BPA affects otolith development
Despite the fact that the effects of BPA on development have been the subject of numerous investigations, few experiments have explored effects during embryogenesis [18, 19, 36]. Here, the distinct advantages of the aquatic, free-living embryos used in this study come to the fore, notably their external and maternally-independent development, and their transparency that together facilitate investigation of morphological phenotypes. Indeed, the main result from our study is that BPA specifically affects otolith development during zebrafish and Xenopus development. The otolith phenotype that we observed is not linked to the cardiac effect previously observed since the timing of the two events differ, the otolith phenotype being dependent on treatments carried out during earlier developmental periods. Our data show that BPA acts early in development, during zebrafish embryogenesis, before 20 hpf to affect otolith formation and that this effect is specific to BPA and closely related bisphenols, since another molecule of the same class (e.g. BPC) does not have the same effect. This observation underscores the possible relevance of the phenotype we describe.
It is interesting to note that the timing of BPA exposure resulting in otolith phenotype actually corresponded to the timing of otic vesicle organogenesis. The otic placode becomes visible at approximately 16 hpf and forms a vesicle with a lumen by cavitation at approximately 18 hpf  (see Figure 2A). Two otoliths appear in the lumen by 19.5 hpf, and at about 24 hpf, the first sensory hair cells are seen, grouped in two small patches, one beneath each otolith, corresponding to the future macula. We found that BPA affects inner ear development in a precise window, at the time of the first sign of morphological appearance of the otoliths, that is at 20 hpf. Indeed, if BPA treatment was delayed till just 22 or 24 hpf, no effects on otolith formation were found. This precise timing of BPA action is indicative of an interaction with specific targets active during early otolith development. These observations emphasize that the effects are not due to acute toxicity but are due to a specific interference with a developmental process. The fact that other bisphenols are inactive also suggests that BPA interacts with a precise selectivity on specific targets. In addition no sign of acute cellular stress such as apoptotic cells or necrotic patches were detected in the embryos after BPA exposure (data not shown).
It is important to note that BPA induces abnormalities of otolith development in two distant vertebrate species, zebrafish and Xenopus. Interestingly, the two phenotypes are different. In zebrafish we observed aggregates of otoliths with sometimes loss of one of the otolith. In contrast, in Xenopus BPA induces a strong dose-dependent reduction in the development of otoliths and a severe reduction of the semi-circular canals. In zebrafish we sometimes observed abnormalities of semicircular canals but with variable penetrance. We consider that this late phenotype may be a consequence of an early defect. In Xenopus, BPA was recently shown to induce malformations of the head region . In this study, treatment with 25 or 50 μM of BPA resulted in scoliosis and malformation of the head region including a shortened distance between the eyes. Here, we observed a similar phenotype when BPA treatment was begun at the egg stage at these concentrations.
Our experiments reveal that only some bisphenols affect otolith development. BPA, BPE and to a lesser extent BPF induced otolith phenotypes whereas BPC was inactive. These differences could be attributed at least in part to dissimilarities in terms of bioavailability and metabolism of bisphenols in embryos, resulting in variations in concentrations of these bisphenols in target tissues. Comparing the fate of BPA and BPF, the bio-concentration and biotransformation of these compounds differed substantially. Residual levels of BPA in embryos were 2.5 fold higher than calculated for BPF. Furthermore, in the case of BPA no break-down products were found, as most of BPF was fully metabolized by the end of the experiment. Due to methodological limitations, these metabolic analyses could not be extended to other bisphenols. However, the data indicate that the extent of the phenotypic effects may be related to the uptake, biotransformation and elimination of these chemicals by zebrafish embryos. Perez et al. , using MCF7 human breast cancer cells in culture, demonstrated that the estrogenicity of bisphenols was influenced not only by the length of the substituents at the bridging carbon, but also by their nature. In a yeast two-hybrid assay, Chen et al.  ranked the estrogenicity of bisphenols as follows: BPB > BPP > BPA > BPE > BPF > BPS. Interestingly we found that some of these compounds also act to reduce pigmentation and that the pharmacology of this effect is clearly different: for pigmentation BPE and BPF are clearly the most active compounds, followed by BPC and BPA that show modest effects. These data suggest that bisphenols interfere with different targets in the otoliths and in the pigmentation developmental pathways.
