Zebrafish as a vertebrate model for high-throughput drug toxicity screening: Mechanisms, novel techniques, and future perspectives

Abstract

Drug toxicity is closely related to both clinical drug safety and new drug development. Therefore, it is vital to understand the mechanisms of drug toxicity fully and to use appropriate research models with advanced technologies. Zebrafish has become an important vertebrate animal model for high-throughput drug screening and toxicity assessment. At the same time, zebrafish has an intact biological complexity, reflecting the whole organism's toxicity, which gives it an advantage over other high-throughput models in toxicity studies. Despite the gradual increase in toxicity studies utilizing zebrafish, a comprehensive and systematic review of the underlying mechanisms and new techniques is still lacking. This review aims to analyze common toxicity mechanisms in zebrafish models, such as oxidative stress, endoplasmic reticulum stress, inflammation, and apoptosis, and macroscopic changes in biological processes like lipid metabolism disorders and neurotransmitter expression abnormalities. It also introduces new technologies applied in toxicity assessment, such as gene editing, novel fluorescence imaging technology, 3D imaging technology, and novel automated technology for high-throughput screening, such as fish capsules. In addition, it also summarizes the advantages and disadvantages of the model. By doing so, it will provide new suggestions for the development and improvement of the model, make it better serve the toxicity study of clinical drugs and provide a more comprehensive perspective for drug toxicity study, thus promoting the development of the field of drug toxicity study.

Graphical abstract

Highlights

• •A systematic summary of the mechanisms and new techniques involved in toxicity assessment of zebrafish model.
• •Highlighting the advantages of the zebrafish model with well-established biocomplexity in toxicity evaluation.
• •Provides novel and unique insights into the improvement of zebrafish models.
• •Prospects for the application of the zebrafish model.

1. Introduction

Toxic side effects of drugs are often a concern. The regulation of the toxicity of medicines and the proliferation of drugs have gained widespread social attention. To date, poisons and drugs have become inextricably linked. For the pharmaceutical industry, studying toxic side effects is an important part of the drug development process. A considerable number of recently developed drugs globally are unsuccessful due to safety concerns, underscoring the importance of rigorous safety testing and monitoring [1]. It is important to recognize that we should always update our understanding of adverse drug reactions and toxicity mechanisms. Traditional drug toxicity evaluation is generally realized through mammalian toxicity testing, and many mammals must carry out safety assessments to develop a new drug. Rodents, such as mice and rats, are commonly used as models for drug safety evaluation. However, the traditional methods of analyzing drug toxicity still have many drawbacks, such as high cost, long cycle time, and poor reproducibility. At the same time, toxicology in the 21st century will pay more attention to the information on the mechanisms by which chemical substances exert toxicity [2]. Therefore, the zebrafish model came into the researchers' view.

Zebrafish have a high degree of genetic similarity to humans, with 71.4% of human genes having at least one zebrafish homologue and 82% of disease genes homologous to zebrafish [3]. Meanwhile, the functions of zebrafish organs, such as the liver, brain, heart, kidney, and pancreas, are highly similar to those of humans. In addition, zebrafish, as a vertebrate model with highly similar organ systems and cell types, has the advantages of being developed in vitro from hyaline eggs, following the 3R's of ethical research, having a wide range of genetic tools that are easy to manipulate genetically, being easy to perform pharmacological experiments, being highly social, and so on [4]. Because zebrafish contains complete biological complexity, it can focus not only on the molecular level but also on the biological processes of toxicity onset in mechanistic studies, which is not the case with other high-throughput models. Zebrafish was mainly used to evaluate environmental pollutants in the early days and is currently the only vertebrate suitable for high-throughput drug screening. In recent years, it has gradually been put into the use of the toxicity assessment of drugs. Therefore, deeply exploring the sensitivity of the model to toxicity and excavating the mechanism of toxicity is very meaningful.

In recent years, much literature has utilized zebrafish for studies, including acute toxicity evaluation, long-term exposure and chronic effects, drug-drug interaction and pharmacokinetic studies, and drug screening. However, the increased frequency of model use has not been matched by a meticulous summary of toxicity mechanisms. At the same time, several new technologies are subtly changing the landscape of toxicity studies. It is time to stop and think about the mechanisms of toxicity behind these studies and the availability of new technologies in zebrafish models. Therefore, in this paper, we will summarize the progress of zebrafish models in drug toxicity evaluation in recent years, discuss in detail the common mechanisms of toxicity and related pathways in toxicity assessment by the model, explore the new technologies conducive to the improvement of the zebrafish model, and consider the current challenges and development trends of the models.

2. Mechanisms of drug-induced toxicity in zebrafish

The study of drug toxicity mechanisms is of paramount importance. From the perspective of drug safety, elucidating the mechanism of adverse effects of drugs within the body, such as the manner of destruction of cellular metabolism or organ cell structure, will enable physicians to assess the efficacy and risk and ensure the safety of patients. In drug development, this represents a pivotal screening stage whereby low-toxicity compounds can be identified at the earliest stages of the research and development process. Furthermore, the study of drug toxicity mechanisms also facilitates comprehension of the physiological and pathological mechanisms of the human body, which provides a foundation for developing novel therapeutic strategies and drives the advancement of the pharmaceutical industry. The mechanisms of drug-induced toxicity in zebrafish encompass both molecular alterations, such as oxidative stress, endoplasmic reticulum (ER) stress, inflammation, and apoptosis, and macroscopic changes in biological processes like lipid metabolism disorders and neurotransmitter expression abnormalities. These macroscopic and microscopic changes are interconnected, collectively providing a comprehensive understanding of the mechanisms underlying toxicity.

2.1. Oxidative stress

Oxidative stress is primarily a loss of balance between oxidative and antioxidant effects in the body, with a tendency for oxidation to occur. When there is a change in the cellular redox balance, it disrupts redox signaling and control. This leads to the accumulation of reactive oxygen species (ROS), which triggers a stress response and the activation of specific genes. The primary mechanism through which oxidative stress occurs in zebrafish is illustrated in Fig. 1. In zebrafish, oxidative stress primarily triggers the activation of NFE2-related factor 2 (Nrf2) and related proteins, leading to the expression and stimulation of antioxidant defenses. This stress also increases the activity of antioxidant enzymes like superoxide dismutase (SOD) and catalase (CAT). Additionally, it affects the glutathione (GSH)/oxidized glutathione (GSSG) system and the thioredoxin (Trx) system. The mitochondrion serves as the primary location for these oxidative processes. Oxidative stress typically works by impeding the respiratory chain.

Fig. 1.

Mechanisms of oxidative stress in the zebrafish model. ARE: anti-oxidant response element; Nrf2: NFE2-related factor 2; ROS: reactive oxygen species; Cyt: cytochrome; TCA: tricarboxylic acid; IL-1β: interleukin-1 beta; IL-18: interleukin-18; GSH: glutathione; GSSG: oxidized glutathione.

Oxidative stress in zebrafish is mainly caused by Nrf2 and other proteins that are part of the body's antioxidant defense system becoming active. These proteins, including Nrf2 and the Nrf2 gene family, activate antioxidant response elements (AREs) to regulate oxidative stress. When oxidative stress occurs, some genes in zebrafish are activated, such as ARE-driven genes, xenobiotic response element (XRE)-driven genes, and peroxisome proliferator-activated receptor (PPAR) response element driver genes [5]. It is through this pathway that many drugs are toxic. For example, flavonoid drugs have been demonstrated to significantly increase the activity of antioxidant enzymes such as SOD, CAT, and glutathione peroxidase (GPx) [6]. This results in the scavenging of free radicals and reduced oxidative damage. In addition, flavonoids improve mitochondrial function and cellular energy metabolism by stimulating the mRNA production of transcription factors for mitochondrial genes, such as NFE2-related factor 1 (Nrf1). Furthermore, they impede mitochondrial cell death and protect neurons from harm by activating the Nrf2 pathway [6].

In addition to the above pathways, the expression of related enzymes also signals the toxicity of the corresponding drugs due to oxidative stress. When fluoxetine was given to zebrafish, the expression of antioxidant-related genes (sod, cat, gpx, nrf1, and nrf2) went up [7]. When zebrafish were exposed to genipin, the levels of reactive oxygen radicals (ROS) and malondialdehyde (MDA) increased significantly. At the same time, the activity of total superoxide dismutase (T-SOD) went down significantly. These results show that the higher levels of oxidative stress and the increased expression of genes related to cell death (apoptosis) are bad for zebrafish development [8]. Aflatoxin FB1 (AFB1) is a highly toxic substance that causes oxidative stress primarily by increasing the production of ROS. In zebrafish embryos, exposure to AFB1 also triggers apoptosis by activating the innate apoptotic pathway [9]. Two chemicals, nicotinamide (NOR-N) [10] and acetyl-11-keto-β-boswellic acid (AKBA) [11], cause oxidative stress in similar ways. Paracetamol, dexamethasone, metformin, and their combinations, which are frequently prescribed for fever or diabetes mellitus, cause an increase in the expression of genes involved in programmed cell death in a zebrafish model. All these drugs also raise the levels of oxidative biomarkers and the activity of antioxidant enzymes, except for paracetamol, which lowers the activity of SOD and CAT enzymes and is the most harmful [12].

Many genes produce and utilize GSH and other antioxidants, which react with sulfhydryl compounds (such as thioredoxins) and regulate them. The names of these genes are gclc, gclm, gss, gr1, ggt1a, and cbsb. There is a significant similarity between genes that are activated by larvae and genes that are activated by oxidative stress in mammals, such as HSP70 and HSP90A2. Another crucial redox system, alongside GSSG, is the Trx system. Nitroglycerin, also known as difurazone, is a compound that promotes the formation of antimicrobial agents. Cell death is caused by targeting an enzyme called thioredoxin reductase 1 (TrxR1) through generating ROS, similar to apoptosis. Nitroglycerin has the potential to be used as a powerful chemical for treating cancer [13].

Mitochondria serve as a primary site for oxidation, and certain medications exert their effects through mitochondrial stress. Succinate dehydrogenase inhibitor (SDHI) based fungicides affect zebrafish's energy metabolism by inhibiting the mitochondrial respiratory chain's succinate dehydrogenase/complex II (SDH/C II). They change how zebrafish use energy by blocking the SDH/C II complex of the mitochondrial respiratory chain [14]. This inhibition reduces energy generation and increases ROS buildup, which could ultimately lead to apoptosis. Furthermore, SDH I has been found to impact various pathways, including fatty acid synthesis. It has been observed to down-regulate genes related to lipid metabolism (srebf1, hmgcra, pparα1, cyp51, and acca1) and up-regulate genes associated with oxidative stress (cat, gpx1a, sod1, and sod2). Additionally, SDH I leads to a decrease in the expression of the sdhb gene and a reduction in the expression of the sdha gene.

