Identification of co-regulators recruited to ERα mutants
Since Y537S, D538G, and ESR1-YAP1 ERα (Fig. 1a) promote E2-independent transcriptional activation of an ERE-dependent reporter (Fig. 1b), we tested whether this activation is through recruiting or repelling distinct co-regulators (coactivators and co-repressors). Using our 4xERE DNA pulldown assay to identify co-regulators recruited to ERα , we first utilized recombinant WT and Y537S ERα, along with HeLa S3 nuclear extract (HNE) to form complexes. Washed complexes were then subjected to label-free, quantitative MS and bound proteins were normalized by the amount of ESR1 N-terminal peptides bound (Fig. 1c, top; Supplementary Table 2).
Compared with unliganded WT ERα, a subset of coactivators were recruited in an enhanced manner to Y537S ERα (Fig. 1c, bottom). Namely, the histone H3 lysine 4 (H3K4) methyltransferase KMT2D complex displayed the greatest enrichment with Y537S ERα, along with SRC-1, -2, and -3, p300, CBP, and KMT2D’s paralog, KMT2C. Immunoblotting validated the enhanced KMT2D and SRC-3 recruitment (Fig. 1d).
As SRC-3 directly binds Y537S ERα [14, 19], we tested whether a KMT2D complex  would directly interact with Y537S ERα. Indeed, the binding of the KMT2D complex was enhanced with Y537S, along with E2-bound WT, compared to unliganded WT ERα (Fig. 1e). We found other potential coactivators with enhanced binding to Y537S ERα (Supplementary Figure 1a). We further validated PELP1 recruitment by immunoblotting, as an inhibitor disrupting this interaction is described  (Supplementary Figure 1b). We found very few co-repressors had reduced recruitment to Y537S ERα (Supplementary Figure 1c).
As purified D538G ERα failed to recruit SRC-3 despite binding EREs (data not shown), we resorted to using extracts from transfected 293T cells as sources of WT, Y537S, and D538G ERα proteins. We again observed enhanced recruitment of KMT2D and SRCs to Y537S compared to unliganded WT ERα (Supplementary Figure 2). The D538G mutant also recruited these coactivators, but three to four times less than that of Y537S. Additional potential co-regulators displayed enhanced binding to both ERα mutants. However, for subsequent functional characterization, we chose to focus on SRCs and the KMT2D complex.
SRCs are critical for ERα LBD mutant activity and cell growth
We next tested the functional consequence of enhanced SRC-ERα mutant interactions on transcription. Knockdown of SRC-3 using published siRNAs  significantly reduced both Y537S and D538G ERα-mediated transcriptional activity on the ERE-Luc reporter (Fig. 2a). Treatment with a small molecule inhibitor (SMI), SI-1, which inhibits the activities of all three SRCs (IC50 = 0.2 μM)  and reduces SRC protein levels (Fig. 2b), severely reduced the transcriptional activities of WT and mutant ERα (Fig. 2c). Cell viability was minimally affected (Supplementary Figure 3a). We also tested whether the combination of an oral SERD and SI-1 would further reduce LBD mutant ERα transcriptional activity. We focused on AZD9496  (AZD) as it: (1) reduced endogenous ERα protein, (2) was significantly more potent than ICI182,780 (ICI, fulvestrant) in reducing mutant ERα transcriptional activities (unlike another SERD GDC-0810  (GDC; Supplementary Figure 3b-d), and (3) AZD was reported as more effective than ICI at inhibiting tumor growth promoted by Y537S ERα . The combination of SI-1 and AZD synergistically reduced both Y537S and D538G activities on the ERE-luciferase (Luc) reporter (Fig. 2d, e, Supplementary Table 4).
