Herbal medicine, as a complementary and alternative medicine, is widely used in Asia as well as the west. Herbal medicine is generally considered to be natural, green, and harmless. In reality, herbal medicine may also have side effects, such as hepatotoxicity, nephrotoxicity, cardiotoxicity, neurotoxicity, and carcinogenicity. Hepatotoxicity is the most common side effect of great concern. The hepatotoxicity of drugs can be intrinsic or idiosyncratic (European Association for the Study of the Liver, 2019). The former refers to a liver injury that is closely related to drug dosage and timing, can be predicted, and has insignificant individual differences; the latter refers to a liver injury that is not correlated with the drug dosage and timing, is unpredictable, seen in a few people, and has a significant individual difference.

Herbal medicine is a common cause of drug-induced liver injury (DILI). A recent study carried out in Asian countries focusing on the causes of DILI reported that herbal medicine accounts for about 25% of all DILI (Byeon et al., 2019; Shen et al., 2019). However, very few studies focus on herb-induced liver injury (HILI), and little is known about species, components, mechanisms, and predisposing factors of herbs that cause liver injury. Common herbs that can cause HILI include Reynoutria multiflora (Thunb.) Moldenke (Polygonum multiflorum Thunb., PM), Cullen corylifolium (L.) Medik., Corydalis yanhusuo (Y.H.Chou and Chun C. Hsu) W.T.Wang ex Z.Y.Su and C.Y.Wu, etc., of which the most common is PM, accounting for 32.3% of all cases (Byeon et al., 2019). PM-related HILI (PM-HILI) is seen only in a few of the patients consuming this herb.

PM, the dried tuberous root of Reynoutria multiflora (Thunb.) Moldenke, when used as a traditional tonic and dietary supplement is believed to possess hypolipidemic, anti-atherogenic, anti-aging, immunity-enhancing, neuroregulatory, hepatoprotective, and anti-inflammatory properties. These potentially beneficial effects contradict their role in liver injury. However, the potential beneficial effects and rare liver injury are common features of many herbs that cause liver injury. An updated review can help in better understanding the predisposing factors, hepatotoxic components, and molecular mechanisms of PM-HILI, and aid in its diagnosis, treatment, and prevention. It can also aid in a better understanding of HILI and help in further research on the same.

Although Polygonum multiflorum Thunb. is not an international standard Latin name, but it is widely used in scientific literature. In order to make it easier for readers to understand and relate to other literature, the abbreviation PM is used in this review.

Reports of PM-HILI from China and other countries (But et al., 1996; Park et al., 2001; Mazzanti et al., 2004; Panis et al., 2005; Jung et al., 2011; Lian et al., 2013; Zhang Y. et al., 2013; Dong H. et al., 2014; Zhu et al., 2015; Luo et al., 2016; Li CY. et al., 2017; Han Z. X. et al., 2019; Fu et al., 2020) have increased in recent years (Table 1). Compared to other DILI, PM-HILI does not present specific clinical and pathological features. Hepatocellular injury is very common and cholestasis and mixed liver injury are rare. Common manifestations are jaundice and markedly increased levels of biochemical parameters such as total bilirubin, alanine aminotransferase, aspartate aminotransferase, etc. (Park et al., 2001; Mazzanti et al., 2004; Panis et al., 2005; Jung et al., 2011; Dong H. et al., 2014; Zhu et al., 2015; Li H. et al., 2017; Wang Y. et al., 2019). Most patients with PM-HILI have a good prognosis, although some patients may present with acute liver failure, chronic liver injury, or even death. PM-HILI with underlying chronic liver disease, particularly alcoholic liver disease, has a worse prognosis and is more likely to develop into chronic DILI (Jing et al., 2019). Studies suggest that PM-HILI is an idiosyncratic drug-induced liver injury (IDILI). Clinical epidemiological analysis has found that PM-HILI only occurs in a very small number of individuals taking PM, which is a low-probability event (Zhang et al., 2020b). PM is usually prescribed at a wide range of doses over large periods in patients, which suggests the lack of a clear dose-dependent relationship (Zhu et al., 2015; Tu et al., 2019a; Zhang et al., 2020b). In clinical practice, PM has been prescribed at a dosage within the recommended range of Chinese Pharmacopeia (3–6 g/day for raw PM, 6–12 g/day for processed PM) in patients with PM-HILI (Lei et al., 2015; Li CY. et al., 2017). PM-HILI is most commonly seen in patients with immune-related diseases such as vitiligo, psoriasis, rheumatoid arthritis, seborrheic alopecia, which are often accompanied by immune stress or mild elevation of inflammatory cytokines (Tu et al., 2019a), suggesting that immune stress may be an important predisposing factor for PM-HILI. Experimental animal studies have shown that only a long-term and large-dose administration of PM can exert hepatotoxic effects in healthy animals, while the corresponding dose is clinically difficult to achieve (Li et al., 2013; Zhu et al., 2014; Wang T. et al., 2015; Yang M. et al., 2016). However, PM can cause liver injury when administrated at a routine clinical dose within a short period in susceptible animal models (Li et al., 2015; Xie et al., 2016). Nevertheless, the phenomenon of high-dose and long-term administration of PM is also observed in clinical practice, and it cannot be ruled out that PM also exerts direct hepatotoxic effects when administered at very high doses.

