Codon adaptation by synonymous mutations impacts the functional properties of the estrogen receptor‐alpha protein in breast cancer cells

Oestrogen receptor‐alpha (ERα) positivity is intimately associated with the development of hormone‐dependent breast cancers. A major challenge in the treatment of these cancers is to understand and overcome the mechanisms of endocrine resistance. Recently, two distinct translation programmes using specific transfer RNA (tRNA) repertoires and codon usage frequencies were evidenced during cell proliferation and differentiation. Considering the phenotype switch of cancer cells to more proliferating and less‐differentiated states, we can speculate that the changes in the tRNA pool and codon usage that likely occur make the ERα coding sequence no longer adapted, impacting translational rate, co‐translational folding and the resulting functional properties of the protein. To verify this hypothesis, we generated an ERα synonymous coding sequence whose codon usage was optimized to the frequencies observed in genes expressed specifically in proliferating cells and then investigated the functional properties of the encoded receptor. We demonstrate that such a codon adaptation restores ERα activities to levels observed in differentiated cells, including: (a) an enhanced contribution exerted by transactivation function 1 (AF1) in ERα transcriptional activity; (b) enhanced interactions with nuclear receptor corepressor 1 and 2 [NCoR1 and NCoR2 (also known as SMRT) respectively], promoting repressive capability; and (c) reduced interactions with SRC proto‐oncogene, non‐receptor tyrosine kinase (Src) and phosphoinositide 3‐kinase (PI3K) p85 kinases, inhibiting MAPK and AKT signalling pathway.

Oestrogen receptor-alpha (ERa) positivity is intimately associated with the development of hormone-dependent breast cancers. A major challenge in the treatment of these cancers is to understand and overcome the mechanisms of endocrine resistance. Recently, two distinct translation programmes using specific transfer RNA (tRNA) repertoires and codon usage frequencies were evidenced during cell proliferation and differentiation. Considering the phenotype switch of cancer cells to more proliferating and less-differentiated states, we can speculate that the changes in the tRNA pool and codon usage that likely occur make the ERa coding sequence no longer adapted, impacting translational rate, co-translational folding and the resulting functional properties of the protein. To verify this hypothesis, we generated an ERa synonymous coding sequence whose codon usage was optimized to the frequencies observed in genes expressed specifically in proliferating cells and then investigated the functional properties of the encoded receptor. We demonstrate that such a codon adaptation restores ERa activities to levels observed in differentiated cells, including: (a) an enhanced contribution exerted by transactivation function 1 (AF1) in ERa transcriptional activity; (b) enhanced interactions with nuclear receptor corepressor 1 and 2 [NCoR1 and NCoR2 (also known as SMRT) respectively], promoting repressive capability; and (c) reduced interactions with SRC proto-oncogene, nonreceptor tyrosine kinase (Src) and phosphoinositide 3-kinase (PI3K) p85 kinases, inhibiting MAPK and AKT signalling pathway.

