The molecular basis of OH-PCB estrogen receptor activation

Polychlorinated bisphenols (PCBs) continue to contaminate food chains globally where they concentrate in tissues and disrupt the endocrine systems of species throughout the ecosphere. Hydroxylated PCBs (OH-PCBs) are major PCB metabolites and high-affinity inhibitors of human estrogen sulfotransferase (SULT1E1), which sulfonates estrogens and thus prevents them from binding to and activating their receptors. OH-PCB inhibition of SULT1E1 is believed to contribute significantly to PCB-based endocrine disruption. Here, for the first time, the molecular basis of OH-PCB inhibition of SULT1E1 is revealed in a structure of SULT1E1 in complex with OH-PCB1 (4ʹ-OH-2,6-dichlorobiphenol) and its substrates, estradiol (E2), and PAP (3’-phosphoadenosine-5-phosphosulfate). OH-PCB1 prevents catalysis by intercalating between E2 and catalytic residues and establishes a new E2-binding site whose E2 affinity and positioning are greater than and competitive with those of the reactive-binding pocket. Such complexes have not been observed previously and offer a novel template for the design of high-affinity inhibitors. Mutating residues in direct contact with OH-PCB weaken its affinity without compromising the enzyme’s catalytic parameters. These OH-PCB resistant mutants were used in stable transfectant studies to demonstrate that OH-PCBs regulate estrogen receptors in cultured human cell lines by binding the OH-PCB binding pocket of SULT1E1.

The SULT1E1 isoform plays a pivotal role in regulating the cellular activities of estrogens. The sulfuryl moiety prevents estrogens from binding to and activating their receptors (14,15). Consequently, SULT1E1 activity is linked to physiological processes in which estrogens are engaged, which positions the enzyme as a potential therapeutic target in circumstances where enhanced estrogenic activity (by inhibiting its inactivation) is desirable (e.g., metabolic syndrome (16), diabetes (17), renal failure (18), and estrogen therapeutic augmentation (19)).
Hydroxylated polychlorinated biphenyls (OH-PCBs) are the most potent known SULT1E1 inhibitors-K i OH-PCB values range as low as 100 pM (20). OH-PCBs derive metabolically from their parent PCBs, which are highly stable, lipophilic, environmental toxins that bioaccumulate (21). PCBs have entered ecosystems and food chains on a global scale where they disrupt the endocrine signaling systems of numerous species (22)(23)(24), including humans (25,26). OH-PCB half-lives in human serum range from 2.6 to 15 years (27), and PCBs will persist in our environment for centuries (27). Approximately 2 billion kg of PCBs were produced between early 1920 and late 1970 (28), when they were banned (29). Remediation is ongoing at 16 PCBcontamination Superfund sites in the United States (30).
To better understand the molecular basis of OH-PCB action, and with the intent to use the findings as a template for the design and synthesis of potent SULT1E1-specfic inhibitors, we determined the solution structure of SULT1E1 bound to PAP, estradiol (E2), and OH-PCB1 (Fig. 1). Structure and binding studies reveal that OH-PCB1 binds to and reshapes the SULT1E1 active site into a high-affinity E2-binding pocket in which E2 interacts directly with OH-PCB1 and cannot access catalytic residues. Finally, structurally guided SULT1E1 mutagenesis is used to demonstrate that estrogen receptor (ER) activation in cultured human cells is regulated by OH-PCB binding to SULT1E1.

OH-PCB selection
The OH-PCBs used in our studies are shown in Figure 1. OH-PCB1 was selected because its exchange rate is well suited to the NMR line-broadening methods used in the structural studies described below. Notably, OH-PCB1 is present at high levels in the serum of individuals living in Japan (21). OH-PCB2 was chosen because it is among the highest affinity SULT1E1 inhibitors known (K i 400 PM (20)) and it derives from one of the most prevalent PCB contaminants in the United States (31,32), PCB77 (33).

