Review on Toxicity Effect of Emerging Contaminants on Trans-/Multi-Generational Fish

Abstract

In recent years, toxicological studies on fish exposed to emerging contaminants (ECs) have been relatively in-depth. However, research on trans-/multi-generational exposure of fish to ECs remains scarce. Limited data indicate that when parental generations (P) are subjected to ECs stress, it can threaten the normal transmission of functions in offspring, such as growth and development, reproduction, physiology, endocrine, neural, and behavioral functions. Even after the exposure is terminated, these negative impacts may persist. Under the long-term presence of ECs, the health of fish offspring may affect the survival of entire populations and the stability of ecosystems. Therefore, this review summarizes studies on trans-/multi-generational effects of ECs on fish and analyzes these research results. Based on the materials collected, more research on trans-/multi-generational ECs effects on fish is urgently needed, especially regarding the F3 generation, combined toxicity, and trans-/multi-generational epigenetic effects. This will enable a comprehensive assessment of the health and ecological risks posed by ECs at environmental concentrations to fish.

1. Introduction

With the rapid globalization of the economy and accelerated urbanization, the demand for synthetic materials in industry and agriculture has increased exponentially. This trend, coupled with relatively relaxed environmental policies, has led to widespread pollution across various ecosystems in the world. In particular, emerging contaminants (ECs) pose critical threats to the health of coastal, estuarine, and marine ecosystems, as well as human beings. ECs are anthropogenic pollutants recently detected in the environment or newly recognized for their adverse effects, and they remain unregulated or inadequately addressed by current laws and standards (Figure 1) [1]. Moreover, environmentally relevant concentrations of ECs can have detrimental impacts on the biota and human health [2]. This category includes Pharmaceutical and Personal Care Products (PPCPs), Micro/Nano Particles, Endocrine Disrupting Compounds (EDCs), Industrial substances, and other synthetic chemicals of emerging concern (Figure 1). Many ECs exhibit concerning properties including bioaccumulation, environmental persistence, reproductive toxicity, and genotoxicity, often at trace concentrations [3,4,5]. These properties not only impact exposed individual organisms but also transfer health risks across generations via food chain dynamics, potentially triggering cascading ecological and public health crises.

ECs are widely distributed in ecosystems. Some of these contaminants are bioaccumulative and can persist for a long time, meaning their toxic effects may continuously interfere with the growth and reproduction of organisms over an extended period [6,7]. Studies have shown that even after exposure ceases, the residual toxicity of EDCs may still trigger long-term physiological changes in organisms, impairing the health and genetic stability of offspring [8,9,10]. What is more challenging is that ECs have complex and diverse chemical compositions, and these substances exhibit multi-dimensional toxic mechanisms of action, making monitoring and governance efforts extremely difficult [11,12]. Therefore, in-depth research on the characteristics of ECs, their toxic effects on organisms, and their mechanisms of action holds urgent scientific value and practical significance.

A critical limitation in current ecotoxicological assessment is the disconnect between laboratory studies and environmental reality [13]. Most toxicological experiments employ concentrations far exceeding environmental levels and focus on short-term exposures, potentially underestimating the true long-term risks of ECs at environmentally relevant concentrations [14,15]. As a result, current regulatory frameworks, which typically assess single compounds, may underestimate the toxic impacts of environmentally relevant concentrations of ECs on aquatic organisms and humans [16]. Persistent trans-/multi-generational pollutant exposure effectively addresses these problems.

Transgenerational and multigenerational exposure studies address this gap by examining continuous, long-term effects across multiple generations at environmentally relevant concentrations (Figure 2) [14,17,18]. It allows extrapolation to humans, helping identify potential hazards of such pollutants to human health. However, transgenerational studies remain severely limited due to their complexity, time-consuming (often >6 months), and numerous methodological challenges. The accumulation of technical variables and potential experimental disruptions over such extended periods increases the risk of confounding factors, resulting in a severe lack of reliable multigenerational data in this field.

ECs due to their environmental persistence and bioaccumulative properties, can accumulate in fish over time even under low concentration exposure. This accumulation eventually reaches toxic levels. Over time, they cause persistent damage to fish physiology, growth, reproduction, and behavior. These effects may also be transgenerationally transmissible. Therefore, this review (1) synthesizes all available transgenerational toxicity data for major EC categories, (2) analyzes mechanisms underlying transgenerational inheritance, (3) identifies critical data gaps, and (4) discusses emerging methodologies that may accelerate research.

2. The Toxic Action of Typical ECs

We will briefly interpret the toxic effects and action mechanisms of emerging pollutants on marine organisms in recent years and discuss them from the aspects of multigenerational, transgenerational, and epigenetic inheritance. While ECs encompass a broad range of substances (Figure 1), transgenerational studies have primarily concentrated on specific categories. The following sections examine those ECs with robust trans-/multi-generational datasets in fish, including micro/nano plastics (MNPs), PPCPs and EDCs. Additionally, it is critical to distinguish between multigenerational and truly transgenerational effects. The F1 generation represents offspring directly exposed to contaminants during their development (either as embryos or as primordial germ cells in the exposed mother). The F2 generation may also experience direct exposure as primordial germ cells during F1 fetal development. Therefore, effects observed in F1 and F2 do not constitute true transgenerational inheritance, as these generations were potentially exposed during critical developmental windows. Only effects observed in the F3 generation and beyond, which were never directly exposed at any developmental stage, represent genuine transgenerational epigenetic inheritance [19,20]. Throughout this review, we use ‘multigenerational’ for studies examining F1–F2 generations and ‘transgenerational’ only for studies extending to F3 or later generations.

2.1. Micro/Nano Plastics (MNPs)

Over recent decades, the growing production and use of plastics have led to increasing accumulation of microplastics and nanoplastics (MNPs) in natural environments, creating a severe environmental problem. In aquatic environments, the concentration of MNPs can range from several micrograms to several milligrams per liter, and they are extremely widely distributed [21,22]. The ocean is widely recognized as a major sink for these pollutants [23,24]. When ingested by marine organisms, MNPs cause significant harm to metabolism, growth, development, and reproduction, with effects potentially cascading to population-level impacts [25,26,27,28].

