Commercial rodent diets are a major source of inadvertent estrogen exposure for laboratory animals (Brown and Setchell 2001
; Thigpen et al. 1987a
). There are many potential sources of estrogenic substances in the diet, such as mycotoxins and pesticide residues. However, it is phytoestrogens that are quantitatively the major source of “estrogen” exposure for rodents. Phytoestrogens include primarily the soy bean isoflavones and coumestans derived from alfalfa, and these compound classes can profoundly influence the results of endocrine-sensitive end points (Boettger-Tong et al. 1998
; Brown and Setchell 2001
). Dietary phytoestrogens are biologically active, possessing many hormonal and non-hormonal properties, and consequently at significant levels of intake can affect growth and development, reproduction, gene expression, and sensitivity to carcinogens. Furthermore, the sensitivity of animals to phytoestrogens is also a function of species and strain, age, sex, dosage, route of administration, and duration of exposure (Ashby et al. 2000
; Odum et al. 2001
; Owens et al. 2003
; Padilla-Banks et al. 2001
; Spearow et al. 1999
; Thigpen et al. 2004a
; U.S. EPA. 2003
; Yang and Bittner 2002
Typical exposure level of rodents to levels of isoflavones contained in most commercial diets that are formulated with soy meal range from 80 to 160 mg/kg body weight/day which is far in excess of what is typically consumed when people eat soy foods on a daily basis (0.5–1.0 mg/kg body weight/day). However, there is wide variation in the phytoestrogen content of commercial rodent diets (Thigpen et al. 2004a
) because purified soy proteins can vary in phytoestrogen content by 3- to 4-fold (Setchell and Cole 2003
), given the large natural variation in the isoflavone content of soybeans (Eldridge and Kwolek 1983
; Hoeck et al. 2000
; Hou and Chang 2002
; Njitu et al. 1999
; Simonne et al. 2000
; Tsukamoto et al. 1995
). Such variation poses a major problem in manufacturing diets with consistent compositions. It is possible to control the soy protein content of the diets, but it is extremely difficult to control or standardize the phytoestrogen content because soy proteins used by industry vary by less than 3% from batch to batch (Setchell and Cole 2003
). Consequently, investigators using rodent diets formulated with soy meal face the prospect of being unable to control the extent of exposure to phytoestrogens; this may affect the results of estrogenic studies and make it difficult to both reproduce and compare results within or between laboratories.
The obvious solution to eliminating batch-to-batch variability in phytoestrogen content is to eliminate significant known sources of phytoestrogens by the removal of soybean meal or soy protein from commercial rodent diets used in studies that can be affected by dietary phytoestrogens. Soy protein or casein stripped of isoflavones and other estrogens by alcohol washing could possibly offer an alternative source of protein. The removal of soybean meal from all rodent diets is another potential option. Such a move would result in standardization of experimental results and an improvement in the sensitivity of bioassays for estrogenic substances. The results presented here reinforce the need to move in this direction and magnify the importance and need to use soy/alfalfa-free diets when conducting studies evaluating hormonal end points that can be affected by phytoestrogens (Brown and Setchell 2001
; Thigpen et al. 1998
). At a minimum, the phytoestrogen content of the diet should be reported.
