Hepatotoxicity due to herbal dietary supplements: Past, present and the future

Abstract

Dietary supplements (DS) constitute a widely used group of products comprising vitamin, mineral, and botanical extract formulations. DS of botanical or herbal origins (HDS) comprise nearly 30% of all DS and are presented on the market either as single plant extracts or multi-extract-containing products. Despite generally safe toxicological profiles of most products currently present on the market, rising cases of liver injury caused by HDS – mostly by multi-ingredient and adulterated products – are of particular concern. Here we discuss the most prominent historical cases of HDS-induced hepatotoxicty – from Ephedra to Hydroxycut and OxyELITE Pro-NF, as well as products with suspected hepatotoxicity that are either currently on or are entering the market. We further provide discussion on overcoming the existing challenges with HDS-linked hepatotoxicity by introduction of advanced in silico, in vitro, in vivo, and microphysiological system approaches to address the matter of safety of those products before they reach the market.

1. Herbal dietary supplements: Introduction

Dietary supplements (DS), which include vitamin, mineral, and botanical extract formulations, have gained widespread acceptance and popularity both in the United States and worldwide. In the U.S., after implementation of the Dietary Supplement Health and Education Act (DSHEA) in 1994, DS sales increased from $4 billion to $40 billion by 2019 and continue to grow (Bailey et al., 2011; de Boer and Sherker, 2017). The estimated number of DS on the U.S. market has risen from ~4,000 in 1994 to more than 80,000 today (Navarro et al., 2017). Data from the National Health and Nutrition Examination Survey (NHANES) and other sources indicates that 50–70% of adult Americans use DS, with 20% regularly consuming herbal dietary supplements (HDS) (Dwyer et al., 2003; Roe et al., 2018; Gahche et al., 2018). The most recent data suggest that $9,602 billion were spent in the United States on HDS in 2019 (Smith et al., 2020).

The United States Food and Drug Administration (U.S. FDA) does not allow for specific DS claims regarding treatment or prevention of certain diseases. However, it does allow for “structure function” claims (e.g., “supports overall wellness”, “provides immune support”, or maintains gut health,” etc.), which are extensively used by manufacturers to attract customer attention. Recent surveys indicate that the COVID-19 pandemic led to a spike in the popularity of select DS, especially those touted to optimize healthy immune function such as elderberry, vitamin C, vitamin D₃, quercetin, and zinc, as well as combinations of these (Aysin and Urhan, 2021; Norton et al., 2022).

2. Evidence of HDS-induced liver injury

The U.S. FDA regulates DS more like it does foods than drugs and, therefore, does not require pre-market assessments of safety and effectiveness. Commonly used dietary ingredients, including many botanical extracts that were in use prior to October 1994, are referred to as “old dietary ingredients” (ODI) and are “generally recognized as safe” (GRAS). For novel nutritive constituents introduced to the market after October 15, 1994, a new dietary ingredient (NDI) notification must be submitted to FDA that should provide convincing safety data on the constituent. However, it must be emphasized here, no authoritative list of dietary ingredients marketed or in use before 1994 exists. On account of their non-drug classification, DS can be legally sold over-the-counter (OTC) without a prescription and, as happens in most cases, without a doctor or pharmacist’s advice. Thus, the toxicity potential of such products is not realized until after their ingestion by the consumer.

This loose regulatory oversight regarding DS safety is concerning as the evidence of serious adverse effects associated with their use, although infrequent, is not without precedent (Roe et al., 2018). To date, hepatotoxicity is among the most frequent adverse events reported through the U.S. FDA Center for Food Safety and Applied Nutrition (CFSAN) Adverse Event Reporting System (CAERS). Furthermore, the Drug-Induced Liver Injury Network (DILIN) reports that about 20% of all drug-induced liver injuries (DILI) in 2013 were attributable to HDS, which gave rise to a separate category known as HDS-induced liver injury, or HILI (de Boer and Sherker, 2017; Navarro et al., 2017; Roytman et al., 2018). Many of those cases required hospitalization, others required liver transplantation, and some resulted in death. For the last, subjects often succumbed to acute liver failure, while others died due to complications related to their liver transplantation status.

To date, most HILI cases can be traced to DS formulations containing unusual and heretofore untested combinations of exotic botanical extracts and/or purified phytochemicals, poorly researched new dietary ingredients, products intentionally adulterated with approved or unapproved drugs, or combinations of these (Miousse et al., 2017; Woo et al., 2021). Multi-ingredient DS are complex mixtures of phytochemicals whose combined pharmacological and toxicological effects are often unknown or can differ markedly when compared to their ingestion as single entities. Such combinations may give rise to unanticipated adverse effects usually not attributable to consumption of an individual constituent or combination of constituents. A number of such events have been reported over the last two decades for multi-ingredient products like Slimquick, Hydroxycut, OxyELITE Pro, and several formulations marketed by the company Herbalife (Adike et al., 2017; Chatham-Stephens et al., 2017; Navarro et al., 2017). Interestingly, all of these products were marketed for exercise performance enhancement and/or weight loss. A problematic subset of this category are DS marketed to the bodybuilding community, many of which have been linked to liver injury with clinical manifestations typical for anabolic steroid abuse (i.e., cholestatic liver injury). In several instances, chemical analysis of these products revealed them to be adulterated with synthetic anabolic steroids, thus confirming the clinical diagnosis (El Sherrif et al., 2013; Navarro et al., 2017; Durazzo et al., 2022).

Still another group of DS that pose potential HILI risks are those designed to fill novel market niches that quickly enter into commerce lacking comprehensive toxicological and pharmacological assessments (Woo et al., 2021). Representative of this category are cannabidiol-containing products. Cannabidiol (CBD) is a biologically active, non-psychotropic phytoconstituent of Cannabis sativa that has gained significant inroads in the U.S. market in recent years, with seemingly endless health claims. Yet these promises may come at a cost, including a range of adverse side effects of which hepatotoxicity and CBD-drug interactions are among the most concerning (Zendulka et al., 2016; Ewing et al., 2019b; Ben-Menachem et al., 2020; VanLandingham et al., 2020).

3. Major contributors to herbal and dietary supplements-induced liver injury

The rapid market growth of DS has been paralleled by a significant increase in multi-ingredient DS-associated adverse effects stemming from deficiencies in premarket safety assessments (Gurley et al., 2018). Recent studies report that increased supplement usage during the COVID-19 pandemic further enhanced HDS-induced liver injury (Nagral et al., 2021). Pharmacodynamic interactions and metabolic activation of toxic natural products among various phytoconstituents are likely important contributors to untoward health effects with these products (Gurley et al., 2018; Wang et al., 2021). In this regard, potential interactions between various botanical ingredients present in multi-ingredient HDS are particularly concerning. For instance, caffeine (CAF), a major constituent of most HDS marketed as weight loss aids or exercise performance enhancers, has long been recognized as a cardiovascular and central nervous system stimulant; however, it is also well-known for its ability to augment hepatotoxicity of other stimulants and sympathomimetics (McNamara et al., 2006; Gurley et al., 2015). Supplement formulas containing synthetic CAF and Yohimbe extract (YOH), a natural source of the hypertensive agent yohimbine, are notably troublesome. According to the California Poison Control System, YOH-containing HDS have been associated with a disproportionate number of severe outcomes (Kearney et al., 2010). Furthermore, CAF and YOH were present in a number of weight-loss products removed from the market following reports of liver injury; these include Hydroxycut and OxyELITE Pro that will be discussed below (Fong et al., 2010; Foley et al., 2014; Roytman et al., 2014; Avigan et al., 2016).

New dietary ingredients (NDIs) that entered the market after October 15, 1994, are responsible for a significant proportion of adverse events associated with DS. In the case of NDIs constituting a single plant extract, these too are complex phytochemical mixtures and, thus, may be perceived as “multi-ingredient” products. For example, green tea extract (GTE) is a common DS ingredient that has been linked to liver toxicity. Green tea is one of the world’s oldest and most popular beverages, but its widespread consumption as a concentrated extract has only come about since 1994. The most probable cause of GTE-mediated liver injury appears to be epigallocatechin gallate (EGCG), which often comprises over 50% of GTE phytoconstituents and whose hepatotoxicity has been demonstrated in various animal models (Lambert et al., 2010; Church et al., 2015). For other NDIs, like Hoodia gordonii, Garcinia cambogia, or Scutelleria, which are representative of many under-researched plant species that make it onto the global market as DS, it is much more challenging to identify potential toxicants.

