Molecular basis and dual ligand regulation of tetrameric estrogen receptor α/14-3-3ζ protein complex

Therapeutic strategies targeting nuclear receptors (NRs) beyond their endogenous ligand binding pocket have gained significant scientific interest driven by a need to circumvent problems associated with drug resistance and pharmacological profile. The hub protein 14-3-3 is an endogenous regulator of various NRs, providing a novel entry point for small molecule modulation of NR activity. Exemplified, 14-3-3 binding to the C-terminal F-domain of the estrogen receptor alpha (ERα), and small molecule stabilization of the ERα/14-3-3ζ protein complex by the natural product Fusicoccin A (FC-A), was demonstrated to downregulate ERα-mediated breast cancer proliferation. This presents a novel drug discovery approach to target ERα; however, structural and mechanistic insights into ERα/14-3-3 complex formation are lacking. Here, we provide an in-depth molecular understanding of the ERα/14-3-3ζ complex by isolating 14-3-3ζ in complex with an ERα protein construct comprising its ligand-binding domain (LBD) and phosphorylated F-domain. Bacterial co-expression and co-purification of the ERα/14-3-3ζ complex, followed by extensive biophysical and structural characterization, revealed a tetrameric complex between the ERα homodimer and the 14-3-3ζ homodimer. 14-3-3ζ binding to ERα, and ERα/14-3-3ζ complex stabilization by FC-A, appeared to be orthogonal to ERα endogenous agonist (E2) binding, E2-induced conformational changes, and cofactor recruitment. Similarly, the ERα antagonist 4-hydroxytamoxifen inhibited cofactor recruitment to the ERα LBD while ERα was bound to 14-3-3ζ. Furthermore, stabilization of the ERα/14-3-3ζ protein complex by FC-A was not influenced by the disease-associated and 4-hydroxytamoxifen resistant ERα-Y537S mutant. Together, these molecular and mechanistic insights provide direction for targeting ERα via the ERα/14-3-3 complex as an alternative drug discovery approach.

Initial studies have shown that NR/14-3-3 PPIs can be modulated using small molecules. Specifically, the natural product FC-A or covalently tethered small molecules have been shown to stabilize the ERα/14-3-3 and ERRγ/14-3-3 PPIs (Figs. 1C and S3) (38,(44)(45)(46). These small molecules, also called molecular glues, increase the binding affinity between the two protein partners (PPI stabilization) by binding in a composite pocket formed at the interface of 14-3-3 and the phosphorylated NR protein. Further, stabilization of the ERα/ 14-3-3 PPI by FC-A suppresses ERα chromatin binding and subsequent transcriptional activity (38). These results illustrate the potential of the stabilization of the ERα/14-3-3 interaction as an alternative drug discovery entry.
Despite the potential of therapeutic targeting of NR/14-3-3 complexes, little is known about NR/14-3-3 interactions on a molecular level. To date, biochemical and structural studies of NR/14-3-3 complexes have been exclusively performed using short phosphopeptide mimics of the NRs (Figs. 1C and S2) (45)(46)(47). Studies of NR/14-3-3 complexes using protein domains or full-length NRs, such as those performed for other 14-3-3 PPIs (48)(49)(50)(51), are needed to gain an enhanced molecular and structural understanding of NR/14-3-3 complex formation. As an example, the use of the entire NR LBD allows studies on the interplay between NR/14-3-3 complex formation and aspects like NR dimerization, ligand binding, and cofactor recruitment. Additionally, small molecule stabilization of full-length NR/14-3-3 complexes can be investigated in the context of clinically relevant NR point mutations which is of high interest (52)(53)(54)(55). Molecular insights into NR/14-3-3 PPIs beyond the phospho-binding groove would also provide the potential to identify novel composite binding pockets for small molecule targeting, thereby expanding the number of entry points for novel PPI modulator design and increasing the potential for selectivity (56).
Within this work, we aim to enhance our understanding of 14-3-3 binding to ERα via in vitro characterization of the protein complex formed by 14-3-3ζ and an ERα construct comprising the LBD and phosphorylated F domain. To gain a robust understanding of protein complex formation and the stoichiometry of binding, several biophysical assays were performed, including analytical size exclusion chromatography (SEC) and analytical ultracentrifugation (AUC). In addition, differential scanning fluorimetry (DSF), fluorescence anisotropy (FA), and hydrogen-deuterium exchange (HDX) experiments determined the role of ERα ligands E2 and 4-OHT on ERα/14-3-3 complex formation and identified 14-3-3 binding to the drug-resistant ERα-Y537S mutant. Finally, we show that the ERα/14-3-3 PPI can be stabilized by the natural product AEGFPA pT V-COOH NTD DBD LBD . The binding of its endogenous ligand E2 (blue sticks) induces the folding of helix 12 (h12) in an active conformation, allowing cofactor binding (yellow cartoon). In contrast, antagonist 4-hydroxytamoxifen (4-OHT, yellow sticks) inhibits this conformational change and cofactor recruitment, PDB: 5WGD and 3ERT. C, crystal structure of 14-3-3 dimer (gray/white surface) and co-crystal structure of 14-3-3σ (white surface) with the ERα derived Cterminal phosphopeptide (green sticks) and small molecule stabilizer Fusicoccin-A (FC-A, pink sticks), PDB: 4JDD. The 14-3-3/ERa-LBD-F-domain complex FC-A and that this PPI stabilization can be achieved independently from ERα ligand binding in both wild-type ERα and the ERα-Y537S mutant, thus presenting a potential orthogonal therapeutic strategy for targeting endocrine resistance in breast cancer.
Co-crystal structures of 14-3-3 bound to ERα-phosphopeptides revealed why the strep-tagged ERα peptide showed a reduced FC-A responsiveness (Fig. S10). The C-terminal streptag occupied the FC-A binding site within the 14-3-3 binding groove, forcing a structural rearrangement of the ERα peptide upon FC-A binding. This makes stabilization by FC-A less favorable compared to the WT ERα peptide. While the ERα(PKA)-strep construct could be used as a representative protein for ERα itself, the protein is less favorable to study 14-3-3 binding and small molecule PPI stabilization. As such, we shifted our focus to an alternative recombinant protein expression approach to obtain the ERα/14-3-3 protein complex.

