The strategy for anthracycline neutralization and detoxification in a cardiac cell is diverse, and failure of involved proteins can result in anthracycline accumulation. The solute carrier (SLC) 28A3 transporters (also known as sodium-coupled nucleoside transporters) are expressed in the human heart and are used by anthracyclines as an entry point (17). The cardioprotective effects of SLC28A3 has been recapitulated in human induced pluripotent stem cell model, and desipramine, a SLC competitive inhibitor, has been identified to exert protective effects on AIC (18). Moreover, ATP-binding cassette (ABC) proteins are involved in exporting multiple chemotherapeutics, such as anthracyclines, from cardiac cells using energy derived from ATP hydrolysis. ABCC1-ABCC2-ABCC5, as prominent proteins of this type, are membrane bound transporters highly expressed in human cardiac cell (9, 19). In addition, the following genes listed in Table 1 are among the most prominently involved.

The proteasome is thought to be involved in transporting anthracyclines from the sarcoplasm to the nucleus (24). Anthracyclines through the planar ring system intercalate the DNA bases and the amino sugar region, which prevents DNA replication and transcription. Inhibition of Topoisomerase-2 (Top2alpaha in cancer cells) interferes with the release of DNA supercoils and the ternary complex of anthracycline, topoisomerase and DNA inhibits the re-ligation of double-stranded breaks leading to their persistence. Further, doxorubicin-induced oxidative stress facilitates phosphorylation and activation of ERK and p53 (25). Subsequently, activated p53 enters the nucleus and up-regulates the expression of Bax and PUMAα. PUMAα can not only directly promote the translocation of Bax to mitochondria, but also indirectly activates Bax by competitive binding of Bcl-xL protein (26). Also, the activated p53 protein facilitated down-regulation of transcription factor GATA-4, which results in the lower expression of the Bcl-2 protein (27). The above-mentioned factors cause Bax to form pores in the mitochondrial outer membrane, leading to release of cytochrome C (Cyt c), which stimulates the activation of caspase-9/-3 and thereby cell apoptosis (28–30).

There is considerable overlap between the mechanisms accounting for anticancer effects and cardiotoxicity though distinct differences exist. Anthracyclines, for example, block topoisomerase-2 beta in cardiomyocytes and alpha in cancer cells and have a preference for mitochondria in cardiomyocytes. Mitochondrial injury appears to be a key component of anthracycline-related cardiotoxicity, and has been hypothesized as the cause of the long-term risk of cardiomyopathy linked with anthracycline exposure (2). According to studies, the expression of the transferrin receptor (TfR) is up to five times higher in cancerous tissue than in healthy tissue (31), DOX-Tf compound should be able to increase the intracellular concentration of medicines in breast cancer cells, assisting in the treatment of chemoresistance (32). As previously reported, DOX conjugation with Tf markedly increased cytotoxicity in human leukemia cells, both DOX-sensitive and DOX-resistant, as well as those derived from solid tumor cells (33). According to earlier studies, Tf-bound DOX triggers the TRAIL-dependent, extrinsic apoptosis pathway, which results in programmed cell death. TNF- and other cytokines are present, demonstrating a connection between the conjugate's pro-inflammatory effects and its cytotoxicity (34, 35).