Other contributing factors to HDS-induced liver injury can be linked to poor adherence to current Good Manufacturing Practices (cGMP). This is often associated with the manufacturer’s usage of unauthenticated raw material as well as violations (unintended or otherwise) of responsible manufacturing principles. Results received under the Freedom of Information Act indicate that over half of all dietary supplement manufacturing facilities inspected by the U.S. FDA were found to be non-compliant on at least one item (ConsumerLab, 2020).

Many of the raw plant materials for HDS manufactured in the U.S. hail from Asian markets, with China and India producing the bulk of the supply base (Xiang et al., 2022). Less stringent controls for agricultural practices in these countries are often associated with incidences of raw materials being contaminated with heavy metals or pesticides whose use has been banned in the U.S. Furthermore, inappropriate processing, storage, or shipping conditions may subject both raw material and finished product to bacterial and fungal contamination. While establishing the causality of liver injury due to the presence of the above-mentioned contaminants in a product labeled as a dietary supplement is a challenge, evidence of such contamination has been documented (Liu et al., 2015). Furthermore, the National Center for Complementary and Integrative Health (NCCIH) warns about such potential (NCCIH, 2019).

Last, but not least, is the problem of product adulteration with either conventional or unconventional medications (Durazzo et al., 2022). As previously mentioned, bodybuilding DS spiked with anabolic steroids present the largest cohort of violators within this group. Another frequently encountered class are those marketed for sexual performance enhancement. Such products may be adulterated with FDA-approved phosphodiesterase-5 inhibitors for treating male erectile dysfunction (e.g., sildenafil, tadalafil, vardenafil), or with various unapproved analogs; however, there is little evidence of these products being associated with hepatotoxicity (Brown, 2017; Skalicka-Woźniak et al., 2017; Yéléhé-Okouma et al., 2021). Still other products may be adulterated with synthetic versions of natural phytoconstituents. A recent example of this practice was evidenced by the multi-ingredient weight-loss product OxyElite-Pro, in which an extract of bael tree (Aegle marmelos) bark, was replaced with synthetic aegeline, a compound not approved for human administration (Miousse et al., 2017).

4. Most prominent historical cases

4.1. Single ingredient (one plant extract) products

4.1.1. “Bush” tea

No HDS can be considered to contain just a single ingredient as each plant extract is itself a diverse collection of concentrated phytochemicals; nevertheless, many HDS contain only one plant extract, and products in this category are often referred to as “single ingredient” products. With that definition in mind, one of the most prominent historical examples of botanical-induced liver injury is the sinusoidal obstruction syndrome that was commonly observed in Jamaica and whose occurrence was attributed to consumption of the so-called “bush tea” in the middle of the 20th century (Bras et al., 1954). This tea was brewed from the Senecio plant and was widely used in the Caribbean to treat various gastrointestinal disorders. However, high content of pyrrolizidine alkaloids (PAs) in the plant caused obliteration of hepatic vein associated with toxic destruction of hepatic sinusoidal endothelial cells with further development of fibrotic scars and cirrhosis (Bras et al., 1954; Fan and Crawford, 2014). Despite the discovery of bush tea’s toxicity in the 1950s, it remained popular in Jamaica for decades, as approximately 18% of hepatic cirrhosis cases reported in Jamaica in the late 20th century were attributed to bush tea consumption (Williams et al., 1997).

Senecio is not the only botanical widely used in traditional medicine that contains PAs. The infamous Symphytum officinale, commonly known as comfrey, has been reported to cause fibrous obliteration and destruction of hepatic veins with the development of cirrhosis, all of which were attributed to high quantities of PAs in the plant (Yeong et al., 1990; Stickel and Seitz, 2000). Furthermore, sporadic cases of sinusoidal obstructive syndrome are being reported, including a prominent case in an 18-month-old child caused by consumption of tea Adenostyles alliariae (Alpendost) that was erroneously gathered in place of another plant (Sperl et al., 1995). Interestingly, reports on hepatotoxicity associated with consumption of other teas, potentially not containing the PAs, such as red bush tea (rooibos tea) – Aspalathus linearis (Burm.f.) R. Dahlgren have also started to appear in the peer-reviewed literature (Sinisalo et al., 2010; Engels et al., 2013; Reddy et al., 2016). The latter case, although clearly associated with overconsumption of tea (approximately 10 cups a day), is evidence that “natural” does not necessarily mean “safe”, and that toxicity margins exist even for natural products that seem to be fairly innocuous.

4.1.2. Green tea extract (GTE)

Today, the most recognized single ingredient HDS often linked to hepatotoxicity is green tea and green tea extract (GTE). Green tea is one of the most widely consumed beverages in the world and has been touted for numerous positive health effects attributable to its principal phytoconstituent: catechin polyphenols. Various HDS whose formulations center around GTE have been developed and marketed as weight-loss aids, and it is with these products that the hepatotoxic potential of green tea has come under scrutiny. Green tea has an excellent safety record when consumed as a beverage, but when ingested as a concentrated extract, exposure to catechin polyphenols is heightened, especially in a fasted state (Sarma et al., 2008; Oketch-Rabah et al., 2020). A large number of green tea-containing DS are comprised only of GTE, whereas others are multi-ingredient products in which GTE is either a major or minor component. It has been speculated that exposure to concentrated catechin polyphenols predisposes some consumers to liver injury.

The GTE extraction process typically utilizes water or aqueous solutions, but sometimes organic solvents such as methanol, chloroform or dichloromethane may be employed, the residues of which may be toxic. Still other methods may utilize exchange resins which can concentrate not only catechins but other toxic impurities, as well. In short, the process of extraction greatly increases catechin content and potential contaminants (i.e., pesticide residues) compared to what is present in brewed green tea.

Liver injury as a result of green tea beverage consumption is extremely rare, and only a few anecdotal cases have been reported, usually after consumption of large volumes of the decoction (LiverTox, 2022; Teschke et al., 2012). On the other hand, the growing popularity of GTE and GTE-containing products over the last two decades has been paralleled by numerous cases of liver injury that, with varying degrees of probability, can be attributed to ingestion of these DS (Javaid and Bonkovsky, 2006; Jimenez-Saenz and Martinez-Sanchez Mdel, 2006; Sarma et al., 2008; Navarro et al., 2013; Zendulka et al., 2016; Navarro et al., 2017). In many such cases, GTE was a constituent of a multi-ingredient product and causality was difficult to establish; however, a large prospective study conducted in post-menopausal women found serum ALT levels elevated in 6.7% of study participants who ingested GTE compared to 0.7% of controls (Dostal et al., 2015). Furthermore, a re-challenge of selected participants with GTE resulted in a rapid recurrence of elevated ALT levels, which returned to normal once the product was discontinued. In general, reported cases of GTE-induced liver injury were often the result of ingesting high quantities of the product (above 800 mg) or due to prolonged GTE consumption (from several weeks to several months). Clinically, pathological findings resembled those of acute hepatitis, mostly of a hepatocellular type with significant elevations of liver enzymes (Navarro et al., 2017; DILIN, 2018; Hu et al., 2018).

The exact causes and mechanisms of GTE-associated liver injury remain poorly understood, but accumulating evidence indicates that the presence and concentration of certain catechins, epigallocatechin gallate (EGCG) in particular, is an important determinant of hepatotoxicity. EGCG usually comprises >50% of total catechin content in green tea and GTE. Increased EGCG bioavailability, as a result of saturable drug metabolizing enzyme and efflux transporter activity, is hypothesized to play a major role in GTE-induced hepatotoxicity (Oketch-Rabah et al., 2020). Other factors may contribute to or exacerbate GTE/EGCG-induced hepatotoxicity. For instance, animal studies and clinical trials have demonstrated that GTE consumption in a fasted state is associated with higher catechin plasma concentrations compared to a fed condition, and this substantially potentiates EGCG hepatotoxicity (Isbrucker et al., 2006; Kapetanovic et al., 2009; Isomura et al., 2016; Shanafelt et al., 2009, 2013). Also, the role of genetic predisposition in GTE-induced hepatotoxicity is becoming increasingly recognized. For instance, Hoofnagle and colleagues reported an HLA-B*35:01 allele to be tightly associated with liver injury among GTE users (Hoofnagle et al., 2020).

