Mouse mammary ERRα DBD (70–170) was cloned in the in-house expression vector pnEAtH. The vector was transformed in Escherichia coli BL21 (DE3) pRARE2 strain, grown at 37°C and induced for protein expression at OD_600nm = 0.6 with 1 mM IPTG at 25°C for 3 h. The cell pellet was resuspended in binding buffer (20 mM Tris pH = 8.0, 400 mM NaCl, 10% glycerol, 4 mM CHAPS) and lysed by sonication. The crude extract was centrifuged at 45,000 × g for 1 h at 4°C. The lysate was loaded on a Ni affinity step on HisTrap FF crude column (GE Healthcare, Inc.), and the protein was eluted at a concentration of 150 mM imidazole. The hexahistidine tag was cleaved overnight using thrombin protease. ERRα-DBD was subjected to a Heparin column using an increasing salt gradient. Finally, the ERRα-DBD was polished by size-exclusion chromatography (SEC) in a SEC buffer (50 mM Bis Tris pH = 7.0, 120 mM KCl, 0.5 mM CHAPS, 4 mM MgCl₂) by using a Superdex S75 16/60 column (GE Healthcare). Complexes were formed by mixing the appropriate DNA solution to the purified ERRα DBD to achieve desired DNA:protein molar ratios. The complex was incubated at least 30 min at 4°C before performing the respective experiments.

HPLC purified single-stranded DNAs were ordered from Sigma-Aldrich, Inc. or Euromedex France. The single-stranded DNA oligonucleotide and its respective anti-sense DNA fragment were mixed in the DNA annealing buffer (10 mM Tris pH = 8.0, 100 mM NaCl, 1% DMSO, 0.1 mM EDTA), heated at 95°C, and gradually cooled down to 4°C using a BIO-RAD PCR machine. 4 mM MgCl₂ was added to the cooled annealed DNA sample.

The protein–DNA complexes were run on an 8% polyacrylamide gel (PAGE) at 2 W constant power after pre-running the gel for 40 min at 4°C. Two different native gel systems were used (i) Tris/CAPS (pH = 9.4) buffer system contained 60 mM Tris base and 40 mM CAPS (3-cyclohexil-amino-1-propane-sulfonic acid) (ii) Ammonia/CAPS buffer (pH = 10.4) system (containing 37 mM ammonia and 60 mM CAPS). Approximately 3–5 µg protein was loaded per lane along with its DNA counterpart at defined molar ratios. The polyacrylamide gels were stained using Instant Blue Protein Stain (Expedeon Protein Solutions) for 15 min and rinsed in water.

Size-exclusion chromatography (SEC)-MALLS/QELS experiments were performed on a multi-angle laser light scattering detector (miniDAWN TREOS, Wyatt Technologies) coupled in-line with SEC and an interferometric refractometer (Optilab T-rEX, Wyatt Technologies). A Superdex S75 or S200 10/300 GL column (total volume 24 mL, GE Healthcare) with a flow rate of 0.5 mL/min was used to separate the sample before performing the MALLS/QELS measurement. Experiments were done with 50–100 µL receptor–DNA complex samples at concentrations between 1 and 3 mg/mL in 20 mM Bis Tris pH = 7.0, 120 mM KCl, 1 mM MgCl₂, 0.5 mM CHAPS, 1 mM TCEP. The molar mass was determined by construction of Debye plot using Zimm formalism [plot of K*c/R(θ) as a function of sin²(θ/2)] at 1-s data interval. The analysis of the data was performed using the ASTRA 6.1software (Wyatt Technologies).

