Part II (also see introduction)

Welshons, WV, KA Thayer, BM Judy, JA Taylor, EM Curran and FS vom Saal. 2003. Large effects from small exposures. I. Mechanisms for endocrine disrupting chemicals with estrogenic activity. Environmental Health Perspectives doi:10.1289/ehp.5494

This page explores one component of the publication by Welshons et al., specifically their analysis of the different mechanisms involved in inducing changes in MCF-7 breast cancer cell proliferation through exposure to estradiol.

The central finding of these experiments is that low level responses to estradiol are mediated by hormone binding to the estrogen receptor, whereas toxicological responses to high level exposures take place independently of binding with the estrogen receptor. That's important because standard toxicological testing is designed in a way that Welshons et al. conclude cannot detect estrogen receptor mediated responses. In other words, countless tests of potentially estrogenic substances have been done without any possibility of revealing whether or not they can interfere with natural estrogen signaling.

The implication is that regulatory testing is full of false negatives: conclusions that contaminants have no effects when in fact they do.

What did Welshons and his coworkers do? When exposed to estrogenic substances, MCF-7 cells increase their rate of cell division and proliferate. Welshons et al. studied this proliferation response over a wide range of doses of 17ß-estradiol, both by itself and in combination with two other types of chemicals, synthetic estrogens and anti-estrogens.

Doses of estradiol ranged from the control, 0 estradiol, through exposures at the parts per quadrillion, parts per trillion, parts per billion, and parts per million range. At the highest doses used, estradiol is cytotoxic.

The other chemicals were diethylstilbestrol (DES), a synthetic estrogen, at 3 parts per trillion, and two different anti-estrogens, raloxifene and ICI 182,780, at doses sufficient to eliminate any response mediated by the estrogen receptor.

They used two forms of the MCF-7 cell. One was the standard MCF-7, with estrogen receptors. The other was a variant of MCF-7 called C4-12-5 cells. These have been genetically engineered to lack estrogen receptors. Hence any response to estradiol (or any other chemical) in the C4-12-5 cells cannot be mediated by estrogen receptors.

What did they find?

As shown in the graph to the right, experiments with estradiol alone in the normal MCF-7 cells (with estrogen receptors) revealed a strongly non-monotonic shape, a classic inverted U.

When they exposed the C4-12-5 variant of MCF-7 cells to the same range of estradiol (left), they observed no enhancement of proliferation at low levels of exposure. The high level decrease in proliferation to levels beneath the control response, however, remained present.

In a third experiment (right), they found that the presence of the synthetic estrogen DES at 3 ppt was sufficient to eliminate the low dose, estrogen receptor mediated response to estradiol, but that it had no effect on the high dose response.

The crucial point to note here is the difference in shape of the first graph compared to the other two. Without the estrogen receptor (second graph), or in the presence of a contaminant that is estrogenic (third graph), the low level response of the system is eliminated.

Welshons et al. carried out a crucial additional set of experiments that serve as standard controls and positive controls. These controls allowed the researchers to separate responses mediated by the estrogen receptors from those that acted via some other (unspecified) mechanism. Without these controls and without explicit testing at extremely low levels (parts per trillion and below), the experiments would have indicated that estradiol itself is not estrogenic!

And the reality is that standard toxicological testing requires neither a full set of controls nor experiments at relevant low doses.

The controls employed by Welshons et al. confirmed that:

• The MCF-7 cell system used in these experiments is responsive to estrogens and the response can be suppressed by anti-estrogens. This is important to show because, while in principal this cell system is known to be responsive to estrogens, it is possible that the specific procedures and materials used by Welshons et al. could have contaminated the experiment with another estrogen that would have rendered the system insensitive to further hormonal stimulation. Inadvertent contamination can lead to false negatives.
• In the cell line lacking estrogen receptors, C4-12-5, estradiol did not elicit a response at low levels and the anti-estrogens did not inhibit the response at high levels. The clear implication is that estrogen receptors are necessary for the low-level response but not necessary for the high level response.

In perhaps the most telling positive control (right), purposeful low level contamination of the experiment by an estrogenic compound, DES, eliminated responsiveness to estradiol. Adding an anti-estrogen reduced the response, indicating that an estrogen (DES) was present in the background. In this case, the "control" is the MCF-7 system plus 3 ppt DES. By adding the anti-estrogen (blocking DES) plus a higher but non-cytotoxic amount of estradiol they show that DES's impact was via the same mechanism as estradiol.

This combination of experiments and controls illustrates how easy it would be to mistakenly classify estradiol—and by implication other estrogenic substances—as non-estrogenic at low levels of exposure.

Thus: in the first experiment (top graph) with functional estrogen receptors, estradiol had both a low-dose effect and a high dose effect. By showing in the second experiment (second graph) that the low-dose effect was not present in the absence of estrogen receptors, they prove that the low dose response is mediated by that receptor. This conclusion is strengthened further by the third experiment (third graph), which shows that contamination by another estrogen, DES, also eliminates the low dose response. This is because at the level of DES used, all estrogen receptors are already occupied even prior to the addition of estradiol. Hence adding estradiol cannot increase the proliferation response because it is already at its maximum level.

Finally, by using anti-estrogens in their positive controls (fourth graph), they show that the effect of the DES contamination can be removed, decreasing the proliferative response from 100% of the "control" to roughly 40%.

What does it mean? This seemingly simple and logical set of experiments raise huge questions about basic assumptions and procedures used in regulatory testing. Two issues loom large:

• Toxicological testing takes place at high levels and is assumed to allow extrapolation to low levels. These experiments show that for estrogens, not only is extrapolation inappropriate, the mechanisms mediating high-level vs. low-level responses are different at the molecular level. Estrogen toxicity (high level) is not the same as estrogen signal disruption (low level). Tests appropriate for examining estrogen toxicity reveal nothing about signal disruption at physiologically relevent levels.
• Background contamination by estrogenic substances can obscure estrogen signaling responses, making them impossible to detect unless extensive negative and positive controls are included in the experimental procedure. Such contamination is highly likely: from food (phytoestrogens and other contaminants) and from caging. For example, Howdeshell et al. report that polycarbonate caging creates the opportunity for estrogenizing animal subjects because it leaches bisphenol A even at room temperatures. It is highly likely that many experiments within the published literature have been misled by this sort of contamination, leading to spurious conclusions.

Taken together, these two factors imply that for toxicological testing of estrogenic substances, the literature is highly likely to be rife with false negatives.

Welshons et al. provide more detail about one specific example of a recent false negative in the toxicological literature, the failure of industry scientists to replicate findings that low level bisphenol A exposure in the womb leads to prostate enlargement in mice.