Developmental exposure to phytoestrogens found in soy: New findings and clinical implications.

Historically, there have been few epidemiological studies investigating the impact of infant soy formula consumption during developmentally sensitive windows. As exposure to soy formula occurs during infancy, it is difficult to connect these exposures with reproductive health challenges that manifest during adulthood. There is little to no evidence that genistein or other phytoestrogens in the postnatal period significantly alter human male reproductive system development, even though it can be impacted by some EDCs ( Bonde et al. , 2016 ; Chin et al. , 2021 ; Patisaul, 2021 ). In contrast, there is accumulating evidence that development of the female reproductive system is sensitive to postnatal exposure to phytoestrogens present in soy-based infant formulas. For this reason, we focused this review on epidemiological studies relevant to female reproduction. The studies described below cover a variety of observational, functional, and molecular endpoints and are grouped by functional impact and ordered by age of the study population. To clarify our language regarding sex and gender, in places where the short-hand term “female” or “male” is used, we are specifically referring to assigned biological sex, generally based on evaluation of external genitalia on prenatal ultrasound or at birth in infant studies, or evaluation of external genitalia and internal reproductive organs during enrollment in adult studies. We make this clarification to prevent unintentionally gendering infants who have not yet had the opportunity to determine their gender identity and expression, and adults who are intersex and/or may define their gender identity and expression differently than assigned sex, but who could still be affected by the exposures and reproductive outcomes discussed. We would like to use more precise language, but as a review of existing literature, we are not sure how sex was defined across all individuals in these studies. This language is not perfect and does not adequately capture the diversity in biological sex. This is an ongoing problem in the field of reproductive biology. Although initial studies suggested that soy formula feeding was not associated with altered female reproductive development, breakthroughs in assessing the impacts of soy formula on reproductive development have come from looking at developmental trajectories instead of observing isolated time points. For example, studies of the development of reproductive organs at a single timepoint found no significant difference in breast bud, uterine, or ovarian volume at 4 months of age in female infants fed soy formula versus cow’s milk formula or breast milk ( Andres et al. , 2015 ; Gilchrist et al. , 2010 ). However, a longer term study comparing breast bud size in breast fed, cow’s milk formula fed, and soy formula fed infants showed no significant difference in breast bud size during the first year of life, but in the second year of life, soy formula fed infants failed to show the decline in breast bud size that was observed in both breast fed and cow’s milk formula fed infants ( Zung et al. , 2008 ). These conflicting findings highlight the point that the time windows of observable outcomes vary over the course of development and depend on the selected organs and developmental markers. A pilot study sought to address the issue of developmental time windows and the impact of environmental estrogens by determining the trajectory of several reproductive developmental markers from less than 48 hours after birth through six months of age ( Bernbaum et al. , 2008 ). Breast bud tissue reached a maximum size and vaginal cells showed maximum estrogenization at birth and both declined with age. However, genital development showed no change with age over the observed time window. Female infants fed soy formula showed re-estrogenization of responsive tissues at 6 months, whereas those fed cow’s milk formula or breast milk did not. This study demonstrates the utility of the trajectory approach and determined that six months of age was a time when tissue-level effects of soy formula exposure could be observed. This trajectory approach was used in the Infant Feeding and Early Development (IFED) Study, which followed 283 infants exclusively given breast milk, cow’s milk formula, or soy formula from birth until 9 months to determine the impact of diet on reproductive developmental markers ( Adgent et al. , 2018 ). This study differed from previous studies not only in using the trajectory approach, but also in the rigor with which feeding groups were established and verified by testing urinary isoflavone levels. There was not a significant difference in breast bud diameter, serum estradiol, or follicle stimulating hormone trajectories between females fed cow’s milk formula and soy formula. However, relative to both breastfed and cow’s milk formula-fed infants, females fed soy formula had a higher vaginal cell maturity index and a slower decrease in uterine volume ( Fig. 2 ). These developmental changes in the reproductive tracts of soy fed female infants were consistent with exogenous estrogen exposure. Although both studies correlated soy formula feeding with altered reproductive tissue development, neither study was designed to determine if these changes could have a long-term impact on tissue function. Long term changes in functional outcomes can be mediated by stable alterations in epigenetic marks such as DNA methylation. To determine if there were persistent epigenetic changes in reproductive tissues induced by soy formula feeding, DNA methylation was examined in vaginal epithelial cells collected from female infants in the IFED study ( Harlid et al. , 2017 ). Epigenome-wide methylation analysis of vaginal cell DNA revealed significantly altered methylation in soy formula fed versus cow’s milk formula fed female infants at CpG sites near the transcription start site of proline rich 5 like ( PRR5L ), an estrogen-regulated gene containing three estrogen receptor alpha (ERα) binding sites. Longitudinal analysis of the two most differentially methylated sites indicated that mean methylation levels in the two feeding groups were similarly increased at birth but that these levels decreased with age in cow’s milk formula fed infants while soy formula fed infants maintained higher mean levels. Although the overall trajectories of the methylation levels did not reach statistical significance, there were statistically significant changes in methylation at each time point measured after 126 days of age. Database analysis showed a negative correlation between methylation at one of these CPG sites and PRR5L mRNA level. This study provides proof-of-principle that stable epigenetic changes influencing gene function occur in response to soy formula feeding during infancy and persist long-term, providing a mechanistic explanation for later alterations in reproductive tissue function. Decreasing age at menarche and health burdens related to menstruation can have a major impact over the course of a female’s life, including earlier menopause, increased risk of breast and endometrial cancer, and psychosocial factors ( Canelón and Boland, 2020 ; D’Aloisio et al. , 2013 ; Mishra et al. , 2009 , 2017 ). Exposure to phytoestrogens may contribute to these adverse outcomes. In a retrospective study of 811 participants, assessment of over 30 self-reported health outcomes in young adults of either sex who were fed soy formula as infants showed few significant changes overall relative to those fed cow’s milk formula ( Strom et al. , 2001 ). However, females fed soy formula as infants reported slightly longer menstrual bleeding and greater menstrual discomfort, but not an increased severity in menstrual flow or alterations in timing of menarche ( Strom et al. , 2001 ). This study was weakened by its retrospective nature and its reliance on self-reporting through phone interview for data