Introduction
As the main mediator of oestrogens, the oestrogen receptor-alpha (ERa) is involved in a variety of physiological processes ranging from the establishment and maintenance of the female sexual differentiation patterns to cardiovascular and neuronal systems, to liver, fat and bone metabolisms [1,2]. ERa influences also several pathological processes of which the most common is breast cancer [3,4]. A majority of studies show that ERa is absent or expressed at very low level in breast cancer stem cells and instead appears in breast cancers derived from luminal cells [5,6]. Approximatively 70% of diagnosed breast cancers express ERa whose activation by oestrogens favours the proliferation of breast cancer cells. For these cancers, the therapy of choice is endocrine therapy, which aims to deprive the tumour of oestrogens or directly block ERa activity, respectively, through the use of aromatase inhibitors or antioestrogens such as the tamoxifen [7,8]. However, endocrine resistance occurs in 30% of cases due to a hormonal escape, with cells continuing to proliferate without oestrogenic stimulation and becoming resistant to endocrine therapy. Breast tumours then switch to less-differentiated phenotypes and undergo invasive and metastatic processes associated with unfavourable vital prognosis [7].
ERa belongs to the nuclear receptor superfamily, and like all members, it is primarily a transcription factor that exhibits genomic activity through the regulation of gene expression. ERa transcriptional activity is mediated by two transactivation functions (AF): AF1, localized in the N-terminal part of the protein, and AF2 in the C-terminal ligand-binding domain. These two functional domains present interface surfaces that allow the recruitment of coactivators or corepressors in an ordered, cyclical and combinatorial manner [9,10]. Both AFs do not participate equally in ERa genomic activity, and it has been showed that their contribution depends on the cell differentiation stage [11]. The loss of AF1 activity in dedifferentiated cells has notably been linked to the loss of cell-cell junctions, which particularly occurs during epithelial mesenchymal transition (EMT) [11,12]. In addition to its genomic activity, ERa is also known to modulate non-genomic membrane signalling pathways through a protein pool located at the cytoplasm and/or plasma membrane and capable of interacting with several kinases such as SRC proto-oncogene, non-receptor tyrosine kinase (Src) and phosphoinositide 3-kinase (PI3K) p85 kinases [13]. Activation of the downstream mitogen-activated kinase (MAPK) and protein kinase B (AKT) signalling pathways ultimately leads to the stimulation of cell proliferation and survival [14].
Several mechanisms are known to promote endocrine resistance. These include ER-activating mutations, imbalance between coactivators and corepressors expression or overactivation of growth factors signalling pathways [15]. Endocrine resistance is generally accompanied by the epithelial-mesenchymal transition of breast cancer cells and functional modifications of ERa [16,17]. We recently demonstrated that, during EMT, the activation of the actin/MRTFA pathway results in functional modifications of ERa due to the impairment of its localization and interactions with partners. These mechanisms ultimately lead to a decrease of its transcriptional activity and an increase of its non-genomic activity, promoting hormonal escape [17,18]. Such modifications in ERa activity are likely to result from changes in environment and ERa partners as mentioned above but may also be due to protein conformational changes that have an impact on its functional activity. Indeed, ERa is a protein whose multifunctional properties are largely regulated by important and varied allosteric changes in the protein that are provoked either by ligand binding, DNA response element recognition or partner interactions [19][20][21][22]. However, other more upstream mechanisms could also be involved in conformational changes of ERa.
Due to the degeneration of the genetic code, amino acids can be encoded by several codons. These synonymous codons are not equally used in coding sequences, leading to a codon usage bias. A correlation between preferred codons and more abundant transfer RNAs (tRNAs) was observed, suggesting that tRNAs pools and codon usage are essential players in the regulation of gene expression. They may differ depending on the cell type, cell cycle phase or differentiation status of the cell [23,24]. For instance, Gingold et al. [24] demonstrated in 2014 a strong dependence between tRNA selection and abundance and the transcriptional programmes involved during cell proliferation and differentiation. In particular, the tRNA pool was correlated to codon usage bias in genes specifically expressed in differentiated or proliferating cells improving expression of cell state specific genes. More recently, specific expression of tRNAs in metastatic breast cancer cells that may promote tumour progression has been observed [25]. In fact, the availability of tRNA appears to be a key parameter impacting translation speed, and translation speed modification can alter the co-translational folding of some proteins [26][27][28][29]. Concerning ERa, previous work notably suggested that its conformation would be sensitive to the cellular environment and codon composition [30,31]. It should be recalled that in healthy breast tissues, ERa is expressed in differentiated cells that do not or rarely proliferate, whereas in ERa-positive breast cancers, it is often expressed in proliferative cells [32]. A conformational change in ERa during tumour transformation or EMT could, therefore, be induced by a change in tRNA abundance and codon usage, altering the translation rate and final conformation of the receptor.
In the present study, we show that ERa codon usage fits the codon usage of genes specifically expressed in differentiated cells. We generated an oestrogen receptor synonymous sequence to adapt the frequency of ERa codon usage to that observed in genes specifically expressed in proliferative cells, and then investigated the functional consequence of these synonymous mutations on ERa activity. We demonstrate that these synonymous mutations improve the AF1 transcriptional activity of ERa and alter ERa interactions with its partners by reducing its interaction with kinases and promoting the formation of complexes with corepressors. Finally, the expression of the synonymous mutant in MCF7 cell line abolishes the anti-apoptotic effect of oestradiol.

Plasmids and transient transfections
The ERE (ERa-responsive element)-tk (thymidine kinase gene promoter)-Luc (luciferase), C3 (complement C3 gene promoter)-Luc, AP1 (activating protein 1 gene promoter)-Luc, SP1 (specific protein 1 gene promoter)-Luc and pTAL-Luc reporter genes as well as the internal control CMV (cytomegalovirus promoter)-bgal (b-galactosidase), the pCR-ERa, pCR-ERa D79, pCR-ERa D173, pCR-GFP, pCR-SRC1 (Steroid receptor coactivator 1) and pCR-NCoR1 (nuclear receptor corepressor 1) expression vectors have been previously described [12,33,34]. pCR-ERa SYN-opt was produced synthetically (GeneArt, Life Technologies, ThermoFisher Scientific, Waltham, MA, USA). pCR-ERa SYN-opt D79 and pCR-ERa SYN-opt D173 expression vectors were generated by subcloning corresponding PCR products into pCR3.1 vector (Invitrogen). To measure ERa mRNA expression level in transient transfection experiments, 500 ng of expression vectors by well was transfected in 6-well plates together with 500 ng of pTAL-LUC used as an internal control. To assess ERa transcriptional activity on reporter genes, 50 ng of control (pCR3.1) or ERa expression vectors were co-transfected with 200 ng of reporter genes and 200 ng CMV-bgal (b-galactosidase) by well in 24-well plates. When pCR-MRTFA DN200, pCR-SRC1 or pCR-NCoR1 expression vectors were added to the DNA mix, 200 ng of plasmid was used. For immunofluorescence and proximity ligation assay, 250 ng of expression vectors was used by well in 24-well plates. For western blot analysis, 500 ng of expression vectors was transfected by well in 6-well plates. Transfections were carried out with JetPEI Ò (Polyplus transfection) according to the manufacturer's instructions. GFP-ERa WT or GFP-ERa SYN-opt expression vectors were generated from the pcDNA4/TO expression vector (T-Rex system, Invitrogen) by subcloning the corresponding PCR products.

