Achieving CLARITY on bisphenol A, brain and behaviour

Abstract

There is perhaps no endocrine disrupting chemical more controversial than bisphenol A (BPA). Comprising a high-volume production chemical used in a variety of applications, BPA has been linked to a litany of adverse health-related outcomes, including effects on brain sexual differentiation and behaviour. Risk assessors preferentially rely on classical guideline-compliant toxicity studies over studies published by academic scientists, and have generally downplayed concerns about the potential risks that BPA poses to human health. It has been argued, however, that, because traditional toxicity studies rarely contain neural endpoints, and only a paucity of endocrine-sensitive endpoints, they are incapable of fully evaluating harm. To address current controversies on the safety of BPA, the United States National Institute of Environmental Health Sciences, the National Toxicology Program (NTP), and the US Food and Drug Administration established the Consortium Linking Academic and Regulatory Insights on BPA Toxicity (CLARITY-BPA). CLARITY-BPA performed a classical regulatory-style toxicology study (Core study) in conjunction with multiple behavioural, molecular and cellular studies conducted by academic laboratories (grantee studies) using a collaboratively devised experimental framework and the same animals and tissues. This review summarises the results from the grantee studies that focused on brain and behaviour. Evidence of altered neuroendocrine development, including age and sex-specific expression of oestrogen receptor (ER)α and ERβ, and the abrogation of brain and behavioural sexual dimorphisms, supports the conclusion that developmental BPA exposure, even at doses below what regulatory agencies regard as “safe” for humans, contribute to brain and behavioural change. The consistency and the reproducibility of the effects across CLARITY-BPA and prior studies using the same animal strain and almost identical experimental conditions are compelling. Combined analysis of all of the data from the CLARITY-BPA project is underway at the NTP and a final report expected in late 2019.

1 |. INTRODUCTION

Synthesised in 1891 by the Russian chemist A. P. Dianin and demonstrated to be oestrogenic in the 1930s,^1,2 there is now quite possibly no chemical more controversial than bisphenol A (BPA). First used in plastics in the 1950s, BPA has become a pervasive component of our manufactured landscape and a high-volume production chemical, meaning that over a million pounds are produced or imported into the US annually. BPA is as inevitable and unavoidable as air. BPA has insidiously become an inextricable part of our lives and a critical catalyst for intense discussions about how and why chemicals are introduced into the commercial landscape with, typically, very minimal and conflicting information about their potential toxicity. Despite decades of research and literally thousands of published studies, messages about safety from regulatory agencies are largely at odds with those from scientific expert groups, creating confusion and divisive animosity, with each side claiming that the data support their position. In an effort to bridge this interpretive gap, an ambitious and uniquely coordinated study was undertaken by governmental and academic scientists in the USA with the common goal of comprehensively assessing the toxicological impact of BPA on multiple organ systems across a wide range of doses.³ Initiated in 2011, CLARITY-BPA (which stands for the Consortium Linking Academic and Regulatory Insights on BPA Toxicity) comprised two main components (Figure 1). The first was a “core” guideline-compliant study, with classical toxicology endpoints, run by US Food and Drug Administration (FDA) scientists at their research facility; the National Center of Toxicological Research (NCTR). The second was a set of more mechanistic studies conceived of, and conducted by, National Institutes of Health (NIH)-funded academic investigators (collectively called the “grantee” studies), using tissues from the same set of animals as the “core study”. The grantee studies included endpoints considered particularly sensitive to BPA. Here, a review is provided regarding the genesis, purpose and outcome of the CLARITY-BPA studies that focused on an organ considered particularly sensitive to BPA: the developing brain.

FIGURE 1.

Organisation chart for Consortium Linking Academic and Regulatory Insights on BPA Toxicity (CLARITY-BPA) participants. Grantees, each in different academic laboratories, were funded with grants by the United States National Institute of Environmental Health Sciences (NIEHS) which is also home to the National Toxicology Program (NTP), the organisation overseeing data integration. The National Center of Toxicological Research (NCTR) is a US Food and Drug Administration (FDA) research facility and was responsible for executing the core study within CLARITY-BPA. Animals and tissues from that core study were then allocated to the grantees. All data was submitted to the NTP which curated the data sets and made them publicly available

Bisphenol A is everywhere. It is used in food can linings, polycarbonate plastics, flame retardants, thermal receipts, epoxy resins and medical devices. Those embracing bottles with a BPA-free sticker are still inevitably exposed from harder to avoid sources, including BPA-containing epoxy resins added to metal-coating paints, such as those used on household appliances and farm equipment, and by handling some kinds of thermal paper receipts. BPA is also found in dental adhesives and resins used for bonding, filling and capping teeth. BPA has been detected in the air, soil, sea and fresh water, landfill leachate and human tissues, including blood, urine, amniotic fluid and cord blood from every population ever tested on Earth.^4,5 Babies are born with BPA in their bodies, along with hundreds of other anthropogenic chemicals, including other bisphenols.^6,7 As early as the 1990s, when public awareness of its presence in baby bottles was magnified, BPA raised an uncomfortable set of questions for the public and scientists alike: are we doing enough to test the safety of chemicals like BPA, and are “real world” exposures meaningfully harmful or not?

