Addressing Cardiovascular Toxicity Risk of Electronic Nicotine Delivery Systems in the Twenty-First Century: “What Are the Tools Needed for the Job?” and “Do We Have Them?”

Cardiovascular System and Disease: Historical and Emerging Relationship with Tobacco Products

Cigarette smoking is positively and robustly associated with cardiovascular disease (CVD) [1], including hypertension [2], atherosclerosis [3], cardiac arrhythmias [4], stroke [5], thromboembolism [6], myocardial infarctions, and heart failure [7, 8]. However, the CVD risk associated with using new and emerging tobacco products such as Electronic Nicotine Delivery Systems (ENDS) remains unclear. After more than a decade of ENDS presence in the U.S. marketplace, uncertainty persists regarding the long-term health consequences of ENDS use, especially for CVD [9, 10]. Numerous studies of ENDS aerosol exposures in humans and animals collectively demonstrate that these exposures induce early biomarkers of cardiovascular harm, including endothelial dysfunction (ED) and platelet activation (PA) in healthy young subjects and animals. A recent study of chronic ENDS users found that not only did they have comparable ED as cigarette smokers, but they also had elevated sera levels of biomarkers of harm, S100A8 and high mobility group B1 (HMGB1), that are not elevated in chronic smokers [11]. This and several other similar studies in humans show that both acute and chronic ENDS use increase biomarkers of CVD risk (e.g., heart rate [12], blood pressure [13], heart rate variability [12], and vascular dysfunction [14, 15]), although not all human studies have demonstrated these acute effects [16]. Exposure of animals to ENDS-derived aerosols induces ED and PA in vivo and ex vivo, adding support and biological plausibility to the findings gathered in human studies [17,18,19,20,21]. Nonetheless, the causal link of these outcomes to ENDS use still raises skepticism despite the application of the ‘conventional benchmark’ biomarkers of cardiovascular harm that are sensitive to the specific use of combustible cigarettes [17]. Similar endpoints are monitored in preclinical animal studies that require less time than conducting and evaluating longitudinal human studies (e.g., Framingham, Population Assessment of Tobacco and Health) yet also include known markers of systemic inflammation (cytokines), oxidative stress, endothelial dysfunction (flow-mediated dilatation, FMD), platelet aggregation, arterial stiffness, and atherosclerosis [17]. Although valuable in principle, preclinical studies suffer from the usual objection that animal models do not use the product as intended, which leaves enormous uncertainty regarding the nature of the relationship between real-world exposure and cardiovascular harm. How do we close that gap so that our screening tools more accurately and efficiently and in a temporally adequate manner reflect acute and chronic risk of using ENDS? We explore the answers to these questions in this paper.

Inherent Challenges: Capturing Disease Risk with New Approach Methods (NAMs) and the 3 R’s (Replacement, Reduction, Refinement)

In the past 5 decades, numerous technological advances have been the equivalents of “moon-shots” that include computers, miniaturization, tools for high-speed imaging and data acquisition, cloning, and gene editing, all with innumerable applications in the biological sciences. This has certainly been the case for hazard assessments of drugs and environmental chemicals in areas such as cardiovascular, respiratory, and developmental toxicity. Minimizing the time between the emergence of risk (e.g., ENDS use) and the administration of well-founded regulatory policy requires thoughtful consideration of the currently available sources of data, their applicability to the prediction of health outcomes, and whether these available data streams are enough to support an actionable decision. This challenge forms the basis of this white paper that addresses how best to reveal potential toxicities of ENDS use in the human cardiovascular system—a primary target of conventional tobacco smoking. To that end, we identify current approaches used to evaluate the impacts of tobacco on cardiovascular health, in particular emerging techniques that replace, reduce, and refine slower and more costly animal models with high-throughput platforms (e.g., can be applied to tobacco regulatory science). We address the limitations of these emerging platforms, and we propose ideas for systems biology approaches to help close the knowledge gap between the older animal models and new approach methodologies (NAMs). It is hoped that these suggestions and their adoption within the greater scientific community will result in fresh data streams that will support and expand the decision-making capacity of tobacco regulatory agencies worldwide.

Specific Challenges of Modelling Emergent Properties of the Cardiovascular System in Risk Assessment

Admittedly, the numerous physiological properties and pathology of intact organisms are difficult to replicate ex vivo—blood pressure, pulse pressure, shear forces, insulin resistance, coronary artery calcium, carotid intima-media thickness, and atherosclerosis—to name but a few. Nonetheless, emerging technology is beginning to address these complex outcomes through the use of surrogate endpoints that are mechanistically linked or invoked by shared systems biology pathways (e.g., adverse outcome pathways or AOPs). Applied use of human-induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs), fibroblasts, and endothelial cells (e.g., iPSC-CMs; described in more depth below) represent an important advance in this area because these are human cells with human-coded physiology as well as an individual human’s specific genetic program that permit probing of drug toxicity via susceptibility genes (i.e., personal or precision medicine). The coupling of differentiated iPSCs with high-throughput platforms such as microelectrode arrays (MEA) allow the acquisition of an electrocardiogram-like signal from the iPSC-CMs and makes possible rapid screening of thousands of compounds for both therapeutic efficacy and toxicity on a compressed timeline. While a single-cell type cannot replicate the complex phenotype of a 3-D heart or blood vessel that contains multiple cell types with their intact interconnections, the predicted development of multi-phenotype, 3-D culture systems is now emerging as a practical reality [22, 23].

Recent advances have exceeded the old limitations of cultured cell experiments by combining multiple differentiated iPSC types to make assembloids and organoids—including cardiac organoids that beat. These new and emerging platforms present exciting opportunities to examine the cardiovascular toxicity of ENDS aerosol and tobacco product smoke, as well as their product additives and constituent ingredients. In the following section, we consider the benefits and limitations of these emerging platforms in order to highlight potential pitfalls and to recognize how best to use these platforms in tobacco regulatory science with respect to CVD and the overall effects on public health.

Cardiac

State of the Art Modeling and Toxicities

Animal and cell-based cardiovascular models have been indispensable in drug discovery because cardiac tissue is difficult to acquire from patients and, requires informed consent, invasive biopsy, and is not without complications [24, 25]. The effects of conventional tobacco products and ENDS on the cardiovascular system have been proposed or attempted using in vivo, ex vivo, in vitro, and in silico models. However, each of these models has its own limitations. Animal models are costly and time-consuming and may not precisely capture genetic or environmental mechanisms of cardiac dysfunction in humans. Traditional cell culture has been used with some success to evaluate probable environmental toxicants and potential drug candidates. However, primary cell lines derived from biological sources have a finite lifespan that limits their widespread use. Immortalized cell lines from tumor tissue or viral transformation are widely used in traditional culture studies; however, these lines express an altered genotype that could potentially impact the responses of their cellular physiology to experimental treatments and the translation of these findings to the clinical population. More importantly, two-dimensional (2-D) surfaces are not the native environment for a cell that is normally surrounded by an extracellular matrix (ECM) and other cells in a three-dimensional (3-D) format. The 3-D organization induces physical constraints, facilitates cellular crosstalk, and allows for biochemical and mechanical transduction from the external environment that is essential for the maturation and development of tissue [26,27,28].

Stem cell biology is one discipline that has the potential to bridge the gap between animal models and traditional cell culture systems [24]. Stem cell biology has already facilitated the differentiation of cellular lines from iPSCs to enhance toxicity screening and drug development [29, 30]. It is likely that iPSC-derived cardiac lineages could be leveraged as an initial screening step to avoid more costly and time-consuming animal studies with the potential to generate more clinically relevant findings. Original techniques for phenotyping iPSC-derived cardiac lineages relied on 2-D monolayers to analyze treatment effects in a standard culture dish. The emphasis was on preparing iPSC-derived tissues with high purity to improve the signal-to-noise ratio in transcriptomic, proteomic, and metabolomic analysis [31,32,33,34]. The emergence of single-cell transcriptomics has facilitated the development of 3-D structures such as assembloids, engineered heart tissue, and organoids [35, 36]. The recapitulation of the cardiac secretome by these 3-D systems uses peptide hormones, growth factors, and cytokines to facilitate crosstalk between different cardiac cell types in a biologically relevant manner [37, 38]. Cardiac-derived exosome cultures also participate in inter-cellular signaling networks similar to those in the intact heart [38,39,40]. As a result, the secretome is a likely source for novel biomarkers that might help elucidate the mechanisms of ENDS-induced cardiovascular toxicity, while an iPSC-derived 3-D model systems coupled with transcriptomic mapping could reveal novel pathways and mechanism(s) of disease (Section “In Vitro Models and Culture Systems”).

