Biosensors Paving the Way to Understanding the Interaction between Cadmium and the Estrogen Receptor Alpha

Cadmium is a toxic heavy metal ubiquitously present in the environment and subsequently in the human diet. Cadmium has been proposed to disrupt the endocrine system, targeting in particular the estrogen signaling pathway already at environmentally relevant concentrations. Thus far, the reports on the binding affinity of cadmium towards human estrogen receptor alpha (hERα) have been contradicting, as have been the reports on the in vivo estrogenicity of cadmium. Hence, the mode of interaction between cadmium and the receptor remains unclear. Here, we investigated the interaction between cadmium and hERα on a molecular level by applying a novel, label-free biosensor technique based on reflectometric interference spectroscopy (RIfS). We studied the binding of cadmium to hERα, and the conformation of the receptor following cadmium treatment. Our data reveals that cadmium interacts with the ligand binding domain (LBD) of the ERα and affects the conformation of the receptor. However, the binding event, as well as the induced conformation change, greatly depends on the accessibility of the cysteine tails in the LBD. As the LBD cysteine residues have been reported as targets of post-translational modifications in vivo, we present a hypothesis according to which different cellular pools of ERα respond to cadmium differently. Our proposed theory could help to explain some of the previously contradicting results regarding estrogen-like activity of cadmium.


Introduction
Cadmium is a heavy metal with no known beneficial physiological function. It is ubiquitously present in Earth's crust, from where it is released by volcanic activity, mining, and use of phosphate fertilizers and fossil fuels. Plants take up cadmium from the soil and form the major source of cadmium intake in nonsmoking, non-occupationally exposed populations [1]. The average daily intake of cadmium through diet is approximately 10-20 mg, which is close to the recently revised tolerable weekly intake of 2.5 mg/kg set by the European Food Safety Authority [1]. Chronic exposure to cadmium causes damage to the kidneys and bone, increases the risk of various cancers, and disrupts reproductive functions in both female and male [2,3,4]. The Interactional Agency for Research on Cancer has classified cadmium as a group I human carcinogen [5]. More recently, studies demonstrating estrogen-like activity of cadmium have raised concerns and led to a classification of cadmium as an endocrine disrupter [6,7,8].
The estrogen receptors ERa and ERb are nuclear hormone receptors that regulate gene expression in response to the female sex steroids estrogens. When the cognate ligand 17b-estradiol (E 2 ) binds to the ligand binding pocket within the ERaLBD, it interacts with specific amino acid residues (glu353, arg394, and his524) leading to a conformational change and the formation of the activation function 2 (AF-2), which is an interaction site for coactivators [9]. The recruitment of co-activators bridges the receptor to the basal transcription machinery and allows the regulation of transcription. Alternatively, the activation of ERa can lead to rapid, extranuclear and thus non-genomic effects like release of secondary messengers and activation of kinases [10]. The mechanisms of the extranuclear activities of the ERs remain less well understood than the genomic activities. Nevertheless, inappropriate regulation of ER activity by environmental endocrine disrupters is believed to be a factor behind the increasing incidence of hormonal cancer in industrialized countries. For example, cadmium exposure is linked to the risk of endometrial and breast cancer in humans [11,12]. These epidemiological connections to increased risk of hormonal cancers in humans render the understanding of the endocrine disruptive mechanisms of cadmium very important.
In mammalian cell culture, cadmium induces the expression of estrogen target genes, triggers activation of cytoplasmic kinases, and promotes proliferation of estrogen responsive cell lines [7,8,13,14,15,16,17,18,19], suggesting that cadmium promotes an agonist conformation of the ER and activates both genomic and non-genomic estrogen signaling. However, these effects are not observed in all studies [19,20,21,22]. Similarly, in rodents cadmium promotes uterine growth, the hallmark of estrogen exposure, in some [6,23,24] but not all studies [23,25,26]. The binding of cadmium to hERa has been studied in two independent reports. Stoica et al. reported cadmium as a strong ligand for the hERa (dissociation constant K D = 5*10 210 M) [8], while Rider et al. classified cadmium as a non-binder [8,27].
In order to help explain these contradictory findings, we decided to apply a novel biosensor technique to study the binding of cadmium to hERa and the conformational consequences of the interaction. Our methodology is based on RIfS, and it represents a label free, time-resolved method for the study of specific interactions between biomolecules. The platform has successfully been applied to ERa before to study the binding to ligands [28,29], DNA [30], and co-activators [31]. The principle of the RIfS assay relies on measuring the change in optical thickness of thin transducer chips that are coated with suitable biomolecules (Figure 1). We applied two different chips in the current study that we have recently thoroughly characterized: one coated with a derivative of the ERa ligand estrone (E 1 ) [29] and another with a peptide that binds to the agonist conformation of ERaLBD [31].