Recently, a study was performed that examined the incidence and severity of GTE hepatotoxicity while minimizing the influences of genetic predisposition, fasting, and co-ingested stimulants (i.e., caffeine) and impurities (Gurley et al., 2019). This study used B6C3F₁ mice, a strain known for average sensitivity to hepatotoxicants. Phytochemical characterization of the investigated decaffeinated GTE (dGTE) product demonstrated that solvent residues and other impurities were well below the level of detection (LOD) and also provided detailed information on catechin type and content. Based on this, dGTE within the 1X – 10X range (65.9–659 mg/kg) of mouse equivalent doses (MED) were gavaged to fed mice for ten days. The highest dose (659 mg/kg dGTE) had 15 mg of catechins, of which EGCG constituted 10.68 mg. Detailed toxicological analysis demonstrated no evidence of hepatotoxicity and a lack of transcriptomic changes in the livers of experimental mice, suggestive of nominal catechin bioavailability in the fed state. It is worth mentioning that a similar approach was taken in studies assessing the toxicity of the multi-ingredient product OxyELITE-Pro and a cannabidiol-rich hemp extract (discussed in detail in section 4.2.2 and 5.5, respectively) in which overt toxicity was observed for the highest doses of each product (Miousse et al., 2017; Skinner et al., 2018; Ewing et al., 2019a).

In 2020, following an extensive review of the medical and scientific literature, the United States Pharmacopeia (USP) proposed a cautionary labelling requirement in their monograph “Powdered Decaffeinated Green Tea Extract” that advises taking the product with food and not to use, or immediately discontinue, if liver toxicity is suspected (Oketch-Rabah et al., 2020).

4.2. Multi-ingredient products

4.2.1. Ephedra and ephedra-free DS

Multi-ingredient HDS are those whose formulations comprise more than one botanical extract. In their heyday, the vast majority of Ephedra-containing HDS were multi-ingredient products. From their initial introduction onto the US market in 1994, Ephedra-containing DS were linked to a plethora of serious adverse health effects that set the stage for a decade of scientific and legal wrangling – the deliberations of which were fueled by a disquieted media – ultimately leading to their removal a decade later by the FDA. While cases of hepatotoxicity were encountered during this period (Nadir et al., 1996; Estes et al., 2003; Bajaj et al., 2003), they were greatly outnumbered by cardiovascular and neurological injuries linked to Ephedra use. This is understandable given that most formulations blended extracts of Ephedra – a natural source of ephedrine alkaloids (e.g., ephedrine, pseudoephedrine, norephedrine, methylephedrine, etc.) with various sympathomimetic effects – with a litany of other botanical extracts (e.g., guarana, green tea, yohimbe, Citrus aurantium, Coleus forskohlii, etc.), many of which are botanical sources of cardiovascular and central nervous system (CNS) stimulants (e.g., caffeine, yohimbine, synephrine, forskolin, etc.). Ironically, blends of purified ephedrine, norephedrine, and caffeine had been available as OTC medications in the 1980s. However, these synthetic combination products – known as “amphetamine look-alikes” on account of their adverse CNS effects – were removed from the market in 1988 owing to significant numbers of heart attacks, strokes, seizures, and psychotic events linked to their usage. Thus, with the decade-long availability of Ephedra HDS, serious adverse health effects should have been antici pated. This startling array of adverse events could be traced to the farrago of botanical extracts found within each formulation, whose mélange of unique phytochemicals begat complex pharmacology, soon followed by the attendant toxicological consequences, both predicted and unanticipated. Ephedra-containing HDS, therefore, epitomized the fallacy of ill-conceived botanical extract combinations; yet, despite their removal from the market in 2004, a new genre of products, although retaining the same flawed formulation concepts, quickly filled the marketplace void: Ephedra-free HDS. It would appear, however, that removing Ephedra as a key ingredient from multi-component weight-loss and exercise performance enhancement formulations, and replacing it with a multiplicity of other natural stimulants, or doubling down on the quantities of those already in place, failed to alleviate the problem. Almost as soon as Ephedra-free HDS made their entrance onto the marketplace, adverse event reports – eerily similar to those of their banned namesake – began appearing in the medical literature and within the files of the FDA’s MedWatch program (Gurley et al., 2015). Prospective clinical research on the safety of Ephedra-free HDS is limited, but published trial results appear to confirm the concerns of many health care providers regarding cardiovascular side effects and product contamination (Haller et al., 2005; Foster et al., 2013). As for the hepatotoxic potential of these HDS, we are relegated to evaluating and interpreting evidence available from published case reports and adverse event reports. In this regard, the product Hydroxycut serves as a prototypical example.

4.2.1.1. Hydroxycut.

The proprietary blend Hydroxycut, marketed for weight loss, exercise performance enhancement, and “fat burning,” was, along with hundreds of other Ephedra-containing HDS, withdrawn from the market in 2004 due to the risk of serious adverse health effects. In 2009, a new Ephedra-free version of Hydroxycut was withdrawn from the U.S. market due to 23 reported cases of liver injury associated with its use. Nevertheless, its reformulated herbal mixes are available on the market, resulting in new cases of Hydroxycut-induced liver injury (Kaswala et al., 2014; Araujo and Worman, 2015; Khetpal et al., 2020).

Hydroxycut-associated hepatotoxicity was diagnosed mostly within 3 months from the beginning of product use and symptoms typically presented as abdominal pain, dark urine, and jaundice. Most cases were characterized as hepatocellular injury with spiking serum levels of aminotransferases (ALT and AST), and little to no increases in alkaline phosphatase (ALP). However, it has to be emphasized that several cases of cholestatic or mixed acute hepatitis with prolonged jaundice were also reported with Hydroxycut.

Different Hydroxycut products with varying formulations were implicated in these cases and, therefore, it is impossible to identify a particular ingredient responsible for its toxicity. In all likelihood, the actions of several ingredients, working in concert, may be to blame. Several of these products contained CAF and YOH, two alkaloids that may potentiate the effects (including hepatotoxicity) of other phytoconstituents within the mixture (Ohta et al., 2007; Chen et al., 2015; Cimolai and Cimolai, 2011; Kearney et al., 2010). YOH was initially proposed as an aphrodisiac, but had little to no clinical effectiveness. It was later touted for its exercise performance enhancement properties, prompting its appearance on many weight-loss supplement labels. However, no clinical or experimental confirmation of these claims could be substantiated (Bucci, 2000). Furthermore, a number of products claiming to source YOH from Yohimbe bark extract likely contained synthetic yohimbine (Betz et al., 1995).

Still other potential contributors to hepatotoxicity associated with Hydroxycut ingestion are Garcinia cambogia and GTE. G. cambogia, apart from its presence in Hydroxycut, has been linked to several other cases of liver injury on its own (Andueza et al., 2021), and GTE’s risk has been outlined above. However, both of these extracts were often present in proportions smaller than other botanicals within the Hydroxycut formulation.

Importantly, despite the removal of initial Hydroxycut formulations from the market in 2009, versions with still further refined formulas are currently on the market. Not surprisingly, new cases of liver injury linked to Hydroxycut, while infrequent, are still being reported, and include cases of cholestatic liver injury and vanishing bile duct syndrome (VBDS) (Adike et al., 2017; Khetpal et al., 2020). Thus, it would appear that the case of Hydroxycut argues strongly for the additive or synergistic influence of multiple botanical extracts on HILI, especially when little to no prospective research is conducted into their safety.

4.2.2. OxyELITE-pro (OEP) OxyELITE-Pro New Formula (OEP-NF)

One of the most recent cases of an HDS-associated outbreak of liver injury involved the product OxyELITE-Pro (OEP). Its initial preparation, as well as a subsequent “Super Thermo” formulation, was removed from the market due to seven cases of acute liver injury reported in previously healthy military personnel (Foley et al., 2014). The toxicity was attributed to an indirect-acting sympathomimetic agent, 1,3-dimethylamylamine (DMAA), which the makers of OEP claimed was present in geraniums in spite of actually being developed by Eli Lilly in 1948 and marketed as a nasal decongestant. Subsequent “New Formulas” substituted DMAA with aegeline, a phytochemical present in the fruit and bark of Aegle marmelos (bael fruit). However, shortly after OEP New Formula (OEP-NF) appeared on the marketplace, a new spate of severe liver injuries was reported, first in Hawaii and then in the continental U. S. (Roytman et al., 2014; Johnston et al., 2016; Chatham-Stephens et al., 2017). In total, 55 cases of liver injury were associated with ingestion of OEP-NF (Klontz et al., 2015; Heidemann et al., 2016; de Boer and Sherker, 2017; Navarro et al., 2017). In October of 2013, the U.S. FDA banned the product and requested the manufacturer recall all products from the supply chain and distribution centers.

