Human Stem Cell‐Derived Endothelial‐Hepatic Platform for Efficacy Testing of Vascular‐Protective Metabolites from Nutraceuticals

Abstract Atherosclerosis underlies many cardiovascular and cerebrovascular diseases. Nutraceuticals are emerging as a therapeutic moiety for restoring vascular health. Unlike small‐molecule drugs, the complexity of ingredients in nutraceuticals often confounds evaluation of their efficacy in preclinical evaluation. It is recognized that the liver is a vital organ in processing complex compounds into bioactive metabolites. In this work, we developed a coculture system of human pluripotent stem cell‐derived endothelial cells (hPSC‐ECs) and human pluripotent stem cell‐derived hepatocytes (hPSC‐HEPs) for predicting vascular‐protective effects of nutraceuticals. To validate our model, two compounds (quercetin and genistein), known to have anti‐inflammatory effects on vasculatures, were selected. We found that both quercetin and genistein were ineffective at suppressing inflammatory activation by interleukin‐1β owing to limited metabolic activity of hPSC‐ECs. Conversely, hPSC‐HEPs demonstrated metabolic capacity to break down both nutraceuticals into primary and secondary metabolites. When hPSC‐HEPs were cocultured with hPSC‐ECs to permit paracrine interactions, the continuous turnover of metabolites mitigated interleukin‐1β stimulation on hPSC‐ECs. We observed significant reductions in inflammatory gene expressions, nuclear translocation of nuclear factor κB, and interleukin‐8 production. Thus, integration of hPSC‐HEPs could accurately reproduce systemic effects involved in drug metabolism in vivo to unravel beneficial constituents in nutraceuticals. This physiologically relevant endothelial‐hepatic platform would be a great resource in predicting the efficacy of complex nutraceuticals and mechanistic interrogation of vascular‐targeting candidate compounds. Stem Cells Translational Medicine 2017;6:851–863


INTRODUCTION
Cardiovascular diseases and stroke, which are among the top causes of mortality, pose major health care burdens. Common risk factors such as high blood pressure and hypercholesterolemia underscore a predisposition to blood vessel dysfunction. Atherosclerosis, characterized by arterial hardening and fatty plaque build-up in vessel walls, is one of the key contributors to vascular pathology. It is multifactorial, involving many cell types in a complex interplay of inflammation and oxidative stress [1]. Long-term medication is often required for secondary prevention of adverse vascular events. Despite beneficial pleiotropic effects of cholesterol-lowering statins on the vasculature, other side effects have been reported [2]. The concept of medical nutrition is on the rise to help modulate chronic diseases [3]. Dietary phospholipids from soybean, eggs, and fish have been explored as nutraceuticals with protective effects on atherosclerosis [4]. Herbal extracts are used as antiatherogenic agents by reducing the production, absorption, or oxidation of cholesterol [5,6]. Unlike small-molecule drugs, the complexity of constituents in nutraceuticals often present challenges in predicting their efficacy in preclinical screening. The liver is integral to biotransformation of complex compounds into new chemical species that may confer either therapeutic or toxic effects. Therefore, a human-relevant vascular model that incorporates liver metabolism would be more effective in assessing bioactivity of nutraceutical ingredients.
Advances in human pluripotent stem cell (hPSC) differentiation offer an unparalleled ability to generate many cell types in adequate quantities for tissue engineering, regenerative medicine, and disease modeling in vitro [7][8][9]. Vascular cell-based phenotypic assays have been demonstrated using endothelial and smoothmuscle cells generated from hPSCs [10][11][12]. In response to inflammatory or biomechanical stimuli, hPSC-derived endothelial cells are able to model athero-susceptible phenotypes and display a breach of barrier integrity to allow leukocyte transmigration. Although incorporation of primary hepatocytes is most likely to replicate human liver function for drug testing in a metabolism-enabled cellular model, their scarcity and the high cost of freshly isolated or cryopreserved human primary hepatocytes limit their extensive use [13]. We and others have created hPSC-derived hepatocytes that express cytochrome P450 (CYP) drug-metabolizing enzymes and show sensitivity to drugs with known hepatotoxicity [14,15]. Interestingly, induced hepatocytes derived directly by lineage reprogramming of human fibroblasts have comparable CYP activities with primary hepatocytes [16]. Vascularized liver microtissue has been proposed, although its utility as a drug-testing platform has not yet been shown [17]. Thus far, there has not been a model developed to recapitulate vasculature-liver paracrine interaction for efficacy testing of complex compounds.
In this work, we developed a vascular-liver model based on hPSCderived endothelial cells (hPSC-ECs) and hPSC-derived hepatocytes (hPSC-HEPs). The hPSC-ECs, generated via a mesodermal precursor population, demonstrated functional characteristics such as tube formation and inflammatory activation. HPSC-HEPs were derived by using our previous protocol [18], and they expressed drug-metabolizing enzymes. To test our hypothesis, we selected two nutraceuticals, quercetin and genistein, that are known to have potential vascular-protective effects [19,20]. We found that the parent compounds elicited minimal anti-inflammatory effects on hPSC-ECs. Conversely, when cocultured with hPSC-HEPs, the inflammatory response of hPSC-ECs was suppressed effectively, suggesting nutraceutical bioactivation by the metabolic activity of hepatocytes. Our data support the hypothesis that hPSC-HEPs could process quercetin and genistein into primary and secondary metabolites, which exerted greater anti-inflammatory effects compared with their parent compounds. Hence, this hPSC-based endothelial-hepatic platform could better predict the efficacy of nutraceuticals and identify beneficial properties of their constituent ingredients.

