Page 792 - The Toxicology of Fishes

772 The Toxicology of Fishes

Identifying Modes of Action Not Represented by Standard Assays
For a variety of reasons, environmental toxicology has come to rely heavily on data generated according
to a limited set of standardized test procedures. Standardization of test procedures is very important in
the context of providing comparability in data and thereby allowing the development of extrapolation
models. Logistical considerations also limit the routine collection of certain types of toxicity data (e.g.,
life-cycle chronic tests, including reproduction). One consequence of this approach is that standard
toxicity tests are not well suited to detect certain types of toxicological effects. For example, the issue
of endocrine-disrupting chemicals caught the world largely by surprise and spawned intensive research
on the endocrine system. It could be argued that one of the reasons this issue did not come to light
earlier is that reproductive and second generation endpoints are rarely a part of aquatic toxicity testing
commonly performed on chemicals. The point here is not to suggest that massive testing of reproductive
endpoints should be undertaken but rather to underscore the importance of considering modes of action
that might lie outside standard testing protocols and developing toxicological tools (such as receptor
binding assays in the case of steroid hormone mimics) that can help identify chemicals that may act
through these alternative modes of action and thereby be incompletely assessed by common toxicity test
protocols.

Better Extrapolation/Reducing Animal Usage
Refining our ability to accurately predict ecological risk puts immediate pressure on having more
extensive toxicological data, but both resource limitations and the desire to reduce reliance on animal
testing provide pressure to pursue less toxicity testing. To meet both desires, greater ability to extrapolate
toxicological data is needed: extrapolation among species, among chemicals, and among endpoints.
Current USEPA ambient water quality criteria for the protection of aquatic life require acute toxicity
data for at least eight species and chronic toxicity data for at least two of those species (Stephan et al.,
1985). Although these requirements were instituted to reduce uncertainty, they also limit the number of
chemicals for which criteria can be developed. Because many more chemicals require assessment than
have this level of toxicity testing, methods are needed to develop risk estimates based on fewer data,
and the means to make this extrapolation without introducing intolerable uncertainty are needed.

Multiple Stressors/Nonchemical Stressors
Although most toxicological data and, for that matter, most ecological risk assessments focus on
individual chemicals, real-world exposures are to multiple stressors, both chemical and nonchemical
(e.g., nutrients, habitat degradation, temperature). For some of the more ubiquitous chemical mixtures,
toxicological models have been developed to assess the potency of aggregate mixtures, such as for
polycyclic aromatic hydrocarbons (PAHs) (Di Toro et al., 2001a,b; Swartz et al., 1995), polychlorinated
biphenyls (PCBs), dibenzodioxins, and dibenzofurans (Van den Berg et al., 1998). Even more challenging
is the development of approaches that integrate the effects of chemical and nonchemical stressors in
terms of their combined influence on populations and communities.

Linkage of Organismal and Suborganismal Responses
to Population and Community Response
We previously discussed the need to link suborganismal measures to organismal-level responses such
as survival, growth, and reproduction. A further challenge is to better link organismal (or suborganismal)
responses to population and community level responses; for example, diazinon exposure has been shown
to affect chemoreception in salmon, measurably reducing a characteristic antipredation response after
an olfactory cue (Scholz et al., 2000). Although it is not difficult to imagine that reduced antipredator
behavior could lead to the decline of an exposed natural population, it is also likely that very small
changes in behavior would not affect the stability of the population as a whole. What is not clear is how
one could quantitatively relate measured behavioral changes to the projected risk to salmon populations.