Homeostasis in the sarcoplasmic reticulum can break down due to various physical and chemical factors. Anthracyclines create ROS by single electron transfer under the action of SR flavoenzymes such as NADPH-cytochrome P450 reductase, which can disrupt Ca²⁺ homeostasis in the SR of cardiomyocytes and impair the function of Ca²⁺-dependent chaperone proteins and enzymes; hence, triggering the unfolded protein response (UPR). In fact, anthracycline-induced ROS and RNS can suppress the expression of sarco/endoplasmic reticulum Ca²⁺-ATPase2 (SERCA) resulting in decreased Ca²⁺ absorption across a concentration gradient from sarcoplasm to SR. As a result, there is an increase in the phosphorylation level of Ca2+/calmodulin-dependent protein kinase II (CaMK II), an essential kinase that regulates Ca²⁺ homeostasis, and over-activated CaMK II abnormally up-regulates the phosphorylation of SR calcium channels and calcium pumps, leading to calcium imbalance (43). Therefore, excessive accumulation of abnormal proteins due to Ca2+-dependent chaperone protein impairment, activates SR transmembrane sensors including protein kinase-like endoplasmic reticulum kinase (PERK), transcription factor 6α (ATF6α) and inositol-requiring kinase 1α (IRE1α). Thereafter, luminal SR chaperone, glucose-regulated protein 78 (GRP78; known as immunoglobulin-binding protein or BiP) is activated and released from these sensors and allows dimerization and auto-phosphorylation of PERK as well as IRE1α, and translocation of ATF6α and further processing in the nucleus. PERK is a kinase enzyme which phosphorylates and inactivates the elongation initiation factor (eIF2a), leading to a broad reduction in protein translation, which in turn leads to the translation activating transcription factor 4 (ATF4) and transcription factor C/EBP homologous protein (CHOP) (44, 45). The abnormal expression of CHOP can change Bax/Bak by inhibiting Bcl-2, leading to increased mitochondrial outer membrane permeability, which makes large amounts of Ca²⁺ leak from the SR lumen into sarcoplasm. Subsequently, calmodulin is activated and then mediates the activation of caspase-12, which further activates caspase-9/-3 and finally initiates apoptotic pathway in cardiomyocyte (44).

Cardiotoxicity in zebrafish embryos can be noted when doxorubicin (DOX) is administered at different time windows ranging from hatching (0–3 dpf) (50) to post-hatching larva (after 3 dpf) (51). To avoid early cardiogenesis, Liu et al. (2014) incubated zebrafish embryos with solutions containing 100 µM doxorubicin at 1 day post-fertilization (dpf), when the heart had formed and circulation had begun (50). Given that zebrafish embryos are encircled within a protective chorion before 3 dpf, and most drugs are enriched inside the chorion (52), some researchers prefer to use either post-hatch larvae that naturally remove the chorion or dechorionated embryos for modeling AIC. An alternative method to incubation is to deliver anthracyclines via intrapericardial injections (53). Consistent cardiac phenotypes have been noticed among different delivery methods, including pericardial edema, impaired cardiac contractility, decreased blood flow, and increased apoptosis (50). We termed these models as embryonic AIC models (eAIC) in this review. Besides doxorubicin (DOX), Han et al. assessed cardiotoxicity incurred by other anthracyclines in zebrafish embryos, including daunorubicin (DAU), pirarubicin (PIRA), and epirubicin (EPI), as well as a liposome encapsulated DOX. They found that cardiotoxicity of DOX can be partially attenuated by the liposome packaging, while DAU caused the least cardiotoxicity. While DOX reduces heart rates significantly, the other four anthracyclines do not (54).

An attractive attribute of the eAIC models is their powerful chemical genetic capacity, which enables rapid assessment and screening of large number of compounds. eAIC models have been used widely to assess candidate therapeutic compounds that have been reported to exert therapeutic effects on other types of cardiotoxicity and/or cardiomyopathy and to elucidate the underlying mechanisms. D006-induced cardioprotection on eAIC was noted, which may be facilitated by maintenance of mitochondrial biogenesis and augmentation of HO-1 expression (55). Derivative of danshensu and tetramethylpyrazine–DT-010 might rescued cardiomyocytes in eAIC by decreasing ROS generation and preventing cell death, which may be facilitated through the stimulation of the PGC-1α/Nrf2/HO-1 pathway (56). Cardioprotective functions of synthesized Isosteviol analogues on eAIC could be exerted through blocking NF-κB signaling pathway and rising the mitochondrial action (57). Cardioprotective effects of allium extracts, including allium flavum and allium carinatum, have been noted (58). Tanshinone IIA, a component of Salvia miltiorrhiza, exerts cardioprotective effects on eAIC by increasing the autophagic flux via autophagosome formation and autolysosome clearance through the Beclin1/LAMP1 molecular pathway (59). A009, a polyphenol-rich Olive mill wastewater (OMWW) extract, was able to counteract the doxorubicin-induced cardiotoxic effects in the zebrafish eAIC model, which is consistent to its cardioprotective activities on rat and human cardiomyocytes. Intriguingly, this waste material from extra-virgin olive oil (EVOO) processing is able to increase the effects of breast cancer chemotherapy via its antiangiogenic and antiproliferative activities (60). While all of these aforementioned compounds are able to rescue cardiac dysfunction to a certain degree, therapeutic target genes for these compounds remain elusive and further mechanistic studies are needed to translate these discoveries to AIC patients. In addition to chemical genetic studies, eAIC has been used as an animal model to assess gene modifying effects. Yamashita et al. used CRISPR–cas 9 technique to knock down the key antioxidant responsive gene–nuclear factor erythroid 2-related factor 2a (nrf2a). They found that Nrf2a-deficient zebrafish showed increased sensitivity to DOX-induced oxidative stress, prompting further studies of this gene as a beneficial modifier to AIC (61).