Generation of Hepatocytes From hPSCs
Hepatocytes were generated from hPSCs by a growth factorbased differentiation protocol described in our previous protocol [15]. After 20 days of differentiation, the cells were harvested by using a serial 23 TrypLE Express treatment and further dissociated into single cells by passing them through a 40-mm cell strainer. These single cells were then seeded at 2.5 3 10 5 cells per well in a collagen I (50 mg/ml, Bio Laboratories, Singapore, http://www.biolab.com.sg, catalog no. 354236)-coated dishes. Attachment and recovery were promoted by seeding them in step IV differentiation medium with hepatocyte growth factor (R&D Systems, catalog no. 294-HGN-005), Follistatin (R&D Systems, catalog no. FS-288), Oncostatin (R&D Systems, catalog no. 295-OM-010), and Y-27632 (Rock Inhibitor) to prevent anoikis in the freshly harvested hPSC-HEPs. The next day, medium was changed to Williams E medium (Sigma-Aldrich, catalog no. W1878) without   serum, and cells were serum-starved overnight before nutraceutical treatments. Nutraceuticals quercetin (Sigma-Aldrich, catalog no. Q4951) and genistein (Sigma-Aldrich, catalog no. G6649) were administered at a single dose of 10 mM. Hepatocytes in this work were derived from H9-ESCs. In hepatic characterization, primary human hepatocytes (PHHs) and HUH7 cells were used as positive controls, and HeLa cells were used as negative controls.
Endothelial-Hepatocyte Coculture Experiments on Ibidi 2 3 9 Wells H9-ESCs were used to generate the ECs and hepatocytes for coculture experiments and assays with nutraceuticals. Both cell types were derived from an isogenic source of hPSCs to ensure a robust endothelial-hepatic model without confounding phenotypic differences due to genetic variations. The Ibidi m-slide 2 3 9   wells (Ibidi, Martinsried, Germany, http://ibidi.com) were used for creating the coculture setup. Each minor well was first coated with collagen I, and then a total of 2.5 3 10 5 hPSC-ECs and hPSC-HEPs were seeded in individual wells in a 1:1 ratio [32] within a major well to allow equivalent contribution of paracrine factors between the two cell types (Fig. 5A). After cell attachment, the medium was changed to 1:1 William's E and EGM-2 without serum, to be commonly shared by hPSC-ECs and hPSC-HEPs in a major well. Upon overnight serum starvation, the cocultures were preconditioned with 10 mM quercetin or genistein for 48 hours before IL-1b stimulation.

Enzyme-Linked Immunosorbent Assay
Conditioned EGM-2 medium was collected, and the concentration of human IL-8 was determined by using the human IL-8 enzyme-linked immunosorbent assay (ELISA) kit (Thermo Fisher, catalog no. KHC0081), according to the manufacturer's instructions.