collection, although approximately 5% of the interviews were verified by cross-referencing with official medical records with a confirmation rate of approximately 87%. Several more recent studies with larger sample sizes and, in one case, data collected in clinical visits have found that infant soy formula consumption is associated with altered timing of menarche (with increased incidence of both early and late menarche), heavier menstrual flow, and increased adverse impact of menstruation on quality of life ( Adgent et al. , 2012 ; D’Aloisio et al. , 2013 ; Upson, Harmon, Laughlin-Tommaso, et al. , 2016 ). It is important to note that two of these studies were from a cohort of females with a family history of breast cancer, which is associated with early age at menarche, but the authors found that the median age at menarche in their cohort was similar to that found in the National Health and Nutrition Examination Survey (NHANES) 2003–2010 ( McDowell et al. , 2007 ). These studies may have been further confounded by their reliance on self-reporting and lack of accounting for exposures to other estrogens. Additionally, the cohorts in these studies were not designed to be representative of the general U.S. population because two studies used a cohort of predominantly white females, one study enrolled only Black females, and another study used a cohort from the United Kingdom. However, the consistency of the findings across cohorts supports a correlation between infant soy formula consumption and increased severity of menstruation. Together, this body of research suggests that exposure to phytoestrogens through soy infant formula increases the burden of menstruation. Given the association between early menarche and disease risk later in life, the finding of altered age at menarche in particular holds implications for overall reproductive health. Exposure to phytoestrogens may also promote both endometriosis and fibroids, which have high prevalence in the U.S. ( Pavone et al. , 2018 ; Peiris et al. , 2018 ). Endometriosis affects approximately 11% of femalesin the U.S. and the lifetime risk of developing fibroids is even higher at 70–80% ( Baird et al. , 2003 ; Buck Louis et al. , 2011 ). Despite the high prevalence of these diseases, relatively little is known regarding how risk factors influence disease progression. Endometriosis and uterine fibroids are both sensitive to estrogen signaling, and endometriosis is an estrogen-driven disease ( Chantalat et al. , 2020 ; Pavone et al. , 2018 ). Recent research suggests that developmental exposures to phytoestrogens in soy infant formula may increase vulnerability to these diseases. In a study of 340 females diagnosed with endometriosis and 741 undiagnosed population-based controls, infant soy formula consumption was associated with over twice the risk of developing endometriosis relative to unexposed females ( Upson et al. , 2015 ). The increased risk in the soy formula-exposed group was even greater than that following gestating parental DES exposure, which has been linked to endometriosis in several epidemiological studies ( Ottolina et al. , 2020 ). Although the Upson et al. studies relied on self-reporting for exposures, medical records were used to confirm the endometriosis diagnosis and researchers took the extra measure of contacting subjects’ parents or another relative to confirm DES exposure and soy formula feeding. Findings on infant soy formula feeding and uterine fibroids suggest a more complex relationship than with endometriosis. Two earlier studies found that infant soy formula feeding increased the risk of developing fibroids early in life ( D’Aloisio et al. , 2010 , 2012 ). However, a more recent study found no association between infant soy formula feeding and overall prevalence or number of fibroids, but did find an association with larger fibroid size ( Upson, Harmon, and Baird, 2016 ). All three of these studies relied on questionnaires to gather data on early life exposures. The two D’Aloisio et al. studies had larger sample sizes ( D’Aloisio et al. , 2010 , 2012 ). One used a cohort of 19,972 non-Hispanic white females, and the other a cohort of 3,534 Black females. The results of these two studies were compared, and the association between infant soy formula feeding and risk of early-onset fibroids was similar in magnitude between these two groups. The consistency of these results across groups gives more weight to their findings. Both studies, however, relied on self-reporting of a physician diagnosis of fibroids. As fibroids are often asymptomatic, many cases go undiagnosed ( Pavone et al. , 2018 ). D’Aloisio et al. tried to account for this by looking only at early-onset fibroids, thus restricting their case population to those who had reported a diagnosis by 35 years of age when the prevalence of fibroids is lower. The Upson et al. study had a smaller sample size, with a cohort of 1,696 Black females, but used transvaginal ultrasound to screen for and measure fibroids in all subjects. This methodology allowed for more accurate findings than in the previous study. D’Aloisio et al. concluded that their findings suggested an increase in early onset fibroids in females fed soy formula as infants. However, because participants with more severe fibroids were more likely to experience symptoms and thus more likely to seek diagnosis, it is possible that these findings reflect an increase in severity rather than an increase in prevalence. This conclusion would be consistent with the findings in the Upson et al. study that there was no difference in prevalence, but there was an increase in fibroid size in females fed soy formula as infants. In summary, these findings suggest that infant soy formula feeding is not associated with an increased prevalence of fibroids but, instead, is associated with increased severity and earlier onset. Epidemiologic studies have examined some associations between genistein in infant soy formula and several infant and later-life outcomes. Significant progress has been made in the IFED studies by repeatedly assessing reproductive development over time and rigorous feeding group verification. However, epidemiological studies often cannot capture a full life-course perspective or control for all confounding and can be limited, compared to animal studies, by the inability to experimentally manipulate human subjects or collect tissue samples. Studies often rely on self-reported infant feeding and health history for data collection via survey and interviews. Infant diet cannot be randomly assigned, and so demographics and exposures to other exogenous estrogens may vary greatly between groups, making it difficult to control all sources of bias. For example, in the IFED study, selected feeding option varied by gestating parental race. The majority of parents who participated in the study were Black and Black infants comprised the majority of the soy formula or cow’s milk formula feeding groups, as compared to the breastfed group, the majority of whom were white ( Adgent et al. , 2018 ). Finally, reproductive health is influenced by a multitude of factors across the reproductive lifespan, and it is not possible to control for these factors in observational studies, especially additional co-exposures. Limitations to research in this field include the age group being investigated, the timing of enrollment prior to delivery or shortly after birth, and the duration of a study exploring long-term effects such as the impact of soy formula on fertility, which would take over two decades to assess. Efforts to enroll infant cohorts for long term follow-up have expanded in the last 5–10 years and continued follow-up studies will provide valuable data for future generations. Nonetheless, these studies are challenging to design and carry out, and until such studies are done, it is important to utilize information gleaned from well controlled animal studies to help guide conclusions regarding the safety of infant soy formula feeding.