Clinical manifestations of OEP-associated hepatotoxicity usually appeared within two to twenty weeks from start of ingestion (Heidemann et al., 2016). The most common symptoms were nausea, fatigue, abdominal pain, and jaundice, often in conjunction with dark urine. Histologically, hepatotoxicity presented as acute hepatitis with necrosis, the degree of which was dependent upon injury severity. In the majority of cases, symptoms resolved and liver function tests returned to normal within two to three months. It is worth mentioning, however, that several patients required liver transplantation and a 10% mortality was noted among patients who developed jaundice (de Boer and Sherker, 2017; Navarro et al., 2017).

Initially, OEP-associated hepatotoxicity was attributed to aegeline. Indeed, some studies with aegeline produced significant toxicity both in vitro and in animal models (Arseculeratne et al., 1985; Mohammed et al., 2016; Manda et al., 2016). Furthermore, phytochemical analysis of OEP-NF paralleled by the criminal investigation revealed that aegeline was not only of synthetic origin but had been illegally imported into the U.S. (Miousse et al., 2017). Thus, it is plausible to hypothesize that aegeline-induced toxicity could be driven by or further exacerbated via potential contaminants from synthetic intermediates or non-naturally occurring enantiomers. At the same time, other potential contributors could not be dismissed. For instance, further phytochemical analysis of the OEP-NF product demonstrated that other phytochemicals claimed on the label were also of synthetic origin (Miousse et al., 2017). Furthermore, being a multi-ingredient product, OEP-NF contained CAF and YOH as well as other indirect-acting sympathomimetics (e.g., higenamine, coclaurine), thus rendering the product’s toxicological profile unpredictable.

Detailed toxicological assessment of OEP-NF using both inbred and outbred mouse strains demonstrated its toxic potential, as evidenced by increased liver/body weight ratios, increased ALT, AST and miR-122 plasma levels, the appearance of apoptotic liver foci and robust transcriptomic alterations in liver parenchyma (Miousse et al., 2017). Further investigations using the NZO/HlLtJ obese mouse model provided potential insights into the role cardiometabolic syndrome might play in higher susceptibility to OEP-NF-associated toxicity, as an over-whelming majority of patients with liver injury associated with OEP-NF ingestion were either overweight or obese. In these mice, besides hepatotoxicity, ingestion of OEP-NF has also led to systemic cardiotoxicity which was evident within 8 weeks of product administration (Skinner et al., 2018).

Like the banned Ephedra-containing supplements (e.g, Xenadrine RFA, Metabolife, Ripped Fuel, etc.) and many currently available Ephedra-free successors (e.g., Xenadrine EFX, Metabolift, Zantrex 3, etc.), OEP-NF is yet another multi-ingredient HDS in which the pharmacological and toxicological repercussions of haphazardly combined botanical extracts and their attendant phytochemical assemblages remain unknown. This is further complicated by the constantly shifting ingredient lists and the variety of products that bear the same name (for instance, OEP can be readily purchased online again; however, its ingredients list differs from the above-mentioned product). That is until unwitting consumers are admitted to emergency rooms with unexplained liver injuries.

5. Products on the market with suspected potential for hepatotoxicity

5.1. Ashwaghandha

Ashwagandha, a popular herb in Ayurvedic medicine, is an extract of the roots of Withania somnifera, an evergreen shrub cultivated in tropical and subtropical zones of the world (Mandlik Ingawale and Namdeo, 2021). Its purported benefits include adaptogenic, anxiolytic, anti-diabetic, anti-inflammatory, and even aphrodisiac effects (Agarwal et al., 1999; Mandlik Ingawale and Namdeo, 2021). Despite its long history of use in Ayurveda, the results of various controlled clinical trials have been equivocal. For instance, the most recent systematic review and meta-analysis of twelve randomized controlled trials reported that Ashwagandha significantly reduced anxiety and stress levels; however, certainty of evidence was low for both outcomes (Akhgarjand et al., 2022). Nevertheless, ashwagandha became a popular constituent of many HDS, either as a single ingredient or a constituent of multi-extract products.

Ashwagandha is generally considered safe; however, evidence of liver injury associated with its ingestion, although rare, is not without precedence. One of the first reported cases appeared in 2017 from Japan and was associated with combining ashwagandha with anxiolytics, with at least a doubling of the ashwagandha dose one month prior to symptom onset (Inagaki et al., 2017). The nature of the liver injury was determined as cholestatic, with high levels of plasma bilirubin, and a histological picture of canalicular cholestasis. All other potential causes were excluded.

A more recent report from Björnsson and colleagues presents a review of five cases of ashwagandha-induced liver injury from Iceland and the US Drug-Induced Liver Injury Network (DILIN) (Björnsson et al., 2020). Similar to the Japanese case above, all five patients developed jaundice associated with hyperbilirubinaemia, whereas two presented with cholestasis, and the other three exhibited mixed patterns of liver injury. Liver function tests gradually normalized over several months after discontinuation of the product and no cases of acute liver failure were reported by the authors. Phytochemical characterization of the available products demonstrated the presence of ashwagandha without evidence of known toxic compounds or heavy metals. The causative role of ashwagandha in the diagnosed liver injury was considered definite in one case, highly likely in two others, and probable and possible for the two remaining patients (Björnsson et al., 2020).

5.2. Extract of Coleus forskohlii/forskolin

Similar to ashwagandha, Coleus forskohlii extract (CFE) has been used extensively in Ayurvedic medicine for the treatment of cardiovascular and respiratory diseases (Ammon and Müller, 1985). It has received much attention recently as a single entity and as a constituent of multi-ingredient HDS weight loss products.

In an animal study designed to compare the hepatotoxic effects of CFE and forskolin, male ICR mice received diets with various concentrations of each. While CFE decreased visceral fat in a dose-dependent manner, dose-related increases in liver weights, plasma ALT, AST, and ALP were detected within 1 week after study initiation (Virgona et al., 2013). Interestingly, no evidence of liver toxicity was observed in mice given forskolin alone, suggesting that other phytoconstituents in CFE are responsible for this effect. In a subsequent study, CFE was shown to induce the development of fatty liver in mice, presumably via enhancement of de novo triglyceride synthesis (Umegaki et al., 2014).

A recent study observed CYP3A induction in both mouse intestine and liver upon CFE administration (Yokotani et al., 2020). This induction, however, was short lived and required substantially high doses, persuading the authors to conclude that the potential for CFE-drug interaction was very low. An earlier study by this group, reported that an attenuation of the anticoagulant activity of warfarin was associated with hepatic induction of CYP2C attributable to CFE exposure both in mice and human liver microsomes (Yokotani et al., 2020).

In a nationwide survey performed in Japan, 10.5% of all study participants who reported CFE use experienced adverse effects (Nishijima et al., 2019). Of those reporting, gastrointestinal syndrome was experienced by 92%, with 81.3% experiencing diarrhea. Data analysis suggested a safe intake of CFE to be less than 250 mg/day.

5.3. Kratom

Mitragyna speciosa is a tree native to Southeast Asia, known in the U. S. by its Thai name: kratom. In traditional medicine, botanical products extracted from its leaves (usually, in the form of tea) or chewable leaves themselves, were used to treat diarrhea as well as being used as an analgesic agent or to produce a stimulant effect. On the U.S. market, kratom has been touted for its potency to treat abstinence syndrome associated with opium withdrawal, as well as both stimulatory and sedative effects. These effects are associated with the major phytoconstituents of kratom, namely mitragynine and 7-hydroxymitragynine that act as μ-opioid receptor agonists and δ- and κ-opioid receptors antagonists (Kong et al., 2017; Olsen et al., 2017). Most users, however, reported the use of the kratom as a painkiller, antidepressant and anxiolytic (Garcia-Romeu et al., 2020).

Accumulating evidence indicates that chronic kratom users develop mild addiction as well as opioid abstinence syndrome and other side effects associated with stimulation of μ-opioid receptors – insomnia, anorexia and constipation, to name a few (Suwanlert, 1975; Rosenbaum et al., 2012). Analysis of the U. S. National Poison Data System revealed over 1,800 cases of reported kratom poisoning occurred between 2011 and 2017, with a number of cases attributable to kratom-associated liver injury (Post et al., 2019). The largest cluster of these cases – eleven – was reported by the Drug Induced Liver Injury Network (DILIN) between 2011 and 2019, eight of which were reported between 2017 and 2019 (Ahmad et al., 2021).

Kratom-induced liver injury is characterized by the high levels of bilirubin and associated jaundice. The levels of ALT and AST, however, are usually only moderately elevated, with only one case known reaching an increase of ALT levels over 1,000 U/L. Interestingly, half of the patients were characterized by mixed patterns of liver injury evident by both hepatocellular and cholestatic liver injury.

