Modeling diabetic endothelial dysfunction with patient‐specific induced pluripotent stem cells

Abstract Diabetes is a known risk factor for various cardiovascular complications, mediated by endothelial dysfunction. Despite the high prevalence of this metabolic disorder, there is a lack of in vitro models that recapitulate the complexity of genetic and environmental factors associated with diabetic endothelial dysfunction. Here, we utilized human induced pluripotent stem cell (iPSC)‐derived endothelial cells (ECs) to develop in vitro models of diabetic endothelial dysfunction. We found that the diabetic phenotype was recapitulated in diabetic patient‐derived iPSC‐ECs, even in the absence of a diabetogenic environment. Subsequent exposure to culture conditions that mimic the diabetic clinical chemistry induced a diabetic phenotype in healthy iPSC‐ECs but did not affect the already dysfunctional diabetic iPSC‐ECs. RNA‐seq analysis revealed extensive transcriptome‐wide differences between cells derived from healthy individuals and diabetic patients. The in vitro disease models were used as a screening platform which identified angiotensin receptor blockers (ARBs) that improved endothelial function in vitro for each patient. In summary, we present in vitro models of diabetic endothelial dysfunction using iPSC technology, taking into account the complexity of genetic and environmental factors in the metabolic disorder. Our study provides novel insights into the pathophysiology of diabetic endothelial dysfunction and highlights the potential of iPSC‐based models for drug discovery and personalized medicine.

complications.This work bridges the gap between preclinical animal studies and human clinical trials, potentially improving patient outcomes and guiding the clinical management of diabetic vascular complications.

| INTRODUCTION
Diabetes mellitus is a growing public health problem with known contributions to lifestyle and genetics. 1In the United States alone, its prevalence has reached about 34 million people as of 2020, with another 88 million people diagnosed as prediabetic. 2The recent COVID-19 pandemic has posed a substantial risk for those living with diabetes, resulting in significantly increased disease severity and higher mortality after the coronavirus infection. 3Diabetes mellitus is associated with increased risks of cardiovascular diseases and vasculopathy, including macrovascular (e.g., coronary and peripheral artery disease) and microvascular complications (e.g., diabetic foot ulcer and retinopathy). 4Despite significant recent advances in blood glucose monitoring and glycemic control, debilitating vascular complications remain in most diabetic patients. 5Chronic exposure to hyperglycemia, hyperlipidemia, and oxidative stress presented in diabetes is believed to cause vascular damage that is persistent even after glucose normalization. 6Unfortunately, no specific clinical treatment currently exists to improve these cardiovascular outcomes.There is, thus, an urgent need to develop effective therapeutics to restore vascular functions for those living with diabetes.
Although the detailed mechanisms of the pathogenesis of diabetic vascular diseases are still not completely understood, all mechanisms converge on the endothelium as an important disease target. 7,8The vascular endothelium serves several functions for the maintenance of vascular tone, mechanotransduction, injury repair, metabolism, barrier integrity, metabolism, angiogenesis, and hemostasis. 9,10Endothelial dysfunction-characterized by a reduction of nitric oxide bioavailability, an increase in reactive oxygen species (ROS) and a proinflammatory, pro-thrombotic milieu-is a key marker in all forms of diabetic vascular complications. 11,12To date, a number of preclinical animal models have been developed for diabetes and diabetic vascular complications. 13However, animal models do not always faithfully recapitulate the human condition, which has contributed to the lack of clinical translational success in this area. 14Development of human disease models that are representative of diabetic endothelial dysfunction in vitro will complement the current preclinical animal models and facilitate therapeutic discovery to restore the endothelial function in diabetes.
The advent of induced pluripotent stem cell (iPSC) technology offers an exceptional opportunity for creating disease-and patientspecific cellular models, investigating underlying mechanisms, and optimizing therapy. 15Our study aims to develop and validate in vitro models of diabetic endothelial dysfunction using iPSC-derived endothelial cells (iPSC-ECs) from diabetic and healthy individuals to dissect the underlying mechanisms and identify new therapeutic targets.
Using iPSC technology, we generated patient-specific ECs from healthy subjects and diabetic patients, evaluated their baseline endothelial functions, and identified molecular signatures of endothelial dysfunction in diabetes with transcriptome analysis.Furthermore, we investigated the role of a diabetogenic treatment in inducing the diabetic phenotype in healthy and diabetic ECs.Lastly, using the patientderived iPSC-ECs as in vitro disease models, we performed drug screening and identified angiotensin receptor blockers (ARBs) that improved endothelial function in vitro for each patient.The findings from our study may provide new insights into the pathophysiology of diabetic endothelial dysfunction and contribute to the development of personalized therapies for diabetic cardiovascular complications.