Luciferase assay
Cells were plated in 24-well plates and incubated for 24 h before medium exchange for serum and steroid starvation. Cells were transfected as previously described and treated with ligands for 24 h before luciferase assay (Luciferase Assay System, Promega, Madison, WI, USA) [35]. b-galactosidase activity was determined by incubating two third of each lysate with 0.7 mgÁmL À1 O-nitrophenyl b-D-galactopyrano side (ONPG) in assay buffer (0.8 mM MgCl 2 , 35 mM b-mercaptoethanol) before measuring the absorbance at 415 nm with an iMark Microplate Absorbance Reader (Bio-Rad, Hercules, CA, USA). Luciferase activity reflecting ERa transcriptional activity was no rmalized to the b-galactosidase activity.

Immunofluorescence and proximity ligation assay (PLA)
Cells were plated on cover slides in 24-well plates and incubated for 24 h before medium exchange for serum and steroid starvation. When cell treatment was completed, phosphate buffered saline (PBS) containing 4% paraformaldehyde was used to fix the cells for 15 min. After PBS washing, cells were permeabilized in PBS containing 0.3% Triton X-100 for 15 min. After washing, an incubation with primary antibodies (1 : 1000) in PBS containing 3% FBS was performed overnight at 4°C. For immunofluorescence experiments, cells were incubated the next day with secondary antibodies (1 : 1000) in PBS-FBS for 2 h at room temperature. For PLA, the Duolink Ò Proximity Ligation Assay reagents from Sigma-Aldrich (Saint Louis, MO, USA) were used to detect ERa interactions with partners, according to manufacturer's instructions, as previously described [17]. Finally, the cover slides were mounted in Duolink Ò In Situ Mounting Medium with DAPI (Sigma-Aldrich). Images were obtained with an ApoTome Axio Z1 Imager microscope (Zeiss, G€ ottingen, Germany) and processed with AXIOVISION software (Zeiss). Fluorescent cells and PLA dots were analysed using IMAGEJ software.

Western blotting
Whole-cell extracts were lysed in RIPA buffer or directly prepared in 39 Laemmli buffer from subconfluent cells in 6-well plates. Subcellar fractionation of HepG2 cells was performed as previously described [17]. Following sonication, the protein extracts were denatured for 5 min at 95°C, separated on 10% SDS polyacrylamide gels and transferred to nitrocellulose membrane (Amersham Biosciences, Amersham, UK). The proteins were then probed with specific antibodies (1 : 1000) as previously described [11] and detected using the Substrat HRP Immobilon Western kit from Millipore (Burlington, MA, USA).

Proliferation assay
Cells were plated in 24-well plates and incubated for 24 h before medium exchange for serum and steroids starvation. Tetracycline induction and pharmacological treatments were applied for 6 days before cell trypsinization and counting.

Flow cytometry analysis
Cells were plated in 10-cm dishes and incubated for 24 h before medium exchange for serum and steroids starvation. After 48 h tetracycline induction and pharmacological treatments, cells were collected in PBS containing 30% IFA buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 4% FBS) and centrifuged for 10 min at 800 g, 4°C. Cells were fixed for 30 min on ice with 70% ethanol and incubated for 20 min in IFA buffer at 4°C. A 30-min incubation with 100 lgÁmL À1 RNAse A was then performed at 37°C, and 50 lgÁmL À1 propidium iodide (Sigma) was finally added for 10 min at 37°C. Finally, the cell cycle was then analysed using the FACSCalibur flow cytometer (BD Biosciences, Franklin lakes, HI, USA).

TUNEL staining
Apoptosis was determined by detecting DNA fragmentation using terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) staining. This was performed with an In Situ Cell Death Detection Kit, Fluorescein (Roche, Basle, Switzerland) according to the manufacturer's instructions, as previously described [34]. Images were obtained with an Apo-Tome Axio Z1 Imager microscope (Zeiss) and processed with AXIOVISION software. TUNEL-positive cells were analysed using IMAGEJ software.