Bisphenol A is an endocrine disrupting chemical (EDC). Of that, there is no question. It has been known for almost a century that BPA is weakly oestrogenic, meaning that, compared to endogenous oestrogens, it has a low but not insignificant capacity to bind with and act as an agonist on nuclear oestrogen receptors in classic in vitro assays assessing binding and transcriptional capacity. BPA has a 10 000-fold lower binding affinity for nuclear oestrogen receptor (ER)α and ERβ than 17β-oestradiol and binds each with roughly equivalent affinity. BPA has also been reported in a handful of studies to act via other receptors, including membrane-bound forms of nuclear ERs, androgen receptor and G protein-coupled receptor 30. Activity of BPA in the “gold standard” toxicity test of oestrogenicity in vivo, the immature rat uterotrophic assay, is decidedly unremarkable. BPA is generally inactive at doses below 8 mg kg⁻¹ body weight (bw) when given by s.c. injection or 40 mg kg⁻¹ orally,⁸ both of which are far beyond what any human would reasonably be exposed to, even in an occupational setting. Because this assay is considered valuable by regulatory toxicologists for assessing potential endocrine disruption in vivo, the results have historically been used to support the conclusion that BPA poses no health risks to humans at the levels to which we are currently exposed. However, confidence in that conclusion hinges on a key assumption: that this and similar assays popular with risk assessors are the most sensitive and reliable way of assessing the whole body’s vulnerability to an oestrogen-disrupting compound at any point in the lifespan. Foundational studies in neuroendocrinology suggest that, instead, the most sensitive organ for endocrine disruption might be one not typically given much scrutiny in formal toxicity testing: the developing brain.

Neuroendocrinologists have long known that hormone signalling is essential to brain development, particularly sexual differentiation. Although there are species differences in the specific timing and mechanisms by which this differentiation occurs, steroid hormones universally play a major role.⁹ In rodents, perinatal oestrogen, aromatised from testicular androgens and acting directly on oestrogen receptors, is essential for brain masculinisation.^10,11 Gonadal hormones also contribute to human neural and behavioural sex differences, although the specific roles for androgen and oestrogen receptors are not as clearly defined.^12,13 In species from rodents to birds, manipulation of steroid hormone signalling, particularly oestrogen, during perinatal brain development profoundly alters sexual differentiation, and permanently alters brain morphology and behaviour.¹⁴ Thus, because BPA has repeatedly been shown to interfere with steroid hormone signalling and metabolism during critical windows of development, there is significant concern about the potential for BPA to alter neurodevelopment and, ultimately, compromise cognitive and other behaviours.^15–23

2 |. HEALTH RISK OR HYSTERIA: RISK IS IN THE EYE OF THE BEHOLDER

Formalised concerns about BPA-related effects on brain development arose in the early 2000s, when two critical expert reviews of the available BPA literature at the time concluded that there was at least some evidence for effects on the developing brain. One was conducted and published by the National Toxicology Program (NTP) in 2008 and cited “some concern” for effects on the brain, behaviour and prostate gland in foetuses, infants and children at current levels of exposure.²⁴ The other was conducted by a panel convened by the World Health Organization (WHO) in collaboration with the Food Agricultural Organization (FAO) of the United Nations in 2011 and cited concern that developmental BPA exposure could alter anxiety-related behaviours specifically.¹⁵ Additional expert panels and structured literature reviews published around that time affirmed the merit of human health concerns for the developing brain.^18,25,26

By contrast, prior and subsequent risk assessments by governmental agencies with the authority to curtail BPA use and exposure generally concluded that current exposure levels are not likely to be harmful.²⁶ For example, in a 2010 statement subsequently updated in 2012, the FDA generally agreed with the NTP panel’s 2008 findings, and expressed “some concern” about the effects of BPA on the brain and behaviour in foetuses, infants and children. However, in 2014, it abruptly reversed course and concluded that, at current exposure levels, there was no risk of harm (CAS RN 80-05-7). So how is it that regulatory agencies can come to such different conclusions than other scientific expert panels? How can conclusions from a common pool of literature be so discordant? The simple answer is that the groups are viewing that literature through a very different lens.

Lack of data is not the problem. The literature on BPA and its possible health effects is enormous by toxicological standards. Approaching 4000 papers have been published regarding its potential effects (in vitro, in vivo and in humans), with more than 500 of those on brain and behaviour. How that literature is compiled and evaluated is what differs so dramatically across expert groups. Inconsistencies in study design, dosing routes, animal models used, housing conditions and other experimental factors make it challenging to essentially compare “apples with apples” and interpret the available literature with high confidence. An additional challenge, as in any scientific field, is that studies vary in quality. Selecting the “best” and “most relevant” studies, qualitatively defined, is fundamental to any evaluation or formal risk assessment, and this is where viewpoints between regulatory and academic scientists can differ dramatically.