The lowest concentration at which an effect was observed was 5 μM, but with low penetrance (less than 5% affected embryos). At 25 μM the majority of the embryos were affected and all at 70 μM. The lowest active concentration (5 μM) corresponds to more than 1 mg/mL several orders of magnitude higher than the levels reported in aquatic environments [4–6] or in human serum [8, 9]. Of note however is the fact that the concentration of BPA found in amniotic fluid was approximately five-fold higher than levels measured in maternal plasma . Even these high concentrations however are markedly lower than the one needed in our assay to generate a phenotype. Thus, the effect we observe in the aquatic species is probably not relevant to human populations.
A previously undescribed otolith aggregate phenotype
The development of the inner ear in fish and amphibians is representative of the development of inner ears of vertebrates in general. Vertebrate inner ear development is a self-contained model system for fundamental research into genetic control of development and, as shown here, for detecting the effects of potential EDCs on early vertebrate development. Whitfield et al.  carried out a large scale screen of zebrafish mutants induced by N-ethyl-N-nitrosurea treatment. Mutations occurring in no less than 58 genes leading to defects in the development of either the semicircular canals and/or the otoliths were identified . In this screen the authors did not examine fine morphology of the inner ear, such as hair cells, and so they estimate that many more genes are potentially involved in the process.
Interestingly, BPA induces a new zebrafish otolith phenotype. Most of the phenotypes of otolith formation are caused by the specific knock-down or mutation of a given gene [47, 68]. Many of them result in a small malformed ear, a defect in the AP or DV axis, or in specific deletion of specific structures (semicircular canals, otoliths) but to our knowledge the phenotype we describe here has not been previously described. The abnormalities we observe included aggregation of up to 18 otoliths (Figure 1B, lower panel), appearance of multiple otoliths or their absence (Figure 1B). It is clear that the majority of the embryos exhibit a bilateral aggregation of both the anterior and posterior otoliths. This is clearly different (or opposite to) to the loss of otolith observed by Blasiole et al.  after treatment with ouabain or morpholino knock-down of the Na+/K+ ATPase α1a.1. In fact the phenotype that we find most similar to the effect of BPA is the one seen after inhibition of the oc90 gene, a member of PLA2-like otoconin family . It is interesting to note that oc90 is clearly upregulated in the otic vesicle after BPA treatment suggesting it may be play an important role in generating the BPA phenotype. Nevertheless we still observed the otolith aggregation induced by BPA following oc90 knock-down, suggesting that this gene is not required for BPA action. Given the fact that BPA up-regulates oc90 expression we cannot exclude that oc90 act as a modifier of BPA action but this remains to be addressed experimentally.
The developmental basis of the phenotype was analyzed by studying the expression of gene markers. We found that the patterning of the otic vesicle placode and the hindbrain segmentation are normal and that the antero-posterior and dorso-ventral axes of the otic vesicle are correctly patterned. Thus, despite the fact that BPA acts prior to 20 hpf we only detect defects with gene markers at later stages, suggesting these modified expressions derive from early action of BPA on an unknown target. The change of gene markers observed suggests that the phenotype extends much further than a simple abnormality in otolith formation. Indeed, with aldh1a3 and/or ugdh we observed a loss of expression in specific regions such as cranial epithelial projections, endolymphatic duct and in the posterior cristae. This result is in accordance with the fact that we often see, with a variable penetrance, morphological alterations of the semicircular canals. Of note are the changes of expression observed in otic vesicle markers after BPA treatment, which are reminiscent of the changes reported by Petko et al.,  after oc90 inhibition, with a normal msxC and dlx3b expression in both cases. The early timing of the effects in our data suggest that BPA exposure can have multiple consequences for the late development of the ear, resulting in abnormal formation of the otolith as well as other defects such as in the semicircular canals. Unfortunately, results from the in situ analysis do not really clarify the causes of the dramatic phenotype in otolith aggregates. Up-regulation of oc90 might be a determinant factor for this phenotype. In fish, otoliths are composed of calcium carbonate crystals condensed on a core protein matrix. oc90, is the ortholog of the mammalian otoconin-90 gene which encodes the major matrix protein of otoconia. Therefore it is tempting to speculate the with an excess of signal that will lead to the development of the otolith the consequences will be an excess of otolith as observed in our BPA treated embryos. The precise mechanisms underlying this effect have yet to be understood.
