Diabetic Endothelial Cells Differentiated From Patient iPSCs Show Dysregulated Glycine Homeostasis and Senescence Associated Phenotypes

Induced pluripotent stem cells derived cells (iPSCs) not only can be used for personalized cell transfer therapy, but also can be used for modeling diseases for drug screening and discovery in vitro. Although prior studies have characterized the function of rodent iPSCs derived endothelial cells (ECs) in diabetes or metabolic syndrome, feature phenotypes are largely unknown in hiPSC-ECs from patients with diabetes. Here, we used hiPSC lines from patients with type 2 diabetes mellitus (T2DM) and differentiated them into ECs (dia-hiPSC-ECs). We found that dia-hiPSC-ECs had disrupted glycine homeostasis, increased senescence, and impaired mitochondrial function and angiogenic potential as compared with healthy hiPSC-ECs. These signature phenotypes will be helpful to establish dia-hiPSC-ECs as models of endothelial dysfunction for understanding molecular mechanisms of disease and for identifying and testing new targets for the treatment of endothelial dysfunction in diabetes.


INTRODUCTION
Vascular endothelial cells (ECs) play an important role in maintaining vascular homeostasis (Rajendran et al., 2013). In addition to serving as a physical barrier between the vessel wall and blood, ECs regulate platelet aggregation, coagulation, fibrinolysis, and vascular tone through secreting nitric oxide (NO), cytokines, and growth factors (Kruger-Genge et al., 2019). Endothelial dysfunction is a condition in which the EC loses its physiological properties, promotes vasoconstriction, fibrinolysis, inflammation, and atherogenesis (Kruger-Genge et al., 2019).
Although primary human ECs are the best choice of cells for studying endothelial dysfunction, senescence, and angiogenesis, they are limited by cell source and cell isolation technique. The discovery of human induced pluripotent stem cells (hiPSCs) has paved the way for EC generation (Patsch et al., 2015;Zhang et al., 2017;Christensen et al., 2019). Disease specific hiPSC lines can be reprogrammed from patient somatic cells and subsequently can be differentiated back into somatic cells for disease modeling and drug screening and discovering in vitro (Liu et al., 2018;Rowe and Daley, 2019;Williams and Wu, 2019).
In this study, we investigated hiPSC lines reprogrammed from dermal fibroblasts derived from patients with T2DM. ECs were generated from diabetic hiPSCs (dia-hiPSCs). We aim to identify signature and specific phenotypes in dia-hiPSC derived ECs (dia-hiPSC-ECs). This will be helpful to establish dia-hiPSC-ECs as models for understanding the molecular mechanism as well as screening and discovering drugs for treatment of endothelial dysfunction and senescence, abnormal angiogenesis, and disrupted glycine homeostasis in diabetes.

MATERIALS AND METHODS hiPSC Culture and Differentiation
Four hiPSC lines used in this study were reprogrammed from dermal fibroblasts using Sendai virus and 4 reprogramming factors (Thermo Fisher, North Logan, UT, United States): OCT4, SOX2, KLF4, and C-MYC, as described previously (Ye et al., 2013). PCBC16iPS and P1C1iPS cells were reprogrammed from human neonatal and adult dermal fibroblasts of healthy subjects (Lonza, Switzerland), respectively . DP2-C8iPS and DP3-C6iPS cells were reprogrammed from dermal fibroblasts of two adult patients with T2DM. PCBC16iPS and P1C1iPS are derived from male donors. DP2 donor was a female patient and 53 years old, while DP3 donor was a male patient and 51 years old. The procedures were approved by the ethics committee of Nanjing Hospital, Nanjing, China and the Centralised Institutional Review Board of Singapore Health Services Pte. Ltd, Singapore. Informed consent forms were signed by both patients. hiPSCs were cultured in a feeder-free system with a 1:1 mixture of E8/mTeSR (STEMCELL Technologies, Canada) and passaged every 4 days with Versene (Thermo Fisher Scientific, United States). PCBC16iPS was an established hiPSC line and described previously (Ye et al., 2013Zhang et al., 2014;Su et al., 2018;Tan et al., 2019;Tao et al., 2020).
For teratoma formation experiment, the animal experimental protocol and procedures were approved by the Institutional Animal Care and Use Committee of Singapore Health Services Pte. Ltd. 2 × 10 6 hiPSCs were injected into the flanks of NOD-SCID mice (Invivos, Singapore). Animals were followedup to 3 months. Teratoma were explanted, processed, and embedded into paraffin. Tissue sections of teratoma were stained with hematoxylin and eosin and viewed under an Olympus microscope to identify endoderm, mesoderm, and ectoderm cells. For karyotyping, hiPSCs were assessed by Cytogenetics Lab. of Department of Molecular Pathology, Singapore General Hospital, Singapore or of KK Women's and Children's Hospital, Singapore.