Native nano-Electrospray-Mass Spectrometry (nanoESI-MS) analyses were performed on an electrospray quadrupole-time-of-flight mass spectrometer (Synapt G2 HDMS, Waters, Manchester, UK) coupled to an automated chip-based nanoESI source (Triversa Nanomate, Advion, Ithaca, NY, USA). The mass spectrometer was calibrated using singly charged ions produced by a 2-g/L solution of cesium iodide (Acros organics, Thermo Fisher Scientific, Waltham, MA, USA) in 2-propanol/water (50/50 v/v). Samples were buffer exchanged in 150 mM ammonium acetate (NH₄Ac), pH 7.1 buffer using 0.5 mL Zeba™ Spin desalting Columns (Thermo Fisher Scientific, Waltham, MA, USA) and centrifuged at 1,500 × g for 2 min. After buffer exchange, concentrations were determined by UV-Vis using a Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Samples were diluted to 5 µM in 150 mM ammonium acetate, pH = 7.1 buffer and directly infused into the mass spectrometer. Instrumental parameters were optimized for the detection of labile non-covalent complexes by raising the interface pressure to 6 mbar and the cone voltage to 140 V. Data treatment was realized with MassLynx 4.1 software (Waters, Manhester, UK).

Nuclear magnetic resonance experiments were recorded at 700 MHz on an Avance III Bruker spectrometer equipped with a TCI z-gradient cryoprobe. NMR data were processed using TopSpin (Bruker) and analyzed with Sparky software packages (36). NMR experiments were performed in 50 mM sodium phosphate buffer (pH 7.0), 5 mM MgCl₂, and 1 mM TCEP in 90/10 H₂O/D₂O. PAGE-purified DNAs were annealed by heating at 95°C and snap-cooled at 4°C. The concentration of DNA samples was measured using a Nanodrop Spectrometer and calculated with molar extinction coefficients. DNA samples volumes were 150 µL in 3-mm NMR tubes. ¹H assignments were obtained using standard homonuclear experiments. NMR data were acquired at 15°C and 20°C. Solvent suppression was achieved using the “Jump and Return” sequence combined to WATERGATE (37–39).

Two-dimensional NOESY spectra were acquired with mixing time of 400 and 50 ms. Base pairing was established via sequential nuclear Overhauser effects (NOEs) observed in 2D NOESY spectra at different mixing times. Complexes were formed by stepwise addition of ERRα DBD to 100 µM DNA samples, to achieve DNA:protein ratios of 0.5, 1.0, 1.5, and 2.0. The concentrations of protein samples ranged from 0.8 to 1.4 mM maintaining minimal dilution of the DNA samples. All titrations experiments were performed at 20°C, monitoring the imino protons region of 1D spectra.

