Rajapakse, N, E Silva and A Kortenkamp. 2002. Combining Xenoestrogens at Levels below Individual No-Observed-Effect Concentrations Dramatically Enhances Steroid Hormone Action. Environmental Health Perspectives 110:917–921.

People never are exposed to just one chemical. We all have multiple contaminants in our bodies at the same time. In this paper, Rajapakse et al. present a sophisticated analysis of one type of interaction that can take place. They found that the presence of multiple contaminants, each known to bind with the estrogen receptor yet, in this experiment, at levels too low to detect an effect when tested individually, caused a dramatic increase in the effect of estrogen itself.

What did they do? Rajapakse et al. performed a series of experiments using the now-classic yeast estrogenicity assay that quantifies the degree to which a compound can stimulate an estrogenic response via binding with the estrogen receptor.

They first ran their experiments one chemical at a time. Using these results, they established dose-response curves for each compound, including estrogen. This allowed them to determine the "no-effect" level, i.e., the concentration at which each compound, by itself, had no detectable estrogenic response.

They then created different mixtures of the natural estrogen, 17ß-estradiol and 11 different xenoestrogens in varying concentrations, and measured the dose-response curves of these mixtures.

The xenoestrogens they used were:

• 2´,3´,4´,5´-tetrachlorobiphenyl-4-ol
• 2´,5´-dichlorobiphenyl-4-ol
• 4´-chlorobiphenyl-4-ol
• genistein
• 2,4-dihydroxbenzophenone
• benzyl-4-hydroxyparabene
• 2,3,4,5-tetrachlorobiphenyl
• bisphenol A
• resorcinol monobenzoate
• 2,3,4-trichlorobiphenyl
• phenyl salicylate

What did they find? In the first phase of their work, Rajapaske et al. studied each compound individually to determine relationships between dose and response specific to each chemical.

Each xenoestrogen had a characteristic dose-response curve describing its ability to provoke an estrogenic response. As expected, all xenoestrogens were significantly weaker than natural estrogen, 17ß-estradiol.

Absorbance curves for 17ß estradiol and 11 xenoestrogens. Weaker compounds plot farther to the right.   adapted from Rajapaske et al. 2002

Each absorbance curve shows the degree of estrogenic response elicited by a given compound at a given concentration. Each of this curves is sigmoidal (S-shaped) in shape. Beneath a certain level (different for each compound), no response is observed. As the concentration of that compound is increased (moving right and upward along a particular curve, the response increases, until it reaches a maximum and then levels out at a plateau. Increasing the concentration then has no additional effect on absorbance.

To compare estrogenic potency among compounds, Rajapaske et al. compare the concentration of the compound necessary to produce 1% of the maximum response observed in the assay.

17ß-estradiol is the strongest, with 1% of the response reached by 2.3 x 10-5 µM (micromolar, a standard unit of concentration). Phenyl salicylate (labeled 12 in the graph) is the weakest. Its concentration must be roughly 400,000 higher than estradiol's to elicit the same level of response. Bisphenol A (#9) is roughly 28,000 times weaker than estradiol. The strongest xenoestrogen used, 2´,3´,4´,5´-tetrachlorobiphenyl-4-ol (#2), is 350 times weaker than estradiol. Clearly all the xenoestrogens in this experiment are dramatically less potent than natural estrogen at provoking estrogenic responses.

This first round of experiments then set up the second phase. Rajapaske et al. worked out dose-response curves for mixtures of the 11 xenoestrogens in combination with 17ß-estradiol. They did this first by combining each xenoestrogen in a xenoestrogen stock mixture with each compound present in proportion to its concentration level necessary to provoke 1% of the maximum response. They then combined that xenoestrogen mixture in solutions in combination with a 17ß-estradiol stock mixture, also at its 1% level. The key variable here was that the ratio of the xenoestrogen stock to 17ß-estradiol stock varied from 25,000 to 1 to 100,000 to 1.

Their interest in these combinations was to test whether it was possible to predict, on the basis of the dose-response curves of the compounds by themselves, what the dose-response curve of the mixture would be.

They found excellent agreement between observed effects and their calculated predictions, using a method called "combination-addition" that is designed specifically for using with sigmoidal dose-response (S-shaped) curves.

They also found that a more traditional method, "effect-addition" led to erroneous predictions that would have been interpreted wrongly as synergistic. In other words, the "effect-addition" approach underestimated the true effect of the mixture, whereas the "combination-addition" approach predicted it correctly. Both these approaches assumed additivity of the compounds, but they were based upon different models of additivity, one corrected for the sigmoidal nature of the dose response curve. They concluded, therefore, that the effect of the combination of estradiol and xenoestrogens in a mixture was additive, not synergistic.

In a final and profoundly important step in their analysis of the importance of mixtures, they looked at set of these experiments in which the levels of the xenobiotics were so low that individually they would not cause any effect.

For each xenoestrogen they had determined a "no-effect" level, that is, the highest concentration of that compound that caused no statistically-significant effect. In other words, at that level or below, the compound's effect was not any different than having no compound.

It is often said by skeptics of the potential impact of endocrine disruption that weakly estrogenic compounds like those used in this experiment won't have an effect because there is so much more natural estrogen around and it is so much more powerful. These skeptics would thus predict that the effect of these contaminants in combination on the assay should be trivial. After all, each of them are present at a level that causes no effect. This is the way that toxicological standards are currently developed: each in isolation.

Rajapakse et al. then looked again, with a different lens, at the mixture experiments and specifically at a case in which the xenoestrogens were all beneath "no effect" levels and 17ß-estradiol concentrations were in the mid-range of its dose response curve.

In this case, they knew what 17ß-estradiol by itself would do at that concentration because they had begun the work with compounds individually (above). They then predicted what the mixture would do based on their two prediction models, "combination-addition" and "effect-addition" (CA and ES in the figure below).

They found (observed effect, figure below) that the mixture more than doubled the effect of 17ß-estradiol. Their CA prediction matched the model, whereas the ES prediction grossly underestimated the mixtures' impact.

Effect of adding 11 xenoestrogens, each at levels beneath effect levels, to a solution with 17ß-estradiol. The effect of the mixture more than doubled the effect of 17ß-estradiol alone. The predictions using a "combination-addition" method were accurate whereas the "effect-addition" model grossly underestimated the observed effect.

What does it mean? Rajapaske et al.'s results are important for three reasons.

First, they clearly and elegantly demonstrate that xenoestrogens at very low levels can dramatically increase the responsivity of a system to endogenous 17ß-estradiol. These results definitively refute the claim that the low potency and low concentrations of xenoestrogens, compared to 17ß-estradiol, make them insignificant players in the control of hormone systems. According to Rajapaske et al.:

Second, the comparisons of the two different types of predictions, "combination-addition" vs. "effect-addition" show that it is important to use the correct predictive models to distinguish between "additive" vs. "synergistic" effects. Both of these predictions were based on assumptions of additivity. One, the CA model, was developed specifically for sigmoidal curves. The "effect-addition" model, however, has been more widely used in toxicology. It may have led to erroneous conclusions about synergy.

Third, they show conclusively that working with mixtures is vital to anticipating the effects of endocrine disrupting compounds. Their mixtures caused effects that normal regulatory science would miss completely, because it is based on work with compounds one-by-one. These results mean it is highly likely—if not virtually certain—that current standards, based on traditional regulatory science, are not strong enough to protect human health.

Many questions remain to be followed-up. Foremost among these is the fact that while some contaminants have estrogenic activity, others have anti-estrogenic impacts. And multiple contaminants of both types are probably ubiquitous in most living organisms. How do they interact in mixtures?