Structure determination strategy
The effects of an unpaired spin-label electron on the solution NMR spectrum of a bound ligand in rapid exchange (34)(35)(36) with bulk solvent are well established. Electron/nuclear spin-spin interactions broaden NMR linewidths in a distance-dependent fashion, which is given by the Solomon-Bloembergen equation (37,38). When distances are measured from three well-spaced spin labels located at defined positions on the protein scaffold, each ligand proton can be triangulated and thus positioned on the scaffold. The structure of the enzyme,ligand complex is then refined using NMRdistance-constrained molecular dynamics (MD) docking and confirmed via mutagenesis.

Spin label attachment sites
Spin label attachment sites were selected from SULT1E1 backbone regions predicted to be stable by MD evaluation of the fully equilibrated E,PAPS,E2 scaffold. Surface residues that are solvent exposed and well isolated from the catalytic machinery were selected from the stable regions using the criterion that they be separated to maximize the spin-label paramagnetic field coverage of the protein surface. Finally, all insertion points satisfied the criteria that their C α RMSF be ≤1.0 Å and that the RMSF of the modeled spin-label nitroxyloxygen be 6.0 Å, suggesting unfettered motion.
Spin labels were attached to the SULT1E1 scaffold at Cys residues that were site-specifically incorporated via PCRbased mutagenesis. Prior to creating spin-label attachment constructs, reactive Cys was removed from the native scaffold. Only one of the four native SULT1E1 Cys (i.e., C69) reacts with DTNB (39), and mutagenic conversion of C69 to ser produced a stable, fully active non-DTNB-reactive scaffold. Five individual cys-attachment constructs were prepared from the C69S mutant (G16C, K25C, N150C, Q163C, and N233C). Mutants were labeled (see, Experimental procedures) and the initial rate parameters (k cat , K m and K i ) of each spin-labeled mutant were determined (see, Experimental procedures, Initial-Rate studies) to evaluate whether the catalytic integrity of the constructs had been compromised by the insertions and/or spin label attachments. The results, compiled in Table 1, reveal that the labeled mutant parameters are nearly identical to those of wild-type (WT) SULT1E1.
The spin labels used in the current study are presented in Figure 2 (white carbon atoms). Interaction between a boundligand proton and an unpaired electron can be detected when the interspin distance is ≤ 25 Å. The large semitransparent spheres seen in Figure 1 are centered on the spinlabel nitroxyl-moiety oxygen atom and their radii (25 Å) correspond to the approximate, maximum detectable interspin distance. As is evident, the distribution "coats" nearly the entire surface of the protein with a paramagnetic field of sufficient strength to broaden ligand 1 H-NMR peaks, thus allowing distances to be determined, regardless of where the ligand binds. Figure 2 shows spin labels at the five attachment sites used in the initial screen. The three spin labels used in the distance studies are labeled according to their residue positions.