Quantitative RT-PCR (RT-qPCR)
For transient transfection experiments, cells were plated in 6-well plates, and 24 h later the medium was replaced with medium containing 2.5% dextran/ charcoal-stripped FBS before transient transfection. RNA was extracted 48 h after transfection. For stable cell lines, cells were plated in 6-well plates and incubated for 24 h before medium exchange for serum and steroids starvation. RNA was extracted after 48 h tetracycline induction and 24 h pharmacological treatments. RNA extraction was performed using the Nucleospin RNA Plus kit (Macherey-Nagel, D€ uren, Germany) according to manufacturer's instruction. Quantitative RT-PCR was performed as previously described [34], with the primers sequences indicated in Table S1. Measurement of control (empty), ERa WT and ERa SYN-opt mRNA expression in transient transfection experiments was performed using primers designed in the common 3 0 UTR BGH sequence of the plasmid. Results were normalized to pTAL-Luc mRNA expression after subtraction of non-reverse transcribed products. mRNA expressions from stable cell lines were normalized to TBP mRNA level.

Protein degradation and limited digestion
Cells were plated in 6-well plates and incubated for 24 h before medium exchange for serum and steroids starvation. Cells were then treated 48 h with tetracycline. Analysis of ERa WT and SYN-opt protein degradation was performed at different times after cycloheximide treatment. For limited digestion experiments, cells were scrapped in PBS and centrifuged for 7 min at 800 g, 4°C after 48 h tetracycline induction. Cell pellet was sonicated and the lysate recovered after a 15-min centrifugation at 12 000 g, 4°C. After protein quantitation with the DC TM Protein Assay kit (Bio-Rad), a same amount of proteins was digested by 0.2-1.6 ngÁlL À1 trypsin or chymotrypsin (Sigma Aldrich) for 30 min at 30°C. Enzymatic reaction was stopped by adding 19 Laemmli buffer, and the samples were denatured for 5 min at 95°C before western blotting.

Bioinformatic analysis
Regulatory elements recognition and promoter and enhancer prediction were performed with Nsite and FPROM respectively [36,37]. Default parameters were used in both cases. Prediction of secondary structures of mRNA was performed with RNAfolder [38].

Statistical analysis
Statistical analyses were performed using Student's ttest. The values are provided as the mean AE standard error of the mean (SEM).

ERa codon usage is adapted to the translation programme of differentiated cells and less adapted to the translation programme of proliferating cells
In order to measure the adaptability of ERa codon usage to the dual translation programme in cellular proliferation and differentiation [24], we compared the frequency of codon usage deduced from the coding sequence of the human ERa gene with those of the two functional gene sets, 'M phase of mitotic cell cycle' and 'pattern specification process' (which we will call hereafter 'proliferation' and 'differentiation') from Gingold et al.'s study [24]. As expected, codon usage frequency for ERa correlated strongly with that of genes specifically expressed in differentiation (R 2 = 0.93) and weakly with that of genes specifically expressed in proliferation (R 2 = 0.46; Fig. 1). Since ERa is increasingly expressed in proliferating cells during mammary epithelial cell carcinogenesis, we generated, based on Gingold et al.'s study [24], an ERa complementary DNA (cDNA) in which codons were mutated to synonymous codons in order to optimize the codon usage of ERa to proliferative cells. Specifically, we chose a codon for each amino acid whose frequency of use in genes specifically expressed in proliferation approximates the frequency of the original codon observed in genes specifically expressed in differentiation. Codons with a similar frequency of use between 'proliferation' and 'differentiation' were maintained. Thus, a total of 82% of the codons (50 of 61) was changed into synonymous codons. This mutant, called ERa SYN-opt, has an identical amino acid sequence to ERa wildtype (WT), but a different DNA sequence ( Fig. 1 and Fig. S1).
Changes in the coding sequence can have major consequences in terms of transcriptional and translational activity by possibly creating or altering enhancer sequences, splicing factor binding sites and/or secondary structures of the messenger RNA (mRNA). Therefore, a predictive analysis of DNA regulatory sequences and mRNA secondary structure was first performed on ERa WT and SYN-opt coding sequences, showing in particular the presence of a TATA box in the coding sequence of ERa SYN-opt and a slightly more unstable secondary structure of the corresponding mRNA compared with ERa WT (Fig. S2). ERa WT and SYN-opt mRNA expression levels were then measured after transient transfection of the corresponding pCR expression vectors in HEK293 cells, selected for this study because of their high transfection efficiency. mRNA expression levels were normalized to Luc mRNA produced from a co-transfected pTAL-Luc reporter gene and expressed as fold changes from mRNA transcribed from empty pCR3.1 expression vector ( Fig. 2A). While the insertion of a coding sequence in the expression vector increases mRNA expression, no significant difference is observed between the expression level of ERa WT and SYN-opt mRNAs. Translation efficiency of these transcripts was next studied by qualitative and quantitative measuring the translated proteins through Western blot and immunofluorescence experiments. Results were normalized to the expression of the green florescent protein (GFP) produced from a co-transfected pCR-GFP plasmid. As illustrated in Fig. 2B-D, ERa SYN-opt protein is properly expressed, has the right molecular weight and is recognized by antibodies targeting both ERa N-and Cterminal domains. As ERa WT, it displays a major nuclear localization (Fig. 2D). However, its expression level is about 30-40% lower than that of ERa WT. With the ERa SYN-opt protein correctly produced both qualitatively and quantitively, the functional properties of the synonymous mutant were then studied.