In Vivo Models and Human Studies

Toxicology studies have traditionally relied on animal models to show cardiac toxicity. Studies in nonmammalian models of cardiac development, such as zebrafish [41], have utility in capturing the potential embryonic toxicity of e-cigarette components. The small mammalian models in mice and rats are genetically closer to humans and also facilitate inhalation studies. Small animal models do not exhibit cardiovascular pathophysiology such as cardiomyopathies, channelopathies and arrhythmias, and atherosclerotic heart disease that causes plaque rupture. However, the use of genetic knock-out rodents such as LDL^−/− and ApoE^−/−, which exacerbate atherosclerotic plaque progression, has enhanced studies of CVD in animals [42, 43]. A final caveat in applying outcomes in small rodent models to human health is that while the absolute differences in heart rate, cardiac function parameters, pharmacokinetics, and pharmacodynamics can be scaled for general toxicology issues, these differences remain problematic for focused cardiovascular toxicity studies [44]. Large animal models more faithfully recapitulate the physiology and hemodynamics of humans and have been successfully used to investigate the efficacy of devices such as percutaneous coronary intervention and, more recently, transaortic valve replacement [45, 46]. However, large animal models are costly and still do not fully capture the spectrum of human disease. Non-human primates (e.g., monkeys) are similar anatomically and physiologically to humans, but these studies are encumbered with ethical dilemmas, in addition to their severely limited accessibility and extreme costs.

Epidemiological studies in human population data can correlate toxicant exposure with CVD, but facilities for establishing the mechanistic role of smoke, vape, or the contribution of individual product constituents (i.e., gases or particulates other than nicotine) in humans are geographically limited [9]. While hypothesis generating, retrospective observational studies have many confounders and biases that are introduced, and that cannot be excluded without a prospective randomized control trial. Confounders or biases occur in retrospective studies [47, 48]. Prospective studies are much better, but the design, execution, and data analysis might take years or decades. Thus, prospective studies are limited by their sample size, cost, and the time it takes tobacco and ENDS to cause cardiopulmonary toxicity/disease and cancer. Therefore, epidemiology-centered studies may take years or decades to detect the adverse cardiopulmonary effects of ENDS that are statistically significant. The ongoing longitudinal Population Assessment of Tobacco and Health (PATH) study is an excellent model for how to organize a study that oversamples tobacco users so that early warning signs of cardiopulmonary injury (biomarkers of harm) can be detected [9, 49,50,51,52]. PATH is also revealing important trends in how people actually use ENDS and tobacco products (dual and poly use) as well as transition between tobacco products which likely have important consequences for the induction of disease and for the development and implementation of regulatory guidelines [53, 54].

Ex Vivo Models

Ex vivo models for assessing human cardiovascular toxicity rely on tissue that is often difficult to acquire. Cardiac tissue can be obtained from biopsies of the myocardium; however, these samples are obtained via informed consent and must pass rigorous internal review board (IRB) justifications that protect patients, which may delay or reject requests in cases of low clinical benefit or any ethical uncertainties. Cardiac biopsy samples are also tissue-type, limited to the right ventricle or second heart field, and may not capture the adverse effects of toxins on the unique electrophysiology of the left ventricular tissue field. In addition, samples from patients are potentially exposed to previous unknown environmental exposures that might obscure the analysis of cardiac biopsy tissue. Deceased donor transplants are often rejected and can be sustained in perfusion instruments for ex vivo cardiac studies. However, rejected donor hearts might lead to spurious results because donor hearts are often rejected because of infection, poor cardiac function on echocardiography, or exposure to toxins such as chemotherapy. All in all, although ex vivo models are in some ways an ideal platform, they are of limited scientific value due to many important complicating issues such as previous exposures and sample size, to name a few. These barriers to utility may be overcome with better records on history and exposures with the latter being addressed with multi-omics approaches.

In Vitro Models and Culture Systems

Expanding on traditional cell culture techniques, stem cell biology has facilitated the generation of cardiac tissues from bone marrow-derived stem cells (BMSCs), embryonic stem cells (ESCs), and iPSCs. BMSC from donor bone marrow are easier to acquire or reprogram than ESCs or iPSCs. However, BMSC research has reproducibility issues, non-uniform differentiation protocols, and equivocal applications in the field of regenerative medicine. It must also be noted that as bone marrow-derived stem cells are not pluripotent, the cellular differentiation options are more limited, and their progeny phenotypes might not be mechanistically receptive to ENDS toxicity.

The pluripotent stem cell sources include ESCs, iPSCs, and somatic nuclear transfer (SCNT) [55]. SCNT involves the transfer of a somatic cell nuclear material to an aneuploid oocyte and thus the potential for cloning or the development of totipotent cells, which is controversial. ESCs were traditionally derived from human embryos, but due to significant moral and ethical dilemmas, there are a limited number of ESC lines (and funding). The development of iPSCs has revolutionized stem cell biology [24]. Reprogramming is carried out using commercially available Sendai virus (SeV) encoding for the Yamanaka factors (Oct4, Sox2, Klf4, and c-Myc) [56]. These recombinant SeV vectors do not integrate into the host genome and can reprogram iPSCs from terminally differentiated tissue-types such as human skin fibroblasts and peripheral blood mononuclear cells [57]. Therefore, iPSCs retain the donor’s original genetic blueprinting, which allows for personalized disease modeling, drug screening, regenerative medicine, and toxicity screening, while avoiding the ethical issues surrounding embryonic stem cells.

iPSC-CMs are generated using a defined, high-yield chemical protocol (Fig. 1) [31], which produces contractile sheets of up to 95% TNNT2⁺ cardiomyocytes with up to 100 cardiomyocytes from each iPSC. These iPSC-CMs have been used to model familial cardiomyopathies, such as dilated cardiomyopathy (DCM), hypertrophic cardiomyopathy (HCM), long QT syndrome (LQT) syndrome, Brugada syndrome, and left ventricular non-compaction (LVNC) [58, 59]. In addition, iPSC-derived endothelial cells (iPSC-ECs) generate consistent yields of 80–95% CD31⁺ cells [60, 61] with cobblestone morphology and have been shown to express CD31, CD144, and eNOS. Studies of iPSC-derived cardiac fibroblasts (iPSC-CFs) report successful expression of mesenchymal and myofibroblast markers with cellular morphology comparable to that of primary cardiac fibroblasts [62]. These iPSC-CFs retain donor susceptibility to doxorubicin toxicity and maintain cellular phenotypes discrete from those of iPSC-ECs, and iPSC-CMs.

Induced pluripotent stem cell (iPSC)-derived cardiovascular tissue models for e-cigarette toxicity testing. iPSC-derived cardiovascular lineages and be utilized in 2-D monolayers or 3-D structures such as assembloids, organoids, or engineered heart tissues (EHTs). iPSC-derived 2-D and 3-D models can be used to test the effects of e-cigarette-derived aerosols or constituents using high-throughput “omics” assays. Moreover, iPSC-derived cardiac tissues can be used in functional assays to determine the effects of e-cigarettes on viability, inflammation, and reactive oxygens species (ROS) production, contractility, calcium handing, and electrophysiological properties.

Full size image

Successful toxicity screening in iPSC-derived cardiac tissue has been performed using viability studies, high-throughput-omics (such as transcriptomics, metabolomics, and proteomics) [63], and functional assay outcome measures. Doxorubicin-induced cardiotoxicity is a clinically heterogeneous entity that requires meticulous imaging to screen for susceptible individuals. However, iPSC-CMs can be used to capture the sensitivity of donors to doxorubicin in a cellular assay methodology [64]. Similarly, iPSC-CMs have been employed in high-throughput screening techniques to evaluate the cardiotoxicity of tyrosine kinase inhibitors to create a novel “cardiac safety index” [65]. Thus, iPSC-CMs have potential uses to identify and characterize the genetic basis and molecular mechanisms of toxicant-induced outcomes.

Cultures of iPSC-CMs could also be used for toxicity studies in tandem with atomic force microscopy, calcium imaging, patch clamping [66], and multielectrode array (MEA) [67, 68]. The application of atomic force microscopy allows the measurement and analysis of iPSC 2-D- and 3-D-derived cardiac structures and any treatment-induced changes. The contractile force imaging system can capture the iPSC-derived 2-D cardiomyocytes’ contractile motion at high resolution to detect toxic effects on systolic or diastolic function. The electrophysiological effects of toxins might be revealed by cellular calcium handling profiles that are readily studied in primary tissues using calcium imaging and optical mapping. The use of MEA could generate real-time electrophysiology mapping for electrical wave propagation and arrhythmogenicity of 2-D and 3-D iPSC-derived cardiac tissues.

iPSC-derived cardiac tissues are a breakthrough in understanding the underlying mechanisms of cardiovascular toxicity in vitro. One limitation of using iPSC-derived cardiac tissues is the cellular immaturity of conventionally derived iPSC-CMs. Currently, iPSC-CMs lack T-tubules and ion channels, are arrhythmogenic, and lack the proper 3-D structure of in vivo cardiomyocytes [69]. However, improvements such as the use of maturation medium [70], electrical stimulation, and 3-D printing of the iPSC-derived cardiomyocytes have improved the maturation phenotype and functionality of these cultures [71]. The recent advances in 3-D bioengineering of heart tissues [72, 73], assembloids [74], and organoids [75] offer a new repertoire of iPSC-derived culture options that may more faithfully recapitulate the mechanisms of CVD and toxicity.