Preparation of the transducer chips
The RIfS transducer chips were cleaned, activated and silanized, and diamino-poly(ethylene glycol) was then immobilized to the surface as described previously [29,31].
The chips coated with E 1 -17-CMO were prepared essentially as described in [29]. Shortly, E 1 -17-CMO was dissolved in dimethylsulfoxide (DMF), N,N-diisopropylethylamine was added, and the mixture was pipetted onto a DAPEG transducer and sandwiched with another DAPEG transducer. The sandwich was then incubated in a DMF-saturated atmosphere for 6 h and finally rinsed with DMF and water.
A detailed reaction scheme of the surface chemistry is provided in Figure S1. RIfS ligand binding assay relies on derivatives of the specific ERa ligand E 1 -17-CMO that are covalently bound to the sensor chip surface. White light is illuminated into the system, and it reflects from the different boundaries of the multilayered chip. In case protein binds to the surface, the reflectivity is changed and the optical thickness, determined from the interference spectra, increases. (B) The dependence of the optical thickness of estrogen content in the sample. In case there are no ERa ligands present in the sample, the receptor is free to bind the surface, and the optical thickness of the surface is high (the situation boxed with green). If there are ERa ligands present, they prevent the receptor from binding the surface leading to decreased optical thickness (the situation boxed with yellow), until all receptor is saturated with the ligand and it can no longer bind the surface (the situation boxed with red). doi:10.1371/journal.pone.0023048.g001

Ligands and reaction mixtures
Stock solutions of 1 mg/mL of E 2 and 4-hydroxy-tamoxifen (4-OHT) were prepared in dimethylsulfoxide and stored at 4uC in the dark until use. CdCl 2 was stored as 0.1 M stocks in the freezer. Different concentrations of CdCl 2 were freshly prepared by dilution with Milli-Q water.
The reactions between ERa and ligands were carried out in a binding inhibition test format, where the receptor is pre-incubated with the ligands before the mixture is injected to the RIfS setup. All incubations were carried out for 60 min at 4uC in the dark.
Ligand binding assay. In ligand binding assays, a constant concentration of the receptor (548 nM for hERa and 208.3 nM for CM-hERaLBD) was incubated with different concentrations of CdCl 2 (2.9 nM-285.7 mM). These mixtures were then investigated via RIfS using E 1 17-CMO modified transducers. When E 2 was titrated to CM-hERaLBD with or without 5 mM CdCl 2 the concentration of E 2 was varied between 3.3 nM E 2 and 1.3 mM E 2 .
A detailed scheme for the binding assay is provided in Figure  S2.
A detailed scheme for the conformation assay is provided in Figure S3.

The principle of the RIfS-based ligand binding assay
The RIfS setup consists of a halogen white-light source and a Yoptical fiber, which guides the light to the transducer chip. The reflected light is travels through the same optical Y-fiber to a diode array spectrometer (Spekol-1100, Analytik Jena, Germany). The liquid handling system consists of a Hamilton dilutor Microlab (Hamilton, Switzerland) with two syringe pumps and a 4-way valve. Data acquisition and evaluation was performed using internal software.
The binding of the receptor to the surface was monitored for 250 s after a 100 s baseline period, followed by a 300-600 s dissociation phase and a regeneration step with 6 M guanidiniumhydrochloride pH 2 and a 240 s baseline period. The optical thickness was followed by recording interference of white light reflected at the interfaces of the chip by a diode array spectrometer. Binding curves were recorded as changes of the apparent optical thickness [nm] versus time [s]. All measurements were carried out in 500 mM Tris buffer containing 100 mM KCl with a pH of 7.4 at room temperature (,25uC). The assays were repeated three times and in figures mean optical thickness with standard deviation is presented.