Western Blot
The cell lysates were collected by using radioimmunoprecipitation assay buffer (Thermo Fisher, catalog no. 89901) containing 13 proteinase inhibitor cocktail (Sigma-Aldrich, catalog no. P8340). Protein quantification was performed by using the Quant-iT protein assay kit (Thermo Fisher, catalog no. Q32210). A total of 80 mg of cell lysates was separated by NuPAGE 10% Bis-Tris Gel (Thermo Fisher, catalog no. NP0303BOX) and transferred onto a nitrocellulose membrane. MagicMark XP Western protein standard (Thermo Fisher, catalog no. LC5602) was used to determine the molecular weight of protein bands. The WesternDot 625 goat anti-rabbit Western blot kit (Thermo Fisher, catalog no. W10142) was used to visualize the protein bands. Blocking was performed at 4°C overnight by using 3% skimmed milk in 13 wash buffer provided by the kit and stained with CDH5 antibody in 3% skimmed milk solution for 1 hour at room temperature. The protein bands were visualized and imaged by using Bio-Rad ChemiDoc MP Imaging System (Bio-Rad, Hercules, CA, http://www.bio-rad.com).

Liquid Chromatography-Mass Spectrometry
The metabolic potential of the hPSC-HEPs and primary rat hepatocytes (freshly isolated according to our previously established protocol [33]) were tested by exposing them to nutraceuticals   Internal standard (IS) (Emodin, 10 ng/ml) was added to the conditioned medium. The solid-phase extraction column (Phenomenex, Torrance, CA, https://www.phenomenex.com, Strata C18-E, 55 mm, 70A) was conditioned by washing with 1 ml of methanol and then 2 ml of deionized water. Conditioned medium was added into the column, and 1.5 ml of 30% methanol was added to elute the impurity such as phenol red in the medium. 0.1% formic acid methanol was added to the column to elute all the metabolites and internal standard out to a 15-ml tube. Liquid in the 15-ml tube was dried under N 2 in a sample concentrator with 30°C heater. After drying the sample, 100 ml of 0.1% formic acid methanol was added to the 15-ml tube and vortex for 30 seconds and transferred to another 1.5-ml tube. The samples were then centrifuged at 13,000 rpm for 10 minutes at 4°C, and 10 ml of the supernatant was injected into liquid chromatography-mass spectrometry (LC-MS). Highperformance liquid chromatography combined with electrospray ionization (ESI) ion trap time-of-flight (IT-TOF) multistage mass spectrometry analyses were performed with a Shimadzu LC-MS-IT-TOF instrument, which was composed of two LC-20AD pumps, a SIL-20AC autosampler, a CTO-20A column oven, a CBM-20A system controller, an ESI ion source, and an IT-TOF mass spectrometer (Shimadzu, Kyoto, Japan, http://www.shimadzu.com).

Statistical Analysis
Data were expressed as mean 6 SD of at least three biological replicates of independent experiments. Statistical comparisons were conducted by Student's unpaired t test with 95% confidence interval for two groups of samples or one-way analysis of variance with Bonferroni's post hoc test in multiple group comparisons. Analyses were carried out with GraphPad Prism 5 software (GraphPad Software Inc., La Jolla, CA, http://www.graphpad.com).