Interaction with nuclear receptor pathways
Much data in the literature show that BPA interferes with the estrogen signalling pathway through direct binding to ERα or ERβ and thereby induce an ER agonist or SERM (Selective Estrogen Receptor Modulator) . We have indeed verified that BPA binds to and positively regulates the activity of the three ERs present in zebrafish, although at higher concentration (ED50 at the micromolar range) than in mammals .
Estrogen receptor signalling is intricately implicated in the development, function and regeneration of the inner ear . It was therefore a reasonable working hypothesis that the effects of BPA implicated interference with ER signalling during otic vesicle development. We tested this hypothesis in three ways. First, we examined the effects of estrogen agonist and antagonist activity on otolith development either alone or in combination with effective doses of BPA. We observed only a weak increase of the cotreatment of BPA and ICI on the number of affected fish (see below). Secondly, we studied if BPA in zebrafish could activate an ERE-Luc transgenic reporter gene . We found a modest effect at the concentrations that clearly induce the phenotype (10 μM). These data thus suggest that BPA does not act in an estrogen receptor-dependent manner during otolith development. Several recent observations in the literature indeed suggest that BPA have a much broader set of targets than previously expected. For example, recently the genes responding in male zebrafish liver to the exposure to 17 β-estradiol, and BPA were identified through a microarray analysis . Interestingly these data reveal that the transcriptional network regulated by 17 β-estradiol and BPA in zebrafish are very different. These observations are fully consistent with estrogen receptor-independent effects of BPA that we describe here through developmental analysis. The fact that we observed a synergy when we treated the fish with BPA and with the anti-estrogen ICI suggests a possible connection of the BPA targets and estrogen signalling. This idea is consistent with several recent observations suggest that both in mammals and fish the estrogen signalling play a role in hearing, suggesting a possible connection with otic vesicle formation [57, 58]. The molecular and developmental basis of this cross-talk remains to be explored.
Several data have recently suggested that both TR isoforms (TRα and TRβ) can be targets of BPA (reviewed in ). In transient transfection experiments, BPA (in the micromolar range) suppressed T3-stimulated transcriptional activity stimulated in a dose dependent manner. Since T3 and TRβ are important for ear development [63, 64] we thus tested if the action of BPA in zebrafish otolith development could be linked to its interference with the TR signalling system. We observed no obvious interaction between BPA, T3 and otolith formation: the addition of 1 μM T3 did not interfere, either positively or negatively, with the phenotype induced by BPA. BPA was tested at an optimal (70 μM) or a suboptimal (35 μM) dose of BPA and in none of these situations did we observe an effect of added T3. Thus, interference with the TH signaling pathway does not seem to be responsible for the BPA-induced phenotype. However, induction of deactivating deiodinases by the exogenous T3 cannot be excluded, a phenomenon that would neutralize the exogenous T3. The lack of BPA interaction with TH signalling is in accordance with our previous observation that altered expression patterns of TRs during otic vesicle development in zebrafish could not be detected .