Estrogen-related receptor binds to IR3 REs encompassing two 6 bps DNA-binding sites separated by 3 bps, as well as to 9 bps extended half-site REs composed of a single specific DNA-binding site of the type TNAAGGTCA (ERRE), the latter REs representing most of the natural REs found in target gene promoters. Here, we first address the issue of complex formation of ERRα DBD-DNA for an extended range of DNA targets, using polyacrylamide Ammonia/CAPS or Tris/CAPS native gel electrophoresis. We considered 33 bp DNA fragments originating from the tralpha (tra) and lactoferrin (lf) (40, 41) gene promoters (Table S1 in Supplementary Material) as representatives of specific natural ERREs (Figure 1A) and reconstituted complexes between ERRα DBD and DNA using increasing ratios of protein to DNA. For these two different natural ERREs, the ERRα DBD–ERRE complexes migrate as a single band. In addition, we considered two tra ERRE fragments of different length, a short one of 13 bp only that solely encompasses the extended 5′-TGAAGGTCA-3′ motif, where only one DBD subunit is allowed to bind and a longer one with 29 bps, which would potentially allow the binding of a DBD dimer (Figure 1B). We observed a single migration band for the 13- and 29-bp ERRE complexes, irrespective of the ratio between DNA and the DBD, from sub-stoichiometric (1:0.5) to higher ratios (up to 1:4). This suggests that a monomer is bound to the 13-bp DNA fragment and remains monomeric in the case of longer DNA fragments (as shown with 29 and 33 bps). No potential dimer formation seems to occur when the amount of DBD subunits is increased. When comparing the migration of ERRα DBD on the ERE/IR3 of the tff1 gene promoter (16, 42) (tff1 ERE-IR3) with that of ERRα DBD-tra ERRE, a similar migration pattern is observed, suggesting a monomer on this IR3 RE (Figure 1C, lanes 1, 2). In contrast, when considering a composite element made of an extended half-site ERRE embedded into an IR3 RE (embERRE/IR3), we observed a delayed migration of the complex into the gel (Figure 1C, lanes 3, 5). Similarly, when the ratio between the embERRE/IR3 (26 bps here) and DBD is increased, from stoichiometric (DNA:protein = 1:0.25) to higher ratios (up to 1:4), the low migration band disappears and a higher migration band is observed (Figure 1D). These data suggest that a different type of ERRα DBD–DNA complex can be formed on the embERRE/IR3 sequence, as compared to the monomeric ERRα DBD–tra ERRE, –lf ERRE or –tff1 ERE-IR3 complexes and likely corresponds to dimeric ERRα DBD on DNA. To gain insight into the unexpected behavior of ERRα DBD on the embERRE/IR3sequence, we explored more sequences belonging to either the IR3-type of REs or to the natural extended ERREs. We found diversity in the migration pattern inside each of the two different types, as shown for tff1 ERE-IR3 (16, 42) and the rb1cc1 IR3 (43) as examples of IR3-binding sites, or tra ERRE and tff1 ERRE as examples of ERRE natural DNA REs (Figures 1C,E). While the tff1 ERE-IR3 and the tra ERRE result into a single low migration band, the rb1cc1 IR3 and the tff1 ERRE give a more complex migration pattern with mainly two bands, a lower migration band and a delayed migration band similar to what is observed in the case of ERRα DBD-embERRE/IR3. On the other hand, different embedded ERRE/IR3 REs do not necessarily lead to the observation of two migrating species, as shown in Figure 1F, for the natural ERRE/IR3 RE of the abcc5 gene promoter. Altogether, these data suggest that the existence of a single monomeric species or the occurrence of higher molecular complexes, possibly dimer, on DNA strongly depends on the exact nature of the DNA sequence. In fact, the DNA sequence is likely to be the key factor for the stabilization of ERRα DBD dimer on DNA.

Based on our results obtained from native polyacrylamide gels, we wish to assess the exact binding stoichiometry of ERRα DBD on different DNA sequences and first relied on electrospray mass spectrometry (ESI-MS) analysis carried out under non-denaturing conditions. The mass measurement (29,737 ± 1 Da) indicates that the complex between ERRα DBD and tra ERRE is composed of a single DBD subunit on the DNA sequence (Figure 2A; Table S2 in Supplementary Material). When ERRα DBD–embERRE/IR3 complexes are considered, MS analysis indicate not only the presence of a monomer but also of a dimer of ERRα DBD on this RE whatever the length of the latter sequence (29 and 33 bps in Figures 2B,C), see Table S1 in Supplementary Material for the sequences and Table S2 in Supplementary Material for the measured masses. The results of the MS analysis are consistent with the biochemical data from the native gel electrophoresis and allow the attribution of the two migration bands observed in the native gels for these complexes as the monomeric and the dimeric ERRα DBD-DNA species. MS analysis further suggests that ERRα DBD can also form a dimer on the IR3 RE of the rb1cc1 gene promoter (Figure 2D), suggesting that the presence of a dimer is not limited to the embedded RE embERRE/IR3, an observation fully confirmed by the other techniques used in this study.