In reality, few to no published studies are used for risk assessment in a regulatory context; a fact that often comes as a shock to the scientists who generate the data. This partly stems from a lack of understanding of how studies are evaluated for utility in risk assessment (Figure 2) or even a literature review conducted for regulatory purposes. Take, for example, the 2014 FDA Updated Review of Literature and Data on Bisphenol A, which, affirmed the FDA’s position that “BPA is safe at the current levels occurring in foods”. For each study considered in the review there is a brief synopsis, along with a list of what the panel considered to be its strengths and weaknesses. The utility of each study for hazard identification (understanding if and how a chemical is toxic, as well as the conditions under which toxicity might appear in exposed humans) and risk assessment (the formal evaluation of safety for human exposure) is then listed. Amongst other things, for inclusion in risk assessment, a study must use human-relevant dosing (in this case, oral), more than 10 animals per group, contain dose-response information, demonstrate biological plausibility and examine “validated” endpoints, which, for the FDA, means “physical, behavioural and pathological observations indicative of a toxicological dysfunction (CAS RN80-05-7)”. Of the 36 studies reviewed by the FDA for neurotoxicity, seven were considered relevant for hazard identification but only one for risk assessment. That single study examined working memory in acutely exposed, middle-aged ovariectomised rats,²⁷ which, arguably, is not a suitably sensitive or appropriate animal model for assessing human-relevant neurotoxicity. Of the 25 published studies reviewed for reproductive effects, none were considered relevant for hazard identification or risk assessment. The same was true for all other categories, sparking widespread criticism that FDA’s evaluative approach is too myopic at best and biased at worst.

FIGURE 2.

The key elements of a risk assessment paradigm. Characterising the risk that a chemical such as bisphenol A (BPA) poses to human health requires an understanding of its potential toxicity, which is termed hazard characterisation, and its potential adverse outcomes. Once identified, dose-response studies determine within what dose range humans are likely to suffer from the adverse outcomes identified in the hazard assessment. For endocrine disrupting chemical (EDCs), that dose-response might be non-linear. An exposure assessment determines the degree to which humans are exposed to the chemical of concern, by which route exposure is likely to occur, and at what levels. In the case of BPA exposure is virtually ubiquitous and primarily oral. A decision about the potential risk a chemical poses to public health is then made. In a 2014 literature review considering the use of BPA in food packaging, the Food and Drug Administration (FDA) determined that, out of 36 published neurotoxicity studies, seven were relevant to hazard characterisation but only one had utility for risk assessment. The report upheld the conclusion that BPA poses no significant health risks at current exposure levels

This narrow focus on a very specific set of data is a direct consequence of the prevailing paradigm for assessing the toxicity of chemicals, which is almost entirely reliant on guideline-compliant research. Guideline studies follow protocols conceived and executed by regulatory agencies themselves or an international body such as the Organization for Economic Co-operation and Development to meet statutory mandates (discussions of the strengths and weaknesses of using developmental neurotoxicity as an example are available elsewhere^28–30). These highly prescriptive studies are typically conducted in accordance with Good Laboratory Practices (GLP); a set of internationally recognised quality assurance and quality control procedures for data collection and archiving.³¹ Critically, GLP does not assure the “best” experimental design but, instead, only how the data will be collected and archived.³² Guideline studies also heavily rely on gross organ-level endpoints such as brain weight to assess toxicity because obviously pathological outcomes such as tumor formation, organ deterioration and death are incontrovertibly “adverse”. These endpoints are what the FDA and other regulatory agencies consider validated endpoints, and were defined as far back as the 1940s, in an era well before neuroendocrinology was an established discipline, let alone considered as a route of possible toxicity. Thus, some would argue that they are woefully inadequate for assessing endocrine disruption, particularly neuroendocrine disruption. Biologically consequential adversity does not always equate with obvious pathology, and would thus be missed by guideline studies. Examples germane to BPA include a loss of brain sexual dimorphisms, early puberty, social dysfunction or higher anxiety.

In classical toxicity testing, chemicals are typically administered at high doses, and a dose-response is extrapolated using linear dose models. The key assumption being that “the dose makes the poison” and small doses do less damage than large ones. Work dating back to the 1990s challenged the soundness of that approach for EDCs, and provided critical evidence that EDCs may have relevant effects at low doses and/or non-linear effects on some endpoints in vivo.^33–35 The seminal work was performed in the prostate, although subsequent investigations have identified similar patterns in other organs including brain.^20,36 Acceptance of these EDC characteristics has been fiercely resisted by regulatory groups who, instead, typically characterise their recurrence across the literature as “spurious”, “not toxicologically relevant”, or “not biologically plausible”. Recognition of non-linear dose-response curves and, especially, low-dose effects have huge implications for chemical risk assessment because it essentially means that the current evaluation paradigm is not appropriate for EDCs.

A final concern about relying on guideline studies is risk of bias because the vast majority of guideline studies are conducted by corporate entities with the intention of swaying regulators, and not always published, let alone peer reviewed. Additionally, drafting of the guidelines themselves is heavily influenced by industry, as is the process for evaluating study quality for use in regulatory decision-making.³⁷. CLARITY-BPA was deliberately designed to contain studies that would be acceptable to risk assessors but push the boundaries of toxicological testing by enhancing transparency, involving an interdisciplinary and multi-agency group of scientists, and including endpoints considered biologically relevant for a healthy life but not typically found in a regulatory-compliant study. These included transcriptional profiles, the structure of sexually dimorphic brain regions and even epigenetic changes across hormone-sensitive tissues. Testing for outcomes across multiple organs, as well as a wide range of doses, including those considered “low dose”, were additional uniquely advantageous elements of CLARITY-BPA.