Derivation of Functional Endothelial Cells From Human Pluripotent Stem Cells
Lateral plate mesoderm is a precursor tissue of vascular lineages. We adopted our established protocol of using fibroblast growth factor 2, bone morphogenetic protein 4, and phosphoinositide 3-kinase inhibitor (LY294002) to induce lateral plate mesoderm for 5 days [21] (Fig. 1A). We then used a combination of factors to drive endothelial specification. Transforming growth factor b (TGF-b) inhibition using small molecule SB431542 has been shown to enhance endothelial differentiation of hPSCs [34,35], possibly by counteracting growth of mural cells, which could arise from a common cardiovascular progenitor. FGF2 and vascular endothelial growth factor are commonly known mitogens for promoting angiogenesis and endothelial development [36,37]. Studies have also shown that hypoxic conditions increase the efficiency of endothelial differentiation [38,39], because upregulation of hypoxia-inducible factor triggers downstream targets that play an important role in early blood vessel development [40]. To induce endothelial differentiation, the day 5 mesodermal population was dissociated and plated down as single cells. We cultured these cells under 1% oxygen (O 2 ) in a chemically defined medium containing SB431542, FGF2, and VEGF (Fig. 1A). Endothelial genes were significantly enhanced in 1% O 2 as compared with 21% O 2 , peaking primarily at approximately day 10 (supplemental online Fig. 1A). Flow-cytometric analysis further supported that endothelial specification was optimal at approximately day 10 in 1% O 2 , with more than 45% of the cells positive for PECAM1 ( Fig. 1B; supplemental online Fig. 1B). This protocol generated a sufficient yield of PECAM1+ cells for cell sorting on day 10 of differentiation, giving rise to a purity of 98.43 6 0.16% (Fig. 1C). This PECAM1+ population was then grown on collagen I coating and expanded by using a commercial endothelial growth media (EGM-2). We hereafter refer to these cells as hPSC-ECs. Western blot demonstrated the presence of endothelial adherens junctions, CDH5, in hPSC-ECs and the positive control, human coronary artery endothelial cells (Fig. 1D). Different glycosylated forms of CDH5 were found in hPSC-ECs, close to the molecular weights of those in HCAEC. We postulated that there might be differences in glycosaminoglycan synthesis enzymes in hPSC-ECs and HCAECs [41]. Our hPSC-ECs formed spontaneous tube structures that stained for the mature endothelial marker, von Willebrand factor (vWF) (Fig. 1E). In addition, hPSC-ECs possessed endothelial nitric oxide synthase (eNOS) (Fig. 1F) and were capable of taking up acetylated low-density lipoprotein (LDL) (Fig. 1G), resembling HCAECs, but not the negative controls. We observed comparable tube-forming capability between hPSC-ECs and HCAECs (Fig. 1H). We reproduced this endothelial differentiation protocol on two other hPSC lines, namely, the BJ-and IMR90-induced pluripotent stem cells. The BJ-and IMR90-derived ECs also expressed endothelial proteins (supplemental online Fig. 2A, 2B), as well as demonstrated tube formation capability (supplemental online Fig. 2C). These functional hPSC-ECs were subsequently used for assay development.

HPSC-Derived Endothelial Cells Respond to Inflammatory Stimulation
Inflammation is a hallmark of atherosclerosis [1]. To recapitulate atherosclerosis-associated phenotypes in hPSC-ECs, we used an inflammatory cytokine, interleukin-1b, which is widely implicated in atherosclerosis. Upon stimulation with human recombinant IL-1b, hPSC-ECs responded with a significant upregulation of inflammatory genes ( Fig. 2A). Nuclear translocation of nuclear factor kB, activating major proinflammatory mediators, has been observed in human atherosclerotic lesions [42]. Likewise, nuclear translocation of NFkB was evident in hPSC-ECs after stimulation with IL-1b (Fig. 2B). The production of interleukin 8 from conditioned media  of IL-1b-stimulated hPSC-ECs was significantly higher than that of the unstimulated cells (Fig. 2C). In addition to H9-ECs, we also validated that BJ-ECs and IMR90-ECs could respond to IL-1b by upregulation of inflammatory genes and increase of NFkB nuclear translocation, as well as elevated IL-8 production (supplemental online Fig. 3). Hence, we were able to monitor hPSC-EC inflammatory activation using a range of phenotypic readouts.