Fluorescence Immunostaining
For characterization of hiPSC-ECs in vitro, cells were fixed with 4% paraformaldehyde for 20 min at room temperature. After blocking, primary antibodies, mouse anti-CD31 or anti-CD144 (both from BD Pharmingen, United States), at 1:100 dilution was added into 10% DPBS and incubated overnight at 4 • C. Then, cells were incubated with PE-conjugated goat anti-mouse IgG secondary antibody (Thermo Fisher Scientific, United States) in 10% DPBS for 1 h at room temperature, labeled with DAPI, washed, and viewed under a fluorescence microscope (Olympus, Japan).

Cell Population Doubling Time
Endothelial cell population doubling time was calculated within 7 days post-sorting. Briefly, ECs were harvested on day 2 postsorting and cultured in 6-well plates (2 × 10 5 cells/well) in EGM. The medium was changed every 2 days. ECs were harvested and counted on day 7.

Western Blot Analysis
Endothelial cells on day-7 post-sorting were used for Western Blot analysis as described previously (Ye et al., 2009;Su et al., 2018). For protein analysis using EC supernatant, 2 × 10 5 ECs/well were cultured in 1 ml EGM for 48 h. Then supernatant was collected to determine the endothelin-1 and Matrix metalloproteinase-1 (MMP-1) proteins secreted from ECs. Cell lysate was prepared using PhosphoSafe TM Extraction Reagent (Merck, Germany) and protein concentrations were determined using Bradford reagent (Bio-Rad Laboratories, United States). Proteins from supernatant or cell lysate were separated, electrophoretically blotted onto nitrocellulose membranes, and washed with 10 mM Tris-HCl buffer (pH 7.6) containing 0.05% Tween-20; then, the membranes were incubated in blocking buffer (5% non-fat dry milk, 10 mM Tris pH 7.5, 100 mM NaCl, 0.1% Tween-20) at room temperature for 1 h and with diluted primary antibodies for overnight at 4 • C. Primary antibodies were detected with secondary antibodies conjugated with horseradish peroxidase (HRP) and visualized with a ChemiDoc TM MP Imaging System (Bio-Lab, United States) and Image Lab 5.1 software (Bio-Lab, United States). The detailed antibody information was provided in Supplementary Table 1.

Endothelial Function in vitro
Dil-Conjugated Acetylated Low-Density Lipoprotein (Dil-ace-LDL) Uptake and Tube Formation Dil-ace-LDL uptake and tube formation were evaluated as described previously Su et al., 2018). For DiI-ace-LDL uptake assay, hiPSC-ECs were incubated with DAPI overnight (1:1,000 dilution) and then in EGM supplemented with 10 µg/mL of DiI-ace-LDL (Life Technologies, United States) at 37 • C for 12 h. For tube-formation assay, cells were seeded in 48-well plate that had been coated with Matrigel (Corning, United States) and incubated at 37 • C for 24 h. Numbers of node, junction, and branches and total branches length per magnification (10×) were quantified using angiogenesis analyzer of Image J.