Moreover, SDH I has upregulated genes involved in mitochondrial DNA replication and transcription (polg1, tk2, and tfam) [14]. Cisplatin is a chemotherapy drug. It produces excess ROS in the cell, which alters the functioning of the mitochondrial respiratory chain complex and exacerbates the disruption of the oxidative phosphorylation process, leading to further ROS production. Cisplatin binds to mitochondrial DNA, causing permanent harm, which induces oxidative stress and inflammatory reactions, resulting in apoptosis and tissue destruction. The researchers also used Brn3C: EGFP transgenic zebrafish embryos to look into how the drug esomeprazole affected the damage that cisplatin did to the ears. The investigation revealed that esomeprazole effectively mitigated the damage caused by cisplatin, specifically reducing the loss of auditory hair cells [15]. Metformin caused an increase in the formation of superoxide in embryos by blocking the action of mitochondrial complex I. This caused oxidative stress, resulting in a significant increase in the activity of antioxidant enzymes such as SOD, CAT, and GPx. The purpose of this increase in enzyme activity was to counteract the oxidative stress that was generated [16]. Ketamine, a different drug, can also cause mitochondrial dysfunction. This dysfunction leads to an increase in hydroperoxide production, which in turn causes intracellular oxidative stress. Ketamine also enhances the expression of p53 and activates genes that it regulates. It also throws off the balance of glutathione redox during the embryonic stage and raises the production of GSSG, which shows that it affects the redox state inside cells [17]. Table 1 provides a comprehensive overview of the specific gene expression and mechanisms by which the abovementioned drugs induce oxidative stress in the zebrafish model [6,[8], [9], [10], [11], [12], [13],[15], [16], [17]].

Table 1.

Genes and mechanisms associated with oxidative stress in the zebrafish model.

Compound	Gene expression	Mechanism	Refs.
Flavonoids	sod, cat and gpx (↑) nf-κb, nrf1 (↑)	Enhancing mitochondrial function and cellular energy metabolism. In addition, they inhibit mitochondria-mediated apoptosis via the Nrf2 pathway.	[6]
Genipin	p53, bax (↑)	Significantly higher levels of ROS and MDA were accompanied by a significant decrease in T-SOD activity and an increase in oxidative stress, further inducing apoptosis.	[8]
Aflatoxin	caspase-9 (↑)	Increased production of ROS induces oxidative stress and apoptosis.	[9]
Nicofloxacin nicotinate	mda, sod, cat and gpx (↑)	Increased production of ROS to induce oxidative stress.	[10]
Acetyl-11-keto-β-boswellic acid	mn-sod, cat and gpx (↓)	Decreased activity of the antioxidant enzymes T-SOD, CAT, and GPx. Increased activity of MDA.This suggests that it promotes the generation of reactive oxygen radicals, which triggers oxygen radical-mediated lipid peroxidation via oxidative stress mechanisms.	[11]
Paracetamol	bax, bcl2 and p53 (↑)	Decreased SOD and CAT activity.	[12]
Dexamethasone	bax, bcl2 and p53 (↑)	Increased SOD and CAT.	[12]
Nitroglycerin (difurazone)	Trxr1 (↑)	Inducing ROS-mediated apoptosis-like cell death by targeting thioredoxin reductase 1.	[13]
Cisplatin	sod, cat (↑)	The overproduction of ROS, in turn, impacts the activity of the mitochondrial respiratory chain complex, exacerbating deviations in the oxidative phosphorylation process and leading to the production of more ROS.	[15]
Metformin	sod, cat and gpx (↑)	Inhibition of mitochondrial complex I increases superoxide production in embryos, triggering oxidative stress.	[16]
Ketamine	p53, caspase-6, caspase-9 and gssg (↑)	Triggered mitochondrial dysfunction, leading to increased production of hydroperoxides, which in turn triggered an intracellular oxidative stress response.	[17]

“↑” represents gene expression upregulation, “↓” represents gene expression downregulation. Nrf2: NFE2-related factor 2; ROS: reactive oxygen species; MDA: malondialdehyde; T-SOD: total superoxide dismutase; CAT: catalase; SOD: superoxide dismutase; GPx: glutathione peroxidase.

2.2. ER stress

Similar to oxidative stress-related events, ER stress can cause metabolic abnormalities, including disruptions in lipid metabolism and aberrant protein expression. These abnormalities can ultimately lead to apoptosis and autophagy, resulting in cytotoxicity. The protein binding immunoglobulin (BIP), also known as GRP78, is found in the ER and controls the unfolded protein response (UPR) pathway. It acts as a chaperone. When unfolded proteins accumulate in the ER and induce ER stress, the stress sensors release BIP. This activates three factors: transcription factor 6 (ATF6), protein kinase RNA-like endoplasmic reticulum kinase (PERK), and inositol-requiring enzyme 1 (IRE1). Then, these factors set off a chain of events that include higher levels of ATF6, IRE1α, X box-binding protein 1 (XBP1), PERK, eukaryotic translation initiation factor 2α (EIF2α), growth arrest and DNA damage-inducible protein 34 (GADD34), activating transcription factor 4 (ATF4), and the transcripts of genes related to autophagy (autophagy-related protein 3 (ATG3), autophagy-related protein 4B (ATG4B), autophagy-related protein 5 (ATG5), autophagy-related protein 7 (ATG7), autophagy-related protein 12 (ATG12), Beclin1, forkhead box protein O3A (FOXO3A), and microtubule-associated protein 1A/1B-light chain 3B (LC3B)) [18]. Additionally, there is a significant increase in the expression of mTOR and p62 genes, which ultimately activate oxidative and inflammatory pathways through C/EBP homologous protein (CHOP). This, in turn, leads to defects in autophagy and apoptosis, resulting in cellular damage. It has been shown in a zebrafish model that higher levels of BIP, ATF6, and ATF4 show more ER stress [19]. The unfolded protein response (UPR) over-activates autophagy when ER stress lasts long and cellular balance is not restored. This leads to cell death and harm to living things. Upstream factors, such as Beclin 1, initiate and form autophagosomes, while downstream factors, including ATG3, ATG5, ATG7, and LC3 (ATG8), promote the elongation and closure of autophagic membranes. Furthermore, activation of the CHOP can induce autophagy. High CHOP expression leads to upregulation of Bcl-2 associated X protein (BAX protein) in the cytoplasm, triggering the release of mitochondrial cytochrome c. This release activates caspase-9 and caspase-3, ultimately leading to apoptosis [20]. As illustrated in Fig. 2, the particular ER stress mechanisms validated in the zebrafish model are presented. ER stress occurs when unfolded proteins accumulate in the ER. In response to this stress, BIP is produced from three stress sensors: ATF6, PERK, and IRE1.

Fig. 2.

Main mechanisms of endoplasmic reticulum (ER) stress in the zebrafish model. ERS: endoplasmic reticulum stress; BIP: the protein binding immunoglobulin; ATF6: transcription factor 6; ATF4: transcription factor 4; XBP1s: spliced X box-binding protein 1; CHOP: C/EBP homologous protein; BAX: Bcl-2 associated X protein; Apaf-1: apoptotic protease activating factor-1; PERK: endoplasmic reticulum kinase; IRE1: inositol-requiring enzyme 1.

Several medicines can induce ER stress, with the PERK-EIF2α-ATF4-CHOP route being the most common. The molecular changes of gefitinib and afatinib in zebrafish larvae resulted in the up-regulation of stress-related genes such as Bip and increased expression of apoptosis-related genes such as CHOP, JNK, caspase-9, and caspase-3 while decreasing the expression of anti-apoptotic genes, such as bcl-2/bax, through activation of ER stress sensors IRE1, ATF4, and PERK, which together triggered the development of hepatotoxicity. Additionally, these drugs decrease the bcl-2/bax ratio of anti-apoptotic genes. These molecular changes collectively contribute to the initiation and progression of hepatotoxicity [21]. Also, plumbagin can be used as a possible drug to treat cancers resistant to other drugs in a zebrafish embryonic model [22]. Traditional Chinese medicine Xiaoaiping led to a significant upregulation of mRNA expression of genes related to ER stress and concomitantly promoted the activation of apoptosis-related genes, which in turn induced an apoptotic response [23]. Nicotinamide riboside (NR) ameliorated 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced ER stress in a zebrafish model of Parkinson's disease through the same pathway [24]. Whereas isoniazid aggravated ER damage in zebrafish livers in inflammatory states, resulting in the manifestation of a variety of autophagosomes, the ERS-related factors GRP78, ATF6, PERK, IRE1, XBP1s, GRP94, CHOP, autophagy-associated factors, and apoptosis-associated factors at significantly elevated expression levels [20]. This, in turn, worsened developmental defects, inflammatory responses, and autophagy by activating ERS-related genes in a similar pathway [25].

Additional ER stress mechanisms, including the mitogen-activated protein kinase-nuclear factor kappa-light-chain-enhancer of activated B Cells (MAPK–NF–κB) pathway, the Toll-like receptor 4/myeloid differentiation factor 2 (TLR4/MD-2) pathway, and the master regulator GRP78 molecular chaperone proteins, may also be critical in triggering stress. In a zebrafish model, 2-amino-3-methylimidazo[4,5-f]quinoline (IQ) was found to cause ER stress, potentially leading to oxidative stress through the MAPK and NF-κB signaling pathways, as well as inflammation [18]. Cryptotanshinone (CPT) effectively inhibited HCT116 colorectal cancer cells by inducing ER stress. Treatment with CPT increased PERK and phosphorylated PERK expression, resulting in mitochondrial dysfunction, elevated levels of cellular autophagy-related proteins, and autophagy induction in HCT116 cells [26]. The study found that 6-hydroxydopamine caused an increase in the production of GRP78 in a zebrafish model [27]. Dried tangerine peel polysaccharide, a citrus peel polysaccharide, inhibited the inflammatory process induced by lipopolysaccharide (LPS) by interfering with the TLR4/MD-2 signaling pathway, thereby reducing ER stress and suppressing LPS-induced inflammation [28].

