Exposure to polyoxyethylene tallow amines (POEAs), glyphosate co-formulation surfactants, in a US pregnant population and their potential endocrine disrupting effects.

The current study is based on a sub-sample of The Infant Development and the Environment Study (TIDES), a multicenter study that collected data between August 2010 to August 2012 to investigate associations between prenatal exposure to endocrine disrupting chemicals (EDCs) and reproductive development. A detailed description of this sub-cohort can be found in our previous publication 5 , 16 . In brief, pregnant participants were recruited from four university-based medical centers: University of Minnesota (UMN), University of Washington (UW), University of Rochester Medical Center (URMC) and University of California, San Francisco (USCF). Eligibility criteria for recruitment included 18 year or older pregnant persons in their 1 st trimester (< 13 weeks of gestation) without any medical complications who are able to read and write in English (or Spanish in UCSF) and planned to deliver at one of the four study centers. Questionnaires were administered to collect their health and reproductive history as well as their demographic and lifestyle information. In each trimester, their urine samples were collected using poly-propene cups. Specific gravity (SpG) of urine was measured within 30 minutes of collection using a hand-held refractometer (National Instrument Company, Inc., USA). Samples were stored at −80°C. For each infant, two ano-genital distance (AGD) measurements were collected after birth; the ano-penile (AGD-AP) and ano-scrotal (AGD-AS) distances for males, and ano-clitoral (AGD-AC) and ano-fourchette (AGD-AF) distance for females, described further in 17 , 18 Each center and coordinating center’s Institutional Review Boards approved all study protocols. All participants signed informed consents prior to the start of the study. In our previous study 5 , 94 participants who had both second trimester urine samples and birth exams were randomly selected from the TIDES cohort. The current study includes 86 participants from the same sub-cohort who have sufficient urine sample (>100 uL) for POEA exposure assessment. We measured concentrations of three POEA homologs with two ethoxylate (EO) units, i.e. POE (2) in maternal urine. We purchased the POE (2) analytical standard (CAS 61791-44-4) from AA Blocks Inc. (San Diego, CA, USA). The chemical formula for POE (2) is C z (s/u)EO x , where z is the number of carbon atoms in the alkyl chain, s/u is the tallow moiety and can be either saturated (s) or unsaturated (u), and x is the total number of (EO) units from the two ethoxylated chains. 19 Values of x usually range from 2 to 15 resulting in z different homologs. 20 In this study, the three POEA homologs measured are C 16 s(EO) 2 , C 18 u(EO) 2 , C 18 s(EO) 2 . We developed and applied a low-volume, sensitive, and selective method for the determination of POE (2) homologs in urine, using solid-phase extraction (SPE) for sample clean-up and targeted isotope-dilution liquid chromatography in conjunction with triple quadrupole tandem mass spectrometry (ID-LC-MS/MS) for quantification. We used a urine multiclass assay 21 , modifying it based on published methods for quantifying POE (2) homologs at biologically relevant concentrations in human urine 19 , 22 and plasma. 23 Detailed experimental protocol can be found in supplemental materials ( Supplement ). All values of analytes below LOD (0.1 ng/mL) were replaced with value of LOD/√2, a common imputation method used for non-detectable measurements 24 . To account for urinary dilution, we adjusted concentrations of all analytes for specific gravity (SpG) using the formula: SpG-adj-Analyte =Analyte [MeanSpG −1/ (SpG-1)] where SpG-adj-Analyte is the SpG adjusted analyte’s concentration (ng/mL), and analyte is the observed concentration and MeanSpG is the mean SpG for all the samples. We previously published the method and levels of glyphosate and AMPA on the urine samples of the current study 5 , 16 . In brief, ultra-high-performance liquid chromatography-tandem with mass spectrometry (UPLD-MS/MS) was used and the LOD for glyphosate and AMPA was 0.014 ng/mL and 0.013 ng/mL, respectively. Measurements were performed at the Collaborative Center for Translational Mass Spectrometry (CCTMS) (TGen, Phoenix, AZ). We also report GLY concentrations among 328 females, between 18 to 45 years of age from the NHANES Glyphosate Urine (Surplus) (2017-2018) dataset as comparison.