Animal models have been widely used to study developmental estrogenic chemical exposure and mechanisms of female reproductive disease ( Ho et al. , 2017 ). These models provide the ability to control the timing of developmental exposure, probe genetic and epigenetic mechanisms, map altered pathways, and ultimately connect these molecular changes with phenotypic outcomes. An important caveat to keep in mind regarding these studies, however, is that they generally use exposures to purified genistein rather than genistein in the context of other components found in soy. It is highly likely that these other components could modify the outcomes observed using genistein exposure alone ( Badger et al. , 2002 ). Nevertheless, these studies are very informative regarding mechanisms of action and potential risks of genistein exposure, and some of these studies have been validated in the context of animal soy exposure. Numerous studies on developmental estrogenic chemical exposure demonstrate that differences in the timing of exposure result in distinct reproductive tract developmental outcomes and later life disease. Most of the studies described here investigate adverse effects resulting from early life exposure to estrogenic chemicals in animals (prenatal exposure or exposure up to postnatal day 5) rather than acute adult exposure and short-term effects on fertility, fibroids, endometriosis, or endometrial cancer. The window of rodent reproductive tract development that occurs shortly after birth encompasses a period of rapid and extensive changes in the female reproductive tract that are exquisitely sensitive to disruption ( Ho et al. , 2017 ). Exposure to exogenous factors that cause significant reproductive tract alterations are often specific to this window in development, whereas exposures and alterations that occur in adulthood often do not have as severe an effect or any measurable impact at all. A myriad of tightly regulated system-wide factors contributes to normal fertility. Molecular perturbations and/or morphologic abnormalities can impact ovulation, fertilization, implantation, and maintenance of pregnancy. Mice exposed neonatally to genistein (50 mg/kg/day) subcutaneously are unable to deliver live pups ( Jefferson et al. , 2005 ). This dose produces a similar circulating genistein level as that measured in infants on soy formula ( Doerge et al. , 2002 ). In this model, failure to produce either a fertilized embryo or carry the pregnancy to term is attributed to exposure-induced alterations at each major step leading to live birth ( Jefferson, Patisaul, et al. , 2012 ). Initial barriers to fertility include disruption of the hypothalamic-pituitary-gonadal axis, altered estrous cyclicity, changes in ovarian follicle morphology, ovulation failure, and earlier reproductive senescence ( Jefferson et al. , 2002 , 2005 , 2006 ). Superovulation and mating, or embryo transfer can bypass these phenotypes. However, altered uterine morphology, disrupted differentiation and function of uterine glands, decreased embryo implantation, and increased reabsorption culminate in early pregnancy loss ( Jefferson et al. , 2005 , 2011 ). Although the aforementioned phenotypes have been well-known for the past 20 years, the mechanisms underlying these changes remain unclear, likely because they are challenging to investigate individually. Several specific ovarian defects have been reviewed previously ( Jefferson, Patisaul, et al. , 2012 ). In brief, neonatal genistein exposure impacts dissociation of clustered oocytes within “nests” during development into individual oocyte-follicle pairings, instead leading to multioocyte follicles ( Jefferson et al. , 2002 , 2006 ). However, the presence of multioocyte follicles does not appear to impact the number of oocytes collected after superovulation, or the oocyte competency and ability to produce a live birth following in vitro fertilization and blastocyst transfer to pseudopregnant control mice ( Jefferson et al. , 2005 ; Jefferson, Padilla-Banks, et al. , 2009 ). More recently, the focus has shifted to understanding uterine defects following neonatal genistein exposure and their role in the infertility phenotype. Normal morphological and functional development of the uterus is required for uterine receptivity and successful postimplantation establishment and maintenance of pregnancy ( Kelleher et al. , 2019 ; Taylor and Gomel, 2008 ). Mammalian uterine development and differentiation are incomplete at birth, and several key stages of tissue growth and organization of the endometrium and myometrium occur postnatally ( Cooke et al. , 2013 ). Importantly, the development of these tissue compartments relies on sequential signaling and communication with the development of neighboring tissue. Endometrial adenogenesis (gland formation), in particular, relies on cues from the stroma to be carried out correctly ( Cooke et al. , 2013 ; Kelleher et al. , 2019 ). Chemical perturbations can therefore affect adenogenesis by targeting gland development indirectly through stromal dysregulation or directly by disrupting differentiation signaling within epithelial cells. Mice neonatally exposed to genistein exhibit disrupted tissue organization and abnormal cell differentiation. In genistein exposed mice, gland number is reduced, and complexity of gland structure is diminished ( Jefferson et al. , 2020 ). Instead of numerous histologically uniform glands that extend into the antimesometrial stroma, when glands do develop in genistein exposed mice they are often shallow, located near the lumen, or present in the myometrium (adenomyosis) ( Fig. 3 ) ( Jefferson et al. , 2020 ; Suen et al. , 2018 ). By mid to late adulthood, genistein exposed mice also exhibit cystic dilation, as well as luminal and glandular squamous and basal cell metaplasia ( Suen et al. , 2018 ). Although abnormal gland development, or lack thereof, and gland architecture are major contributing factors impacting fertility, further studies on molecular changes and fundamental shifts in tissue development underlying the alterations in uterine receptivity are ongoing. Several studies have