Co-expression and co-purification of ERα/14-3-3ζ complex
To circumvent proteolytic cleavage of ERα or the use of Cterminal purification tags, we used bacterial co-expression of ERα, 14-3-3ζ, and PKA to obtain the ERα/14-3-3ζ protein complex. Specifically, N-terminally His-SUMO-tagged ERα(PKA) was expressed together with PKA and strep-tagged 14-3-3ζ in E. coli ( Fig. 2A). Co-expression of 14-3-3ζ and ERα with PKA allowed 14-3-3 to bind in situ to the phosphorylated ERα protein shielding the disordered F domain from proteolytic degradation. A similar co-expression approach was previously applied to successfully identify 14-3-3 binding to Ataxin-1 and to obtain a purified Tau/14-3-3 complex (64,65). Notably, the 14-3-3 protein comprises seven human isoforms which have proven to feature highly similar biochemical and structural features (66,67). Yeast two-hybrid studies have previously shown that ERα is able to interact with all seven isoforms of 14-3-3 (38). Therefore, we have selected here one 14-3-3 isoform, zeta, as a representative for the class of 14-3-3 proteins. 14-3-3ζ was specifically selected as it expresses well in E. Coli, because of its relatively high abundance in the human body (68), and because it has been successfully used in similar biochemical studies of larger 14-3-3 protein complexes, such as the BRAF/ 14-3-3 complex (50,51,69).
The ERα/14-3-3 protein complex was co-purified using three subsequent chromatography methods. First, a Ni-NTA column was used to select for His-tagged ERα protein (Fig. 2B). Elution fractions contained both His-tagged ERα and 14-3-3ζ protein, indicating strong binding of 14-3-3ζ to ERα, as 14-3-3 did not contain a His-tag itself. After hydrolysis of the N-terminal His-SUMO tag of ERα, the complex was purified with a streptavidin-column, which again showed coelution of 14-3-3ζ and ERα (Fig. 2C). The ERα/14-3-3ζ protein complex was finally purified by SEC which resulted in the elution of a uniform peak that contained both proteins (Fig. S11). High-resolution mass spectrometry (LC-QToF-MS) analysis showed two protein peaks of which the first corresponded to the mass of 14-3-3ζ and the second to that of the phosphorylated ERα LDB-F protein (Fig. 2D). Notably, truncation of ERα was not observed after co-expression and -purification with 14-3-3ζ. Quantitative MS experiments showed the approximately equimolar presence of both 14-3-3ζ and ERα within the purified protein complex sample, indicating a 1:1 binding of the two proteins (Figs. S12 and S13). Furthermore, ERα binding to 14-3-3 proved to be phosphorylationdependent as co-expression in the absence of PKA did not result in co-elution of 14-3-3ζ with His-tagged ERα (Fig. 2E).