The current in vitro culture systems used for cardiac toxicology studies include the 2-D culture of cardiomyocytes (CMs), 3-D organoids assembled from CMs and other cells, as well as engineered tissues made of CMs. Most of the CMs are derived from mesenchymal stem cells (MSCs), iPSCs, and human embryonic stem cells (hESCs). Recently, iPSC-CMs were used to profile the cardiotoxicity of exposure to ENDS-derived aerosol. Treatment of CMs with ENDS aerosol extract-induced significant oxidative stress and lowered cellular viability in CMs, with no difference in magnitude of effect compared with combustible cigarette smoke extract, and in addition to the altered cellular viability, ENDS aerosol extracts inhibited beat rate and increased field potential duration (FPD) within the cell layer [76, 77]. Interestingly, the FPD corresponds to the cardiac QT interval, and the extended duration of FPD or QT interval is predictive of the development of abnormal heart rhythms such as tachycardia and fibrillation. The increase in FDP seen in the CMs after exposure to the ENDS extracts, suggests that ENDS vaping may increase the risk of arrhythmogenesis in users of this product [78]. Among different flavored products, ENDS with vanilla and apple flavor showed higher cardiotoxicity than a “Hawaiian POG” flavored ENDS product, with lower beating frequency and higher FPD [77]. One of the chemical constituents of cinnamon flavorings, cinnamaldehyde, was found to decrease cellular viability, and inhibit both beat rate and the action potential [79]. Another study reports that direct treatment of CMs with just PG or VG had no effects on either beat rate or contractility [80]. DNA methylation activity (increased gene expression of DNMT3A and DNMT3B) has been reported in CMs after ENDS aerosol exposure, which is known to lead to adverse effects on the contractility of cardiomyocytes and also to induce mitochondrial damage in cardiac cells [76, 81].

In addition to stem cell-derived CMs, commercially available cell lines, such as AC16 (human cardiomyocytes cell-line), H9c2 (rat cardiomyoblast), and HL-1 (mouse atrial myocytes), have been used in toxicology studies. HL-1 cells exposed to ENDS e-liquid with or without flavorings are reported to decrease cellular viability in a concentration-dependent manner [77]. Similar to effects in iPSC-CMs, apple, and vanilla flavors were more toxic in HL-1 cells than was “Hawaiian POG” ENDS liquid [77]. In the H9c2 rat cardiomyoblast cell-line, cardiotoxicity was evaluated using different flavored ENDS aerosol extracts [82]. Among 21 flavored ENDS products, those prepared from tobacco leaves showed higher cardiotoxicity than others but less than the toxicity of Marlboro cigarette smoke extract [82]. Interestingly, ENDS aerosol extract generated with higher voltage during atomizing induced more cell death than lower voltage within the same flavor category [82]. It is well appreciated that the quantity of HPHCs (e.g., carbonyls) generated from e-liquids is a function of power, coil resistance, voltage, user topography, and, thus, e-liquid temperature [83,84,85,86,87,88]. Higher levels of ENDS-derived HPHCs per unit volume can induce greater cytotoxicity [88, 89]. This, however, does not necessarily apply universally to individual constituents of e-liquids. For example, analytical grade cinnamaldehyde is directly toxic to iPSC-CMs, whereas its heated aerosol condensates (at 200 °C or 700 °C) were less toxic in iPSC-CMs [79] despite the fact that their heating generated numerous secondary thermal degradation products [90]. Nonetheless, because of this inherent heating function of ENDS, detailed methods regarding generation (e.g., aerosol extracts vs. condensates), collection, storage, and application of these aerosols in ex vivo and in vitro (e.g., cell culture) settings are necessary for conducting a thorough and quality assessment of the relationship between cytotoxicity and individual or collective HPHCs as well as for the meaningful extrapolation of these findings to an overall estimate of CVD risk [91].

Although 3-D cultured/engineered cardiac tissue has been developed, no study has focused on the cardiotoxicity of ENDS using in vitro 3-D model. Cardiomyocytes in a 3-D (soft) microenvironment (heart-like organ) could be a useful model in the cardiotoxicity study of ENDS liquids or aerosol extracts, which may better reflect the toxicological impact of ENDS use. Nonetheless, biomaterial development has aided in achieving realistic in vitro cardiovascular models. Self-assembly and the use of composite hydrogels have been attractive techniques for in vitro modeling in cardiac toxicology. A self-assembling nanofiber hydrogel was developed to mimic the properties of collagen without variability between batches, which is a shortcoming of using natural biomaterials [92]. Composite hydrogels have also been engineered to withstand mechanical forces from cells [93]. Engineering of substrates may fine-tune differentiation of stem cells into cardiomyocytes. Öztürk-Öncel et al. (2021) modified the surface of poly(dimethylsiloxane) (PDMS) with conventional and amino acid-conjugated self-assembled monolayers to guide differentiation of cells into cardiomyocytes [94]. The development of 3-D culture techniques has allowed primary cardiomyocytes to maintain their sarcomeric structure for up to 14 days in vitro [95]. Innovation in tissue engineering enhances the possibilities of accurately modeling cardiac tissue in vitro. A novel differentiation protocol reports the use of insulin-like growth factor 1 (IGF1) and neuregulin-1β (NRG1), two growth factors important in cardiac development, to stimulate maturing of human ESC-derived cardiomyocytes in both a 2-D culture system and an engineered 3-D cardiac tissue culture [96]. In the 3-D model, the addition of IGF1 increased proliferation while NRG1 increased metabolic activity. Such protocols may be useful in future work to produce more robust 3-D tissue models of cardiac tissue.

Embryonic Stem Cells (ESC)

While the early 2000s yielded a plethora of studies that used ESCs differentiated into cardiomyocytes, the first successful guided de-differentiation of cells into iPSCs was reported in 2007. This created a backlash that impeded the advancement of protocols for even ethically sourced ESC [97]. Because of the ensuing use restrictions and tightly regulated guidelines for the creation of ESCs in several countries, the use of ESCs for cardiovascular in vitro models are not widespread, even in the current literature.

A large body of work describing the differentiation of ESCs into cardiac cell types for clinical applications has been published in scientific literature [40, 98]. In vitro studies pertaining to the cardiotoxicity of ENDS products are less common, but the differentiation of ESCs into cardiomyocytes is well-reported. The H7 human embryonic stem cell-line has been differentiated into beating cardiac tissue through co-culture with END-2 cells, with the first signs of beating normally occurring around day 14 [90, 92]. These ESC-derived cardiomyocytes were used to demonstrate proof of concept of a 3-D tissue model using a 3-D nanofiber hydrogel to support cardiac tissue models [90, 92]. In 2015, a study reported the use of RUES2 ESCs to study the effect of ENDS aerosol extract exposure on cardiac development [99]. The authors did not observe any effects from ENDS aerosol extract on differentiation, but after cells were differentiated into cardiomyocytes for fetal-stage cardiac tissue, a reduced expression of sarcomeric genes was observed. A similar investigation was reported in a more recent study that exposed H9 (WA09) human ESCs to nicotine and found that nicotine exposure prevented the differentiation of ESCs into cells with cardiac lineage [100].

Mesenchymal Stem Cells (MSC)

Mesenchymal stem cells, derived from a variety of tissues such as adipose tissue and bone marrow, are multipotent cells with less differentiation capacity than ESC, but more so than terminally differentiated cells. MSCs derived from bone marrow, i.e., BMSC, retain a progenitor cell lineage that includes cardiac cell phenotypes, including cardiomyocyte-like cells [101]. The primary focus of MSC research in the field of cardiology has been on therapeutic applications to mitigate tissue damage due to drug-induced toxicity or disease by leveraging the non-immunogenic properties of MSCs as transplant tissue, as sources for an allogenic tissue transplant [102, 103]. A variety of protocols for differentiating MCSs into cardiomyocytes in vitro have been reported using chemicals, microRNAs, and cytokines [104]. As of 2023, there are no studies that utilize bone marrow MSC-derived cardiac tissue to study the cardiotoxicity of ENDS that utilize bone marrow MSC-derived cardiac tissue. However, the differentiation protocols intended for transplant tissues could be utilized to provide differentiated culture tissue for in vitro toxicology studies.

Induced Pluripotent Stem Cells (iPSCs)

The iPSCs are reverse-engineered to a pluripotent state with a wide differentiation potential, and offering the ability to better model disease states, patient phenotype-specific drug response, and generally more realistic in vitro models of cardiovascular toxicity [95, 105,106,107]. The use of iPSCs to derive cardiomyocytes is widely used and well-established [108]. A 2013 study demonstrated the utility of iPSCs with the publication of disease-specific patterns of cardiotoxicity revealed using a diverse library of iPSC-derived cardiomyocytes created for drug screening [66]. More recently, multi-cell type models of iPSC-derived cells have even been used in vitro to recreate the complex structure-function relationships in cardiac tissue [95].

Tobacco control researchers have begun to publish studies on ENDS products using iPSC-derived cardiac cells [79, 109]. A 2022 study from Imperial Brands PLC compared heated tobacco product smoke and ENDS aerosol extract with 1R6F smoke using a VITROCELL VC 10 S-TYPE smoke robot that bubbled product smoke through phosphate-buffered saline (PBS) [109]. iPSC-CMs were exposed to the PBS (with smoke extract) added into the culture medium each 24 h up to 72 h, and then metabolite changes (4 metabolites) and cell viability assessed. Results were used to generate prediction models of cardiotoxicity. Several products tested did not yield a detectable effect in the exposure ranges tested.