In addition to the pathways mentioned above, the ER can generate stress and even lead to lipid peroxidation by activating specific pathways. Arecoline causes stress in the ER by raising ROS levels inside cells and stopping hydrogen sulfide production [29]. Crotinine Bavachin (BV) is critical to causing epithelial-mesenchymal transition (EMT) and renal fibrosis in zebrafish by improving the pathways for fibrosis signaling controlled by TGFβ1 and Notch1 [30]. It is also crucial for controlling the combination of curcumin and tumour necrosis factor-related apoptosis-inducing ligand (TRAIL). ER stress controls apoptosis through the JNK-CHOP pathway, which makes TRAIL-induced apoptosis work better in renal cancer cells. The study demonstrated that it enhanced the susceptibility of kidney cancer cells to chemotherapy [31]. ER stress can activate SREBP-1, a transcription factor that controls the expression of lipid genes. This activation can disrupt lipid metabolism [32]. Alcohol metabolism produces reactive aldehydes, which cause protein adduct accumulation. This accumulation leads to ER stress and further impairment of lipid metabolism. Researchers have discovered that magnesium isoglycyrrhizinate (MgIG) can reduce the expression of genes related to ER stress. As a result, MgIG can effectively protect against alcohol-induced hepatic steatosis [33]. Hesperidin also stopped alcohol-induced ER stress by managing the expression of certain genes, such as CHOP, gadd45a, and edem1. Following alcohol exposure, there was a notable rise in the expression of these genes in zebrafish fry. Conversely, after hesperidin treatment, the expression levels of these genes were dramatically reduced. Like hesperidin, naringenin can block the ethanol-induced ER stress signaling pathway to improve the condition of ER stress [34]. Table 2 illustrates the specific gene expression of the drugs mentioned above in the zebrafish model and the mechanisms or corresponding mechanisms these drugs target [18,[20], [21], [22], [23], [24], [25], [26], [27],[29], [30], [31],33,35].

Table 2.

Genes and mechanisms associated with endoplasmic reticulum stress in the zebrafish model.

Compound	Gene expression	Mechanism	Refs.
2-Amino-3-methylimidazo [4,5-f] quinoline	grp78, chop (↑)	MAPK and NF-κB signaling pathways induce oxidative stress and inflammation. Coffee also reduces the expression levels of grp78, chop and P62, and increases the expression levels of atg5, atg12, beclin 1, lc3-II and bcl-2, thereby preventing ERS caused by IQ.	[18,35]
Isoniazid	grp78, atf6, perk, ire1, xbp1s, grp94, chop, Beclin 1, lc3, atg3, atg12, caspase-3, caspase-8, caspase-9, bax, p53 and cyt (↑)	ER injury in the zebrafish liver under inflammatory conditions allows the expression of multiple autophagosomes.	[20]
Gefitinib and afatinib	ire1, atf4 and perk (↑) chop, jnk, caspase-3 and caspase-9 (↑) bcl-2/bax (↓)	Activation of the endoplasmic reticulum stress sensors IRE1, ATF4 and PERK resulted in the up-regulation of stress-related genes such as bip and increased expression of apoptosis-related genes such as chop, jnk, caspase-3, and caspase-9, while decreasing the anti-apoptotic gene bcl-2/bax ratio.	[21]
Plumbagin	perk, eif2α, atk4 and chop (↑)	Increasing ROS levels to activate ER stress.	[22]
Xiaoaiping	chop, hspa5, hsp90b1 and perk (↑)	The mRNA expression of endoplasmic reticulum stress-related genes was significantly up-regulated and promoted the activation of apoptosis-related genes, such as bax and caspase-3, which induced apoptotic responses.	[23]
1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine	perk, eif2α, atk4 and chop (↑)	MPTP-influenced motor dysfunction, survival time, dopamine neurons, peripheral neurons, and NAD+ levels in a zebrafish model of Parkinson's disease.	[24]
Gemcitabine and paclitaxel	hspa5, chop, ire1, xbp1s and atf6 (↑)	Significant ERS response, which in turn exacerbates developmental abnormalities, inflammatory response and autophagy.	[25]
Cryptotanshinone	perk→(p-perk) (↑) lc3b, Beclin-1 (↑)	Increased expression of perk and phosphorylated perk in HCT116 cells caused mitochondrial dysfunction and induced autophagy.	[26]
6-Hydroxydopamine	eif2α→(p-eif2α) (↑) grp-78 (↑)	Suramin, which inhibits protein tyrosine phosphatase 1B, which in turn affects the IRE1-mediated endoplasmic reticulum stress signaling pathway, resulting in altered transcription of XBP-1 shear and α-mannosidase-like proteins, which modulate the cellular response to endoplasmic reticulum stress.	[27]
Arecoline	p53 (↑)	Increased intracellular ROS levels and interfered with endogenous hydrogen sulphide production.	[29]
Bavachin	tgfβ1, notch1 (↑)	Induced EMT and renal fibrosis in zebrafish.	[30]
Curcumin	jnk-chop (↑)	Combined treatment with curcumin and TRAIL is involved in the regulation of apoptosis through the JNK-CHOP pathway, which enhances the effect of TRAIL-induced apoptosis in renal cancer cells.	[31]
Ethanol	srebp-1(↑) chop, gadd45a and edem1(↑) xbp1(↑)	Endoplasmic reticulum stress activates SREBP-1, a transcription factor responsible for lipid gene expression, leading to lipid metabolism impairment.	[33]

“↑” represents gene expression upregulation, “↓” represents gene expression downregulation, “→” represents phosphorylation. ERS: endoplasmic reticulum stress; IQ: 2-Amino-3-methylimidazo [4,5-f] quinoline; ER: endoplasmic reticulum; IRE1: inositol-requiring enzyme 1; ATF4: transcription factor 4; PERK: protein kinase RNA-like endoplasmic reticulum kinase; ROS: reactive oxygen species; MPTP: 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; NAD⁺: nicotinamide adenine dinucleotide; XBP-1: spliced X box-binding protein 1; EMT: epithelial-mesenchymal transition; TRAIL: tumour necrosis factor-related apoptosis-inducing ligand; JNK: c-Jun N-terminal kinase.

2.3. Inflammation response

Inflammation is a non-specific response of the body to stimulation or injury. Its primary role is to protect the body from damage and promote repair. The onset of inflammation is intricately linked to the modulation of diverse cellular and molecular pathways [36]. This complex process encompasses phenomena such as cell membrane disruption, the induction of oxidative stress, programmed cell death (apoptosis), and the activation of immune responses. These events are triggered by many factors that converge to instigate inflammatory reactions [37,38].

Inflammation-induced zebrafish drug toxicity helps us better understand drugs' potential effects on organisms. When the drug triggers an inflammatory response in zebrafish, it may lead to toxic reactions. These reactions may involve multiple organs and systems, including cardiovascular, liver, kidney and nerve. MAPK pathway is an important signal transduction pathway in regulating cell proliferation, apoptosis and inflammatory response. The MAPK pathway includes branches, such as extracellular regulated protein kinases (ERK), JNK and protein 38 (p38), which play a crucial role in inflammation [39]. In the MAPK signaling pathway, MAPK kinase kinase (MAPKKK) is first activated, and then MAPK kinase (MAPKK) is activated. Finally, MAPKK activates MAPK by double-site phosphorylation of threonine (T) and tyrosine (Y). This phosphorylation process enables MAPK to obtain activity, participating in downstream signal transduction and gene expression regulation. Activated MAPK can enter the nucleus and affect the production and function of inflammation-related proteins by regulating the transcription and expression of specific genes [40]. These proteins include inflammatory mediators, chemokines, and adhesion molecules, which play a key role in inflammation.

Moreover, MAPK can also interact with NF-κB signaling pathway to regulate the expression of inflammation-related genes [[41], [42], [43]]. In addition, AKT, also known as protein kinase B (PKB), is a serine/threonine protein kinase that participates in and affects the inflammatory response mainly by regulating intracellular signaling pathways. AKT is involved in the inflammatory mechanism by regulating a variety of downstream molecules. For example, it can affect the activity of NF-κB. When activated, AKT can phosphorylate and inhibit IκB kinase (IKK), thereby preventing IKK from degrading NF-κB inhibitors. This causes NF-κB to be released and transferred to the nucleus, activating inflammation-related gene expression. It can also affect the mTOR signaling pathway, thereby regulating cell autophagy and metabolic processes. The regulation of MAPK and AKT signaling pathways is crucial during development. Recently, changes in these two pathways have been recorded in the evaluation of the developmental toxicity of Mevinphos. The inflammatory response-related genes cox2a, il-1β, and tnfα were significantly up-regulated in zebrafish larvae exposed to Mevinphos, and the anti-inflammatory response genes cat, sod1 and sod2 were significantly down-regulated [44]. Mevinphos affects the MAPK signaling pathway by up-regulating the phosphorylation of ERK and reducing the phosphorylation of p38 and JNK. At the same time, the phosphorylation level of AKT also changed. In addition, thiabendazole causes developmental defects by mediating changes in the phosphatidylinositol 3-kinase (PI3K)/AKT and MAPK pathways [45]. Acaritone can also damage the liver by down-regulating the PI3K/AKT signaling cascade in zebrafish and affecting the downstream signaling factor glycogen synthase kinase 3β (Gsk3β)/β-catenin [46].

Wnt signaling pathways include Wnt/β-catenin, Wnt/programmed cell polarity (PCP) and Wnt/Ca²⁺ pathways. The Wnt/β-catenin signaling pathway is a kind of β-catenin-dependent Wnt signal transduction, also known as the canonical pathway, which mainly controls cell proliferation. In classical Wnt signal transduction, in the absence of Wnt ligands, β-catenin is degraded by the multiprotein complex β-catenin disrupting complex. In the presence of Wnt protein ligands, as shown in Fig. 3, Wnt binds to its core receptor complex and activates Wnt signaling by recruiting Dishevelled (Dvl) proteins and blocking or destroying the formation of Axin/glycogen synthase kinase 3 (GSK3)/adenomatous polyposis coli (APC) complexes, thereby inhibiting the degradation of β-catenin and leading to the accumulation of β-catenin in the cytoplasm. Accumulated cytoplasmic β-catenin translocates into the nucleus and activates proliferation and differentiation-related genes by interacting with the transcription factor T cell factor (TCF) family and activating coactivators [[47], [48], [49]].

Fig. 3.

Mechanisms of Toll-like receptor (TLR) and Wnt/β-collagen signaling pathways. LSP: lipoprotein receptor related protein; TNF: tumour necrosis factor; MyD88: myeloid differentiation factor 88; IRAK1/4: interleukin-1 receptor-associated kinase 1/4; TRAF-6: TNF receptor associated factor 6; TAK-1: transforming growth factor beta-activated kinase 1; IKK: IκB kinase; NF-κB: nuclear factor-κB; NLRP3: NACHT, LRR and PYD domains-containing protein 3; GSK3: glycogen synthase kinase 3; APC: adenomatous polyposis coli.