ERα/14-3-3 bind in a 2:2 stoichiometry
Native PAGE analysis of the purified ERα/14-3-3ζ complex showed a single band protein complex indicating stable complex formation between ERα and 14-3-3ζ without the presence of any major excess of either protein (Fig. S14). The addition of a high-affinity competitive 14-3-3 binder (70) disrupted the ERα/14-3-3ζ complex as apparent by two distinct bands that ran in the native PAGE gel at the same heights as 14-3-3ζ and ERα individually, testifying to the stable complex formation between ERα and 14-3-3ζ.
Analytical SEC and sedimentation velocity AUC (SV-AUC) were used as orthogonal approaches to determine the ERα/14-3-3ζ protein complex size and the stoichiometry of ERα and 14-3-3ζ binding (Fig. 3, A and B). Analytical SEC results showed a single peak for the individual 14-3-3ζ (theoretical M w 29 kDa) and ERα (theoretical M w 33 kDa) proteins which eluted similar to monomeric BSA (theoretical M w 66 kDa), indicating that both proteins themselves were present as stable homodimers in solution (60 kDa), which is in line with reported observations (Fig. 3A) (59,67,71,72). The ERα/14-3-3ζ complex eluted as a single peak with a clear shift in molecular weight compared to ERα and 14-3-3ζ individually, approaching the dimeric BSA peak around 132 kDa. Together with the quantitative MS studies, which showed the equimolar presence of both ERα and 14-3-3ζ in the purified mixture (Figs. S12 and S13), these results indicated the tetrameric complex formation of an ERα and a 14-3-3ζ homodimer.
SV-AUC sedimentation coefficient distributions c(s) validated results observed with SEC. ERα and 14-3-3ζ showed individual peaks with weight-averaged sedimentation coefficients (corrected to 20.0 C and to the density of water), s w(20,w) , of 4.0 S and 3.8 S, corresponding to M w 55 kDa and 58 kDa, respectively (theoretical M w of dimeric ERα is 66 kDa; and of dimeric 14-3-3ζ is 58 kDa) (Fig. 3B). The ERα/ 14-3-3ζ complex showed a main peak with a weight-averaged sedimentation coefficient of 6.3 S, corresponding to a M w 124 kDa, thus supporting the 2:2 stoichiometry of the ERα:14-3-3ζ complex (72,73). Notably, a minor peak was observed at peak positions of the 14-3-3ζ and ERα indicating traces of the non-interacting proteins present in the solution under these conditions.

Dimer-to-dimer binding enhances the ERα/14-3-3 protein complex affinity
Multivalent dimer-to-dimer complex formation was further accessed using a competitive SV-AUC experiment where 14-3-3ζ was added to ERα/14-3-3ζ protein complex to obtain various molar ratios of ERα and 14-3-3ζ. The addition of up to six equivalents of 14-3-3ζ to ERα resulted only in species representing the ERα/14-3-3ζ tetramer and the 14-3-3ζ dimer but did not show any 2:1 14-3-3ζ:ERα complex formation (Fig. 3C). This result showed that ERα binds to 14-3-3ζ as a stable dimer. The dimeric state of ERα is hypothesized to cooperatively enhance ERα binding affinity to 14-3-3ζ, similar to earlier described multivalent 14-3-3 binders CFTR and LRRK2 (74,75). The binding of one ERα monomer to the 14-3-3 dimer brings the second ERα monomer in proximity to 14-3-3, thereby increasing the effective molar concentration of ERα to 14-3-3ζ. Furthermore, dissociation of the protein complex is hypothesized to be slower since two binding interfaces need to dissociate in close succession for 14-3-3ζ and ERα to dissociate. Therefore, dimer-to-dimer binding of ERα and 14-3-3ζ is expected to increase the affinity of ERα for 14-3-3ζ and the stability of the tetrameric protein complex. The dissociation constant K D of dimeric 14-3-3ζ and dimeric ERα was estimated using SV-AUC results of the ERα/14-3-3 complex (Fig. S15). Based on the area under the curve, the amount of ERα and 14-3-3ζ in complex and alone was quantified from which the K D was calculated using the steady state equilibrium binding equation ). This calculation provided a K D 32 ± 6 nM. This affinity is almost 10fold higher than that of the ERα phosphopeptide binding to 14-3-3, indicating stronger binding of the phosphorylated ERα LBD-F domain protein due to the dimer-to-dimer binding mechanism.
deg. C and 59.0 C ± 0.2 deg. C, respectively (Fig. 3, E and F). ERα phosphorylation at T594 did not influence its melting temperature (Fig. S16). The ERα/14-3-3ζ protein complex showed two distinct melting peaks (Fig. 3E), with 2 C increased thermal stability when compared to the individual protein partners (Fig. 3F). A similar increase in thermal stability for 14-3-3ζ was observed upon the addition of an ERα phosphopeptide (Fig. S16). These results thus show a mutual stabilizing effect of 14-3-3ζ and ERα upon tetramer formation.