Nitric Oxide (NO) Produced by ECs
To assess NO produced, hiPSC-ECs on day 7 post-sorting were seeded in 12-well plate and cultured in 1 ml EGM for 48 h. Supernatant was collected and stored at -80 • C. Cells were washed with DPBS and lysed in 100 µl PhosphoSafe TM Extraction Reagent. The concentration of NO was determined using Nitric Oxide Detection Kit as per instruction (SKT-212, StressMarq Biosciences Inc., United States) and normalized by protein quantity and expressed as µM/mg.

Endothelin-1 Secreted by ECs
To determine secreted endothelin-1, a vasoconstrictor, hiPSC-ECs supernatant was collected for Western Blot analysis as described above. Endothelin-1 protein expression was presented as fold change after comparing with that of PCBC-EC which was considered as 100%. The detailed information on endothelain-1 antibody was provided in Supplementary Table 1.

Adhesion Molecule Expressed by ECs
Protein expression of adhesion molecule, including ICAM-1 and VCAM-1, was determined using Western Blot as described above. The protein expression level was normalized by GAPDH and expressed as percentage of GAPDH. The detailed information on ICAM-1, VCAM-1, and GAPDH antibodies was provided in Supplementary Table 1.

Senescence of ECs
To determine cell senescence, ECs on day 14 post-sorting were stained for β-galactosidase (β-gal) expression using Senescence β-galactosidase staining kit (98605, Cell Signaling, United States). EC lysates were used to assess the protein expression of p16, p21, and p53 using Western Blot. The protein expression level was normalized by GAPDH and expressed as percentage of GAPDH. The detailed information on p16, p21, and p53 antibodies was provided in Supplementary Table 1.

Mitochondrial Function of ECs
To determine EC mitochondrial membrane potential, after overnight cultured with EGM supplemented with DAPI, ECs on day-7 post-sorting were cultured with 1:1 ratio of fresh EGM and JC-1 dye solution (Mitochondria staining kit, Sigma-Aldrich, United States) for 30 min in an incubator at 37 • C. Then, cells were washed with DPBS and cultured in fresh EGM. Images of red fluorescence were randomly taken at 50 ms at 20× magnification using Olympus IX73 microscope and Cell Sens Standard software (Olympus, Japan). The fluorescence intensity of each cell was calculated as the overall fluorescence intensity divided by EC number in each image using Image J. Each cell line had three biological replicates and each replicate had 6-8 images analyzed.
To determine ATP synthesis efficiency, ECs were cultured in white 96-well plate. On the day of assay, cell culture medium was removed and 100 µl of ATPlite 1 step Luminescence Assay buffer (6016731, PerkinElmer, United States) was added into each well. The intensity of luminescence was determined using Infinite M200 (TECAN, Switzerland) and was normalized to cell number.

Intracellular Glycine Content
Methods of extraction and metabolic profiling for amino acid were described (Newgard et al., 2009;Muoio et al., 2012). ECs on day-7 post-sorting were washed DPBS and were scraped off wells using 300 µL of 0.6% formic acid (Sigma-Aldrich, United States). 100 µL mixture was used for protein content measurement. The remaining mixture was mixed with 200 µL acetonitrile (Sigma-Aldrich, United States). The amino acid extracts were derivatized with 3 M Hydrochloric acid in methanol or butanol (Sigma-Aldrich, United States), dried, and reconstituted in methanol for analysis in Liquid chromatography-mass spectrometry.
Methods of amino acid analysis were modified from Newgard et al. (2009). Briefly, amino acids were separated using a C8 column (Rapid Resolution HT, 4.5 × 50 mm, 1.8 µm, Zorbax SB-C8) on an Agilent 1290 Infinity LC system (Agilent Technologies, CA, United States) coupled with quadrupoleion trap mass spectrometer (QTRAP 5500, AB Sciex, DC, United States). Mobile phase A (10/90 Water/Acetonitrile) and Mobile phase B (90/10 Water/ Acetonitrile) both containing 10 mM of Ammonium formate were used for chromatography separation. The liquid chromatography run was performed at a flow rate of 0.6 mL/min with initial gradient of 20% B for 0.5 min, then ramped to 100% B in 2.5 min, maintained for 0.5 min, followed by re-equilibrated the column to the initial run condition (20% B) for 2 min. All compounds were ionized in positive mode using electrospray ionization. The chromatograms were integrated using MultiQuant 3.0 software (AB Sciex, DC, United States). The amino acid concentration was normalized by protein concentration and presented as µM/mg/ml.

