiting multiple biopsies. When direct reprogramming aims for an autologous approach, the source cells will be collected from a patient. Collecting methods of source cells would better if they are less invasive and more reliable. For example, blood drawing for blood cell collection is less invasive compared to skin biopsy for skin fibroblasts collection. Even several groups studied cellular reprogramming of urine cells [69, 70] from collected urine, which is much easier and safer to collect compared to drawing blood and performing skin biopsies. The plasticity of cells is also an important element to consider. For EC direct reprogramming, mesodermal lineage cells such as smooth muscle cells and blood cells can be more easily reprogrammed. Not only lineage, but also the age of patients will affect the plasticity of source cells, which may eventually influence reprogramming efficiency. At present, the effect of the age of donor on reprogramming efficiency has been demonstrated in murine cells [71]. Meanwhile, studies on human cells have not consistently found a significant effect of aging on iPSC generation [72, 73]. However, Mertens et al. demonstrated that induced neurons generated from fibroblasts retained transcriptional aging signatures from aged donors [74]. Moreover, reprogramming efficiency has shown to be affected by cellular senescence [75, 76]. Since cellular senescence increases with donor age [77, 78], donor’s precondition may affect the reprogramming outcome. Thus, generally, reprogramming efficiency seems influenced by pre-existing signatures (or memories) of donor cells. In the case of direct EC reprogramming, although Bersini et al. did not find significant differences in reprogramming efficiency based on differential gene expression analysis, they reported heterogeneity in reprogrammed cell phenotypes caused by different ages and preconditions of patients [32]. iVECs from old donors had higher expression of GSTM1 and PALD1, which are associated with oxidative stress, inflammation, and vascular junction stability. In addition, BMPs overexpressed in iSMCs from HGPS were shown to be linked with endothelial barrier damage. This study indicates reprogrammed cells retain signature from original cell type and may affect their functionality.

3.2. Delivery of reprogramming factors

One of the main purpose of direct reprogramming is clinical application. As described in the earlier section, the previous studies demonstrated the potential of reprogrammed ECs as a source for cell therapy. However, when we considered the route of gene transduction or overexpression among previous studies, they are undesirable due to the integration of ectopic genes into host genomes by retroviral [23] or lentiviral [22, 24–27, 30, 79] vectors. These viral vectors are acceptable for biological studies or drug screenings but problematic for cell therapy. However, for clinical application, more favorable vectors for gene transduction should be developed for reprogrammed EC generation. For the choices of viral vectors, non-integrating properties should be considered first. The usage of non-integrating DNA viruses such as replication defective adenovirus or AAV have already been approved for gene therapy and adenovirus has been demonstrated for its feasibility for EC direct reprogramming by Chen et al. [31]. Sendai virus, a cytoplasmic RNA virus, has been shown to deliver transcription factors successfully for iPSC generation [80–82] and induction of cardiomyocytes [83] without the risk of genome integration. Owing to the non-integrating property, ectopic gene expression is not sustained with these viral vectors. Interestingly, the transient expression of transduced genes with these viral vectors is well suited for EC direct reprogramming using ETV2 because it is required in the early phase of reprogramming [26, 30].

Also, we can consider the direct delivery of genetic materials such as DNA in episomal vectors or modified mRNA. These methods are non-integrative and non-viral, and thus safe, clinically compatible, and able to keep host genome integrity. Gene expression through the introduction of DNA plasmids is a classical method of ectopic gene expression. Some early studies of iPSC generation [44, 45] and EC reprogramming via partial pluripotent state [23] used episomal vectors harboring pluripotency transcription factors suggesting that episomal vectors are capable of induce cellular reprogramming. However, when used for direct reprogramming, efficiency via this method is challenging. Modified mRNAs were used to replace viral vectors for iPSC generation a decade ago [84–86]. They were produced in in vitro transcription systems as possessing modified 5’-cap structure and nucleosides to evade the innate immunity and for better stability and translational efficiency. Later, direct reprogramming studies adopted modified mRNA to convert human fibroblasts into myoblast-like cells [87] and hepatocyte-like cells [88]. Although the feasibility of modified mRNA has been proved through these studies, mRNA methods have an advantage and a disadvantage over episomal vectors: its short half-life makes it free from vector clearance problem unlike episomal vectors, whereas mRNA needs to be transduced multiple rounds to maintain its expression at the desired level. Both mRNA and episomal vectors require aids of mechanical or chemical methods such as electroporation and transfection reagents for efficient delivery of genetic materials. Therefore, technical development for nucleic acid transfection must precede the clinical application to ensure cell viability and to maintain the expression of transcription factors, for example, synthetic self-replicative RNAs [89] and self-assembled mRNA nanoparticles [90].