Mapping of interactions between ERα and 14-3-3ζ using HDX-MS
Hydrogen-deuterium exchange (HDX-MS) assays were performed on the individual proteins 14-3-3ζ and ERα, and the ERα/14-3-3ζ protein complex, to identify which regions within the ERα and 14-3-3ζ protein were involved in protein complex formation (Figs. 4 and S17-S20). The proteins were incubated in deuterium-containing buffers which were quenched after 20 s, 2 min, 20 min, and 2 h. Proteins were digested using pepsin after which the amount of deuteration of individual peptides was determined using mass spectrometry. HD exchange kinetics of ERα was followed for 248 peptides, covering 98.6% of the protein sequence, and 240 peptides for 14-3-3ζ which cover 100% of the sequence.
An HD exchange profile was made to visualize the deuteration kinetics of each ERα residue for apo-ERα and ERα within the ERα/14-3-3ζ complex (Figs. 4A and S17). Fast deuterium exchange kinetics (>30% deuteration after 20 s) were mainly observed in helix 1 to 2 (H1-2), the beta sheets (B), helix 7 (H7), residues between helix 9 and 10, helix 12 (H12), and the entire F-domain. Residues in H5-6, H8-9, H10, and H11, on the other hand, showed low amounts of deuterium exchange. These results corresponded nicely with the ERα crystal structures where regions with fast deuterium exchange kinetics are typically present in flexible and/or solvent-exposed regions of ERα. Notably, the C-terminal F-domain of ERα has never been crystallized before or studied with alternative structural biology techniques. Within this study, fast HD exchange kinetics were observed within the entire F-domain indicating that the ERα F-domain is most probably unstructured and solvent-exposed, as observed in the predicted Alphafold structure (76, 77).
14-3-3ζ binding to ERα decreased the deuteration kinetics of several regions within the ERα protein (Fig. 4, A-C; Figs. S17 and S18). Shielding effects were determined by calculating the difference in HDX between 14-3-3ζ-bound ERα and ERα by itself, after which this difference profile was visualized on the ERα Alphafold structure ( Fig. 4C and S17). Most pronounced shielding effects were observed in the N-terminal side of H3 (residues 328-354), the beta sheets through H7 (residues 397-420), the C-terminal side of H11 (residues 521-528), and the C-terminal end of the F-domain (residues 583-591) (Fig. 4, A-C; Figs. S17 and S18). Smaller effects were observed in H1-2 (residues 302-327), H8 (residues 421-444), H12, and the F- domain (residues 531-571). H4-6, H9-10, and the N-terminal part of H11 seemed unaffected, although it should be mentioned that these regions showed minor deuteration in the first place. Interestingly, although the 14-3-3 binding groove is known to primarily bind the C-terminal end of the ERα Fdomain, multiple regions additional to the F-domain seem to be affected by 14-3-3 binding. Shielding effects were observed for almost all flexible and solvent-exposed regions within ERα, indicating an overall stabilization of the ERα fold upon tetramer formation. Most pronounced effects were clustered on the 'bottom' of the ERα LBD structure (H7, H3 N-term, H11 C-term), indicating the proximity of 14-3-3ζ to this side of ERα, albeit with the dynamic movement of the ERα LBD dimer, facilitated by the long and flexible F-domain, leading to mild shielding effects at all sides of the ERα LDB. Notably, ERα/14-3-3ζ dimer-to-dimer binding might result in differential binding of one ERα monomer in comparison to the other ERα monomer, potentially further explaining why many regions within the ERα protein are mildly affected upon 14-3-3ζ binding. Deuteration kinetics for 14-3-3ζ in the absence and presence of ERα protein were similarly analyzed (Fig. 4, D-F, Figs. S19 and S20). The regions of high and low exchange rates of 14-3-3ζ alone corresponded well to known crystal structures and previously published HDX of 14-3-3 (78,79), with high deuteration in loops between H2-3, H3-4, H4-5, and H8-9, and none to minor amounts of deuteration in H2, H3, H4, H5, H7, and H9. ERα binding to 14-3-3ζ led to shielding effects on various parts of the 14-3-3ζ protein, which was most pronounced after 2 h of incubation (Fig. 4, D-F; Figs. S19 and S20). The largest shielding effects were observed within peptides in the C-terminal end of H3 (residues 50-68), H6-7 (residues 154-174), H8 (residues 180-199), and the C-terminus in H9 (residues 217-230). These shielded regions strongly correlated with the 14-3-3 binding groove (H3, H7 and H9) where the ERα C-terminus binds. Interestingly, also H8 on "top" and H6 on the "back" of the 14-3-3 protein were partially shielded, indicating that these 14-3-3 regions are potentially in close proximity to other parts of ERα. In contrast, the "base" of the 14-3-3 protein seems to be less affected by ERα binding, indicating the absence of direct contact with ERα (Fig. S19). Crystallography and HDX studies of other 14-3-3 PPIs typically show similar binding interfaces involving the 'top' of 14-3-3 (H8-9), whereas the "base" of 14-3-3 is not involved with the PPI formation (48, 69, 79-81).