The identification of hepatotoxic components is important to reveal the hepatotoxic effect of herbal medicine. The chemical composition of PM is complex and includes stilbene glycosides, anthraquinones, tannins, flavonoids, phospholipids, micronutrients, etc. The current focus of literature reports has been on the hepatotoxic effects of stilbene glycosides and anthraquinones (Lin et al., 2015; Li et al., 2016; Zhang et al., 2017a; Lin et al., 2017; Zhang et al., 2018; Han L. et al., 2019; Lin Y. et al., 2019; He et al., 2019; Li H.-Y. et al., 2020; Zhang et al., 2020a; Li RL. et al., 2020; Liu et al., 2020).

Li et al. (Li C. et al., 2017) found that ethyl acetate (EA) extract is responsible for the idiosyncratic hepatotoxicity of PM using the "knock-out" and "knock-in" strategies (Figure 1). Using mass spectrometry, it has been identified that the EA extract of PM consists of trans-2,3,5,4'-tetrahydroxystilbene-2-O-β-glucoside (trans-SG), cis-2,3,5,4'-tetrahydroxystilbene-2-O-β-glucoside (cis-SG), and emodin-8-O-glucoside (EG). Li et al. (Li C. et al., 2017) used monomers and confirmed that cis-SG, and not trans-SG, is the important component responsible for the idiosyncratic hepatotoxicity in PM. Cis-SG shows a definite dose-toxicity relationship in the MIS model and the dose is nearly equivalent to the clinical dosage by weight. The content of cis-SG in remaining samples of PM from patients with PM-HILI is significantly higher than that of general samples of PM from different provinces further proves that cis-SG is a hepatotoxicity-related component of PM-HILI (Figure 2) (Zhang et al., 2017a).

The naturally occurring component in PM is largely trans-SG, and the content of cis-SG in PM is low. Trans-SG is the component responsible for beneficial effects and has properties of anti-oxidation (Tao et al., 2011; Zhang et al., 2012), anti-inflammation (Zhang F. et al., 2013; Huang et al., 2013), anti-senescence (Han et al., 2012; Büchter et al., 2015), and enhancing memory (Wang et al., 2011; Chen et al., 2016). Zhang et al. (Zhang et al., 2017a) studied the relationship between levels of cis-SG and idiosyncratic liver injury of PM and found that the higher the cis-SG content, the greater the hepatotoxicity. They also showed that PM does not cause liver injury on the MIS model when the cis-SG content is less than 0.10%, suggesting an upper safety limit for quality control of cis-SG in PM. Therefore, cis-SG is the component responsible for idiosyncratic hepatotoxicity in PM. However, ultraviolet irradiation can induce the transformation of trans-SG to cis-SG (Dong L. H. et al., 2014), resulting in the transformation of effective components into toxic components (Figure 3) and increasing the risk of liver injury. Therefore, exposure to light should be avoided during the preparation and storage of PM.