One of the interesting features of kratom-induced liver injury is the continuously rising level of bilirubin even for some time after discontinuation of the product (Fernandes et al., 2019). This may be explained by the prolonged half-life/slow rates of metabolism of kratom, as in one study it was reported that kratom metabolites could be detected in urine for at least two weeks after discontinuation of its ingestion (Kapp et al., 2011).

While the potential for kratom-drug interaction is not well understood due to kratom’s nature of being an extract of a plant rather than being presented as a single phytoconstituent, some clinical evidence suggests that it cannot be negated. For instance, in the Fernandes et al. (2019) report, a patient was taking 800 mg acetaminophen (APAP) twice daily at time of presenting clinical manifestation of liver injury. This dose of APAP itself is generally considered to be safe; however, a recent example of CBD/APAP-induced liver injury reported in an animal model with the doses of both ingredients being below levels of toxicity may provide evidence of such interaction between APAP and other phytochemicals (discussed in details in section 5.5) (Ewing et al., 2019b).

Due to its numerous side effects, sales of kratom are prohibited in many Asian countries and Australia (Schimmel and Dart, 2020). In the U.S., the initial attempt by the U.S. Drug Enforcement Administration in 2016 to list kratom as schedule I drug was postponed due to strong opposition from kratom lobbyists (Griffin and Webb, 2018). This delay resulted in uncontrolled sales of kratom in the U.S. and a spike of reported side effects associated with kratom use recorded in the last several years (Post et al., 2019; Ahmad et al., 2021). Further delays in regulating kratom may lead to serious consequences and lead to development of a public health emergency with many users experiencing abstinence syndrome, liver injury, and other negative health effects associated with its ingestion.

5.4. Turmeric

Turmeric, an herbal product derived from Curcuma longa, has a long history of use as a spice (i.e., in preparation of curry) and dye, with some records of use as a medicine, particularly in India and Southern Asia. Most of the medicinal effects of turmeric are attributed to curcumin, the major curcuminoid constituent of turmeric rhizome. Over the last decade, a significant body of research has credited a number of beneficial health effects to turmeric, including anti-inflammatory, antioxidant, antimicrobial, and even anti-cancer activity (Aggarwal et al., 2013a, 2013b; Gupta et al., 2013). It should be noted, however, that most of these findings were demonstrated using in vitro systems, with less research being conducted in vivo or via randomized controlled clinical trials. Furthermore, curcumin is characterized by poor oral bioavailability, with doses of at least 4 g per day needed to achieve detectable plasma levels (Lao et al., 2006; Kurien et al., 2015). Altogether, these findings raise concerns with regard to the clinical effectiveness of turmeric-containing DS.

Despite these shortcomings, turmeric has experienced significant growth in sales within the U.S. market, ranking it among the five top-selling HDS in 2019 (Smith et al., 2020). Turmeric is considered to be generally safe as a food (i.e., spice), with no reports of liver injury associated with its ingestion to our knowledge. At the same time, a recent analysis of published data from 20 clinical trials suggests a 5% overall incidence of abnormal liver function with prolonged use of turmeric-containing DS (Lukefahr et al., 2018). This was evidenced by elevated plasma levels of transaminases, LDH, alkaline phosphatase, and bilirubin. Furthermore, fulminant liver injury associated with consumption of turmeric-containing DS has been reported (Lukefahr et al., 2018; Abdallah et al., 2020; Luber et al., 2019; Lee et al., 2020; Lombardi et al., 2020). Hepatotoxicity in these cases was characterized by pain or discomfort in the upper right abdominal quadrant, nausea, fatigue, jaundice, and dark urine. Clinically, significant elevations of ALT and AST along with increased bilirubin were observed. Histologically, non-specific inflammatory acute hepatitis with preserved liver architecture and no fibrotic changes were noted. Two cases of classic autoimmune hepatitis (AIH) were also associated with ingestion of turmeric DS (Lukefahr et al., 2018; Lee et al., 2020). In each of the above cases, other potential causes of liver injury (e.g., viral hepatitis, drug-induced liver injury, etc.) were ruled out and discontinuation of turmeric products resulted in normalization of liver function.

It should be noted that the clinical reports involving turmeric lacked phytochemical characterization of the products consumed by patients. Therefore, it is difficult to determine whether or not turmeric was the cause, or if other factors, such as adulteration, presence of heavy metals, pesticides, solvent residues, and/or microbiological contamination (i.e., fungi) were the triggers of liver injury.

Recently, various attempts have been made by DS manufacturers to increase phytochemical bioavailability, especially that of curcumin (Gurley, 2011). These include phytosome complexation, nanoemulsion and nanoparticle formulations, liposomal encapsulation, and incorporation of natural inhibitors of enteric CYPs (i.e., piperine). While these approaches may significantly aid in elevating curcumin plasma concentrations, they may, in turn, increase the risk of liver injury. For instance, co-ingestion with piperine, an alkaloid derived from Piper nigrum L (black pepper), has been shown to increase curcumin’s bioavailability 20 to 200-fold (Shoba et al., 1998; Skiba et al., 2018). Indeed, in one of the most recently reported cases of liver injury associated with a turmeric-containing DS, the patient had a history of ingesting a product containing curcumin/black pepper (Luber et al., 2019). Furthermore, a recently reported review from Italy of seven cases of turmeric-associated acute liver injury demonstrated that highly bioavailable curcumin products (i.e., containing Piper nigrum L dry extract) were involved in all seven cases (Lombardi et al., 2020). Therefore, close monitoring of the rapidly evolving turmeric market, along with careful documentation of each purported case of turmeric-induced liver injury and accurate phytochemical characterization of the suspected products are clearly warranted.

5.5. Cannabidiol (CBD)

Perhaps the most controversial product among those currently present on the market is CBD, one of the principal phytocannabinoinds in Cannabis sativa L. Mediated by aggressive industry lobby and marketing as well as significant attention in the media, CBD has been touted for numerous health and treatment claims, resulting in high interest among the general population (Walker et al., 2020). Despite these often egregious claims, the only clinically confirmed beneficial effects of CBD were reported for two types of rare pediatric syndromes – Dravet and Lennox-Gastaut – that are characterized by severe and often therapy-resistant seizures (Devinsky et al. 2016, 2017; Huestis et al., 2019). In 2018, when the U.S. FDA approved this purified form of CBD as a drug, Epidiolex^® (Greenwich Biosciences), it automatically closed the door for consideration of CBD as a potential DS – as, according to the U.S. FDA, the product marketed as a drug cannot be marketed as a DS afterwards.

This, however, did not preclude the appearance of CBD on the market in its various forms – from oil to ointments and vaping products. The lack of a legislative oversight resulted in the appearance of substandard and adulterated products that rapidly flooded the market. As a result, reports of side effects started emerging, with products spiked with synthetic cannabinoids (Cogan, 2019; Kleis et al., 2020; Mohr et al., 2022; Simon et al., 2022). This is of no wonder, as a recent study demonstrated that most products that were on the market were either substandard (that is containing more or less than is stated on the label) or spiked with either THC or synthetic cannabinoids (Bonn-Miller et al., 2017; Gurley et al., 2020).

Besides this, it is important to mention that CBD itself, despite showing quite high levels of tolerance (i.e., high doses are needed to cause toxicity), is far from being an innocuous molecule. For instance, the data obtained from clinical trials show that 5%–20% of study participants demonstrated elevated levels of liver enzymes (Devinsky et al. 2016, 2017; Huestis et al., 2019). The latter effect was usually observed after several months of therapy and suggests potential bioaccumulation of CBD and/or its metabolites. Indeed, a study using the mouse model demonstrated that quite high doses of the cannabidiol-rich cannabis extract (CRCE, CBD content 58.4%) were needed to cause hepatotoxicity; however, much lower doses of CRCE were needed to induce hepatotoxicity in a chronic ingestion scenario (Ewing et al., 2019a). The mechanisms of this toxicity are yet to be investigated, but accumulating evidence suggests that they can be multifactorial and involve many molecular pathways. One of the recent studies also demonstrated that chronic ingestion of CRCE triggered a number of pro-inflammatory responses in the mouse gut, leading to potential development of a leaky gut syndrome – one of the driving mechanisms of alcohol-induced liver injury when hepatocytes are under constant attack by bacterial species and their metabolites (Skinner et al., 2020).