Protein Quantification of Glycine Transporters (GlyT) and Serine Hydroxymethyltransferase (SHMT)
Endothelial cell lysates were used to assess the protein expression of GlyT1, GlyT2, cytoplasmic SHMT (cSHMT), and mitochondrial SHMT (mSHMT) using Western Blot. The protein expression level was normalized by GAPDH and expressed as percentage of GAPDH. The detailed information on GlyT1, GlyT2, cSHMT, and mSHMT antibodies was provided in Supplementary Table 1.

Paracrine Factor Profile of ECs
To assess angiogenic factors released from hiPSC-ECs, 2 × 10 5 hiPSC-ECs/well were seeded in 12-well plate and cultured in 1 ml endothelial basal medium (EBM) (Lonza, Switzerland) supplemented with 100x B27 for 48 h. The supernatant was collected and used to determine the profiles of angiogenic factors released from hiPSC-ECs using Human Angiogenesis Arrays Q3 kit (Raybio Tech., United States) as described .

Murine Hind-Limb Ischemia (HLI) Model and Treatment
The animal experimental protocol and procedures were approved by the Institutional Animal Care and Use Committee of Singapore Health Services Pte. Ltd. The animal model was developed as previously described (Su et al., 2018). Briefly, 8-week-old NOD-SCID mice (InVivos, Singapore) were anesthetized with 1.5-2% isoflurane and the femoral artery of the right limb was ligated with 6-0 polypropylene sutures. Animals were randomly assigned to treatment with 1.2 10 6 hiPSC-ECs in 0.1 mL EBM (the hiPSC-EC groups, n = 6 for each group) or with 0.1 mL EBM (the basal medium group, n = 6). The hiPSC-ECs differentiated from the 4 hiPSC lines or basal medium were administered 3 days after HLI induction via 4 intramuscular injections into the center of the ligated area and the surrounding region along the femoral artery.

Laser Doppler Imaging
Mice were anesthetized with 1.5-2% isoflurane, their hind limbs were shaved, and laser Doppler imaging was performed using a PeriScan PIM 3 System (Perimed, Sweden) as described (Su et al., 2018). Measurements in the ischemic (right) limb were normalized to measurements in the non-ischemic (left) limb and expressed as a percentage.

Immunohistochemistry
To identify transplanted hiPSC-ECs, a primary antibody specifically against human CD31 (hCD31, mouse anti-human CD31-Biotin) was used and visualized by mouse anti-Biotin-VioBright 515 (both from Miltenyi Biotec, Germany). Fluorescence images were taken with an Olympus IX71 fluorescence microscope.
Neovascularization in ischemic limb was determined as described (Su et al., 2018). Cryosections were stained for CD31 expression (rabbit anti-CD31, Abcam, United States), which targets both human and mouse ECs), to evaluate total vessel density, and smooth muscle actin (SMA) expression (Cy3conjugated mouse anti-SMA antibodies, Sigma-Aldrich), which targets SMCs, to evaluate arteriole density. Vascular structures that were positive for CD31 expression and for both CD31 and SMA expression were counted for all animals in each group.

Statistics
Data are presented as mean ± standard deviation (SD). Comparisons among groups were analyzed for significance via one-way analysis of variance (ANOVA) with the Tukey correction. Analyses were performed with SPSS software. A value of p < 0.05 was considered significant.

Characterization of hiPSCs
P1C1iPSCs, DP2-C8iPSCs, and DP3-C6iPSCs displayed typical morphological features of human embryonic stem cells (hESCs): colonies with a tightly packed appearance and a well-defined edge and cells have prominent nucleoli (Supplementary Figure 1A). They all expressed pluripotent cell markers, including Oct3/4 and TRA-1-60 (Supplementary Figures 1B,C). Karyotyping demonstrated that all three hiPSC lines had normal chromosome number and structure (Supplementary Figure 2A). Teratoma formation assay showed that all three hiPSC lines were able to form teratoma in NOD-SCID mice. Hematoxylin and eosin staining showed that mesoderm (cartilage), ectoderm (epidermis), and endoderm (gastrointestinal glands) lineage cells were formed in teratoma (Supplementary Figure 2B).