3 |. FRAMING CLARITY-BPA FOR BRAIN AND BEHAVIOUR

A critical goal of the grantee studies in CLARITY-BPA was to replicate prior, related work. We and the other CLARITY-BPA grantees focused on endpoints observed by us and others to be particularly sensitive to EDC exposure and thus considered robust, reproducible and complementary to the data slated for collection by the core study. Over a decade of work by our laboratory and others, using a diversity of species from quail to mice to non-human primates, has repeatedly shown that developmental exposure to BPA disrupts a plethora of sexually dimorphic brain endpoints, especially those shaped by oestrogens.^20,38 Particularly sensitive regions include the anterioventral periventricular nucleus (AVPV), amygdala, medial preoptic area and hippocampus, as well as aspects of the mediobasal hypothalamus including the ventromedial and arcuate nuclei.^39–46 In the AVPV, for example, we have shown that perinatal BPA alters its physical size (in rats), as well as the sexually dimorphic density of dopaminergic and kisspeptin neurones within it, resulting in a more “masculinised” phenotype in females.^39,47,48 Also well documented is the capacity for developmental BPA exposure to impact non-reproductive sexually dimorphic behaviours, especially anxiety.^20,39–43 The available human data largely corroborate the animal data, with multiple lines of evidence linking prenatal exposure to heightened risk of deleterious childhood behaviours, including anxiety^49–52 and hyperactivity.⁵³

Of most relevance to CLARITY-BPA, we and others have repeatedly shown that developmental BPA exposure can alter the expression of ERs in multiple brain regions that coordinate reproductive and other sexually dimorphic behaviours, including the AVPV, amygdala and surrounding structures.^43,48,54 The age-specific location and the density of oestrogen and other hormone receptors, particularly across perinatal development, are critically important for the coordination of brain sexual differentiation and, accordingly, sexually dimorphic physiology and behaviour. Thus, hormone receptor expression radically changes across early life, especially perinatally when much of the hypothalamus undergoes sexual differentiation.^43,55,56 Studies using a diversity of ER knockout mouse models and other tools have painstakingly revealed the celland ER subtype-specific mechanisms by which sexual differentiation occurs. A classic example is the role that ERα plays in the sexual differentiation of the AVPV and neighbouring sexually dimorphic nucleus (SDN).^57,58 Ultimately, perinatal oestrogen in males, derived via the aromatisation of testicular androgens and, possibly, de novo synthesis,^59,60 induces a litany of morphological and functional sex dimorphisms not the least of which is physical size; males have a larger SDN and a smaller AVPV than females.^10,11 Morphological sex differences also emerge and, in some cases, are maintained, in response to oestrogens and other steroid hormones elsewhere in the rodent brain as well, including the amygdala. Generally speaking, ERα is the most important ER isoform for most aspects of rodent brain sexual differentiation,^61,62 with ERβ and membrane ERs playing accessory roles.^63–65 Critically, two of the studies we published showing that developmental BPA exposure could alter brain ER expression conducted as a prelude to the CLARITY-BPA studies, under almost identical experimental conditions (Figure 3).

FIGURE 3.

Measuring oestrogen receptors (ERs) in the brain. A, ERβ expression in the neonatal rat medial amygdala (MePD) is sexually dimorphic at birth but this sex difference switches by postnatal day (PND)4, which is an example of how dramatically ER expression can change across development. Significant differences in expression compared to PND 0 levels are represented by **p<0.01; Significant sex differences are represented by †<0.05. B, An autoradiogram depicting the sex-specific expression of ERβ in the MePD and the neighbouring central portion of the ventrolateral region of the ventromedial nucleus (cVMNvl) on PND2. As the sex difference in MePD expression decreases, the one in the cVMNvl is still pronounced. C, An example from the pre-Consortium Linking Academic and Regulatory Insights on BPA Toxicity (CLARITY-BPA) PND study assessing ER expression on PND1 revealing how dramatically different ERβ expression differed between the gavaged (vehicle) and naïve controls. ***p<0.001; ###<0.001 compared to male vehicle controls; §§§<0.001 compared to female vehicle controls. D, For CLARITY-BPA, tissue was obtained by micropunch. The amygdalar punch is depicted with remnants of the hippocampal and hypothalamic punches visible. Images are adapted from Cao and Patisaul⁵⁶ and Cao et al^56,66