Early microplastic research revealed that MNP-induced mortality in marine organisms primarily resulted from physical damage. Specifically, excessive ingestion of MNPs caused blockages in the esophagus, stomach, and intestines, leading to death. This mortality was not attributed to the chemical properties of MNPs [29,30,31,32]. However, advances in understanding have revealed additional toxicity mechanisms beyond physical obstruction. Microplastics (MPs) and nanoplastics (NPs) can accumulate in fish gills and other internal tissues [33]. Due to their small size, they can also cross the gastrointestinal barrier and ultimately enter fish brains via the blood–brain barrier [34,35]. Furthermore, recent studies have found that MNPs easily interact with organic pollutants, heavy metals, pharmaceuticals, and other substances due to their large specific surface area and strong lipophilicity, acting as carriers that enhanced their toxic effects on marine organisms [28].

Microplastics demonstrate multigenerational effects in fish, though true transgenerational (F3+) data remain scarce. Limited data indicate that when parental zebrafish are exposed to nanoplastics, their offspring embryos show reduced hatching rates and increased malformation rates. This effect can persist into the F2 generation. Notably, microplastics often enhance the toxicity of other pollutants through carrier effects. For example, one study found that zebrafish offspring exhibited transgenerational lipid metabolism disorders after parental combined exposure to bisphenol A (BPA) and NPs [36], suggesting that microplastics may act as vectors amplifying co-contaminant toxicity across generations. However, the mechanisms underlying this enhancement remain unclear, as studies have not determined whether microplastics increase bioavailability, alter metabolic pathways, or act through independent additive toxicity.

Furthermore, studies have found that microplastics at environmental concentrations can affect the growth, lipid storage, and external coloration of adult fish. At low-concentrations (100 particles·L⁻¹), environmentally derived microplastics exert further endocrine-disrupting effects on parental fish and their offspring, resulting in delayed onset of spawning, reduced egg survival rate, and a higher malformation rate in the offspring [37].

Methylation changes provide potential mechanisms for these multigenerational effects. Wade et al. reported a microplastic exposure experiment on two generations of fathead minnows. The results showed that methylation was more significant in parental males than in parental females and F1 offspring. This sex-specific difference is unexplained and warrants investigation. Meanwhile, in an experiment where only the maternal generation was exposed (and offspring unexposed), researchers found that changes in Differentially Methylated Loci (DMLs) and Differentially Methylated Regions (DMRs) persisted in offspring. This led to cross-generational effects that interfered with the offspring’s gene transcription and molecular processes [38]. However, the study examined only F1 offspring, so whether these methylation changes represent true transgenerational inheritance (persisting to F3+) or transient F1 effects remains unknown.

Environmental context critically modulates microplastic toxicity, as demonstrated in marine medaka multigenerational studies. Microplastics can also reduce the growth performance of F1-generation marine medaka. Under high salinity conditions, MPs increase the MDA content and decrease the antioxidant capacity of F2 larvae. Cortisol level of F1-generation larvae decreases while that of F2-generation larvae increases. In contrast, under low salinity conditions, both the MDA content and antioxidant capacity of F1 and F2 larvae increase. The mRNA expression levels of genes in the NF-E2-related factor 2 (Nrf2) pathway are dysregulated. This phenomenon is associated with the transcriptional regulation of genes in the hypothalamic-pituitary-interrenal (HPI) axis [39]. These opposing patterns suggest that environmental context critically modulates microplastic effects, but the lack of mechanistic investigation prevents determining whether salinity alters microplastic bioavailability, particle aggregation, or physiological stress responses.

The field faces critical data gaps: (1) few of studies extend beyond F2, preventing confirmation of true transgenerational inheritance; (2) mechanistic endpoints (specific gene targets, pathway analysis) are rarely measured alongside phenotypic effects, limiting causal interpretation; (3) dose–response relationships are poorly characterized, with most studies testing single concentrations that may not reflect environmental exposure.

2.2. Pharmaceutical and Personal Care Products (PPCPs)

While microplastics represent physical pollutants with associated chemical toxicity, pharmaceuticals and personal care products constitute a distinct class of emerging contaminants designed for biological activity, raising different concerns for multigenerational fish toxicity. PPCPs represent a diverse group of chemicals widely applied in human healthcare and personal care, including antibiotics, analgesics, hormones, and preservatives. Global PPCP usage has increased annually, driven by population growth and improved living standards worldwide. A significant fraction of incompletely degraded PPCPs enters aquatic ecosystems through pathways such as domestic sewage, hospital effluents, and livestock breeding wastewater [40,41]. PPCPs have been detected in surface waters globally at concentrations ranging from nanograms per liter (ng·L⁻¹) to micrograms per liter (μg·L⁻¹) [42,43]. Their environmental persistence varies based on chemical structure, aqueous solubility, biodegradability, and ambient conditions. Regional differences in public health policies further contribute to variations in PPCP composition and concentrations [41,44]. Traditional wastewater treatment systems cannot achieve complete PPCP removal, resulting in their progressive accumulation in water columns, sediments, and aquatic biota [45,46,47].

PPCPs can disrupt fish endocrine systems, impairing key physiological processes including growth, development, reproduction, and behavior [48,49]. Endocrine disrupting compounds in PPCPs induce specific adverse effects such as sex reversal, reproductive organ malformations, and reduced fecundity in fish [50]. Antibiotic-containing PPCPs interfere with fish immune function and associated microbial communities, increasing susceptibility to pathogens while promoting the emergence and dissemination of antibiotic resistance genes [51,52]. Certain PPCPs also cause structural and functional damage to critical organs including the nervous system, cardiovascular system, and liver, compromising overall physiological homeostasis in fish [53].

Recent years have witnessed growing scholarly attention to long-term and multi-generational fish exposure studies on PPCPs [54,55]. Existing research confirms that PPCPs exposure induces multi-generational toxic effects in offspring, though mechanistic understanding remains limited. For instance, parental zebrafish exposed to low concentrations of carbamazepine produce offspring with significant alterations in growth, development, and behavioral effects that persist across generations [55,56]. Similarly, parental zebrafish exposed to antibiotic mixtures (15 commonly detected compounds) at nominal concentrations of 1 and 100 μg·L⁻¹ exhibit transgenerational transfer of these antibiotics, resulting in cardiotoxicity in offspring [57]. However, these studies do not specify whether effects extend to F3 generation, leaving uncertain whether observed impacts represent true transgenerational inheritance or transient multigenerational responses to direct exposure during germ cell development.