The concentration of the receptor in the reaction mixture was adjusted so that the binding signals revealed a linear shape. By providing an excess of possible interaction partners on the surface, it could be ensured that every free receptor molecule could bind to a superior number of free immobilized ligands on the surface. Under these conditions, the obtained sensor signal is dependent on the diffusion of receptors to the sensor surface expressed by Fick's law, where J D is the diffusion flux, D the diffusion coefficient of the biomolecules in solution, c the concentration of these biomolecules, and d the thickness of the diffusion layer. The slope of the binding curve is proportional to the concentration of free receptor in solution. Assuming that Cd binds to the same binding pocket as the steroidal estrogens, such as E 2 , the equation for the affinity constant deducted from the mass equation results in the following formula, where c RL is the concentration of the receptor-ligand complex, c R the concentration of receptor, c L the concentration of ligand, c R,0 the initial concentration of receptor, and c L,0 the initial concentration of ligand. The concentration of receptor-ligand complex is accordingly The concentration of free receptor can therefore be expressed as Knowing this relation between the equations it is possible to determine the affinity constant directly as a parameter of the fit function.

Measurement of biomolecular interactions using RIfS
Biosensors represent relatively young analytical methodology, but have proven to be very powerful in combination with nuclear receptors [32]. Among the large group of biosensors, the label-free techniques provide particular advantages. Their ability to trace biomolecular interactions in a time-resolved manner without a need for labeling minimizes the possibility of artifacts. In addition, in contrast to other more popular biosensors, such as surface plasmon resonance, RIfS uses glass instead of metal films as transducers. In other words, the RIfS surface chemistry does not rely on thiol/gold chemistry which is a great advantage in the present study considering the possible interaction between thiolcontaining polymers and cations such as cadmium.
We applied two different sensor chips in our studies, one coated with the ER specific ligand E 1 -17-CMO, and one coated with a peptide called a/b I that binds to the AF-2, i.e. the agonist conformation of ERaLBD. Both assay formats have been developed and characterized by us [29,31]. The principle of the RIfS based binding assay is presented in Figure 1 (and discussed in detail e.g. in [33]). We first pre-incubated the receptor protein in solution with CdCl 2 or control ligands, and then introduced the mixture to the sensor chips. The binding of the receptor to the surface was monitored over time with the help of white light passing through the chip, and the reflected light was recorded with a diode array spectrometer. Association (and dissociation) of the receptor to the surface changes the reflectivity of the multilayered chip, which is seen as changes in the interference spectrum, and can be calculated as optical thickness of the chip ( Figure 1A). Using the curve showing the change in optical thickness over time we then calculate the relative slope, which is simply defined as the change of optical thickness over time in a given treatment. Finally we normalize the results in one series of experiments to the highest observed slope, which is normally seen in the blank sample containing untreated receptor. For example, in case cadmium does not bind to ERa ligand binding pocket, the receptor is free to bind to the E 1 -17-CMO coated surface, which we observe as an increase in the optical thickness, which gives a high relative slope ( Figure 1B). Similarly, in the conformation assay, if cadmium induces an agonist conformation, the receptor will bind to the a/b I coated surface, and the optical thickness is increased.

Binding of cadmium to hERa
Currently, there is little doubt that cadmium has an effect on the estrogen signaling. However, it is striking that the nature of the reported effects varies from strong agonist to antagonist in the different model systems [6,8,21,22,26,27]. To understand these inconsistencies, it is essential to know how exactly cadmium interacts with the ER. To our knowledge, only two studies have assessed the binding of cadmium to hERa: Stoica et al. classified cadmium as a strong agonist while Rider et al. concluded that cadmium is a non-binder that, however, disrupts the binding properties of E 2 [8,27]. In both reports, cell based and cell free radioligand binding assays were utilized side by side. To gain new insights into the binding mechanism, we applied the RIfS methodology developed in our laboratory. We incubated different concentrations of CdCl 2 with a constant concentration of recombinant full length hERa for 1 h, after which we quantified the relative amount of free ERa with the E 1 -coated RIfS sensor chips. Our results show that CdCl 2 dose-dependently decreases the optical thickness of the surface (Figure 2A). In other words, cadmium binds to hERa. In the cell free binding assay format, Other additional minor details, such as buffer composition, incubation times and reaction temperatures could further affect the result. Although the two K D values differ, they both suggest that cadmium has marked affinity towards the hERa.
Although Rider et al. also observed interaction between cadmium and hERa in their cell free assay, they could not calculate a dissociation constant, because the data did not fit the one-site competitive model. In fact, also Stoica and coworkers noted that the binding of cadmium to hERa displayed noncompetitive features. In agreement with these observations, we also observe that the shape of our data fits poorly to the binding curve ( Figure 2A). The RIfS binding assay assumes competitive binding of ligands to the same ligand binding cavity following the mass equation [R]+[L]«[RL]. Therefore, our data could reflect non-competitive binding mode in agreement with the earlier reports [8,27]. The non-competitive binding mode, and subsequent difficulty to fit the data into a one-site model, could further explain the difference in the K D value in our work compared to that of Stoica et al.