In a 2019 study, iPSC-CMs were used to evaluate the effects of cinnamaldehyde, a common flavor additive in commercial e-liquids [79]. Cells were exposed to the flavoring chemical in liquid form and to its aerosolized thermal products at concentrations estimated to be present in the blood of ENDS users. The cinnamaldehyde exposure of iPSC-CM yielded time- and concentration-dependent functional cardiotoxic effects and reduced viability after 24–48 h of incubation, with effects more pronounced when the compound was in its liquid form [79].

A review of in vitro models for studying cardiovascular toxicity of chemotherapy agents published by Pinheiro and colleagues in 2019 has proposed a workflow for the use of iPSCs in cardiovascular toxicity testing [105]. This workflow suggested dose-response testing using iPSC-CMs to study cardiovascular toxicity related to chronic heart failure and cardiac dysfunction, viability and proliferation assays of iPSC-derived endothelial cells/vascular smooth muscle cell co-cultures to study atherosclerotic phenotypes, and iPSC-derived immune cell/cardiomyocyte co-cultures with viability, electrophysiology, and immune cell assays to study effects such as hyper-eosinophilic cardiac toxicity, myocarditis, and pericarditis. These recommendations are relevant for tobacco research studies of adverse cardiovascular effects that may be amenable to e-liquid and ENDS aerosol evaluation.

While iPSC use marks a significant advancement for cardio-toxicology in vitro, the use of iPSCs has limitations. Immature cell state, small sample sizes, and variance between replicates continue to be a challenge for this methodology [105]. Future work needs to address these issues while continuing to expand the use of iPSC-derived cells for testing of the next-generation tobacco products, especially with regards to specific effects by cellular phenotype. Finally, use of iPSCs in heterogeneous tissue models that recreate the complex hierarchical structure of the heart needs to be encouraged to more fully capture the toxic potential and mechanisms that result from ENDS use in vivo.

Toxicity Profiling

A variety of cell viability and cytotoxicity assays have been used to study the toxicity of cardiac cell types exposed to ENDS products in vitro. According to a 2020 review of available in vitro models for evaluating cardiotoxicity due to chemotherapeutic agents [105], the two most common techniques for assessing viability in cardiomyocytes are ATP detection luminescence assays and resazurin assays (Alamar Blue). Both of these assays use alteration of metabolic function as a marker of cell viability [105]. Specific to tobacco control publications, an article by Simms et al., used the Cell Titer-Fluor Cell Viability Assay (Promega, Madison, WI) to measure the cardiotoxic effects of newly marketed nicotine products on iPSC-CMs [109]. Use of simpler colorimetric assays have also been used to assess the toxicity of tobacco product constituents [79, 100]. In a study by Nystoriak et al., an MTS reduction assay was used to test the cytotoxicity of a common e-liquid flavor additive [79], while He et al. used an MTT assay of the ESC line H9-derived cardiomyocytes to successfully predict the adverse effects of nicotine on cardiogenic differentiation and fetal heart development [100]. While cell viability and cytotoxicity assays can capture the toxicity hazard potential of a chemical or mixture, additional toxicity profiling techniques are available that provide more granular evidence of toxic effects beyond acute cell death. One study in 2019 using MEA was able to detect early adverse effects in cardiomyocytes after short-term exposure to a chemical that appeared to be mechanistically independent of decreased cell viability observed after a longer exposure to the test substance [79].

Toxicity profiling techniques and outcomes may differ between 2-D and 3-D culture settings, with some unique benefits of using 3-D tissue models for cardiac toxicity profiling. For example, one study compared the effect of NRG1 on the differentiation of ESCs and found that cell area increased with the administration of NRG1 in 2-D cultures, but this did not occur in 3-D cultures [96]. Additionally, compared with the same cells in 2-D, a sophisticated 3-D model expressed more cardiomyocyte-specific proteins, better sarcomere organization, greater production of extracellular matrix, and more persistent and stable spontaneous contractions for 2 months [110]. Similarly, a contractility assay coupled in a 3-D culture setting was more sensitive to drug toxicity than with cells in a 2-D setting. Moreover, this assay utilized an optical technique and real-time video recording, and automated cell tracking software to study the contraction behavior of the 3-D tissue model to detect changes at levels of a known toxic substance below those that disturbed ATP levels in 2-D culture. Thus, the functional contractility assay in 3-D culture provides a more sensitive assay than that in a 2-D setting.

With the rapid development of multi-omics technologies, such as low-cost and high-throughput gene sequencing technologies, profiling of chemical exposures through genomics is on the rise. In a 2015 study, exposure to ENDS constituents reduced expression of the sarcomere genes MLC2v and MYL6 during embryonic development [99]. Transcriptomics is another area of rapid development that may provide a tool for profiling cardiotoxicity and the discovery of the molecular mechanisms that influence genetic influence on disease. A 2020 publication reports using single-cell RNA sequencing (scRNA-seq) to successfully detect the adverse transcriptional effects (confirmed by qPCR and Western blotting) of nicotine on cardiac cell differentiation in the H9 line (human ESC) [100]. As these long-read, RNA sequencing technologies and the statistical and bioinformatics tools become more accessible in the greater scientific community, opportunities for more genome-wide toxicity profiling will continue to grow [111].

In addition to multi-omics approaches, novel microsensor techniques have been deployed to study the influence of compounds on cardiac cell activity. In 2020, Kim et al. integrated a piezoresistive sensor into PDMS to measure the effect of drugs on contraction force and beating frequency of cardiomyocytes [112]. Nystoriak and colleagues used a cellular impedance assay to study changes in contractility and rhythmicity of iPSC-derived cardiomyocytes after exposure to a commercial ENDS e-liquid flavor additive and reported their findings in 2019 [79]. The results of the Nystoriak study demonstrate that the cellular impedance assay of iPSC-CMs is an achievable tool for measuring cardiotoxic effects. As nicotine is both a neural stimulant and a depressant with known effects on heart rate and myocardial contractility [113], examination of the effects of ENDS constituents on the mechanical function of cardiac tissue in this way may also provide insight into ENDS risk profile for cardiovascular toxicity and disease [114].

In Silico Analysis

The majority of in silico methods for cardiac toxicity assessment are those used to predict outcomes for drug screening assays and environmental toxicant exposures. Most computational models are based on chemical structure similarities and utilize complex probabilistic or linear regression algorithms to predict potential outcomes for chemicals in commonly used mechanistic toxicity assays. However, few if any, of the outcome assays in these software platforms are specific for cardiac toxicologic endpoints.

One commonly used in silico tool for computer predictions of toxicity is the Toxicity Estimation Software Tool (TEST) published by the Center for Computational Toxicology & Exposure (CCTE) in the United States Environmental Protection Agency (USEPA) [115]. Describing the TEST software is useful for introducing Quantitative Structure Activity Relationship (QSAR) models, which are commonly used is the computational prediction of chemical hazard. QSAR models seek to predict one or more toxicological endpoints based on similarities of chemical structure between the chemical of concern and chemicals that have actual assay findings available in the library.

QSAR programs rely on complex combinations of probability and regression algorithms and use multiple fit coefficients to predict the toxicological endpoint. The toxicological endpoints offered in TEST include estimated LC₅₀ points of departure in water for sentinel species of fish, lethal dose, LD₅₀, for oral dose to rats, bioaccumulation factors in fish, and predictive outcomes for the Bacterial Reverse Mutagenesis or AMES Assay (OECD TG 471) [116], which utilizes 5 strains of Salmonella sp. While these endpoints are relevant for system-level environmental exposures, they are not specific to cardiac toxicity assessment. The TEST software offers five distinct QSAR algorithms for deriving a predictive outcome, many of which are common to other computational platforms [117]. The simpler QSAR method applies a single-multivariate linear regression algorithm with preselected regression parameters (fit coefficients), that are hard-coded in the program. The “group contribution” QSAR method advances the simple QSAR approach by linking the chemical of interest with individual chemical fragments of other screened chemicals in the library, based on the shared similarity of structural fragments or sidechains. The “nearest neighbor” method uses triangulation to predict assay outcomes for the chemical of interest as a probabilistic average from the three most structurally similar chemicals with experimental data in the program’s library. The “nearest neighbor” method is considered both a hybrid regression and classification model. The “hierarchical clustering” method is similar to the nearest neighbor approach but employs a weighted response (e.g., the Tanimoto Score) to define chemical similarity and predict a more sophisticated activity. Unique to the TEST software platform is a “consensus” option that incorporates all these described QSAR algorithms into a mixed-ensemble algorithm to calculate an overall predictive outcome for the chemical of interest. The Pred-hERG software package is a cardiac-specific software platform that uses QSAR to predict specific toxicological endpoints associated with lethal cardiac arrhythmia [118]. This program includes a library of over 5900 compounds and provides dichotomous, multi-class, and probabilistic predictive outcome descriptors.