The BPA-induced phenotype and Na+/K+ATPase activity
Na+/K+ ATPase are key enzymes implicated in cell homeostasis by the maintenance of electrochemical gradients. Na+/K+ ATPase activity has been implicated in otic vesicle formation and otolith development shown by treatment with a chemical inhibitor and morpholinos that inhibit a subunit of this enzyme [54, 66]. The Na+/K+ ATPase enzyme is composed of functionally distinct subunits each with multiple isoforms encoded by different genes . In the case of inner ear development, inhibition of Na+/K+ ATPase subunits interferes with otolith and semicircular canal biogenesis, though the mechanisms implicated are still unclear . Numerous cell types are implicated in the formation of otoliths . In fish, otoliths are secreted into the otocyst (the part of the inner ear containing the otoliths) then captured by cilia of specialised cells that tether the anterior and posterior otoliths in place at each pole of the otocyst. The first crystals tethered serve as seeds for further growth. Na+/K+ ATPase could be required for determining electrochemical gradients during various parts of these processes. In the growth of the semicircular canals, Na+/K+ ATPase subunits are expressed in the areas that give rise to the protrusions that determine their growth.
In our experiments both BPA and ouabain, the inhibitor of Na+/K+ ATPase, induced opposite phenotypes and combination of lower doses of each chemical rescued otolith formation. This result could indicate that the effects of BPA are dependent on an Na+/K+ ATPase activity and one of the heterodimeric enzymes expressed in ear. To test this hypothesis knock-down of α1a.1, the major Na+/K+ ATPase subunit expressed in the otic vesicle was carried out and showed that, indeed, the BPA phenotype cannot be observed without this enzyme. Nevertheless this effect is difficult to interpret since it may be simply be linked to the inability of α1a.1-morphants embryo to form otoliths properly. Combined with the pharmacological interaction noted above (BPA rescues the effect of ouabain; Figure 6), a possible scenario is that α1a.1, and more generally Na+/K+ ATPase activity act downstream of BPA target genes in the otic vesicle. However, this is not the case as we saw no effects of BPA exposure on the expression of α1a.1 or on other Na+/K+ ATPase subunits (Figure 6 and data not shown). It is also possible that the two pathways are independent. If α1a.1 is strictly required for otolith formation, the morphogenesis defects induced by BPA cannot be observed in the absence of otolith. By injecting a sub-optimal dose of α1a.1 the signalling cascade dependent on α1a.1 can be activated causing a small otolith to form. Once formation is initiated, this otolith treated with BPA will still form an aggregate but due do a low level of α1a.1protein present in the embryo, due to morpholino injection, the aggregate phenotype is mild compared to wild type. This scenario implies that the effect of BPA on Na+/K+ ATPase and otolith formation is indirect. However, if this were the case one would not see a rescuing effect of BPA and ouabain combination that we observed. Taken together, these results suggest a link between BPA and Na+/K+ ATPase during induction of otolith malformation, but the precise molecular nature of this link and the exact degree of this dependency remains to be explored.
In this manuscript we show that BPA affects otolith development in two vertebrate species, zebrafish and Xenopus. In both cases the deleterious effects are limited to a time window corresponding to the period of otic vesicle development. Moreover, neither estrogen nor estrogen antagonists, nor TH modify the BPA response, underscoring the concept that the effects observed are estrogen and TH-independent. We conclude that vertebrate organogenesis can be modified by BPA through novel mechanisms. A clear link with the Na+/K+ ATPase system was observed. Our experiments provide an example of a developmentally relevant defect generated by BPA treatment that is independent of its main mode of action which is interaction with the estrogenic or TH signalling pathways. These results suggests that the classical way of studying endocrine disruptors, that is by defining precise "end points" based on classical risk assessment data can some time be misleading because the focus on a specific type of defect [62, 72]. In these cases, phenotypes based on unknown mode of action of the molecule will be overlooked. We believe that zebrafish and Xenopus, that allow medium-throughput screening of EDCs, will be powerful systems to test the effects of these compounds in an unbiased manner.