The first of these studies⁶⁶ was initiated by Sherry Ferguson and colleagues at NCTR with the same in-house strain of Sprague-Dawley rats (NCTR-SD) ultimately used in CLARITY-BPA. Via a fortuitous conversation at the PPTOX II meeting in 2009, she invited my team to examine ER expression in one of her ongoing studies. Because NCTR operates under GLP, we agreed to strict rules about blinding and other experimental design elements designed to minimise risk of bias; an experience that would ultimately prove invaluable for the CLARITY-BPA studies. Exposure was entirely prenatal, with dams exposed from gestational day (GD)6 through to the day of birth. As is typically performed in toxicological studies to ensure precise dosing, BPA administration was by oral gavage through a stomach tube. Two doses of BPA (2.5 and 25 μg kg⁻¹ bw), two doses of ethinylestradiol (EE) (5 and 10 μg kg⁻¹ bw), a vehicle (carboxymeth-ylcellulose) control and, uniquely, a naïve control group (which underwent the same handling as gavage but the gavage needle was not inserted) were included. Rapidly frozen heads from the postnatal day (PND)1 offspring were shipped to us for quantification of ERα and ERβ expression in the hypothalamus and amygdala by in situ hybridisation, as we had performed in prior studies mapping post-natal ER expression in the rat brain.^55,56 As would ultimately be carried out for CLARITY-BPA, all raw data was provided to NCTR prior to unblinding and statistical analysis.

As expected, BPA-related effects were region, dose and sex-specific, with some known sex differences in ER expression eliminated at the lowest dose of 2.5 μg kg⁻¹ bw BPA.⁶⁶ Notably, BPA-and EE-related effects were directionally similar, with exposure resulting in the up-regulation of ER (α or β) in most circumstances. Most striking, however, was the substantial difference in ER levels between the vehicle and naïve controls. In the gavaged controls, ER levels were markedly lower, particularly ERβ in the amygdala (Figure 3). Although not the first study to provide evidence that gavage can induce stress-related consequences,^67,68 to our knowledge, it was the first to clearly demonstrate a neural effect on offspring that were not themselves directly subjected to the procedure but, instead, exposed prenatally. As such, the deleterious effects are entirely in line with a massive literature on offspring consequences of prenatal stress, which also consistently finds sex-specific outcomes including risk of neurodevelopmental and psychiatric disorders.^69–71 BPA and EE generally returned ER expression to levels typical of the naïve control, leading to the conclusion that heightened expression levels in the exposed animals likely reflected an interaction of exposure and stress (although an effect of vehicle cannot be conclusively ruled out).

The second study conducted as a prelude to CLARITY-BPA⁷² was a subchronic exposure study designed and carried out by NCTR using a protocol highly similar to the CLARITY-BPA core study.^73,74 For our portion of the project, only females were examined, although vehicle controls of both sexes were included to ensure known sex differences could be reliably detected, and establish the degree to which BPA and EE “masculinise” the female brain. Four doses of BPA (2.5, 25, 260, 2700 μg kg⁻¹ bw) and two doses of EE (0.5, 5 μg kg⁻¹ bw) were used. Because it was initiated prior to the completion of the study described above, naïve controls were not included. Similar to CLARITY-BPA, there were two experimental arms, each having a different exposure period: GD6 through PND21 and GD6 through PND90. Offspring brains sent to us were isolated and frozen on PND21 or PND90 and underwent ER in situ hybridisation as was performed in the prior, smaller study. Concordant with the PND1 study, the effects were region and sex specific (Figure 3). Directionally, however, the two studies differed. In both the PND21 and PND90 animals, BPA exposure generally decreased ER expression, opposite to that found on PND1. This difference is not entirely surprising given the dramatic age-dependent differences in baseline ER expression levels across the rodent brain (Figure 3).^43,55,56 As in the PND1 study, ERβ expression was more responsive. The effects of BPA and EE were, again, generally concordant in direction but not necessarily with respect to dose. Notably, low-dose BPA effects were not always recapitulated by the lowest dose of EE, an outcome consistent with the concept that EDC-related effects are not always linear.^35,75 Whether or not these non-monotonic effects are reproducible in brain and other tissues was of central interest in the CLARITY-BPA study.

4 |. CLARITY-BPA: DESIGN AND DETAILS

The CLARITY-BPA programme, including the rationale for its genesis, key goals (in addition to those already cited in the present review) and the overarching experimental framework, has been comprehensively described elsewhere by its architects.^3,76,77 The primary impetus for the project was the desire to collaboratively generate data, from multiple sources and over a broad range of endpoints, on a harmonised experimental framework, which could ultimately be used for human risk assessment. Despite a wealth of published literature and other information on BPA, uncertainties and inconstancies in the data were repeatedly cited by the FDA and other regulatory bodies as a substantial impediment to comprehensive evaluation of human risk.²⁶ In their reports, the NTP and FAO/WHO experts echoed this concern about experimental discordance with the latter stating, “The main problem encountered was a body of literature that reported conflicting results and used experimental protocols that prevented the comparison of results from one experiment to another”. Critical elements of concern included different exposure routes, lack of consistency among related endpoints or across studies, questions regarding the relevance of some animal models to humans and differences (age, sex, species) in the metabolism (and detoxification) of BPA. CLARITY-BPA sought to resolve these uncertainties by fleshing the backbone of a classic GLP-compliant, guideline-type “core” study with hypothesis-driven, extramural research using multiple tissues and animals obtained from the core study. Thus, by drawing on the strengths and expertise of the NCTR, the United States National Institute of Environmental Health Sciences (NIEHS), the NTP and academic scientists (chosen from applications submitted to RFA-ES-10–009), the CLARITY-BPA consortium was designed to “enhance risk assessment by resolving scientific uncertainties about BPA’s health effects to better inform regulatory decision‐making”.3