Another concern is CBD’s potential for interaction with other xenobiotics. From clinical trials, it is known that CBD can interact with other anti-epileptic drugs (Ben-Menachem et al., 2020; VanLandingham et al., 2020; Patsalos et al., 2020; Gilmartin et al., 2021; Dial et al., 2022). Based on the available in silico and in vitro data, it was postulated that medications that are substrates for CYP2C19, CYP2C9, CYP1A2, and CYP3A4 may be of particular risk of altered disposition when used concomitantly with CBD (Zendulka et al., 2016; Qian et al., 2019). Animal research confirmed that CBD can serve as a substrate for many drug metabolizing enzymes, particularly for CYP1A2, CYP2E1, CYP2B10 (homolog of CYP2B6 in humans), and CYP3A4 (Ewing et al., 2019a; Kutanzi et al., 2020). Subsequent research has further demonstrated this interaction potential when the pre-treatment of aged female CD-1 mice with CRCE resulted in the development of sinusoidal obstruction-like liver injury after administration of a single dose of acetaminophen (APAP) (Ewing et al., 2019b). In the latter study, the dose of APAP delivered to mice itself caused only mild transitory elevations in ALT and AST, without increasing plasma bilirubin levels and lack of histomorphological findings and the dose of CBD in the administered product was below the toxicity level (mouse equivalent dose of 10 mg/kg). Yet, the combination of CRCE and APAP resulted in the development of fulminant liver injury and mortality rates of nearly 40% (Ewing et al., 2019b).

At the moment of preparation of this article, the regulatory status of CBD remains largely undetermined, allowing hundreds of unregulated (and thus of questionable quality) products to be available on the market. On the other hand, evidence of toxicological effects and high potential for CBD-drug interaction pose serious concerns as to the safety of products labeled “CBD”, “Hemp full spectrum extract”, and many others and warrants both appropriate regulation and understanding of pharmacological and toxicological properties of this popular cannabinoid.

5.6. Tinospora cordifolia (Giloy)

The COVID-19 outbreak stimulated sales of various HDS touted to boost the immune system in order to prevent or treat COVID and its symptoms. Soon enough, the first reports of immune booster-induced liver injury followed, among which a set of cases involving Tinospora cordifolia, more commonly known as Giloy, were of particular interest. First, six patients were reported between September and December 2020 in India (Nagral et al., 2021). They all presented with symptoms resembling autoimmune hepatitis, with five patients demonstrating resolution of liver function and symptoms and autoimmune serological markers after product withdrawal and steroid treatment; one patient succumbed after steroid tapering. Soon, another case study reporting two additional patients was published (Gupta et al., 2013; Gupta et al., 2022), just to be followed by a large, nationwide study covering thirteen health centers in India and reporting 43 additional cases of Tinospora cordifolia-induced liver injury (Kulkarni et al., 2022). The latter patients presented with acute hepatitis, acute worsening of chronic liver disease (a feature observed in 4 out of 6 patients in the study published by Nagral et al., 2021), or acute liver failure, with the median time from initial product consumption to onset of symptoms being 46 days. Further liver biopsy in the subset of patients confirmed autoimmune hepatitis (Kulkarni et al., 2022).

While it is hard to establish the causal relationship in each of the abovementioned cases, ingestion of Giloy products in all cases paralleled by the resolution of symptoms after product withdrawal, suggests Tinospora cordifolia to be a main suspect in this spate of liver injuries. While it is difficult to elucidate why the latter was observed only recently despite the long history of usage of Giloy in Ayurveda, several reasons were suggested, such as potential contamination, as well as individual and predisposing factors (e.g., dose, comorbidities, or herb-drug interaction) (Björnsson et al., 2022). Interestingly, the most recent report indeed indicated severe contamination with mercury, arsenic, and lead in several Giloy samples retrieved from the patients (Kulkarni et al., 2022). Whether or not similar reports will follow from other countries is clearly one of the questions to be answered.

6. Ensuring safety

While the safety of DS continues to pose considerable threats to public health and occasionally causes “serious bleeding to the industry” (in the form of loss of trust from the consumer and litigation cases), primary efforts from the industry are focused on meeting the cGMP requirements and filing NDI notifications. Indeed, there is little concern for the safety of single extract HDS that have been present on the market for decades. However, botanical mixtures and newly emerging products clearly represent two categories of concern.

The handful of examples discussed above represent only a subset of known cases of liver injuries related to consumption of DS. Furthermore, many cases go undiagnosed as patients often do not report usage of DS to clinicians or do not seek medical assistance due to the lack of clinical manifestation of symptoms (de Boer and Sherker, 2017). For instance, the case of the red bush tea-induced hepatotoxicity discussed above became evident only after the patient was admitted to the hospital for an unrelated emergency condition (Reddy et al., 2016). Preoperative examination revealed elevated levels of liver enzymes that precluded surgical procedure under general anesthesia and, potentially, saved the patient’s life. Many other cases, however, required expensive treatment plans and are associated with long-term recovery, or further complicated by the necessity of liver transplantation, and, occasionally result in death. Altogether, these facts argue strongly for the introduction of quality standards and pre-marketing safety assessments for multi-ingredient dietary supplements. Therefore, it is plausible to hypothesize that the 1994 DSHEA will undergo revisions aimed at strengthening requirements for ensuring the safety of marketed products. This, in turn, will require the development of new standards and approaches designed to perform rapid and reliable toxicological product evaluation. One of the options to overcome these challenges could be implementation of mandatory product listing with the U.S. FDA – registration of every new product before it can reach the market. Another potential avenue is an introduction of pre-clinical safety assessment of HDS, especially multi-ingredient HDS, where the combination of numerous constituents poses unknown risks. This assessment clearly should differ from the assessment of pharmaceuticals. A lengthy and expensive process required for the pre-market assessment of drugs cannot be pertinent for HDS as patents cannot be claimed for the latter.

It must be emphasized that several such strategies have been proposed recently, in part due to the efforts of the DS industry and their collaboration with government and academic units. For instance, the National Toxicology Program (NTP) demonstrated usefulness of using the Tox21 platform for high-throughput screening of DS, including HDS (Hubbard et al., 2019). Further, the U.S. FDA, the National Institute of Environmental Health Sciences (NIEHS), and the Health and Environmental Sciences Institute (HESI) established the Botanical Safety Consortium (BSC) – “a public-private partnership aimed at enhancing the toolkit for conducting the safety evaluation of botanicals” (Mitchell et al., 2022). This partnership, built on expertise and contribution from scientists that represent government, academia, industry, consumer health groups, and non-profit organizations, works collectively toward adaptation and integration of novel approaches and tools into routine botanical safety assessment. The overarching goal of the BSC is “integration of these tools into a framework that can facilitate the evaluation of botanicals” (Mitchell et al., 2022).

Below, we will discuss some efforts and specific models that hold tremendous potential to advance and modernize the safety assessment of HDS which would be beneficial for both the trusting consumers and industry.

6.1. In silico models

Advances in computational toxicology allowed for the development of an in silico decision tree methodology for botanical safety assessment that was initially proposed by Little and colleagues (Little et al., 2017). This approach considers the dietary intake level of the botanical (or botanicals, in the case of mixtures) and sufficient margins of exposure that can be obtained from existing safety information. It relies on comprehensive characterization of the botanical with sophisticated chemical analysis, and individual constituents are used to compare to levels found in food or to establish a reasonable margin of exposure. By constructing a botanical decision tree, the authors then proposed a tiered approach that will potentially aid in resolving the safety endpoint gaps as well as instructing on study design for the cases that could not be resolved without utilization of experimental systems (i.e., in vitro or in vivo) (Roe et al., 2018). Indeed, this approach showed promising results in the assessment of red clover herb, Pelargonium sidoides root, and Scutellaria baicalensis Georgi root (Little et al., 2017).

6.2. Cell culture models

There are at least three major considerations when selecting a cell culture model for sensitive hepatotoxicity screening. First is expression of drug-metabolizing enzymes (DMEs). Many drugs and HDS ingredients are subject to Phase I and/or II metabolism in the liver. In fact, the hepatotoxicity of many xenobiotics is due to conversion to reactive metabolites by DMEs. It is critical to select a model that expresses the full complement of DMEs at levels similar to primary human hepatocytes (PHH) or whole liver tissue. The second consideration is the ability of the model to reproduce drug-drug, HDS-drug, and HDS-HDS interactions. Many xenobiotics alter the metabolism of others by inducing expression and/or inhibiting activity of DMEs (Gurley et al., 2012; Brewer and Chen, 2017). Such pharmacokinetic interactions can increase the toxicity of a drug or HDS ingredient by increasing exposure to a toxic parent compound or a reactive metabolite. Thus, a sensitive screening tool should respond to well-characterized DME inducers (e.g., rifampin, phenobarbital, St. John’s Wort) and inhibitors (e.g., azole antifungals, piperonyl butoxide) in the same way as PHH and whole liver. The final consideration is susceptibility to common mechanisms of hepatotoxicity. For example, many drugs are known to cause liver injury in animals and humans through mitochondrial damage (Jaeschke et al., 2012; Pessayre et al., 2012). Therefore, any model chosen for screening purposes should be known to be susceptible to mitochondrial dysfunction, which can be determined through mechanistic studies.