Besides testing known candidate therapeutic compounds and genes, eAIC models have been successfully utilized to discover new therapeutic antidotes. Dr. Peterson's laboratory completed two non-biased screens of compound libraries. Using an eAIC model whereby 1 dpf zebrafish embryos was treated with DOX for 48 h, they assessed 3,000 small molecules from the Prestwick and Spectrum chemical libraries and identified visnagin (VIS) and diphenylurea (DPU) as two hits. Through studying effects of mitochondrial malate dehydrogenase 2 (MDH2) inhibitors on the zebrafish eAIC model, the authors propose that visnagin's cardioprotective effect is associated with MDH2 modulation (50). Further studies uncovered cytochrome P450 family 1 (CYP1) as an important genetic factor for AIC (62). This conclusion was later confirmed by screening 2,271 small molecules from a proprietary, target-annotated tool compound collection. A total of 120 compounds with anti-cardiotoxicity effects were identified with seven of them being extremely efficacious (63). Interestingly, all seven highly effective compounds exhibited inhibitory activity towards cytochrome P450 family 1 (CYP1), despite their structural differences. In addition to the cardioprotective effects of CYP1 inhibitors, a zebrafish CYP1A mutant was found to be resistant to DOX cardiotoxicity, providing genetic evidence to support CYP1 as a therapeutic target gene for eAIC (63).

Despite a convenient in vivo model with high throughput, eAIC cannot fully recapitulate the progressive pathological process that often occurs months or years post chemotherapy in human patients. To address this shortcoming, Ding and co-workers reported the first DOX-induced cardiotoxicity model in an adult zebrafish (64). They injected a single bolus of DOX intraperitoneally into zebrafish of two months to one-year of age. Significant mortality was noted 8–12 weeks later, when their hearts showed cardiomyopathy hallmarks including activation of fetal gene expression, muscular disarray, and myofibril loss (64). Ma et al. presented a detailed summary of their injection protocol, and described two intraperitoneal injection methods, aiming to reduce the variation that might be associated with internal organ damage (65). Later, Wang et al. introduced a third intraperitoneal injection method, whereby the needle penetrated the ventral midline between the pectoral fins (66). Using this new method, significant mortality is avoided, while the fish still manifests severe cardiac dysfunction at 8 weeks post injection.

Because aAIC was among the first cardiomyopathy models in adult zebrafish, this model has been used extensively to develop phenotyping tools. When the aAIC model was first reported in 2011, high-frequency echocardiography (HFE) was not available and survival rate at 8 week post injection (wpi) or later has been used as a noninvasive index to assess aAIC. The other non-invasive index was critical swimming speed (Ucrit) that was measured using a swimming tunnel, which corresponds to exercise capacity measured by treadmill testing in rodent models or patients (67, 68). Later, Zhang et al. developed a Langendorff-like ex vivo system to measure cardiac function in an isolated adult zebrafish heart, which was used to quantify ejection fraction (EF) in aAIC models (69). A transient decline of EF was noted during the first week post-injection (wpi), reflecting early stage acute cardiac damage. EF then recovered after 1 wpi, only to drop again after 4 wpi, with a statistically significant decease at 8 wpi, reflecting a later cardiomyopathy phase of the aAIC model. Packard et al. developed an automated segmentation approach based on histogram analysis of raw axial images acquired by light-sheet fluorescent imaging (LSFI) to establish rapid reconstruction of the 3-D zebrafish cardiac architecture (70). They noted reduction of heart size in the early phase of the aAIC model, which is consistent with a recent report in human AIC patients (71, 72). Later, the same laboratory developed a semiautomated, open-source method—displacement analysis of myocardial mechanical deformation (DIAMOND)—for quantitative assessment of 4D segmental cardiac function. They demonstrated that basal ventricular segments adjacent to the atrioventricular canal display the highest 3D displacement and are also the most susceptible to doxorubicin-induced injury (73). The introduction of high frequency echocardiography enables reliable non-invasive quantification of ejection fraction in an adult zebrafish heart (74). The technology has been applied to the aAIC model, which was able to non-invasively report reduced EF at the late phase of aAIC (66).