| ECs could be generated from diabetic patients and healthy subjects with iPSC technology
To generate patient-specific ECs, we reprogrammed the peripheral blood mononuclear cells (PBMCs) from three diabetic patients and three age-matched healthy subjects (Table 1) into iPSCs, followed by directed differentiation into ECs under xeno-free conditions (Figure 1a).All human subject studies have been approved by the Northwestern University Internal Review Board (IRB), and written informed consents were obtained from each participant prior to the T A B L E 1 Human subject information.S1).DNA fingerprinting verified that each patient-specific iPSC line was derived from each individual parental PBMC line without cross-contamination (Table S1).All iPSC lines were able to form embryoid bodies in vitro, and the differentiated cells expressed markers for the endoderm (Alpha Fetoprotein), mesoderm (Desmin), and ectoderm (ß-3-Tubulin).
(Figure 1c) There was no characteristic difference observed between the HiPSCs and DiPSCs.
ECs were differentiated from HiPSCs and DiPSCs following a previously published protocol, 16 which resulted in similar differentiation efficiencies (HiPSC-ECs: 46.4% ± 7.8%; DiPSC-ECs: 51.6 ± 6.9%, p = 0.4307), assessed via flow cytometry for VE-Cadherin + and PECAM-1 + cells (Figure 1d).VE-Cadherin + cells were isolated via magnetic-assisted cell sorting (MACS) and cultured in EC medium for subsequent experiments.All iPSC-ECs exhibited typical endothelial cobblestone morphology and expressed endothelial markers VE-Cadherin, PECAM-1, and Flk-1 (Figure 1e).Western blotting verified the expression of the endothelial marker VE-Cadherin in both HiPSC-ECs and DiPSC-ECs, as well as the expression of the pluripotency marker ALP in both HiPSCs and DiPSCs (Figure 1f).DNA fingerprinting confirmed that each patient-specific iPSC-EC line was derived from the same parental PBMC and iPSC lines (Table S1).

| DiPSC-ECs exhibit baseline endothelial dysfunction in the absence of a diabetogenic microenvironment
We next evaluated endothelial functions for HiPSC-ECs and DiPSC-ECs at the baseline level under normal EC culture conditions at early passages (passages 2-3).DiPSC-ECs from all diabetic patients in the study expressed inflammation-related markers, including VCAM-1, ICAM-1, and P-Selectin, in contrast to HiPSC-ECs from healthy subjects, which did not express those proteins (Figure 2a). of the HiPSC-ECs (Figure 2c).In contrast, platelets aggregated on a subpopulation of DiPSC-ECs from diabetic patients, suggesting compromised thrombo-resistant functions of the cells.A quantitative analysis revealed a significantly higher percentage of platelet-adherent cells from DiPSC-ECs than HiPSC-ECs ( p = 0.0056) (Figure 2d).A tubulogenesis assay with HiPSC-ECs seeded onto a basement membrane hydrogel (Matrigel™) formed interconnected tubular networks (Figure 2c), indicating functional angiogenetic potential.However, DiPSC-ECs seeded onto Matrigel™ formed disorganized networks with a significantly lower number of master junctions, when compared with HiPSC-ECs ( p = 0.0028) (Figure 2c,e).DiPSC-ECs also exhibited significantly higher amount of intracellular ROS when compared to HiPSC-ECs ( p = 0.0024) (Figure 2f).However, no significant difference was detected in nitric oxide production with an intracellular DAF2-DA assay (Figure 2c) or an extracellular nitrite assay ( p = 0.6056).Lastly, monolayers formed with DiPSC-ECs exhibited disrupted barrier function and thus increased permeability, which allowed for a significantly higher amount of HRP diffusion, when compared with HiPSC-EC monolayer ( p < 0.0001) (Figure 2g).All experiments were performed under a normal endothelial culture condition in the absence of a diabetic-like culture condition.