Fundamentally, QSAR is a method of predictive toxicology that attempts to predict experimental outcomes for chemicals of interest based on structural similarities to other chemicals with known experimental findings. Therefore, QSAR methods are inherently empirical. All QSAR methods are limited by: (1) the existence of chemicals structurally similar to the chemical of interest in the library; (2) the amount and type of empirical data for toxicologic endpoints available in the program library; (3) the degree of similarity between the chemical of interest and the library chemicals; and, (4) the heuristic algorithms employed internally by the software program to make the QSAR predictions. QSAR methods are valuable for leveraging the existing body of toxicology knowledge in data-poor situations, but they do not eliminate the need for direct in vitro, in vivo and ex vivo evaluation for chemicals of concern.

Future development of cardiac-specific in silico toxicity ensemble models that incorporate systems biology techniques into QSAR predictive platforms have the potential to further minimize future reliance on in vivo testing by linking the relationships between chemical QSAR and specific pathways of toxicity in cardiac tissue. A glimpse of this possibility is provided by Yang et al., in their publication A Computational Pipeline to Predict Cardiotoxicity: From the Atom to the Rhythm [119]. Their proposed approach employs Molecular Dynamic (MD) computer simulations to describe the potential interactions between chemicals and the hERG channel to define the molecular initiating event for a mechanism of arrhythmia at the cellular and tissue-level. Using reaction rate kinetics, the authors then modeled the predicted current flow and transient electric potential in computer-simulated, heterogeneous cardiac tissue. Both two- and three-dimensional (a.k.a., geometric) parabolic time diffusion models were used for applied current and ion concentration to predict the QT prolongation and action potential distributions from the chemical interaction. While complex multimedia ensemble platforms such as this are not in common use or validated in technical guidance documents for regulatory applications, they offer great promise for reducing the need for in vivo animal testing for an area of specific concern to cardiac physiology. Moreover, with the increasing sophistication in culture, experimental data from human-derived and differentiated cardiomyocytes will make the relationship between chemical toxicity and in vitro more defined and in silico approaches more theoretically translatable.

Vascular Toxicity

In Vivo and Animal Models

With respect to drug safety testing, there is no question that pharmacological drug development and safety testing employs a strict regimen of pro-arrhythmogenic testing including screening in cultured cardiomyocytes and animal testing in vivo for aberrant activation of the ECG (see Cardiac Toxicity sections above). However, there is not a similar industry ‘standard of testing’ for drug-induced vascular injury (DIVI) despite the long and well-known association between cigarette smoking and vascular injury, and the influence of this injury on the subsequent development of atherosclerosis. The closest outcome vascular-focused researchers have to the ECG “gold standard” for DIVI is the diagnostic for endothelial dysfunction (ED) by non-invasive ultrasound-based measurement of flow-mediated dilation (FMD). Briefly, FMD quantifies a physiological dilatory response to an acute (e.g., 5 min) ischemia usually induced using a blood pressure cuff above or below the site of measurement, most commonly the brachial artery. A healthy brachial artery will dilate by 15–20% to restore flow following the removal of the ischemic insult. This response can be measured by ultrasound as a continuous event before and after an exposure (e.g., smoking of a single cigarette). This methodologic approach has multiple advantages. It is: (1) non-invasive; (2) each subject serves as their own matched control; and (3) FMD is an accepted early biomarker of vascular harm that is predictive for DIVI pathologies associated with hypertension, stroke, and atherosclerosis. In a landmark paper of 1968, atherosclerotic progression was associated with defined levels of cigarette smoking and independent of alcohol intake levels. Due to the utility and ease of use for this early indicator of DIVI, many human studies in the last eight years have employed FMD in human-subject testing to examine for potential DIVI due to ENDS use. In an initial study in 2016, Carnevale et al. reported that acute (10 min) use of an ENDS impaired FMD response to the same degree as smoking a single cigarette, regardless of the smoking history of the subject [15]. Subsequent publications have reported measurable alterations in vascular stiffness in addition to an impaired FMD response [14, 16]. In all the studies to date where cigarette smoking was included as a “positive control,” cigarette smoking impaired FMD [14, 16]. FMD remains a “gold standard” for predicting DIVI, but its use is limited to investigators or laboratory staff trained in the sonography technique and access to clinical experiment facilities to perform controlled toxicant/drug exposures and ultrasound measurements in human subjects [120,121,122]. Because these are often significant barriers to human-subject research, alternatives include the use of animal surrogates.

An important feature of the FMD is that it is measured in a physiologically complete, living animal or human (e.g., immune, nervous, clotting, etc.), which is why it can provide a robust prediction of health or disease state. Because the physiological processes in the heart and vasculature are largely conserved across mammal species, animal surrogates can provide both biological plausibility and means to elucidate the underlying mechanisms of DIVI. However, there are several issues with using animal surrogates for FMD. Animal surrogates are much smaller than humans, which can make ultrasound measurements of even large conduit blood vessels challenging, and they are unable to “hold still” for the measurement or independently smoke a cigarette or ENDS. Thus, animals need to be anesthetized for FMD measurements to permit careful, technical measurements. Notably, few laboratories in the world routinely use this established, validated approach, which highlights the accessibility barriers of the methodology. The laboratory of Dr. M. Springer, University of California at San Francisco, has published several papers showing that acute (10 min) exposures of anesthetized rats to ENDS-derived aerosols [18, 19, 123], cigarette smoke [19, 124], and, marijuana smoke [125] impairs the FMD response. While FMD can be measured as a localized response at a single point, other vascular parameters such as vascular stiffness (measured by pulse wave velocity) and blood pressures (systolic, diastolic) are systemically regulated by integrated changes locally and across the complete vascular bed. Although not as technically limiting as FMD, vascular stiffness and blood pressure do require specific training, special equipment, technical know-how, can be expensive, and require the use of animal surrogates. Nonetheless, all these outcome measurements are highly predictive for vascular diseases such as atherosclerosis and continue to be used within animal and human-subject studies, as the highly integrated vascular response has not yet been validated in ex vivo or in vitro models.

A more common in vivo technique for DIVI studies is the use of transgenic mice (LDL^−/− and ApoE^−/−). These genetically engineered animals, when fed a high-fat diet, develop atherosclerotic-like lesions in the great vessels and aortic valves which can be quantitatively measured. These animal models show promise as a quantitative measure of systemic vascular harm from chronic exposures to ENDS or air pollutants. Moreover, these models are amenable to switching, repeated exposure, or recovery delay designs for ENDS exposures that would permit the assessment of exposure design-driven changes in lesion area and composition. From a physiologic and translational medicine perspective, atherosclerosis lesion development in mice is a robust, cumulative, integrated, pathological endpoint that parallels human atherosclerosis, but with a greatly accelerated time course. Admittedly, rodent atherosclerosis models are not inexpensive, and have limitations with unknown mechanistic issues surrounding the advanced lesions in the coronary arterial tree that rupture in human patients but not in mice. Several published studies between 2016 and 2020, funded by the tobacco industry, reported that 6–8 month exposures of apoE-null mice to smoke from a heat not burn tobacco product had significantly less atherosclerosis and aortic transcriptional and functional changes than those mice exposed to traditional combusted cigarette smoke [126,127,128,129]. Another long-term exposure study in apoE^−/− mice on a high-fat diet demonstrated that exposures to ENDS aerosols (with or without nicotine) increased atherosclerosis and endothelial dysfunction in a similar direction and magnitude as exposure to smoke from 3R4F reference cigarettes [130]. Such models provide mechanistic insights into the role of vascular inflammation, immune system activation, and lipid transport for the greater scientific community.

A variety of non-mammalian in vivo models can be used to assess the effects of environmental pollutants as well as nicotine and other tobacco product ingredients on vascular development and function, in particular angiogenesis. These include the zebrafish and chick embryo [131,132,133], which provide less expensive, genetically manipulable, relatively static models in which alterations of vascular development and function can be documented. Leveraging lower-order species as surrogates has widespread use in the environmental field due to their close similarity with threatened wildlife species, however, their use in studies that translate findings to human health issues is increasing in the published literature [134, 135].

Ex vivo

Ex vivo studies of isolated blood vessels, endothelial cells and platelets following ENDS exposures have been used in animal models and humans for precise measurement of DIVI pathology when controlled settings for testing can be maintained. In human subjects, post-exposure blood sampling provides a wealth of targets for testing including endothelial progenitor cells such as cardiovascular angiogenic cells [136], platelets for thrombotic endpoints, endothelial cell microparticles, and plasma/sera for measuring cytokines/chemokines, lipids, and other biomarkers of harm [17]. Additionally, vascular endothelial cells isolated by J-wire scraping of the brachial vein lining have also been obtained from human volunteers with informed consent [14]. These biological sampling options provide numerous cardiospecific endpoints that can be evaluated using flow-cytometry and imaging microscopy techniques to detect functional outcomes such as eNOS-dependent nitric oxide (NO) production or eNOS phosphorylation. As blood is fairly easily drawn from human subjects, the potential for longitudinal follow-up and pre-post-study designs remains viable [11]. Research in 2018 from Dr. N. Hamburg’s laboratory at Boston University reports that ENDS use not only impairs FMD response but suppresses eNOS-dependent NO formation in isolated brachial vein endothelial cells [137]. This parallel finding provides both external validity to the FMD measurements and potentially provides some insight into the causal mechanism(s) [14, 137]. In a study of healthy donor platelets in 2022, it is also reported that 14 of 15 common flavorants present in e-liquids including the chemicals, cinnamaldehyde, menthol, and vanillin, had no effect on ADP-induced platelet aggregation at a concentration of 100 μM, which implies that these may have little or no direct effect on platelet function in vivo [138]. While there are definite uncertainties in the findings of these studies, they emphasize the potential power of ex vivo approaches for human health studies and as an efficient approach in the use and evaluation of animal surrogates.