This study included 86 participants who were randomly selected from the TIDES cohort contingent on availability of 2 nd trimester maternal urine samples and birth exams. The demographic characteristics are summarized in Table 1 . The average age of the participants was 31.2 years (range: 18.5-43.8 years) and they were predominantly white (77.9%). Maternal urine samples were collected on average at 21 weeks of gestation (range: 14.9-28.9 weeks) and among the participants 70.9% were college graduates or more. For the 86 infants (40 females and 46 males), the mean birthweight was 3.43 kg (range: 1.65-5.44 kg) and the average gestational age was 39.2 weeks (range 32.9-42.3 weeks). Infant characteristics are also shown in Table 1 . The distribution of the POEA homologs, C 16 s(EO) 2 , C 18 u(EO) 2 , and C 18 s(EO) 2 , are shown Figure S1 and summary statistics are reported in Table 2 . We selected these homologs with two (EO) units to be analyzed because they are most likely to be excreted in urine. These homologs were detected in the majority of the urine samples; the proportions of detectable samples (LOD of 0.10 ng/mL) were 79.1% for C 16 s(EO) 2 , 96.5% for C 18 u(EO) 2 and 59.3% for C 18 s(EO) 2 , with corresponding mean concentrations of 1.87 ng/mL (range: 0.00-10.98 ng/mL), 1.88 ng/mL (range: 0.03-12.99 ng/mL) and 1.65 ng/ml (range: 0.00-9.65 ng/mL), respectively. From our previous study 5 , we also measured GLY and AMPA levels on these samples where 95.3% and 93.0% of samples had detectable levels of glyphosate (LOD of =0.014 ng/mL) and AMPA (LOD of 0.013 ng/mL) with corresponding mean concentrations of 0.36 ng/mL (0.00 – 3.30 ng/mL) and 0.43 ng/mL (0.00 - 11.69 ng/mL), respectively. Our study population’s GLY levels are comparable to NHANES 2017-18 surplus dataset, where in women of reproductive age, mean concentration of GLY is 0.36 ng/mL (range: 0.00-5.25 ng/mL) even at a higher LOD of 0.1ng/mL Spearman correlations among analytes are illustrated in Figure 1 . The three POEA homologs were highly correlated with each other with Spearman coefficient (rho) ranging from 0.73 to 0.87. While GLY and AMPA showed strong correlation (rho =0.67) with each other, their correlations with POEA were much weaker (rho = 0.26 - 0.44). Given the sources of individual POEA homologs and their biological activity is unknown and the POEA homologs are highly correlated with each other, we summed the three individual POEA homologs as a single exposure variable (POEA sum ) for subsequent analysis. The mean concentration of the POEA sum was 5.39 ng/ml (range: 0.03 - 26.1 ng/mL) in our study population. We also explored associations between POEA sum levels and demographic variables of our study population ( Figure S2 , Table 3 ). POEA sum levels did not differ by infant sex, study center, maternal education, season or time of urine collection. However, levels of POEA sum differed by race ( p = 0.047) where White participants had the lowest mean POEA sum levels (4.71 ng/mL), in comparison to Black participants (7.28 ng/mL) and others (8.99 ng/mL). We previously reported associations between urinary GLY/AMPA and AGD measures in the offspring of our study population 5 ; herein we explored similar associations of infant AGD with POEA levels. We observed positive associations between POEA sum and AGD-AP in male babies (β=2.46, p =0.057) in a logistic regression model adjusted for age at exam and weight-for-length Z-score ( Table 4 ). Such relationship remained unchanged after adjusting for GLY (β=2.45, p =0.055) or AMPA (β=2.46, p =0.058) or ( Table 4 ). No significant association was seen in female infants ( Table 4 ).