We next determined the effect of cadmium on the conformation of hERa. We incubated hERa with vehicle (buffer without ligand or CdCl 2 but with corresponding amount of solvent), CdCl 2 , E 2 , the antiestrogen 4-OHT, or combinations of CdCl 2 with E 2 or 4-OHT. As expected, E 2 treated hERa bound to the a/b I coated surface, indicating an agonist conformation, while vehicle and 4-OHT treated hERa did not ( Figure 2B). To our surprise, CdCl 2 treated hERa did not bind to the surface. Moreover, when hERa was co-incubated with E 2 and CdCl 2 , it no longer bound to the surface ( Figure 2B), indicating that cadmium destroys E 2 induced agonist conformation of ERa.
In summary, these results suggest that cadmium binds to hERa with high affinity in a non-competitive manner without inducing an agonist conformation, and that cadmium disrupts the agonist conformation induced by E 2 . Based on these observations, it is more likely that cadmium would trigger antiestrogenic effects in vivo than estrogen agonist effects. Indeed, some reports in the literature support this hypothesis. Silva et al. did not observe effects on estrogen-dependent breast cancer cell proliferation with cadmium alone, but in the presence of E 2 cadmium inhibited the proliferation [22]. Similar results were reported in yeast hERa transactivation assays [21,22]. Furthermore, three independent in vivo studies suggest that cadmium does not promote uterine growth in rodents, suggesting lack of estrogen activity [23,25,34].

Binding of cadmium to hERa with protected cysteine residues
Although the above results gain some support from the literature, we do realize that they contradict several other publications reporting estrogen-like effects for cadmium [6,7,8,13,14,15,16,17,18,23,24]. We were concerned that the high concentration of preservatives present in the commercial ERa preparation might have interfered with our assay. According to our calculations, the 537 nM hERa solution used in our assay contained a total concentration of 343 mM of ethylenediaminetetraacetic acid and dithiothreitol -both well known to complex Cd 2+ . Hence, we wanted to repeat the assays without preservatives. However, native ERa quickly destabilizes in the absence of additives due to unprotected cysteine residues. In order to solve this problem, we chose to use an hERa preparation consisting of the LBD (amino acids 301-553) with carboxymethylated cysteine residues [9]. This receptor variant is expected to maintain its normal ligand binding properties, as none of the LBD cysteine residues are in direct contact with the ligand [9]. In fact, we have previously utilized this receptor in the RIfS binding assay without problems [29].
We repeated the ligand binding and conformational assays using CM-hERaLBD instead of full length hERa. Unexpectedly, cadmium did not affect the binding of CM-hERaLBD to the sensor surface ( Figure 3A). In the case of full length hERa, 300 mM CdCl 2 completely blocked the receptor from binding to the surface (Figure 2A), but in the case of the CM-hERaLBD, the same concentration had no impact on the receptor, and with a 10fold higher concentration only a minor reduction in the optical thickness was observed ( Figure 3A). To control that the batch of CM-hERaLBD used in the assay functioned as it should, we measured the binding affinity of E 2 to it. The obtained binding curve displayed a good fit (Figure 4, black curve) and the calculated K D value of 2.0*10 210 M is in good agreement with reported values for full length, unprotected hERa [8], demonstrating that the batch of CM-hERaLBD at hand behaves like the native receptor. To further test the effect of cadmium on this receptor, we repeated the binding assay with E 2 in the presence of cadmium. The E 2 binding curves were essentially identical in the presence and absence of CdCl 2 , indicating that cadmium does not affect the binding of E 2 to the CM-hERaLBD (Figure 4). These observations suggest that the LBD cysteine residues may have a central role in the binding of cadmium to hERa.