investigated how genistein-induced oviductal and uterine phenotypes contribute to early pregnancy loss. Mice neonatally exposed to 50 mg/kg/day genistein can be superovulated and mated successfully, but only about half of the resulting embryos survive in transit through the oviduct when normally most embryos survive ( Jefferson, Padilla-Banks, et al. , 2009 ). When the oviduct of these mice is bypassed by transferring blastocysts from control donors directly into the uterus of pseudopregnant neonatally genistein-exposed mice, the embryos begin to implant but are reabsorbed shortly thereafter. These findings indicate that both the oviduct and uterine endometrium of mice neonatally exposed to genistein are not capable of supporting pregnancy ( Jefferson, Padilla-Banks, et al. , 2009 ). Within the oviduct, several essential transcription factors in the Hoxa (homeobox A cluster) family, Wnt (wingless-related MMTV integration site) signaling genes, and Hedgehog family genes are dysregulated and expression of downstream pathway genes is more consistent with patterns normally observed in cervical and vaginal tissue than in the oviduct ( Jefferson et al. , 2011 ). Consistent with this finding, genistein exposure also alters the oviductal mucosal immune response during pregnancy, reflecting an immune response more similar to that normally observed in the cervix ( Jefferson, Padilla-Banks, et al. , 2012 ). This type of altered tissue patterning, in which more anterior tissues take on characteristics of posterior tissues, is known as “posteriorization” ( Fig. 4 ). Taken together, these findings highlight the importance of appropriate tissue patterning during development and the risks inherent in estrogenic chemical exposure during female reproductive tract development and differentiation. The majority of the available genistein exposure data uses a 50 mg/kg/day dose administered by subcutaneous injection. However, in the few studies using either lower doses of subcutaneously administered genistein or orally administered genistin, genistein, or isoflavone mixtures, similar reproductive outcomes were observed. For example, neonatal mice given genistein in a soy formula emulsion by gavage had increased uterine wet weight, altered progesterone receptor expression in the uterine epithelium, and altered estrous cyclicity, all of which are changes normally observed in response to estrogenic chemicals, including subcutaneously administered genistein ( Cimafranca et al. , 2010 ). Effects using these alternative dosing compounds and/or strategies generally presented a dose-response effect where low doses led to less severe phenotypes ( Jefferson, Doerge, et al. , 2009 ; Kaludjerovic et al. , 2012 ). The number of individual phenotypes that arise in mice following neonatal genistein exposure contribute to infertility and compromise live birth. However, even when each phenotype is considered individually, its potential translation to human reproductive effects following genistein exposure through soy-formula consumption is a cause for concern. This question could be addressed with “time to pregnancy” studies in healthy females previously exposed (or not) to soy formula who are attempting to conceive, as well as adverse pregnancy outcomes such as pregnancy loss and preterm birth. Uterine adenocarcinoma development is one of the most concerning disease phenotypes found in animal models following early life genistein exposure. In an animal model of hormonal carcinogenesis, mice are exposed to an estrogenic chemical during the first five days of life and uterine cancer incidence is assessed in adulthood. This model was initially developed to investigate the adverse effects of exposure to the potent estrogenic chemical diethylstilbestrol (DES) ( Newbold et al. , 1990 ), a human transplacental carcinogen with a myriad of adverse human reproductive effects ( Decherney et al. , 1981 ; Hatch et al. , 2011 ; Herbst et al. , 1971 ). This model has been widely used over the past three decades to assess DES effects, and depending on the mouse strain assessed, adult endometrial cancer incidence ranges from 50–90% ( Davis et al. , 2012 ; Newbold et al. , 1990 ; Suen et al. , 2016 , 2019 ). Interestingly, several other endogenous, environmental, and industrial estrogenic chemicals have also been tested in this model, resulting in dose-dependent estrogenic effects. Approximately 20 years ago, the neonatal estrogenic chemical exposure model described above revealed that when mice were exposed to genistein neonatally, approximately 35% developed uterine cancer as adults ( Newbold et al. , 2001 ; Suen et al. , 2016 ). More recent studies have documented histopathologic changes in the reproductive tract from the time of exposure to late adulthood ( Suen et al. , 2016 , 2018 ). These findings include basal cell metaplasia, squamous cell metaplasia, uterine muscle dysmorphogenesis, adenomyosis, and endometrial carcinoma ( Fig. 3 ). To date, mechanistic studies targeting specific pathways that link genistein exposure with cancer development have not been completed. Nonetheless, genistein-induced uterine cancer phenocopies cancer induced by DES in histopathologic appearance, strongly suggesting a similar mechanism of disruption of uterine programming and differentiation induced by a short term exposure during a critical period of development ( Ashby, 2001 ; Newbold et al. , 2001 ). Although effects of genistein and other phytoestrogens are largely mediated by estrogen receptor action, there is little information available regarding the downstream pathways responsible for the endometrial cancer phenotypes, even in rodents. As mentioned above ( Section 6.1 ), developmental drivers are dysregulated in the female reproductive tract when genistein exposure occurs during development ( Jefferson et al. , 2011 ). Proteins that normally regulate development can lead to cancer development when re-activated in adults; these are referred to as “oncofetal” proteins. Neonatal genistein and DES exposures result in aberrant upregulation of the oncofetal protein sine oculis homebox1 (SIX1) ( Jefferson et al. , 2011 , 2013 ; Suen