| Transcriptome-wide differences exist between cells from healthy and diabetic patients
We wondered if there were any global transcriptional differences between cells isolated from healthy individuals and diabetic patients.
To answer this, we isolated RNA and performed RNA sequencing of iPSCs and Ipsc-ECs from the two groups (total 24 samples; see Methods 4.8) for identifying transcriptome-wide differences.We first asked whether we see differences between the iPSCs and iPSC-ECs differentiated populations for both healthy and disease patients.
Indeed, with both principal component analysis (PCA) (Figure S2) and a conservative filtering of differential expressed genes (log2FC > 2 and p adj < 0.01), we found a total of 3464 genes being differentially expressed (Figure 3a; Figure S3; Table S2).Importantly, as expected, our Gene Ontology (GO) analysis 17 revealed that these genes upregulated in differentiated samples corresponded to angiogenesis, vascular development, and muscle tissue development (Figure 3b,c).This result establishes the power and validity of our bulk sequencing approach.
We next asked what kind of transcriptome-wide differences exist between cells from healthy individuals and diabetic patients.We performed differential expression analysis separately for iPSCs and iPSC-ECs and found significant differences between cells from healthy individuals vs. diabetic patients (Figure 3d,e; Figure S4; Table S3).For example, with a log2FC > 2 and p adj < 0.01, we found 245 (and 378 for a log2FC > 1) total genes differentially expressed between HiPSC-ECs and DiPSC-ECs (Figure 3d).Our GO analysis revealed that consistent with our protein-level analysis, these genes corresponded to extracellular matrix structure and organization, among others (Figure 3e).Similarly, we also found differences, albeit to a lesser degree (total 37 different genes log2FC > 2 and p adj < 0.01), between HiPSCs and DiPSCs (Figure 3f; Figure S5; Table S4).Together, our experiments reveal extensive transcriptomewide differences between cells from healthy and diabetic patients.TNF-α (20 pg/mL).These conditions were chosen based on clinical serological studies of diabetic patients. 18,19After the treatment, HiPSC-ECs exhibited significant upregulation of adhesion molecules, including ICAM-1, VCAM-1, and P-Selectin (Figure 4a), in comparison to HiPSC-ECs under normal culture condition (Figure 2a).However, DiPSC-ECs exhibited limited response after the diabetogenic treatment and maintained the expression of inflammatory markers (Figure 4a), similar to DiPSC-ECs at the baseline level (Figure 2a).
Quantitative measurement of ICAM-1 expression with in-cell ELISA demonstrated a significant elevation of ICAM-1 expression after the diabetogenic treatment for HiPSC-ECs (Figure 4b).However, no significant difference was detected in DiPSC-ECs before and after the diabetogenic treatment (Figure 4c).Similar results were observed for expression of VCAM-1 and P-selectin (Figure 4d-g).
Cell viability after the diabetogenic treatment decreased significantly for HiPSC-ECs (Figure 4h), but not for DiPSC-ECs (Figure 4i).This suggests that HiPSC-ECs are more sensitive toward the environmental change while DiPSC-ECs have been pre-conditioned to tolerate the diabetic-like microenvironment.
Lastly, diabetogenic treatment resulted in significant variations in ROS generation for both HiPSC-ECs (Figure 4j) and DiPSC-ECs (Figure 4k), without statistically significant difference relative to the baseline ROS.Overall, HiPSC-ECs were more responsive toward the diabetogenic treatment with significant changes in several of the evaluated parameters compared to the baseline, while DiPSC-ECs maintained a similar status as the baseline level without significant response to the changes.

| Endothelial function of DiPSC-ECs could be improved by individualized pharmacological treatment
Next, we explored the use of patient-derived DiPSC-ECs as in vitro disease models for diabetic endothelial dysfunction to identify potential pharmacological agents for their functional improvement.
Given that DiPSC-ECs all exhibited significantly elevated adhesion molecule expression relative to HiPSC-ECs, we performed a small-scale drug screening for compounds that have been reported to improve EC functions: pioglitazone (a PPAR-γ agonist), 20 telmisartan (an angiotensin receptor blocker), 21 valsartan (an angiotensin receptor blocker), 22 heparin (an anticoagulant), 23 aspirin (a nonsteroidal anti-inflammatory drug and blood thinner), 24 and resveratrol (an antioxidant polyphenol). 25 5c).Additionally, DiPSC-ECs exhibited significantly improved tubulogenesis after each treatment (Figure 5d), including the number of master junctions (Figure 5e) and other parameters evaluated (Figure S7).However, no significant improvement was found in other endothelial functions of DiPSC-ECs after the pharmacological treatment, such as barrier function or antiplatelet adhesion.