The Wnt/β-catenin signaling pathway is highly conserved in vertebrates and invertebrates. Cardiac developmental toxicity induced by benzophenone and anti-tumour drug adriamycin (ADR) is thought to be related to it. Some scholars have evaluated the gene expression of axin2, lef1 and β-catenin related to the Wnt signaling pathway in zebrafish after ADR induction and found that the expression of these three genes was significantly up-regulated. The gene expression of axin2, lef1 and Wnt/β-catenin was also observed in zebrafish with benzophenone-induced cardiac dysplasia. It can be speculated that the Wnt/β-catenin signaling pathway is associated with cardiac dysplasia [50,51]. The inhibitor IWR-1 is used to verify whether the drug affects the expression of immune system-related genes, such as inflammation and apoptosis, through the Wnt/β-catenin signaling pathway. Inhibition of the Wnt pathway may represent a new treatment for heart disease [52].

TLR also mediates the activation of NF-κB signaling [49]. TLR is a kind of receptor that recognizes exogenous pathogens and plays a vital role in inflammation. As shown in Fig. 3, after recognizing pathogen-associated molecular patterns (PAMPs), the conformation of TLR molecules changes, and the intracellular TIR domain recruits adaptor proteins to activate downstream signaling pathways. Most TLRs activate NF-κB through the myeloid differentiation factor 88 (MyD88)-dependent pathway, thereby inducing the production of pro-inflammatory cytokines and further inducing inflammation [[53], [54], [55]]. Clethodim can induce TLR signaling activation through TLR4-mediated and MyD88-dependent signaling pathways, increasing the expression of TLR4, MyD88, and NF-κB p65 at the mRNA level. In contrast, another adaptor of TLR activation, TRIF, decreased sharply. Interferon regulatory factor (IRF) is a transcriptional regulator that plays a vital role in inflammatory activation. With the increase in clethodim, IRF4, IRF5, and IRF8 significantly up-regulated the mRNA level of IRF3 and significantly down-regulated the mRNA level of IRF7. However, no significant change was observed in the mRNA level of IRF7 in zebrafish embryos [56,57]. Spinetoram elicited the activation of TLR4/NF-κB-dependent immune signals in embryonic zebrafish. The expression of TLR4 and MyD88 was initially up-regulated and subsequently down-regulated, significantly surpassing that of the control group. The expression levels of NF-κB and p65 progressively escalated, while the pro-inflammatory cytokines interleukin 6 (IL-6), IL-8, and interferon-γ (IFN-γ) exhibited an initial uptick followed by a decline. Conversely, CXC motif chemokine ligand 1 (CXCL1) demonstrated a gradual increase, whereas the anti-inflammatory cytokine IL-10 was downregulated, its levels significantly reduced compared to those in the control group. This dysregulation collectively contributed to the immunotoxic effects observed in zebrafish embryos [57]. Wang et al. [58] employed the NF-κB inhibitor QNZ in rescue experiments and discovered that it impeded the nuclear translocation of NF-κB p65 by suppressing its phosphorylation within the cytoplasm, thereby inhibiting the TLR4/NF-κB signaling pathway activation. Notably, the NF-κB inhibitor QNZ demonstrated efficacy in mitigating the immunotoxic effects observed in zebrafish.

The Janus kinase-signal transduction and transcriptional activation (JAK-STAT) pathway regulates biological processes such as cell growth, differentiation and immune response. And it also plays an essential role in the inflammatory response. Inflammatory mediators such as interferons and interleukins can regulate the inflammatory response by activating the JAK-STAT pathway. Specifically, cytokines bind to plasma membrane receptors, resulting in receptor conformation changes and homodimers' formation. This conformational change makes the respective binding JAKs close to each other, resulting in cross-phosphorylation and activation of JAK activity. The activated JAK further phosphorylates the tyrosine residues at the intracellular end of the receptor, providing an anchor site for STAT or other cytoplasmic proteins with the SH2 domain. STAT binds to the phosphorylated tyrosine residues of the receptor through the SH2 domain, and then its C-terminal tyrosine residues are also phosphorylated by JAK. Phosphorylated STAT molecules are dissociated from the receptor, forming homodimers and exposing their nuclear localization signals. The dimerized STAT translocates into the nucleus and binds to the regulatory sequences of specific genes to regulate the expression of related genes, thereby triggering an inflammatory response [[59], [60], [61]]. The JAK-STAT pathway activated by growth factors and cytokines has been shown to play a crucial role in signal transduction from cell membrane receptors to the nucleus and other developmental processes. Cheng et al. [62] found that the JAK-STAT signaling pathway mediates cyhalofop-butyl-induced immunotoxicity, and tetrabromobisphenol A bis (2-hydroxyethyl ether) (TBBPA-DHEE) cell motility, survival and motor activity are controlled by cytokines through the JAK-STAT pathway. Overexpression of the JAK-STAT protein may affect development and cause neurotoxicity [63]. The up-regulation of inflammatory cytokines is also one of the reasons for the cardiac toxicity of zebrafish larvae caused by clozapine [64].

2.4. Apoptosis

Apoptosis constitutes a pivotal mechanism underlying drug-induced cytotoxicity, as illustrated in Fig. 4, with mitochondria serving as central actors in this process. The cascade of apoptotic triggers can be bifurcated into extrinsic and intrinsic signaling pathways. The external signaling pathway mainly comes from the death ligand tumour necrosis factor (TNF) signaling pathway, the binding and agonism of the signaling molecules CD95L (also known as FasL) and TNF-related apoptosis-inducing ligand (TRAIL) with their receptors CD95 (also known as Fas) and TRAIL-R, and the release of cell perforin by natural killer (NK) cells [[65], [66], [67], [68]]. Intracellular signals, conversely, involve crosstalk between apoptosis and oxidative stress, ER stress, inflammatory effects, trophic factor deprivation, calcium inward flow, and kinase inhibition [69,70]. These upstream signals travel deeper into the cell, prompting the translocation of BAX to the mitochondrial membrane, which subsequently interacts with tBID (the truncated BH3-interacting domain death agonist) on the outer mitochondrial membrane and induces enhanced mitochondrial outer membrane permeabilization (MOMP) [69].

Fig. 4.

BAX and BAK are involved in the mechanism of apoptosis. Red arrows indicate up-regulation and green arrows indicate down-regulation. When the expression of key genes is altered, or the content of related proteins is changed, the drugs may cause apoptosis and thus induce zebrafish-related toxicity through this pathway. TNF: tumour necrosis factor; TRAIL: tumour necrosis factor-related apoptosis-inducing ligand; BID: BH3-interacting domain death agonist; tBID: truncated BID; BAD: Bcl-2-associated death promoter; BCL-xL: B-cell lymphoma-extra large; BAX: Bcl-2 associated X protein; BAK: BCL-2 homologous antagonist/killer; BOK: Bcl-2-related ovarian killer; MOMP: mitochondrial outer membrane permeabilization; MIMP: mitochondrial inner membrane permeabilization; DIABLO: direct inhibitor of apoptosis protein-binding protein with low pI; HTRA2: high-temperature requirement protease A2; XIAP: X-linked inhibitor of apoptosis protein; APAF1: apoptotic protease activating factor 1; ENDOG: endonuclease G; AIF: apoptosis-inducing factor.

Where tBID is generated by shearing BID (BH3-interacting domain death agonist) located in cytoplasmic lysate by caspase 8. The sheared BID is transferred to the mitochondrial membrane and further triggers the downstream apoptotic pathway. It is believed that BCL-2 homologous antagonist/killer (BAK) and BCL-2 ovarian killer (BOK) have similar actions as BAX, but the antagonistic inducers of BOK need to be further investigated [69]. In addition, gtBID and jtBID have also been associated with enhanced apoptotic signaling, such as BAX. gtBID production involves target cell destruction by cytotoxic T lymphocytes (CTLs) and is produced by shearing of granzyme B (GZMB). In contrast, the production of jBID is associated with activated JNK. Certain anti-apoptotic proteins of the BCL-2 family (BCL-2, BCL-XL, MCL-1, etc.) are essential for MOMP inhibition. The core of the apoptotic pathway lies in the formation of MOMP, which is induced by a series of complex changes in the membrane BAX, followed by mitochondrial contents such as cytochrome c, direct inhibitor of apoptosis protein (IAP)-binding protein with low pI (DIABLO) and mitochondrial high-temperature requirement protease A2 (HTRA2), apoptosis-inducing factor (AIF) and mitochondrial endonuclease G (ENDOG) [69], which penetrate out of the intermembrane space and enter the cytosol [71]. Cytochrome c in the cytosol combines with apoptotic protease activating factor 1 (APAF1) and caspase 9 to form apoptotic vesicles, which in turn induce the activation of a series of apoptotic factors (e.g., caspase 3, caspase 7 and caspase 6), which leads to a series of cellular function and morphology alterations, including microscopic function disruption, cellular crumpling, and DNA fragmentation, and ultimately causes apoptosis. DIABLO and HTRA2, on the other hand, indirectly cause apoptosis by antagonizing the caspase-inhibitory function of the X-linked inhibitor of apoptosis protein (XIAP), which blocks its inhibitory effect on apoptotic factors [72]. AIF and ENDOG induce apoptosis by inducing DNA fragmentation. Further, mitochondrial inner membrane permeabilization (MIMP)-induced mtDNA overflow will generate inflammatory vesicles via the cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS/STING) pathway and induce cellular focal death [72,73]. Notably, not all members of the caspase family of enzymes are directly involved in apoptosis. Only caspase 3, caspase 7, and caspase 6 act as executioners of apoptosis [71], whereas inflammation-associated caspases 1, 4, 5, 11, and 12 do not play a role. In addition, the P53 signaling pathway, a cell cycle regulatory pathway, has been implicated in apoptosis, and ER stress-induced calcium overload also agonizes caspase 12, which in turn activates caspase 3, caspase 7, and caspase 6. In contrast, activation of the PI3K/AKT signaling pathway may block or delay apoptosis [74].

Intuitively, the expression of BAX and BCL-2 is highly correlated with apoptosis, and BCL-2/BAX has been considered the most significant marker of apoptosis [75]. The expression of caspase 8, caspase 3, caspase 7 and caspase 6 is also an indicator we often pay attention to. Upstream ligand receptors such as TNF receptor subfamily members and their signaling systems, as well as downstream mitochondrial contents such as cytochrome c, DIABLO and HTRA2, AIF and ENDOG, and P53 signaling pathway-related proteins, are also evidence of apoptosis. Recent studies have utilized transgenic zebrafish to observe the number of immune cells in zebrafish embryos after drug interventions, which, in conjunction with alterations in apoptosis-related genes, can be used to verify whether apoptosis is one of the causes of developmental toxicity induced by drugs [52,76]. On the other hand, roxithromycin, an antimicrobial drug, interferes with the differentiation of motor neuron progenitor cells and behavioral deficits in zebrafish by up-regulating the bax/bcl-2 ratio and the expression of caspase-3a.

