Response to comments on: Perinatal exposure to a glyphosate-based herbicide impairs female reproductive outcomes and induces second generation adverse effects in Wistar rats

MARÍA MERCEDES MILESI; VIRGINIA LORENZ (PRIMERA AUTORÍA COMPARTIDA); PABLO M. BELDOMENICO; STELLA VAIRA; JORGELINA VARAYOUD; ENRIQUE H. LUQUE

ARCHIVES OF TOXICOLOGY.

SPRINGER

Lugar: Berlin; Año: 2019 vol. 93 p. 3635 - 3638

0340-5761

Dear Editors,We thank Prof. Plewis and Prof. Paumgartten for their interest in our article and their comments related to the application of statistical analyses accounting for litter effects in both first- and second-generation rats (Plewis 2019; Paumgartten 2019). To address this issue, we have re-analyzed our data using Linear Mixed Models (LMMs), which allow accounting for the lack of independence of observations from the same litter (Lazic and Essioux 2013). Two sets of LMMs were constructed using the statistical software R version 3.6.1 (The R foundation for statistical computing, https://www.r-project.org/), package lme4, functions lmer and glmer (Bates et al. 2015). In the first set, the response variable was the ´pre-implantation loss rate of F1 dams´, the fixed effect was ´treatment´ (three-level categorical variable), and the random intercept included to adjust for the litter effect was the ´ID of the F0 dams´. In the second set of LMMs, the response variables were the feto-placental parameters of F2 offspring (fetal length, fetal weight, placental weight, and placental index), the fixed effect was ´treatment´, and the random intercept included to adjust for the litter effect was the nested random effect ´F1 dam ID/F2 dam ID´. In addition, variables deemed a priori to be potential confounders were individually included in both sets of models to adjust for their influence on the relationship of interest (McNamee 2005), and were removed from the model if they did not contribute to the model´s goodness of fit (GOF). The potential confounders assessed were: food consumption of F0 dams during treatment, weight of F0 dams at the beginning of pregnancy, number of F1 pups per litter, F1litter weight at birth, F1 female pup weight at birth, age of F1 females at the time of becoming pregnant, number of F2 pups per litter. The results of our statistical re-analysis are shown in Tables 1 and 2. In agreement with our previous reports (Milesi et al. 2018), we found a significant increase in the rate of pre-implantation loss in the GBH-HD group compared to controls. None of the potential confounders assessed improved the GOF of the LMMs (Table 1).Table 2 shows that the GBH-HD treatment induced a significant decrease in fetal weight and length, as well as a significant increase in placental index, which are in accordance with our previous results reported in Milesi et al. (2018). On the other hand, the GBH-LD treatment induced lower fetal weight compared to the control group, in agreement with our previous analysis, and a trend to lower fetal length of borderline significance (p-value = 0.056262). Regarding fetal weight and length variables, an improvement in the GOF of the LMM was observed when including the confounding factor ?number of F2 pups per litter?.In addition to the above mentioned feto-placental parameters, we have re-analyzed the data from the occurrence of structural congenital anomalies in GBH-HD F2 offspring, taking into account the litter effect. In accordance with our previous results, we found a significant difference using Fisher?s exact test (p-value = 0.008791). It is worthy to highlight, as mentioned in our article, that fetuses with anomalies came from different F1 mothers, which in turn came from different F0 dams (Milesi et al. 2018).In summary, the re-analysis of our data using LMMs shows that the differences previously detected in both fertility (first- generation) and fetal parameters (second- generation) between control and GBH-exposed groups are mostly maintained, and therefore do not significantly affect the conclusions drawn in Milesi et al. (2018). It is important to acknowledge, however, that ignoring ´litter effects´ and other sources of lack of independence in the statistical analyses is very common in the biomedical literature, and that proper analysis to control such phenomenon should replace the traditional statistical approaches whenever they are inappropriate.The following responses correspond to issues raised from Prof. Paumgartten (2019). Prof. Paumgartten stated that ?the conclusion that GBH increased the number of preimplantation losses in F1 females, for instance, may not be true? since ?uteri were not immersed in ammonium sulfide (10%), or Salewski?s method, to visualize the implantation sites (IS)?.The method we used to evaluate reproductive performance in