et al. , 2016 ). SIX1 is overexpressed in the uteri of genistein and DES exposed mice, its transcript expression persists and increases over the animals’ lifetime, and SIX1 protein localizes to abnormally differentiated endometrial epithelial cells and neoplastic lesions ( Suen et al. , 2016 , 2018 ). We have shown that Six1 upregulation is associated with persistent changes in epigenetic marks at the Six1 gene locus ( Jefferson et al. , 2013 ). In humans, SIX1 is not expressed in normal human endometrium but is detected in about 25% of endometrial cancers and is more prevalent in late stage disease, but whether it is a disease driver or just a disease marker is unknown ( Suen et al. , 2016 ). Further studies are needed to continue clarifying the mechanistic underpinnings of developmental exposure-induced effects on the female reproductive tract. All of the genistein exposure studies described above use oral or subcutaneous routes of exposure to isoflavone mixtures or purified genistein and result in serum genistein levels similar to serum genistein levels present in human infants consuming soy-based formulas ( Cao et al. , 2009 ; Cimafranca et al. , 2010 ; Jefferson, Patisaul, et al. , 2012 ). It is difficult to predict the translational relevance of animal studies to human biology. However, the comparable dose and overlapping window of development, the data presented above, and the significant incidence of severe adverse health outcomes (infertility and cancer) observed in animal models highlight that animal data should be more heavily weighted in safety recommendations.

Tissues undergoing changes in form or function are exquisitely sensitive to disruption of the differentiation pathways that control these changes. A classic example of this phenomenon is that the fetal period of organogenesis is highly susceptible to teratogenic chemicals (e.g., thalidomide, retinoic acid), whereas adult exposures to the same chemicals may cause no adverse effects. Although fetal development is clearly a time of extensive differentiation and growth, certain periods of postnatal development are also critical for later adult function. Notably, development of the central nervous system and reproductive system are modulated by steroid hormone signaling both prenatally and postnatally, leaving these tissues particularly sensitive to disruption by EDCs ( Ma, 2009 ; Park and Jameson, 2005 ; Sisk and Zehr, 2005 ; Young et al. , 1964 ). Reproductive development begins in utero with the development of internal and external genitalia when the fetus is exposed to low levels of gestating parental estrogen ( Adgent et al. , 2018 ; Kuiri-Hänninen et al. , 2014 ; Lanciotti et al. , 2018 ). Male reproductive tract development is dependent on testosterone produced by the developing testis, but based on animal studies, early female reproductive development normally occurs independent of endogenous estrogen signaling ( Krege et al. , 1998 ; Lubahn et al. , 1993 ). Even so, ERα is detected in the human fetal female reproductive tract beginning at 12 weeks’ gestation and in the vagina beginning around 21 weeks’ gestation ( Cunha et al. , 2017 ). Expression of estrogen receptors leaves this biological system poised for later activation, or in the case of EDC exposure during this time, susceptible to abnormal activation of signaling and downstream biological disruption ( Chianese et al. , 2018 ; Herbst et al. , 1971 ). Hypothalamic-pituitary-gonadal (HPG) axis activation also occurs during fetal life, peaking at mid gestation followed by a decline towards total suppression at birth ( Kuiri-Hänninen et al. , 2014 ). This pattern of fetal HPG axis activation is associated with increasing levels of circulating estradiol over the course of fetal development in females, with the highest levels around birth ( Lanciotti et al. , 2018 ). After birth, infants experience a period of withdrawal from gestating parental estrogen exhibited by declining estrogenization of reproductive tissues ( Bernbaum et al. , 2008 ). This decline is followed by a period of postnatal HPG axis activation termed “minipuberty” ( Kuiri-Hänninen et al. , 2014 ; Lanciotti et al. , 2018 ). Minipuberty begins in the first week of life and lasts for 7–9 months in males and up to 4 years in females. In females, minipuberty leads to variable endogenous estrogen production, which is highest in the first 3–6 months. Endocrine signaling that occurs during this time period is essential for maturation of the sexual organs and lays the groundwork for future fertility. Changes in reproductive development may be obscured when observations are made only at isolated time points, particularly when the observations are made during a time of physiological change such as sensitive windows of development. For example, a reproductive tissue may only show abnormal estrogenization in response to endocrine disruption during the brief postnatal window when tissues from unexposed individuals are undergoing withdrawal from gestating parental estrogen, and this effect may be hidden by the ensuing minipuberty ( Kuiri-Hänninen et al. , 2014 ). To address these caveats, recent epidemiological studies have incorporated the use of developmental trajectories to identify subtle, time-dependent changes. Less is known regarding windows of developmental sensitivity in humans in adolescence and adulthood. Relative to the short rodent developmental window, human female reproductive tract development occurs over a much longer period of time, with endometrial gland formation beginning in the second trimester of fetal development and continuing for many years until puberty ( Fig. 1 )( Gray et al. , 2001 ). In addition, cyclic loss of endometrial tissue with menstruation followed by endometrial regeneration opens another potential window of sensitivity in females of childbearing age. As a result, the window of sensitivity in humans is likely far longer than in rodents, but because development is slower, the reproductive tract could be less susceptible to short-term disruption. This question will be an important one to answer in future human epidemiological or clinical studies.