Nutraceuticals Are Not Effective in Suppressing the Inflammatory Response of hPSC-Derived Endothelial Cells
Next, we tested whether administration of nutraceuticals quercetin and genistein could suppress inflammatory responses in IL-1b-stimulated hPSC-ECs. Quercetin, a naturally occurring flavonoid compound, is found commonly in food, such as tea, onions, berries, and apples. It exerts various beneficial effects through its anti-inflammatory [19] and antioxidant [43] properties. Quercetin intake is also correlated with lower incidence of coronary heart disease and stroke [44]. Genistein, a potent phytoestrogen, is effective in mitigating endothelial dysfunction [20] and exerts an anti-inflammatory effect by downregulating the NFkB pathway [22]. Plasma concentrations for genistein can range from 0.03 to 16.34 mM [30,45], in line with the dosage commonly used for in vitro studies [22,[26][27][28][29][30][31]. Quercetin has very low bioavailability in human plasma, where the concentrations range between 0.3 and 3.5 mM [46][47][48][49]. Nonetheless, higher concentrations of quercetin are known to be safe and well tolerated [50,51]. Previous studies in human hepatocytes [24,25] and human C-reactive protein mice [23] have used quercetin at a concentration of 10 mM. Therefore, we chose to treat the stimulated hPSC-ECs with quercetin or genistein at a concentration of 10 mM for up to 72 hours. However, gene expression of inflammatory markers did not show substantial reduction from the IL-1b-stimulated levels at various time points (Fig. 3A). NFkB nuclear translocation levels remained elevated despite administration of quercetin and genistein (Fig. 3B). There was also no significant reduction of IL-8 protein levels from the conditioned media of stimulated hPSC-ECS after nutraceutical treatment for 48 hours (Fig. 3C).
We further investigated whether the hPSC-ECs could metabolize the nutraceuticals. By using liquid chromatography-mass spectrometry to analyze conditioned media from hPSC-ECs, the level of quercetin (blue line) was found to decrease over time, whereas the levels of metabolites (black lines) from both nutraceuticals did not increase remarkably (Fig. 3D, 3E). Our data showed that first, the parent compounds may not be effective at eliciting anti-inflammatory effects on endothelial cells. Second, limited capacity of hPSC-ECs to break down the nutraceuticals into metabolites could have compromised the bioactivity of these compounds. Therefore, we explored whether hepatocytes derived from hPSCs were capable of metabolizing the nutraceuticals.

HPSC-Derived Hepatocytes Bioactivate Nutraceuticals Through Metabolism
Liver has been shown to metabolize quercetin into its bioactive metabolites, which in turn exert greater beneficial effects compared with their parent compound [24,25]. We generated hepatocytes from hPSCs following our established protocols [15,18]. The stepwise differentiation protocol recapitulates embryonic liver development because hPSCs progressively turn from primitive streak/mesoendoderm, definitive endoderm, and hepatoblasts to become hepatocytes (hereafter referred to as hPSC-HEPs). Our hPSC-HEPs stained positive for albumin, characteristic of functional hepatocytes (Fig. 4A). In accordance with our previous findings [15], we produced hPSC-HEPs that expressed cytochrome P450 genes (Fig. 4B), some of which were comparable to the positive control PHH. CYP enzymatic activities are necessary for the metabolism of complex compounds.
Indeed, LC-MS analysis demonstrated that the levels of both quercetin (blue line) and genistein (green line) gradually declined over time in the presence of hPSC-HEPs (Fig. 4C, 4D). Correspondingly, the levels of metabolites (black lines) increased over time in the hPSC-HEPs. Most of the metabolites peaked at 48 hours and dropped by 72 hours. Thus, hPSC-HEPs were capable of converting quercetin and genistein into their metabolites, with 48 hours being the optimal duration based on the metabolic profiles of these nutraceuticals. When we compared the metabolic activity of our hPSC-HEPs to freshly isolated primary rat hepatocytes [33], both demonstrated that quercetin (blue bars) and genistein (green bars) declined over time, giving rise to metabolites (gray patterned bars) (supplemental online Fig. 4). Because the primary rat hepatocytes could have retained some in vivo characteristics, their metabolic kinetics was apparently faster because substantial metabolites had emerged by 6 hours of treatment. We then investigated the effects of nutraceuticals on IL-1b-stimulated hPSC-HEPs. We observed a significant reduction in the panel of inflammatory gene expression upon nutraceutical treatment (Fig. 4E). The production of IL-8 protein from conditioned media of hPSC-HEPs was also significantly suppressed after treatment with nutraceuticals for 48 hours (Fig. 4F). Hence, the ability of hPSC-HEPs to process nutraceuticals into their bioactive metabolites could have resulted in their efficacy in abrogating inflammatory activation.

