Ideally, decisions regarding environmental health should be made with an understanding of the effects that xenobiotics have on population growth rates (Forbes and Calow 2002; Schmidt 2004); however, this can be challenging for amphibians because they have relatively long generation times (often ≥ 3 years) and are frequently not amenable to life-cycle completion under tractable conditions. Moreover, the sheer number of registered chemicals (tens of thousands) and species to be tested and the complexities of population growth rate analyses may make this approach too time-consuming considering the urgent need to identify causal factors in the global decline of amphibians (Houlahan et al. 2000; Stuart et al. 2004). Researchers have suggested that a more practical and efficient alternative may be to study the effects of contaminants across life stages using amphibians reared under semi-natural conditions where postexposure effects and density-dependent regulation are permitted (Rohr et al. 2004; Sih et al. 2004). The proxy for population-level effects would be the net effect of the contaminant at some point near reproductive age, where the net effect is defined as the sum of exposure effects, carryover effects, and density-mediated compensation (Figure 1). We used this approach to investigate the net effects of the herbicide atrazine (0, 4, 40, and 400 ppb) on the streamside salamander, Ambystoma barbouri.

We focused on atrazine for several reasons. Atrazine is the most widely used pesticide in the United States, and possibly the world, and is one of the most common contaminants in groundwaters and surface waters [U.S. Department of Agriculture (USDA) 2002; U.S. Environmental Protection Agency (U.S. EPA) 1994]. Thus, many amphibians are likely exposed to atrazine. Atrazine concentrations seldom exceed 20 ppb for long but are likely to remain low (parts per billion) for extended time periods, especially considering that reported half-lives in fresh water exceed 100 days (de Noyelles et al. 1980, 1989; Klaassen and Kadoum 1979). Near agricultural areas, atrazine levels have been reported as high as 500 ppb in ponds (de Noyelles et al. 1982), 200 ppb in streams (U.S. Geological Survey 2000), and 40 ppb in rain (Nations and Hallberg 1992).

Recent studies have raised concerns regarding the effects of atrazine on amphibian fitness. The endocrine-disrupting properties of atrazine have been implicated in exposures between 25 and 0.1 ppb inducing gonadal abnormalities and hermaphroditism in leopard frogs (Rana pipiens) and African clawed frogs (Xenopus laevis) (Carr et al. 2003; Hayes et al. 2002a, 2002b, 2003; Tavera-Mendoza et al. 2002a, 2002b). Exposure to 3 ppb has increased tadpole mortality (Storrs and Kiesecker 2004) and susceptibility to infection (Kiesecker 2002).

We employed a 4 × 2 ×2 complete factorial design in which each aquarium was randomly assigned one of four ecologically relevant atrazine concentrations (0, 4, 40, or 400 ppb; 98% pure; Chemservice, West Chester, PA), one of two food-level treatments (low or high), and one of two hydroperiod treatments (presence or absence of a water level reduction). Atrazine was predissolved in acetone; all treatments contained 0.004% acetone, and concentrations were confirmed by flame ionization detection gas chromatography. Similar solvent concentrations were shown to have no effect on A. barbouri behavior, life history, or survival (Rohr et al. 2003a), and so a negative control was not included in this experiment. Low-food aquaria received 2.24 g (± 0.005 g) of live black worms (Lumbriculus variegates) twice a week, and larvae in the high-food aquaria were fed black worms ad libitum. In half the aquaria, the volume of water was incrementally reduced during water changes to simulate summer drying of water bodies, but atrazine concentration remained the same. Water volume did not drop below 2 cm. Mortality was quantified every other day throughout the exposure period, and animals were exposed to treatments until metamorphosis (mean ± SD, 59.9 ± 4.52 days; range, 47–70 days; for effects on embryos, larval growth, behavior, and timing of metamorphosis, see Rohr et al. 2004). In summary, there were three replicate aquaria of each treatment combination and a total of 12 aquaria exposed to each of the four atrazine concentrations.