Isoflavones are soluble aromatic compounds that have structural homology to estradiol and are classified as phytoestrogens because they occur naturally in plants and have the ability to interact with estrogen signaling pathways ( Kurzer and Xu, 1997 ; Vyn et al. , 2002 ). Initially produced in plants, isoflavones serve diverse functions necessary for plant health including mediating pathogen resistance and host-symbiont interactions ( Sugiyama, 2019 ). Genistein (4’, 5, 7-Trihydroxyisoflavone) and daidzein [7-Hydroxy-3-(4-hydroxyphenyl) chromen-4-one] are the predominant plant isoflavones and are particularly abundant in soybeans and other leguminous species. These isoflavones can exist in the aglycone form or as the glycosidic conjugates, genistin and daidzin, which have lower bioactivity but higher water solubility ( Van Rhijn and Vanderleyden, 1995 ). In addition to acting as a precursor for other antimicrobial molecules, genistin and daidzin possess innate antimicrobial activity and are stored constitutively in plant root tissue ( Dakora and Phillips, 1996 ; Graham et al. , 1990 ). The isoflavone content of leguminous plants is strongly influenced by environmental factors and concentrations can vary widely between plant species, cultivation season, growth climate, and anthropogenic treatment such as fertilizers ( Vyn et al. , 2002 ). Isoflavone content in soybeans cultivated in different countries can range from ~100 mg (Brazil) to ~160–180 mg (U.S. and Korea, respectively) per 100 g soybean ( Rizzo and Baroni, 2018 ). As a result, quantifying isoflavone intake based on unprocessed soy intake is challenging. However, the generally increased intake and subsequently elevated serum levels of genistein relative to daidzein following soy protein consumption ( van Erp-Baart et al. , 2003 ; Kolátorová et al. , 2018 ; Peeters et al. , 2007 ; Setchell et al. , 1997 ; Valentín-Blasini et al. , 2003 ), despite cultivation conditions, makes genistein the phytoestrogen of most concern to human health; therefore, the remainder of this review will focus on genistein. Genistein pharmacokinetics in adults has been reviewed extensively and will be summarized briefly here ( Cassidy et al. , 2006 ; Franke et al. , 2014 ; Whitten and Patisaul, 2001 ; Woods and Hughes, 2003 ). When humans and other animals consume genistein or genistin, microbes in the oral cavity and intestine hydrolyze the glycosidic bonds, converting the conjugated form to genistein ( Kolátorová et al. , 2018 ; Walle et al. , 2005 ). Genistein is readily absorbed across the intestinal epithelium, and undergoes biotransformation predominantly by phase II metabolism ( Yang et al. , 2012 ). Genistein aglycone and phase II conjugates enter circulation and travel to tissues where unconjugated and conjugated forms undergo dynamic inter-conversion to genistein aglycone or glycosylated conjugates ( Yang et al. , 2012 ; Yuan et al. , 2012 ). In addition to genistein, other isoflavones such as daidzein go through a similar absorption, distribution, metabolism, and elimination processes, adding to the variety of derivatives with estrogenic activity ( Mayo et al. , 2019 ). Interindividual differences in specific ingested foods, the intestinal microbiome, and expression/activity of metabolic enzymes can alter in vivo exposure to the bioactive forms of xenobiotics, including isoflavones ( Cassidy et al. , 2006 ; Moon et al. , 2006 ). Hence, it is difficult to directly relate dietary exposure levels to bioactivity ( Di Lorenzo et al. , 2021 ). Less is known about genistein pharmacokinetics in infants. The gut microbiota in infants is not fully developed for the first 2–3 years of life ( Yatsunenko et al. , 2012 ), so microbiome influences on genistein metabolism could differ from that of adults. There are differences between infants and adults in genistein absorption, distribution, metabolism, and elimination properties. Notably, when fed soy products containing comparable weight-based doses of isoflavones, infants exhibit increased plasma and urine isoflavone concentrations demonstrating increased bioavailability as compared with adults ( Franke et al. , 2006 ; Halm et al. , 2007 ). A combination of continuous daily exposure (due to feedings every few hours) and decreased biotransformation, metabolism, and elimination may explain why plasma genistein is higher in infants compared with adults consuming similar amounts of genistein ( Setchell et al. , 1998 ). Furthermore, recent studies suggest that exposure to genistein in utero or through postnatal soy infant formula feeding may alter the composition of the developing microbiota ( Baumann-Dudenhoeffer et al. , 2018 ; Vázquez et al. , 2017 ). Variation in the microbiome, along with individual differences in metabolic enzymes or other factors impacting detoxification, could further influence genistein metabolism in soy formula-fed infants. Further studies are needed to develop a better understanding of the pharmacokinetics of genistein metabolism and interactions with the intestinal microbiome in infants. The ability of genistein to activate human estrogen receptor signaling depends on several factors including binding affinity toward the receptor and serum proteins and the level of endogenous estrogens. In vitro studies indicate that relative to estradiol, genistein has decreased binding affinity for both estrogen receptor alpha (ERα) (50-fold lower) and beta (ERβ) (7-fold lower) ( Choi et al. , 2008 ). However, biological activity is modulated in vivo by binding to serum proteins such as sex hormone binding globulin. Estradiol binds serum proteins far better than genistein, resulting in a 10-fold lower availability of estradiol for estrogen receptor binding relative to that of genistein ( Nagel et al. , 1998 ). This finding indicates that in vivo , genistein can have significant biological activity when present in concentrations similar to that of estradiol. In infants exclusively consuming soy formulas, plasma isoflavone levels are 13,000–22,000 times higher than plasma estradiol levels, a scenario where genistein is likely to activate estrogen receptor-mediated responses ( Setchell et al. , 1998 ). Genistein and other isoflavones may also have biological effects that are not mediated by estrogen receptors, but the mechanisms of non-estrogenic action are still being elucidated.