Disrupted Glycine Homeostasis in Dia-hiPSC-ECs
Metabolomic analysis of amino acid profiles showed that intracellular glycine was significantly lower in dia-hiPSC-ECs compared to healthy hiPSC-ECs ( Figure 3A). The intracellular glycine concentration in DP2-EC was 29.6 or 36.2% and DP3-EC was 45.2 or 55% of PCBC-ECs or P1C1-ECs. As intracellular  glycine content is maintained by uptake from the media by the transporters GlyT1 and GlyT2 and also by conversion from serine by cSHM and mSHMT, we further measured their protein expression levels. Western Blot showed that the major isoform of GlyT1 is GlyT1A (70.5 kDa), not GlyT1C (78 kDa). Protein expression levels of GlyT1A, GlyT2, and cSHMT were similar among four cells lines (Figures 3B-E).

Dia-hiPSC-ECs Had Slow Growth Rates and Increased Senescence
Cell doubling time increased by 26 and 30% in DP2-ECs and DP3-ECs, respectively, compared to PCBC-ECs and P1C1-ECs, indicating slower cell growth rate in dia-hiPSC-ECs than healthy hiPSC-ECs ( Figure 4A). A cell senescence assay showed that β-gal protein expression significantly increased by 145 and 137%, respectively, in DP2-EC and by 88.9 and 82.7%, respectively, in DP3-EC compared to PCBC-EC and P1C1-EC (Figures 4B,C).

dia-hiPSC-EC Have Poor Angiogenic Potential for Treatment of HLI
The potential of dia-hiPSC-ECs for treatment of ischemic disease was evaluated in a mouse model of HLI (Figure 7A; Su et al., 2018). Perfusion was barely detectable in mouse limbs of the basal medium group at 14 days post-treatment. However, perfusion was recovered to 45 or 35% in the ischemic limbs of the PCBC-EC or P1C1-EC group, which was not only significantly greater than the extent of recovery in the basal medium injected mice (5.7 ± 1%, p < 0.001 or p < 0.01), but also than that in the DP2-EC (15.8 ± 8%, p < 0.001 vs. PCBC-EC group) or DP3-EC group (12.7 ± 3.8%, p < 0.001 vs. PCBC-EC group or p < 0.05 vs. P1C1-EC group; Figures 7B,C).
Tissue sections stained for the specific hCD31 and SMA indicated that the transplanted hiPSC-ECs were present both in capillaries and smooth muscle cell coated arterioles in all animal groups that received hiPSC-EC transplantation (Figure 7D), suggesting that transplanted hiPSC-ECs can contribute to capillary and arteriole formation in ischemic limbs.