Notably, current research on PPCP-induced trans-/multi-generational toxicity remains limited, with most studies focusing on single-compound exposures. The toxic effects and underlying mechanisms of multi-compound PPCP exposures in complex environmental settings require further systematic investigation, including F3+ generation assessment to confirm true transgenerational inheritance, mechanistic endpoints (methylation profiling, histone modifications, miRNA expression) paired with phenotypic measurements, environmentally realistic mixture compositions and concentrations, comparative studies across fish species to identify susceptibility determinants and recovery experiments to determine whether effects are reversible or permanent.

2.3. Endocrine Disrupting Chemicals (EDCs)

The health hazards of EDCs to organisms have been widely recognized. Meanwhile, EDCs have been detected in multiple environmental media, including water, soil, aerosols, the atmosphere, and living organisms [58]. Due to their high hazards (such as potent toxicity, teratogenicity, carcinogenicity, mutagenicity, etc.), high environmental persistence, and bioaccumulation, EDCs are regarded as a typical type of ECs [59]. In recent years, studies on the effects of EDCs on trans-/multi-generational fish have been increasing [60]. Therefore, understanding the trans-/multi-generational toxic effects and mechanisms of EDCs is of great significance for a comprehensively assessing the ecological risks of endocrine disruption.

EDCs can cause disorders in the reproductive, developmental, and endocrine systems of fish [61,62]. In trans-/multi-generational studies, it was found that when female viviparous eelpout (Zoarces viviparus) were exposed to octylphenol (OP) at concentrations of 25 and 100 μg·L⁻¹ for 17 days or 35 days, the expression levels of estrogen receptors and vitellogenin in F1-generation embryos increased significantly. Meanwhile, intersex phenomena occurred during the process of gonadal differentiation [63]. Similar studies have also found that nonylphenol (NP) at concentrations of 20 and 200 μg·L⁻¹ can inhibited the growth of zebrafish and caused gonadal damage [6]. Meanwhile, gnrh2 and gnrh3 were upregulated, and fshβ and lhβ were downregulated in the hypothalamic-pituitary-gonadal axis (HPG). For female zebrafish, the expression of fshβ and lhβ was upregulated in the parental (P) generation and F1 generation but showed a downward trend in the F2 generation. In male zebrafish, the expression of the cyp19a1a gene in the gonads was downregulated, while the expression of fshr, lhr, and esr genes in females showed a decreasing trend. Compared with the P generation, the F2 generation exhibited stronger tolerance to high concentrations of NP (20 and 200 μg·L⁻¹), but higher sensitivity to low-concentration NP (2 μg·L⁻¹) [6].

EDCs can cause malformations and induce the development of cancer in trans-/multi-generational fish. In inland silverside (Menidia beryllina), early-life exposure to 3–10 ng·L⁻¹ EDCs (bifenthrin, levonorgestrel, ethinylestradiol, trenbolone) caused significant trans-/multi-generational (F0–F2) DNA methylation differences (via reduced representation bisulfite sequencing, RRBS) in promoters/genes, enriched key pathways, inherited EDC-responsive genes (EDCRG) methylation, showing low EDCs alter epigenetics across unexposed generations [64]. These results illustrate the potential reasons why fish may develop teratogenicity and carcinogenicity upon exposure to EDCs. This represents the most mechanistically rigorous EDC multigenerational study to date, directly linking exposure to heritable epigenetic marks. However, even this study examined only F0–F2 generations, leaving uncertain whether methylation patterns persist to F3 (true transgenerational inheritance) or dissipate after removal of direct germ cell exposure.

Major knowledge gaps persist: (1) No EDC studies extend to F3+ to confirm transgenerational versus multigenerational effects; (2) the mechanistic basis for generation-specific gene expression reversals (upregulation in F1, downregulation in F2) remains unexplored; (3) non-monotonic dose–response patterns across generations are documented but unexplained, requiring investigation of whether different concentration ranges trigger distinct molecular pathways; (4) sex-specific differences in multigenerational susceptibility are observed but not mechanistically characterized—do males and females show differential epigenetic reprogramming, EDC metabolism, or germline vulnerability?

2.4. Mixture Effects

Real-world aquatic organisms face simultaneous exposure to complex contaminant mixtures, yet multigenerational toxicity research has focused almost exclusively on single-compound exposures (Table 1). The limited mixture studies reveal critical knowledge gaps regarding synergistic and antagonistic interactions across generations.

Evidence for mixture enhancement of multigenerational effects is emerging but mechanistically unexplored. Nanoplastics combined with BPA produced F1–F2 lipid metabolism disorders in zebrafish [40], suggesting microplastics may act as vectors amplifying co-contaminant transgenerational toxicity. Similarly, environmental microplastics (100 particles·L⁻¹) enhanced endocrine-disrupting effects in exposed parents and offspring beyond what would be expected from microplastics alone [65]. The antibiotic mixture study [61] documented direct transgenerational transfer of 15 pharmaceutical compounds from exposed parents to embryos, causing F1 cardiotoxicity. In inland silverside, early-life exposure to an EDC mixture (bifenthrin, levonorgestrel, ethinylestradiol, trenbolone at 3–10 ng·L⁻¹) produced DNA methylation changes persisting through F0–F2 generations [64].

However, neither study formally tested whether effects were synergistic (greater than additive), antagonistic (less than additive), or simply additive, limiting interpretation of interaction mechanisms. Most of studies examined only two chemical classes simultaneously, far fewer than the dozens of contaminants co-occurring in real environments. Whether low-dose environmental mixtures produce different multigenerational toxicity patterns than the high-dose single compounds dominating current research remains unknown. Most critically, no studies extend beyond F2 generation, leaving uncertain whether mixture exposures cause true transgenerational (F3+) effects or only multigenerational responses. The near-absence of mixture multigenerational data represents a critical gap given that wastewater effluents contain 50+ pharmaceutical compounds, urban runoff combines pesticides and microplastics, and agricultural areas expose fish to complex pesticide cocktails.

Table 1. The endpoints and effects of ECs on fish across trans-/multi-generational exposures.