Additional advantages to the ex vivo approach include the potential availability of tissues from multiple organ systems for parallel testing and assessment. In animal surrogate studies, multiple blood vessels can be isolated post-exposure and then tested in parallel to assess for variations in vascular bed sensitivity to different exposure types or concentrations. In human subjects, multiple blood vessels can be obtained as surplus tissues with informed consent following a coronary artery bypass graft (CABG) surgery. In these cases, however, the donor is undergoing surgery to improve flow to the heart usually because of vascular insufficiency due to atherosclerosis or other cardiac pathology, and the findings from these isolated blood vessels can be confounded by disease status and risk factors, which typically include smoking, diabetes, hypertension, and/or dyslipidemia. Nonetheless, the status of endothelial function in these tissues can be measured, and response to tobacco product constituents, including aldehydes, have been reported [139, 140]. Acquiring donor blood vessels from healthy individuals post-mortem, is challenging at best, and at present is not an option for most studies. Less common is the exteriorization of a vascular bed in the living subject for assessment of function in situ. This has been done successfully to examine the effects of air pollutant exposures in rat cremaster and spinotrapezius arteries [141], but in situ approaches are not common due to the surgical skill required and stipulations for dedicated equipment.

Where isolated blood vessels have been assessed for DIVI following animal exposures to ENDS-derived aerosols acutely, short-term, or chronically, ENDS exposure is reported to induce vascular injury, specifically endothelial dysfunction [18, 19, 18, 19]. This finding is supported by studies in different blood vessels (e.g., aorta, middle cerebral artery, etc.) and replicated across multiple species including mice and rats [142,143,144]. There are numerous advantages to adopting such an approach for the assessment of DIVI (as well as other endpoints such as platelet activation, immune activation, etc...), however it requires the use of animal surrogates which conflicts with the mandate to “replace, reduce and refine” animal models in biomedical research. Nonetheless, animal studies with ex vivo assessment of DIVI remain valuable because of the breadth and wealth of scientific data that can be gathered post-exposure and the conserved biological plausibility of these findings as a shared consequence of mammalian biology.

In Vitro

To reduce and refine animal use, the development and application of in vitro approaches are needed to fill the physiological gaps between an intact animal and isolated organs or cultured cells. In vitro models of cardiovascular tissue have been proposed and used for toxicity testing with a variety of optical, -omics, and biochemical techniques as outcome measurements [105, 108]. In mammals, cardiac tissue exhibits an intricate hierarchical structure that consists of several tissue types and is sensitive to external stimuli and a variety of regulatory pathways [145]. Primary isolation of cardiomyocyte culture has historically been the in vitro platform of choice in cardio-toxicology [105, 108, 146]. However, this single-cell type, of course, does not completely recapitulate the structure and function of cardiac tissue or an intact heart [145].

In the past decade, organ-on-a-chip platforms, biomaterial development, and innovation in the use of stem cells have enabled more sophisticated tissue culture models to investigate mechanisms of toxicity in which cardiac endothelial cells, cardiac fibroblasts, extracardiac cells, and vascular cells may play a joint role [147]. Three-dimensional (3-D) tissue culture enables contraction of tissue, realistic tissue structure, accurate ratio of cells types, and phenotypes similar to in vivo tissue [95, 146]. For example, cardiac spheroids, which may use co-cultures of cardiomyocytes, endothelial cells, and fibroblasts organized to mimic cardiac tissue, have been used to study the cardiotoxicity of cancer drugs [95]. Organ-on-a-chip platforms permit toxicological research in an in vitro engineered microenvironment that well captures the structure and physiologically of in vivo tissue through careful selection of cells, substrates, and cues. Engineering of extracellular matrix, geometry, substrate stiffness, and mechanical stress such as fluid flow contribute to the realism of organ-on-a-chip models [148]. Microvasculature-on-a-chip models have been used to study endothelial barrier dysfunction, vascular obstruction, and drug delivery [149, 150]. Heart-on-a-chip has been predominately used for electrophysiology and disease modeling studies [107, 151, 152]. Cardiac tissue organ-chip models have become so sophisticated that chamber-specific tissue models can be fabricated [153]. There is a clear opportunity to utilize such platforms to study the toxicity of environmental particulate matter, including those in ENDS aerosol [148]. In our efforts to replace, reduce, and refine animal use, we can adapt these in vitro cardiac tissue systems in the study of vascular disease. In this section, we will be describing what types of cell sources can be used and the different types of assays that can be performed on these in vitro models to better understand vascular disease, without the need to use animal models.

Sources of Vascular Cells

Vascular cells for use in culture models of disease can be sourced from various lineages. We can classify these into four main progenitor lines: (a) Human endothelial cells (HECs), (b) Human smooth muscle cells (HSMCs), (c) Human endothelial progenitor cells (HEPCs), and (d) Human-induced pluripotent stem cells (iPSC) cells. Within the first two lineages, those obtained from the umbilical cord and the tissue of an adult individual are considered discrete sub-lineages [154]. All of these cell lineages can be purchased commercially, making them easy to obtain and replenish as needed. They can also be used to model a diverse range of vascular diseases (Table 1) as they retain the ability to recapitulate human disease states to a certain extent under specific conditions. There are, however, a few limitations to be considered when using these cells. Human endothelial and smooth muscle cells are primary cell lines collected from a human donor (e.g., aorta, pulmonary artery, coronary artery, lymphatic vessels, umbilicus, etc.) which limits the availability and variability for certain demographic populations. Primary cells can be challenging to maintain, and they have a fixed lifespan in culture. HEPCs can be sourced from bone marrow, peripheral blood, and umbilical cord blood [155], and are commercially available; however, they are heavily used for studies of vascular homeostasis and vascular repair processes. iPSCs-derived cells, first obtained by Yamanaka in 2007, have come a long way in their use for modeling human disease [156]. Several protocols exist that differentiate them into diverse cell types for use in vascular disease models in vitro. Studies of vascular disease in vitro, can be done with endothelial cells (ECs) [157], smooth muscle cells (SMCs) [158], fibroblasts (FBs) [159], pericytes [160], or macrophages [161]. The advantage of iPSCs is that they can be obtained in a minimally invasive way (blood draw or skin biopsy) from any demographic to capture population-specific pathologies for a wide range of diseases, as well as guide applications in personalized medicine. However, iPSC-derived cell lines cannot be purchased commercially and require a significant investment of time and resources by the researcher to phenotypically induce the desired cell types, as well as optimize protocols to the modeling of a specific condition/disease. Traditional cell culture techniques are severely limited in their ability to replicate vascular tissue physiology particularly as single-cell monolayers in a dish. The microanatomy of the vascular system is constructed of multiple cell types in a specific three-dimensional assembly where discrete cell types interact to influence physiologic outcomes. To address this, significant effort over the past decade has been made towards the creation of 3-D vascular organoids that will more faithfully replicate the structural cellular interactions that drive vascular biology [162].

Toxicity Profiling

Toxicity profiling in vitro has become important in modeling vascular disease and assessing the potential effects of drugs, chemicals, or environmental exposures on vascular tissues. In vitro profiling permits a more rapid screening for multiple chemicals of concern (i.e., high-throughput) in addition to insight into cell-specific effects and disease progression and is seen as an ethical alternative to animal models. Various assays and techniques already exist which could be used to address issues specific to the vasculature. Many of the cell types listed in Table 1, have potential applications to assess xenobiotic responses that affect vascular disease development. In the following sections, we discuss the molecular and functional assays that might be performed in an in vitro approach to characterize toxicity in the vasculature for xenobiotics and chemicals of concern.

Endothelium and smooth muscle cells are the major cell types currently used in ENDs toxicity studies of the vasculature. While individual cell monolayers are hard-pressed to recreate the microenvironment of an actual blood vessel, vessel-on-a-chip systems have been created to study the physiological changes present in the microvasculature [163]. These bioengineered tissue chips are assembled as multiple cell types in a 3-D structure with soft-hydrogel/matrix gel as the extracellular matrix, which allows for more realistic tissue forces (e.g., elasticity). Some of these chips also incorporate a microfluidic channel to mimic blood flow for both exposure and effect measurement. This is an important advance for vascular culture techniques as traditional cell culture lacks a consistent flow rate over the cells to mimic blood circulation [164]. The microfluidic vessels-on-chips have been specifically developed to capture circulating blood interactions and microenvironments present in living tissues. In addition to the organ-on-a-chip platform, precision-cut slices of coronary vessels from donors or animals can serve as an ex vivo approach to look for toxicant injury or vascular dysregulation [165]. Although there are multiple 3-D engineered tissues currently in use, none of them have been used to study the cardiovascular toxicity of ENDS aerosols. Advancements to the exposure capability of these testing platforms that reproduce the microenvironments in the lung, microvasculature and coronary vessel transition zones are needed to comprehensively assess for cardiovascular risks of ENDS use, and to bridge the mechanistic data gap between animal and in vitro methods.