After metamorphosis, A. barbouri from each aquarium were placed into terraria (18.5 cm in diameter, 7.5 cm high) so that the 48 aquarium replicates were preserved. Thus, density at metamorphosis carried over into the terrestrial stage, simulating what occurs in nature and allowing us to quantify any density-mediated compensation. Each terrarium contained approximately 6 cm of homogeneously mixed, organic top soil. Every week, we fed the salamanders vitamin- and mineral-enriched crickets ad libitum and added water to the terraria (if needed) to ensure that each had sufficient moisture. Terraria were maintained at room temperature on a 12:12-hr photoperiod. We recorded survival on 13 July 2004, a mean of 421 days (SD = ± 4.52; range, 410–433 days) since metamorphosis and last atrazine exposure. Survival was not recorded between metamorphosis and 13 July 2004 because postmetamorphic A. barbouri burrow into the substrate and digging them out greatly disturbs both the salamanders and the terraria. When possible, salamanders were treated humanely and with regard for alleviation of suffering.

We used the general linear model to test for the main and interaction effects of atrazine (continuous predictor and log₁₀-transformed), food level, and dry down on angularly transformed percent survival at metamorphosis (during atrazine exposure), after metamorphosis (carryover effect plus density-mediated compensation), and on 13 July 2004 (net atrazine effect; Figure 1). To test for carryover effects, we controlled for starting density in the postexposure phase (by incorporating survival to metamorphosis as a covariate in the statistical model) when examining postexposure survival. Because density-mediated compensation is driven by the different starting densities in the postexposure period, controlling for starting density removes density-mediated compensation, leaving any carryover effects of the contaminant (plus error; Figure 1). Where the main effect of atrazine was significant, we examined all pairwise comparisons among concentrations using Fisher’s least significant difference (LSD) tests.

To compare the magnitude (slope of regression) of atrazine’s exposure, post-exposure, carryover, and net effects, we examined all pairwise comparisons of these four effects using the general linear model, treating pairs of effects as repeated measures variables (i.e., we compared these effects within aquaria/terraria). An interaction between effect type and atrazine would indicate that mortality associated with atrazine depended upon the type of effect considered. The α-value for these tests (α = 0.0083) was modified for the number of comparisons using a Bonferroni adjustment. Comparisons with the carryover effect were conducted using the standardized residuals of survival at metamorphosis regressed against survival after metamorphosis.

During the exposure period, atrazine concentration was positively associated with mortality (F_1,40 = 13.80, p < 0.001), with survival in control aquaria differing from that in aquaria containing 40 ppb (LSD, p = 0.027) and 400 ppb atrazine (LSD, p < 0.001; 0 vs. 4 ppb, p = 0.094; Figure 2A). Exposure concentration was also positively associated with mortality for the carryover effect (F_1,39 = 10.71, p = 0.002). Survival of salamanders not exposed to atrazine was, once again, greater than the survival of animals exposed to 40 ppb (LSD, p = 0.039) and 400 ppb (LSD, p = 0.015; 0 vs. 4 ppb, p = 0.056; Figure 2B). The carryover effect analysis revealed that starting density in the postexposure phase was positively associated with postexposure mortality (F_1,39 = 8.76, p = 0.005), indicating that the heavy mortality imposed by atrazine during the exposure period was ameliorated after exposure by density-mediated compensation (Figure 2B). Although the carryover effect was greater than density-mediated compensation, this compensation was substantial enough to offset a sizable portion of the carryover effect, resulting in a nonsignificant relationship between exposure concentration and net postexposure survival (F_1,40 = 3.55, p = 0.067; Figure 2C).

Although density-mediated compensation neutralized the bulk of the atrazine-related carryover effect, it was not substantial enough to fully counteract the mortality associated with both the exposure and carryover effects. Consequently, exposure concentration was related positively to the net mortality across both the exposure and postexposure stages or mortality at the end of the study (F_1,40 = 8.91, p = 0.005; Figure 2D). A. barbouri that were not exposed to atrazine had greater net survival than did A. barbouri exposed to 4 ppb (LSD, p < 0.013), 40 ppb (LSD, p < 0.007), or 400 ppb atrazine (LSD, p < 0.001; Figure 2D).