Results of the present study clearly demonstrate that the 3- to 4-fold variability in phytoestrogen content between different mill dates of the same diet produce statistically significant differences in the time of VO in CD-1 mice and F344 rats but not in S-D rats. Furthermore, our results show that the S-D rat is less sensitive to dietary phytoestrogens than either the CD-1 mouse or the F344 rat. Our findings that the dynamic response window or range for the mean time of VO for the three rodent strains fed different diets varied ~ 10 days for the CD-1 mice (PNDs 20–30), and F344 rats (PNDs 32–42) and only ~3 days for the S-D rats (PNDs 29–32) provides additional evidence that the S-D rat is less sensitive to dietary phytoestrogens and is not the most sensitive strain for conducting VO bioassays. This observation has significant implications for the selection of the most appropriate rodent species and strain to be used in testing for EDCs. Using three different batches of the PMI 5002 diet, we found an inverse relationship between the time of VO and the dietary phytoestrogen content for F344 rats. Mean VO times ranged from 32.6 days (431 μg/g diet) to 35.5 days (98 μg/g diet) for diets containing phytoestrogens; this was significantly earlier than the mean time of VO of 38.2 days when this strain was fed a diet essentially devoid of phytoestrogens. For S-D rats, the same batches of PMI 5002 diets resulted in VO times (32.4–32.7 days) that were no different from the control PMI 5K96 diet (31.8 days), even though plasma isoflavone concentrations in S-D rats fed different mill dates of the PMI 5002 diet were much higher than the total isoflavone plasma levels in S-D rats fed the PMI 5K96 control diet (). The much higher plasma concentrations in the S-D rat versus the F344 rat and the CD-1 mouse may simply reflect the higher rate of food intake by this strain ().
You et al. (2002)
reported that an acceleration in the time of VO in exposed female offspring was the only observed effect of dietary genistein at 300 ppm (micrograms per gram of diet) or approximately 30–39 mg/kg/day. This difference in the S-D rats’ response to VO is probably caused by the difference in design of our study and that of You et al. (2002)
. In their study dams were fed the test diet (300 ppm) during gestation and weaning, and the female offspring were maintained on the test diet until VO was recorded. In our study, the dams, with their 8-day-old female pups, were placed on a phytoestrogen-free diet. After weaning (PND19), pups were placed on different mill dates of the same PMI 5002 test diet or on the control PMI 5K96 diet.
Strain differences in estrogen sensitivity were further evident from studies in which the AIN-76A diet (with a high ME level) was spiked with 0, 150, 300, or 450 μg/g diet of pure genistein. The F344 rats in study II showed significant differences in the time of VO in animals fed the diets with the two highest doses of genistein. A slight effect was also observed with the diet spiked with the lowest dose of genistein. This study showed that the mean time of VO in F344 rats varied from 36.75 to 26.75 days with increasing levels of genistein in the diet. On the other hand, the mean time of VO in S-D rats was only advanced by the diet containing the highest dose of genistein (450 μg/g diet). In contrast, F344 rats showed an advanced time of VO when fed the 300- or 450-μg genistein/g diet. S-D rats consumed more food per day than F344 rats, but surprisingly, based on the estimated dose of genistein, the F344 rats received more genistein from PND19 to PND26. However, this was reversed from PND26 to PND33 when S-D rats consumed a higher dose of genistein than F344 rats ().
When we looked at the plasma concentration of genistein for the CD-1 mouse and F344 and S-D rats (), we found a much higher dose-related response of genistein concentration in the plasma of S-D rats compared with CD-1 mice and F344 rats. A possible explanation for the marked difference in response to the different diets between S-D rats and F344 rats or CD-1 mice is the apparent inability of S-D rats to efficiently metabolize genistein. This idea is supported by Helton et al. (1977), who suggested that isolated intact liver parenchymal cells from S-D rats were less efficient than the C3H mouse cells in their ability to covert 17α-ethynyl-estradiol into its metabolites. The slower clearance of genistein by female S-D rats was also reported by Sfakianos et al. (1997); their data showed that genistein is metabolized by the liver and absorbed by the intestinal wall, but a small amount appears in the urine. Genetic differences in the inability of S-D rats to metabolize isoflavones at the same rate as CD-1 mice and F344 rats may be contributing to the higher plasma level of free genistein. The plasma concentration in the S-D rat exposed to the PMI 5002 diet in study I shows a similar pattern of response to D&G in the diet, although this data could not be compared to plasma from the F344 rat or the CD-1 mouse.
Study III was designed to discern the relative role of ME versus phytoestrogen content in influencing the time of VO. When the diet had an ME in the 3.04–3.2 Kcal/g range, differences in the phytoestrogen content of the diet influenced the time of VO in F344 rats (i.e., diets with higher phytoestrogen contents resulted in consistently earlier VO times in this rat strain). However, data from study I for the four diets with comparable ME (3.10–3.15 Kcal/g) indicated that S-D rats showed no difference in time of VO, even though the phytoestrogen contents of the diets were significantly different. Again, these results indicate the relative insensitivity of the S-D rat to dietary phytoestrogens compared with the CD-1 mouse and the F344 rat.