The HepaRG cell line is one of the best available options for screening (McGill et al., 2011; Andersson et al., 2012; Jaeschke et al., 2021). A great deal of research is devoted to the development of cell culture models for studies of drug hepatotoxicity. PHH are the gold standard, but have major disadvantages: limited availability, variation in quality due to donor health, and high cost. Hepatoma cell lines are an alternative to PHH. Unfortunately, a major obstacle to the use of hepatoma cells is poor expression of DMEs. Numerous studies have demonstrated that expression of most DMEs is far lower in commonly used hepatoma cells (e.g. HepG2, Huh7) than in primary human hepatocytes or liver tissue (Rodríguez-Antona et al., 2001; Hart et al., 2010; McGill et al., 2011; Kvist et al., 2018). The HepaRG cell line overcomes many of the challenges presented by both PHH and common hepatoma cells. Unlike PHH, they are renewable and capable of being passaged many times before a new culture must be established. Importantly, the profile of DME expression in HepaRG cells resembles that of PHH and intact liver tissue (Hart et al., 2010; Kvist et al., 2018). HepaRG cells also respond to inducers and inhibitors of DMEs in much the same way as PHH (Kanebratt and Andersson, 2008; Turpeinen et al., 2009; Anthérieu et al., 2010). Finally, there is strong evidence that susceptibility to toxicants and the basic mechanisms of drug toxicity in HepaRG cells are similar to PHH and animal models (McGill et al., 2011; Xie et al., 2015; Woolbright et al., 2016; Kamalian et al., 2018). Finally, sensitivity can be further increased using 3D HepaRG cultures (Mueller et al., 2014).

There are several other hepatocyte culture methods that show promise. Hepatocyte-like cells have been prepared from induced pluripotent stem cells (iPSC). However, overall DME expression in cultures of iPSC-derived hepatocyte-like cells is generally low when compared to mature human hepatocytes (Song et al., 2009; Sancho-Bru et al., 2011; Takayama et al., 2018). Strategies for enrichment of hepatocyte-like cells with high DME expression have been developed (Mallanna et al., 2016; Takayama et al., 2018), but are typically laborious and time-consuming and require specialized equipment. Interest in 3D culture systems using primary hepatocytes or even hepatoma cell lines, such as spheroid cultures, 3D bioreactors, and organ-on-a-chip technology, has grown in recent years (Godoy et al., 2013; Andersson, 2017). An advantage of 3D culture systems is that the architecture of the liver can be modeled (Godoy et al., 2013). However, the development of these systems is still in early stages, and a clear path forward has not yet emerged, which may be due in part to problems in the way different cell culture models have been compared (McGill, 2018). Future work should focus on better characterizing these systems and developing common methods to compare different models.

Finally, careful consideration should be given to the endpoints that are measured in these models. Unfortunately, many different endpoints are used in toxicology studies performed in vitro, which makes it difficult to compare the utility of different models (McGill, 2018). End points like oxidative stress and mitochondrial respiration offer high sensitivity, but poor specificity for toxicity because not all hepatocytes that experience oxidative stress or mitochondrial dysfunction will actually die. On the other hand, cell death markers such as enzyme release offer high specificity, but may not have sufficient sensitivity to identify all xenobiotics with potential for hepatotoxicity during preclinical testing. As stated above, a screening test should maximize sensitivity. Therefore, screening studies should include indicators of oxidative stress, mitochondrial respiration, and biomarkers known to have high sensitivity for hepatocyte damage such as microRNA-122. miR-122 has been shown to increase in plasma from humans and in cell culture medium of HepaRG cells before the standard liver injury marker ALT, and even in some cases when ALT never increases (Dear et al., 2014; McCrae et al., 2016; Gill et al., 2017; Ten Berg et al., 2018).

6.3. Animal models

6.3.1. Importance of animal models

Despite strong trends towards the development and utilization of alternative models and systems, it is still challenging to fully mimic in vivo architecture and physiology in cell culture. Indeed, regulatory agencies like the U.S. FDA still require animal testing in at least two species for premarket drug approval. As long as animals continue to be viewed as the gold standard model, it would behoove manufacturers and researchers in the herbal and dietary supplement area to include them in safety assessments. Clearly, utilization of robust full-scale animal models for every HDS product would not be cost-efficient or even necessary at this point. However, moving forward, confirming the results of in silico and in vitro studies for the most concerning botanicals and, especially, their combination (mixtures) will be an important step in precluding potentially dangerous products entering the market and applying conventional regulatory standards for safety assessments. In addition, animal models such as the Uetrecht-Pohl model (discussed below) are vital for the exploration of hepatotoxicity mechanisms, since idiosyncratic hepatotoxicity usually involves an adaptive immune response that is currently difficult to reproduce in vitro..

6.3.2. Towards utilization of appropriate animal models: mimicking heterogeneity of human populations

In most cases, DILI and HILI are idiosyncratic by nature, and utilization of conventional inbred mouse strains for pre-clinical safety assessment is not feasible. In these regards, outbred mouse (e.g., CD-1 or Swiss) or rat (i.e., Sprague-Dawley) strains and other recently developed rodent models, such as the Collaborative Cross (CC) and Diversity Outbred (DO) mice, provide potentially better opportunities for mimicking the genetic diversity of the human population. The utilization of such models, however, is often associated with a high inter-individual variability in the levels of basic end-points, thus presenting a considerable challenge for data analysis and interpretation. On the other hand, animal models consider usage of control (naïve, or untreated) animals, as well as the attainment of multiple serum/plasma and other biological specimen before, during, and after treatment. This allows for comparison of pre- and post-treatment levels of various end-points as well as an analysis of their dynamics. Furthermore, simultaneous analyses of a number of end-points, such as liver pathology coupled with plasma levels of liver enzymes, bilirubin, and miR-122 increase chances for identification of genuine responders after treatment with a particular xenobiotic. For instance, using this integrative approach, we have demonstrated inter-individual variability in susceptibility to OEP-NF in outbred CD-1 mice (Miousse et al., 2017). In this study, we selected mice with endpoint values greater than 2 SEM from the control group mean. After this, we identified that a number of mice exhibited a spectrum of findings that included increased liver/body weight ratio, elevated serum levels of ALT, AST, and miR-122, and appearance of hepatocellular mitotic foci. This approach allowed us to render these animals as genuine responders to a toxic HDS that was removed from the market after causing serious liver injury (Miousse et al., 2017).

6.3.3. Towards utilization of appropriate animal models: modelling particular pathological states

While the law does not allow HDS to be marketed for treatment of particular diseases or conditions, general claims are allowed and are being used extensively by the industry. Among the most popular claims for HDS are the claims “for general well-being”, “performance enhancement”, and “for weight loss.” The last of these represents a particularly interesting case as the targeted users for these products are overweight and obese subjects. Indeed, as rising obesity levels have been observed in recent decades, with over 60% of Americans either overweight or obese, DS for weight loss and management have gained popularity (Ogden et al., 2012; Wharton et al., 2019). While efficacy of such DS and their long-term effects are often questionable, there is an alarming aspect that requires attention. Obesity is often linked to metabolic syndrome and associated diseases, such as non-alcoholic fatty liver disease (NAFLD), diabetes, and cardiovascular disease, all of which can exacerbate toxicity exerted by constituents of multi-ingredient DS (Kopelman, 2000). Consequently, metabolism of botanicals can be significantly affected by these underlying conditions that, in turn, may predispose these subjects to higher sensitivity to xenobiotics. For instance, most of the subjects who experienced severe liver injury after ingestion of OEP-NF used either as a performance enhancer or weight-loss product were either overweight or obese (Klontz et al., 2015; Heidemann et al., 2016; Chatham-Stephens et al., 2017). Other studies indicate a number of severe side effects such as liver and kidney failure in obese subjects (Wharton et al., 2019).

In these regards, utilization of appropriate animal models that will take into consideration specific conditions of target consumer populations is essential. For instance, in regards to obesity, a number of mouse models have been developed; however, most of these models either require feeding mice with a high-fat diet (HFD) or are the result of gene editing technologies (i.e., transgenic mice). While these models are invaluable for mechanistic studies, they cannot be effectively utilized for safety assessment since both the diet and genetic modification may influence the metabolism and toxicity of the substance in question. For instance, feeding mice with HFD in order to achieve obesity or a particular liver pathological state (i.e., NAFLD) may also significantly increase the bioaccessibility of an investigated botanical’s phytoconstituents and, therefore, influence the potential toxicity or efficacy.

