Efficient and robust differentiation of endothelial cells from human induced pluripotent stem cells via lineage control with VEGF and cyclic AMP

Blood vessels are essential components for many tissues and organs. Thus, efficient induction of endothelial cells (ECs) from human pluripotent stem cells is a key method for generating higher tissue structures entirely from stem cells. We previously established an EC differentiation system with mouse pluripotent stem cells to show that vascular endothelial growth factor (VEGF) is essential to induce ECs and that cyclic adenosine monophosphate (cAMP) synergistically enhances VEGF effects. Here we report an efficient and robust EC differentiation method from human pluripotent stem cell lines based on a 2D monolayer, serum-free culture. We controlled the direction of differentiation from mesoderm to ECs using stage-specific stimulation with VEGF and cAMP combined with the elimination of non-responder cells at early EC stage. This “stimulation-elimination” method robustly achieved very high efficiency (>99%) and yield (>10 ECs from 1 hiPSC input) of EC differentiation, with no purification of ECs after differentiation. We believe this method will be a valuable technological basis broadly for regenerative medicine and 3D tissue engineering.


Introduction
Blood vessels play essential roles in the generation of higher tissue structures, especially large tissue and organ structures.The importance of endothelial cells (ECs) has already been shown in the formation of various organs such as heart [1][2][3], liver [4][5][6][7], kidney [8], bone [9], and skin among many others [10][11][12][13].Thus, efficient EC preparation methods that provide scalable and stable supply are necessary for three-dimensional (3D) tissue engineering and organ regeneration.Human pluripotent stem cells are one of the most suitable sources for such purpose.
Previously, using mouse embryonic stem cells (ESCs), we established a method for systematic induction of cardiovascular cells from vascular endothelial growth factor (VEGF) receptor-2 (VEGFR2)-positive mesoderm cells as cardiovascular progenitors [14,15].VEGF/ VEGFR2 signaling is essential for inducing EC differentiation from VEGFR2-positive mesoderm cells.Furthermore, we also found that cyclic adenosine monophosphate (cAMP) signaling potently enhances EC differentiation [16,17] and that activation of a major downstream molecule of cAMP, protein kinase A (PKA), increased the expression of VEGFR2 and another VEGF receptor, neuropilin1, which together form a specific receptor for the VEGF-A 165 isoform.The binding of VEGF-A 165 to VEGFR2 and neuropilin1 is reported to enhance VEGFR signaling by approximately a factor of ten.Coincidently, PKA activation increased the sensitivity of VEGFR2 + progenitors to VEGF, which increased the appearance of ECs also by a factor of ten [17].PKA is also directly involved in the EC commitment process.Etv2/ER71, an ETS transcription factor, plays an indispensable role in EC and hematopoietic lineage commitment from early mesoderm [18,19].We previously showed that PKA-activated CREB (cyclic AMPresponsive element (CRE) binding protein) bound to CRE on the Etv2/ER71 promoter region and directly induced Etv2/ER71 expression [20].In that same report, we also observed that PKA activation during ESC differentiation triggered EC differentiation and induced early commitment to EC lineage.In addition, we reported that Notch and β-catenin signaling are simultaneously activated in the downstream of cAMP and protein complex formation with Notch intracellular domain and β-catenin induced a set of arterial EC gene expressions resulted in arterial EC differentiation [21].These results indicate that VEGF is critical for EC differentiation and growth, while cAMP is critical for EC commitment and specification.
With regards to human induced pluripotent stem cell (hiPSC) differentiation, we previously reported an efficient cardiomyocyte (CM) differentiation method based on a 2D monolayer, serum-free condition [22], that was modified from a directed differentiation protocol from human ESCs [23].In our method, we first induce mesoderm cells with Activin-A, bone morphogenic protein 4 (BMP4), and basic fibroblast growth factor (bFGF), and then induced CM commitment with a wnt inhibitor, Dickkopf-related protein 1 (DKK1).Inferring from our mouse ESC results, we anticipated that fate control of mesoderm stage cells to EC lineage should provide an efficient source of ECs.Therefore, in the present study, we investigated an EC differentiation method from hiPSCs that combined differentiation stage-specific supplementation of VEGF and cAMP.We further demonstrated that purification of EC-committed cells at peri-EC stage, which can eliminate non-responder cells to EC differentiation ques, achieved highly pure and efficient EC differentiation.