The current study is the first report on human exposure to POEA in a human population. The study subjects are participants of a US-based multi-center pregnancy cohort recruited across the country. The exposure was assessed by analyzing three POEA homologs with two (EO) units (i.e. C 16 s(EO) 2 , C 18 u(EO) 2 , and C 18 s(EO) 2 ) in maternal urine collected at second trimester. We focused on short-chain POEAs, such as POE (2), because POEAs with shorter (EO) chains are more frequently detected in human plasma than those with longer chains 23 . Based on POE (2) homologs’ structural similarities, we speculate that POE (2) would have a similar elimination route as aliphatic alkyl polyethoxylates (a.k.a. alcohol polyethoxylates) which are known to be quickly excreted to urine 25 . Additionally, POEAs with shorter (EO) chains are more likely to be added to GBHs and like alkyl polyethoxylates in humans, long-chain POEAS may go through non-oxidative metabolism to form shorter EO chains 23 . Given ubiquitous exposure to GLY or GBH in the general population 8 as well as in our study population 5 , 16 , it is not surprising that all three POEA homologs were detected in majority of our samples (range: 59.5% – 96.6%). Because of limited information regarding toxicity of individual compounds, persistence of individual POEA homologs and level of correlation, a summation of the three POEAs was used to create a single variable, POEA sum . The mean concentration of POEA sum was at 5.39 ng/mL in our population. In comparison to the concentrations of GLY and AMPA, i.e. 0.36 ng/mL and 0.43 ng/mL, in the same samples, the level of POEA appears to be ~10 fold higher. Given that these three POEA homologs only constitute a fraction of the POEA, it is likely the total POEA levels would be higher in urine. While the three individual POEA homologs were highly correlated with each other (rho > 0.7), the correlations of individual POEA and POEA sum were weakly correlated with GLY and AMPA with rho of 0.26 and 0.40, respectively. However, these results may not necessarily translate to exposure levels; instead, they may reflect differences in pharmacokinetics in terms of absorption and disposition between GLY and POEA. We did not observe significant differences in POEA levels with respect to locations of the study, seasonality, timing of the urine collection, education of the participants, etc. Instead, we observed a racial difference in POEA levels in our study where the lowest level was observed in White participants (4.38 ng/mL) compared to Black participants (6.72 ng/mL) and others (9.92 ng/mL). However, no such racial differences were apparent for GLY or AMPA in the same population. A report from the most recent National Health and Nutrition Examination Survey (2013–2014) also did not report significant racial differences for GLY 8 . The reason for racial difference of POEA levels in our study population is unknown and it may be a result of both physiological and/or lifestyle factors. Numerous studies have demonstrated greater toxicity of GBH compared to pure GLY and POEAs appear to contribute much of the observed toxicity (reviewed in 26 ). A review of GLY by Munoz and colleagues 27 suggest that GLY is an endocrine disrupting chemical (EDC) as it processes eight out of ten key characteristics proposed in La Merrill et al 28 . Ano-genital distance (AGD), is a sexually dimorphic measurement, increasingly used in human studies as an early-life biomarker of fetal androgen exposure where shortened AGD in male offspring reflects decreased in utero androgen exposure whereas in female reflects increased in utero androgen exposure 29 , 30 . Children with hypospadias and cryptorchidism have significantly shorter AGD 31 - 33 . Males with shorter AGD have been associated with poorer semen quality 34 , 35 and reduced testosterone and testicular volume 36 . Females with longer AGD has been associated with increased androgenization symptoms including congenital adrenal hyperplasia 37 , multifollicular ovaries 38 polycystic ovary syndrome (PCOS) 39 , and shorter AGD with endometriosis 39 . In the current study, we observed a positive trend in the association between POEA sum and AGD (i.e. AGD-AP) in male offspring (β=2.46, p =0.057) and this observation remains unaltered when adjusted for GLY (β=2.45, p=0.055) or AMPA (β=2.46, p=0.058). Surprisingly, this finding contradicts our previous results in the same population where urinary GLY and AMPA levels were positively associated with AGD in female offspring only 5 . These findings suggest that POEAs are potential EDCs but their biological activities may be different from those of GLY and AMPA; more studies are needed to evaluate developmental toxic effects of POEAs on endocrine and reproductive systems. As with most EDCs, although phenotypic changes associated with these chemicals may be subtle, the public health implications can be substantial given the prevalent exposure and potential for lifetime and inter-generational impacts. Our study has multiple strengths. Our environmentally exposed multicenter study of normal-risk pregnant people, i.e. TIDES, is geographically and demographically diverse, thus increasing the potential generalizability of the study. The parent study was designed to examine relationships between several environmental exposures in utero in relation to reproductive outcomes where study protocols were standardized to collect reliable AGD measurements. The study population is well characterized so that potential bias can be evaluated and minimized. We also acknowledge limitations of our study which is exploratory by nature given the small sample size and limited power. A single second trimester urine sample was used, which may not reflect the POEA exposure throughout life or during pregnancy, especially given the limited information on pharmacokinetics of POEA. However, as we observed ~10 times higher POEA levels in urine compared to GLY and AMPA may partially attributed to longer half-lives of POEA. It points to the possibility that urinary POEA level may be a better biomarker for long-term GBH exposure. Another limitation is that we used the sum of only three POEA homologs as proxy for total POEA. Although we expect these POEA homologs with short (EO) chain to be most likely excreted via urine, whether their concentrations reflect the total POEA exposure is challenging to evaluate. Lastly, we don’t have direct measurement of the source of POEA exposure as POEA surfactants may be used in other pesticides. Although, we suspect food consumption to be the primary route of exposure because our study subjects come from a non-agriculturally exposed population (who likely consume food produced nation- or even world-wide) and displayed no geographical difference. Additionally, consumer foods currently are not being assessed for pesticide co-formulants beyond active ingredients. Our study has significant implication in the public health debate on the biosafety of GBH. In 2016, with emerging evidence of broad range POEA toxicity to the eco-system as well as to potential human health, the EU Commission recommended to Member States that POEA-type co-formulants be banned from use. As a result, manufacturers of GBHs in Europe have been replacing POEA co-formulant with other surfactants, but the usage continues in the US and other part of the world 12 . It is challenging to evaluate human exposure to POEA because the identity and concentration of the co-formulants in GBH are rarely disclosed as they are considered confidential information by industry and protected by regulation. Without access to neither compositional information nor structural characteristics of POEA surfactants, evaluation of their toxicity and ultimate impact on human health is impossible. In conclusion, in a normal risk pregnant population, we reported prevalent exposure to POEA surfactants found in GBH, at a urinary concentration higher than the active ingredient, GLY. We also observed an indication of a positive association between POEA and AGD in male offspring, suggesting possible endocrine disruptive property of POEA. Given that human exposures to glyphosate are ubiquitous and increasing and POEA continues to be used in the US and globally, biomonitoring POEA in the general population should be implemented. POEA’s possible impacts on endocrine/reproductive and other health outcomes, as well as possible programming consequences to lifelong health, warrants timely investigation.