ECs	Fish Species	Exposure Dose	Exposure Duration	Generations	Endpoint in Offspring	Reference
Single ECs on trans-/multi-generational fish
Microplastics	fathead minnows (Pimephales promelas)	100, 2000 particles·L−1 (150–500 μm)	P: 178 dpf F1: 12 dpf	P, F1	P: Affect the growth, lipid storage, and external coloration. F1: In the low-concentration treatment (100 particles·L−1), microplastics can delayed onset of spawning, reduced egg survival rate, and a higher malformation rate in the offspring.	[37]
	fathead minnows (Pimephales promelas)	100, 2000 particles·L−1 (150–500 μm)	P: 178 dpf F1: 12 dpf	P, F1	F1: Methylation was more significant in parental males than in parental females and F1 offspring.	[38]
Carbamazepine	Zebrafish (Danio rerio)	10 μg/L	P: 6 weeks	P, F1, F2, F3, F4	P: Reproductive output, courtship and aggressive behaviors, 11-ketotestosterone (11 KT), and sperm morphology decreased. F1-F4 (only paternal fish exposed): 11 KT, reproductive output, altered courtship, aggression, and sperm morphology decreased.	[55]
OP	viviparous eelpout (Zoarces viviparus)	25, 100 μg·L−1	P: 17 and 35 dpf	P, F1	F1: the mortality of embryos increased, and the body length and weight decreased; the expression levels of estrogen receptors and vitellogenin of F1 embryos were increased; abnormal gonadal differentiation occurred.	[63]
PFOS	Zebrafish (Danio rerio)	5, 50, 250 μg·L−1	P: 150 dpf F1: 7 dpf	P, F1	F1: The mortality rate of F1 larvae was 100% at 250 μg·L−1; the locomotor activity of F1 larvae were abnormally enhanced in 5 μg·L−1.	[66]
BPA	rare minnow (Gobiocypris rarus)	15, 225 μg·L−1	P: 21 dpf	P, F1	P: induced poor quality of the embryos. F1: The fertilization, hatching, and spontaneous coiling rates decreased; the malformation rate increased; ossification in craniofacial cartilage were delayed; embryonic heart rate increased.	[60]
NP	Zebrafish (Danio rerio)	2, 20, 200 μg·L−1	P: 32 dpf F1: 120 dpf F2: 140 dpf F3: 7 dpf (recovery)	P, F1, F2, F3	P: the expression of fshβ and lhβ were upregulated; F1: inhibit the growth of zebrafish and cause gonadal damage. the expression of fshβ and lhβ were upregulated. fshr, lhr, and esr were decreased. F2: inhibit the growth of zebrafish and cause gonadal damage. exhibits stronger tolerance to high concentrations of NP (20 and 200 μg·L−1), but higher sensitivity to low-concentration NP (2 μg·L−1) F3: Fertility ratio and hatching ratio were decreased.	[6]
Combined ECs on trans-/multi-generational fish
Microplastics and EE2	fathead minnows (Pimephales promelas)	MPs, MP + EE2 (10 ng·L−1), MP + EE2 (50 ng·L−1)	P: 44 dpf F1: 14 or 21 dpf	P, F1	F1: reduced embryos locomotor activity; stronger swimming ability.	[65]
Microplastics and salinity	marine medaka (Oryzias melastigma)	180 μg·L−1 polystyrene ( 5.0 μm), 5‰ and 25‰ salinity	F1: 25 dph F2: 120 dph (recovery)	F0 (unexposed), F1, F2 (recovery)	F1: high salinity treatment increases in MDA content, TAC, CAT activity and cortisol levels. Additionally, also upregulated the mRNA levels of antioxidant genes (nrf2, keap1a, keap1b, sod1, cat, mt2) and HPI axis genes (crhr1, crhr2, pomca, mc2r, cyp11c1). Reduced the growth performance.	[39]
Nanoplastics and bisphenol	Zebrafish (Danio rerio)	BPA 20 μg·L−1, NPs 1 mg·L−1, NPs 1 mg·L−1 + BPA 20 μg·L−1 (N + B)	F0: 28 dpf F1: 4 months (recovery)	F0, F1	F0: N + B groups induced the gut damage and microbiota dysbiosis; Differentially expressed metabolites were mainly enriched in purine metabolism, mTOR, FoxO pathways, and all enriched in mTOR in control/NPs/N + B and control/BPA/N + B groups. The N + B group had the most such metabolites; nanoparticle + BPA co-exposure exacerbated F0 zebrafish liver metabolic disturbance, with mTOR possibly the mechanism. F1: Dead egg rate was increased at NPs, and N + B groups; the TC and TG level increased at N + B groups.	[36]
Bifenthrin, Levonorgestrel, ethinylestradiolt, renbolone	inland silverside (Menidia beryllina)	bifenthrin (3.02 ng·L−1), EE2 (6.79 ng·L−1), levonorgestrel ( 9.27 ng·L−1), trenbolone ( 9.60 ng·L−1).	F0: 8–21 dph F1: unexposed F2: unexposed	F0, F1, F2	F0, F1, F2: Through RRBS, which were attributed to strict inheritance of DNA methylation alterations and dysregulation of epigenetic control mechanisms. differential methylation was more frequent in gene bodies than in promoter regions across all treatments and generations	[64]
Antibiotic	CAM, CIP, CTC, EYR, ENR, LCM, NOR, OFL, OTC, ROX, SDZ, SMX, SMZ, TC, trimethoprim	0, 1, and 100 μg·L−1	F0: 150 dpf F1: eggs	F0, F1	F1: All antibiotics were detected in F1 eggs in 1 and 100 μg·L−1 treatment. Cardiotoxicity, and the apoptosis of cardiac cells. let-7a-5p regulated the cardiac hypertrophy signaling pathway.	[57]

Table 1. The endpoints and effects of ECs on fish across trans-/multi-generational exposures.