Flow cytometry
On d(9), we dissociated the cells using Accumax (Innovative Cell Technologies, San Diego, CA) and stained them with the cell surface markers listed in S1 Table .For cell surface markers, staining was carried out in PBS with 5% FCS.To eliminate dead cells, cells were stained with 4',6-diamidino-2-phenylindole (DAPI) for surface marker staining or with the LIVE/ DEAD fixable Aqua Dead Cell Staining Kit (Thermo Fisher Scientific) for intracellular staining.For intracellular proteins, staining was carried out on cells fixed with 4% paraformaldehyde (PFA) in PBS.Cells were stained with the anti-cardiac isoform of Troponin T (TNNT2) (clone 13211, Thermo Fisher scientific) labeled with Alexa-488 using Zenon technology (Thermo Fisher Scientific).The staining was performed in PBS with 5% FCS and 0.75% Saponin (Sigma-Aldrich, St. Louis, MO).Stained cells were analyzed on an AriaII flow cytometer (Becton Dickinson, Franklin Lakes, NJ).After the selection of FSC/SSC gate, we additionally eliminated the doublets by SSC-W/SSC-H gate and FSC-W/FSC-H gate.Data were collected from at least 10,000 events.

Evaluation of Vascular Endothelial (VE)-cadherin + ECs at d(9)
At d( 9), cells were dissociated using Accumax and stained with DAPI and VE-Cadherin.Staining was carried out in PBS with 5% FCS.Stained cells were examined and purified on an AriaII flow cytometer.Viable cells were counted with the trypan blue-exclusion test.The yield of ECs was calculated by multiplying the viable cell count with the percentage of ECs.
Purification and re-culture of VEGF Receptor-2 (VEGFR2) + cells at d (6) At d(6), cells were dissociated using Accumax and stained with DAPI and VEGFR2.VEGFR2-positive cells were purified on an AriaII flow cytometer and then recultured at 10,000 VEGFR2 + cells/cm 2 in RPMI+B27 medium with 1 mM 8bromo-cAMP, 100 ng/mL VEGF and 10 μmol/L of a Rho-associated coiled-coil forming kinase inhibitor (Y-27632; Cal-Biochem) on thin-coated Matrigel dishes.The culture medium was replaced to that supplemented with VEGF alone on d (7).On d (9), viable cells was counted by the trypan blue-exclusion test.The purity of VE-cadherin and CD31 double-positive ECs was analyzed with a flow cytometer.

Immunofluorescent imaging
Cells were fixed with 4% PFA and stained with antibodies shown in S1 Table .Nuclei were visualized with DAPI (Thermo Fisher Scientific).Stained cells were photographed with an all-inone fluorescent microscopic system, Biorevo BZ-9000 (Keyence, Osaka, Japan).

Tube formation assay
Tube formation assay was performed as described previously [26].Briefly, differentiated ECs (7 x 10 4 ) at d13 were plated on a 24-well plate (Thermo Fisher Scientific) coated with 300 μl Matrigel Basement Membrane Matrix GFR (BD Biosciences), and cultured for 24 hrs.

Quantitative reverse-transcription Polymerase Chain Reaction (qPCR)
Total RNA was extracted from purified ECs using RNeasy (QIAGEN, Hilden, Germany) according to the manufacturer's instructions.GAPDH was used to normalize gene expressions.Quantitative PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) on a StepOnePlus system (Thermo Fisher Scientific) with Delta Delta Ct method.Forward and reverse primer sequences are shown in S2 Table.

