Fig 1.
A cell line model for MR and GR co-expression.
(A) The repeating unit in the MMTV array comprises the MMTV LTR driving expression of viral Harvey Ras (HaRas) cDNA. Mammary adenocarcinoma line 3617 contains approximately 200 copies at one location in chromosome 4. (B) 3617ChMR cells express fluorescently tagged receptors that localize to the cytoplasm in the absence of hormone but undergo nuclear translocation and bind the array structure (arrows) in the presence of corticosterone. (C) Dexamethasone (5 nM) induces array loading of both receptors, complete GFP-GRC656G translocation, but partial mCherry-MR translocation. Aldosterone (5 nM) induces mCherry-MR array loading, complete mCherry-MR translocation, but partial GFP-GRC656G translocation. 1 nM corticosterone favours mCherry-MR translocation, likely due MR’s higher affinity for corticosterone relative to GR. Treatments were 30 min. (D) The GRC656G mutation has little effect on the corticosterone sensitivity of EGFP-GR. COS-1 cells containing no endogenous GR or MR were transfected with equivalent amounts of wildtype (wt) or C656G mutant rat EGFP-GR prior to treatment with corticosterone for 24 hrs at the doses indicated. Co-transfected pFC31-Luc provided a MMTV-driven firefly luciferase reporter and pRL-CMV a Renilla transfection control. Two-way ANOVA of Renilla-corrected luminescence showed a significant effect of dose (F(5,24) = 190.4, p<0.001), and of mutation (F(1,24) = 7.54, p = 0.011), but no interaction (dose × mutation, F(5,24) = 1.43, p = 0.251). The effect of mutation was minimal, independent samples t-testing produced no significant differences between wild type and mutant receptor output at any dose (p = 0.625, 0.717, 0.702, 0.086, 0.103, and 0.350 for 0, 4, 8, 15, 30, and 100 nM respectively). Mean ± SEM, n = 3.
Fig 2.
Co-immunoprecipitation supports MR-GR interaction in a cell line and whole rat hippocampus.
(A) Anti-GR immunoprecipitation additionally pulls down mCherry-MR when 3617ChMR cells are treated with vehicle or corticosterone. Negative controls include a non-immune IgG as the immunoprecipitating antibody and a MOCK immunoprecipitation containing antibody and buffer but no lysate. Parent 3617 cells do not express mCherry-MR but contain GFP-GR and endogenous mouse GR. mCherry-MR expected at 136 kDa. (B) Immunoprecipitation of rat GR from whole hippocampus additionally captures MR supporting interaction following stress exposure. Mean plasma corticosterone for these animals was 366 ng/ml (range 326–434 ng/ml). Endogenous MR expected at 107 kDa. (C) MR does not co-immunoprecipitate with GR when animals are bilaterally adrenalectomized (ADX, mean corticosterone 17 ng/ml, range 6–36 ng/ml). Compare signal to positive control (i.p. CORT) animal. (D) Intraperitoneal corticosterone injection (i.p., 3 mg/kg) of ADX rats restores MR co-immunoprecipitation indicating hormone dependence. Mean corticosterone was 617 ng/ml (range 516–717 ng/ml) when killed 30 min after injection at the corticosterone peak [8]. Animals for i.p. CORT controls were adrenally intact and injected with 3 mg/kg corticosterone 30 min before death (mean plasma corticosterone 586 ng/ml, range 375–948 ng/ml). Smeared MR bands with the anti-MR 1D5 antibody may suggest MR post-translational modifications. Faint bands around 50 kDa represent the non-specific labelling of the immunoprecipitating IgG common to many co-IPs.
Fig 3.
MR and GR binding during simulated ultradian corticosterone pulsatility.
(A) Representative 3617ChMR cells following 20 min, 100 nM corticosterone pulse (p) + washout (w). MMTV array formation (arrows) provides measurable loading of fluorescently tagged GR and MR at chromatinised DNA. (B) Percentage total cells over multiple fields of view forming arrays loading GFP-GRC656G, mCherry-MR, or both receptors on the same array. Mean ± SEM, n = 6 from two independent experiments, ≥130 cells per replicate. Counting 0 min samples was not routinely performed (cytoplasmic receptors) but a baseline of 2–3% cells formed weak intranuclear arrays in the absence of corticosterone. (C) ChIP targeting fluorescent receptor tags. Transient EGFP-GRC656G and prolonged mCherry-MR loading were observed at the MMTV array structure and at GRE-containing endogenous gene promoters Sgk1 and Per1. Mean ± SEM, n = 3. (D) Percentage of total cells forming arrays loading GFP-GRC656G and/or mCherry-MR. An ultradian pulse in the physiological range was simulated with 10 nM corticosterone. Mean ± SEM from two independent experiments. (E) Three successive 20 min pulses of 30 nM were applied 1 hr apart to mimic the ultradian rhythm (arrows). Each pulse produced a transient increase in GFP-GRC656G array loading, while mCherry-MR array loading was largely unresponsive to ultradian stimulation. The downward drift in occupancy over the time course likely relates to accumulated media changes. Mean ± SEM, n = 6 from two independent experiments.
