X. birchmanni and X. malinche are two species of swordtail fish from the Río Panuco Basin in northeastern Mexico. Males of both species exhibit divergent sexually dimorphic visual traits (figure 1a), and hybrids between the two species display easily recognized recombinant phenotypes (Rosenthal et al. 2003; figure 1b). Despite the conspicuousness of hybrids, Rauchenberger et al. (1990) noted ‘no evidence of hybridization’ in 1988 among sympatric populations in the Río Calnali in the state of Hildago, eastern Mexico despite an extensive survey based on morphometrics and isozyme markers. In 1997, Rosenthal et al. (2003) discovered a population of hybrid origin at the same site based on similar morphometrics and the same isozyme markers. Taken together, these studies suggest a recent hybridization event. Continued sampling of the Río Calnali suggests that individuals of hybrid origin are ubiquitous, and pure-species individuals are non-existent or rare (Rosenthal et al. 2003). A survey of mitochondrial haplotypes indicates that most of the hybrids are descended from X. birchmanni females (hybrid individuals sampled from three sites in 2003 from the Río Calnali, X. birchmanni haplotypes, N=62; X. malinche haplotypes, N=18; 435bp of the control region D-loop) suggesting that disruption of species-recognition mechanisms in X. birchmanni females may have played a role in hybridization. We therefore focused on the ability of female X. birchmanni to discriminate male conspecific cues from those of X. malinche in an altered chemical environment.

We began our investigation by testing female mating preference in the water collected directly from the stream where hybrids occur. During the last two decades, in the time that the X. birchmanni/malinche hybrid zone is likely to have arisen, the Río Calnali has been subject to increased disturbance, most notably due to the influx of untreated sewage and agricultural waste. It is well established that exogenous substances, including synthetic products, can have damaging effects on chemical communication (Park et al. 2001; Park & Propper 2002). Human impacts, however, may also cause changes in the abundance of naturally occurring substances. We next considered the potential effect of one such substance, humic acid (HA). HA results from degrading organic matter, and thus processes that increase plant and algal growth, such as eutrophication, are likely to be followed by high levels of HA before it is degraded by ultraviolet radiation (Thomas 1997). Natural levels of HA reflect this process and have been observed to fluctuate as much as 5mgl⁻¹ over a matter of hours, and can range from trace amounts to over 200mgl⁻¹ (Steinberg 2003). Although its chemical structure is unknown, HA has been shown to produce slight feminization in swordtail fishes (Steinberg et al. 2004), and to affect a variety of biological processes (Sato et al. 1987; Maggioni et al. 1992; Lovley et al. 1996; Qiao & Farrell 2002). In particular, HA has an affinity for two teleost steroidal reproductive pheromones (Mesquita et al. 2003) and causes a decrease in olfactory response to those pheromones in goldfish (Carassius auratus; Hubbard et al. 2002). All known fish sexual pheromones are either steroids or prostaglandins, both of which possess hydrophobic properties (Stacey et al. 2003). In water, HA forms microvesicles containing a hydrophobic core (Simpson et al. 2002), which may bind to pheromones and leave them undetectable to chemoreceptors (Hubbard et al. 2002). Furthermore, HA has been shown to inhibit membrane-bound receptors (Yang et al. 2002), suggesting that HA may also block or damage chemoreceptors.

In the experiments conducted on HA, 15 females were exposed to water containing supplemental HA, and 13 control females were tested in water without HA. All females responded to both stimuli in all clear-water trials (N=15), but significantly fewer subjects tested in HA responded to male chemical cues at all (N=9, Fisher's exact test: p=0.018). Female latency to conspecific cues did not differ significantly across trials in the control group (repeated measures ANOVA, F=1.891, p=0.149, d.f.=3, 12); however, latency to conspecific cues in the trial conducted in HA was significantly greater than in all other trials (repeated measures ANOVA, F=11.098, p<0.001, d.f.=3, 14; Bonferroni-corrected post hoc tests, p<0.05), females responded more slowly only when in water treated with HA (figure 3a). In the control group, there were no significant differences in female association time to the conspecific cue among trials (repeated measures ANOVA, F=0.755, p=0.526, d.f.=3, 12), but in the HA experimental group, females responded to conspecific odour cues significantly less during the trial conducted in HA and the trial conducted 2 days following exposure, than in the other trials (repeated measures ANOVA, F=16.913, p<0.001, d.f.=3, 14, Bonferroni-corrected post hoc tests, p<0.05; figure 3b). The results measuring the frequency that females spent within one body length of conspecific versus heterospecific stimulus outflows closely mirror the association time data: in the control group, there was no significant difference in female response to the conspecific cue among trials (repeated measures ANOVA, F=0.108, p=0.955, d.f.=3, 12), but in the HA experimental group females responded to conspecific odour cues significantly less during the trial conducted in HA and the trial conducted 2 days following exposure, than in the other trials (repeated measures ANOVA, F=17.324, p<0.001, d.f.=3, 14, Bonferroni-corrected post hoc tests, p<0.05; figure 3c).

Females tested at 20mgl⁻¹ HA also showed a loss of preference for conspecific cues (paired t-test N=15, t=0.273 p=0.788) and their net preference (time with conspecific cue minus time with heterospecific cue) was significantly different from clear-water trials (two-sample t-test N=15, t=2.24 p=0.041). Females showed a significant preference for conspecific cues in water containing only 2mgl⁻¹ HA (paired t-test N=15, t=2.36, p=0.033) and the net preference did not differ from clear-water trials (two-sample t-test N=15, t=0.805, p=0.434).