Statistical analysis
At least three independent experiments were performed.Statistical analysis of the data was performed with ANOVA.p < 0.05 was considered significant.Values are reported as mean ± SD.

Optimization of VEGF and cAMP supplementation
Previously, we established a monolayer high-density culture-based CM differentiation protocol from hiPSCs [22,23].In the present work, we attempted to induce ECs from hiPSCs by directing the fate of mesoderm-stage cells toward ECs mainly by VEGF and cAMP activation.Briefly, mesoderm-stage cell induction process is as follows: undifferentiated hiPSCs maintained on feeder-free condition were collected as single cells, and seeded on Matrigel-coated multi-well plates 4 days prior to the initiation of differentiation (d(-4)).Matrigel overlay was conducted when the culture became fully confluent (d(-1)) for 24 hours, then we initiated the differentiation culture (d(0)).Mesoderm cells were induced with the addition of Activin-A (100 ng/mL) from d(0) to d(1) followed by BMP4 (10 ng/mL) and bFGF (10 ng/mL) from d(1-5).
Then, various concentrations of VEGF and a cAMP analogue, 8-bromo-cAMP, were supplemented at time points around the possible mesoderm stage (d( 5)), and the efficiency of vascular endothelial (VE)-cadherin-positive EC induction on d( 9) was evaluated with flow cytometry.First, we examined the effects of VEGF.The addition of VEGF on d(5-9) dramatically induced EC appearance compared with no VEGF.VEGF (100 ng/mL) stimulated efficient VE-cadherin + EC appearance whereas VEGF at 200 ng/mL showed no apparent difference with VEGF 100 ng/mL (S1a ) and yield (16.1×10 4 cells/cm 2 ), suggesting that cAMP acts on EC commitment and early EC differentiation processes, but has less effect at later stages.In our CM differentiation protocol, cells are covered with Matrigel (Thermo Fisher Scientific) solution at d(-1) to stabilize cell attachment [22].Because cells became unstable after Activin-A treatment and massive cell detachment often occurred during EC differentiation that caused the termination of experiments, we added a secondary Matrigel overlay at d(1).This process successfully prevented massive detachment and loss of cells during EC differentiation and resulted in high EC yield.Small modification in the Activin-A treatment (125 ng/mL, 18 hours treatment) further improved cell attachment and EC appearance (data not shown).We adjusted the starting day of VEGF treatment to d(4) (the same date as cAMP treatment) to simplify the protocol.Following these modifications, we established the first endothelial differentiation protocol called "stimulation method" (Fig 1a).In summary, single dissociated hiPSCs are plated on Matrigel-coated dishes at d(-4) (60,000 to 87,500 cells/cm 2 ; input hiPSCs) and cultured in the maintenance condition.The first Matrigel overlay is on d(-1), and on d(0), the medium is changed to differentiation medium (RPMI1650 with B27 supplement).Then, Activin-A (125 ng/mL) is added for 18 hours followed by medium change with BMP4 (10 ng/mL) and bFGF (10 ng/mL) treatment and the second Matrigel overlay.VEGF (100 ng/mL; d4-9) and 8bromo-cAMP (1.0 mM; d(4-6)) treatments start on d(4).EC appearance is evaluated on d (9).