14-3-3ζ can bind the ERα Y573S drug-resistant mutant
Point mutations in the ERα LBD are known to modulate ERα conformation and transcriptional activity. ERα Y537S is one of the most prevalent somatic mutations in patients with breast cancer, typically acquired after antiestrogen treatment (54). Structural and biophysical characterization has shown that the Y537S mutation places H12 in an agonistic, constitutively active, conformation (Fig. S21), causing this ERα mutant to be resistant to antiestrogen treatment (54,82). Therefore, it is of high interest to study 14-3-3ζ binding to ERα-Y537S as it may provide a new approach to modulate the transcriptional activity of the drug-resistant ERα-Y537S mutant. The ERα-Y537S/14-3-3ζ complex could successfully be co-expressed and co-purified using the aforementioned methodologies for WT ERα/14-3-3ζ complex purification, indicating that the Y537S point mutation did not impede 14-3-3 binding. SV-AUC confirmed this as similar sedimentation distributions were obtained for the ERα-Y537S/14-3-3ζ complex and the WT ERα/14-3-3ζ protein complex (Fig. S22A). Furthermore, DSF studies showed a 2 C enhancement of thermal stability of both ERα-Y537S and 14-3-3ζ upon complex formation, similar to WT ERα (Fig. S22, B-D). Interestingly, ERα-Y537S in the absence of 14-3-3 showed melting temperatures of 47.1 C ± 0.5 deg. C, which is 2 C higher than wildtype ERα, indicating higher thermal stability of ERα when mutated. All data together confirmed ERα-Y537S/14-3-3ζ complex formation, providing an interesting entry point of targeting this mutant via its interaction with 14-3-3.

ERα ligand binding is independent of ERα/14-3-3 complex formation
Small molecule ligands play an important role in the regulation of ERα transcriptional activity in both healthy and diseased state, making it highly valuable to study their effects on ERα/14-3-3 complex formation. Therefore, we set out to study the effect of ERα/14-3-3 complex formation on ligand binding to both WT and the Y537S mutant of ERα. Here, we specifically studied the endogenous ERα agonist E2 and therapeutic partial antagonist 4-OHT (Fig. 5A) (83,84). In both SV-AUC and native PAGE, E2 and 4-OHT did not show any effect on the ERα/14-3-3 protein complex size, apparent by the similar c(s) distribution of the ERα/14-3-3ζ complex in the presence and absence of ligands (Figs. 5B and S23). Furthermore, ligand-dependent DSF studies were performed to determine the effect of ERα ligand binding on the thermal stability of ERα in complex with 14-3-3ζ. (Fig. 5 Similar effects of E2 and 4-OHT were observed for experiments with the ERα-Y537S/14-3-3ζ complex (Fig. S25). Overall, the SV-AUC and DSF data indicated that ERα ligands did not disrupt ERα/14-3-3ζ complex formation and ERα was fully ligand responsive when bound to 14-3-3ζ.
Notably, the overall effects of E2 appeared to be smaller (max. shielding effect 30% instead of 40%) when ERα was bound to 14-3-3ζ which can be explained by the partial shielding effects that 14-3-3ζ already has on ERα, making the effect of E2 less pronounced. The affected regions of ERα upon E2 binding correlated well with published ERα-E2 crystal structures (Fig. 6, B and D), where E2 binds in a pocket formed by H3, H6, H8, the beta sheets, and H11 to 12 (84). Interestingly, these results indicated that the ERα F-domain conformation was not significantly affected by E2 binding, which has not been studied on a structural level previously. The similar shielding effects upon E2 binding to ERα alone and the ERα/ 14-3-3ζ complex, clearly showed that 14-3-3ζ does not influence the E2-induced conformational changes in ERα. This suggests that the ERα F-domain is sufficiently long and flexible to accommodate ligand-induced conformational changes to helix 12, without affecting 14-3-3 binding to ERα.
FA assays with the mutated ERα-Y537S construct showed improved cofactor recruitment of apo ERα-Y537S in comparison to WT ERα (Fig. S31). Apo ERα-Y537S provided a binding affinity for the LXXLL peptide of 1.7 ± 0.1 μM, which was almost similar to the amount of cofactor recruitment in the presence of agnostic ligand E2 (K D = 0.8 ± 0.1 μM). These results align with previously published data where the Y537S mutation resulted in a constitutively active conformation of ERα in the absence of agonistic ligands (54). Furthermore, ERα-Y537S was found to be less sensitive to 4-OHT inhibition of cofactor recruitment (Fig. S31). Similar to WT ERα, 14-3-3 binding did not significantly influence SRC-1 recruitment to the ERα-Y537S protein (Fig. S31).