rats is an old and very well-known method; we have used this protocol in many of our previous publications, and also it has been applied by several other researchers (as examples Borges et al. 2017; Hu et al. 2011; Lima de Barros et al. 2016; Xie et al. 2018). The aim of our work (Milesi et al. 2018) was to perform an integral evaluation of both fertility and feto-placental parameters in the same rat. Under this assumption, the application of Salewski?s method (Salewski 1964) is not compatible with the evaluation of some feto-placental parameters (such as placental index and placental weight), because, during surgery, the uterine tissue must be cut to quickly and carefully remove fetuses and placentas. Both the results obtained in our study (Milesi et al. 2018) and others published from our Lab increase the evidences that low doses of GBH exposure during critical developmental periods cause adverse effects on uterine development and female reproductive capability. Some of these evidences include that 1) postnatal exposure to GBH during the first week of life disrupts uterine morphology and differentiation in prepubertal rats (Guerrero Schimpf et al. 2017), 2) using the same experimental model, adult pregnant females exposed to GBH show a significant increase in the number of fetal reabsorptions associated with an altered decidualization response on gestational day (GD) 9 (Ingaramo et al. 2016; Ingaramo et al. 2017), 3) exposure to GBH enhances the sensitivity of the rat uterus to estradiol and induces histomorphological and molecular changes associated with uterine hyperplasia (Guerrero Schimpf et al. 2018), 5) applying the same exposure model as in Milesi et al. (2018), GBH epigenetically deregulates the expression of the uterine estrogen receptor alpha gene during the pre-implantation period, which could be related to GBH-induced implantation failures (Lorenz et al. 2019), and 6) in a different animal model, the ewe lamb, neonatal exposure to GBH induces histomorphological changes in the uterus (Alarcón et al. 2018), which may affect the normal development of the organ. All these published results support the fact that exposure to low doses of GBH in a sensitive window of development affects the functional development of the uterus, with long-term detrimental consequences on female reproduction. Together, our results also stress the importance of assessing different species, doses, administration routes, and windows/lengths of exposure since effects on fertility may differ. In relation to our animal model, the inbred Wistar-derived strain is generated and maintained in our Lab. We periodically perform rigorous controls of developmental and reproductive parameters in order to keep a data record and select our animals. Several features such as feto-placental parameters, incidence of congenital anomalies, number of live pups, ano-genital distance, vaginal canal opening, estrous cycle, reproductive performance, incidence of tumors, etc. are examined and recorded periodically. As stated in our article, structural congenital anomalies have never been detected in our own animal facility, suggesting that the developmental disorders found in F2 offspring from GBH-exposed females are not the result of spontaneous changes. We agree with Prof. Paumgartten on the fact that one cannot rule out the possibility that the observed effects were due to co-formulants of the GBH formulation other than glyphosate. For that reason, in the discussion section of our article, we clearly stated, ?we cannot ascribe the reproductive and developmental adverse effects found to glyphosate, the co-formulants or both acting together?. Understanding the contribution of both glyphosate alone and/or co-formulants to GBH effects is a matter of high relevance, and for that reason, we are currently working to give light to the subject.After re-analyzing our data and considering all the comments raised by Profs. Plewis and Paumgartten, we reaffirm the conclusions of our work (Milesi et al. 2018), in which we found that perinatal exposure to a GBH formulation impairs reproductive performance in F1 females rats and induces fetal growth retardation and structural congenital anomalies in their progeny (F2 generation).Finally, we would like to express our conviction that no paper is the ´absolute´ truth, but instead a small contribution to approximate that truth. On this basis, and taking into account the interest on the subject shown by Profs. Plewis and Paumgartten, we invite them or other researchers to reproduce the experiments in their labs, for which we offer our strain of Wistar rats. Repeating experimental results by different groups allows science to progress, adds robustness to the evidences and allow being more certain about a hypothesis.