We tested whether SRC inhibition would affect growth of stably expressing WT or Y537S ERα MCF-7 cell lines . SI-1 at 400 nM reduced viability in both WT and Y537S ERα-expressing cells by 91% (Fig. 2f). When combinations of SI-1 with AZD or ICI were tested, a synergistic reduction in viability of Y537S ERα-expressing cells was found at the two highest combined doses or highest combined dose, respectively (Fig. 2f and Supplementary Figure 3e; Supplementary Tables 5 and 6). Thus, SI-1 combined with AZD was most effective in reducing both transcription and cell growth mediated by mutant ERα proteins.
Inhibiting SRCs and mutant ERα most effectively reduces patient-derived xenograft tumor growth
We next tested the efficacy of an improved pan-SRC inhibitor (SI-2) , AZD, or the combination in a patient-derived xenograft (PDX) expressing Y537S ERα (WHIM 20 ) for tumor reduction (Fig. 3a). SI-2 was chosen instead of SI-1, given its reduced IC50 (3.4 nM) and ability to reduce ER– tumor growth . After tumors grew to 350 mm3, mice were randomized, E2 was withdrawn to mimic AI treatment, and mice were then treated with control vehicle, SI-2, AZD, or the combination. After 4.5 weeks, SI-2 alone significantly reduced tumor volume, AZD gave a larger reduction, but the combination gave the most significant reduction in tumor growth.
We next confirmed that the drugs indeed affected their intended targets and tested for effects on apoptosis and proliferation. First, as expected, AZD treatment significantly reduced ERα expression in tumor lysates. Unexpectedly, AZD treatment upregulated SRC expression (Fig. 3b, Supplementary Figure 4a), which may have relevance for patients receiving AZD monotherapy in clinical trials (NCT02248090/NCT03236974). Second, we tested the expression of a classical ER target gene, PR (Fig. 3c, Supplementary Figure 4a). While SI-2 did not affect PR expression, AZD clearly did. Third, we found that SI-2 increased an apoptosis marker (cleaved PARP protein), while AZD decreased proliferation as measured by BrdU incorporation (Fig. 3d, e, Supplementary Figure 4b). Finally, we did not observe any significant toxicity with any drug treatment after examining recipient mouse livers by histochemistry and measuring body weights (Supplementary Figures 4c-d). Thus, our PDX data support a potential new treatment regime for breast cancers bearing ESR1 LBD mutations, which is to combine a SRC inhibitor with an oral SERD.
KMT2C/2D are novel coactivators for Y537S ERα
From above, we found that the KMT2C/2D complexes were preferentially enriched with Y537S ERα (Fig. 1c, Supplementary Figure 2a). To determine the functional role of KMT2C/2D, we tested whether their depletion would affect Y537S ERα-mediated reporter expression. Knocking down KMT2C, KMT2D, or both together reduced Y537S ERα transcriptional activity (Fig. 4a). Upon double KMT2C/2D knockdown, WT ERα transcriptional activity was also reduced (Fig. 4b). However, ESR1-YAP1 transcriptional activity was not affected, ruling out a general transcriptional effect.
We additionally found that KMT2C/2D knockdown reduced anchorage-independent growth of WT, Y537S, and D538G ERα-expressing cells (Supplementary Figure 5), with significant differences observed between WT and mutant ERα data. More importantly, knockdown of KMT2C/2D significantly sensitized the partially resistant Y537S ERα cells to anti-estrogens currently given in the clinic (Fig. 4c). Overall, our data reveal an important functional role for KMT2C/2D with Y537S ERα in transcription and cell growth.