In these regards, of particular interest is an NZO/HlLtJ mouse model that proved to be useful in the case of OEP-NF (Skinner et al., 2018). This non-transgenic mouse strain is characterized by marked obesity involving both visceral and subcutaneous fat depots and is independent of dietary fat intake (Fig. 1). Feeding mice with OEP-NF resulted in markedly lower tolerance to the product and showed potential for the development of cardiotoxicity (Skinner et al., 2018). Recently, this strain was proposed as a model for the studies on the role of metabolic syndrome in acute radiation toxicity (Ewing et al., 2020).

Fig. 1.

Selection of an appropriate model is the pillar of successful research. A 4-months old NZO/HltJ female mouse, weighing in at 93 g, the embodiment of a diet-independent obesity and fatty liver disease. This is an excellent model to investigate the efficacy and safety of products targeted for weight management, performance enhancement, and treatment of fatty liver disease.

6.3.4. Towards utilization of appropriate animal models: the Uetrecht-Pohl model

A recent hypothesis is that severe idiosyncratic hepatotoxicity is the result of loss of immunotolerance. Many xenobiotics that cause severe DILI in a small proportion of those exposed also cause minor, transient transaminase elevations in a much greater proportion (a phenomenon known as Temple’s Corollary) (Watkins, 2005; US Food and Drug Administration, 2009; Vazquez and McGill, 2021). Because idiosyncratic hepatotoxicity is thought to be largely immune mediated, it has been suggested that the injury resolves in the latter cases due to development of immunotolerance, while loss or prevention of immunotolerance results in severe injury. Consistent with that hypothesis, it has been reported that blocking immunotolerance in mice leads to liver injury that appears strikingly similar to DILI in humans in response to well-known idiosyncratic hepatotoxicants (Metushi et al., 2015; Chakbraborty et al., 2015). This innovative approach, dubbed the Uetrecht-Pohl model (McGill and Jaeschke, 2019) was recently used to great effect to explore the role of the immune system in the idiosyncratic hepatotoxicity of green tea extract (Cho et al., 2021) and could be applied in similar studies of HILI in the future.

6.4. Microphysiological systems

Despite all the ample opportunities animal models provide in hepatotoxicity testing, a number of associated limitations often impairs translation of experimental data to humans. Indeed, translation of a mouse dose to a human one and back presents significant challenges considering rodents’ much faster metabolism and smaller body surface area, as well as shorter life spans compared to humans; taken together, these factors bring sufficient uncertainty into validation of animal findings. Therefore, the development of novel approaches and models that can effectively address these deficiencies is critical. These models should utilize human cells, preferably in the context of physiological systems that emulate particular organs and systems, rather than isolated cell cultures. Recent progress in the development of microphysiological systems, more commonly known as “organs-on-chip,” offer new opportunities in safety assessment of DS. Instead of a single hepatocyte culture, these alternative models are built in a three-dimensional environment on microfluidic channels in which the primary hepatocyte is co-cultured with other cell types that construct liver architecture (e.g., sinusoidal endothelial cells, stellate, and Kupffer cells) (Fig. 2). (Shi et al., 2022).

Fig. 2.

Hepatocytes and sinusoidal endothelium in a two-cell liver-on-chip model. Primary human hepatocytes were seeded in one channel of an organ-on-chip system and human liver sinusoidal endothelial cells were seeded in the other. (A) Hepatocytes. (B) Sinusoidal endothelial cells. Dashes outline one representative cell in each channel. 400× magnification.

A number of recent studies have successfully demonstrated the vast potential these systems hold in predicting DILI (Jang et al., 2019; Messelmani et al., 2022). Furthermore, the US Center for Food Safety and Applied Nutrition has recently evaluated the usefulness of the liver-on-chip system for assessment of chemical toxicity (Eckstrum et al., 2022), clearly indicating the direction government agencies are taking in risk assessment. In the long run, it is expected that microphysiological systems will be able to replace animal testing, offering more physiological relevance to human models. At the moment, however, this area is still in its infancy and will require further development in construction of simultaneously functional multi-organ systems; such systems must also become more cost-efficient in order to be used more widely.

6.5. Selection of appropriate end-points

Diagnosis of DILI and HILI, as well as toxicological profiling of HDS for potential liver injury, remain significant medical and scientific challenges. While a number of traditional end-points, such as serum ALT, AST, and total bilirubin (TBil) exist, they 1) can be affected by a palette of non-hepatic factors, 2) often do not characterize the degree of liver injury, and 3) provide limited mechanistic insights. For instance, serum aminotransferase elevations can be observed after strenuous exercise (Pettersson et al., 2008). Despite enthusiasm after initial experimental studies and clinical investigations, a novel plasma biomarker of liver injury, miR-122, showed high levels of inter-individual variability (Church et al., 2019). Furthermore, recent studies indicate that plasma levels of miR-122 exhibit substantial variations within the same individual and can be released by normal hepatocytes (Lin et al., 2017; Momen-Heravi et al., 2015; Rivkin et al., 2016; Church et al., 2019; Vogt et al., 2019). Furthermore, it must be considered that different end-points will need to be assessed with different systems in use. For instance, while HepaRG cells show promise for botanical safety assessment, they contain very low levels of ALT activity, thus precluding utilization of the latter biomarker for hepatoxicity screening (McGill et al., 2011).

Custom-built or pre-developed commercially available hepatotoxicity gene expression arrays may provide useful information regarding the toxicity potential of a botanical, as well as magnitude and mechanistical insights into its toxicity. Successfully utilized in testing for toxicity of many chemicals and drugs, this approach also proved useful in a number of toxicological assessments of several phytochemicals performed by our group and others. For instance, using this approach we could identify a spectrum of molecular responses as a result of administration of hepatotoxic doses of cannabidiol (CBD) (Ewing et al., 2019a), pro-oxidative effects and altered lipid metabolism in the case of OEP-NF toxicity (Miousse et al., 2017; Skinner et al., 2018), and lack of toxicological responses after administration of dGTE to non-fasted mice (Gurley et al., 2019). In the latter case, only one gene out of 84 investigated targets was significantly deregulated (Mcm10), further highlighting that decaffeinated GTE delivered in non-fasting condition lacks hepatotoxicity and that previously reported hepatotoxicity was mediated rather by EGCG, but not the GTE per se. Finally, an integrative approach that takes into consideration a set of end-points characteristic of liver injury may aid in identifying an herb (or mixture’s) potential to target certain populations.

7. Conclusions

The rapid progress and evolution of both industry and consumer markets alongside increasing cases of adverse events associated with ingestion of untested and adulterated products speaks to the necessity of revisiting the existing principles of regulation of DS. Development of a DS-specific path for their pre-clinical assessment and establishment by the regulator of a list of products that enter the market promise to improve consumer safety and drive bad actors from the market. Introduction of novel approaches for hepatotoxicity prediction (for instance, microphysiological systems) as well as modernization of current systems together with the development and validation of novel and robust biomarkers of liver injury will further improve the safety of the products on the market. The recent litigation involving USP Labs, the manufacturer of the infamous OEP-NF, resulted in nearly $60 million in fines and imprisonment of the guilty parties and is a logical and unavoidable consequence of the insufficiently regulated market. Unfortunately, several trusting consumers paid with their lives for this sad and eye-opening example of existing challenges in the world of botanicals and their regulation.

Abbreviations

ALT

alanine aminotransferase

AST

aspartate aminotransferase

HDS

herbal dietary supplement

CAF

caffeine

CBD

cannabidiol

CC

Collaborative Cros/s

dGTE

decaffeinated green tea extract

DILI

drug-induced liver injury

DILIN

Drug-Induced Liver Injury Network

DMAA

1,3-dimethylamylamine

DMEs

drug-metabolizing enzymes

DO

Diversity Outbred

DS

dietary supplement

DSHEA

Dietary Supplement and Health and Education Act

HILI

herbal-induced liver injury

iPSC

induced pluripotent stem cells

NAFLD

non-alcoholic fatty liver disease

NASH

nan-alcoholic steatohepatitis

NHANES

National Health and Nutrition Examination Survey

PHH

primary human hepatocytes

OEP-NF

OxyELITE Pro New Formula

YOH

Yohimbe extract

Data availability

No data was used for the research described in the article.