In addition to the protected cysteine residues, the CM-hERaLBD differs from the full length ERa in one more important aspect: it lacks the domains outside the LBD. The lack of the N-terminal A/B and central DNA binding domains could have an impact on the behavior of the receptor in our assay. Although both receptors variants we use behave normally in response to E 2 , the possibility remains that unorthodox ligands such as cadmium have a binding mode so distinct, that multiple domains are involved. The impact of cadmium on the DNA binding domain has been addressed before. Exchanging the zinc finger Zn 2+ ions with Cd 2+ has no functional consequence in terms of ERa DNA binding and transcriptional activity [35,36]. We find it therefore unlikely that the presence or absence of this domain in our assay causes the dramatic change in the affinity of the receptor towards cadmium. The N-terminal A/B domain harbors the ligand independent activation function 1 (AF-1) that contributes to ER activity. Upon ligand binding, AF-1 and AF-2 can act together to synergistically boost ERa activity [37], and on the other hand, growth factors can activate ER in the absence of ligands through phosphorylation of The binding of cadmium to hERa was studied by incubating different concentrations of CdCl 2 (2.9 nM-285.7 mM) with a constant concentration (c R 548 nM) of hERa. The mixtures were then guided to the sensor surface and the optical thickness monitored. Cadmium dose dependently reduces the thickness, indicating that it binds to hERa. The affinity constant (K A ) is given as a fit-parameter while O is an offset allowing the fit to be adjusted in y-axis direction. (B) The conformation of hERa was assessed after the receptor was incubated with vehicle, CdCl 2 , E 2 , 4-OHT, CdCl 2 +E 2 , or CdCl 2 +4-OHT. The average binding curves (recorded for 700 sec) of three independent assays is shown with the standard deviation depicted as dotted lines. Only E 2 alone triggered an agonist conformation recognized by the surface. Cadmium alone and in combination with other ligands prevented the formation of agonist conformation. doi:10.1371/journal.pone.0023048.g002 serine residues in the AF-1 [38]. The synergistic activities of AF-1 and AF-2 involve co-regulator proteins that bridge these domains together, and the ligand independent activation through AF-1 relies on kinase activity. Our experimental setup does not contain co-regulators or kinases, and therefore we think that it is unlikely that the absence of AF-1 in the CM-hERaLBD is the reason for the discrepancy between the full length and truncated receptor affinities towards cadmium. It is more likely that this difference is due to the availability of the cysteine tails in the LBD, which have been shown to be crucial for cadmium-hERa interaction by other too [8,39].
We further tested the effect of cadmium on the carboxymethylated receptor in the conformation assay. As expected, E 2 induced agonist conformation while vehicle and 4-OHT did not ( Figure 3B). However, to our great surprise, CdCl 2 had marked effects on the conformation of the CM-hERaLBD. The binding curves show that the presence of CdCl 2 in the reaction mixtures promotes the formation of the agonist conformation of the receptor in all environments: alone, in the presence of an agonist (E 2 ), and in the presence of an antagonist (4-OHT) ( Figure 3B). In summary, these observations suggest that cadmium does not bind to CM-hERaLBD in a manner that would prevent binding of E 2 (i.e. not to the same ligand binding pocket), but cadmium does affect the conformation of the receptor. It seems as if cadmium, working outside the ligand binding pocket in the ERaLBD, can pull the receptor towards an active conformation. This would predict estrogen-like activities for cadmium in biological systems. Indeed, several studies have suggested that cadmium promotes estrogen signaling in cell culture and in experimental animals [6,7,8,13,14,15,16,17,18,23,24].
In conclusion, carboxymethylation of the cysteine residues in the LBD of hERa did not change the behavior of the receptor towards E 2 , but dramatically altered the behavior towards CdCl 2 . It is unlikely that these differences depend on the additives present in the commercial full length hERa, or domains outside the LBD. Instead, the results suggest fundamental differences in the binding mechanism between E 2 and cadmium. Cadmium can coordinate with several amino acid tails, and a model has previously been proposed where cadmium interacts with cys381, cys447, glu523, his524 and asp538 in the hERa LBD [39]. The fact the mere carboxymethylation of the cysteine residues had such a great effect in our assays further emphasizes the role of these residues.

A working hypothesis
The data presented here suggests that cadmium may interact with hERa in different manners depending on the state of the cysteine residues in the LBD of the receptor. The two models, one suggesting antagonistic and the other agonistic activity, both gain support from the literature. The burning question now is, whether both models could function in biological systems. In other words, can the state of the cysteine residues in the ERa LBD be affected in living cells?
The properties of many proteins are modified by posttranslational modifications, where amino acid tails are for instance phosphorylated, sumoylated, or acetylated [40]. Various posttranslational modifications have been reported to affect ERa, and although most of them are centered around the variable A/B domain in the N-terminus of the protein [41], the cys447 in the LBD is a target of palmitoylation [42]. Palmitoylation is a mechanism to associate proteins to cell membranes. Interestingly, cadmium has been shown to activate markers of membrane associated estrogen signaling in breast cancer cell lines in vitro as well as in rodents in vivo [14,15,24,34]. The cysteine residues could also be modified in redox processes, a phenomenon that has so far been studied only in the DNA binding domain of ERa [43].