We calculated univariate summary statistics for each of the exposure variables (e.g., C 16 s(EO) 2 , C 18 u(EO) 2 , C 18 s(EO) 2 ) and examined their distributions using histograms and scatterplots. We used Spearman correlations to assess relationships between these analytes. Individual levels of the three POEA homologs were summed up into POEA sum as a single exposure variable because these were highly correlated. We performed bivariate analyses to explore relationships between exposures and covariates of interest using Spearman correlations for continuous variables and Kruskal-Wallis test for categorical variables. Correlation plots, scatterplots and box plots were used to visualize these results. We also examined AGD’s relationship with POEA sum using linear regression models. POEA sum exposure was examined as dichotomous categorical variable where high is above the population median and low is below the median. We fitted both unadjusted and adjusted linear regression models including age at exam, weight-for-length Z-score, glyphosate and AMPA. Since AGD measures are sex-specific, analyses exploring relationships between exposures and AGD were stratified by infant sex.

Glyphosate [GLY; N-(phosphonomethyl)glycine]-based herbicides (GBHs), most notably Roundup ® , are the most widely used pesticides in the world 1 . Agricultural uses of GBHs have risen drastically since 1996, when genetically modified GLY-tolerant crops were introduced 1 , 2 leading to a dramatic increase of detectable GLY/AMPA levels in urine worldwide 3 , 4 , including pregnant women 5 , 6 and children 7 . In the most recent NHANES report (2013-2018 cycle), GLY was detected in 70.0-81.7 % of the US population 8 , 9 . Emerging evidence links GLY and GBHs to a myriad of adverse outcomes in both ecosystem and human health 10 . Our team previously reported a positive association between urinary GLY levels and ano-genital distance (AGD) 5 which is a marker of gestational androgen exposure 11 . However, exposure to pure GLY alone is unlikely, since GLY is applied to crops in GBH formulations that contain adjuvant ingredients. The major co-formulants in GBHs are surfactants that make up 5–15% of concentrated products by weight 12 . The most common surfactants in GBH formulations over the last 40 years have been polyethoxylated tallow amine amines, i.e. POEAs 12 . POEA surfactants are usually synthesized from fatty acids, derived from animal fat and then ethoxylated to tallow amines by adding ethylene oxide. The resulting product is a mixture of tertiary amines are consisting of one alkyl chain (R) primarily composed of saturated or unsaturated carbon moieties of 16 or 18 carbon in length (e.g. C 16 H 33 , C 18 H 35 , or C 18 H 37 ) and two ethoxylate chains containing 2 to 15 (EO) unit 13 . These co-formulants are considered “inert” even though their toxicity is not well established or studied. Emerging evidence suggests that GBH is more toxic than GLY alone 14 , 15 . However, it is challenging to tease out the effect of GBH adjuvants from its active ingredient (i.e. GLY) in population studies. Under US federal law, adjuvants are classified as “inert” additives in GBHs as they enhance the herbicides effectiveness without directly causing herbicidal effects; thus their identities are not required to be labeled on commercial products such as Roundup ® . The ambiguous labelling and lack of established bioassays for these compounds leads to POEA being ignored by herbicide safety evaluations and risk assessments. In the current study, we leverage samples from our previous study where we reported urinary GLY levels in pregnant women and a positive association with AGD in their offspring 5 . We used a sensitive and quantitative assay to measure urinary levels of three POEA homologs that are most likely to be excreted via urine and explored their associations with AGD.