SRCs and KMT2D are crucial for growth of inducible LBD mutant ERα cells
As our above cell lines stably overexpressed LBD ERα mutants , we created MCF-7 cell lines supporting conditional (doxycycline (Dox)-inducible) expression of FLAG-tagged WT, Y537S, or D538G ERα. The FLAG tag did not impair the transcriptional activities of ERα proteins (Supplementary Figure 6). Dox addition to cells grown in charcoal-stripped media induced expression of these ERα proteins, with Y537S and D538G ERα accumulating to a similar extent but less than WT (Fig. 5a). We next performed co-immunoprecipitations to test whether SRC-3 and KMT2D displayed enhanced association with inducible mutant ERα proteins under hormone-depleted conditions. We found greatest KMT2D and SRC-3 association with the Y537S mutant, with less recruited to D538G ERα (Fig. 5b). Importantly, the Dox-inducible mutant ERα proteins activated endogenous ERα target genes (GREB1 and TFF1) in a hormone-independent manner and D538G ERα had a weaker effect than Y537S, in accordance with other models [7, 9, 29,30,31] (Fig. 5c and Supplementary Figure 7a). We next asked whether ablation of these key co-regulators would inhibit the viability of these cells. Consistently, SI-1 significantly reduced the viability of all ERα-expressing cells (Fig. 5d), while knockdown of KMT2C/2D selectively affected WT and Y537S ERα-expressing cells (Fig. 5e).
Knockdown of KMT2C/2D modulates Y537S ERα direct target gene expression
We next tested the effect of KMT2C/2D depletion on GREB1 and TFF1 gene expression in our Dox-inducible MCF-7 cells. KMT2C/2D knockdown reduced Y537S ERα-mediated transcription of both genes (Fig. 5f and Supplementary Figure 7b). In WT ERα cells, the loss of KMT2C/2D also reduced TFF1 mRNA, but GREB1 or ESR1 were unaffected. We extended our analysis to 10 total ERα target genes by depleting only KMT2D using a validated siRNA , as KMT2D had greater recruitment than KMT2C to Y537S ERα (Fig. 1c). We observed that Y537S mutant expression regulates select ERα targets, a subset of which is reduced by KMT2D depletion (Supplementary Figure 7c).
Chromatin immunoprecipitation (ChIP) was used to examine whether the ERα mutants bound EREs [33,34,35] located upstream of the GREB1 and TFF1 gene transcription start sites (TSS) in a Dox-dependent manner (Fig. 5g and Supplementary Figure 7d). All ERα proteins displayed Dox-dependent enrichment on EREs upstream of the GREB1 and TFF1 TSS, but not on a negative control region . Thus, the binding of the LBD mutant ERα proteins to multiple EREs suggests direct regulatory roles in regulation of these genes.
Chromatin occupancy of Y537S ERα and KMT2D are positively correlated
As SRC-3 and p300 are co-localized with Y537S ERα on chromatin , we next tested whether KMT2D chromatin occupancy correlated with Y537S ERα. We performed ChIP using a validated KMT2D antibody [37, 38] (Fig. 5h and Supplementary Figure 7e). KMT2D occupancy increased in Y537S ERα cells in a Dox-dependent manner for all EREs assayed, but not for the control region, without increased KMT2D protein expression (Supplementary Figure 7f). Thus, Dox-induced occupancy of EREs by Y537S ERα is correlated with increased recruitment of KMT2D and enhanced transcription of GREB1 and TFF1.
Co-regulators recruited to the ESR1-YAP1 fusion
We next investigated co-regulator recruitment to ESR1-YAP1, which contains the N terminus and DBD, but lacks the LBD, of ERα fused in-frame to the C terminus of the Yes-associated protein 111 (Fig. 1a). ESR1-YAP1 promotes high levels of expression of an ERE-dependent reporter without E2 (Fig. 1b and Supplementary Figure 6a). We compared co-regulator recruitment to either purified WT ERα or ESR1-YAP1 bound to ERE DNA in the absence of E2 (Fig. 6a). After normalization for ERα-binding differences, the ESR1-YAP1 protein did not recruit more SRC-3 and KMT2D as compared to WT ERα, unlike the LBD ERα mutants (validated in Fig. 1d). Instead, the ESR1-YAP1 protein recruited many subunits of the 26S proteasome (Fig. 6a). Immunoblotting validated enhanced proteasome recruitment (SUG1/PSMC5 and 20S C2/PSMA1) of ESR1-YAP1 vs. WT ERα, even with E2 (Fig. 6b). We further validated that the 26S proteasome was recruited to ESR1-YAP1 in E2-deprived T47D stable cell line  extracts (Supplementary Figure 8a).