ECs	Fish Species	Exposure Dose	Exposure Duration	Generations	Endpoint in Offspring	Reference
Single ECs on trans-/multi-generational fish
Microplastics	fathead minnows (Pimephales promelas)	100, 2000 particles·L−1 (150–500 μm)	P: 178 dpf F1: 12 dpf	P, F1	P: Affect the growth, lipid storage, and external coloration. F1: In the low-concentration treatment (100 particles·L−1), microplastics can delayed onset of spawning, reduced egg survival rate, and a higher malformation rate in the offspring.	[37]
	fathead minnows (Pimephales promelas)	100, 2000 particles·L−1 (150–500 μm)	P: 178 dpf F1: 12 dpf	P, F1	F1: Methylation was more significant in parental males than in parental females and F1 offspring.	[38]
Carbamazepine	Zebrafish (Danio rerio)	10 μg/L	P: 6 weeks	P, F1, F2, F3, F4	P: Reproductive output, courtship and aggressive behaviors, 11-ketotestosterone (11 KT), and sperm morphology decreased. F1-F4 (only paternal fish exposed): 11 KT, reproductive output, altered courtship, aggression, and sperm morphology decreased.	[55]
OP	viviparous eelpout (Zoarces viviparus)	25, 100 μg·L−1	P: 17 and 35 dpf	P, F1	F1: the mortality of embryos increased, and the body length and weight decreased; the expression levels of estrogen receptors and vitellogenin of F1 embryos were increased; abnormal gonadal differentiation occurred.	[63]
PFOS	Zebrafish (Danio rerio)	5, 50, 250 μg·L−1	P: 150 dpf F1: 7 dpf	P, F1	F1: The mortality rate of F1 larvae was 100% at 250 μg·L−1; the locomotor activity of F1 larvae were abnormally enhanced in 5 μg·L−1.	[66]
BPA	rare minnow (Gobiocypris rarus)	15, 225 μg·L−1	P: 21 dpf	P, F1	P: induced poor quality of the embryos. F1: The fertilization, hatching, and spontaneous coiling rates decreased; the malformation rate increased; ossification in craniofacial cartilage were delayed; embryonic heart rate increased.	[60]
NP	Zebrafish (Danio rerio)	2, 20, 200 μg·L−1	P: 32 dpf F1: 120 dpf F2: 140 dpf F3: 7 dpf (recovery)	P, F1, F2, F3	P: the expression of fshβ and lhβ were upregulated; F1: inhibit the growth of zebrafish and cause gonadal damage. the expression of fshβ and lhβ were upregulated. fshr, lhr, and esr were decreased. F2: inhibit the growth of zebrafish and cause gonadal damage. exhibits stronger tolerance to high concentrations of NP (20 and 200 μg·L−1), but higher sensitivity to low-concentration NP (2 μg·L−1) F3: Fertility ratio and hatching ratio were decreased.	[6]
Combined ECs on trans-/multi-generational fish
Microplastics and EE2	fathead minnows (Pimephales promelas)	MPs, MP + EE2 (10 ng·L−1), MP + EE2 (50 ng·L−1)	P: 44 dpf F1: 14 or 21 dpf	P, F1	F1: reduced embryos locomotor activity; stronger swimming ability.	[65]
Microplastics and salinity	marine medaka (Oryzias melastigma)	180 μg·L−1 polystyrene ( 5.0 μm), 5‰ and 25‰ salinity	F1: 25 dph F2: 120 dph (recovery)	F0 (unexposed), F1, F2 (recovery)	F1: high salinity treatment increases in MDA content, TAC, CAT activity and cortisol levels. Additionally, also upregulated the mRNA levels of antioxidant genes (nrf2, keap1a, keap1b, sod1, cat, mt2) and HPI axis genes (crhr1, crhr2, pomca, mc2r, cyp11c1). Reduced the growth performance.	[39]
Nanoplastics and bisphenol	Zebrafish (Danio rerio)	BPA 20 μg·L−1, NPs 1 mg·L−1, NPs 1 mg·L−1 + BPA 20 μg·L−1 (N + B)	F0: 28 dpf F1: 4 months (recovery)	F0, F1	F0: N + B groups induced the gut damage and microbiota dysbiosis; Differentially expressed metabolites were mainly enriched in purine metabolism, mTOR, FoxO pathways, and all enriched in mTOR in control/NPs/N + B and control/BPA/N + B groups. The N + B group had the most such metabolites; nanoparticle + BPA co-exposure exacerbated F0 zebrafish liver metabolic disturbance, with mTOR possibly the mechanism. F1: Dead egg rate was increased at NPs, and N + B groups; the TC and TG level increased at N + B groups.	[36]
Bifenthrin, Levonorgestrel, ethinylestradiolt, renbolone	inland silverside (Menidia beryllina)	bifenthrin (3.02 ng·L−1), EE2 (6.79 ng·L−1), levonorgestrel ( 9.27 ng·L−1), trenbolone ( 9.60 ng·L−1).	F0: 8–21 dph F1: unexposed F2: unexposed	F0, F1, F2	F0, F1, F2: Through RRBS, which were attributed to strict inheritance of DNA methylation alterations and dysregulation of epigenetic control mechanisms. differential methylation was more frequent in gene bodies than in promoter regions across all treatments and generations	[64]
Antibiotic	CAM, CIP, CTC, EYR, ENR, LCM, NOR, OFL, OTC, ROX, SDZ, SMX, SMZ, TC, trimethoprim	0, 1, and 100 μg·L−1	F0: 150 dpf F1: eggs	F0, F1	F1: All antibiotics were detected in F1 eggs in 1 and 100 μg·L−1 treatment. Cardiotoxicity, and the apoptosis of cardiac cells. let-7a-5p regulated the cardiac hypertrophy signaling pathway.	[57]

2.5. Population-Level Implications of Multigenerational Effects

Individual-level multi-/trans-generational effects have potential population consequences, though empirical population-level data remain scarce. Reduced F2 fecundity and altered sex ratios observed in multiple studies could theoretically decrease population growth rates if exposure is sustained and widespread. For example, ethinylestradiol exposure causing skewed sex ratios across generations could reduce effective breeding population size, while decreased hatching rates in F2 offspring would lower recruitment [67]. However, whether these individual-level effects translate to population decline in wild settings remains uncertain for several reasons:

First, laboratory studies use genetically uniform populations lacking the genetic diversity that might buffer wild populations against transgenerational effects through varied susceptibility or adaptive capacity. Second, density-dependent compensation mechanisms may offset reduced fecundity, if fewer offspring survive to reproduction, remaining individuals may experience reduced competition and higher per-capita reproductive success. Third, recovery potential after exposure cessation is unknown, if epigenetic changes are reversible, populations might recover once contamination is removed. Fourth, most studies examine isolated single-species effects rather than community-level interactions where predators, competitors, and prey might also be affected.

Critical data gaps preventing population risk assessment include: (1) no field studies documenting population decline attributable to transgenerational contaminant effects; (2) lack of population modeling integrating multigenerational data with demographic parameters; (3) unknown threshold exposure durations or concentrations causing irreversible population effects versus recoverable impacts; (4) absence of data on whether natural selection can lead to rapid adaptation, potentially mitigating population-level consequences.

Among the studies reviewed above, the majority examine multigenerational effects in F1-F2 generations. True transgenerational effects (F3+) have been documented for only a limited number of compounds. For example, NP showed persistent effects in F3 zebrafish [6], while benzo[a]pyrene (BaP) demonstrated F3 impacts in marine medaka [8]. The scarcity of F3+ data reflects the practical challenges of conducting multi-year studies spanning three or more generations, representing a significant knowledge gap in understanding true transgenerational inheritance of emerging contaminant effects in fish.