Codon usage optimization of ERa coding sequence by synonymous mutations (ERa SYNopt) impacts the behaviours of ERa on ERE-, AP1-and SP1-driven reporter genes
ERa SYN-opt transcriptional activity was first characterized following transient transfection in HepG2 and 'pattern specification process' (differentiation) from the study of Gingold et al. [24]. Each dot corresponds to one codon. SYN-opt ERa complementary DNA (cDNA) was generated by selecting for each amino acid a codon whose frequency of use in genes specifically expressed in 'proliferation' approximated the frequency of the original codon observed in genes specifically expressed in 'differentiation.' Codons with a similar frequency of use between 'proliferation' and 'differentiation' were maintained (see Fig. S1).
cells. This cell line was chosen for its differentiated phenotype, providing an AF1-permissive context with no endogenous ERa expression, as previously described [11,12]. As in HEK293 cells, ERa SYN-opt protein is properly expressed in HepG2 and displays a major nuclear localization with a slight decrease in Interestingly, ERa SYN-opt exhibited a lower ligandindependent activity than ERa WT on the C3-Luc reporter gene (Fig. 3C). This activity was then induced nearly three-or five-fold in the presence of OHT or E2, respectively, whereas it was completely inhibited by ICI. Besides binding directly to ERE-driven genes, ERa can also regulate gene expression indirectly through protein-protein interactions with other transcriptional regulators such as activating protein 1 (AP1) and specific protein 1 (SP1). OHT and ICI were previously described as potent agonists on this ERa tethering pathway [39][40][41]. We, therefore, compared the transactivation activity of ERa WT and SYN-opt proteins on AP1-Luc and SP1-Luc reporter genes after treatment of cells with E2, OHT or ICI (Fig. 3C). As expected, ERa WT was capable to transactivate both reporter genes after an OHT or ICI treatment and E2 treatment was inefficient on the AP1-Luc. Surprisingly, ERa SYN-opt protein was unable to mediate a transcriptional activation of both reporter genes, regardless of the ligand used. Slight residual activity was observed with ICI treatment only. Thus, synonymous mutations abolish ERa transactivation on AP1-and SP1-driven reporter genes.
In conclusion, these results show that adapting the codon usage frequency of ERa by synonymous mutations impacts the ability of ERa to transactivate both classical (direct ERE binding) and non-classical (tethering) reporter target genes.