Zhang et al. (Zhang et al., 2020a) studied EG in the EA extract of PM and found that EG causes liver injury in the MIS rat model in a dose-dependent manner and the dose is nearly equivalent to the clinical dose by weight. It is indicated that EG is an important component in PM that is responsible for idiosyncratic hepatotoxicity. The upper safety limit of quality control of EG in PM is 0.17%, determined using the method described earlier by Zhang et al. (China Association of Chinese Medicine, 2020).

Equimolar amounts of emodin do not cause liver injury, and the toxicity of combined anthraquinone is significantly higher than free anthraquinone in idiosyncratic hepatotoxicity, in contrast with the results of direct hepatotoxicity caused by the two compounds at very large doses (Yang M. et al., 2016). This helps in better understanding the fact that the toxicity of PM decreases after processing.

The finding of Ma et al. (Ma et al., 2015) and Gao et al. (Gao et al., 2017) showed that both the toxicity as well as the content of EG is reduced after processing (Figure 4), which indicates the hepatotoxicity of PM is highly correlated with EG. Reducing the idiosyncratic hepatotoxic components (or the susceptible components) of PM by processing is one of the important approaches to reduce toxicity.

Anthraquinones such as emodin (Lin et al., 2015; Zhang et al., 2018; Li H.-Y. et al., 2020), rhein (Li H.-Y. et al., 2020), and other components such as torachrysone-O-hexose (Han L. et al., 2019), dianthrones (Li H.-Y. et al., 2020), polygonnodes C1-C4 (Yang JB. et al., 2016), catechin (Li et al., 2016), etc. with intrinsic hepatotoxicity have also been investigated. Although the above components cause liver injury when administrated in large quantities, the quantity of the above components is very low in the clinical dosage of PM. Therefore, PM-HILI cannot be independently caused by the above components. However, they may play a secondary or synergistic role in the occurrence of PM-HILI. Recognition of this may contribute to a more complete understanding of the complex multi-component mechanisms of PM-HILI.

Current studies have shown that PM-HILI is considered to be IDILI and only occurs in a small number of susceptible individuals. However, in addition to IDILI, direct hepatotoxicity may be seen in specific PM-HILI patients who received a large dosage of PM. PM-HILI may be a result of a synergistic effect of multiple components and multiple mechanisms. Studies have shown that the molecular mechanism of PM-HILI is related to the inflammatory reaction, alterations of drug-metabolizing enzymes, apoptosis, and disorders of bile acid metabolism. A summary of the related mechanisms of PM-HILI around the two main susceptible components (cis-SG and EG) is presented below.

Several hypotheses about the mechanism of IDILI have been established (Roth and Ganey, 2011; Bai et al., 2017), including inflammatory stress hypothesis, genetic polymorphism hypothesis, hapten hypothesis, mitochondrial dysfunction hypothesis, etc. Inflammatory stress hypothesis refers to that a state of inflammation stress of the body may increase the susceptibility of the liver to some drugs, resulting in IDILI. Several drugs of idiosyncratic nature like diclofenac (Deng et al., 2006), ranitidine (Luyendyk et al., 2003), chlorpromazine (Buchweitz et al., 2002), trovafloxacin (Shaw et al., 2007), etc. have been proven to induce IDILI in LPS models. HLAs- and metabolic enzyme-related gene polymorphism has been closely associated with IDILI. Studies have shown that HLA-B*57:01 (Monshi et al., 2013), HLA-A*33:03 (Hirata et al., 2008), HLA-DQA1*02:01 (Spraggs et al., 2011), etc. are related to IDILI caused by flucloxacillin, ticlopidine, lapatinib, respectively. The gene polymorphism of UGT2B7, CYP2C8 (Daly et al., 2007), and CYP2E1 (Vuilleumier et al., 2006) has been reported to be related to IDILI caused by isoniazid. Hapten hypothesis refers to that some reactive drug metabolites bind to cellular proteins and act as haptens, which may induce an adaptive immune response (Uetrecht, 2007) and damage tissue by activating cells to release ROS, proteases, cytokines, etc. Studies have found that IDILI induced by multiple drugs including halothane (Bird and Williams, 1989), sulfamethoxazole (Callan et al., 2009), piperacillin (El-Ghaiesh et al., 2012), etc. is related to the hapten hypothesis. Multiple drugs (troglitazone, diclofenac, nimesulide, etc.) associated with IDILI exhibit a clear mitochondrial hazard (Boelsterli and Lim, 2007), which can lead to cell death. These hypotheses are derived from understanding IDILI from different perspectives or stages, and may not be mutually exclusive. IDILI is likely the result of multiple factors. Most of these hypotheses involve the immune system of the body, which plays a key role in the occurrence and development of IDILI.