Over a series of meetings and telephone calls during 2011 and 2012, the details of CLARITY-BPA were decided collaboratively. Extensive descriptions of the experimental design,^3,77 including critical analyses of its strengths and weaknesses/limitations, have been published elsewhere.^78–80 Key elements included the use of the NCTR-SD rat, a uniform protocol for animal housing and husbandry that minimised inadvertent exposure to other EDCs, oral dosing by gavage, analytical quantification and certification of dose, diet and background BPA, and the amalgamation of all raw data in a shared repository (Chemical Effects in Biological Systems [CEBS]) for public distribution (https://ntp.niehs.nih.gov/results/areas/bpa). Use of gavage was controversial not only because of concerns regarding stress, but also because it is far more invasive than any route by which humans are exposed. Ultimately, it was adopted because it is standard practice for guideline toxicology studies. Most significantly, all academic investigators worked under blinded conditions, with “unblinding” only occurring following the deposition of the raw data into CEBS for quality assessment and assurance, and approval by the CLARITY-BPA Executive Committee. All animal work was performed at NCTR with the dissected tissues distributed from there. All behavioural work was conducted on site by visiting academic scientists and was over-seen by NCTR scientists and staff.

The dose range deliberately included “low” doses and was: 2.5, 25, 250, 2500 and 25 000 μg kg⁻¹ bw d^–1. For reference, the current No Observed Adverse Effect Level (NOAEL) is 5 mg kg⁻¹ bw d⁻¹ by oral exposure, and the US reference dose (the level considered “safe” for human exposure) is 5 μg kg⁻¹ d⁻¹. Two doses of ethinyl oestradiol (EE; 0.05 and 0.5 μg kg⁻¹ bw d⁻¹) were included as controls for oestrogenicity. A subset of studies had additional positive controls. For example, a propyl-2-thiouracil group was included for the thyroid-specific studies (research still in progress). As in the sub-chronic exposure study that preceded it, CLARITY-BPA had two “arms:” continuous and stop dose. For the continuous dose arm, animals were gavaged daily beginning from GD6 until sacrifice (up to 2 years). For the stop dose arm, animals were only gavaged from GD6 until weaning (PND21) to specifically examine the developmental effects of BPA. Age at death occurred at multiple points thereafter (up to 2 years). The core study was released in September 2018 (https://ntp.niehs.nih.gov/results/pubs/rr/reports/abstracts/rr09/index.html). A summary and review of all study outcomes, include both “core” and “grantee” results, published as of 1 April 2019 is also available,^78,79 with the official NTP comprehensive final report being due in August 2019.

5 |. CLARITY-BPA: BR AIN AND BEHAVIOUR

In addition to us, two other grantee laboratories examined brain and behaviour-related endpoints under CLARITY-BPA, with data now published by us and one other team, and some tissues shared between these and other teams. The outcomes are summarised in Table 1. Our collective set of studies used two different groups of animals. The first was used for behavioural analyses by us and a research team led by Cheryl Rosenfeld from the University of Missouri in collaboration with Sherry Ferguson at NCTR. Exposure spanned GD6 to PND21 (the stop dose arm) and, given the laborious nature of the studies, included only a subset of the available dose groups (vehicle, 2.5, 25 and 2500 μg kg⁻¹ bw BPA, 0.5 μg kg⁻¹ bw EE). To facilitate behavioural testing, study animals were transferred at weaning from the main facility to a separate building, and placed on reverse light. They were then subjected to a battery of well-validated assays assessing anxiety-related behaviours, exploratory behaviour and spatial navigation. Some animals were tested as juveniles and some were tested in adulthood.

TABLE 1.

Effects of developmental bisphenol A (BPA) exposure on oestrogen receptor (ER) expression in the hypothalamus (Hyp) and amygdala (Amyg) are consistent across studies directly related to and including Consortium Linking Academic and Regulatory Insights on BPA Toxicity (CLARITY-BPA)

Dose(μg kg−1 bw day−1)	PND1 In situ hybridisation		CLARITY-BPA PND1 Micropunch/PCR		Subchronic study PND 21/90 In situ hybridisation		CLARITY-BPA PND 90 Rosenfeld Lab
	Hyp	Amyg	Hyp	Amyg	Hyp	Amyg	Hyp	Amyg
2.5	Yes↑	Yes↑	Yes↑	No	Yes↓	Yes↓	–	–
25	Yes↑	Yes↑	No	Yes↑	Yes↓	Yes↓	–	–
250	–	–	Yes↑	Yes↑	–	–	–	–
260	–	–	–	–	Yes↓	–	–	–
2500	–	–	No	No	–	–	Yes↓	–
2700	–	–	–	–	Yes↓	–	–	–
25000	–	–	Yes↑	No	–	–	–	–

Note:: Exposure generally increases ER expression pre-weaning but decreases it thereafter. Effects are consistent, even at the lowest dose of 2.5 μg kg⁻¹ bw, which is lower than the oral reference (“safe”) dose for human exposure.

PND, postnatal day; PCR, polymerasechain reaction.