ERa SYN-opt protein enhances the transcriptional activity mediated by AF1
As previously mentioned, the transcriptional activity of ERa relies on two transactivation functions, AF1 and AF2 located in the N-and C-terminal domains of the protein respectively. Specifically active in differentiated cells [11,12], the AF1 has several different structural and functional units referred to as boxes 1-3 (Fig. 4A). We, therefore, evaluated the respective contribution exerted by both AF2 and AF1 boxes towards the transcriptional activity of ERa SYN-opt protein.
For that purpose, we generated for each ERa WT and SYN-opt proteins, two N-terminal deleted forms, ERa Δ79 (deletion of AF1 box 1) and ERa Δ173 (additional deletion of AF1 boxes 2/3). We then compared the transcriptional activity of these deleted forms with that of the full-length ERa in different cell lines with various differentiation stage and AF1 permissiveness [11,12,35]. A similar activity measured for full-length ERa and ERa Δ79 indicates the absence of an AF1 box 1 activity when a strict AF2-mediated transcriptional activity is revealed by the same activity for fulllength ERa and ERa Δ173. The expression of the different forms was first controlled by Western blotting (Fig. 4B). The full-length and deleted forms of ERa SYN-opt were generally slightly less expressed that those of ERa WT. In the strict AF1-permissive cell lines HepG2 and MCF7, ERa WT activity is mediated by both AF1 box 1 and box 2/3 as expected (Fig. 4C). Interestingly, ERa SYN-opt favours the contribution of AF1 box 1 in these cells, since the difference of activity between the full-length and Δ79 form is much higher in the context of the SYN-opt receptor as compared to the WT ERa. In MDA-MB-231 cells, ERa WT activity does not require the AF1 box 1 activity, while ERa SYN-opt activity remains primarily driven by this subdomain (Fig. 4C). Finally, whereas the transcriptional activity of ERa WT relies strictly on the AF2 function in the less-differentiated cell line SUM159PT, ERa SYN-opt additionally exhibits an AF1 boxes 2/3 activity, but AF1 box 1 activity remains quenched in these cells.
AF1 activity is also known to be regulated by growth factors such as the epithelial growth factor (EGF), through the phosphorylation of the serine residue at position 118 of ERa by the mitogen-activated protein kinase MAPK [42]. We, therefore, evaluated the impact of EGF treatment on the transcriptional activity of ERa WT and SYN-opt proteins. Results show that ERa SYN-opt is slightly more sensitive to EGF stimulation than ERa WT (Fig. 4D). Notably, the effect of EGF appears to affect more the AF1 box 1 activity in ERa SYN-opt protein when it depends more on box 2/3 activity in ERa WT protein.
Altogether, these results demonstrate that optimizing the codon usage frequency of ERa by synonymous mutations improves the transcriptional activity of AF1 function.

ERa SYN-opt protein improves tamoxifenmediated repression in the presence of NCoR1 corepressor
SERMs such as OHT act as either agonists or antagonists of ERa functions through the recruitment of coactivators or corepressors, respectively, in a tissue-, celland promoter-specific manner [43,44]. In an attempt to study the consequences of ERa codon optimization on the SERM activity of OHT, MCF7 was transiently transfected with either ERa WT or ERa SYN-opt expression vectors in the presence or absence of expression vectors for SRC1 (steroid receptor coactivator-1) or NCoR1 (nuclear receptor corepressor-1), and the activity of the reporter gene C3-Luc was then measured after treatment or not with E2 or OHT. We did not include ICI in these analyses since this molecule has no SERM activity on ERE-driven genes. To better visualize the expected repressive activities, the results of these assays, shown in Fig. 5, were expressed in log2 fold change (FC) relative to the basal transcriptional activity of the reporter gene. Overexpression of SRC1 similarly enhanced the agonist activity of OHT on ERa WT and SYN-opt proteins. In contrast, the overexpression of NCoR1 allowed only the ERa SYN-opt protein to behave as a transcriptional repressor in the presence of OHT, but also, interestingly, in the absence of ligand. These results indicate that optimizing the codon usage frequency of ERa by synonymous mutations enhances the ability of ERa to repress transcription in the presence of OHT, presumably through improved NCoR1 corepressor recruitment.