ERα/14-3-3 PPI stabilization by FC-A is orthogonal to ERα ligand binding
The natural product FC-A is a known stabilizer of the ERα/ 14-3-3ζ PPI. This small molecule binds at the interface of ERα/ 14-3-3 protein complex (Fig. 7A) and thereby increases the affinity between the binding partners (38,46). DSF studies were used to determine the effect of FC-A on 14-3-3ζ, ERα, or the ERα/14-3-3ζ protein complex, for both WT ERα and ERα-Y537S ( Fig. 7B; Figs. S24 and S25). As expected, FC-A had no effect on the T M of 14-3-3ζ or ERα alone (Fig. S24) but increased the T M of the 14-3-3ζ protein in complex with WT ERα from 61.1 to 67.6 C (+6.5 C) (Fig. 7, B and C). Similarly, FC-A was found to stabilize the ERα-Y537S/14-3-3ζ complex as apparent from the increase 14-3-3ζ melting temperature from 60.2 to 66.3 C (+6.1 C) (Fig. S25). FC-A did not affect the T M of the ERα protein, in the ERα/14-3-3ζ complex, indicating a local effect of FC-A, confined to the composite binding pocket. This is in line with the previous observations, where the ERα F-domain acts as a long and flexible linker between the most C-terminal ERα residues binding in the 14-3-3 binding groove, and the globular ERa LBD dimer.
Interestingly, the FC-A-induced increase of 14-3-3ζ thermal stability was fully orthogonal to E2 or 4-OHT binding to both the WT ERα and the Y537S mutant. In the presence of E2 and 4-OHT, FC-A still increased the melting temperature of 14-3-3ζ in the ERα/14-3-3ζ protein complex with +6.4 C and +6.5 C, respectively (Figs. 7C and S24). In reverse, the earlier described increase in ERα melting temperature upon the addition of E2 and 4-OHT (Fig. 5, D and E) was not influenced by ERα/14-3-3ζ stabilization by FC-A (Figs. 7D and S24). E2 and 4-OHT increased the ERα T M , in the ERα/14-3-3ζ complex, with +12.8 C and +13.7 C, respectively, which was even slightly increased in the presence of FC-A (+14.1 C E2; +15.4 C 4-OHT) (Figs. 7D and S24). FC-A thus clearly stabilized the ERα/14-3-3ζ complex, for both WT ERα and Y537S mutant, and showed to be independent of ERα ligand binding.

Discussion
NR drug discovery approaches have mainly focused on targeting the NR endogenous ligand binding pocket present within the LBD. Despite great successes using this approach, significant interest has developed in alternative manners to modulate NRs. An orthogonal entry for NR modulation is offered by their PPIs with the 14-3-3 protein. However, the highly relevant molecular understanding of these NR/14-3-3 PPIs is often lacking, while this is necessary to identify new entry points for NR drug discovery. Here we studied the NR ERα and its interaction with 14-3-3 on a molecular level. Copurification of the intact complex of 14-3-3ζ and the ERα LBD and F domains revealed high-affinity binding between ERα and 14-3-3ζ via the formation of a tetrameric complex between an ERα homodimer and a 14-3-3ζ homodimer. Furthermore, the The 14-3-3/ERa-LBD-F-domain complex binding of 14-3-3ζ to the disease-relevant Y537S-ERα was confirmed, highlighting the possibility of targeting the ERα/14-3-3ζ PPI as an alternative drug discovery approach for drugresistant mutants of ERα.
Both agonist (E2) and antagonist (4-OHT) binding to the ERα LBD did not disrupt 14-3-3 binding to ERα, as apparent from SV-AUC and the native PAGE. Furthermore, natural ligand E2 induced activating conformational changes of the ERα LBD and subsequent cofactor peptide recruitment, in a similar fashion for ERα in isolation and ERα in complex with 14-3-3ζ. Similarly, synthetic ligand 4-OHT showed antagonistic behavior for ERα alone and ERα in complex with 14-3-3ζ, as observed by reduced cofactor recruitment. Finally, the ERα/14-3-3ζ PPI stabilization by FC-A was shown to be functional and not impeded, nor dependent, on ERα ligand binding in both wild-type ERα and the Y537S mutated protein.
Combined, these results inform that 14-3-3ζ binding to ERα, and stabilization of this PPI by FC-A, function independently of conformations, mutations, and liganded state of the ERα LBD. This orthogonality is most likely facilitated by the long and flexible 42-residue F-domain of ERα, accommodating ERα conformations to not affect 14-3-3 binding. Molecular stabilization of the ERα/14-3-3 protein complex with molecular glues like FC-A would therefore be a potential entry point for targeting ERα and its drug-resistant variants. The orthogonality of the molecular events within the ERα would even bode for dual targeting of both the PPI interface and the classical ERα ligand binding pocket. Furthermore, whereas orthosteric drugs such as 4-OHT also show binding to ERβ (85) and ERRγ (86), next to ERα, 14-3-3ζ binding to ERα occurs at the ERαunique C-terminus, providing the possibility to target ERα in a highly selective manner.
The concept of therapeutic targeting of the ERα/14-3-3 PPI could be envisioned to be translatable to other NR/14-3-3 protein complexes. So far, 14-3-3 has been identified as the binding partner of eight NRs, for which 14-3-3 binds to each NR in a unique manner. The NR/14-3-3 interactions form a potential entry point for targeting 'hard-to-drug' NRs due to, for example, drug resistance or the absence of an orthosteric pocket in the LBD. A potential example is the Androgen Receptor (AR), an established prostate cancer target. Although prostate cancer is initially often successfully targeted with androgen deprivation therapy or AR antagonists such as enzalutamide, drug-and castration-resistant AR mutants or splice variants are often developed within patients with prostate cancer (25,87,88). The most prevalent AR splice variant, AR-V7, even lacks the entire LBD while remaining constitutively active, making it extremely challenging to target this drug-resistant variant of AR (87,88). The binding of 14-3-3 to the NTD of AR, which remains present in the AR splice variants, provides an alternative entry point of targeting AR in drug-and castration-resistant patients with prostate cancer. In all cases, mechanistic and structural insights into the formation of the NR/14-3-3 complex, such as those obtained in this study for the ERα/14-3-3 complex are urgently needed.