Evaluation of the stimulation method
Next, we evaluated the stimulation method by measuring EC differentiation.The addition of both VEGF and 8bromo-cAMP were compared to those with VEGF alone or with neither VEGF nor 8bromo-cAMP (vehicle).The efficiency of EC induction was evaluated with flow cytometry on d (9).The VE-cadherin + EC population was significantly increased with the addition of VEGF and 8bromo-cAMP compared to VEGF alone or vehicle (56.2±12.5% vs. 11.8±7.2% vs. 2.3±2.4% of total cells, P = 0.000017, n = 4) (Fig 1b).The calculated EC count was also notably increased following the VEGF and 8bromo-cAMP treatment (1.66±0.70×10 5 vs. 4.9±3.3×10 4 vs. 9.8±10.4×10 3 cells/cm 2 culture surface, P = 0.0022, n = 4) (Fig 1c).On the other hand, the cardiac troponin T + CM population decreased with VEGF and 8bromo-cAMP treatment (20.0±13.2% vs. 47.6±25.7% vs. 55.9±12.1% of total cells, P = 0.049, n = 4), while the populations of Tra-1-60 + undifferentiated hiPSCs and platelet-derived growth factor receptor β (PDGFRβ)-positive vascular mural cells were unchanged (S2 Fig) .Overall, the cell populations on d(9) using the above protocol consisted of 60% VE-Cadherin + ECs, 20% cTnT + CMs, 5% PDGFRβ + vascular mural cells, and 10% TRA1-60 + undifferentiated hiPSCs.Induced ECs (up to 70%) were purified by FACS or MACS using anti-VE-cadherin and/or CD31 antibodies, which provided ECs at more than 99% purity (Fig 1d Despite the good effects of VEGF and cAMP on EC commitment, we noticed that not all mesoderm cells responded and committed to EC lineage when we examined the time course of VEGFR2 + mesoderm population (Fig 2e).VEGFR2 + mesoderm cells were induced by Activin-A/BMP4/bFGF treatment during d(0-4).In the control, VEGFR2 + cell population reached maximum on d( 5) and then decreased with almost no VE-cadherin + EC appearance.On the other hand, VEGF and 8bromo-cAMP treatment induced VE-cadherin + ECs that maintained VEGFR2 expression after d (6).While many cells started to express VE-cadherin on d(6), a small population of cells remained negative for VEGFR2.These VEGFR2-negative cells never disappeared even after EC differentiation (d( 10))(arrows in Fig 2e), indicating that they did not respond to EC commitment signaling.We speculated that these "non-responder cells" should be a main cause of the contamination of non-ECs after EC differentiation with the stimulation method.

Second EC differentiation protocol: Stimulation-elimination method
To further improve the EC induction efficiency, we therefore examined the elimination of these non-responder cells at an early EC differentiation stage.We purified VEGFR2 + cells on d6 with 99.0±0.5% purity (n = 7) and re-cultured them with VEGF and 8bromo-cAMP (Fig 3a).8bromo-cAMP was withdrawn within one day (d( 7)).On d( 9), the responder cells gave rise to ECs with more than 99% purity (Fig 3b).Moreover, the yield of ECs increased by a factor of four compared to the stimulation method, which did not include the elimination (0.68 ±0.13 ECs vs. 4.20±0.83ECs from 1 input hiPSC, n = 3) (Fig 3c).These results indicate that the selection and re-culture of VEGFR2 + cells on d( 6) successfully eliminates non-responder cells and achieves almost complete EC induction from the responders.This second method, which combines VEGF and cAMP stimulation with non-responder elimination ("stimulation-elimination" method), showed highly specific and efficient EC differentiation.Purified ECs on d( 9) induced with the stimulation-elimination method or stimulation method were similarly expanded approximately 2.5 times after an additional 5-day culture in endothelial serum free medium (Human Endothelial-SFM) (Fig 3d).This additional culture resulted in 11.1 ECs and 1.74 ECs per 1 input hiPSC on d( 14) using the stimulation-elimination method and stimulation method, respectively.ECs were replated at 10,000 cells/cm 2 following dissociation with Accumax (Innovative Cell Technologies) and grew to 80-90% confluence within 5 to 7 days of culture.In many cases, ECs showed a tendency to cease growing following an additional one to three passages in the human endothelial serum free medium condition.Occasionally, ECs were able to continuously proliferated reaching approximately 10 3 −10 4 times increase in 2 months of culture period.Induced ECs were cryopreserved using a freeze-preserving liquid (Cellbanker III; Nippon Zenyaku Kogyo Co, Koriyama, Japan) on both d(9) and d (14).Approximately two thirds of the plated cells survived the day after thawing and grew healthily afterwards.