E2 anti-apoptotic effect is abolished in MCF7 expressing ERa SYN-opt protein, resulting in a weaker oestrogen-induced proliferation
To investigate the impact of ERa SYN-opt protein on hormone-dependent breast cancer cells, we generated MCF7 subclones with a tetracycline-inducible vector (D) HepG2 cells were transfected with the expression vectors previously described together with the C3-Luc and CMV-bgal reporter genes, and then treated with 10 nM E2 in the presence or the absence of 50 ngÁmL À1 of epidermal growth factor (EGF). Luciferase activity was normalized to b-galactosidase activity. Data correspond to the mean values AE SEM of three separate transfection experiments and are expressed as fold change from pCR3.1 control (*P-value < 0.05 and **P-value < 0.01, Student's t-test).
system to stably expressed ERa SYN-opt protein. We selected two clones, MCF7 ERa SYN-opt clones 1 and 2. MCF7 subclones expressing either the empty expression vector (MCF7 control) or the expression vector encoding ERa WT (MCF7 ERa WT) were also established in parallel and used as controls. ERa WT and SYN-opt proteins were fused to the GFP tag in their N-terminal part in order to discriminate artificially expressed forms of ERa from endogenous ERa in MCF7 cells. Tetracycline-induced expression of GFP-ERa WT and SYN-opt proteins was first controlled by western blot and immunofluorescence approaches (Fig. 6A,B). Their expression was predominantly localized in the nucleus, like endogenous ERa, with an approximately 1.5-fold higher level for GFP-ERa WT than for GFP-ERa SYN-opt protein. This difference is very likely due to changes in translational efficiency since GFP-ERa WT and SYN-opt mRNAs have similar expression levels and identical lengths with no detected splicing events in MCF7 subclones (Fig. S3). Interestingly, while the nuclear expression level of GFP-ERa WT varies greatly from one cell to another in accordance with what is observed with the endogenous ERa, GFP-ERa SYN-opt shows a very similar expression between the different cells (Fig. S3). Furthermore, nuclear expression of GFP-ERa SYN-opt remains stable in mitotic cells identified by histone 3 serine 10 (H3S10) phosphorylation staining whereas that of GFP-ERa WT drops sharply. We then assessed the impact of GFP-tagged protein expression on cell proliferation. For that purpose, MCF7 control, GFP-ERa WT and GFP-ERa SYN-opt cells were cultivated during 6 days in the presence or absence of tetracycline and under oestrogen or antioestrogen treatment, before counting the resulting cell population. As expected, E2 induced a threefold increase in cell numbers after 6 days of treatment when OHT and ICI slightly inhibited basal cell proliferation. For control and GFP-ERa WT MCF7 cells, tetracycline treatment had no major effect. In contrast, cell counting of GFP-ERa SYN-opt MCF7 subclones clearly showed a twofold reduction in cell number following tetracycline treatment, indicating that GFP-ERa SYN-opt expression disrupts E2-induced cell proliferation (Fig. 6C). Because similar results were obtained with GFP-ERa SYN-opt subclones 1 and 2 on cell proliferation, subsequent experiments were performed on subclone 1 only. Cell number results from the balance between cell cycle entry and apoptosis. These two parameters were, therefore, measured in order to determine the mechanism responsible for the limited E2-induced proliferation of breast cancer cells following GFP-ERa SYN-opt expression. First, the relative proportion of cells in each phase of the cell cycle was analysed by flow cytometry (Fig. 6D). Results show a similar dynamic in the different cell cycle phases whatever the MCF7 subclones used and tetracycline treatment status. Hence, as expected, as compared to untreated cells, E2 enhanced twofold the percentage of cells in S phase and OHT and ICI treatments slightly reduced this cell sub-population. The expression of GFP-ERa WT or GFP-ERaSYN-opt proteins had no impact on the proportion of MCF7 cells in the different phases of the cell cycle. The percentage of apoptotic cells was then determined by TUNEL staining in each condition. In control and GFP-ERa WT MCF7 cells, E2 treatment leads to a decrease in the number of apoptotic cells, whereas OHT and ICI treatments either had no strong effect or at most weakly stimulated apoptosis (Fig. 6E). Interestingly, when GFP-ERa SYN-opt protein was expressed in MCF7 cells, almost no reduction in the number of apoptotic cells was observed in presence of E2. Therefore, E2 loses its anti-apoptotic properties in the presence of ERa SYN-opt protein.
In conclusion, these results demonstrate that the low E2-induced proliferation of MCF7 cells expressing of ERa SYN-opt protein is not due to a defect in cell cycle entry but rather to an increased apoptosis.

Expression of ERa SYN-opt protein in MCF7 cells alters MAPK and AKT signalling pathway
In MCF7 cells, ERa is known to exhibit both genomic and non-genomic effects [14]. We, therefore, undertook the study of these two modes of action of ERa in the different MCF7 subclones. First, the genomic activities of the GFP-ERa forms were investigated by studying the consequences of their expression on the activity of endogenous E2-target genes. To do so, cells were treated or not for 48 h with tetracycline in the presence or absence of E2, OHT or ICI, and the expression of six main E2-regulated genes, TFF1, PGR, CXCL12, GREB1, AREG and EGR3, was assessed by RT-qPCR (Fig. 7A). Results show some variations in gene expression between the different subclones but which were not attributed to tetracycline-induced expression of the proteins of interest. Indeed, while the expression levels of the studied genes were induced by E2 and slightly repressed in the presence of OHT and ICI, the expression of either GFP-ERa WT or GFP-ERa SYN-opt proteins had clearly no impact on these regulations. These data suggest that ERa WT and SYNopt proteins expression in MCF-7 cells do not impact endogenous ERa activity on endogenous E2responsive genes.
Non-genomic activities of GFP-ERa forms were then investigated by examining the phosphorylation status of the extracellular signal-regulated kinases (ERK) and AKT by western blot in control, GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones after 24 and 48 h treatment with tetracycline alone or with E2. No significant alteration of ERK nor AKT phosphorylations was observed in control and GFP-ERa WT MCF7 cells following tetracycline and E2 treatments. On the contrary, in GFP-ERa SYN-opt MCF7 subclone, the addition of tetracycline resulted in the decreased phosphorylations of ERK or AKT in an E2-independent or E2-dependent manner respectively. Accordingly, in contrast to ERa WT, the ERa SYN-opt represses MAPK and AKT signalling pathway (Fig. 7B). Although non-genomic activity of ERa is primarily characterized by processes affecting components of signal transduction pathways, the subsequent cellular response may ultimately resume to genes expression regulations through the phosphorylation of transcription factors, coregulators or structural components of chromatin. Notably, histone H3 phosphorylation at serine 10 (H3S10p) is one of the main nucleosome responses to mitogen or growth factor signalling pathways through the activation of the mitogen and stress-activated kinase 1 (MSK1) [45]. Since steroid hormones including progestin and oestrogen have been shown to induce this signalling pathway [46,47], we investigated histone H3 phosphorylation status at serine 10 by immunofluorescence in the different MCF7 subclones treated or not 24 h with E2. Results showed that E2 increased H3S10 phosphorylation in all three MCF7 subclones, nevertheless with a lower basal level of H3S10p in GFP-ERa SYN-opt MCF7 cells than in control and GFP-ERa WT MCF7 cells (Fig. 7C). Taken together, these results lead to the conclusion that the repression of MAPK signalling pathway as occurs in GFP-ERa SYN-opt MCF7 cells converges at the chromatin level through repression of H3S10 phosphorylation.