| DISCUSSION
In this study, we investigated the feasibility of using iPSC technology in combination with diabetogenic treatment to generate in vitro disease models of diabetic endothelial dysfunction.We show that iPSC-ECs can be successfully generated from the blood of healthy and diabetic patients, where DiPSC-ECs exhibit impaired endothelial tubulogenic, anti-thrombotic, and barrier functions with elevated inflammatory status and ROS production, in contrast to the normal endothelial functions of HiPSC-ECs.This dysfunctional phenotype of DiPSC-ECs is consistent with the pathophysiology of various diabetic vascular complications, which typically exhibit low angiogenic potential, high risk of thrombosis, and increased inflammation. 26It is likely that the functional impairments of DiPSC-ECs observed in the study reflect the clinical status of the patients at the time of the blood donation.Our results are also consistent with a recent study by Su et al.,   where ECs differentiated from the iPSCs of two adult patients with type 2 diabetes also displayed disrupted glycine homeostasis, increased senescence, and impaired mitochondrial function and angiogenic potential, in comparison to iPSC-ECs derived from two healthy individuals. 27Taken together, these studies suggest that diabetic endothelial dysfunction can be successfully recapitulated in vitro with iPSC technology.
Previous in vitro models of diabetic endothelial dysfunction mostly utilized hyperglycemia or other diabetic-like conditions to stimulate healthy ECs to induce a diabetic-like phenotype. 28Intriguingly, our results indicate that a diabetic-like microenvironment is not necessary to induce diabetic-like phenotypes for DiPSC-ECs.This is also consistent with a previous study of diabetic cardiomyopathy by Drawnel et al., where cardiomyocytes derived from diabetic patientspecific iPSCs exhibited a cardiomyopathic phenotype in the absence of a diabetic stimulus. 29These results suggest that cells derived from diabetic patients with iPSC technology are genetically and/or epigenetically predisposed to dysfunction related to diabetes.Genetically, meta-analyses of genome-wide association studies (GWAS) have identified more than 240 loci that are associated with type 2 diabetes (T2D). 30Those genetic variations, if present in the diabetic patients in our study, will be retained during the reprogramming and differentiation processes from patient blood into DiPSC-ECs.Additionally, glycemic memory-associated epigenetic changes such as DNA methylation have been postulated to play a role in the pathogenesis and progression of diabetic complications. 31Previous studies have indicated that iPSCs reprogrammed by Yamanaka factors harbor residual DNA methylation signatures characteristic of their somatic tissue of origin. 32,33Therefore, it is likely that iPSCs generated from diabetic patients in our study also retain some of their epigenetic memory as a consequence of epigenetically modified gene-expression patterns.Future studies will investigate the roles of genetic and epigenetic factors contributing to the diabetic endothelial dysfunction of DiPSC-ECs.Our analysis with bulk RNA sequencing reveals several transcriptional differences between healthy and disease patients, both in iPSCs and iPSC-ECs.Since even iPSCs from a single individual can result in highly heterogeneous differentiated populations, 34 future single-cell studies can unravel within-sample and across-sample differences between individual stem or differentiated cells within a population.
In this study, we also recreated the complex microenvironment of diabetes in vitro by exposing iPSC-ECs to hyperglycemia, elevated inflammation, and hyper-urea levels, which are characteristic features of the diabetic condition.Although previous models have primarily used hyperglycemia as the only condition replicated, our approach more closely reflects the complex nature of the diabetic microenvironment.For example, approximately one in three adults with diabetes have been diagnosed with chronic kidney disease, 35 resulting in elevated levels of blood urea above 25 mg/dL. 18Chronic inflammation and elevated levels of TNFα have been shown to be associated with diabetes. 36It was found that diabetic patients have significantly higher serum levels of the inflammatory TNFα. 36Therefore, to recapitulate the complexity of diabetic microenvironment, we have incorporated hyperglycemia, hyper-urea, and elevated inflammation in our system.Interestingly, we found that an acute 5-day exposure to diabetic conditions was sufficient to induce a dysfunctional phenotype in healthy iPSC-ECs, while the same treatment did not affect the already dysfunctional diabetic iPSC-ECs, suggesting potential epigenetic memory.Our approach of incorporating multiple relevant disease conditions may provide a more accurate model for understanding the complex pathophysiology of diabetic endothelial dysfunction and aid in the identification of novel therapeutic targets.
We also show that DiPSC-ECs could be used as an in vitro screening platform for diabetic endothelial dysfunction.Suppression of the pro-inflammatory phenotype was used as a screening endpoint to identify molecules that improve the endothelial function of DiPSC-ECs.Currently, no specific drug treatment exists to prevent or rescue diabetes-induced cardiovascular complications clinically.
Clinical treatment options typically aim to establish glycemic control and delay the progression of the disease. 37As endothelial dysfunction is a key mediator for diabetic vascular complications, therapeutics that aim to restore endothelial function in diabetes could potentially improve the clinical outcomes for those patients.To that end, a number of promising therapeutic candidates have been identified in preclinical studies using various animal models of diabetes.
However, many of the promising therapeutic candidates with success in preclinical evaluations have failed to show efficacy in human randomized control trials.One possible reason for this inability to translate findings to patients could be due to the inability of the preclinical animal models to recapitulate the clinical condition.In vitro humanized disease models, known as "disease-in-a-dish," have a strong potential to fill the gap between preclinical studies and clinical trials.The iPSC technology enables creating disease-specific cellular models, investigating underlying mechanisms surrounding different diseases, and optimizing therapy and treatment options for patients. 38 demonstrated by our study, DiPSC-ECs can serve as in vitro human models of diabetic endothelial dysfunction.Although our current study only uses a limited amount of patient samples (n = 3), we have shown the functional differences between healthy control subjects and diabetic patients are consistent across all the patient samples analyzed.As iPSCs have the ability to continuously proliferate, this will also provide a viable cell source for large-scale, high-throughput screening in future drug discovery.Furthermore, we identified individualized treatment with the ability to improve the phenotype and function of DiPSC-ECs from each patient at least partially.Diabetes is a complex, polygenic condition with known contributions of lifestyle and genetics and significant variations among individual patients.Therefore, the traditional "one-size-fitsall" approach may not be feasible in addressing the challenge.
Precision medicine approaches such as pharmacogenomics that are common now in cancer treatment are likely needed for safe and effective treatment of diabetic vascular complications in the future.
We acknowledge that our study is based on a small cohort of diabetic patients, with an unbalanced gender distribution.Further studies with a larger cohort of patients are needed to validate our findings and to generalize the results to a broader diabetic human population.
Interestingly, the pharmacological agents identified in our study to significantly improve endothelial function in DiPSCs-ECs, namely valsartan and telmisartan, are both classified as Angiotensin II (AngII) receptor blockers (ARBs).AngII (Asp-Arg-Val-Tyr-Ile-His-Pro-Phe) is a peptide hormone that is part of the renin-angiotensin-aldosterone system (RAAS) to regulate blood pressure.ECs express AngII receptor type I (AGTR1), which upon activation, initiates an inflammatory cascade of NAD(P)H oxidase activation and mitochondrial ROS generation, leading to increased expression of adhesion molecules (e.g., VCAM-1, ICAM-1). 39Therefore, pharmacological blockade of AT1 with valsartan or telmisartan in DiPSC-ECs is expected to effectively reduce VCAM-1 expression, as demonstrated in our study.
ARBs have been used in diabetic patients to treat hypertension and to reduce the rate of diabetic nephropathy progression. 40However, ARBs are not routinely used to treat other vascular complications of diabetes, such as ischemic heart disease, stroke, retinopathy, and diabetic foot ulcers.It is likely that the administration of ARBs could be beneficial to diabetic patients with various vascular complications, given our results in the reduction of inflammatory markers and the improvement of tubulogenesis in DiPSC-ECs.For example, a recent clinical report indicates that ARB use is independently and positively associated with wound healing in patients with diabetic foot ulcers. 41evious preclinical studies also demonstrated the topical application of valsartan accelerated wound healing in diabetic mice and porcine models. 42Taken together, these evidence suggest a potential benefit of expanding the use of ARBs to improve vascular functions in diabetes.