Fig 4.
Cross-correlation number and brightness assay (ccN&B) reveals MR-GR interactions in the nucleoplasm of 3617 cells and at chromatin.
(A) Representative cells (scale bars 10 μm). 3617 cells +tetracycline transfected with constructs expressing the proteins indicated 24 hrs before imaging. Use of GRC656G mirrored co-immunoprecipitation experiments. Nuclear localisation was driven by at least 30 min treatment with 1 μM testosterone (EGFP-AR-mCherry) or 300 nM corticosterone (all other conditions). (B) An androgen receptor (AR) tagged at one end with EGFP and at the other with mCherry provided a positive interaction control. AR does not bind the MMTV array but nucleoplasmic Bcc values were significantly higher than negative controls EGFP + mCherry, EGFP-GRC656G + mCherry, and EGFP + mCherry-MR. Bcc values were additionally obtained from GR+GR and MR+GR pairings in nucleoplasm and at the MMTV array. (C) Interaction of mCherry-MR and EGFP-GRwt at 100 nM corticosterone in 3617 nucleoplasm. Bcc values between groups were significantly different (F(3,48.17) = 14.57, p < 0.001). Inset compares Bcc values obtained using 100 nM or 10 nM corticosterone, group means were significantly different (F(2,35.6) = 77.46, p < 0.001). There was no significant difference (ns) between groups with different corticosterone doses. (D) Neuronal cell line N2a shows significant MR-GR interaction in the nucleoplasm relative to negative controls. Group means were significantly different (F(3,56.44) = 14.95, p < 0.001). EGFPA207K prevents EGFP dimerization at higher concentrations. Welch’s F test for main effects with Games-Howell post hoc testing. Means ± SEM, number of cells indicated within or over the bar. Means significantly different from all negative control group(s) where **p < 0.01, ***p < 0.001. # Significantly different from each other, p = 0.023.
Fig 5.
Using proximity ligation assay to detect interaction between MR and GR.
(A) Structural prediction for MR and GR interaction through the classical dimerization domain in the D-loop of the DNA binding domain. Linear protein sequence highlighting D-loop amino acids using rat amino acid numbering for GR (green) and MR (red). The PDB ID for the rat GR DBD is 3G9MA. Rat MR DBD is a homology model based on human MR DBD (PDB ID 4TNTA). The interaction interface template matched by PRISM is 3g9mAB, which is a rat GR DBD homodimer. The energy score predicted by ZRANK is -53, the solvation free energy predicted by PDBePISA ΔiG is -4.6 kcal/mol, interface area is 644Å2, and the ΔiG P-value is 0.39, which passes the significance threshold of 0.5. Interacting hot spots on GR and MR are provided below the structure with complete hot spot and interface residues in S1 Table. (B) Representation of selected D-loop interface contacts of the second zinc finger mediating the interaction in (A). (C) Proximity ligation assay (PLA) identifies MR-GR interaction in 3617 nuclei transfected with untagged receptors and corticosterone treated (100 nM) for 45 min. Nuclei stained with DAPI (blue) while red dots are focal amplifications of ligated DNA probe marking interaction sites. Scale bar is 50 μm (large panels), or 10 μm (lower panels/individual cells).
Fig 6.
MR-GR interaction in 3617 nucleoplasm does not require the classical heterodimerization interface in the D-loop.
(A) Positive PLA signal is obtained only when both receptors and specific antibodies are present. Non-specific background was assessed by replacing primary antibodies with non-immune IgG from the appropriate species. Scale bar = 25 μm aside from the first panel (50 μm). (B) Mutations in the D-loop targeting the expected heterodimer interface do not prevent MR-GR interaction. Schematics show the extent to which A477K (GR)/A639K (MR) mutations combined with mutations in the salt bridge disrupt the expected interface. For GR+A477K+ and MR+A639K+ constructs the salt bridge changes made were D481R and D643R (GR and MR respectively). For GR-A477K- and MR-A639K- constructs R479D (GR) and R641D (MR) mutations were made. Scale bars = 25 μm.
Fig 7.