In contrast, activation of neural progenitor protein 1 (ngn1) genes attenuates the motor neuron abnormalities in zebrafish caused by exposure to roxithromycin [77]. Inflammatory response and oxidative stress are intracellular signals that cause apoptosis. ROS content and inflammation key gene nf-κb are often detected along with apoptosis-related genes to explain the cause of drug-induced apoptosis and further corroborate the occurrence of apoptosis. Tan et al. [78] found that Bajitian (Morinda officinalis) extract would reduce the number of neutrophils in zebrafish. However, gene expression analysis showed that it did not significantly change caspase 3/7 activity, nor did it up-regulate the expression of the bax and nf-κb genes, which indicated that the altered neutrophil counts after Bajitian extract treatment are not due to apoptosis. Expression of pi3k, akt, nf-κb, p65, iκbα, il-8, tnfα, p53, caspase 3, caspase 8, and bax was significantly elevated in zebrafish under exposure to evodiamine, one of the main ingredients of Evodiae fructus. It also inhibited the expression of bcl-2 and thus induced severe cellular damage resulting from inflammatory response, ultimately leading to apoptotic cell death [79]. On the other hand, chelerythrine induced ROS and inflammatory responses, which activated apoptosis by intracellular signaling. Apoptosis-related genes bax, caspase-9 and p53 were significantly elevated in the study, while bcl-2 was decreased considerably, along with significant changes in the expression levels of keap1, sod1, nrf2, tnf-α, stat3 and nf-κb [76]. Being a BAX agonist, BTSA1 can be utilized in studies to verify whether the BAX signaling pathway is a possible cause of toxicant-induced developmental toxicity. Yuan et al. [80] first reported the developmental toxicity of esketamine in zebrafish and found that BTSA1 could reverse the anti-apoptotic and growth-slowing effects of esketamine. This implies that esketamine may inhibit apoptosis through the BAX/Caspase 9/Caspase 3 pathway, hindering embryonic development and adversely affecting children. In addition, other BAX agonists such as OICR766A, BIF-44, and BAX inhibitors such as iMAC1/BAI1, MSN compounds and DAN004 have been reported and are available for researchers to select and utilize [69]. Studies of drug-induced toxicity through apoptosis in zebrafish are listed in Table 3 [58,[76], [77], [78], [79], [80], [81], [82], [83]].

Table 3.

Related toxicity of drugs through induction of apoptosis.

Drugs	Gene expression	Mechanism	Refs.
Sulfoxaflor	p53, caspase3, bax, bax/bcl-2 (↑) bcl-2 (↓)	Inducing apoptosis in zebrafish embryonic cells.	[58]
Chelerythrine	bax, caspase-9 and p53 (↑) bcl-2 (↓)	Activation of apoptosis by intracellular signaling ROS and inflammatory response.	[76]
Roxithromycin	bax/bcl-2 ratio and caspase-3a expression (↑)	Interference with motor neuron progenitor cell differentiation through oxidative stress-induced apoptosis induces motor neuron malformations and behavioral deficits in zebrafish.	[77]
Bajitian extract	No significant change in caspase 3/7, bax and nf/7 expression	Altered neutrophil counts after bajitian extract treatment are not due to apoptosis.	[78]
Evodiamine	pi3k, akt, nf-κb p65, iκbα, il-8, tnfα, p53, caspase 3, caspase 8, bax (↑) bcl-2 (↓)	Inducing apoptosis through inflammation.	[79]
Esketamine	caspase9, caspase3b, caspase3a, baxb, baxa and caspase6 (↓)	Esketamine may inhibit brain cell apoptosis by affecting BAX-mediated endogenous signaling pathways, which results in developmental delays.	[80]
Piperlongumine and piperine	bax and bcl-2 (↑)	Early developmental toxicity in zebrafish induced by apoptosis.	[81]
Dimethyl fumarate	p53, caspase3, bax/bcl-2 (↑)	Causing zebrafish heart developmental delay and induces apoptosis in zebrafish cardiomyocytes.	[82]
Flunitrazepam and its metabolites	caspase-3, caspase-9, bax, p53 and bcl-2 (↑)	Oxidative stress, apoptosis, and histone hypoacetylation, neurotoxicity.	[83]

“↑” represents gene expression upregulation, “↓” represents gene expression downregulation. ROS: reactive oxygen species, BAX: Bcl-2 associated X protein.

2.5. Lipid metabolism disorders

Lipid metabolism is a finely regulated biochemical process in the human body, which covers the key links of lipid synthesis, storage and decomposition. Many genes precisely regulate this process and involve many metabolic pathways and physiological mechanisms [84,85]. When lipid metabolism is disordered, it may lead to abnormal accumulation of fat in the body, leading to obesity and a series of related health problems, including cardiovascular disease [86], type 2 diabetes and nonalcoholic fatty liver disease [87].

The liver plays a central role in lipid metabolism and is responsible for synthesizing a variety of critical biomolecules. Bile acid (BAs), a cholesterol derivative synthesized in the liver, is a highly conserved molecule during evolution and plays a crucial role in regulating bile metabolism and lipid metabolism [88]. BA exerts its regulatory function by activating specific receptors, such as the farnesoid X receptor (FXR). FXR is the earliest discovered endogenous BA receptor, which plays a crucial role in maintaining the balance of lipid metabolism. According to Fig. 5, FXR effectively reduces BA concentration in hepatocytes by inhibiting the expression of genes cyp7a1 and cyp8b1 related to BA synthesis [89]. In addition, the activation of FXR can also limit the accumulation of BAs in hepatocytes by inhibiting the expression of BA membrane transporter sodium taurocholate cotransporting polypeptide (NTCP) and promoting the excretion of BAs from the liver by enhancing the expression of BA efflux pump BSEP, thereby maintaining the balance of BAs. The FXR/small heterodimer partner (SHP) signaling pathway formed by FXR and SHP is vital in regulating cholesterol and BA homeostasis in the enterohepatic circulation. When BA levels rise, FXR can inhibit the transcription of SHP, which in turn affects the metabolism of cholesterol and BAs [90]. FXR is also involved in regulating lipid metabolism, reducing triglyceride (TG) levels by inhibiting steroid regulatory element binding protein 1c (SREBP-1c) and inhibiting key genes in the process of adipogenesis, such as stearoyl-CoA desaturase 1 (SCD-1). In addition, FXR can promote β-oxidation of fatty acids by activating peroxisome proliferator-activated receptor α (PPARα), thereby preventing the accumulation of triglycerides in the liver [89]. In zebrafish model, different concentrations of aqueous extract of Psoralea corylifolia (AEFP) treatment increased the expression of downstream genes such as shp, cyp7a1, cyp8b1, bsep, mrp2, pparγ, me-1, scd-1, NTCP, lipoprotein lipase (LPL), cpt-1 and cpt-2, while decreased the expression of pparα. When FXR and PPARα agonists were combined with AEFP, the decrease of fluorescence intensity and area index of zebrafish liver caused by AEFP was reversed. On the contrary, the application of FXR and PPARα inhibitors exacerbated these changes, indicating that AEFP may interfere with its downstream pathways by down-regulating the expression of fxr and pparα genes, thereby causing hepatotoxicity [91]. The antibiotic sulfamethoxazole (SMX) also affects molecular signal transduction and pathways such as the PPAR signaling pathway and BA synthesis. Reducing the transcription of related genes destroys fatty acid β oxidation, synthesis and transport processes, disrupts fatty acid homeostasis, and ultimately leads to excessive deposition of triglycerides, resulting in liver damage [92].

Fig. 5.

Mechanisms of farnesoid X receptor (FXR)-mediated lipid metabolism. BA: bile acid; NTCP: sodium taurocholate cotransporting polypeptide; TG: triglyceride; FA: fatty acid; SHP: small heterodimer partner; SREBP-1c: steroid regulatory element binding protein 1c; SCD-1: stearoyl-CoA desaturase 1; PPARα: peroxisome proliferator-activated receptor α; CPT-1: Carnitine palmitoyltransferase-1.

In the field of lipid metabolism, the type and function of lipoproteins are the key to understanding how lipids are transported and metabolized in the body. The lipoprotein types involved included chylomicrons (CM), very low density lipoprotein (VLDL), low density lipoprotein (LDL), and high density lipoprotein (HDL). These lipoproteins carry apolipoproteins (ApoA and ApoB) and triglycerides, which are important biomarkers for assessing lipid metabolism. Chen et al. [93] studied the side effects of cysteamine and revealed its potential effects on lipid metabolism. Studies have found that cysteamine can cause a significant increase in alanine aminotransferase (ALT), aspartate aminotransferase (AST), total triglycerides and total cholesterol levels, and cysteamine-induced zebrafish liver lipid deposition is serious. At the same time, the levels of factors related to adipogenesis increased while those related to lipid transport decreased. This suggests that cysteamine may interfere with lipid metabolism by inhibiting the Wnt signaling pathway, resulting in hepatotoxicity. The lipoprotein lipase (LPL) gene plays a vital role in lipid metabolism. It encodes a lipase that plays a role in the liver, intestines and adipocytes, responsible for the hydrolysis of circulating triglycerides and releasing fatty acids for tissue absorption [94]. LPL activity is critical for lipid transport and metabolism. In addition, the effect of Lenvatinib on lipid metabolism has also received attention as a thyroid cancer treatment [95]. Studies have shown that Lenvatinib exposure can lead to liver development defects in zebrafish models, and the level of lpl mRNA involved in fatty acid transport is down-regulated. This finding suggests that Lenvatinib may affect the transport and metabolism of fatty acids by affecting the expression of lipid metabolism-related genes.