Based on our results, we have formulated a working hypothesis where the effect of cadmium on estrogen signaling is a net result of different cellular pools of ERa responding to this heavy metal differently ( Figure 5). The ERa cysteines could be affected for instance by oxidative stress (that cadmium can cause) or post-   (marked with a stick). The cysteine residues may become unavailable for interaction with cadmium for instance by posttranslational modifications (PTM) or oxidation. When a total population of ERa is exposed to cadmium, the metal binds close the ligand binding pocket of unmodified ERa and prevents the formation of active conformation recognized by AF-2 interacting co-activators, but in the case of modified ERa (the diamond indicates modification of the cysteine residues) cadmium binds outside the ligand binding pocket and pulls the receptor towards the active conformation that allows interaction with co-activators. (B) In living cells, membrane-associated ERa and nuclear unmodified ERa could be examples of receptor pools with different reactivity towards cadmium. doi:10.1371/journal.pone.0023048.g005 translational modifications, creating pools of ERa with a distinct reactivity towards cadmium ( Figure 5A). This could lead to a hypothetical scenario where, for instance, palmitoylated ERa at the membranes is stimulated by cadmium whilst nuclear, unmodified is not ( Figure 5B). Interestingly, support for this hypothetical model comes from two independent recent reports that have studied the effects of cadmium on both rapid, membrane associated estrogen signaling as well as on the nuclear estrogen signaling. Using human breast cancer cells, Zang et al. showed that cadmium promotes rapid ERK1/2 phosphorylation, which is a typical membrane associated estrogen effect, while not having an effect on the transcription of estrogen target genes pS2 and PgR [19]. Essentially the same pattern of activity was observed in vivo in mice by Ali and coworkers: they treated estrogen reporter mice with CdCl 2 and observed phosphorylation of ERK1/2 in liver but no effects on transcriptional activity of the nuclear ERa [26]. It is worth noting that under suitable conditions extranuclear kinase activation can lead to transactivation of nuclear ERa [44]. The possible differing responsiveness of distinct cellular ERa pools to cadmium, combined with the inherent connection between the extranuclear and nuclear estrogen signaling, could help to explain some of the contradicting results on the estrogenicity of cadmium.
Future studies should vigorously test our working hypothesis by examining the effect of post-translational modifications and oxidative stress on the responsiveness of ERa towards cadmium in the nuclear and extranuclear parts of the estrogen signaling pathway. Figure S1 Surface chemistry reaction scheme. Two differently functionalized transducers were used. Both were prepared in the same way for the first two steps: 1) Coupling of 3-glycidyloxypropyl-trimethoxysilane to the activated surface (in acetone for 1 h at room temperature) followed by 2) coupling of DAPEG (in dichloromethane for 18 h at 70uC). The transducers were then either 3) covalently modified with E 1 17-CMO (in DMF for 6 h at room temperature) using diisopropylcarbodiimide activation or 4) covalently modified with biotin (in DMF for 1 h at room temperature). The biotin-modified transducers were then put into the RIfS setup and rinsed with 5) Streptavin (1 mg/mL for 250 s at room temperature) and subsequently rinsed with 6) biotinylated a/b I peptide solution (1 mg/mL for 250 s at room temperature). The last two steps were also monitored online using RIfS. (TIF) Figure S2 Ligand binding assay scheme. A constant amount of hERa (548 nM for hERa and 208.3 nM for CM-hERaLBD) is incubated with different concentrations of estrogenic ligands (top row). With increasing amount of estrogenic ligand the binding of the receptor to the transducer chip is inhibited, resulting in a smaller increase in optical thickness and smaller relative slope. By plotting the relative slope values as a function of ligand concentration, and applying a non-linear fit (described in detail in the materials and methods section), the affinity of the ligand towards the receptor can be calculated. (TIF) Figure S3 Conformation assay scheme. A constant amount of hERa (3.3 mM) is incubated with either E 2 (33 mM) or 4-OHT (24 mM) in the presence or absence of CdCl 2 (3.3 mM). After the incubation phase of 1 h at 4uC, helix 12 of the receptor LBD is either in a typically agonistic conformation (symbolized by the cylinder in the green receptor, right) or in an antagonistic conformation (symbolized by the cylinder in the red receptor, left). When rinsing these mixtures over the a/b I modified transducer chip and monitoring the optical thickness as a function of time, the different conformations result in the different binding curves (higher green curve for agonists and lower red curve for antagonist) in RIfS. (TIF)