NMR distance measurements
The distance-dependent effects of protein-coupled spin labels on NMR linewidths of ligands in solution are well understood (40)(41)(42). The distance (r) between the unpaired electron and nucleus is given by the following equation (37,38): where R 2 is the transverse relaxation rate of the nuclear spin, μ o is the permeability of a vacuum, γ is the proton gyromagnetic ratio, g is the electron g-factor, μ B is the magnetic moment of the free electron, S is the electron spin quantum number, τ c is the rotational correlation time of the protein, and ω is the Larmor frequency of the proton.
Transverse relaxation rates are calculated from NMR linewidths, which, for protein-bound ligands, are typically too broad to determine accurately. If the ligand exchanges between the protein and solution at a rate comparable with or greater than the difference in Larmor frequency between the bound and free species, observed R 2 values (R 2 obs ) for the bound species can be obtained from the slopes of Solution-Phase-Linewidth versus Fraction-Ligand-Bound plots (i.e., LWversus-FB plots) (43,44). The LW-versus-FB plot for the H3  H5 NMR peak of OH-PCB1 is presented in Figure 3A. The OH-PBC1, PAP, and E2 concentrations used are reported in the Figure 3 legend. The full OH-PCB1 1 H-NMR spectrum and the line-broadening effect of spin label 233 on the H3 H5 peak width as a function of percent of bound ligand are given in Figure 3, B and C, respectively. R 2 obs values contain contributions from relaxation caused by the unpaired electron (the paramagnetic contribution) and the protein (the diamagnetic contribution). Interspin distance calculations (Equation 1) depend only on the paramagnetic contribution (R 2 ), which is obtained by subtracting the diamagnetic contribution from R 2 obs . The diamagnetic contribution is given by the slopes of LWversus-FB plots constructed using control constructs in which the spin-label PROXYL-moiety (2,2,5,5-tetramethyl-1-pyrrolidinyloxy) is replaced by the cyclohexyl-group (42) (see, Experimental procedures). OH-PCB1 exhibits four 1 H-NMR resonances, three of which correspond to pairs of chemically indistinguishable protons (H3/H5, H2'/H6', and H3'/H5'). The 12 interspin distances needed to triangulate the protons associated with the four resonances were obtained from LW-versus-FB plots (see, Fig. 3A and Fig. S1) and are given in Table 2.
Refining the structure NMR triangulation distances were used in conjunction with MD docking simulations (45,46) to obtain the structure of the E,PAPS,E2,OH-PCB1 complex. During the simulations, each proton is constrained by a restoring force (50 kJ mol −1 Å −1 ) to move within an ellipsoid whose centerpoint is given by the intersection of the three NMR distance vectors that position it and whose principal axes lengths correspond to the standard errors (±1 σ) of the distance measurements. The restoring force (applied using distance_restraints in GROMACS) drives a proton toward the center of its ellipsoid if it lies outside of the ellipsoid surface (45,46). Given that OH-PCB1 contains three proton pairs that are indistinguishable by 1 H-NMR, distance constraints were applied not to each proton in a pair, but to the midpoint of the chord that connects the pair. As is appropriate for NMR distance measurements (37,38,47), distance _restraints was parameterized to use time-averaged, (1/r 6 )weighted distance restrains, which were applied simultaneously to the 12 distances that constrain the four OH-PCB1 positions. Docking simulations were run long enough to allow structures to achieve equilibrium. The results of ten simulations were analyzed using g_cluster in GROMACS. Only a single structural cluster (≤2.0 Å RMSD) was detected. The resulting structures were virtually identical and are overlain in Figure 4. The structures can be downloaded at ModelArchive (48) (accession # ma-xnec7). No significant structural changes were detected over 10 ns once the distance constraints were removed.

The structure
The structure of OH-PCB1 bound to its binding site in the E,PAPS,E2,OH-PCB1 complex is presented in Figure 5A. OH-PCB1 is situated in a highly hydrophobic pocket and is in direct contact with the three residues highlighted in blue (P75, P80, and Y239). The OH-PCB1 benzyl rings are sandwiched on one side by the rings of P75 and P80 and on the other by those of E2. The ring plane of Y239 contacts edges of both OH-PCB1 and E2. Remarkably, the binding of OH-PCB1 establishes a new E2-binding site in which it prevents catalysis by intercalating between E2 and the catalytic machinery. The structure reveals how an inhibitor (OH-PCB1) that sterically prevents a substrate (E2) from binding its active-site pocket can be transformed from a competitor to an allostere whose effects cannot be diminished by increasing substrate concentrations.
In moving from Panel A to B, one sees the effect of withdrawing OH-PCB1-the olive residues cluster into direct- Figure 2. The spin-labeled SULT1E1 constructs. SULT1E1 subunits are shown in blue and red, and ligands, PAP, E2, and OH-PCB1, are colored teal, yellow, and red, respectively. Spin labels (white) are shown attached at five positions chosen to completely "coat" the dimer in a paramagnetic field of sufficient strength to detect its effects on the solution NMR spectrum of ligands without compromising the catalytic integrity of the enzyme. The experimental constructs incorporate one spin label per subunit. The three spin labels selected for structural studies are labeled according to their attachment-site sequence position. Semitransparent spheres are centered on spin label nitroxyl-oxygen atoms and their radii are set at 25 Å-the approximate maximum distance over which ligand/spin label interactions can be detected. The indicated 90 rotation transforms the left-into the right-hand structure.
contact positions with E2, which rotates longitudinally about the axis connecting its 3-and 17-OH moieties (causing its C18-methyl group to face the viewer) and vertically, which enhances 3-OH reactivity via H-bonding with His107 (49) and reactive proximity (3.4 Å) to the sulfuryl-moiety. The structure seen in Panel B was MD-generated as previously described (50) and is virtually identical to the E,PAP,E2 structure from which it was derived (49).