2.6. Abnormal neurotransmitter expression

In addition to a high degree of physiological and genetic homology to humans and the ability to operate genes, zebrafish have a central nervous system morphology similar to humans [96] and a high degree of sociality [4]. It has a good reference value in evaluating the toxicity of brain tissue and nervous system [97]. Neurotransmitter expression is often of interest in the neurotoxicity evaluation of drugs. Disturbances in the release or metabolism of various neurotransmitters may cause a range of behavioral abnormalities, including seizures, depression, anxiety, and others. Neurotransmitter assays used in toxicity studies assess neuroprotection or neurotoxicity in the central nervous system by measuring the concentrations of individual neurotransmitters or groups of neurosteroids in various neurotransmitter systems and their metabolism-related substances [98]. Neurotransmitter systems include the noradrenergic nervous system, the histaminergic nervous system, the cholinergic nervous system, the dopaminergic nervous system, the 5-hydroxytryptophanergic nervous system, the gamma-aminobutyric acid nervous system, the oxytocinergic nervous system and the hypothalamic-pituitary-adrenal axis, etc., which work together in precise collaboration and interplay to jointly dominate the physiological activities of the organisms, such as behavior, emotions and memory. As shown in Fig. 6, various neurotransmitter systems work in concert with each other to maintain the functional stability of the nervous system. When one or more neurotransmitters are abnormally expressed, this stability may be disrupted, thus causing abnormal behavior in zebrafish.

Fig. 6.

Abnormal neurotransmitter expression. ERK: extracellular regulated protein kinases; JNK: c-Jun N-terminal kinase; E/I: excitability/inhibitory; HPI: hypothalamus-pituitary-interrenal; IT: isotocin; OXT: oxytocin.

Excitatory/inhibitory (E/I) balance in synaptic transmission and neural circuits is a common hallmark of various neurodevelopmental disorders [99]. The excitatory neurotransmitter glutamate (Glu) and the inhibitory neurotransmitter gamma-aminobutyric acid (GABA) act synergistically and play a decisive role in maintaining the E/I balance [98]. Meanwhile, the inhibitory neurotransmitter 5-hydroxy tryptamine (5-HT) can partially localize GABA receptors through cAMP-protein kinase A (PKA) anchored to a kinase-anchoring protein (AKAP), which interacts with GABA to co-regulate E/I balance in neural circuits [100]. Especially in the midbrain-limbic system, the combined effects of GABA and 5-HT on the behavioral and emotional performance of individuals are significant and play a crucial role in causing emotional abnormalities in animals and drugs to counteract anxiety, epilepsy, convulsions, and so on. The GABA pathway-related genes we usually focus on include glutamic acid decarboxylase gad2, glutamic acid decarboxylase 1b (gad, gad1b), γ-aminobutyric acid transaminase abat, γ-aminobutyric acid type A receptor subunit β2 gabrb2, GABAA receptor α1 and α2 (gabra1 and gabra2), GABA transporter 1 (gat1, slc6a1a), vesicular GABA transporter (vgat, slc32a1), vesicular glutamate transporter 1 (vglut1, slc17a), glutamate ionotropic receptor N-methyl-D-aspartate (NMDA)-type subunit 1a (grin1a) glutamate ionotropic receptor, AMPA-type subunit 1 gria1a, and lysosomal carrier series 12 Member 2 slc12a2, and others. Genes related to the 5-HT pathway include tryptophan hydroxylase 1 (tph1, tph1b) and TPH2 (tph2), the 5-hydroxytryptamine transporters (sert, slc6a4a and slc6a4b), monoamine oxidase (mao), the two inhibitory receptors for 5-HT, 5-HT1A (htr1aa and htr1ab), and the inhibitory receptor for 5-HT1B (htr1b), among others [98,99]. Wu et al. [101] found that changes in GABA levels in the brains of zebrafish exposed to diazepam and its recovery patterns were centrally specific. Diazepam may cause persistent abnormalities in zebrafish motor activity by affecting GABA levels in the brain. Yang et al. [102] evaluated the brain and liver toxicity of three antiepileptic drugs, namely oxcarbazepine (OCBZ), lamotrigine (LTG) and carbamazepine (CBZ). The results showed that LTG and CBZ significantly increased brain acetylcholinesterase (AChE) activity, and the three antiepileptic drugs significantly increased brain GABA levels, while there were no significant changes in MDA and Glu brain levels. Cortisol and pregnanolone are important neurosteroid markers associated with photosensitivity and memory deficits [103]. Xie et al. [98] found that carbamazepine treatment resulted in the expression levels of the genes gad2, abat, gab2, gabrg2, gria1a, and slc12a2 in the GABA-Glu pathway in zebrafish brains being significantly down-regulated, the expression levels of genes thp2, htr1aa and htr1b in the 5-HT pathway were changed, and the mRNA levels of genes crha, actha, pc1 and pc2 in the hypothalamus-pituitary-interrenal (HPI) axis were significantly up-regulated, respectively. Meanwhile, GABA levels in zebrafish brains decreased while 5-HT and plasma cortisol levels increased, suggesting that environmentally relevant low concentrations of carbamazepine are sufficient to interfere with the anxiety response in zebrafish via the GABA/5-HT system and the HPI axis.

The expression of several genes associated with neural development and circadian rhythms is often measured together in response to nerve cell damage. Studies on cortisone have pointed out that cortisone not only represses some of the genes in the GABA synthesis pathway but also affects the expression of genes involved in the visual cycle (rpe65a, opn3, and vsx1) and the circadian network (tefa, per2, and clocka). It was shown that mRNA levels of gabra1, gabbb1b, slc6a1a and zfkcc2 were significantly down-regulated, while the mRNA of gabr1b was suppressed in the retina, suggesting that cortisone exposure will cause anxiety and retinal dysfunction in zebrafish [104]. Syn2a and gap43 are associated with neuronal development and neurotransmitter release. In studies for ephedrine and codeine, these genes were up-regulated, along with a decrease in GABA levels, down-regulation of gad2, abat, slc12a5b, and slc12a2, and up-regulation of slc6a11b, indicating that ephedrine and codeine exposure may impair neuronal development and neurobehavior in zebrafish larvae [105]. In the study of some neurodegenerative diseases, such as Parkinson's disease, and Alzheimer's disease, the monitoring of genes related to neuronal growth and development is necessary. Xie et al. [77] examined changes in the expression of genes associated with zebrafish motor neurons (ngn1, olig2), axon growth (cd82a, mbpa, plp1b, sema5a), neuroimmune responses (aplnra, aplnrb), and development, and found that roxithromycin impeded motor neuron by inducing aberrant apoptosis and progenitor cell differentiation.

The NMDA receptor is a subunit of the ionotropic glutamate receptor, which plays an essential role in the development of the nervous system, formation of neural circuits, and transmitter transmission. Neurodevelopmental toxicity induced by hygromycin may be related to neuronal apoptosis regulated by NMDA receptors. Ma et al. [106] showed that hygromycin inhibited the expression of several subtypes of NMDA receptors (grin1a, grin1b, grin2bb, grin2ca), which induced apoptosis in the brain of zebrafish embryos. When the NMDA receptor agonist NMDA was administered, inhibitory neurobehavioral changes, as well as brain NMDA receptor expression and apoptosis, were completely antagonized in zebrafish larvae. They, therefore, concluded that hygromycin induced inhibitory neurobehavioral toxicity in zebrafish larvae early in development by inhibiting NMDA receptor activity and expression.

Oxytocin (OT) is a more primitive regulator that affects all aspects of physiological activity, and the regulation of animal social behavior by the oxytocinergic system is now widely recognized [107]. Zebrafish isotocin (IT) is highly similar to mammalian oxytocin (OXT), and experiments often focus on estrogen receptors (esr1, esr2a, and esr2b), IT (it) and ITR (itr), solute carrier family 12 members 5a and 5b (slc12a5a and slc12a5b), and solute carrier family 12 member 2 (slc12a2), to monitor the expression and regulation of oxytocinergic nerves in zebrafish [99]. At the same time, no single system of neuromodulators acts independently of others in a complex neuromodulatory network, and they tend to interact with each other and exert potentiating or inhibitory effects on each other. Currently, neuromodulatory interactions between OT and opioids, as well as between OT and 5-HT, have been demonstrated [108], whereas 5-HT is a modulator of noradrenergic (NA) neurons [109], and NA is involved in influencing behavioral activity in zebrafish through the ERK/activator protein 1 (AP-1) pathway [109,110]. As shown in Fig. 7, briefly, Ang II binding to AT1R leads to the activation of signaling kinases (ERK, FRK, and JNK), and the activated signaling kinases phosphorylate downstream transcription factors (Elk-1, Fos, Jun, and ATF-2), which in turn form a heterodimer of dual phosphorylated Fos and Jun or a homodimer of dual phosphorylated Jun, and together constitute AP-1. Subsequently, these dimers bind to a common DNA site, initiating the transcription of tyrosine hydroxylase (TH). The generated TH will induce the conversion of l-tyrosine to L-3,4-dihydroxyphenylalanine, which ultimately produces norepinephrine and causes anxiety.

Fig. 7.

Mechanism of extracellular regulated protein kinases (ERK)/activator protein-1 (AP-1) pathway involved in inducing anxiety in zebrafish. Ang II: angiotensin II; AT1R: angiotensin II type 1 receptor; JNK: c-Jun N-terminal kinase; FRK: Fyn-related kinase; ATF-2: activating transcription factor 2; Elk-1: Ets-like protein-1; Jun, Fos: subunits of activated protein-I; TH: tyrosine hydroxylase.

The dopaminergic nervous system is also a critical neuromodulatory system that regulates various behavioral activities. In zebrafish, dopamine neurons form on the first day after fertilization and fully develop within 4 days. Studies have examined the dopamine (DA) content in brain tissue and the expression of genes related to the dopaminergic nervous system, including DA receptors (drd1b, drd2b, drd2a, drd3 and drd4b), TH (th1 and th2), DA neurotransmitter transporter protein (dat), and so on, to evaluate the effects of drugs on the dopaminergic nervous system of the zebrafish [[111], [112], [113]]. Exposure of zebrafish to isoniazid resulted in a significant shortening of the length of zebrafish DA neurons, a reduction of fluorescence intensity, and a substantial reduction in the expression levels of th1 that were observed under fluorescence microscopy, suggesting that the DA signaling pathway may be involved in isoniazid-induced neurodevelopmental toxicity in zebrafish [111]. Quinpirole, a selective agonist of DA D2/D3 receptors, on the other hand, was found to potentially cause anxiety-like behaviors, stereotypic behavior, and memory impairment in zebrafish [114]. Venlafaxine is a pentraxin-norepinephrine reuptake inhibitor, and Su et al. [112] observed that Valbenazine exposure led to an increase in DA and Glu levels in zebrafish larvae, which in turn induced developmental toxicity in the zebrafish. Tang et al. [113] then found that zebrafish th1 and th2 mRNA abundance significantly increased. In contrast, drd1b and drd2b mRNA abundance decreased after Venlafaxine exposure, along with an increase in pentraxin levels and a decrease in DA levels, suggesting that the increase in serotonin levels seems to inhibit the function of the dopaminergic reward system and affects the courtship behavior of adult zebrafish.