Analytical SEC
Protein samples were diluted in 20 mM Tris pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM TCEP to a final concentration of 5 to 10 μM. All analytical SEC experiments were performed on an Agilent 1260 bio-inert HPLC in combination with a Superdex200 increase 3.2/300 column at a flow rate of 0.075 ml/min and 20 mM Tris pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM TCEP as running buffer. Peak detection was performed by absorbance measurements at 280 nM.

Sedimentation-velocity AUC
Protein samples were dialyzed into 20 mM HEPES pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM TCEP before all AUC measurements to obtain the best buffer match between the blank and the sample. Protein samples were diluted to their final concentrations in dialysis buffer and ligands were added where described. Samples were placed into double sector titanium centerpieces with 12-mm optical path length. SV-AUC experiments were performed using a ProteomLabTM XL-I analytical ultracentrifuge (Beckman Coulter) at 20 C and at 43.000 to 45.000 rev./min rotor speed (An-50 Ti rotor, Beckman Coulter). All sedimentation profiles were collected by absorbance measurements at 280 nm. The calculated distributions were integrated to establish the weight-average sedimentation coefficients corrected to 20 C and to the density of water (s w(20,w) ).

QToF-MS quantification
Dilution series of 14-3-3ζ-strep and ERα-pT594-strep were prepared in MQ (0.1% FA) to final concentrations of 0.025, 0.020, 0.015, 0.010, and 0.005 mg/ml. Furthermore, a 500×, 750× and 1000× dilution of the ERα/14-3-3ζ protein complex were prepared. The final samples (100 μl) were transferred to a 200 μl LC-MS vial. UPLC-QToF-MS analysis was performed on a Waters (Milford, MA, USA) Acquity I-Class UPLC system coupled to a Waters Xevo G2-XS quadrupole time-offlight (QToF) mass spectrometer. The devices were controlled by MassLynx Software (version 4.2, Waters). Full scan in positive electrospray ionization (ESI+) mode was used as MS acquisition mode with an acquisition range from 150 to 2000 m/z. A 3 μm, 100 × 2.0 mm Polaris 3 C8-A column (Agilent, Middelburg, the Netherlands) was placed inside a column oven at 40 C and used for chromatographic separation. Flowrate was set at 0.3 ml/min, and a gradient of water containing 0.1% (v/v) formic acid (A) and acetonitrile con- Data were analyzed using MassLynx software. Chromatograms were background subtracted (polynomial order 1, below curve 40%, tolerance 0.010, flatten edges). The area under the peak was determined using integration in the MassLynx software with a relative area threshold of 10. The obtained area under the curve was then plotted against the protein concentration, after which a linear regression was determined between the five data points. Using the equation of the linear regression, the concentration of 14-3-3 and ERa was determined in each protein complex sample.
To perform mass analysis of the individual peaks deconvolution was performed on m/z spectra of each individual peak. After visual inspection of the m/z spectrum, the spectrum was zoomed to the five most abundant peaks from which the mass spectrum was determined using MaxEnt1 (mass ranges 27-30 kDa or 32-36 kDa; resolution 0.10 Da/channel, Simulated Isotope Pattern with Spectrometer Blur width 0.32-0.38 Da, minimum intensity ratios left 33%, right 33%, iterate to converge). Mass spectra were centered and errors of the deconvolution process were determined.

HDX -peptide mapping
100 pmol of 14-3-3ζ-strep or ERα(PKA)-strep was mixed in 1:1 (v/v) ratio with 1 M glycine at pH 2.3 and injected on a mixed Pepsin/Nepenthesin-2 acidic protease column. Generated peptides were trapped and desalted by a Micro trap column (Luna Omega 5 um Polar C18 100 Å Micro Trap 20 × 0.3 mm) for 3 min at a flow rate 200 μl min −1 using isocratic pump delivering 0.4% formic acid in water. Both protease column and trap column were placed in an icebox. After 3 min, peptides were separated on a C18 reversed-phase column (Luna Omega 1.6 μm Polar C18 100 Å, 100 × 1.0 mm) with a linear gradient 5 to 35% B in 26 min, where solvent A was 2% acetonitrile/0.4% formic acid in water and solvent B 95% acetonitrile/5% water/0.4% formic acid. The analytical column was placed in an icebox. TimsToF Pro mass spectrometer (Bruker Daltonics) operating in positive MS/MS mode was used for the detection of peptides. Data were processed by DataAnalysis 5.3 software (Bruker Daltonics). MASCOT search engine was used for the identification of peptides using a database containing the sequence of 14-3-3ζ or ERα.