Mass spectrometric results suggest the existence of monomeric and dimeric ERRα DBD species on the embedded ERRE/IR3 REs and native gels indicate that the existence of larger, possibly dimeric species is not limited to this specific RE, but to other DNAs, as shown for rb1cc1 IR3 and tff1 ERRE. Since mass spectrometric analysis in non-denaturing conditions is performed in ammonium acetate, we decided to undertake multi-angle laser light scattering experiments (MALLS) coupled to SEC in more physiological buffer conditions. SEC-MALLS allows the quantitative in-line measurements of the molar mass of the species in solution that are first separated on a SEC column. We first considered ERRα–DBD complexes with 26-bps DNA fragment of either tra ERRE, tff1 ERE/IR3, or embERRE/IR3 and added a rather large excess of DNA while reconstituting the complexes, resulting in a chromatogram with two overlapping peaks, one belonging to the DBD–DNA complex and the other one, on the right, to the free DNA. As shown in Figure 3A, the two complexes ERRα DBD–tra ERRE and ERRα DBD–tff1 ERE/IR3 are composed of one monomer on DNA, while the measured molar mass of ERRα DBD-embERRE/IR3 corresponds to the mass of the dimer. As observed in the native gels (Figure 1F) and shown in Figure 3B, ERRα DBD is a monomer on the embedded ERRE/IR3 RE of the abcc5 gene promoter. While the binding stoichiometry of all the complexes was rather unambiguous, we wondered how the complexes between ERRα DBD and rb1cc1 IR3 or tff1 ERRE would behave since we saw that their migration pattern on the gels was rather complex. As shown in Figure 3C, these measured molar mass of the two complexes reflects the coexistence of dimer (at the left-hand side of the SEC peak) and monomer (toward the center of the peak) on DNA. This suggests that the dimer is less stable on these DNAs when compared to embERRE/IR3 and that both species are in equilibrium in solution. The results of SEC-MALLS experiments are thus consistent with ESI-MS data, and strongly suggest that the dimerization of the isolated ERRα DBD on DNA is possible, but strongly depends on the DNA sequence of the target binding site.

Nuclear magnetic resonance was used to investigate the binding properties of ERRα DBD onto different types of DNA: tff1 ERE/IR3, lf ERRE, tra ERRE, rb1cc1 IR3, and embERRE/IR3. NMR spectra display imino protons resonances (found between 12 and 14 ppm) provided that they are protected from exchange with the solvent. Their observation indicates that the corresponding protons are involved in hydrogen bonds, generally due to the formation of base pairs. All base paired imino protons of free DNAs were first assigned via sequential NOEs observed in 2D NOESY (Figure 4). The A:T Watson–Crick base pair were discriminated from G:C base pair by the strong correlation observable between T H3 imino proton and the H2 proton of adenine. In a G:C Watson–Crick base pair, two strong NOEs are observable between the guanine H1 and the cytosine amino protons. The chemical shifts of imino protons in 1D spectra were then monitored as a function of ERRα DBD concentration and were used to map the DNA regions in contact with the protein. Complexes were formed by stepwise addition of ERRα DBD to DNAs. Successive additions of ERRα DBD on DNAs resulted in several changes in the imino protons region. Addition of ERRα DBD to tff1 ERE/IR3 DNA induces chemical shifts of the resonances corresponding to T3, T46, G42, and G48 imino protons, indicating that the protein binds the first extended site. In contrast, the resonances corresponding to T16 and T32 imino protons remain unaffected, suggesting that ERRα DBD does not bind the second site (Figure 4A; Figure S1 in Supplementary Material). With lf ERRE DNA, most of spectral changes involve nucleotides located in the first extended site. Seven imino protons, T3, T10, T13, T41, G42, T46, and G51 undergo clear chemical shifts, while resonances corresponding to G25, T29, and T32 remain unchanged (Figure 4B; Figure S2 in Supplementary Material). These observations indicate that ERRα DBD binds only the first site. The ERRα DBD titration performed on tra ERRE DNA also induces changes in the imino protons region. Despite the complexity of the spectra due to overlapping of several resonances, clear chemical changes are observable for T2, G8, G9, G42, and G48 residues, while G21 and G28 do not undergo variations (Figure 4C; Figure S3 in Supplementary Material). This suggests that ERRα DBD binds the first extended site and not the second. With rb1cc1 IR3 DNA, similarly to tra ERRE DNA, due to overlapping of resonances, we only considered the imino protons that were well resolved. Addition of ERRα DBD to rb1cc1 IR3 DNA induces a clear chemical shift of the resonance corresponding to G5 and G36 residues (Figure 4D; Figure S4 in Supplementary Material), indicating that the protein contacts the two sites. Finally, titration performed on embERRE/IR3 DNA also induces perturbations in the imino protons region of 1D spectra. Resonances corresponding to T21, T2, T4, T10, T16, T35, and T41 undergo clear chemical shifts upon ERRα DBD binding (Figure 4E; Figure S5 in Supplementary Material). This shows unambiguously that ERRα DBD binds the two sites within the embERRE/IR3 DNA. Taken together, NMR data suggest that the binding of ERRα DBD affects the first extended site in tff1 ERE/IR3, lf ERRE, and tra ERRE DNAs, while the protein binds two sites in the case of rb1cc1 IR3 and embERRE/IR3 DNAs, in perfect agreement with the other biophysical and biochemical data.