One of the most clinically relevant aAIC models in mice is created by administering low dosage doxorubicin (5 mg/kg) intravenously (IV) four times per week (75, 76). This dosage was chosen based on the pharmacokinetics of DOX in mice following a single dose of 5 mg/kg, which is equivalent to the pharmacokinetics of DOX in cancer patients after a normal dose of DOX therapy (60 mg/m²) (75). In mice, the IV route was preferred to the intraperitoneal (IP) route because IP injection of DOX causes damage, fibrosis, and consequent malaise, anorexia, weight loss, and noncardiac decease (75, 77–79). Because 20 mg/kg doxorubicin was administered in the zebrafish aAIC model ages 2 months up to 1 year, this dose is the same with the accumulative dose in the mouse aAIC model (64). In the zebrafish aAIC model, reduced ejection fraction can be first noted at 4 weeks post injection (wpi) and the end functional readout is typically carried at 8 wpi, which is similar to the mouse aAIC model, whereby reduced ejection fraction is first noted at 2 wpi and the end functional readout is typically done at 7 wpi when cardiac function has become stable. Because adult zebrafish have strong regenerating abilities in their heart muscle, which mammals lack (80), it was unclear if the zebrafish aAIC model has distinct responses to DOX at the cellular level. To address this concern, Ding et al. revealed that while cardiomyocyte hypertrophy contributes to cardiac remodeling, cardiomyocyte hyperplasia does not, as evidenced by unaltered PCNA index and 5-bromo-2-deoxyuridine (BrdU) incorporation (64). Based on these data, they concluded that the pathogenesis in the zebrafish aAIC model is likely conserved with that in mouse aAIC, without incurring significant compensational cardiac regeneration.

The generation of aAIC models and the development of phenotyping tools in zebrafish paved the way towards leveraging powerful zebrafish genetic tools to study AIC. A non-biased forward genetic approach has been recently established, which enables the discovery of modifying genes for AIC, i.e., genes that either accelerate or slow down the progression of AIC (82, 83). The screen was built on a gene breaking transposon (GBT)-based insertional mutagenesis system that is proven to effectively disrupt tagged genes (84, 85). Besides efficient gene disruption in each GBT mutant line, the expression pattern of the endogenous gene can be reported by a RFP fluorescent protein reporter. Around 1,000 GBT mutant lines have been generated at Mayo Clinic, ∼20% of which manifest visually noticeable cardiac expression, resulting in the generation of a zebrafish insertional cardiac (ZIC) mutant collection (86). Ding et al. stressed heterozygous ZIC lines with DOX to search for genetic factors that modify the survival and/or cardiac function of aZIC (86). By screening 609 GBT lines, they identified 4 candidate aAIC-modifying genes, including GBT0002/sorbs2b, GBT0136/ano5a, GBT0411/dnajb6b (L) and GBT0419/rxraa.