The high variability in phytoestrogen content of commercial diets, evident from the differences in total isoflavone content of the three different mill dates of the same PMI diets tested, means that it would be difficult to obtain reproducible results in hormonal studies between different laboratories and, for that matter, within the same laboratory over time. Currently, few diets are certified for the phytoestrogen content. Compounding the problem is the fact that different rodent species or strains show differing responses to different diets. Using the least sensitive rodent strain or the wrong diet may lead to inaccurate results when assessing the estrogenicity of a substance. As early as 1987, we (Thigpen et al. 1987a
) reported that rodent diets significantly differ in estrogenic activity and concluded that a “standardized diet” with minimal estrogenic activity should be used when comparing the effects of estrogenic compounds. Our findings presented here and earlier (Thigpen et al. 2003
) establish the importance of using a standardized phytoestrogen-free diet with low ME levels (3.0–3.1 Kcal/g diet) for the VO and uterotrophic bioassays to enhance the sensitivity of the assay.
Another critical factor to consider when selecting rodent species/strain for conducting VO bioassays is the variation in baseline data in the mean time of VO when animals are fed different diets with variable levels of phytoestrogens and ME. For example, our data confirm that the F344 rat reached puberty later than the S-D rat. The difference in the mean time of VO between F344 rats and S-D rats was 6.2 days when animals were fed the PMI 5K96 low-ME control diet and approximately 7.2 days when they were fed the AIN-76A high-ME control diet. In study III, the variation in the mean time of VO in F344 rats fed 12 different diets was approximately 10 days. This wider dynamic response window in the mean time of VO in the F344 rat suggests that using this model provides a greater opportunity to detect a weak estrogenic response to weak-acting EDCs than does the S-D rat.
Differing opinions exist regarding the choice of optimal diet and rodent strain for the uterotrophic bioassay used in the OECD program designed to evaluate the estrogenic activity of approximately 87,000 potential EDCs (Kanno et al. 2003b
; Owens et al. 2003
). The rat and mouse have been routinely used in uterotrophic bioassays for years. The OECD validation studies were performed primarily using S-D or Wistar rats. This was based on an understanding that both species are expected to be equivalent, and therefore one species should be acceptable for the worldwide validation in order to save time and money. The OECD proposed that it was acceptable to use diets with up to 350 μg/g TGE in this testing program when conducting uterotrophic assays in ovariectomized or immature S-D or Wistar rats (Kanno et al. 2003b
; Owens et al. 2003
). The OECD guidelines (OECD 2006
) state that in some cases mice may be used instead of rats. Thus, modification of the protocol may be necessary for mice because the food consumption of mice on a body weight basis is higher than that of rats. Therefore, the phytoestrogen content of the diet should be lower for mice than for rats (Owens et al. 2003
; Thigpen et al. 2002
). We have shown that the S-D rat is clearly less sensitive to estrogens. It is difficult to comprehend why an important testing program such as the OECD would adopt an assay using an animal species that is relatively insensitive at detecting estrogen activity, and at the same time would compound the problem by using a test diet with a significant level (< 350 μg/TGE) of background phytoestrogens. The diet seems especially problematic because of the wide batch-to-batch variability in the phytoestrogen content of rodent diets and the fact that vendors of these diets presently do not assay for known dietary estrogen and the phytoestrogens. Establishing threshold values for immature and adult mice and rats or setting limits, even < 350 μg/g TGE, for phytoestrogen content in the diet is not feasible given that it is impossible to manufacture diets that have constant levels of phytoestrogens. Any program designed to determine the estrogenicity of a chemical should use the most sensitive assay possible. This is an important consideration because some EDCs may have very long half-lives and thus, even with low-exposure levels, these compounds could accumulate in tissues. Our data on timing of the onset of VO confirms that the S-D rat is less sensitive to phytoestrogens than either the CD-1 mouse or the F344 rat. Therefore, it seems logical that one of the latter two strains would be a more appropriate model to ensure a more sensitive VO or uterotrophic assay. Although the proposed OECD acceptable dietary level of TGE (350 μg/g diet) (OECD 2006
) may not have profound effects on uterine growth in the S-D rat, there is sufficient published data to show that phytoestrogens do have other genomic and nongenomic effects in this species (Brown and Setchell 2001
). Furthermore, the use of diets containing phytoestrogens would not be appropriate for other strains of rats or for mice. In the VO or uterotrophic assays it is imperative that the test rodents be fed a standardized diet essentially devoid of phytoestrogens or one that has extremely low levels, because this will maximize the sensitivity of the assay. Based on our findings, we have suggested that the diet should contain no more than 20 μg/g TGE (Thigpen et al. 2004b
), which is approximately at the detection limit (10 μg/g diet) of most HPLC assays for phytoestrogens. Additionally, the diet should ideally have a low level of ME, approximately 3.1 Kcal/g diet, because higher levels, independent of the phytoestrogen content, can influence uterine weight (Odum et al. 2004
; Thigpen et al. 2002
) and time of VO.