Despite numerous prior studies by us and others showing robust and reproducible effects of developmental BPA exposure on anxiety and exploratory behaviours, the effects in the CLARITY-BPA study were subtle and sporadic.^81,82 For example, in the open field test, which assesses anxiety and exploratory drive, statistically significant effects of 2.5 and 25 mg kg⁻¹ bw d⁻¹ BPA were identified in the juveniles on a few interval endpoints, such as time resting in the second 5 minutes of the test. In the Barnes Maze, which tests spatial navigation, the adult 2500 μg kg⁻¹ bw BPA females, were less capable than control females of locating the escape box in the allotted time.⁸² Overall, evidence for BPA-related effects on non-reproductive behaviour was minimal and underwhelming. Notably, in the unexposed controls, some anticipated sex differences were either not detected or the opposite of expected, suggesting that some behavioural sex differences may be uniquely different in the NCTR-SD strain compared to other SD strains,⁸¹which is a not atypical finding for inbred rodent strains, and best documented in mice.⁸³

The brains from the juvenile animals were then sent to our laboratory for analysis by unbiased stereology to probe for evidence of abrogated volumetric sex differences in the AVPV, SDN, posteriodorsal portion of the medial amygdala (MePD) and the locus coeruleus (LC).⁸⁴ Expected volumetric sex differences (AVPV, SDN and MePD) were detected in the unexposed controls, and BPA did not eliminate those differences. Unanticipated, however, was that EE did not eliminate them either, which is surprising because oestradiol is well established to be potently masculinising.^10,11 Although one research team had previously reported a volumetric sex difference in the LC sensitive to neonatal steroid manipulation and BPA, neither phenomenon was observed in the CLARITY-BPA animals.^85,86 This dimorphism may be strain-specific and thus not a universally applicable endpoint for endocrine disruption. As in prior studies, however, the AVPV was particularly BPA-sensitive, with all doses of BPA enlarging the female AVPV and a similar enlargement observed in males at the 25 and 2500 μg kg⁻¹ bw dose levels, which are outcomes consistent with our prior work.³⁹

MePD volume, a novel endpoint for BPA studies, was also increased by BPA, although only in the right MePD of 2500 μg BPA kg⁻¹ bw d⁻¹ exposed males. A lateralised effect is biologically plausible because the MePD has numerous structural and functional asymmetric differences, some of which are maintained by circulating androgens.^87,88 That the effect was only observed at a single dose could invite dismissal but, because nuclear volume is typically less sensitive than other region-specific dimorphisms, the phenomenon merits follow-up, particularly in light of our pre-CLARITY evidence of disrupted ER expression in the amygdala (discussed above) (Figure 3). Additionally, work conducted in collaboration with Andrea Gorend scientific statement on endocrine-disrupting chemical′s research team also found evidence of endocrine disruption by BPA in the juvenile rat amygdala, including not only the down-regulation of Erβ, but also the sex-specific disruption of Mc4r, Mc3r and Tac2.⁴² Although the functional significance of MePD disruption is not clear, this subregion of the amygdala integrates olfactory and pheromone information with hormonal, social and other cues to facilitate appropriate adult reproductive behaviours.⁸⁹

The second group of CLARITY-BPA animals analysed by our research team were prenatally exposed and collected on PND1 (Table 1). Thus, this study was akin to the first of the two preCLARITY-BPA studies discussed above and, accordingly, ER expression was of primary interest. Exposure was to one of five doses of BPA (2.5, 25, 250, 2500 and 25 000 μg kg⁻¹ bw), vehicle or two doses of EE (0.05 and 0.5 μg/kg bw). Our pre-CLARITY studies used in situ hybridisation, which has exceptionally high anatomical resolution but is decidedly not high throughput and can only label one gene at a time. To increase the number of genes examined, for CLARITY-BPA, brains were analysed by a combination of targeted and untargeted transcriptomics, with the hypothesis that ER expression and other targets in the ER signalling cascade would be disrupted. The three regions examined were the hypothalamus, hippocampus and amygdala. Each was isolated by microdissection and the RNA analysed by RNA sequencing and also a quantitative reverse transcriptase-polymerase chain reaction for a list of predetermined targets including ERα and ERβ. The greatest number of differentially expressed genes was found in the male hypothalamus and female amygdala.^90,91 In the hypothalamus of both sexes, elevated ERα and ERβ expression was observed at 2.5, 25 and 2500 μg BPA kg⁻¹ bw. In the hippocampus, the only evidence of ER disruption was heightened ERβ expression in the 25 000 μg kg⁻¹ bw males. Similarly, only ERβ was altered in the amygdala with expression levels non-monotonically heightened in both sexes. Pathway analysis in the amygdala of both sexes revealed enrichment for corticotrophin-releasing hormone signalling, which an outcome concordant with extensive prior data suggesting BPA-related effects on anxiety and other stress-related behaviours. Similarly, gonadotrophin-releasing hormone signalling was also identified as a perturbed pathway, and was consistent with prior work by us others showing BPA-related disruption of the AVPV and hypothalamic-pituitary-gonadal axis, even at low doses.

