Hepatotoxicity Testing in Larval Zebrafish - Zebrafish: Methods for Assessing Drug Safety and Toxicity - page 92

performed on all larvae to confirm this classification on a case-by-case basis. In this
study, the overall correlation with mammalian in vivo data was good at 82%, with
sensitivity (hepatotoxic-specific compounds correctly identified) at 88% and spec-
ificity (nonhepatotoxic compounds correctly identified) at 67% (Hill et al., 2008). Of
these compounds, tamoxifen and danazol, both of which require metabolism in
mammals to cause toxicity, and troglitazone, which was withdrawn from the market
(Fung et al., 2001), were correctly identified in the zebrafish model, indicating that
zebrafish larvae may have the equivalent detoxification pathways. Supporting evi-
dence for this comes from numerous studies that have been cataloging zebrafish
metabolizing enzymes including various cytochrome P450s such as CYP3A, CYP1A,
CYP19, and CYP26 (Carney et al., 2004; Bresolin et al., 2005; Tseng et al., 2005;
Rubenstein, 2006). A couple of nonhepatotoxic compounds that also caused adverse
effects on the zebrafish liver were attributed to secondary toxicity induced when the
LOEC (lowest observed effect concentration) was exceeded, such as for nephrotox-
icity in the case of gentamycin, but this needed further investigation. In addition,
certain toxicants including ketoconazole and sodium valproate were incorrectly
identified as nontoxic, demonstrating again that this phenotypic model alone may
not be infallible. However, when bioanalytical techniques based upon LC-MS were
later incorporated into this assay, these compounds were shown to have only achieved
very low body burdens of compound in the larvae and hence should have been
unclassified pending additional tests with either higher aqueous concentrations or
microinjection, to ensure there were sufficient levels to cause an adverse effect. The
need for bioanalysis has also been discussed in the context of other zebrafish assays
such as for developmental toxicity (Gustafson et al., 2008; Hill, 2008b) and
cardiotoxicity (Hill, 2009) and at present no single physiochemical property has
been identified that can suitably predict the uptake of different classes of compounds
into zebrafish larvae (Doshna et al., 2009). In addition to demonstrating when
uptake has been poor, bioanalysis has also been useful to identify those compounds
that are readily absorbed and only cause toxicity at a body burden well in excess
of the expected therapeutic window for that drug. When these factors were later
considered for the hepatotoxicity assay, the sensitivity, specificity and overall
predictivity were revised to 97, 77 and 91% respectively (Jones et al. 2009).
Some further examples of zebrafish phenotypic hepatotoxicity screens are shown
in Table 8.1. Included are compounds that only cause overt adverse effects to the liver
as part of a severe acutely toxic response, such as for astemizole and haloperidol that
affect cardiac function as a primary endpoint of toxicity, and therefore hepatotoxicity
must be interpreted accordingly.
8.6 SECONDARY ANDMECHANISTIC LIVER ASSAYS
Although the phenotypic assay is a valuable, relatively high-throughput assay for the
determination of hepatotoxic liability, as the endpoints are indiscriminant for specific
modes of action, additional screens may be utilized to help elucidate the underlying
mechanisms involved.