The interaction between a pair of species can be variable in space and time, but understanding of this variability is not well developed.  We focused on the interaction between zebra mussels (Dreissena polymorpha), freshwater filter-feeders exotic to North America, and the toxic phytoplankter, Microcystis aeruginosa, because studies have suggested opposite responses of M. aeruginosa to mussel invasion in different habitats.  Given indications from previous observational work, we hypothesized that this two-species interaction is highly sensitive to variation in ecosystem productivity (i. e., nutrient loading).  We tested this hypothesis with a large-scale enclosure experiment (40,000 L) conducted in oligotrophic Gull Lake, which experienced a large increase in M. aeruginosa after mussel invasion.  We specifically asked whether the response of M. aeruginosa to mussel manipulation could be reversed by nutrient fertilization.  Nutrient treatments were established that simulated conditions in mesotrophic and eutrophic lakes versus an ambient, oligotrophic control.  The response of M. aeruginosa biomass was quantified three weeks after mussels were introduced to half the enclosures.  One eutrophic enclosure had phosphorus levels that were two times greater than the target, and was excluded from the experiment (N = 29).  Nitrogen was not added since ambient nitrate in Gull Lake is high.
Results/Conclusions

We were able to maintain elevated and roughly-constant levels of phosphorus in the mesotrophic and eutrophic treatments via weekly monitoring of phosphorus concentrations coupled with targeted additions of phosphate.  There was a statistically significant interactive effect of nutrients and mussels on the biomass of M. aeruginosa after three weeks.  As seen in previous experiments and lake surveys, mussels had a strong positive (and significant) effect on M. aeruginosa biomass at ambient (oligotrophic) nutrient levels.  The effect of mussels in the mesotrophic treatment was non-significant, with elevated among-replicate variability.  Under eutrophic conditions, the mussel effect on M. aeruginosa biomass was reversed, i. e., significantly negative.  The positive and negative effects of mussels were very similar in magnitude, as quantified by the log ratio (0.48 versus -0.51).  Thus, we were able to reverse the sign of a two-species interaction via fertilization.  Mechanisms underlying the effect reversal are not known, but parallel research suggests large genetically-based variation within the M. aeruginosa population of Gull Lake with respect to vulnerability to mussel grazing.  The consequences of large intraspecific trait variation as a driver of variation in the direction and magnitude of species interactions is largely unexplored.