Although we do not know when the post-exposure mortality occurred, we saw no evidence that there was substantial mortality soon after the salamanders were transferred to their terraria. Further, in a separate study on this species with similar exposure and rearing conditions, atrazine had no significant effect on larval survival, yet exposure was associated with hyperactivity and increased desiccation risk 8 months after exposure, with no detectable recovery from atrazine exposure (Rohr and Palmer 2005). These data suggest that exposure to atrazine early in development may have permanent effects on these salamanders. This conclusion is supported by our finding that the effect of atrazine during exposure did not differ from the magnitude of its carryover effect, once again suggesting that there was no recovery from atrazine exposure. Atrazine has been shown to disrupt neuroendocrine processes in amphibians (Hayes et al. 2002b, 2003; Larson et al. 1998), and this certainly is a possible mechanism for the observed long-term effects on behavior and survival.

Various studies have reported nonlinear relationships between atrazine concentration and amphibian responses, and many of these relationships have been nonmonotonic, with the largest change in response occurring at low exposure levels (Hayes et al. 2002b, 2003; Larson et al. 1998; Storrs and Kiesecker 2004). Endocrine disruptors commonly induce non-monotonic dose–response curves (Welshons et al. 2003), and thus the endocrine-disrupting potential of atrazine has been suggested as the cause of the documented nontraditional non-monotonic dose responses (Hayes et al. 2002b, 2003; Larson et al. 1998; Storrs and Kiesecker 2004). The relationship between atrazine concentration and survival in this study was nonlinear (logarithmic) but monotonic. However, we cannot rule out the possibility of a non-monotonic relationship because detection depends upon selecting the right concentrations to reveal any nonmonotonicity. The responses to atrazine recorded in A. barbouri have many consistencies with endocrine disruption. The largest adverse change in survival occurred at low exposure concentrations. Further, exposure to endocrine-disrupting compounds during critical developmental stages often induce irreversible effects (Bigsby et al. 1999), not unlike the long-term, postexposure effects observed here and in a previous study (Rohr and Palmer 2005). Virtually every response to atrazine we have quantified in this species using these concentrations has had the greatest response change at low concentrations (Rohr et al. 2004; Rohr and Palmer 2005; Rohr et al. unpublished data), suggesting that A. barbouri may be sensitive to marginal atrazine inputs into aquatic systems.

The deleterious, long-term effects of larval atrazine exposure suggest that encounters with contaminants after metamorphosis may not be necessary for contaminants to harm postmetamorphic amphibians, a life stage that often disproportionately affects population dynamics (Biek et al. 2002, Schmidt et al. 2005, Vonesh and De la Cruz 2002). This is important because exposure to substantial concentrations of contaminants is probably more likely before metamorphosis, because most amphibian embryos and larvae are strictly aquatic and cannot readily escape water bodies where many contaminants accumulate and concentrate.

Although the data in this laboratory study suggest that atrazine may have adverse effects on A. barbouri, we caution against extending this interpretation to populations in nature or to other amphibian species for the following reasons: a) causes of postmetamorphic mortality in the laboratory may not match those in the field; b) there is documented annual variation in the effects of atrazine on amphibians (Rohr et al. 2004); c) the strength of density-dependent processes may be different in the wild and for other species (Hellriegel 2000); d) species can differ in their susceptibility to contaminants (Bridges and Semlitsch 2000; Forbes and Calow 2002); and e) effects of contaminants can depend upon community structure (Relyea and Mills 2001; Rohr and Crumrine 2005), which was not incorporated into this study. Further, density-mediated compensation can be delayed by a generation or more; this, however, seems unlikely for atrazine because it is applied as a preemergent and thus enters ponds and wetlands in agriculture landscapes annually at relatively consistent times and quantities (Hayes et al. 2003, Huber 1993, Solomon et al. 1996). Consequently, larvae of many amphibian species are likely to be exposed to atrazine each spring, reducing the chances of cross-generational compensation. Nevertheless, the possibility of cross-generational compensation suggests that combining experiments that span life stages with demographic models might be particularly insightful for evaluating the population-level effects of environmental stress. Despite the aforementioned admonishments, this study, and evidence that exposure of amphibians to low levels of atrazine can increase mortality (Storrs and Kiesecker 2004), impair reproductive development (Hayes et al. 2002a, 2002b, 2003; Tavera-Mendoza et al. 2002a, 2002b), and adversely interact with natural stressors (Boone and James 2003; Kiesecker 2002; Rohr and Palmer 2005), must raise concerns about the role of this widespread, persistent, and mobile herbicide in the international decline of amphibians. Certainly, more research on the effects of atrazine on amphibians is necessary.