Validating the structure
In the absence of inhibitor, the olive residues are in direct contact with E2, and the blue residues lie slightly beyond E2contact distance, suggesting that their contribution to substrate binding and catalysis might be slight. If the OH-PCB1binding and catalytic functions of the site are separable, the effects of mutating the OH-PCB1 binding site on the OH-PCB1 inhibition constant (K i OH-PCB1 ) can be used to validate the structure. To assess whether mutagenizing OH-PCB1 binding-site residues influences the catalytic functions of SULT1E1, the effects of such mutations on the initial-rate parameters (K m and k cat ) of 1-hydroxypyrene (1-HP, a fluorescent E2-like substrate (51, 52)) were determined (see, Experimental procedures). OH-PCB1 direct-contact residues (F75, F80, and Y239) were mutated to alanine (43,44) individually and in combination (F75/F80), and the results of the initial-rate studies are given in Figure 6, A-C and Table 3. As is evident, the mutations have no measurable effect on catalytic function (i.e., K m 1-HP and k cat ). In contrast, the mutations had pronounced effects on K i OH-PCB1 . Mutating Y239, which contacts an edge of OH-PCB1, causes a 5.7-fold decrease in affinity, while mutating residues that sandwich the inhibitor   (3) 25 (4) 22 (3)  2 6 22 (3) 24 (3) 16 (3)  3 5 28 (4) 25 (4) 14 (2) a Values in parentheses indicate 95% confidence interval. (F75 and F80) result in larger decreases (11-and 12-fold, respectively). The effect of the F75/F80 double mutation, 113fold, is within error equal to the product of the individual mutations, suggesting that these residues operate largely independently. These findings fully support the structure seen in Figure 5A, and confirm that the OH-PCB1-binding and catalytic functions of SULT1E1 are indeed separable.

The energetics of OH-PCB/reactant interactions
To better understand the inhibition mechanism, interactions between OH-PCBs and reactants were probed using equilibrium binding studies. OH-PCB-binding titrations were performed at saturating reactant concentrations and monitored via ligand-binding induced changes in SULT1E1 intrinsic fluorescence (see, Experimental procedures). OH-PCB1 titrations are presented in Figure 7, A and B, and the associated dissociation constants are given in Table 4. Consistent with the structure, OH-PCB1 and PAP do not detectibly interact-OH-PCB1 affinities for E and E,PAP are essentially identical (750 ± 27 and 780 ± 40 nM, respectively). In contrast, and as is supported by their contact in the structure, OH-PCB1 and E2 influence one another's affinities-OH-PCB1 binds 12-fold more tightly to E,E2 than to E (62 ± 2 versus 750 ± 27 nM). Notably, PAP binding does not alter the energetics of OH-PCB1/E2 interactions-OH-PCB1 affinities for E,E2 and E,E2,PAP are experimentally indistinguishable (62 ± 2 and 66 ± 3 nM).  . Rate measurements were taken during conversion to product of ≤5% of the concentration limiting substrate consumed at the reaction endpoint. Rates were normalized to the rate in the absence of inhibitor (i.e., V/V max ). Each point is the average of three independent measurements and the associated standard deviations are smaller than the dot diameters. Averaged data were least-squares fit to a noncompetitive inhibition model (see, Experimental procedures) and the lines passing through the data represent the behavior predicted by the best-fit constants reported in Table 3.