3.3. Reprogramming efficiency

For clinical application, reprogramming efficiency is also important. Although iPSC and direct reprogramming strategies share similar concept of ectopically expressing specific transcription factors, iPSCs have the capacity of clonal expansion to produce an unlimited number of cells even from a single colony, whereas cells produced through direct reprogramming are somatic cells with limited capacity of proliferation. Therefore, the efficiency of direct reprogramming is much more critical. Aforementioned studies attempted to increase reprogramming efficiency employing additional reprogramming factors including novel transcription factors, signaling molecules to activate or inhibit specific signaling pathways, non-coding RNAs or epigenetic modulators. However, the most efficient method of reprogramming is unclear, due to different markers used to define reprogrammed cells and the varying time points at which the cells were collected. Han et al. was the first to make a direct comparison of the reprogramming efficiency between their method and methods from Ginsberg et al., Morita et al., Wong et al., and Lee et al. [34]. They derived rECs from human fibroblasts and noted that the proportion of CDH5/PECAM1 double-positive populations at D42 stayed below 15% regardless of which methods were used.

3.4. Subtypes of ECs

Furthermore, there should be an investigation of different phenotypes within reprogrammed cells to construct a better clinical model for subtype-specific cell therapy. Currently, there is not a clear understanding of how different endothelial subtypes can be generated from direct reprogramming strategies. During endothelial lineage development, arterial and venous endothelium are emerged from primitive vasculature. COUP-TFII has shown to play a critical role in venous development, and the knockdown of the TF resulted in inappropriate activation of arterial markers like Nrp1 and Notch1 in veins [91, 92]. You et al. suggested that COUP-TFII is a key factor for venous lineage specification, and the down regulation of Nrp1 expression inhibits Notch signaling and subsequent suppression of arterial genes [92]. Conversely, Hey1 and Hey2 are important modulators for Notch signaling, and knockdown of Hey genes suppressed expression of arterial markers while expression of venous markers increased [93, 94]. So far, studies used notch signaling inhibitor, such as γ-secretase inhibitor L-685, 458 (GSI), or agonist, such as resveratrol, to promote venous or arterial ECs differentiation from PSCs, respectively [95, 96]. Several studies also suggested that VEGF specify EC fate by binding at VEGFR2. High concentration of VEGF induced arterial-specification, and low concentration of VEGF induced venous-specification. Lymphatic ECs are another endothelial subtype that originates from venous endothelium. Prox1 has shown to be a master regulator for lymphatic development [97]. Prox1 knockdown caused failure in developing lymphatic vasculature [98], and overexpression of Prox1 increased expression of lymphatic genes like, Podoplanin and Vegfr3, while downregulating blood endothelial genes [99, 100]. Sox18 also has been considered as important factor that acts upstream of Prox1 [101]. Sox18 knockdown mice died with severe edema, while overexpression of Sox18 increased expression of lymphatic EC specific markers including Prox1. Lymphatic endothelial cells were successfully generated from iPSC with VEGF-C supplement [102]. Overall, an understanding of the transcription factors, growth factors, and signaling pathways involved in the specification of different endothelial subtypes can provide valuable insights for the development of direct reprogramming methods aimed at generating specific endothelial subtypes. Recent advance in sequencing technology for transcriptome, epigenome, proteome, and metabolome will provide clues to decipher the reprogramming mechanisms and to identify candidates for additional factors for reprogramming. Also, recent efforts to elucidate the heterogeneity of ECs with scRNA-seq analysis will be able to identify unique transcriptional programs specific to each subtype of ECs, which can be harnessed to enhance efficiency by further specification of EC subtypes during EC direct reprogramming [103, 104].