The 26S proteasome plays a role in WT ERα degradation that is linked with the receptor’s ability to activate ERE-driven reporter genes [39,40,41]. We thus tested whether a proteasome inhibitor, MG132, would similarly inhibit ESR1-YAP1 transcriptional activity. Indeed, MG132 treatment of cells transfected with an ERE-driven reporter and an ESR1-YAP1 expression plasmid reduced Luc activity, as compared to the vehicle control (Fig. 6c). Interestingly, MG132 or the FDA-approved proteasome inhibitor, bortezomib, could inhibit the transcriptional activity of a GAL4 DBD-YAP1 fusion (Supplementary Figure 8b).
Proteasome activity is important for ESR1-YAP1-mediated cell growth and gene expression
We next tested the functional significance of the ESR1-YAP1: 26S proteasome interaction in T47D stable lines expressing HA-tagged YFP, WT ERα, or ESR1-YAP1 proteins (Fig. 7aii and Supplementary Figure 8c). Increasing concentrations of bortezomib treatment for 3 days efficiently reduced the growth of all three T47D lines compared to the vehicle control (Fig. 7a,i).
We next wanted to define the effect of ESR1-YAP1 expression in T47D cells grown in an E2-deprived state on ERα target gene expression. The ESR1-YAP1 fusion promoted the expression of two target genes, TFF1 and PGR, significantly above the level of unliganded WT ERα-expressing cells (Fig. 7b), even though much less ESR1-YAP1 fusion was expressed (Fig. 7a, ii and Supplementary Figure 8c). The regulation of these two genes is likely direct, as ChIP assays revealed ESR1-YAP1 occupancy at two defined ERE enhancer-like sequences [35, 42] (Fig. 7c and Supplementary Figure 8d).
The effect of proteasome inhibition on TFF1 and PGR gene expression in E2-deprived ESR1-YAP1-expressing cells was tested by treating cells with bortezomib (Fig. 7d). Proteasome inhibition had both a dose-dependent and gene-specific effect on the ESR1-YAP1 targets, decreasing PGR and increasing TFF1, which resembles the effect of proteasome inhibitors on these genes after E2 treatment of MCF-7 cells [43,44,45,46]. Bortezomib decreased endogenous ESR1 mRNA expression, consistent with prior reports [43, 47]. Finally, we observed that bortezomib stimulated ESR1-YAP1 mRNA expression (driven by the CMV promoter), which may explain why bortezomib did not severely reduce ESR1-YAP1 cell viability as compared to overexpressed WT ERα (Fig. 7a). Thus, the proteasome modulates ESR1-YAP1 target genes and cell growth, suggesting a new approach for treating tumors bearing this class of resistance mutation.
Potential clinical relevance of mutant ERα-binding coactivators
To investigate whether the expression levels of the coactivators identified in this study correlate with patient outcomes, we queried two existing expression data sets—the Symmans Breast 2 ER-positive tamoxifen-treated patients  or ER-positive breast cancer patients treated with endocrine therapy (in KM plotter ; Supplementary Figure 9). In the Symmans data set, we found significantly higher KMT2D mRNA levels in patients who had a metastatic occurrence after 3 years of tamoxifen treatment vs. those that had not (panel a). We also found that higher SRC-3 and proteasomal subunit mRNA levels correlated with reduced survival from distant metastasis (panels b, c). In the KM plotter analysis, we found that higher KMT2D, but not KMT2C, mRNA significantly correlated with reduced recurrence-free survival (panel d). Finally, we assayed a metastatic breast cancer mutation/amplification database  and found that the ESR1 LBD mutations were not present in patients with mutations in either KMT2C or KMT2D genes (panel e).