OH-PCB inhibition of human sulfotransferase 1E1
The energetics studies are consistent with the structure and reveal an OH-PCB1/E2 interaction energy that enhances the affinities of both ligands 12-fold. E2 affinities for E,PAP and E,PAP,OH-PCB1 were determined (13 ± 1 and 0.98 ± 0.10 nM), respectively (Fig. 7, C and D, and Table 4) and reveal that E2 binds more tightly as an inhibitor than as a substrate (K m E2 = 5 nM (4)).
To assess whether OH-PCBs can add directly to the E,E2 complex, and vice versa, the order of binding OH-PCB and E2 was determined. To do so, the affinity of each ligand was assessed at two saturating and tenfold different concentrations of the partner ligand. If binding is ordered, the ligand's affinity will appear to increase tenfold as the partner draws it onto the enzyme. If instead, binding is random, the ligand affinity will not vary with the change in partner concentration. The affinity of OH-PCB1 did not vary at 32-and 320 × K d concentrations of E2 (0.40 and 4.0 μM, respectively); similarly, E2 affinity did not vary at 100-and 1000 × K d concentrations OH-PCB2 (0.70 and 7.0 μM, respectively)-see, Figure 8, A-C and Table 4. Hence, binding is random and the enzyme can bind either ligand and rearrange from the reactive to inhibited configuration without first dissociating its partner. Notably, OH-PCB2 was used in these studies because, unlike OH-PCB1, its solubility does not preclude the high [OH-PCB]/K d ratios they require.

PCB-based regulation of the estrogen receptor
The theory that SULT1E1 inhibition plays a meaningful role in the OH-PCB-based disruption of endocrine function has not been tested directly. The ability to weaken OH-PCB affinity for SULT1E1 without influencing its catalytic properties provides an opportunity to substantiate this theory using the estrogen-response systems in human cells. Toward this end, stable transfectants that express WT or double mutant (F75/F80) SULT1E1 were constructed (see, Experimental procedures) from Ishikawa cells-an immortalized endometrial adenocarcinoma cell line (53) with undetectably low intrinsic levels of SULT1E1 expression (53). To ensure that the F75/F80 double mutation does not affect the E2 initial-rate parameters (as is the case with 1-HP), the WT and mutant E2 parameters (K m and k cat ) were determined and proved to be identical within error (see , Table 5 and Fig. 9). A (-) SULT1E1 control strain was

OH-PCB inhibition of human sulfotransferase 1E1
constructed using the transfection vector without a SULT coding-region insert. Transfectants were selected for ER activation studies based on SULT1E1 activity levels in cell extracts (see, Experimental procedures). Extract activities in the WT and mutant (MT) transfectants selected for further study (4.8 and 5.0 pmol min −1 mg −1 extract, respectively) were comparable with those reported for human mammary epithelial cells (4.8 pmol min −1 mg −1 (54)) and 52-fold higher than the (-) SULT1E1 control strain extracts.
The E2 and OH-PCB concentration dependence of transfectant ER activation was measured via alkaline phosphatase activity and the resulting titrations are presented in Figure 10A. Endogenous alkaline phosphatase levels are coordinated with ER activation and measured via paranitrophenyl phosphate hydrolysis (55). Consistent with SULT1E1 inactivation of E2, the ER-activation EC 50 of E2 is 80-fold higher in the WT (red dots) and MT (blue dots) strains, which express SULT1E1, than in the control strain (black dots), which does not (see Table 6). The arrow seen in Figure 10A pinpoints the ER response levels at the fixed E2 concentration (10 μM) used in the OH-PCB-titrations shown in Panels B and C. The OH-PCB1 and OH-PCB2 ER-activation patterns are similar across the three strains. OH-PCBs do not affect activation in the control strain; hence, any effects on activation in the WT and MT strains are likely linked to SULT1E1 expression. OH-PCB effects on the WT strain are pronounced-activation begins at background and increases to a maximum comparable with that associated with the control strain. OH-PCB1 and OH-PCB2 EC 50 values (38 and 0.87 nM, respectively) are similar to their K i values (60 and 0.50 nM), suggesting that the plots are reporting increases in E2 activity as SULT1E1 is inhibited. To establish that the OH-PCB effects are due to binding at the OH-PCB-binding site seen in Figure 2A, activation in the MT strain was tested. As is evident, activation in the MT strain remains near background throughout the entire OH-PCB concentration range with the exception of slight elevation at the highest OH-PCB concentrations, which may be due to relatively weak inhibition of the mutant and/or low-affinity OH-PCB activation of the receptor (56,57).