4. Conclusions

We have reviewed the studies of EC generation using direct reprogramming strategies. These studies not only demonstrated the feasibility of direct reprogramming as a method to produce functional ECs but also paved a new path to tackle cardiovascular diseases. These reprogrammed ECs exhibited phenotypic and functional EC characteristics in vitro and in vivo. Also, they displayed therapeutic potential in cardiovascular animal models. In addition, simpler procedures and shorter processing time compared to the generation of PSC derived cells as well as reasonable conversion efficiency with viral transduction make direct reprogramming a promising option to acquire autologous ECs for treating CVD via cell therapy.

As a reprogramming agent, not only new viral vectors such as Sendai viruses, adenoviruses, or AAV are clinically applicable but modified mRNA could be a great option. Various source cells need be explored to find more reliable and feasible options. Information obtained from the mechanism studies with high throughput sequencing and bioinformatics, especially single cell sequencing, will contribute to the identification of novel factors to improve direct reprogramming. In addition to generating reprogrammed cells in vitro, this direct reprogramming concept can be applied to in vivo cells and tissues. While such approach was demonstrated in mouse cardiomyocyte and neurons using lineage tracing, no studies clearly demonstrated such possibility for endothelial cells [105–109]. Direct cellular reprogramming also shed new light on generating a cardiovascular tissue in vitro which include cardiac and vascular cells using microRNAs and chemicals [46]. Despite such advance, there remains multiple questions to be answered for clinical application such as the long-term fate of reprogrammed cells and potential side effects. However, with rapid progress in this field, EC direct reprogramming strategy carries a great potential for treatment of cardiovascular diseases, disease modeling and drug discovery.

Highlights.

• Cell therapy has been considered an attractive strategy to treat advanced ischemic cardiovascular diseases as it supplies cardiac and vascular components.

• Direct reprogramming strategy, which aims to generate a target cell type using lineage-specific factors without undergoing pluripotency, received remarkable attentions for regenerative therapy.

• Somatic cells from different origins have been successfully converted into functional endothelial cells with various direct reprogramming methods.

• To enhance reprogramming efficiency, studies used combinations of various reprogramming factors such as transcription factors, angiogenic growth factors, and small molecular inhibitors or modulators for signaling pathways.

Abbreviations

AAV

adeno-associated viral

AC

amniotic fluid-derived cell

bFGF

basic fibroblast growth factor

BMP4

bone morphogenetic protein 4

BSL I

Griffonia (Bandeiraea) Simplicifolia Lectin I (GSL I)

CM

cardiomyocyte

CVD

cardiovascular disease

DOX

doxycycline

DPSCs

Dental pulp stem cells

EC

endothelial cell

EGM-2

Endothelial Cell Growth Media-2

EiEC

ETV2-induced endothelial cell

EMT

epithelial-to-mesenchymal transition

eNOS

endothelial nitric oxide synthase

ESC

embryonic stem cell

Fsk

forskolin

hADSCs

human adipose-derived stem cells

HDF

human dermal fibroblasts

HGPS

Hutchinson-Gilford Progeria Syndrome

hiPSC

human induced pluripotent stem cell

HLI

hindlimb ischemia model

HUVEC

human umbilical vein EC

iEC

induced endothelial cell

iEnd

induced endothelial

ILB4

isolectin B4

iN

induced neuron

iPSC

induced pluripotent stem cell

iSMC

induced smooth muscle cell

iVEC

induced vascular endothelial cell

LDPI

laser Doppler perfusion imaging

MET

mesenchymal-to-epithelial transition

MI

myocardial infarction

MIM

mesoderm induction medium

miRNA

microRNA

modRNA

modified mRNA

OSKM

OCT4, SOX2, KLF4, c-MYC

PiPSC

partial-induced pluripotent stem cell

PSC

pluripotent stem cell

rEC

reprogrammed endothelial cell

rCVT

reprogrammed cardiovascular tissue

SCAP

stem cells from apical papilla

SCID

severe combined immunodeficiency

shP53

p53 short hairpin RNA

siRNA

small interfering RNA

SMC

smooth muscle cell

TGF

transforming growth factor

TF

Transcription factor

UBC-MSC

umbilical cord blood-derived mesenchymal stem cells

UEA I

Ulexeuropaeus agglutinin

VEGF

vascular endothelial growth factor

VPA

valproic acid