| Statement of novelty
While the reprogramming and differentiation of iPSCs from diabetic patients have been previously reported in the literature, 27 our study contributes to the field by providing new insights and data that expand upon previous findings.First, we have performed transcriptome analysis of iPSCs and iPSC-ECs to compare the global gene expression between diabetic and healthy cells, which has not been reported in previous studies to the best of our knowledge.Our analysis revealed several differentially expressed genes that are involved in endothelial dysfunction and diabetes-related pathways, which provide insights into the mechanisms underlying diabetic vascular complications.Moreover, we have also used diabetic iPSC-ECs as a disease model for drug screening and identified individualized pharmacological treatments for each patient cell line.This approach provides a more personalized and precise approach to drug discovery for diabetic vasculopathy, which has not been explored in previous studies.This information could be valuable for developing new therapeutic strategies for diabetic vasculopathy.
All antibody manufacturer information and dilutions are listed in Table S1.Oligos for Illumina ® (Dual Index Primers Set 1) (E7600S), as described previously. 43,44Prior to library preparation, the samples were randomized to avoid experimental and human biases.We started our experiments with a total of 24 samples (iPSCs and differentiated iPSCs-ECs) from healthy and disease patients.During library preparation, two DiPSC and 2 HiPSC-EC patient samples were excluded from sequencing due to insufficient RNA isolation.

| Bulk sequencing and analysis
RNA-seq reads were pseudo-aligned to the human transcriptome (GRCH38 cDNA) utilizing Kallisto v0.48.0 to generate estimated gene count matrices. 45Count matrices were normalized and differential gene expression analyses were done with DESeq2 v1.38.3 using R v4.2.3 for each subgrouping comparison. 46The count matrix was further filtered for low expression counts such that any gene less than a total of 10 counts across samples cutoff were removed from downstream analysis.To compare between samples, differential gene expression data was extracted.The top 50 most up-and downregulated genes (total 100, unless the number of differentially expressed genes were <100) were filtered utilizing an ordered Log2FC gene list and then used to generate corresponding comparisons within heat maps.GO analysis was performed using clusterProfiler in R. Scripts for all extraction and downstream differential expression analysis are provided at the Goyal laboratory GitHub: https://github.com/GoyalLab/Bulk-RNA-Jiang-Analysis.git.
Raw data are uploaded to the NCBI GEO repository with accession number GSE236430 and are available to the public.