DISCUSSION
We characterized hiPSC-ECs derived from patients with T2DM and identified feature phenotypes in two dia-hiPSC-EC lines: disrupted glycine homeostasis, increased senescence, and impaired mitochondrial function and angiogenic potential.
hiPSCs generated human ECs not only can be used for treatment of ischemic diseases, but also can be used for modeling diseases and drug screening and discovering in vitro (Liu et al., 2018;Rowe and Daley, 2019;Williams and Wu, 2019). Although ECs differentiated from hiPSCs of pulmonary hypertension and Hutchinson-Gilford progeria syndrome have been studied (Zhang et al., 2011;Gu et al., 2017;Sa et al., 2017;Matrone et al., 2019), few studies have characterized the function of iPSC derived ECs in diabetes or metabolic syndrome in rodents (Takenaka-Ninagawa et al., 2014;Gu et al., 2015) and human (Vila Gonzalez et al., 2020). The endothelial dysfunction of mouse iPSC derived from obesity is due to decreased NO production (Gu et al., 2015). Only one study has investigated ECs derived from hiPSCs of patients with diabetic macular edema (Vila Gonzalez et al., 2020). The results of this work were able to recapitulate clinical evaluations of anti-VEGF responsiveness in vitro. Thus, feature phenotypes are largely unknown in hiPSC-ECs from patients with diabetes. In this study, we characterized dia-hiPSC-ECs and found that they not only recapitulated several phenotypes of endothelial dysfunction, such as increased vasoconstrictor (endothelin-1) and adhesion molecule (ICAM-1) production and impaired angiogenesis, but also showed disrupted glycine homeostasis, increased senescence, and impaired mitochondrial function.
Although decreased NO production is a feature phenotype of endothelial dysfunction, significantly decreased eNOS activity and NO production were only found in DP3-ECs in this study. However, endothelin-1 and ICAM-1 production significantly increased in both dia-hiPSC-EC lines. These results indicate that increased vasoconstrictor and adhesion molecule expression may be more common phenotypes than reduced NO production in dia-hiPSC-ECs.
One signature phenotype we found in this study is the disrupted glycine homeostasis, including low intracellular glycine content. This is consistent with clinical findings of serum glycine level which is lower not only in T2DM patients, but also in patients with impaired glucose tolerance (Wang-Sattler et al., 2012;Drabkova et al., 2015;Palmer N.D. et al., 2015). Serum glycine concentration was negatively associated with steady-state plasma glucose (Seibert et al., 2015) and has been identified as a serum biomarker of both insulin sensitivity and regional fat mass in functionally limited older adults (Lustgarten et al., 2013). Low intracellular glycine level may compromise vascular endothelial growth factor signaling in ECs and lead to poor angiogenic potential and mitochondrial function (Guo et al., 2017). The observed low intracellular glycine content may be due to low protein expression levels of mSHTM. Intracellular glycine homeostasis is modulated by GlyT (GlyT1 and GlyT2) and SHMT (cSHMT and mSHMT). Except for mSHMT, the other three protein expression levels are similar between the healthy hiPSC-ECs and dia-hiPSC-ECs. Future studies are needed to explore the mechanism of how disrupted intracellular glycine homeostasis causes endothelial dysfunction in dia-hiPSC-ECs for a better understanding the role of glycine homeostasis in endothelial function in diabetes.
Increased senescence is found in dia-hiPSC-ECs. Feature characteristics of senescent dia-hiPSC-ECs include up-regulated β-gal, p21, and p53 protein expression and increased secretion of MMP-1 and VEGF which are components of a SASP (Coppe et al., 2010), and impaired mitochondrial function. Mitochondrial dysfunction is composed of up-regulated Fis-1 protein expression, decreased mitochondrial membrane potential, ATP synthase protein expression, and ATP production in dia-hiPSC-ECs. Although senescence may be a causal role in mitochondrial dysfunction (Korolchuk et al., 2017), it may be that mitochondrial dysfunction itself leads to senescence (Wiley et al., 2016). Increased senescence in dia-hiPEC-ECs may be due to the high protein expression levels of p53 and p21 as the p53/p21 pathway leads to the inhibition of Cdk4/6 activity and contributes to senescence (Sharpless and Sherr, 2015). Poor cell viability associated with senescent dia-hiPSC-ECs may result in poor angiogenic potential in vitro and in vivo.
One limitation of the current study is the number of hiPSC lines used. We only used 2 hiPSC lines each of healthy subjects and patients with T2DM. Although robust difference was observed in glycine homeostasis, endothelial senescence, mitochondrial function, angiogenic potential in dia-hiPSC-ECs as compared with healthy hiPSC-ECs, they may be limited in 2 dia-hiPSC-EC lines used in this study.
In conclusion, we identified feature phenotypes in dia-hiPSC-ECs: disrupted glycine homeostasis, increased senescence, impaired mitochondrial function and angiogenesis. Efforts are needed to find the key causal factors and the molecular mechanisms contributing to the onset and development of disrupted glycine homeostasis and senescence in dia-hiPEC-ECs. The features identified here may serve as targets for new therapeutic strategies for the treatment of endothelial dysfunction in diabetes.