For some of the DNA sequences, the ERRα DBD can bind a dimer. This is clearly observed for embERRE/IR3, and to some extend for tff1 ERRE and rb1cc1 IR3, but neither for the tff1 ERE/IR3 nor for tra ERRE. Thus, the question arises as to which region is key for the stabilization of a DBD dimer on DNA, since the mere presence of two half-sites is not enough to explain dimerization. Therefore, we considered two DNA sequences, one which is always bound by a monomer (tff1 ERE/IR3) and another one (embERRE/IR3) which leads to dimer stabilization. We then considered chimeras of these two DNA-binding sequences (Figure 5A), where the 5′ region, including the flanking region and the first half-site belongs to one of the sequence, whereas the 3′ region, including the spacer, the second half-site and the flanking region comes from the other DNA. This results into two chimeras, 5′-tff1 ERE/IR3_3′embERRE/IR3 (5′tff) and 5′embERRE/IR3_3′-tff1 ERE/IR3 (5′emb). The native ammonia/CAPS native gels of the ERRα DBD–DNA complexes indicate that the two chimeras behave differently. The ERRα DBD-5′emb (Figure 5B, lanes 5, 6) behaves as a monomer on DNA, in a way similar to ERRα DBD-tff1 ERE/IR3 (Figure 5B, lanes 3, 4), while the ERRα DBD-5′tff (Figure 5B, lanes 7, 8) allows the formation of a dimer on DNA, like embERRE/IR3 (Figure 5B, lanes 1, 2). SEC-MALLS experiments confirmed these observations (Figure 5C), where the measured molar masses for the peak are close to the expected values of the monomer or the dimer on DNA. Similarly, the MS analysis performed under non-denaturing conditions confirmed the existence of dimeric species for both ERRα DBD-embERRE/IR3 and ERRα DBD-5′tff complexes (Figure 5D; Table S2 in Supplementary Material). For the latter samples, we noticed that an additional species is present that corresponds to two concatenated DNA molecules. These species come from the initial preparation of DNA (as shown in Figure S6 in Supplementary Material). Taken together, the analysis of the chimeras indicate that the key elements in the DNA sequence for the binding of an ERRα DBD dimer are located in the 3′ region and cannot be attributed to the 5′ 3bp extension present in the embERRE/IR3 sequence.