Conclusions
A set of five SULT1E1 constructs that permit disulfidebased attachment of R-groups at catalytically innocuous sites distributed roughly uniformly across the protein's surface has been established. Attachment of spin labels at these sites allows the entire surface of the protein to be "coated" in a paramagnetic field of sufficient strength to detect its effects on the NMR spectrum of ligands in rapid exchange, regardless of where they bind. Three spin-labeled constructs were used to determine the structure of OH-PBC1 bound to SULT1E1 in a complex with E2 and PAP. The structure, which was confirmed via mutagenesis, reveals that OH-PCB1 binds at the active site and forms a new E2-binding pocket in which it is sandwiched between E2 and the protein, and E2 is stabilized in a nonreactive position.
Ligand interactions and binding order were assessed in equilibrium-binding studies. Consistent with the structure,

OH-PCB inhibition of human sulfotransferase 1E1
OH-PCB1 shows no net energetic interaction with PAP, but interacts significantly with E2. OH-PCB1 and E2 bind 12fold more tightly in one another's presence, leading to the conclusion that E2 binds more tightly as an inhibitor than as a substrate. Ligand binding is random, which reveals that SULT1E1 can bind either ligand and restructure from its reactive to its inactive configuration without dissociating the partner ligand. The structure predicted that residues in direct contact with OH-PCB1 could be mutagenized to weaken OH-PCB1 binding without compromising catalysis. These predictions proved to be accurate and led to a catalytically normal, OH-PCB1resistant double mutant that was used to demonstrate that either OH-PCB1 or OH-PCB2 can regulate estrogen receptor activation in a cultured human cell line by binding to SULT1E1. To our knowledge, this is the first direct demonstration of OH-PCB-based activation of the ER, and it supports the contention that OH-PCB inhibition of SULT1E1 contributes meaningfully to PCB-based endocrine disruption.
Computer and software MD simulations were performed on a Parallel Quantum Solutions QS32-2670C-XS8 computer. PQS Molecular Builder software was purchased from Parallel Quantum Solutions (59). Source code for GROningen MAchine for Chemical Simulation (GROMACS) 4.5 was downloaded from http://www. GROMACS.org under the GROMACS General Public License (GPL) (45,46,60). Automated Topology Builder (ATB) is maintained by the National Computational Infrastructure (NCI) at Australia National University and is freely available at https://atb.uq.edu.au/ (61,62). A Genetically Optimized Ligand Docking (GOLD) license was obtained from the Cambridge Crystallographic Data Center (63).

SULT1E1 plasmids
The SULT1E1 expression plasmid consists of an E. coli codon-optimized SULT1E1 coding region inserted into the PreScission Protease cleavage site of a triple-tagged pGEX-6P expression vector containing an (N-terminal)-His/GST/MBP tag (43,64). The cys-insertion mutants used for regiospecific attachment of maleimide-based labels were constructed as follows: cys69 (the only DTNB reactive cysteine) was replaced with ser, and three single-cys mutants were then created by inserting cys into the nonreactive (C69S) scaffold at G16, N150, and N233. The mutations used to test the SULT1E1 structural model (Y239A, F75A, F80A, and F75A/F80A double mutants) were inserted into the WT coding region. All mutations were generated using site-directed PCR mutagenesis (43,65).

Covalent labeling
Labels (N-cyclohexylmale-imide or 3-maleimido-PROXYL) were added in 20-fold excess over reactive cysteine to a solution containing SULT1E1 construct (50 μM active sites), 0.50 mM PAP, and 50 mM KPO 4 (pH 7.4), 25 C ± 2 deg. C. PAP was added to enhance enzyme stability. The reactions were monitored by using DTNB to measure unreacted cysteine and were considered complete when >98% of the cysteine had reacted (3 h).