| Western blotting
Cells were grown on 6-well plates and lysed in RIPA buffer supplemented with protease inhibitors.Total protein concentration was determined using a BSA assay kit.Equal amounts of protein per sample (5 μg) were separated on a 4%-12% Bis-Tris gel by electrophoresis and transferred onto a nitrocellulose membrane.The membrane was then blocked with BSA and incubated with the primary antibodies overnight at 4 C.After washing with PBS-T, the membrane was incubated with secondary antibodies conjugated to horseradish peroxidase (HRP) for an hour at room temperature.
Protein bands were visualized after reacting with a chemiluminescent substrate and imaged with Azure c300 Gel Imager (Azure Biosystems, Dublin, CA).Quantitative analysis was performed with ImageJ.All antibody information and dilutions are listed in Table S1.

| Cell viability assessment
A resazurin-based assay was used to assess cell viability after diabetic microenvironment treatment.On day 5 of the diabetic microenvironment treatment, iPSC-ECs were incubated with resazurin (0.025 mg/mL, Sigma, R7017) for 1.5 h before fluorescence was read at 560/590 nm on the Cytation 3 microplate reader (BioTek).Readings for quantitative assays were adjusted to the cell viability per experiment.

|
In-cell enzyme-linked immunosorbent assay (ELISA) For quantitative analysis of cellular expression of proteins in iPSC-ECs, the cells were seeded onto a 96-well plate with fibronectin coating at 5000 cells/well.After treatments, the cells were fixed with 4% formaldehyde, permeabilized with 0.1% Triton-X and blocked with 5% FBS.Primary antibodies were incubated with each well at 4 C while the secondary antibody with HRP conjugate was incubated at room temperature for 1 h.3,3 0 ,5,5 0 -tetramethylbenzidine (TMB, 100 μL/well) was added to each well for 20 min, before stopped with 0.16 M sulfuric acid (50 μL/well).Absorbance at 450 nm was read with Cytation 5.All readings were normalized to cell numbers based on resazurin assay.

| Platelet adhesion
Human platelets were isolated from anticoagulant whole blood samples of healthy subjects, and diluted to 10 8 platelets mL À1 in platelet suspension buffer as previously described. 47The platelet suspension was incubated at 37 C for 30 min with monolayers of patient-specific iPSC-ECs.The adherent platelets were probed with platelet-specific antibody CD41, while the patient ECs were counter-stained with CD144 and Hoechst 33258.The images were acquired with Cytation 5 Cell Imaging Multi-Mode Reader (BioTek Instruments, Winooski, VT) and analyzed with Gen5 Data Analysis Software (BioTek).A total of four images were acquired for each well, and a total of 4 wells were analyzed for each patient (n = 4).

| Permeability assay
A transwell permeability assay was used to gain a quantitative understanding of the iPSC-EC monolayer's barrier function.iPSC-ECs were seeded directly onto transwell cell culture inserts (0.4 um pore size) at a seeding density of 50,000 k/cm 2 .Horseradish peroxidase (HRP, 0.125 μM, ProspecBio) was added to the top chamber of transwell insert and incubated with the endothelial monolayer for 45 min, after which samples were obtained from the lower chamber to react with TMB solution.Absorbance was read at 450 nm after adding stop solution.
Manufacturer recommended protocol was closely followed.Briefly, CM-H 2 DCFDA was diluted to a final concentration of 10 μM in prewarmed PBS supplemented with calcium and magnesium for 60 min, before obtaining fluorescent measurements at 485/528 nm.

| Nitric oxide generation assay
The intracellular production of nitric oxide (NO) was measured with the nitric oxide indicator, DAF-FM diacetate (4-amino-5-methylamino-2 0 ,7 0 -difluorescein diacetate).Manufacturer recommended protocol was closely followed.Briefly, DAF-2-DA (1 μM in PBS) was incubated with 50,000 iPSC-ECs/well in a 24-well plate for 30 min, and the fluorescence images were taken at 485/528 nm.To measure nitric oxide released to the extracellular environment, the cell culture medium (24 h after cell seeding) was collected, and the nitrate was converted to nitrite with nitrate reductase/NADPH.Total nitrite concentration was detected with a Measure-IT High-Sensitivity Nitrite Assay Kit and normalized to cell number using a Quant-iT PicoGreen dsDNA Assay Kit, following manufacturer's instructions.expression and were used to treat each cell lines for 48 h.Tubulogenesis was re-evaluated with and without optimized drug treatment as previously described.

| Statistical analysis
Three subjects (n = 3, biological replicates) were included in each group.For each experiment, n ≥ 3 experimental replicates were carried out per biological replicate.All data were plotted and analyzed with GraphPad Prism 9.2.0 (GraphPad Software, San Diego, CA, USA).
For all statistical analysis, a normality test was performed to confirm normal distribution prior to statistical comparison.Nested 2-tail t-tests were used to compare between the diabetic patient group and the healthy control group, and p < 0.05 was considered statistically significant.Paired t-tests were used to compare the diabetic microenvironment treatment to untreated cells within the same cell line, and p < 0.05 was considered statistically significant.For the drug screening study, a heatmap was plotted after normalization to the baseline.
One-way ANOVA with Tukey post hoc test were used for each cell line individually, and p < 0.05 was considered statistically significant.