The above mechanisms of drug toxicity reflect the advantages of zebrafish in applying to specific organs and systems as well as having an intact organism. These mechanisms not only help us understand the specific pathways of action of drugs in zebrafish but also facilitate the rapid determination of the toxic effects of drugs at the early stage of research to better utilize the zebrafish model for the evaluation of drug toxicity and to provide an important reference for the prediction of the potentially toxic effects of drugs on human beings, and to improve the Research efficiency.

3. Application of new technologies in toxicity evaluation

New technologies have enhanced the utility of zebrafish modeling in toxicity evaluation in three main ways: improving prediction accuracy, broadening the model's applicability, and reducing the complexity of the operation. Multi-omics has led to a more comprehensive and clearer understanding of toxicity. At the same time, rodent models and in-silico tools have made the zebrafish model more accurate and reliable in predicting toxicity. With gene editing technology, more disease models have been developed, facilitating researchers to choose appropriate disease models. Meanwhile, using transgenic fluorescent zebrafish with fluorescent probes, new imaging techniques, and automated tools has dramatically improved research efficiency. In conclusion, the application of new technologies in zebrafish toxicity evaluation is quite extensive, and the development of new technologies will make the zebrafish model more perfect and recognized by more laboratories.

3.1. Multi-model collaboration and multi-omics applications

Recently, multi-omics studies have emerged as a prevalent method for assessing safety. By employing multi-omics research, we can better understand toxic substances' harmful impacts and mechanisms. A growing focus is on enhancing the precision of zebrafish model prediction through collaborative efforts among several models. For instance, Zhang et al. [115] used high-content screening (HCS) and ultra-high performance liquid chromatography coupled with triple quadrupole tandem mass spectrometry (UHPLC-QqQ-MS/MS) to detect six cytotoxic profiles and 13 constituents of Psoraleae Fructus (PF) extracts and fully verified the hepatotoxic equivalence in primary hepatocytes, zebrafish and mice. In addition, integrating zebrafish models with computer simulation assessments has garnered significant interest. Web-based pharmacological analyses commonly provide a substantial quantity of target genes and pathways. Zebrafish models bridge the gap between web-based analyses, which generate significant data, and small-scale rodent models. Gao et al. [116] integrated the ADMET prediction platform ADMETlab 2.0 with zebrafish models. The findings from both approaches demonstrated that brucine and brucine N-oxide could cause liver injury. Furthermore, the likelihood of brucine generating hepatotoxicity was higher than brucine N-oxide. Gao et al. [117] successfully used the zebrafish model to test the pathways suggested by network pharmacology. They did this by using RT-qPCR and Western blotting analysis. The MyD88/NF-κB signaling pathway is used to help fix the liver damage that thioacetamide (TAA) causes in zebrafish. Li et al. [118] used zebrafish as a model organism to study the harmful effects of 10 halogenated aromatic disinfection by-products (DBPs). They also used molecular docking and quantitative structure-activity relationship (QSAR) modeling to anticipate potential causes of toxicity. English et al. [119] used many tools to look at the transcriptome of zebrafish that were exposed to ifosfamide. These included Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis, gene set enrichment analysis (GSEA) analysis, and Pathway Studio cellular network enrichment analysis. They discovered that the toxicity caused by ifosfamide might be linked to molecular pathways associated with cardiovascular function.

3.2. Construction of new models and application of gene editing technology

In the past, we have obtained a variety of zebrafish models, but these models still have deficiencies and limitations. In recent years, people have constantly been exploring the development of new zebrafish models to meet more needs. For example, common zebrafish kidney injury models include two methods of injection of gentamicin, causing extensive damage to renal elements and tubular destruction and infection with Edwardsiella spp. Morales Fénero et al. [120] developed a new kidney injury model using cisplatin in an attempt to minimize the impact of modeling on drug screening. Nwagbo et al. [121] developed a new model for studying very long-chain polyunsaturated fatty acids (VLC-PUFA) depletion and elongation of very long chain fatty acids protein 4 (ELOVL4)-related dysfunction. In addition, gene editing technology has been widely used in modifying zebrafish models. Currently, many kinds of fluorescence-labeled zebrafish models have been put into experimental use, in which the toxicity of drugs and the tissues and organs affected are usually judged according to the area and intensity of the labeled fluorescence. With the deepening of technology, we can establish more fluorescent labeling models or even label different zebrafish organs with different fluorescence to make the experiments more convenient. Parvez et al. [122] utilized multiplexed intermixed CRISPR droplets (MIC-Drop) to perform a large-scale reverse genetic screen in zebrafish. They utilized microfluidics to produce droplets of hundreds to thousands of genes injected by a single needle into many zebrafish embryos for high efficiency and throughput. When embryos exhibiting the target phenotype are present in these zebrafish, the identity of the disturbed gene is rapidly discovered by searching and sequencing barcodes, which avoids the need to individually rear each injected animal or deconvolute the mutant gene by genome sequencing. Recently, Kent et al. [123] established a null mutation in the zebrafish CRISPR/Cas9 her3 gene, a model expected to be useful due to the study of her3/HES3-mediated neurodevelopment and its role in cancer. Xie et al. [124] established a new transgenic zebrafish line Tg (GAcyp1a: eGFP/Luc) and used it to monitor dioxin/dioxin-like compounds (DLCs) in the environment. Naturally, we can also obtain zebrafish with different tissue or systemic defects as research models for related diseases, and these models will provide the basis for related studies such as drug interactions and pharmacokinetics. In the future, zebrafish as a class of non-rodent experimental animal models will continue to be of interest, and there will be an increasing variety of models available for experimentalists to select and use.

3.3. New methods and techniques based on oxidative stress and ER stress

In toxicity studies, it is important to have appropriate methods to monitor or visualize the onset of toxicity, as this not only helps the researcher to identify the mechanism of toxicity more quickly but also improves the efficiency of research and development. As shown in Fig. 8, emerging proteomics techniques based on zebrafish models have shown significant potential in studying oxidative and ER stress. Proteomics, due to the transparent and rapid development of zebrafish embryos, can directly observe the dynamic changes of protein expression, especially for studying protein interaction and modification during early development. Therefore, it can be applied to proteomics analysis under different stress conditions to reveal which proteins are responsive to specific stress responses and explore the potential effects of these changes on biological phenotypes [125]. In addition, the fluorescein ubiquitinated cell cycle indicator (LUCCI) is a new experimental method developed in zebrafish, which fuses the nano-luciferase (NLuc) with the first 110 amino acid sequences of zebrafish geminin under the zebrafish insulin promoter, so that the cells produce a fluorescence signal during the S/G2/M phase. When zebrafish is exposed to fluorescein substrates, NLuc releases light signals, and the intensity reflects the degree of β cell proliferation to achieve chemical screening in the process of β cell regeneration [32].

Fig. 8.

New techniques based on oxidative stress and endoplasmic reticulum stress. (A) Proteomic processes and techniques that can be used in zebrafish. (B) A newly developed experimental method for β cell screening in zebrafish-fluorescein ubiquitination cell cycle indication (LUCCI). (C) Different imaging techniques have been developed based on different response ions, and their main reaction mechanisms are as follows: ClO⁻: (a) NPCC two-photon fluorescent probe imaging. It can be used to monitor the changes in the production of the three ions during stress accordingly. (b) Ratiometric fluorescent probes Rx-NE and Rx-NCE. (c) Ratiometric Fluorescence AIE/AIEE/AIRE Active Probes. O₂⁻: (d) Two-photon fluorescent probe for imaging endogenous superoxide anions in living cells and zebrafish. ONOO⁻: (e) Hyperbranched polysiloxane probes. (f) Mitochondria-endoplasmic reticulum dual-targeted red-emitting fluorescent probe, a novel red-emitting fluorescent probe. NLuc: nano-luciferase; GSH: glutathione; HOCl: hypochlorous acid; gmnn: geminin; ELISA: enzyme-linked immunosorbent assay; ICAT: isotope-coded affinity tag; SILAC: stable isotope labeling with amino acids in cell culture; ITRAQ: isobaric tags for relative and absolute quantitation; ER: endoplasmic reticulum; MCSA: mitochondria-endoplasmic reticulum dual-targeted red-emitting fluorescent probe; AIE/AIEE/AIRE: aggregation-induced emission/aggregation-induced emission enhancement/aggregation-induced ratiometric emission; Mito: mitochondria.

In addition to proteomics techniques and new experimental methods, targeted probe bioimaging techniques for ER in zebrafish have also been continuously developed. Different imaging techniques have been developed based on different response ions (such as ClO⁻, ONOO⁻, O₂⁻). Hypochlorite ion-based two-photon fluorescent probe (TP probe) for ER reduction stress imaging. The new probe NPCC has the potential to distinguish normal cells from cancer cells, especially by monitoring the levels of GSH, cysteine (Cys), and hypochlorous acid (HOCl). The effectiveness and selectivity of the NPCC (a new chemical structure combining coumarin and pyrazoline derivatives) probe were verified by cell and zebrafish experiments, which provided a new tool for future research on the synergistic effect of reductive stress with ROS and reactive sulfur (RSS) [126]. However, most probes usually have the aggregation-induced quenching (ACQ) problem, which reduces their emissivity. The ratiometric fluorescent aggregation-induced emission/aggregation-induced emission enhancement/aggregation-induced ratiometric emission (AIE/AIEE/AIRE) active probe can resist environmental interference through the built-in two-emission correction function, showing specific localization of ClO⁻ in ER for ratiometric detection of ClO⁻ in living cell ER. A benzothiazole-containing salicylaldehyde acetyl hydrazone AIE probe modified by a p-toluenesulfonamide molecule as an ER positioning group. The probe can be used for imaging ClO⁻ in living cells and zebrafish ER [127]. The novel ER-targeted ratiometric fluorescent probes Rx-NE and Rx-NCE that can activate ClO⁻ are used to image foam cells under ER pressure and have excellent optical properties for classical rhodamine dyes and coumarin dye bridge groups. The Rx-NCE probe innovatively tracks the atherosclerotic blood vessels of atherosclerotic transgenic (tg) (flk1: eGFP) zebrafish. The probe Rx-NCE is of great value in studying the pathological features of ER stress and atherosclerotic diseases [128]. A hyperbranched polysiloxane probe based on ONOO⁻ imaging has good biosafety and ideal photostability, which can accurately locate the ER. Si–Er–ONOO made an effective fluorescence response to the fluctuation of ONOO⁻ concentration in ER and zebrafish [129]. The mitochondria-ER dual-targeted red-emitting fluorescent probe (MCSA) is a novel red-emitting fluorescent probe for imaging and detecting ONOO⁻ in living cells and zebrafish. This sensor enables visual detection of foreign and endogenous ONOO⁻ in vivo in living cells and zebrafish [130]. The ER-specific two-photon fluorescent probe (ER-Rs) based on O₂⁻ imaging was used to image endogenous superoxide anion (O₂⁻) in living cells and zebrafish. The ER-Rs used near-infrared light as an excitation source. It has the advantages of less light damage to biological samples, weak autofluorescence interference, and deep tissue penetration ability, which fills the gap in the application of ER-specific two-photon probes in living cells and animals and provides new tools and methods for studying the role of superoxide anions in physiological and pathological processes [131].