Published literature clearly indicates that HUVEC cultures have increased oxidative stress and inflammatory responses, in addition to increased LDH leak and inhibited metabolic activity after exposure to ENDS aerosol, aerosol extracts and their component chemicals [166,167,168]. Treatment of HUVEC with ENDS aerosol extracts has also induced mixed apoptosis and necrosis responses and demonstrable DNA damage [169]. Because of this acute localized toxicity, questions still remain regarding the use of HUVEC as an appropriate screening model for endothelium dysfunction in the vascular bed, or if a shift to a more complex multicellular system is required [170]. At levels found in blood of smokers, nicotine is known to induce proliferative, migratory, and pro-angiogenic responses in HUVEC whereas at higher levels, nicotine is anti-proliferative and cytotoxic [171]. Nicotine is also known to activate the DDAH/ADMA/NOS signaling pathway in HUVEC, which is associated with endothelial dysfunction [172]. Chemical flavorants found in ENDS e-liquids and aerosol, such as cinnamaldehyde and vanillin, are known to induce cell death and inhibit metabolic activity [167]. Similar findings have also been reported for HCAEC, another endothelial cell-line commonly used to screen for toxic effects on the vascular system [173]. A finding unique to HCAEC is that treatment with cigarette smoke extracts induced activation of inflammatory responses, but this induction was not seen after treatment with extracts of ENDS aerosol [173]. Another study by Wu et al. in 2018, using HAEC, reports treatment with nicotine caused LDH leakage, oxidative stress, and a robust inflammatory response that triggered apoptosis [174]. In the same study in 2018, atherosclerotic mice given the same nicotine in vivo demonstrated enhanced formation of vascular lesions, which was considered supportive of the findings and effects seen in the HAEC [174]. Nicotine treatment has produced similar effects in other types of epithelium as well, including oxidative stress and inflammation activation, which are associated with endothelial dysfunction. Decreased eNOS expression in HUVEC and inhibited eNOS phosphorylation in HAEC have also been reported after treatment with either ENDS aerosol condensate or sera of chronic ENDS users [11, 175]. Inhibited eNOS coupling activity results in a reduction of nitric oxide levels in the vascular system blunting a protective mechanism for vascular regulation and disease prevention [176]. Published studies have suggested that decreased expression of eNOS may lead to the endothelial dysfunction associated with ENDS use through altered PI-3K/Akt signaling [177]. In addition to these studies in primary cells used in vascular toxicology studies, iPSC-derived endothelium (iPSC-EC) have been used in ENDS aerosol exposure studies, and treatment-induced inflammation and oxidative stress were both reported in the iPSC-EC as well [178].

The vascular smooth muscle cell (e.g., HASMC) is another terminal cell phenotype used in vascular toxicity models. Treatment of HASMC with ENDS aerosol condensate from 25 W, but not 8 W, setting and with cinnamon flavor increased IL-8 secretion (a pro-inflammatory marker) without a change in cell viability or in ROS [179]. In additional vascular smooth muscle cell lines (i.e., bovine and human aortic), nicotine stimulated cell proliferation and migration [180,181,182]. In one of these studies, nicotine treatment is noted to increase the secretion of bFGF and TGFβ by VSMC in a concentration-dependent manner [180]. Increased levels of bFGF and TGFβ are linked with vascular remodeling, i.e., stiffening and thickening of the vascular wall [183]. Nicotine treatment is also known to increase arterial stiffness, collagen overexpression, and local remodeling of vascular tissues [184, 185]. In addition to HASMC, harvested aortic vascular rings have also been used ex vivo to directly test the toxicity of e-liquids and their constituents including flavorants. Vascular rings possess both functional endothelium and vascular smooth muscle cells (media), and thus, the sensitivity of each layer can be discerned under appropriate conditions. Minimal toxic changes to the endothelium and VSMC layers of rat aorta were observed with the direct addition of nine individual ENDS e-liquid flavorants [186]. These notable differences in outcomes between numerous published studies, likely result from the many differences that exist in the incorporation of these toxicity platforms including choice of species, blood vessel selection, cell origin, duration of toxicant exposure, and the use of individual flavorant constituents, flavorant extracts, or aerosol condensates as the test article.

High-Throughput “-Omics” Assays

One of the newer tools we can use to assess the toxicity profile of ENDS and other tobacco products is the high-throughput “omics” platforms [187]. These tools rely on the integration of biomolecular technology data streams into an algorithm that can create a comprehensive molecular fingerprint for specific chemical exposures and predict the functional impacts in a biological system. In the context of tobacco science research, the use of multi-omics could provide a global view of the molecular changes and functional impact in these in vitro vascular models after exposure to ENDS aerosol, e-liquid or their chemical constituents. Techniques are applicable to studies of vascular toxicity include transcriptomics, proteomics, metabolomics, and epigenomics, and are classified and described in Table 2. Obvious advantages of a multi-omics approach are: (1) generation of comprehensive molecular profiles that can be chemical- or agent-specific as well as cell phenotype driven; (2) detection of systems biology level interactions and crosstalk that can be used in mechanistic inquiry; and, (3) support hypotheses generation to identify constituents, novel targets, pathways, and/or mechanisms involved in adverse effects of tobacco products on vascular function.

Even though these techniques could provide an insightful understanding of how gene and protein expression, protein as well as metabolite changes evolve after exposure to the ENDS aerosol (particle or aerosol phase) or its constituent chemicals, there are some use limitations. While the molecular biology techniques that feed data into the analysis are now in common use, the data analyses are challenging as they demand generous amounts of time, are not inexpensive, and require specific computer software or service licenses to integrate and interpret the multi-omic data streams. Despite these hurdles, multi-omics approaches are tools providing rich biological data that help examine our general hypotheses regarding chemical and/or cell-specific response patterns/toxicity, as well as mechanistic and systems biology investigations.

Functional Assays

In addition to the advanced multi-omics assays, there are more traditional cardiovascular-specific techniques and function assays that can be performed in the laboratory setting to examine the effects of ENDS and other tobacco products. These are techniques that primarily focus on detecting changes in biological or functional parameters of vascular cells in vitro. Some of the outcome measurements are described in more detail in Table 3, and include relatively gross measurements such as cell death [188], proliferation [189], cell migration [190], angiogenesis [191, 192], membrane permeability [193, 194], as well as more granular parameters such as NO release [195], and ROS production [196].

The use of functional assays provides a direct assessment of functional outcomes relevant to vascular mechanisms of pathophysiology which are known to occur at the cellular level to bridge exposure responses from in vitro platforms to the whole organism. Functional assays are often more simplistic and accessible compared to advanced techniques such as ‘omics approaches, as functional assays are often available as commercial kits with no need for optimization or a significant depth of understanding for the chemistry, enzymology, or immunology on which they are based. However, functional assays, even though widely used, have use limitations. Functional assays provide preselected, targeted information more oriented towards characterizing a singular functional parameter. Therefore, they may fail to capture the complexity of molecular interactions and regulatory networks operating in response to treatment, even if the preselected parameter is part of the cellular response mechanism. These assays will provide snapshots in time of molecular activity, and thus, a time-dependent array of functional assays is needed to track mechanistic progression or interactions as with any appropriately planned experimental design.

Both multi-omics and functional analysis are valuable tools for studying in vitro vascular models, each with its strengths and limitations. Multi-omics provide a data-rich molecular profile and systems-level insights, while, on the other hand, functional analysis is recommended for a direct assessment of singular outcomes. The selected approach is most often driven by the research question(s) as well as the depth and breadth of data needed to answer it, available laboratory resources, and the amount of time and money that can be allocated to the study. In the best-case scenario, integrating multiple approaches will provide the most comprehensive assessment of the interplay between molecular pathways and functional outcomes in vascular biology research.

In Silico

As described above for investigating specific cardiac toxicity markers, in silico approaches are similarly applied in vascular injury research or sometimes together as was done to explicate COVID-19 as a CVD [197]. The in silico process can work in a reversed engineered fashion by interrogating databases with biomarkers of harm or transcriptional changes across a spectrum of related CVD landscape (e.g., atherosclerosis due to high-fat, or hypertension) to derive a subset of overlapping markers/genes that point to perhaps known or novel targets (Graphical Abstract), and analogous to a drug-repurposing algorithm.[198] Abundant publicly available databases (clinical, animal models, cells, etc.) on DIVI provide ample starting material for understanding how drugs of a specific class may initiate DIVI. An obvious goal of drug design is to avoid the costly development of a drug that ultimately fails in expensive animal studies or human Phase I trials because of unexpected DIVI. For assessing ENDS risk, a similar approach can be attempted by first defining an element of DIVI that is searchable, e.g., “endothelial dysfunction”, coupled with a tobacco-related treatment, e.g., “nicotine,” and then searching relevant databases. The rise of artificial intelligence (AI) and machine learning apply high-throughput capacity coupled with analytical proficiency to improve prediction regarding both the sensitivity and specificity of upstream DIVI biomarkers but also the nature of the underlying mechanism of the disease process itself. Because the sheer quantity of data yielded by the high-throughput ‘omics’ assays (e.g., single-cell RNAseq; described above) can be coupled with AI and machine learning algorithms such that there can be a major reduction in the use of costly screens done in vivo. When such screens are done, the power of comparison with well-characterized disease-based databases will enhance the quality and quantity of data collected. Although there appears a need for coordinated experimental validation of these approaches, this is also in progress, so that predictive algorithms are not being created without biologic confirmation.