Females showed a strong preference for food odour cues in clear-water trials (paired t-test, N=12, t=4.83, p=0.0003), and no preference for food cues in water containing 20mgl⁻¹ HA (paired t-test, N=12, t=0.422, p=0.679) or 200mgl⁻¹ HA (paired t-test, N=12, t=1.14, p=0.883). The net preference of females in trials conducted with HA differed significantly from clear-water trials (two-sample t-tests, 20mgl⁻¹ HA N=12, t=3.71, p=0.002; 200mgl⁻¹ HA N=12, t=3.35, p=0.005).

In laboratory tests without supplemental HA, X. birchmanni females strongly preferred conspecific male chemical cues over heterospecific cues (figure 3). When we measured female preferences in water containing 200mgl⁻¹ HA, we found that they responded to both conspecific and heterospecific stimuli more slowly than in tests without HA (figure 3a). All females responded to both stimuli in all clear-water trials, but significantly fewer subjects tested in HA responded to male chemical cues at all. Furthermore, the females that did respond in HA trials, did not exhibit a difference in preference between conspecific and heterospecific male odour cues (figure 3b,c). Females tested at a markedly lower HA concentration, 20mgl⁻¹, also showed a loss of preference for conspecific cues; female preference, however, was retained in water containing only 2mgl⁻¹ HA.

In order to determine whether the loss of female preference was due to specific effects on chemical communication, rather than on females' overall motivation to mate, we tested females with computer animations of courting X. birchmanni and X. malinche males, in water with and without 200mgl⁻¹ of supplemental HA. Females showed nearly identical, significant preferences for conspecific visual cues in both treatments (figure 4). HA therefore inhibited females' species recognition based on chemical, but not visual cues. We also found that while females showed a strong preference for food odour cues in clear-water trials, they showed no preference for food cues in water containing 20 and 200mgl⁻¹ HA. These results further support the notion that HA disrupts detection of chemical cues in aquatic animals.

All experimental females were exposed to a single, brief (approximately 45min) bout of elevated HA concentration during a preference trial. In order to evaluate the long-term effects of this brief exposure, we retested the females in water without supplemental HA 2 days after the 200mgl⁻¹ HA treatment. While response latency returned to prior levels (figure 3a) and females again showed a preference for conspecific male odour (figure 3b,c), the strength of the response was significantly weakened: females continued to associate significantly less with conspecific odour when compared to their responses before exposure and 10 days after exposure, and to the responses of control females never exposed to HA (figure 3b). These results suggest that the effects of exposure to high levels of HA can persist even after a pulse of disturbance, likely due to a damaging effect of HA on the binding properties of chemoreceptor cells. In goldfish, HA reduces the response of olfactory epithelium and olfactory bulb to sexual pheromones (Hubbard et al. 2002). Furthermore, HA is known to inhibit binding and interfere with the function of mammalian receptor cells (Yang et al. 2002). Ten days following exposure to 200mgl⁻¹ HA, we found that females once again showed a strong preference for conspecific male odour (figure 3b,c). These results suggest that if HA did inflict chemoreceptive damage, full functional recovery was obtained after 10 days. Chemoreceptor regeneration time is unknown in Xiphophorus, but full regeneration may not even be necessary for certain olfactory discrimination tasks. In goldfish, behavioural responses and amino-acid discrimination ability persisted even after the surface of olfactory epithelium was reduced by 85% (Zippel 2000). Control females that were tested in water without supplemental HA throughout the series of trials consistently showed a strong significant preference for conspecific male odour, suggesting that the response profile of experimental females was due to the effects of HA rather than to habituation to the stimuli.

In addition to affecting chemoreception, HA may also be interfering with chemical cues themselves. Females' latency to respond to both conspecific and heterospecific cues was significantly greater in the trials conducted with supplemental HA than in any of those without, and latency did not differ across trials without HA (figure 3a). Humic substances are known to bind to hydrophobic chemicals and thus reduce their aqueous concentrations. Available concentrations of a steroid pheromone in goldfish were strongly reduced by the similar concentrations of HA we used in our study (Mesquita et al. 2003). We therefore cannot exclude the possibility that HA may interfere with the concentration of available pheromone in the swordtail signalling environment.

Regardless of the mechanisms whereby HA interferes with the female reproductive behaviour in X. birchmanni, the consequences are clear: elevated levels of HA, a natural and ubiquitous substance, compromises females' ability to discriminate between conspecific and heterospecific male odour cues without disrupting their motivation to mate (as shown in the visual preference tests). And although we cannot exclude the hypothesis that other compounds in Río Calnali may have influenced swordtail hybridization, our laboratory experiments directly test the effects of HA of species recognition of X. birchmanni. Despite the benefits that HA may provide (Kullberg et al. 1993), our results suggest that its impact could also be detrimental for some species. HA concentrations vary broadly in nature (Steinberg 2003). Particular environments such as the Río Negro in the Amazon River Basin, however, contain consistently high natural HA levels (Wood et al. 2003), and are home to diverse assemblages of animals that rely on chemical cues in social communication (Weitzman & Fink 1985). Furthermore, other animals seem unaffected in environments contaminated with sub-lethal amounts wastewater and agricultural effluent. Such animals may have evolved mechanisms for coping with their environments.