Despite the variety in source, content, and composition, soy continues to be a potential source of estrogenic chemical exposure during developmentally sensitive windows. Given the large body of evidence from animal studies indicating that genistein exposure has adverse health outcomes later in life and the growing support from epidemiological studies that similar adverse outcomes may occur following soy-based infant formula exposure in humans, updated recommendations on soy formula feeding are needed. This is particularly true now that there is direct evidence that soy formula feeding is associated with estrogen-induced changes in human female reproductive tract development ( Adgent et al. , 2018 ). Because of the difficulties in conducting epidemiological studies connecting exposures during infancy to health outcomes in adulthood, these recommendations will need to give more weight to findings from animal studies unless epidemiological and/or clinical studies can conclusively demonstrate that soy formula feeding during infancy does not pose a risk. Until then, consideration should be given to a requirement for guidelines or warnings on soy formula product packaging regarding phytoestrogen content so that parents can make informed choices.

Shifts in health care toward preventive medicine highlight the importance of identifying and characterizing preventable environmental exposures with adverse health effects. Assessment and regulation of both synthetic and naturally-occurring chemicals that impact normal biology, particularly during periods of rapid development, are needed at multiple levels — exposure prevention, risk prediction, and disease treatment. Endocrine disrupting chemicals (EDCs), a class of chemicals that interfere with the endocrine system, are of particular interest because the dose required to elicit an effect may be low, the effects may be subtle resulting in delayed intervention, and exposure may lead to permanent biological changes ( Kahn et al. , 2020 ; Vandenberg et al. , 2012 ). EDCs that activate estrogen receptor signaling are known as estrogenic chemicals. This class of EDCs includes plasticizers (e.g., phthalates and bisphenol A), pharmaceuticals (e.g., diethylstilbestrol and ethinyl estradiol), naturally occurring mycoestrogens (e.g., zearalenone), and plant-derived phytoestrogens (e.g., genistein and daidzein). As with many toxicants, the timing of EDC exposure plays a major role in the potential for significant biological impact. Developmentally sensitive windows, such as prenatal development, comprise periods of rapid tissue formation, growth, differentiation, and development of functional competency that are highly susceptible to disruption leading to long-term health impacts. Following birth, the endocrine system goes through a period of functional activity that impacts development and is known as “minipuberty” (see Section 4). During minipuberty, gonadal steroid hormones including estradiol and testosterone are released into the circulation and influence development of many tissues including the reproductive system, mammary gland, and the brain ( Kuiri-Hänninen et al. , 2014 ). These developmental changes in response to hormonal signaling make minipuberty particularly susceptible to endocrine disruption. Despite being derived from natural sources, phytoestrogens are potent estrogen receptor activators and can induce biological effects similar to endogenous and synthetic estrogens ( Patisaul and Jefferson, 2010 ). The isoflavone genistein is the most abundant phytoestrogen in soy products ( Adlercreutz and Mazur, 1997 ), and therefore has an increased potential to impact human health. Diet is the most common route for phytoestrogen exposure, and adult dietary phytoestrogen exposure varies based on available food resources, cultural norms, and dietary choices. However, even adult vegetarians who consume moderate amounts of soy-containing foods have relatively low phytoestrogen levels ( McCarver et al. , 2011 ; Peeters et al. , 2007 ). Like adults, infants who are breast-fed or fed cow’s milk-based formulas have minimal phytoestrogen exposure. In contrast, soy-based infant formulas are a source of concentrated phytoestrogen exposure and their use leads to high serum genistein levels in infants ( Cao et al. , 2009 ; Setchell et al. , 1997 ). Use of soy-based infant formula exposes infants to an estrogenic EDC during the developmentally sensitive window of minipuberty. The U.S. National Toxicology Program Center for the Evaluation of Risks to Human Reproduction monograph on soy infant formula in 2010 concluded that “there is minimal concern for adverse effects on development in infants who consume soy infant formula” ( McCarver et al. , 2011 ; National Toxicology Program, 2010 ). However, several new human and animal studies have been published supporting a need for reevaluation and reconsideration of past recommendations. This review focuses on phytoestrogens, particularly genistein, and potential exposures during infancy due to consumption of soy-based infant formulas. We provide background information on the sources and quantities of human exposure to phytoestrogens. We then evaluate epidemiological and animal-model findings related to reproductive outcomes following early life phytoestrogen exposure, primarily considering studies published in the past 10 years. We identify the strengths and weaknesses of these studies and explore the implications for clinical recommendations and regulations for the use of soy formula in infant feeding.