HDX
All proteins were dialyzed and diluted into 20 mM Hepes, 150 mM NaCl, 10 mM MgCl2, and 0.5 mM TCEP pH7.5 to a final concentration of 20 μM. E2 or DMSO was added to a final concentration of 150 μM. Hydrogen deuterium exchange was initiated by 10-fold dilution of the proteins under different conditions in a deuterated buffer. Fifty microliter aliquots (100 pmol) were taken after 20 s, 2 min, 20 min and 2 h of incubation in deuterated buffer, quenched by 50 μl of 1 M glycine, pH 2.3 and snap frozen in liquid nitrogen. Aliquots were quickly thawed and analyzed using the same system as described above. Peptides were separated by linear gradient 10 to 30% B in 18 min. Mass spectrometer was operated in positive MS mode. Spectra of partially deuterated peptides were processed by Data Analysis 5.3 (Bruker Daltonics) and by inhouse program DeutEx.

Native PAGE
Samples were prepared in 20 mM Tris pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM TCEP with protein at a final concentration of 2.5 to 5 μM. Ligands were added at a final concentration of 100 μM. All samples were 1:1 diluted into native PAGE loading dye (62.5 Tris pH 7.1, 75 mM NaCl, 5 mM MgCl 2 , 20% glycerol, 0.01% bromphenolblue). after which 12 μl of each sample was loaded on a 4 to 20% Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad). Gels ran at 130 V for 2.5 h at 4 C in running buffer (25 mM Tris, 192 mM Glycine, pH 8.3). Gels were washed in MilliQ (20 min), stained with Coomassie Brilliant Blue G-250 (Bio-Rad), and destained in MilliQ until bands were clearly visible. Gels were imaged and analyzed with ImageJ.

Differential scanning fluorimetry
Proteins were diluted (in 20 mM Hepes pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 500 μM TCEP) to obtain 40 μl samples containing 5 μM 14-3-3ζ-strep, 5 μM ERα-strep or 10 μM ERα/14-3-3ζ complex, with either 1% DMSO (negative control) or 100 μM ligand (E2, 4-OHT, FC-A). All samples additionally contained 10x ProteoOrange dye (Lumiprobe, 5000x stock in DMSO) and were heated from 35 to 79 C at a rate of 0.3 C per 15 s in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Fluorescence intensity was determined using excitation 575/30 nm and emission 630/ 40 nm filters. Based on these melting curves, the (negative) first derivative melting curve was obtained, from which the melting temperature T M could be determined. Reported T M values were obtained from three independent experiments from which the average and standard deviations were determined using excel.

Fluorescence anisotropy
All FA dilution series were prepared in polystyrene (nonbinding) low-volume Corning Black Round Bottom 384-well plates (Corning 4514 or 4511). FA measurements were performed directly after plate preparation, using a Tecan Infinite F500 plate reader at room temperature (l ex : 485 ± 20 nm; l em : 535 ± 25 nm; mirror: Dichroic 510; flashes: 20; integration time: 50 ms; settle time: 0 ms; gain: 60; and Z-position: calculated from well). Wells containing only fluoresceinlabeled peptide were used to set as G-factor at 35 mP. All data were analyzed using GraphPad Prism (7.00) for Windows and fitted using a four-parameter logistic model (4PL) to determine apparent binding affinities (K D app ). All results are based on two independent experiments from which the average and standard deviations were calculated to obtain the final values.
Both soaked and non-soaked crystals were fished and flashfrozen in liquid nitrogen. X-ray diffraction data were collected at the p11 beamline of PETRA III facility at DESY (Hamburg, Germany) with the following settings: 1440 image, 0.25 /image, 100% transmission, and 0.1 s exposure time. Initial data processing was performed at DESY using XDS after which preprocessed data was taken to further scaling steps, molecular replacement, and refinement.
Data were processed using the CCP4i2 suite (version 7.1.18). XDS-preprocessed data were scaled using AIMLESS. The data were phased with MolRep, using protein data bank (PDB) entry 4JC3 as a template. A three-dimensional structure of 3 0 dAc-FC-A was generated using AceDRG, which was thereafter built in the electron density based on visual inspection Fo-Fc and 2Fo-Fc electron density map. Sequential model building (based on visual inspection Fo-Fc and 2Fo-Fc electron density map) and refinement were performed with Coot and REFMAC, respectively. Finally, alternating cycles of model improvement (based on isotropic b-factors and the standard set of stereo-chemical restraints: covalent bonds, angels, dihedrals, planarities, chiralities, non-bonded) and refinements were performed using Coot and phenix.refine from the Phenix software suite (version 1.20.1-4487). Pymol (version 2.2.3) was used to make the figures in the manuscript. All structures were deposited in the protein data bank (PDB) and obtained IDs: 8C40, 8C42, 8C3Z and 8C43. See Table S1 for x-ray crystallography data statistics.

Data availability
Crystal structures described in this manuscript have been deposited to the PDB. They have the following PDB codes: 8C40, 8C42, 8C3Z, and 8C43.
Supporting information-This article contains supporting information.