Renewal of Nutraceutical Metabolites in the Presence of Hepatocytes Protects Endothelial Cells From Inflammatory Activation
To enable accurate assessment of complex compounds in vascular health, we examined two configurations of endothelial-hepatic paracrine interaction. First, we allowed 48-hour preincubation of each nutraceutical with hPSC-HEPs for metabolism to take place (Fig. 5A). Subsequently, hPSC-HEP-conditioned media were collected and treated on hPSC-ECs under IL-1b stimulation. Alternatively, we cocultured hPSC-ECs and hPSC-HEPs on IBIDI m-slide 2 3 9 wells, where each cell type could be seeded separately into minor wells and shared a common medium by filling up the major wells (Fig. 5A). The coculture was pretreated with each nutraceutical for 48 hours, followed by IL-1b stimulation. Our data showed that hPSC-HEP-conditioned media with either quercetin or genistein did not seem to inhibit NFkB nuclear translocation in IL-1bstimulated hPSC-ECs (Fig. 5B). In contrast, when cocultured with hPSC-HEPs, stimulated hPSC-ECs displayed a significant suppression of NFkB nuclear translocation. Furthermore, the secretion of IL-8 in the coculture setting was remarkably decreased, but not in the conditioned media configuration (Fig. 5C). The metabolite profiles of quercetin (blue bar) and genistein (green bar) in each of the two configurations showed that there were detectable levels of various metabolites (gray bars) in the endothelial-hepatic coculture (supplemental online Fig. 5B), but not in the conditioned media setting (supplemental online Fig. 5A). This supports that the metabolites in hPSC-HEP-conditioned media could be degraded to a certain Narmada, Goh, Li et al. extent when they were subsequently treated on hPSC-ECs under IL-1b stimulation. This might lead to insufficient anti-inflammatory effects. Conversely, we also investigated whether hPSC-ECs could, in turn, impact the metabolic function of hPSC-HEPs in a coculture setting. It has been reported that endothelial cells could improve hepatic function [52][53][54][55] and provide some levels of hepatoprotection from acetaminophen toxicity [56]. In our study, cocultured hPSC-HEPs had comparable albumin levels with monocultured hPSC-HEPs (supplemental online Fig. 6A). Notably, there was significant increase of CYP gene expressions in cocultured hPSC-HEPs (supplemental online Fig. 6B), suggesting that the presence of hPSC-ECs could promote metabolic activity in hPSC-HEPs. Quantification of NFkB nuclear translocation shows that quercetin or genistein significantly decreased levels of NFkB nuclear colocalization in coculture of hPSC-ECs with hPSC-HEPs. (C): IL-8 protein levels were significantly reduced in endothelial-hepatic coculture but not in conditioned media setup. Statistical differences were compared with their respective stimulated groups without nutraceutical treatment. ppp, p # .001 (n = 3 independent biological replicates). Abbreviations: hPSC-ECs, human pluripotent stem cell-derived endothelial cells; hPSC-HEPs, human pluripotent stem cell-derived hepatocytes; IL, interleukin; NFkB, nuclear factor kB.  Endothelial-Hepatic Model for Efficacy Testing Therefore, a coculture of hPSC-ECs and hPSC-HEPs could better recapitulate the in vivo vascular-liver systemic interactome, with renewal of metabolites by liver metabolism.