Characterization of hiPSC-derived ECs
hiPSC (201B6)-derived pure ECs were cultured on a gratin coated dish and positively immunostained for the endothelial markers CD31 and VE-cadherin (Fig 4a and 4b).The mRNA expression levels of the endothelial markers CD31, VE-cadherin and endothelial nitric oxide synthase (eNOS) in hiPSC-derived ECs on d(9) were comparable to those in HUVECs (human umbilical vein endothelial cells) and significantly higher than those in undifferentiated hiPSCs (negative control)(Fig 4c -4e).CD34 [28] and CD133 [29] are markers of ECs as well as multipotent progenitor cells, including immature hematopoietic stem cells and precursors of endothelial cells [30,31].CD34 and CD133 expression levels of hiPSC-derived ECs were higher than those of HUVECs (Fig 4f and 4g).Though the expression levels had decreased during the culturing of ECs, they were still higher than in HUVEC, even on d(30) (data not shown).These results suggest that our methods achieve induction and maintenance of an early stage ECs.We next tested tube formation assay.ECs (d( 13)) successfully formed tube-like networks on Matrigel (Fig 4h).These cells configuring tube formation were positive for CD31 (Fig 4i).Uptake of low-density lipoproteins (LDLs), another EC function, was clearly displayed in hiPSC-derived ECs (Fig 4j and 4k).The tube formation capacity and morphology were almost comparable with those of HUVECs (S5 Fig) .Those results suggest that hiPSC-derived ECs with stimulation-elimination method are functional ECs.

Arterial and venous specification of hiPSC-derived EC
Previously, we reported that cAMP signaling is involved in arterial endothelial specification through Notch and β-catenin activation in a mouse ES cell system [21].Notch signal related- genes, including Notch1 and 4, Dll4, RBP-J, and Hey1/Hey2, are essential for arterial formation and vasculature development [32][33][34][35][36]. On the other hand, chicken ovalbumin upstream promoter-transcription factor II (COUP-TFII) suppresses Notch signaling and regulates vein identity [37].We recently demonstrated that angiopoietin 1 induces venous ECs in coronary vein formation in vivo and EC differentiation from mouse embryonic stem cells [38].
Then, we examined arterial-venous specification in human ECs induced from iPSCs.As being speculated from that ECs are induced with VEGF and cAMP stimulation, ECs that induced with our method were found to be initially induced with arterial features that show 13.5 times more mRNA expression for the arterial EC marker ephrinB2 [39] than do human umbilical artery endothelial cells (HUAECs) on d (9).However, following an additional 5-day culture for EC expansion with no cAMP, ephrinB2 expression (d( 14)) significantly decreased but still higher than that in HUAECs (Fig 5a).Other arterial marker expressions, such as Dll1, Dll4, and Notch1, were similarly high in ECs on d (9) and showed a tendency to decrease by d ( 14) (S6 Fig) .On the other hand, whereas the mRNA expression of a venous EC marker, COUP-TFII [37] on d (9) was significantly lower than that in HUVECs, it returned to a comparable level after EC expansion (Fig 5b).We previously showed that arterial-venous specification was unstable in the early stage of ECs in a mouse ESC system [21].Early ECs still possess plasticity between arterial and venous phenotypes, and they can change their features from arterial to venous and vice versa according to the culture conditions.Therefore, we speculated that the ECs induced from hiPSCs in this study remained in an undetermined state for arterial and venous ECs.Higher expressions of early stage EC markers CD34 and CD133 (Fig 4f and  4g) further support our speculation that ECs retain plasticity in EC diversity.
Next, we tried to control arterial and venous specification with various combinations of small molecules or growth factors in our stimulation-elimination method (Fig 5c -5e).The ratio of NRP1-positive cells at d6 in stimulation-elimination method or NRP1 and CXCR4 double positive ECs at d (9) were significantly increased compared to those with VEGF alone (no cAMP) condition.At d( 14), ECs with VEGF alone condition completely lost arterial maker.Moreover, extended cAMP supplementation after d(7) had beneficial effect for arterial specification at d(9) and d (14).The positive ratio of NRP1 and CXCR4 at d(9) was almost comparable with those in HUAECs (human umbilical artery endothelial cells) (Fig 5e).On the other hand, Supplementation of angiopoietin 1 after d(6) together with cAMP administration during d(4) to d( 6) significantly decreased the ratio of arterial EC population at d (9).Thus, arterial-venous EC fates were able to be controlled during and after EC differentiation.