Equilibrium binding studies
The binding of inhibitors to WT and mutant SULT1E1 was monitored via ligand-induced changes in the intrinsic fluo- Initial-rate studies

NMR-distance-restricted molecular dynamics modeling
As described previously (34)(35)(36), a ligand-free model of SULT1E1 was constructed from the SULT1E1,PAPS (PDB 1HY3 (71)) structure using SWISS-MODEL. The model was protonated (pH 7.4) and energy minimized using GROMACS. GROMAS57 energy-parameter files were created using Automated Topology Builder (62) for OH-PCB1, PAPS, E2, and a spin-labeled cysteine analogue in which the nitroxylmoiety was replaced by a hydroxyl group. Spin-labeled cysteine analogues were added as noncanonical amino acids to the GROMAS57 energy field and used to create a triply spin-labeled model in which G16, N150, and N233 are replaced by the analogue. PAPS and E2 were positioned in the active site of the spin-label model using GOLD (72,73) and the system was equilibrated (298 K, NaCl (50 mM), pH 7.4) in 100 psec increments using GROMACS. Once equilibrated, OH-PCB1 was randomly positioned in a simulated cube of water (52 × 52 × 52 Å) containing the spin-labeled construct and docked using GROMACS (73). OH-PCB1 docking was constrained using the NMR-determined spin-label/PCB1proton distances as described in Results and discussion, Refining the structure.
Transfection protocol pcDNA 3.1 constructs harboring either the SULT1E1 WT or double mutant (F75A/F80A) coding regions were generated using Gibson Assembly (74). The vectors were linearized using BGIII. Ishikawa cells were grown at 37 C ± 2 deg. C in growth medium (MEM media containing 10% v/v FBS). At confluence, the cells were washed (3×) with PBS media before coating with an Opti-MEM solution containing a linearized construct (50 ng μg −1 ) and lipofectamine (2.5 units ml −1 ). After 24 h at 37 C ± 2 deg. C, cells were washed (3×) with PBS before adding growth medium containing neomycin (400 μg ml −1 ) to select transfectants. Selective growth medium was replenished every 48 h until single colonies could be isolated (3--6 weeks). Single colonies were transferred using trypsin digest (0.05%) to 12-well plates and grown at 37 C ± 2 deg. C to confluence for further experimentation and storage.

Transfectant SULT1E1 levels
Transfectants were grown at 37 C ± 2 deg. C to 60 to 70% confluency in 12-well plates, washed (3×) with PBS (25 C), and lysed using RIPA buffer (0.50 ml) (75). Lysate was centrifuged (15,000g for 10 min, 25 C) and the supernatant was collected, assayed, flash-frozen, with liquid nitrogen, and stored at −80 C. SULT1E1 activity was stable following freeze/thaw. Extract protein concentrations were determined using the Bradford assay (76) and SULT levels were determined by measuring turnover at saturating 1-HP and PAPS. Assay conditions were identical to those described in Initial-Rate Studies except that extract (1-3 μg) was added in lieu of pure enzyme.

Estrogen receptor activation studies
Transfected Ishikawa cells were grown at 37 C ± 2 deg. C to 60 to 70% confluency in MEM containing 10% v/v FBS. Cells were then transferred to 96-well tissue culture plates and grown at 37 C ± 2 deg. C to 80% confluency in MEM containing 10% v/v charcoal filtered FBS. The plate was then washed with PBS and grown at 37 C ± 2 deg. C for 24 h in MEM without FBS. Estradiol (0-3.0 μM) and OH-PCB (0-1.0 μM) solubilized in neat DMSO were added and cells were incubated at 37 C ± 2 deg. C for 5 days. The final concentrations of DMSO were ≤0.10%. Cells were then washed twice with PBS before adding pNpp (5.0 mM) in PBS (75). Following incubation with pNpp for 3 h at 25 C ± 2 deg. C, wells were read at 405 nm using a Synergy HT BioTek Plate Reader.

Data availability
All data and materials are available upon request at tom. leyh@einsteinmed.org. All SULT1E1,E2, PAPS,OH-PCB1 spin-label directed docking models and the NIH SAVE v5.0 structure validation report are available for download at www. model.archive.org (accession no. ma-xnec7).