Experimental evidence suggests higher order MR-GR complexes in 3617.
(A) Number and brightness assay (N&B) from the green channel only to compare EGFP-rGR average molecular brightness in the presence of mCherry or mCherry-rGR. The ε value of EGFP-rGR reduces in the presence of mCherry-rGR consistent with lower stoichiometry and heterodimer formation. (B) The ε value recorded from the green channel (EGFP-rGR) in the presence of mCherry-rMR is slightly but significantly higher than in the presence of mCherry alone. (C) The ε value for mCherry-rMR in the red channel is not significantly different when co-expressed with EGFP or EGFP-rGR. (D) The molecular brightness (stoichiometry) for both EGFP-rGR and mCherry-MR in co-transfected 3617 cells was significantly higher at the MMTV array than in the surrounding nucleoplasm. Measurements were obtained from the channel in which the box plot is coloured (green, EGFP-rGR; red, mCherry-rMR). Independent samples t-tests, equal variances not assumed. All cells measured had corticosterone added at 100 nM (A-C) or 300 nM (D). Data shows median (line), mean (+) and the 25-75th percentiles at the box boundaries. Whiskers are the 10-90th percentiles with outliers as black dots.
Fig 8.
Structural predictions for MR-GR interactions through the DBD or LBD.
The interaction interface template is indicated above each structure. The following information is provided below each model; interaction energy scores predicted by ZRANK (Z), interface area in Å2, the solvation free energy ΔiG and the ΔiG P-value predicted by PDBePISA. Each model is below the significance threshold of 0.5 P-value, indicating high hydrophobicity and significance according to PDBePISA. Each interface also has energy levels comparable to the template interfaces from PDB (S4 Table). (A) A possible mode of DBD-mediated interaction of a GR+GR homodimer with a MR+MR homodimer. Structure could theoretically occur by looping intervening DNA between palindromes. (B) A possible mode of DBD-mediated interaction between two MR-GR heterodimers forming a tetramer at DNA, potentially looping out DNA between binding sites. The DBD structures from PDB are 1R4R and 3G6P for GR, and a rat homology model of 4TNT for MR. Complete hot spot and interface residues for parts A and B provided in S3 Table. (C) MR and GR LBDs may associate through multiple alternative interfaces from nuclear receptor family crystal structures indicated in parentheses. GR LBD structures used are rat models based on 4p6x (cortisol) and MR structures are rat models based on 2aa2 (aldosterone). Residues provided below each structure are interacting hot spots and/or those with highest pair potential. Complete hot spot and interface residues are provided in S5 Table. (D) Assortment of predicted assemblies through the LBD illustrating the range of structural complexes and various interfaces available for MR and GR to interact.
Fig 9.
MR and GR can interact independently of the ligands in the receptor binding pockets.
(A) Experimental comparison of Bcc values determined by ccN&B for MR-GR where different endogenous or synthetic ligands have been applied to activate both receptors (100 nM). The Bcc value for corticosterone (known interaction) was set to 100%. The greyed-out region represents the maximum mean Bcc value obtained for non-interacting controls in previous experiments. (B) Experimentally determined Bcc values for combined treatments expressed relative to 100 nM corticosterone (known interaction). Doses were varied in accordance with expectations around ligand selectivity to obtain a likely dominant occupation of each receptor by the appropriate ligands. Dexamethasone 10 nM + aldosterone 10 nM, spironolactone + RU486 (1 μM each), aldosterone + RU486 and corticosterone + RU486 (10 nM MR-targeted agonist, 1 μM GR-targeted antagonist). Mean ± SEM, number of cells shown within bar. *Significantly different, p<0.05; ns, not significant; nptd, not possible to determine. (C) ZRANK binding energy scores of computational predictions of alternative interfaces between MR and GR. Different agonist and antagonist combinations and structural conformations demonstrate the potential for multiple modes of interaction through the LBD regardless of the bound ligand, with varying interface preference. Note the different trends in interface preferences, especially with the open antagonist bound form. cortico., corticosterone; aldo, aldosterone; dexa, dexamethasone; spiro, spironolactone; “–”indicates no predictions were found by PRISM for the indicated pair of proteins through the template interface. (D) Structural examples of MR-GR interactions through the LBD when both are agonist bound, and when GR or MR is antagonist bound. For each pair of interactors, the most favourable interface is displayed with the PDB IDs of the templates for rat homology modelling provided below each structure. The following information is provided below each model; interaction energy scores predicted by ZRANK (Z), interface area in Å2, the solvation free energy ΔiG and the ΔiG P-value predicted by PDBePISA.