Based on the inflammatory stress hypothesis, Rao et al. (Rao et al., 2020) considered PM-HILI to be an immune-mediated idiosyncratic liver injury. Further research has found that PM is more likely to cause liver injury in patients with immune activation. Also, individual components in PM have a potential role in promoting immune activation and causing liver injury. Wang et al. (Wang JB. et al., 2016; Wang J. B. et al., 2016) proposed the immune stress-mediated tri-element injury hypothesis of idiosyncratic toxicity of PM (Figure 5). It states that during an abnormally activated state (MIS) of the body’s immune system, the immune-promoting components in PM (trans-SG) may enhance the body’s immune activation, and at the same time increase the susceptibility of the liver to certain components (cis-SG, EG), resulting in the overexpression of inflammatory cytokines, leading to liver injury. It is also called the "firewood-oil-spark hypothesis", in which the abnormally activated immune system is equivalent to "firewood", the immune-promoting components in PM are equivalent to "oil", and the potentially susceptible components for liver injury is equivalent to "spark". A single risk factor may not catch fire, but when three risk factors are present at the same time, a small spark may cause a raging fire (acute liver injury). This hypothesis has been experimentally confirmed with animal models, providing new ideas for understanding PM-HILI. Indeed, this hypothesis systematically reveals the characteristics of the multi-component, multi-target effect, and immune synergy of the idiosyncratic HILI, providing new strategies for studying the mechanism of idiosyncratic HILI. In addition, it also suggests that PM should be avoided in combination with immune-promoting herbs and encouraged in combination with immune-suppressing herbs, which may be conducive to reduce liver injury. HLA-B*35:01 allele and rs1055348 may also be potentially involved in PM-HILI, but evidence on the experimental animal models is not yet available and needs to be explored.

PM-HILI is a typical representative of HILI. Studying the risk factors, hepatotoxic components, and molecular mechanisms of PM-HILI is essential for a thorough understanding of the pathogenesis, treatment, and prevention of HILI. PM-HILI is considered to be IDILI and occurs only in a small number of susceptible individuals. The risk factors of PM-HILI mainly include MIS and gene polymorphism. The specific components responsible for the idiosyncratic hepatotoxicity are mainly cis-SG and EG. The processing technology of PM should be improved for quality control of cis-SG (<0.10%) and EG (<0.17%). Exposure to light should be avoided during the preparation and storage of PM for preventing the transformation from trans-SG to cis-SG. Underlying diseases, immune status, genetic background, etc. should be considered comprehensively when using PM and related preparations in clinical practice. The usage of biomarkers, such as HLA-B*35:01 allele, inflammatory cytokines like TNF-α, MCP-1, VEGF, etc. to accurately identify the susceptible populations of PM-HILI is necessary for safe medication. PM should be encouraged in combination with immune-suppressing herbs for preventing liver injury. New mechanisms of PM-HILI need to be found and validated by in vivo and in vitro studies and complete the current hypothesis in the future.