In an era where reproducibility is of paramount significance across biomedical genres following some spectacular failures in preclinical research,⁹² it is remarkable how unequivocally consistent the CLARITY-BPA transcriptome data are with the data obtained in the first study that we conducted in conjunction with NCTR. Both found heightened ER expression in BPA-exposed PND1 animals. Both used strict blinding and other procedures intended to minimise the risk of bias, which makes the results particularly valuable. This reproducibility is especially striking because the two studies used two different techniques. The CLARITY-BPA study used microisolated tissue containing the entire region of interest, which delivered the capacity to assess the entire transcriptome but lacked the anatomical resolution obtainable via in situ hybridisation. Nevertheless, both studies showed that prenatal BPA exposure disrupts neonatal ER expression in the hypothalamus and amygdala (Figure 3).

The transcriptomics analysis also identified BPA-related effects in other hormone sensitive pathways critical for sociosexual behaviours. These included expression of oxytocin and GABA vesicular transporter (Slc32a1) in the hypothalamus, oxytocin in the hippocampus, and androgen receptor, oxytocin and vasopressin receptors in the amygdala. Numerous genes involved in GABA and glutamate signalling were also found to be disrupted in the amygdala. Disruption of oxytocin and vasopressin signalling had previously been identified by us and in other studies as being sensitive to BPA in multiple species.^42,93–96 Similarly, interference with GABAergic and glutamatergic signalling and neurone differentiation has also been shown in several different capacities, especially in the amygdala. Linked functional outcomes vary but include behavioural perturbations and a shift in the age at pubertal onset.^97–101

A CLARITY-BPA study independently conducted by Cheong et al¹⁰² using adult rats from the behavioural studies produced highly concordant trancriptomic results. Only the 2500 μg kg⁻¹ BPA, 0.5 μg kg⁻¹ EE and vehicle animals were examined (exposed from GD6 through PND21 and killed at 3 months of age). Evidence of disrupted hippocampal oxytocin and vasopressin gene expression was found in the 2500 BPA μg kg⁻¹ BPA, with hypothalamic ERα down-regulated in males exposed to BPA or 0.5 μg kg⁻¹ bw EE. Hypothalamic ERβ was only reduced in the EE males. Directionally, these effects are consistent with that found in the PND90 animals from the subchronic exposure study that pre-dated CLARITY-BPA (Figure 3) and at a dose in between those found to be disruptive.⁷² Collectively, the CLARITY-BPA data independently produced by our two teams are consistent with a robust literature originating from a multitude of laboratories showing that BPA impacts oestrogen, oxytocin and vasopressin pathways throughout the brain of multiple species.^{42,48,93,96,103,104}

6 |. SUMMARY AND CONCLUSIONS

The CLARITY-BPA studies, and the two collaborative NCTR studies that preceded them, unequivocally show that ER expression in the rat brain is altered by developmental exposure to BPA at doses as low as 2.5 μg kg⁻¹ bw (Table 1). In neonates, ER expression is generally increased, comprising a phenomenon that likely makes the brain more sensitive to endogenous oestrogen. This may explain why BPA is so often observed to be “oestrogenic” in vivo, despite its limited binding affinity for ERs in vitro.^105–107 Region-specific up-regulation of ERs is also particularly significant given that it is now widely recognised that the brain can synthesise its own steroid hormones, including oestradiol, and thus is not necessarily dependent on circulating levels.^108,109 Disruption of brain ER (both mRNA and protein levels) is one of the most consistently observed outcomes of developmental BPA exposure. Additionally, the CLARITY-BPA studies provide further compelling evidence that developmental BPA exposure alters oxytocin and vasopressin related signalling pathways, and the volume of the AVPV. Evidence of disrupted GABA and glutamate signalling is also suggested, which adds to a small but growing literature on BPA-related effects on neurotransmission.

What remains to be seen is how the data generated by the unprecedented and landmark attempt to produce an “end all, be all” study on BPA will be used by FDA and other regulatory bodies for making human health decisions. Collectively, the data reveal the plausibility of non-linear dose effects, multiple modes of action, and the pressing need to incorporate more endocrine-sensitive organs (such as the brain) and endpoints in guideline studies. Summarised data from the core and grantee studies published to data show evidence of effects at 2.5 μg kg⁻¹ in multiple organs, including the brain, mammary gland, ovary and heart,^78,79 which is an outcome that un-deniably shows that the currently established NOAEL of 50 mg kg⁻¹ bw d¹ is nowhere close to a true “no observed adverse effect level” and urgently needs re-evaluating.

Bisphenol A is now a household name with BPA-free bottles, including baby bottles, readily available, and the “BPA-free” sticker on a multitude of products being commonplace. This outcome, however, is primarily attributable to consumer demand rather than regulatory action. Regulators remain reluctant to limit BPA and other EDCs from consumer products and struggle to synthesise relevant data for EDCs. CLARITY-BPA offers critical solutions by identifying particularly sensitive endpoints, as well as reproducible effects that can readily be added to regulatory testing schemes. The final report for CLARITY-BPA is due in August 2019. Our group remains optimistic that the lessons learned in CLARITY-BPA will help protect developing brains from BPA and other EDCs that are so insidiously pervasive in our built environment.

DATA AVAILABILITY STATEMENT

All data from the CLARITY-BPA study is fully available via the online data bank CEBS maintained by the NTP.