| CONCLUSIONS
In conclusion, we have evaluated the function of both healthy and diabetic patient-derived iPSC-ECs under normal and diabetic conditions while considering inflammation, oxidative stress, permeability, and thrombosis.Overall, we observed a pro-inflammatory and prothrombotic phenotype in cells derived from diabetic patients, with implications in the progression of atherogenesis and thrombosis.
Exposure to a diabetic-like microenvironment induced a phenotypic surrogate of diabetic endothelial dysfunction in cells obtained from healthy subjects but did not have an effect on the cells derived from diabetic patients.Our in vitro disease models, which consider the complexity of genetic and environmental factors in diabetes, provide a screening platform for identifying candidate molecules that can rescue the disease phenotype.Future work will focus on expanding the drug screening to a high-throughput platform using our in vitro models to identify therapeutics that could restore or prevent the dysfunctional phenotype.Moreover, our disease model will be leveraged Western blotting study verified the elevated expression of VCAM-1, ICAM-1, and P-Selectin in DiPSC-ECs relative to HiPSC-ECs (Figure 2b).Interaction of HiPSC-EC monolayers with healthy human platelets exhibited limited platelet adhesion, indicating functional anti-thrombogenic activity F I G U R E 1 Generation and characterization of iPSCs and iPSC-ECs from healthy subjects and diabetic patients.(a) Schematic representation of the general process of iPSC-EC generation from the patient blood samples using iPSC technology; (b) Representative images of iPSC colonies derived from three age-matched healthy control subjects (HiPSCs) and three diabetic patients (DiPSCs) expressing pluripotency markers OCT4, SSEA4, and TRA-1-60; (c) Representative images of embryoid bodies formed from HiPSCs and DiPSCs in vitro expressing markers of the three germ layers: Alpha Fetoprotein (endoderm), Desmin (mesoderm) and β-3-Tubulin (ectoderm); (d) Flow cytometry analysis after EC differentiation from HiPSCs and DiPSCs for EC markers VE-Cadherin and PECAM-1 prior to cell soring with no statistically significant difference in EC differentiation efficiency; (e) MACS sorted HiPSC-ECs and DiPSC-ECs exhibiting cobblestone morphology and expressing endothelial markers VE-Cadherin, PECAM-1 and Flk-1; (f) Representative western blots and quantitative analysis of iPSCs (healthy and diabetic) expressing alkaline phosphatase (ALP) and iPSC-ECs (healthy and diabetic) expressing VE-Cadherin.Scale bar = 100 μm.
Functional assessment of HiPSC-ECs and DiPSC-ECs at the baseline level.(a) Representative images of HiPSC-ECs staining negative for adhesion molecules (ICAM-1, VCAM-1 and P-Selectin) while DiPSC-ECs staining positive for the adhesion molecules; (b) Representative western blots and quantitative analysis of DiPSC-ECs expressing elevated levels of ICAM-1, VCAM-1 and P-Selectin relative to HiPSC-ECs; (c) Representative images of HiPSC-ECs with low platelet adhesion, organized tube formation, and intracellular nitric oxide production, whereas DiPSC-ECs exhibiting high platelet binding, disorganized tubulogenesis, but still with intracellular nitric oxide production; (d) Quantification of platelet adherent cells indicating DiPSC-ECs exhibit higher thrombogenicity than HiPSC-ECs; (e) Quantification of master junction density in tubulogenesis analysis indicating DiPSC-ECs exhibit lower angiogenic potential than HiPSC-ECs; (f) Quantification of intracellular ROS production with DCF-DA assay indicating DiPSC-ECs exhibit higher oxidative stress than HiPSC-ECs; (g) Quantification of permeability by an HRP diffusion assay indicating DiPSC-ECs have higher permeability than HiPSC-ECs thus impaired barrier function.Scale bar = 200 μm.*p < 0.05.**p < 0.01.n.s.not significant.

2. 4 |
Exposure to a diabetogenic environment induces endothelial dysfunction in HiPSC-ECs without affecting DiPSC-ECsTo further evaluate the role of a diabetic-like microenvironment on the endothelial function of HiPSC-ECs and DiPSC-ECs, we treated the cells for 5 days with a diabetogenic medium that contains high glucose (25 mM), urea nitrogen (9 mM), and inflammatory cytokine F I G U R E 3 RNA-seq differential gene expression analysis comparing iPSCs and iPSC-ECs between healthy and diabetic groups.(a) Heatmap of the 100 most up-and downregulated genes comparing iPSCs and iPSC-ECs; (b,c) Gene ontology (GO) analysis of both upregulated and downregulated genes between iPSCs and iPSC-ECs; (d) Heatmap of the 37 most up-and downregulated genes comparing HiPSC-ECs and DiPSC-ECs; (e) Joint GO analysis of both upregulated and downregulated genes between HiPSC-ECs and DiPSC-ECs; (f) Heatmap of the 37 most upand downregulated genes comparing HiPSCs and DiPSCs.