To query where the screen is effective in discovering meaningful genes in AIC and/or cardiomyopathy, follow up studies of these 4 GBT mutant lines and affected genes were carried out. (1) ANO5 is a known dilated cardiomyopathy (DCM) causative gene (87) (2) Studies of a mouse Sorb2 knock out (KO) mutant confirmed the modifying effects of heterozygous Sorb2 KO on AIC (88). Interestingly, Sorbs2 encodes an intercalated disc (ICD) protein and homozygous Sorb2 KO mutant manifests arrhythmogenic right ventricular cardiomyopathy (ARVC)-like phenotypes, including enlarged right ventricle and cardiac arrhythmia. Preliminary human genetic studies identified likely pathogenic variants from two ARVC patients, supporting SORBS2 as a candidate ARVC gene (88). (3) The DNAJB6 gene encodes a member of the J protein family, which serves as a chaperone molecule to aid folding and quality control of the protein (89). Alternate splicing of the same gene results in two DNAJB6 isoforms: a short DNAJB6(S) isoform that is encoded by the first 6 exons, and a longer DNAJB6 (L) isoform that adds two additional exons to the 3′ end of the gene (86). Different from the short DNAJB6(S) isoform that is a causative gene for limb girdle muscular dystrophy (LGMD) (90, 91), DNAJB6(L) is a cardiac-enriched isoform (86). Pathogenic sequence variants were found in dilated cardiomyopathy (DCM) patients, and a transgenic fish line harboring the human variant DNAJB6 (L)-S316W exerts deleterious modifying effects on aAIC in zebrafish. In contrast to deleterious modifying effects of GBT411/dnajb6(L), Tg(DNAJB6(L)) exerts therapeutic effects on aAIC (86). The therapeutic benefits of Dnajb6 (L) overexpression were confirmed in a mouse AIC model by using a AAV9-based gene delivery system. (4) In contrast to the other 3 GBT lines that exert deleterious modifying effects, GBT419 exerts salutary modifying effects on aAIC. Detailed molecular studies unveiled that the salutary modifying effects of GBT0419 can be ascribed to gain-of-function of retinoid x receptor alpha a (rxraa) in endothelial cells, but not in cardiomyocytes or in epicardial cells (67). The statement that RXRA is a new susceptibility gene for AIC was supported by a GWAS analysis of 1,191 patients with early-stage breast cancer treated with DOX (67).

In summary, follow-up studies of 4 GBT lines from a pilot screen confirmed important roles of these genes in AIC. In addition to modifying effects on AIC, many genes turned out to be susceptibility genes in cardiomyopathies, and some genes could be potential therapeutic target genes (Figure 1). This series of work established a new method for discovering genetic factors for AIC, which differs from the ongoing technologies such as genome-wide association studies (GWASs) in humans and quantitative trait locus analyses in rodents. There are significant limitations with these statistics-based methods, which often point to sizable genomic regions that cover a large number of candidate genes, making it difficult to confidently establish genotype-phenotype relationships (92). Conversely, the identity of each modifier gene identified from an aAIC-based forward genetic screen can be unambiguously determined. Therefore, the zebrafish aAIC model potentially opens the door to systematic discoveries of genetic modifiers of AIC, which could be achieved by scaling this forward genetic strategy to the whole genome.

Using the aAIC model, Ding et al. demonstrated that a heterozygous mTOR mutant exerts therapeutic benefits on aAIC, providing the first genetic evidence on cardioprotective effects of mTOR inhibition on AIC (64). During this study, dynamic changes in mTOR signaling were noted over the dynamic course of aAIC. Consistent with this notion, they found that rapamycin, a mTOR inhibitor, exerts deleterious effects in the early stage of aAIC, which is opposite to cardioprotective effects in the late stage (66). Similar to mTOR signaling, biphasic changes in autophagy activity were noted in the aAIC model: increase in the early and decrease in the late stage. In line with time-dependent signaling changes, overexpression of Atg7, a core autophagy protein, exerts deleterious effects in the early stage, but salutary effects in the late stage of aAIC. Next, Ding et al. leveraged to use the efficient zebrafish model for screening FDA-approved drugs with autophagy activating function (FAA). They reported that top-ranking FAAs are able to revert declined cardiac function in the late phase of aAIC, but exert deleterious effects in the early phase (66).

Besides autophagy, a time-dependent effect was also reported for RXRA-based therapy (67). The therapeutic effects of RXRA agonists only occur when administered during the early, but not the late, phase of AIC. Together, these studies in the in vivo zebrafish model raised an important concept that pathological signaling in the late phase of AIC can be quite different from, or even opposite to, the signaling in the early phase. Different types of therapeutic strategies must be designed to treat the early phase and late phase of AIC, respectively (Figure 2).