In order to get a more detailed molecular insight into the key determinants of ERRα DBD dimerization on DNA, we hypothesized that the 3′-region that differs between tff1 ERE/IR3 and embERRE/IR3 is crucial for dimer binding and designed a series of mutants in this region (Figure 6A). The nucleotides at positions +7 to +10 that include the last base pair of the half-site and the flanking region differ between tff1 ERE/IR3 and embERRE/IR3. The sequence at these positions is C/G rich with A/T at position +7 for tff1 ERE/IR3, while for embERRE/IR3 it is composed of three T/A base pairs followed by A/T. This suggests that the complementary strand of embERRE/IR3 between position +5 and +9 contains only purines (+5-GGAAA-+9), while the related sequence of tff1 ERE/IR3 is G rich with a T in the middle of the stretch at position +7 (+5-GGTGG-+9). Therefore, we first changed the (+5-GGAAA-+9) stretch of embERRE/IR3 to (+5-GGTGG-+9), resulting in the destabPu-embERRE/IR3 (destabPu) DNA sequence. In parallel to this, the tff1 ERE/IR3 GGTGG was modified to GGAAA to mimic locally the sequence seen in embERRE/IR3, leading to tff1 ERE/IR3-Pu sequence (Pu). A significant destabilization of dimer formation is observed for the ERRα DBD–destabPu complex, as seen in the native polyacrylamide gel when comparing the intensity of the dimer band with that of ERRα DBD-embERRE/IR3 (Figure 6B, lanes 2, 5) and measured with SEC-MALLS (Figure 6C), where the SEC peak is shifted to the lower mass range (arrow in Figure 6C). However, the sample fraction collected under the maximum of the SEC peak and run on a native gel indicates that some dimer is present in the sample (lane 5 in Figure 6D). Even more remarkable is the presence of a slight amount of dimer for ERRα DBD-Pu, while the complex with the parent tff1 ERE/IR3 DNA only shows monomer on DNA (Figure 6B, lanes 6, 9), strongly suggesting that this is the critical region for the stabilization of the second subunit and the dimer formation. The SEC-MALLS of the ERRα DBD-Pu sample shows that the majority of the complexes is composed of monomer on DNA (Figure S7 in Supplementary Material), and no dimer band is seen after the SEC step (Figure 6D, lane 8), suggesting that the stabilization of the dimer is weak on this sequence. We tried to be even more conservative in changing the sequence by mutating only one bp in either tff1 ERE/IR3 DNA or embERRE/IR3. We chose to mutate the base pair at position (+7) from A/T to T/A for tff1 ERE/IR3 (IR3mutA7T) and from T/A to A/T for embERRE/IR3 (embmutT7A) (Figure 6A). The consequences of this mutation are that the complementary DNA stretch between positions (+5) and (+9) becomes GGAGG for IR3mutA7T and GGTAA for embmutT7A. This mutation has already remarkable effects on the dimer association, either slightly destabilizing the dimer for embERRE/IR3 (Figure 6B, lane 8 compared to lane 2 and Figure 6C) or promoting a slight dimer association for tff1 ERE/IR3 (Figure 6B, lane 10 compared to lane 9). Notice that the effect persists upon SEC-MALLS measurements and sample collection at the peak maximum (Figure 6D, lanes 6, 7). This change in the nature of the nucleotides, either by breaking or restoring the purine stretch observed in the complementary strand between positions (+5) and (+9) is critical. Notice that no dimer is seen to appear, when a TAA extension is present at the 5′end of the second half-site (3′IR3TAA) (on the complementary strand and shown in Figure 6B, lane 7). Altogether, the data from native polyacrylamide gels and SEC-MALLS suggest that small variations in the DNA sequence at the 3′-end are linked to the propensity of the DBD to dimerize. Thus, subtle variations in the sequence of this region lead to dramatic effects and support the hypothesis that dimerization of ERRα DBD on DNA is intrinsically linked to the nature of DNA sequence. Since DNA groove dimensions depend on the DNA sequence (44–46), we used a high-throughput, experimentally well validated method to predict the DNA shape features of the sequences considered in our studies (47). As Figure 7 indicates, the analysis shows large differences in the minor groove widths between the tff1 ERE/IR3 and embERRE/IR3 right in the region between positions (+5) and (+9). The embERRE/IR3 exhibits a pronounced minimum around the nucleotide at position (+6), while the tff1 ERE/IR3 behaves at contrary at this location with a peak maximum at positions (+6) and (+7). These observations suggest that a narrow minor groove between positions (+6) and (+9) is essential for the stabilization of ERRα DBD homodimer and that wide and shallow minor groove at this location is not favorable to dimer formation. Next, we repeated the DNA shape prediction for the series of mutant and natural sequences. As shown on Figure 7, the mutants embmutT7A and IR3mutA7T follow a quite similar profile with the expected trend, the former exhibiting a shallower minimum as compared to embERRE/IR3 and the latter a strong decrease in the minor groove width as compared to tff1 ERE/IR3. These predictions are consistent with the observations of the destabilization of the dimeric form for embmutT7A, or on the contrary of stabilization of a fraction of the complex into the dimeric form for IR3mutA7T. Remarkably, the DNA shape predictions for the various individual ERR binding sites considered in our studies are remarkably consistent with the observations of dimer versus monomer-only behavior (see Figure S8 in Supplementary Material).