4
Characterization of HiPSC-ECs and DiPSC-ECs after exposure to a diabetogenic extracellular microenvironment.(a) Representative images of HiPSC-ECs and DiPSC-ECs staining positive for adhesion molecules (ICAM-1, VCAM-1 and P-Selectin) after 5-day diabetogenic extracellular treatment; (b) Elevated ICAM-1 expression observed in HiPSC-ECs after the diabetic treatment; (c) No significant change in ICAM-1 expression observed in DiPSC-ECs after the diabetic treatment; (d) Elevated VCAM-1 expression observed in HiPSC-ECs after the diabetic treatment; (e) No significant change in VCAM-1 expression observed in DiPSC-ECs after the diabetic treatment; (f) Elevated P-selectin expression observed in HiPSC-ECs after the diabetic treatment; (g) No significant change in P-selectin expression observed in DiPSC-ECs after the diabetic treatment; (h) Reduction in cell viability observed in HiPSC-ECs after the diabetic treatment; (i) No significant change in cell viability observed in DiPSC-ECs after the diabetic treatment; (j) No significant change ROS production observed in HiPSC-ECs after the diabetic treatment; (k) No significant change ROS production observed in DiPSC-ECs after the diabetic treatment.Scale bar = 200 μm.*p < 0.05.**p < 0.01.***p < 0.001.n.s.not significant.

1 (
Specifically, we screened for the safety and efficacy in reducing inflammation status via VCAM-1 expression with those compounds at various doses for each patient's DiPSC-ECs.Interestingly, the patientderived cells exhibited a difference in tolerance toward the drug treatment (Figure5aand FigureS6).For instance, drugs at all doses, except for telmisartan at 100 μM, were not cytotoxic toward DiPSC-ECs from DM07, whereas almost all six compounds exhibit dose-dependent cytotoxicity toward DiPSC-ECs from DM02.The efficacy in reducing VCAM-1 expression was mostly consistent among the patients, with a reduction in VCAM-1 expression when treated with pioglitazone, telmisartan, and valsartan, whereas treatment with heparin, aspirin, or resveratrol resulted in no effect or slight increase in VCAM-1 expression (Figure5band FigureS6).An optimized therapeutic treatment was identified for each patient (DM02: Valsartan at 10 μM; DM07: Telmisartan at 1 μM; DM08: Valsartan at 1 μM) that resulted in a significant reduction in VCAM-1 expression without causing cytotoxicity.DiPSC-ECs were treated with individually optimized pharmacological treatment for 48 h and evaluated for their endothelial functional recovery.Western blotting study confirmed reduced VCAM-1 expression in DiPSC-ECs after treatment with Angiotensin II receptor blockade, and that Angiotensin II Receptor Type-AGTR1) is expressed by DiPSC-ECs with and without pharmacological treatment (Figure

F I G U R E 5
Drug screening with DiPSC-ECs as in vitro models of diabetic endothelial dysfunction.(a) Heatmap of cytotoxicity at 48 h when DiPSC-ECs were treated with potentially therapeutic compounds at various concentrations; Purple: high cytotoxicity; White: no cytotoxicity.(b) Heatmap of VCAM-1 expression relative to cells with no drug treatment when DiPSC-ECs were treated with potentially therapeutic compounds at various concentrations; Blue: reduced VCAM-1 expression; White: baseline VCAM-1 expression; Red: elevated VCAM-1 expression.(c) Representative western blots and quantitative analysis of DiPSC-ECs expressing AngII receptor AGTR1 and reduced levels of VCAM-1 after drug treatment; (d) Re-evaluation of DiPSC-EC tubulogenesis after the optimized drug treatment resulted in improved tube formation; (e) Quantification of master junction density in tubulogenesis analysis indicating improved tube formation after drug treatment with DiPSC-ECs.Pio: pioglitazone; Tel: telmisartan; Val: valsartan; Hep: heparin; Asp: aspirin; Res: resveratrol.*p < 0.05.**p < 0.01.***p < 0.001.

bated at 37 C
to form gels. Patient-specific iPSC-ECs at passage 2 were dissociated into single-cell suspension and seeded onto the matrix at 40,000/well in EC medium, with four wells per patient (n = 4).Eighteen hours after cell seeding, images were taken with phase contrast microscopy (Nikon TE2000U, Melville, NY) with a 4Â objective.All images were analyzed with ImageJ with the Angiogenesis Analyzer plugin.