DISCUSSION
We have developed an endothelial-hepatic system to predict the efficacy of nutraceuticals in vascular protection. Insights from developmental studies guided our hPSC differentiation strategy. We established a protocol for efficient generation of functional endothelial cells from a lateral plate mesoderm precursor. Endothelial specification was induced by using FGF2, VEGF, and 1% O 2 , all of which play roles in blood vessel development and angiogenesis [57]. Furthermore, small-molecule SB431542, a potent antagonist of activin receptor-like kinase, could enhance the efficiency of endothelial differentiation by inhibiting TGF-b signaling [35], which would otherwise promote mural cell specification from mesoderm. These hPSC-ECs were responsive to IL-1b-stimulated inflammation, but treatment with either quercetin or genistein was not able to offset the inflammatory activation. This was because of limited metabolic activity of hPSC-ECs to break down the nutraceuticals into bioactive metabolites. Hence, it led us to postulate that hepatocytes from hPSCs possess metabolic capacity to enhance bioavailability of metabolites from quercetin and genistein. We generated functional hepatocytes according to our previously established protocol [15]. The hPSC-HEPs, with high expression of CYP enzymes, were able to effectively metabolize quercetin and genistein into primary and secondary metabolites. Similar metabolites were detected when we compared the nutraceutical treatment on our hPSC-HEPs with primary rat hepatocytes, as well as those metabolites described in primary human hepatocytes [24-28, 58, 59]. We recognize that the primary rat hepatocytes required shorter time to metabolize the parent nutraceutical, because they were freshly isolated and hence could have retained most of their in vivo functionality. Depending on the structural complexities, it is also likely that different nutraceuticals would have distinct metabolic profiles. The hPSC-HEPs may require different treatment durations to release optimal levels of metabolites. This also highlights the importance of dosage response in hPSC-HEPs, where a range of physiologically relevant concentrations could be tested.
Incorporation of hepatocytes to endothelial culture underpins a key novelty of this work. Notably, hPSC-HEP-conditioned media containing nutraceutical metabolites were not effective in suppressing inflammation in hPSC-ECs. There could be a decline in the potency of metabolites from hPSC-HEP-conditioned media due to degradation. Instead, when hPSC-ECs were cocultured with hPSC-HEPs in a shared medium, we noticed that there was a significant reduction in inflammation. Continuous replenishment of metabolites in coculture setup recapitulated the systemic setting of liver paracrine effects on the vasculatures. Conversely, endothelial cells are known to improve hepatic function by promoting cell viability, synthesis of albumin and urea, and efficiency of the drug transporter system [52][53][54][55]. Our hPSC-ECs could, in turn, increase the metabolizing CYP enzyme activity in hPSC-HEPs. Another advantage of this endothelial-hepatic crosstalk is to recapitulate human-relevant response where certain drug metabolism dynamics may not be accurately reproduced in animals [60]. Our current system could still have limitations just as other in vitro cell-based platforms. The need for evaluating chronic exposure (i.e., .3 weeks) to drugs will involve further optimization to ensure viability and sustainable functionality of cells in long-term culture. For high-throughput drug-testing efforts, scalability of hPSC-derived cells may require improved conditions, such as supportive extracellular matrices or automation in bioreactors [61][62][63][64], for robustness of production.
Multicellular coculture models [55,65,66] are gaining momentum for various applications. A recent study described the use of human umbilical vein endothelial cells to stabilize hepatoblastoma C3A cells and modulate drug-induced hepatotoxicity [56]. Endothelial cells from hPSCs were used to vascularize liver constructs [52]. These are part of the advances to model liver vasculatures, using multiple stromal cell types alongside hepatocytes to develop liver lobule and sinusoid-like structures [52-57, 65, 67-70], for hepatotoxicity screening and tissueengineering applications. We are the first to report a human stem cell-derived endothelial-hepatic platform for efficacy testing of complex compounds. Nonetheless, we could capitalize on our endothelial-hepatic model for further development. Cells-onchip in microfluidic-based systems could provide the benefits of different flow dynamics, minimizing reagents used, and optical suitability for high-content imaging of cells. Initial studies showed that endothelial cells [71,72] and hepatocytes [17,73] have improved functionality in perfusion cultures. Phenotypic assays could also be developed to capture different pathological readouts for efficacy testing, as well as toxicology assessment. A spectrum of assay endpoints for vascular injury and atherosclerosis may include endothelial dysfunction, oxidative stress, apoptosis, matrix remodeling, etc. Multiplexing of phenotypic readouts in multicellular models [17,74] would add great value to the applications of coculture systems. In addition, our endothelial-hepatic platform could be used for disease modeling involving paracrine crosstalk between liver and vasculature. Because the liver is integral to normal or dysfunctional lipid homeostasis, this interplay could influence vascular function [75]. Moreover, it is likely that lipid-modifying nutraceuticals may involve liver metabolism to exert their actions on vascular tissue.

CONCLUSION
Our hPSC-based endothelial-hepatic model represents a physiologically relevant cellular system for nutraceutical screening, enabled by the paracrine crosstalk between the two cell types. Such a coculture system will help dissect the modes of actions of complex compounds without the need to purify into their constituent ingredients. Furthermore, this human-relevant platform could enable mechanistic interrogation of candidate compounds with liver-and vascular-targeting therapeutic effects or toxicity prediction.