Globally, there is a consensus in nutritional recommendations that human breast milk is the ideal source of nutrition for infants and that exclusive breast feeding should be done for the first six months of life followed by food introduction with continued breast feeding for one year or longer ( Bhatia et al. , 2008 ; National Academies of Sciences, Engineering, and Medicine, 2020 ). In situations where formula must be used, the guidelines consistently recommend cow’s milk-based formula. According to these resources, soy-based formulas are only recommended in special circumstances, which include infants with galactosemia and hereditary lactase deficiency or in situations where a vegan diet is preferred by parents ( Bhatia et al. , 2008 ; National Academies of Sciences, Engineering, and Medicine, 2020 ). Primary lactase deficiency in infants is a rare and severe autosomal recessive disorder, and secondary lactase deficiency can be managed with lactose free or lactose reduced cow’s milk formula ( Bhatia et al. , 2008 ; Wanes et al. , 2019 ). Therefore, according to institutional guidelines across the globe, use of soy formula should be quite limited. Despite the limited indications, soy-based formula is widely used. The most recent U.S. survey data from NHANES 2003–2010 indicated that of the 81% of infants who are fed formula, 13% are fed with soy-based formulas ( Rossen et al. , 2016 ). Soy formula use differs demographically from the use of other infant formulas. Compared with breast-fed only infants, a higher percentage of infants fed with any type of infant formula come from families with a low ratio of income to poverty and a household education level of less than high school ( Rossen et al. , 2016 ). However, infants fed specifically with soy formula tend to come from families with higher income to poverty ratios and a household education level of at least some college. Based on data from the IFED study, the most common reasons for choosing soy formula are (1) successful use with a previous child (2) perceived healthfulness or (3) anticipated lactose intolerance based on family history ( Adgent et al. , 2018 ). The findings of the IFED survey suggest that soy formula is not medically necessary for most infants to whom it is fed. Agencies and working groups in France, Ireland, Switzerland, the United Kingdom, and Europe at large have expressed concerns over the risk posed by phytoestrogen exposure through soy infant formula consumption ( AFSSA, 2005 ; Agostoni et al. , 2006 ; Woods and Hughes, 2003 ). In their position statement, the UK Committee on Toxicity of Chemicals in Food, Consumer Products, and the Environment (COT) gave more weight to studies on soy formula use in animals than do expert recommendations released by U.S. agencies ( Woods and Hughes, 2003 ). The COT suggests that the concerns raised from animal studies are grave enough that they should be considered until disproven in human studies ( Woods and Hughes, 2003 ). The COT therefore recommends breast milk and cow’s milk as the preferred source of nutrition for infants unless otherwise advised by a health professional. This represents a divergence from the methodology of recommendations made by agencies in the U.S. and Canada, which generally follow a rationale that animal studies do not merit enough concern to restrict consumption until they are conclusively upheld through studies in humans. Guidelines from Australia and New Zealand are similar. The Royal Australasian College of Physicians (RACP) Paediatrics and Child Health Division does not take a stance on the safety of phytoestrogen exposure via soy formula but does restrict indications for soy formula use to the management of galactosemia ( RACP, 2006 ). The Australian College of Paediatrics released a position statement on soy in 1998 that recommended against the use of soy formula in the treatment of cow’s milk protein allergy, but their concern was based on potential sensitivity to soy protein ( Simmer, 1998 ). A more recent publication by the Royal Australasian College of Physicians upheld these recommendations, but did note that conflicting evidence had emerged on the safety of infant exposure to phytoestrogens ( RACP, 2006 ). Researchers from the Ministry of Health in New Zealand came out more strongly against the use of infant soy formula due to concerns over phytoestrogen exposure, and recommend against its use except when required in cases of galactosemia ( New Zealand Ministry of Health, 1998 ; Tuohy, 2003 ). The recommendations of the Canadian Paediatric Society (CPS) published in 2009 largely align with those of other countries ( Leung and Otley, 2009 ). Based on the available animal and human data, as well as recommendations provided by experts from other regions, the CPS concluded that “There are concerns based on animal and in vitro data regarding the phytoestrogen content of soy-based formulas, and the potential risks for those infants who receive their sole sources of nutrition from these formulas. However, based on available human data, no overt harm has been proven with the use of currently available soy-based infant formulas as the sole source of nutrition for infants” ( Leung and Otley, 2009 ). Given the high use of soy formula without clinical indications, the CPS does recommend that “Physicians should consider limiting the use of soy-based formulas to those infants with galactosemia or those who cannot consume dairy based products for cultural or religious reasons” ( Leung and Otley, 2009 ). In the U.S., formal guidelines from the American Academy of Pediatrics (AAP) also recommend limited use of soy formula; however, both the AAP and the United States Food and Drug Administration consider the use of soy-based formulas in full term infants safe ( Bhatia et al. , 2008 ; Rossen et al. , 2016 ). Despite numerous animal studies highlighting reproductive concerns of soy protein and purified isoflavone exposure during development, experts in the U.S. have largely focused safety decisions for soy infant formula on available human data ( National Toxicology Program, 2010 ). In 1998, the AAP released guidelines on soy protein-based formula feeding in infants that contained a statement on phytoestrogens and indicated a low risk from exposure based on human data available at the time of publication ( Klish et al. , 1998 ). The AAP updated these guidelines in 2008 to include more data on phytoestrogen exposure research, but these guidelines again stated that the current data does not conclusively demonstrate adverse effects from dietary exposure to soy isoflavones ( Bhatia et al. , 2008 ). In 2011, the National Toxicology Program’s Center for the Evaluation of Risks to Human Reproduction held an expert panel to conduct a review on the risks associated with the use of infant soy formula ( McCarver et al. , 2011 ; National Toxicology Program, 2010 ). After reviewing data from animal and human studies, the panel concluded that it was unclear whether animal studies using genistein treatment could be extrapolated to infants fed soy formula, and that there was insufficient data from studies in humans to increase the recommended level of concern. Based on the animal and epidemiological studies available at the time of the panel, the committee voted to update the level of concern related to the use of soy infant formula from “negligible concern” (recommendation in 2006) to “minimal concern”. Minimal concern represents a “2” on a five-level scale ranging from negligible concern (1) to serious concern (5). The panel called for further studies in both humans and animals stating that “Soy infant formula may or may not cause reproductive toxicity in boys and girls based on current evidence. Preliminary data addressed in the panel’s report do not allow firm conclusions on this effect.” ( McCarver et al. , 2011 ; National Toxicology Program, 2010 ). Overall, experts in the U.S. support the use of infant soy formula as a safe and cost-effective alternative to cow’s milk formula.