3. Mechanisms of the Trans-/Multi-Generations Toxicity of ECs on Fishes

To date, most studies on aquatic organisms have focused on the effects of a single toxic compound on one or more generations and their offspring. These results may have cognitive deficiencies and inadequacies in understanding the toxic effects and mechanisms of endocrine-disrupting chemicals (EDCs) on aquatic organisms [68]. Based on the current research findings, the following three types of toxicological mechanisms have been summarized (Figure 3).

3.1. Trans-Generational Transmission, Bioaccumulation and Oxidative Damage

ECs can accumulate in fish across generations to induce toxic effects, a common toxicological mechanism. ECs have been detected in various fish species across different trophic levels [69], and this long-term accumulation may reach teratogenic and lethal levels. When ECs enter organisms for metabolism, the parent compounds are often the most toxic throughout the biological metabolic process, also referred to as the ultimate toxicant [70]. As the concentration of ECs in fish increases across generations, more reactive oxygen species (ROS) are generated [71]. ROS can disrupt intracellular redox balance, causing oxidative damage to lipids, proteins, and nucleic acids. Organisms initially respond to ROS accumulation through compensatory upregulation of antioxidant defense enzymes (SOD, CAT, GPx, GR). Zhou et al. demonstrated that F0 zebrafish exposed to low concentrations (10 μg·L⁻¹ PS-NPs) successfully mounted adaptive responses with significantly increased SOD and CAT activities [72]. However, this compensatory capacity is concentration- and time-dependent. Li et al. documented a biphasic pattern in rainbow trout where antioxidant enzymes initially increased at 7 days but became exhausted and decreased below control levels at 21–42 days under prolonged carbamazepine exposure, accompanied by elevated oxidative damage [73]. Most critically, this compensatory failure is inherited across generations. Pitt et al. showed that F1 zebrafish offspring, despite never being directly exposed to nanoplastics, exhibited depleted glutathione reductase and could not mount the GPx upregulation observed in their parents, indicating inherited impairment of compensatory mechanisms [74]. Zhang et al. further demonstrated progressive deterioration in F2 marine medaka, where oxidative stress worsened compared to F1, with increased MDA and decreased antioxidant capacity under high salinity [39]. When compensatory antioxidant systems fail or become exhausted in this manner, unopposed ROS accumulation proceeds to activate downstream damage pathways.

Excessive ROS can activate the nuclear factor κB (NF-κB) signaling pathway in acute and chronic exposures. NF-κB is a key regulator of immune and inflammatory responses. Once activated, it migrates into the nucleus and upregulates the expression of a series of pro-inflammatory cytokines, including tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and interleukin-6 (IL-6) [75]. Continuous activation of this pathway, induced by ECs-generated ROS, may lead to chronic inflammation in fish. Long-term inflammation can disrupt normal cellular functions, trigger cell death, and ultimately result in fish mortality or the emergence of teratogenic phenotypes [76,77]. While ROS can activate NF-κB signaling in directly exposed organisms [75,76,77], whether this pathway contributes to transgenerational effects in F2/F3 offspring remains undemonstrated, as multigenerational studies have not measured NF-κB activation, inflammatory cytokines, or downstream inflammation markers across unexposed generations. The persistent transgenerational phenotypes observed may result from inherited compensatory antioxidant failure, epigenetic inheritance of oxidative damage, continued inflammation signaling, or a combination of these mechanisms.

3.2. Endocrine Interference and Signaling Pathways Disorder

After entering the bloodstream, ECs migrate to target organs or tissues. During this process, they transform into multiple intermediate contaminants. These contaminants, including PPCPs and pesticides, may disrupt fish endocrine systems. For example, some PPCPs, such as bisphenol A (BPA) and 17α-ethinylestradiol (EE2), can mimic or antagonize natural hormones [78,79]. BPA binds to estrogen receptors in fish, disrupting the normal regulation of the hypothalamic-pituitary-gonadal (HPG) axis [80]. This interference may cause abnormal sexual development, such as feminization in male fish. Manifestations include reduced sperm production, abnormal gonadal morphology, and altered secondary sexual characteristics [81]. Additionally, exposure to per- and polyfluoroalkyl substances (PFAS) can imbalance estrogen receptor (ER) and androgen receptor (AR) expression in adult zebrafish and reduce sperm quality [82].

Beyond endocrine disruption, ECs may interfere with fish metabolic pathways. For instance, triclosan inhibits fatty acid synthase (FAS) activity, leading to hepatic fat accumulation in zebrafish. This likely occurs via interactions with short-chain fatty acids and miR-101a [83,84]. Pesticides can inhibit key enzymes in fish metabolism. Some organophosphorus pesticides target acetylcholinesterase, an enzyme critical for neural function and involved in certain metabolic regulatory processes [85]. Disrupted metabolic pathways cause energy imbalance and abnormal nutrient metabolism. These effects ultimately impact fish growth, development, and survival. In severe cases, they may result in lethal or teratogenic outcomes, such as growth retardation, developmental malformations, and increased mortality [86].

Studies show that different EC types target specific tissues and organs. For example, persistent organic pollutants (POPs) and EDCs primarily act by interfering with fish endocrine systems. They bind to hormone receptors, disrupt normal hormonal signal transduction, and thus affect various fish physiological functions [87,88,89]. Antibiotics may alter the normal fish microbial community, disrupt physiological balance, and induce oxidative stress through certain metabolic pathways [90]. When organisms are exposed to ECs over multiple generations or transgenerationally, these tissues and organs sustain continuous damage. Particularly during fetal, neonatal, and embryonic stages, when detoxification mechanisms are underdeveloped, ECs more easily affect offspring via germ cells, the placental barrier, etc. This increases multi-generational or transgenerational toxicity. Such long-term accumulation may reach teratogenic and lethal levels.

3.3. The Transgenerational Inheritance of Epigenetic Modifications

Emerging contaminants can establish heritable epigenetic changes that persist across unexposed offspring generations, representing a critical mechanism of transgenerational toxicity. Understanding which epigenetic modifications are causal versus correlative, when contaminant exposure establishes inheritable marks, and why some effects persist to F3 is essential for risk assessment.