3.4. New imaging techniques and automation in zebrafish modeling

A female zebrafish can lay about one hundred eggs at a time, which provides the possibility of high throughput for experiments. However, the large amount of imaging and analysis work required for morphological observations, vascular and neural development, tissue sections, etc., also makes high throughput difficult, so having a more convenient imaging and analysis system seems imminent. The combination of computer and artificial intelligence with zebrafish modeling holds promise. Toxicopathology is one of the most important methods in nonclinical toxicology testing of drugs and is an important part of hazard identification and risk assessment. High-throughput analysis using zebrafish often accompanies extensive image processing during toxicity evaluation. Researchers have used computers to perform behavioral analysis and morphological analyses. For example, they connect video cameras to computers to easily obtain behavioral tracks of zebrafish or use computers to observe morphological changes in the internal organs of zebrafish, calculating fluorescent areas and intensities of the liver, heart, kidneys, and so on. In contrast, the production and reading of tissue sections during histopathological evaluation is often time-consuming and cumbersome, and the standard measurements vary from operator to operator. Suppose artificial intelligence (AI) systems can be fully utilized. In that case, the time of non-clinical toxicology tests can be shortened to a certain extent, and the drug development process can be accelerated [132].

Chen et al. [133] developed a 3D reconstruction pipeline based on multi-view images, and they utilized the acousto-hydrodynamic rotary tweezers (ART) system to perform high-speed morphophenotyping of zebrafish larvae in a non-contact manner, reflecting the morphology and visceral changes of zebrafish before and after the dosing of drugs. Westhoff et al. [134] developed an automated imaging pipeline for analyzing the kidneys of zebrafish larvae, and Kotiyal et al. [135] demonstrated a method to rapidly and accurately quantify the size of zebrafish livers.

In the cardiovascular system, Yang et al. [136] reported a light-sheet-flow imaging system (LS-FIS) based on light-sheet illumination and a continuous-flow imager. The automated loading and dispensing of LS-FIS allowed 3D imaging of whole larvae from dozens of zebrafish embryos within half an hour. They then used LS-FIS to study the vascular development process in 3–9 dpf (days post-fertilization) zebrafish by running about 50 transgenic fluorescent zebrafish embryos labeled with EGFP (enhanced green fluorescent protein) in the vascular system through LS-FIS once a day. The results provided clear images of vascular development and data on the length and number of blood vessels visualized in zebrafish. They then introduced an improved 3D U-Net (MS-3D U-Net) to more accurately segment zebrafish intersegmental and dorsal longitudinal anastomotic vessels in 3D photomicrograph images [137]. Naderi et al. [138] developed and validated a zebrafish automatic cardiovascular assessment framework (ZACAF) based on a U-net deep learning model for automated assessment of cardiovascular indices. Chen et al. [139] established an FE-Unet network to obtain continuous and columnar vascular structures in transgenic zebrafish lines by comparing angiographic images, which allowed image threshold segmentation to be closer to the results of manual segmentation, thus accurately and automatically extracting the topological parameters of the cerebral vascular network in developing zebrafish.

J.W. Kenney et al. [140] created the first fully segmented 3D digital adult zebrafish brain atlas (AZBA) for neurological and behavioral analysis. Colón-Rodríguez et al. [141] developed a high-throughput phenotyping solution for automated imaging (VAST) and behavioral assays (DanioVision), making this platform not only fast for the detection of morphologic defects but also greatly facilitating behavioral analyses and extremely valuable for application. For imaging the immune system, Kaveh et al. [142] established an in vivo imaging technique that can comprehensively map immune cell migration and analyze the immune cell migration pattern of zebrafish larvae in detail, which has a certain practical value. In addition, Katz et al. [143] combined micro-computed tomography (micro-CT) with a novel application of ionic silver staining, which resulted in images of melanin distribution throughout zebrafish larvae and facilitated computational analysis of their melanin content and morphology. Recently, Tang et al. [144] designed a fish capsule (FC) system by combining automated zebrafish pod technology and droplet microarray strategy. Its use of a microfluidic chip, controlled by hydrodynamics, allows for the automatic and anesthesia-free loading, trapping, orientation and fixation of hundreds of fish larvae. This system can bridge the gap between in vivo large-scale screening and resource-limited individual laboratories and is extremely helpful in accelerating drug discovery and pharmacological studies. Fig. 9 is a simple diagram of the process.

Fig. 9.

Schematic diagram of the rapid generation of zebrafish capsules for organ-specific microscopic imaging. PDMS: polydimethylsiloxane.

4. Concluding remarks and future perspectives

Zebrafish have become an important tool for assessing drug toxicity, screening and preclinical testing because they are genetically similar to those of higher vertebrates, have well-preserved physiological pathways, and can perform large-scale experiments and quickly accumulate large sample data. Its feasibility in toxicity mechanism studies has been recognized by an increasing number of laboratories. From cardiovascular to hepatic and neurological systems, it can be said that zebrafish sensitivity to drug toxicity is considerable. Here, we focused on oxidative stress, ER stress, inflammatory response, impaired lipid metabolism and abnormal neurotransmitter expression. In general, oxidative stress, ER stress and inflammatory responses are upstream of apoptosis, and they can crosstalk with each other to jointly regulate cellular activity tendencies. They are also the cause of drug toxicity frequently reported in studies. Abnormalities in lipid metabolism and abnormalities in the expression of neurotransmitters, on the other hand, are alterations in life processes at the macroscopic level of living organisms. Drugs that cause abnormalities in lipid metabolism usually cause toxicity in the liver, while abnormalities in neurotransmitter expression are responsible for drug-induced neurotoxicity, such as depression and anxiety.

As various technologies continue to be iterated, they will be more closely linked and compatible with the zebrafish model, which will help the zebrafish model improve its performance and facilitate research. Updates in imaging systems and automated devices and the development of new fluorescent probes and biosensors are more favorable for obtaining sharper imaging and for more precise detection of specific biological processes, such as protein-protein interactions, ion fluxes, and metabolic changes. Zebrafish provide an excellent platform for multi-omics studies, enabling researchers to reveal drug toxicity more comprehensively and in-depth. The zebrafish model is an ideal transition from in-silico tools to rodent models. The optimal balance of modeling throughput and reliability makes this an invaluable tool for multi-omics studies. Because of this, it can also be used with many different models to improve experimental accuracy and comprehensiveness. Gene editing technology has enabled the application of zebrafish models in personalized and precision medicine. Research can establish zebrafish models with specific genetic variants to emulate personalized drug responses across diverse populations or develop various zebrafish models with different genetic backgrounds and developmental stages to more accurately mimic individual differences and toxicity responses in humans. This approach is instrumental in advancing precision medicine and tailored pharmaceutical services.

Although zebrafish have the advantage of a relatively short growth cycle and rapid tissue development, differences in fish age, exposure time, surface area of the well plate, and solution dissolved oxygen all contribute to discrepancies in the resulting analysis. Also, while zebrafish genes significantly resemble human genes, around 13% of the genome displays variations [3]. These variances could potentially influence the outcomes of safety assessments and contribute to inconsistencies between experimental and clinical findings. For instance, zebrafish erythrocytes possess a nucleus [145]. A single ventricle and atrium, with no pulmonary circulation, characterize the zebrafish heart. Regarding ocular toxicity, the structure of zebrafish and human retinas is comparable. However, the zebrafish retina can regenerate [146]. With regard to the gastrointestinal system, it is noteworthy that zebrafish lack a stomach and that the pH of their gastric organ is neutral. Additionally, zebrafish have no specific gastrointestinal structures or cell types [147]. In the context of kidney toxicity, the kidneys of mammals are more intricate. Hentschel et al. [148] discovered that when puromycin was administered to zebrafish larvae, it had unintended effects on the central blood flow, unlike in rodent models. This could potentially result in severe episodes of fluid accumulation, known as oedema.

Furthermore, the unique method of administering drugs to zebrafish hinders the efficient screening of drugs that are not soluble in water. The type of solvent used can also affect the model's accuracy. It has been proposed that a concentration of less than 1% dimethyl sulfoxide (DMSO) does not affect zebrafish development [149]. Exposing zebrafish to easily broken down and evaporated drugs can result in differing administered drug concentrations from the intended concentration. Similarly, if certain drugs dissolve in water and produce darker hues, this can pose challenges to the experiment, limiting the range of applications for the zebrafish model. Ultimately, it is imperative to establish a standardized criterion for assessing the toxicity of zebrafish models. This calls for a more thorough standard that covers how to breed and feed zebrafish, how old the fish is, choosing the proper solvent, and even using anesthetics in lab settings. Future research must delve deeply into the question of which kind of toxicity assessment is better suited to the zebrafish model and which assessments are not adequately addressed by this model. It is also necessary to establish which zebrafish model and solution criteria should be employed in the context of disparate toxicity evaluations and to identify and address any potential interferences and issues associated with this choice. This will ensure that the full potential of zebrafish in toxicity evaluations is realized.

In summary, the application of zebrafish in the study of drug toxicity mechanisms has become more and more mature, but the shortcomings of the model behind the high throughput should not be ignored. Fortunately, there have been more and more new techniques to improve the model. Therefore, the great application value of the zebrafish model in toxicity mechanism research still deserves our continuous investment.

CRediT authorship contribution statement

Wenhao Wang: Writing – original draft, Visualization, Conceptualization. Xuan Gao: Writing – original draft, Visualization. Lin Liu: Writing – original draft, Visualization. Sheng Guo: Writing – review & editing, Funding acquisition. Jin-ao Duan: Supervision, Resources, Funding acquisition. Ping Xiao: Writing – review & editing, Supervision, Resources, Funding acquisition.

Declaration of competing interest

The authors declare that there are no conflicts of interest.