Discussion
In this study, we describe efficient and scalable EC differentiation methods based on a 2D monolayer, serum-free culture system.Potent induction of EC commitment and differentiation with cAMP and VEGF at the mesoderm stage dramatically shifted the fate of cells toward ECs and achieved highly efficient EC induction.Moreover, the elimination of cells that had not responded to EC commitment signals (non-responder cells) at the peri-EC stage (d( 6)) resulted in almost 100% differentiation efficiency to ECs.The combination of differentiation stage-specific signal stimulation and non-responder elimination (stimulation-elimination method) is a highly efficient and scalable strategy for the pure induction of target cell populations.
To generate a large-sized organ or tissue, blood vessels that are mainly formed by ECs are necessary.For example, cardiac regeneration in humans after myocardial infarction is estimated to require 10 8 to 10 9 cells [23,[40][41][42][43]. Given that the percentage of the EC population to total cells in an adult mouse heart is around 7%, 10 7 to 10 8 ECs would be required to realize human heart regeneration by cell transplantation.Our method is calculated to be amenable to prepare 10 8 ECs from just two 6-well plates robustly from several hiPSC lines, suggesting the possibility to scale the delivery of ECs even at an industrial level.
ECs are key to 3D tissue engineering, because blood vessels are necessary for mature and functional tissues and organs.ECs in vivo have been reported to have diverse phenotypes that closely relate to the tissue and organ function, such as fenestrated ECs in bone marrow, sinusoidal ECs in liver [44], ECs in glomeruli and podocytes in kidney [45], and ECs involved in the blood-brain barrier and astrocytes in brain [46].Previously, various methods for induction of ECs from human pluripotent stem cells were reported [47][48][49][50][51]. Lineage control and enhanced EC differentiation with cAMP signaling combined with VEGF is our original method [16,17].ECs induced by VEGF and cAMP showed plasticity between arterial and venous phenotypes [21].In addition, hiPSC-derived ECs induced with our method showed expression of immature EC markers (Fig 4f and 4g), suggesting that these ECs should possess a higher ability to adjust to tissue-specific environments and diversify with appropriate function than do fully differentiated EC sources such as HUVECs.Thus, our hiPSC-derived ECs should be more suitable for the reconstitution of organ functions in vitro and in vivo.These results suggest that our stimulation-elimination method, which efficiently and robustly differentiates hiPSCs into ECs, is a potential technological basis for regenerative medicine and tissue engineering.

Fig 2 .Fig 3 .
Fig 2. Time course of endothelial cell and pre-endothelial cell marker.(a) Representative expression time course of VE-cadherin (VECad) and CD31 under stimulation method (VEGF+cAMP) or vehicle without VEGF and cAMP by FACS.(b) Time course of VE-Cadherin-positive cell ratio in two groups.(c) Yield of VE-Cadherin positive endothelial cells per 1cm 2 in two groups.(d) Time course of total cell counts in two groups.(e) Representative expression time course of VEGF receptor 2 (VEGFR2) and VE-cadherin in stimulation method (VEGF+cAMP) or vehicle without cAMP and VEGF.Arrows: non-responder cells to VEGF and cAMP stimulation.https://doi.org/10.1371/journal.pone.0173271.g002