3.3.1. Epigenetic Modifications

Not all contaminant-induced epigenetic changes transmit transgenerationally or directly cause phenotypes. Three contaminant studies illustrate this distinction. Promoter methylation shows strongest causal links. PFOA exposure in rainbow trout induced promoter methylation of lipid metabolism genes (fabp, pparα), directly reducing transcript levels and disrupting lipid homeostasis in both exposed fish and their offspring [91]. Cadmium exposure increased promoter methylation of detoxification genes in zebrafish embryos, causally silencing transcription [92,93,94]. This establishes the causal pathway that contaminant exposure induces promoter methylation, leading to transcriptional silencing, thus exhibiting phenotypic outcome. Distal DMRs may be correlative. MEHP phthalate exposure (30 µM, 0–6 dpf) in zebrafish caused 410 DMRs, with 44% overlapping conserved non-genic elements. However, locus-specific validation across 103 CpG sites at 10 loci revealed only 20% showed consistent F2 transmission [95], indicating genome-wide discovery overestimates true transgenerational inheritance by ~5-fold. Many distal DMRs represent bystander marks without direct functional consequences. Absence of methylation-phenotype correlation indicates alternative mechanisms. Benzo[a]pyrene exposure caused skeletal malformations persisting to F3, yet no DNA methylation correlation was detected [96]. This suggests histone modifications or ncRNAs may drive transgenerational effects for some contaminants independently of DNA methylation, which is a critical consideration when interpreting negative methylation results.

3.3.2. Critical Developmental Windows

Contaminant exposure timing determines whether effects transmit transgenerationally. Gametogenesis exposure establishes most stable marks. Adult marine medaka exposed to hypoxia (1.4 mg/L dissolved oxygen) during spermatogenesis established sperm methylome changes with 409 differentially methylated gene promoters transmitted to unexposed F2 offspring [97]. Marks incorporated during active gametogenesis bypass embryonic reprogramming, explaining why adult exposures can cause F2 effects [98,99]. Early embryonic exposure during sex determination creates delayed-onset effects. BPA/EE2 exposure during medaka days 0–7 (encompassing germ cell differentiation at days 5–7) showed no apparent F0 or F1 abnormalities, yet F2 exhibited reduced fertilization and F3 showed decreased embryo survival [54]. The mechanism underlying is that contaminants disrupting epigenetic programming during germ cell specification establish marks that manifest only when those cells complete gametogenesis in subsequent generations, explaining why F0/F1 can appear normal while F2/F3 show effects. Species-specific reprogramming influences contaminant sensitivity. Zebrafish maintain 80–85% methylation throughout PGC development with no global erasure [98], making them highly permissive for transgenerational transmission. Medaka shows bi-phasic demethylation yet retains 25% of non-promoter CpG islands hypermethylated [99]. Contaminant exposures during low-methylation windows (15 dpf nadir in medaka) may establish more persistent marks than exposures outside these windows.

3.3.3. Methodological Approaches

Genome-wide methylation profiling identifies contaminant-induced DMRs but requires validation. WGBS provides single-base resolution but assesses only ~50% of fish genomes [100]. The methylmercury study used MeDIP-Seq on F2 sperm, revealing increased epimutations compared to F0 and generation-specific DMR patterns [101]. RRBS enables population-scale studies with lower DNA input, the endocrine disruptor study used RRBS to identify 6–11% of genes differentially methylated in F0-F2 silverside following bifenthrin, EE2, and trenbolone exposure [64]. Teleost genome duplications complicate alignment (54–75% versus >85% in mammals), and zebrafish RRBS shows only 1.7-fold CpG enrichment versus 3–4-fold in mammals [100]. Locus-specific validation is essential. The MEHP phthalate study demonstrated that exposed zebrafish exhibited 410 genome-wide DMRs, but only 20% validated at F2 [95]. Bisulfite PCR amplicon sequencing of specific candidate loci provides higher confidence than genome-wide discovery alone. Integration with transcriptomics and functional validation establishes causation. The hypoxia study integrated MeDIP-Seq, RNA-seq, and proteomics, revealing that EHMT2 promoter hypomethylation lead to an increased expression, followed by elevating H3K9me2, thus resulting in silencing of reproductive genes across F0-F2 [97]. The dnmt1 mutant study provided strongest causal proof, where knockdown of hypermethylated genes (runx3, rptor) recapitulated transmitted phenotypes [102]. Fish-specific methodological limitations include: (1) F2/F3 larvae tissue amounts often insufficient for WGBS (1–5 µg DNA required); (2) ChIP-seq antibody cross-reactivity issues; (3) lack of comprehensive histone modification data across F0-F3 in any contaminant study. No microplastic, pharmaceutical, or EDC studies have measured complete histone modification profiles transgenerationally.

3.3.4. Transmit Contaminant-Induced Marks

Fish exhibit fundamentally permissive transgenerational epigenetic inheritance due to absent germline erasure. Unlike mammals undergoing massive PGC demethylation (71% to 7–14%), zebrafish PGCs maintain 84% methylation throughout specification [98]. This creates vulnerability that contaminant-induced methylation changes face no global erasure checkpoint and can persist across generations. Paternal template dominance explains male-biased transmission in many contaminant studies. The maternal genome reprograms to match paternal methylation patterns during early cleavage, establishing sperm methylome as the embryonic template [98]. Contaminant exposures affecting male germline therefore have high transgenerational transmission probability. Limited active demethylation creates default methylation maintenance.

3.3.5. Mechanisms Enabling Contaminant Effects to Persist Through F3

The F3 generation represents the first truly unexposed generation (F1 was exposed as embryos, F2 as germline within F1). Four mechanisms enable contaminant-induced marks to persist through F3. First, imprint-like resistance to demethylation. Contaminant-induced repressive histone marks (H3K9me2, H3K27me3) [97] shield CpG sites from TET enzyme access, creating local environments resistant to demethylation. The marine medaka salinity/microplastics study showed F2 oxidative stress worsened compared to F1 [39], suggesting stable epigenetic inheritance rather than gradual erasure. Second, genetic-epigenetic coupling. Cadmium exposure selected for SNP-methylation coupling in cep19: G allele frequency increased from 0.17 to 0.34, with G/G genotype showing 83.48% methylation versus 7.08% in A/A [94]. Once established, genetic variants maintain methylation patterns across generations independently of continued contaminant exposure. Third, epigenetic hub amplification. Contaminant-induced methylation of chromatin-modifying enzymes (e.g., EHMT2 in hypoxia study) drives genome-wide histone changes affecting hundreds of downstream targets [97]. Reversing hub effects requires coordinated demethylation across multiple loci, which is a substantial barrier. Fourth, environmental stress impairs erasure. Contaminant co-exposure with environmental stress (salinity, temperature, hypoxia) suppresses TET expression [103] or enhances DNMT activity [102], stabilizing induced marks. The marine medaka microplastics study showed salinity-dependent F2 effects [72], demonstrating environmental context modulates epigenetic persistence.