Coagulation/Thrombosis Pathways

One of the issues surrounding ENDS risk assessment is that the production of aerosol from an ENDS relies on chemically created mixtures (the e-liquid) with formulations that vary between manufacturers as well as by individual products; therefore assessment of CVD risk is highly impacted by the individual constituents ENDS e-liquids such as flavorants [178]. Some studies based on these flavor elements have been published that directly or indirectly address effects on the physiologic chemistry of the cardiac system—the coagulation and thrombosis pathways. It has been reported that exposure of cultured human endothelial cells to any one of nine flavorant chemicals, including menthol and cinnamaldehyde, led to concentration-dependent reductions in NO production, with increases in markers of oxidative stress, inflammation, and apoptosis [137]. Because NO levels in vivo inhibit platelet activation and serve as a biological brake on clot formation [199], it has been suggested that these flavorants diminish NO production and increase the risk of platelet activation and thrombotic stroke [200]. This proposed mechanism for CVD morbidity is supported by acute studies in mice that demonstrated exposure to ENDS aerosol-induced systemic prothrombotic effects [21]. Because of the marked chemical variability between e-liquid formulations in marketed ENDS products, and the slow pace of testing using animal inhalation models, investigators are considering the use of high-throughput in vitro assays for platelet toxicity to accurately and more expeditiously detect and characterize hazards to normal platelet function due to ENDS e-liquid chemical constituents. One of the potential options is the Børn method of turbidimetric aggregation, which relies on use of a specially designed piece of medical equipment, the Chrono-log Lumi-aggregometer 480VS (Chrono-Log, Havertown, PA) [201]. This system can perform aggregometry assays with freshly isolated human or animal platelets, and has the capacity for limited multiple channels of detection. There are design considerations for this approach, such as the preparation time of platelets, and multiple agonists of platelet activation (e.g., ADP, collagen, PMA) that need to be run one agonist at a time with each e-liquid constituent or mixture. Nonetheless, because of the immense importance of NO to the control of platelet activation, this relatively simple detection methodology has great potential as a biomarker of hazard for medically significant cardiovascular pathophysiology and stroke.

Gaps in Knowledge

The knowledge gaps in cardiovascular toxicity are barriers that limit the use of new approach methodology (NAM). While 2-D and 3-D iPSC-derived cardiac and vascular tissue are significant improvements in platforms for toxicology testing, they remain costly methods that still have deficiencies in capturing the context of the intact myocardium or a multi-organ system interaction. Repositories for iPSCs do exist, however, there is a fundamental need to create cardiac and vascular organoid repositories with open access so more researchers can use this technology to examine the cardiac and vascular effects of the expanding number of new tobacco products and their multitude of constituents. The multi-media ensemble platforms with computer-generated virtual tissue biology are becoming more accessible, however, platform acceptance by industry and importantly the regulatory community is still uncertain and takes time. In silico approaches such as QSAR analysis of toxicity is affordable and have the benefit of reducing material resources, e.g., animal use in many cases. However, even in silico findings have limitations such as applicability and validation, i.e., the need for more experimental data, particularly for emerging approaches and novel assays and cell lines that better reflect the unique cellular physiology of myocardial and vascular tissues.

How Do We Make a Better “Mousetrap” Specific for Cardiovascular Disease Risk?

A key element in any assessment in regulatory science is the ability to make a complete hazard assessment for a chemical, chemical mixture or byproduct such as combusted smoke or ENDS aerosol. To do that, one must not only have an adequate ‘weight of evidence’ supportive of a regulatory conclusion, but this evidence must be from generally accepted methodologies with outcome measures that are relevant to the target tissues, organ systems or pathophysiology(ies) of concern in the end user. The hazard assessment of new tobacco products, such as ENDS, as well as traditional combusted products for CVD outcomes, suffers from completely “fixable” deficiencies in both areas.

As described in the previous body of this manuscript, multiple new and traditional methodologies can be used to measure therapeutic or adverse effects of relevance to cardiac and vascular physiology. In addition, there are multiple new primary and inducible cell lines accessible to provide specific cardiac and vascular cell phenotypes that can be used for traditional and 3-D or chip fluidics culture platforms. The authors have acknowledged that some cost and informed consent barriers exist which could affect widespread use, but there are regional academic and private institutions that have the ability now to overcome these obstacles and publish findings using CVD-relevant culture systems. If history holds true, increasing use by regional centers will reduce these barriers for other researchers. The same can be said for the familiarity and training issues for both new approach methodologies (NAM) and cell-line husbandry. Like the bacterial gene manipulation techniques of the 1980s and 1990s, mammalian cell culture and even stem cell culture are now taught as academic laboratory courses in colleges and universities across the United States. The sharing of ‘omics data in the public domain has enabled both public and private firms to offer advanced data-mining and pattern recognition services with an accessible fee structure, so that collaborative laboratory partnerships and even investigators without wet-lab access can contribute to the study of relevant disease biomarkers [202, 203].

At present, there are multiple experimental design components and outcome measurement technologies that are waiting to be coupled together as a CVD or cardiac toxicity testing platform by investigators. As with any new approach or technique, there remains uncertainties as to how robustly these potential assay platforms will predict population-relevant effects in humans—but this will never be known until the work is attempted. It is also true that these techniques in their current state are not fully optimized for CVD, but until they are examined as study protocols and the results published, the optimization cannot truly begin. In addition to publications fully devoted to screening and evaluation of tobacco product toxicity, properly constructed methods development publications are also supportive to the weight of evidence needed for hazard documentation and regulatory evaluation.

This leads us back to the regulatory consideration of tobacco product effects on CVD [204]. Both the deficiencies in the weight of evidence for documentation of CVD hazard and the general acceptance of new platforms for CVD hazard detection require visible numbers of publications in peer-reviewed literature. Waiting for the biologically perfect model system before we begin is not a solution. As British statistician George E. P. Box stated many times—All models are wrong, but some are useful. True to form in that regard, even method development publications are useful additions to building the body of evidence surrounding ENDS use and its potential contributions to CVD. Building the body of evidence may require stepping away from the comfortable traditional methods and exploring collaborations with specialists such as engineers and cell biologists in the greater scientific community. However, the cardiac and vascular experimental tools of today were built yesterday with similar uncertainties; it is not a question of if, but when, we will do the same.

CVD remains the leading cause of morbidity and mortality in the world. Tobacco smoking remains the number one preventable risk factor for the development of CVD. There remains uncertainty about the long-term health consequences of ENDS [204]. Do we want to continue to just “roll the tape” until it becomes clear whether or not ENDS use contributes to the burden of diseases for CVD or other diseases including cancer and pulmonary diseases? Thus, it is imperative that we utilize our intelligence and resources to help us “peer into the crystal ball of the future” and correctly predict the future with respect to CVD risk. How are we doing so far? Conventional exposure studies using “gold standard” biomarkers of harm (e.g., FMD) provide numerous examples of ENDS-induced harm in people and in animal studies in vivo [17, 204]. In addition, there are increasing clinical case reports of acute cardiopulmonary disease in ENDS users, but lifestyle confounders and a lack of published scientific data limits our ability to declare a mechanistic link in many of these cases. Similarly, exposure studies using emerging in vitro platforms and ‘omics studies provide an unprecedented wealth of data that in many ways exceed what we could measure 10–20 years ago. However, as these new technologies have not been widely applied in cardiac- or vascular-specific cell and tissue culture platform settings, it remains to be seen how well these data-rich studies predict CVD risk in the human population at large. This likely will require the integration of multiple systems such as the immune system (e.g., iPSC-derived monocytes, T-cells, and B-cells) and nervous system in co-culture with cardiovascular cells to better understand pro- and neuro-inflammatory processes induced by ENDS use, i.e., ‘organ systems on a chip’. The ultimate way to answer these questions is to actively incorporate these new techniques and approaches into our studies, submit the findings under the lens of peer review across the fields of CVD and tobacco research, and the rigorous studies shall emerge. Some may argue that these new technologies are “not ready” for use in cardiovascular toxicity studies, but we cannot know this for sure until the use of these technologies becomes more prevalent in peer-reviewed studies of cardiac and vascular toxicity. The power to make this happen is in our own hands. Without a general acceptance of NAMs as reliable sources of biological outcome measures in our own field paired with independent published reports of detectable tobacco product hazards, cardiovascular toxicity will remain a relatively silent voice in the tobacco regulatory arena.