The rationale for using the ovariectomized female rat or mouse in the uterotrophic bioassay is to reduce the levels of endogenous estrogens to an absolute minimum and to increase the sensitivity and reliability of the uterotrophic assay. It is therefore difficult to understand why similar efforts to minimize exogenous sources of estrogenicity in the diet have not been adopted. Eliminating dietary sources of phytoestrogens as much as possible and using a “sensitive” rat or mouse strain when conducting the uterotrophic bioassay would serve to increase the accuracy, sensitivity, and reproducibility of the assay. Additionally, this would greatly increase the ability of researchers to compare studies across time and within or between laboratories.
Currently, most experimental animals are fed a range of diets with variable concentrations of phytoestrogens and energy levels. When purchasing rodents from different animal vendors, the investigator has little or no knowledge of the diet or the concentration of phytoestrogen in the diet used during gestation, weaning, and prior to delivery of the animals to the research laboratory (). Studies indicate that exposure early in life and during gestation to levels of phytoestrogens typically found in most commercial rodent diets alters the sensitivity of rodents to carcinogens (Cotroneo et al. 2001
; Lamartiniere et al. 2002
) and also influences gene expression and phenotype (Dolinoy et al. 2006
; Naciff et al. 2004
; Wang et al. 2005
). One example of this is the Agouti mouse in which exposure to genistein during pregnancy leads to changes in gene expression that alters coat color of the offspring. Eliminating sources of phytoestrogens in the diet fed to animals during gestation and in the pre- and post weaning periods would lead to greater consistency in experimental designs. We recognize that removing significant sources of phytoestrogens from the usual diet for some animal models may significantly influence biochemical, molecular, and genetic markers, leading to a changed phenotype that more closely resembles that seen before soy meal was used to formulate rodent diets. This will result in a need to reestablish baseline characteristics of the animal model, and for the VO end points or the uterotrophic bioassay it will undoubtedly improve the sensitivity of the assay. Currently, most manufacturers of rodent diets have commercially available diets that are formulated by omitting soybean meal and alfalfa meal, and thus contain only trace levels of phytoestrogen. In the future, animal vendors rearing and supplying animals should consider using diets that contain only trace levels of phytoestrogens, especially for use in estrogenic studies or for studies that measure estrogen-responsive elements as the end points. For testing of EDCs using the uterotrophic bioassay, it is logical to eliminate, as much as possible, known sources of phytoestrogens from the diet. Furthermore, to maximize the sensitivity of the assay, consideration should be given to the use of animal species or strains that are the most sensitive to dietary estrogenic substances. In this regard, we have clearly shown that the S-D rat is not the most sensitive rodent for such testing.
Figure 10 Future dietary considerations. Currently most experimental animals are fed at least two different diets containing variable levels of phytoestrogens and ME. Diet(s) should contain higher levels of ME for gestation and growth and lower levels of ME for (more ...)