4. Prospects and Challenges of ML and AI for Toxicity of ECs to Multi-Generational Fish

The transgenerational effects of emerging contaminants reviewed above, from microplastic-induced reproductive dysfunction to EDC-mediated epigenetic inheritance in fish, present a fundamental challenge: how can we predict and mechanistically understand such effects for the thousands of untested chemicals entering aquatic ecosystems? Machine learning and AI technologies offer a complementary approach, with demonstrated success in predicting toxicity endpoints and elucidating mechanisms [104,105,106,107,108]. While traditional trans-/multi-generational fish toxicity research depends on multi-year in vivo experiments observing mortality, growth inhibition, and reproductive impairment across generations [6,109,110], ML approaches can potentially predict F2/F3 outcomes from F0 data, identify mechanistic biomarkers, and reveal structure-toxicity relationships.

Machine learning approaches have demonstrated substantial success in predicting various toxicity endpoints in fish [106,107,111], evolving from simple structure-activity relationships to sophisticated deep learning architectures [112]. Gustavsson et al. developed transformer-based deep learning models that significantly advanced chemical toxicity prediction in aquatic organisms, achieving high accuracy across fish, algae, and aquatic invertebrates while substantially outperforming traditional QSAR methods [113]. Similarly, Fan et al. successfully applied support vector machines to predict reproductive toxicity of endocrine disrupting chemicals in aquatic species, demonstrating that ML algorithms could capture complex relationships between chemical structure and reproductive endpoints [114].

While no published studies yet apply ML to fish transgenerational toxicity, mammalian research provides a computational roadmap. Fish transgenerational inheritance, like mammals, operates primarily through DNA methylation changes [105], suggesting mammalian ML approaches can transfer to aquatic systems. Liu et al. developed comprehensive machine learning models for rat multigenerational reproductive toxicity, analyzing 275 chemicals using seven different algorithms, achieving balanced accuracies of 58–65% [115]. Beyond reproductive endpoints, ML has proven capable of predicting the underlying epigenetic mechanisms. Haque et al. achieved 100% accuracy in predicting genome-wide locations of transgenerational epimutations using Active Learning and Imbalanced Class Learning [116], while Beck et al. employed deep learning to identify exposure-specific transgenerational DNA methylation patterns [117].

The absence of ML applications to transgenerational toxicity in fish stems from disciplinary isolation rather than technical limitations. The experimental data on microplastics, PPCPs, and EDCs generate the type of data ML requires, such as exposure conditions, epigenetic measurements, and multi-generational phenotypes (Table 1). However, these datasets remain scattered, incompletely characterized for ML purposes (missing chemical descriptors, inconsistent endpoints, limited metadata), and siloed within communities lacking computational expertise. The biological foundation for ML application is strong which fish transgenerational inheritance operates through DNA methylation mechanisms [98,99], the same pathway where mammalian ML models achieved 95–100% prediction accuracy [116].

Several challenges must be addressed using existing experimental data. Data scarcity presents the primary obstacle. Approximately 30–50 chemicals tested across zebrafish, medaka, and other species represent a small fraction of datasets typically used in successful ML applications [104,105,106]. Mechanistic complexity compounds the difficulty, as transgenerational effects manifest through multiple epigenetic pathways (as discussed in Section 3.3) and affect diverse endpoints across generations, requiring models integrating multiple biological levels. Species-specific variations in susceptibility add another layer, differential responses between zebrafish and medaka for similar exposures demonstrate that phylogenetic differences complicate broadly applicable models [36,37,38,39,47]. Finally, regulatory acceptance requires transparent, explainable AI approaches that can mechanistically justify predictions of effects in unexposed generations.

Despite these limitations, ML offers specific capabilities for mechanistic understanding. The diverse contaminants reviewed share no obvious structural similarities, yet some produce transgenerational effects while others do not. ML could identify cryptic structural features or physicochemical properties that predict transgenerational toxicity across chemical classes. Some chemicals cause immediate F0 effects that diminish by F2 [6,72], while others show minimal F0 impact but severe F2/F3 consequences [39,54]. ML could identify which molecular features predict transient versus persistent effects. By integrating methylation data with chemical structures and phenotypic outcomes, ML could predict which genomic regions are susceptible to heritable modifications, explaining mechanistically why certain contaminants cause transgenerational inheritance while structurally similar chemicals do not.

5. Concluding Remarks and Future Perspectives

Previous studies indicate that ECs induce multiple effects in trans-/multi-generational fish, including decreased fertilization and hatching rates, increased teratogenicity and carcinogenicity, abnormal development of tissues (e.g., bones and heart), neurotoxicity, and endocrine disruption. While these effects persist through several generations, evidence for population-level or ecosystem impacts remains hypothetical. The documented trans-/multi-generational effects suggest potential risks to population dynamics, though field validation and longer-term studies are needed to confirm ecological significance.

Based on critical knowledge gaps identified in this review, we propose the following prioritized research agenda. First, current studies mainly focus on short-term exposure to high pollutant doses, which far exceed environmentally relevant levels (ng·L⁻¹ to low μg·L⁻¹ range), limiting our understanding of biological changes under long-term stress. Therefore, future research should prioritize to conducted on the toxic effects and mechanisms of environmentally relevant concentrations on fish. Second, studies on ECs exposure to trans-/multi-generational fish, especially those spanning more than three generations, remain very limited. Most reviewed studies examine F1-F2 generations, which, while valuable for understanding multigenerational impacts, cannot definitively establish transgenerational epigenetic inheritance. Existing results show that the damage caused by ROS and endocrine disruption is limited to a single generation. Future research priorities should include expanding F3+ assessments to verify which emerging contaminants cause genuine transgenerational inheritance versus transient multigenerational effects. Third, Combined EC exposures in fish remain mechanistically uncharacterized, with no studies formally distinguishing synergistic from antagonistic interactions or extending mixture effects beyond F2 generations (Section 2.4). Finally, the number of ECs shows an annual increasing trend, and existing experimental toxicology cannot fully analyze the toxic effects and mechanisms of ECs on fish. Therefore, efforts can be initiated from the field of ML and AI to further strengthen investment in this area.