ERa SYN-opt interacts more with corepressors and less with kinases than ERa WT in MCF7 cells
Because ERa SYN-opt protein was found to impact E2 signalling pathway in MCF7 cells, we extended our study to analyse the interaction of ERa with some of its main partners such as (a) members of the steroid coactivator P160 family, SRC1 (steroid receptor coactivator 1) and SRC3 (steroid receptor coactivator 3; AIB1); (b) the corepressors NCoR1 (nuclear receptor corepressor 1) and SMRT (NcoR2; nuclear receptor corepressor 2); and (c) the kinases Src and PI3K. After treating GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones with the different ligands E2, OHT or ICI, protein interactions were measured by proximity ligation assay (PLA) using antibodies targeting either the GFP-ERa protein through the GFP tag or Era's partners. Confirming our previous western blot analyses, quantification of the GFP fluorescence intensity shows that GFP-ERa SYN-opt expression is almost 1.5-fold lower than that of GFP-ERa WT and that a short-term treatment with either ligand has little impact on the protein expression level (Fig. 8C). As expected, the interaction profiles of GFP-ERa WT with the tested coactivators, corepressors and kinases were similar to those previously described with the endogenous receptor [17]. For instance, as previously reported, the interaction of ERa with P160 family coactivators is not further enhanced by E2 but strongly inhibited in the presence of ICI. Additionally, the interaction of ERa with NCoR1 and SMRT corepressors was almost abolished by a treatment with E2 as well as ICI but persistent after cells exposure to OHT. Finally, the ERa/kinase complex formation was not or only partially affected by ligands [17]. The similarity of GFP-ERa WT interaction profiles with those of the endogenous receptor indicates that the fusion of GFP to the N-terminal domain of ERa does not affect the interactions of the receptor with the tested partners. Interestingly, notable differences were observed when assessing the interaction of the GFP-ERaSYN-opt with its partners (Fig. 8B). First, as compared to the GFP-ERa WT, the GFP-ERaSYN-opt interacted less with SCR1 and SRC3 coactivators when in a basal state. Further changes in these interactions with coactivators occurred only when cells were treated with E2. The interactions of GFP-ERa WT and GFP-ERaSYN-opt with NCoR1 and SMRT corepressors and Src and PI3K kinases were similarly affected by the three tested ligands. However, as compared to the GFP-ERa WT, we detected more ERa/corepressor complexes and less ERa/kinase complexes in the case of the GFP-ERaSYN-opt protein. These differences in interaction intensity cannot fully be attributed to differences in the expression level of the corresponding proteins involved in the interaction (Fig. 8C,D). Compared with a same amount of ERa protein, the interaction of GFP-ERa SYN-opt with the two corepressors increases further and its interaction with Src and PI3K kinases still remains twofold weaker than that observed with GFP-ERa WT. Therefore, taken together, these results indicate that optimizing the codon usage frequency of ERa by synonymous mutations affects how ERa protein interacts with its partners, specifically favouring interaction with NCoR1 and SMRT corepressors over interactions with Src and PI3K kinases.

Limited digestions of ERa WT and ERa SYNopt proteins suggest conformational differences
Changes in the functional properties of ERa SYN-opt compared with ERa WT, both in transcriptional activity and partner interactions, strongly suggest differences in the conformation of the two proteins. In this continuity, a potential modification of its conformation was further investigated by first studying protein degradation kinetics and then by performing limited digestions. Protein degradation kinetic was analysed by western blot after blocking the protein synthesis in GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones with cycloheximide (Fig. 9A). Results shows a significant increase in the stability of the GFP-ERa SYN-opt as compared to the GFP-ERa WT protein, suggesting possible different protein conformations. Importantly, these results additionally indicate that the observed lower amounts of ERa SYN-opt protein produced in the cells (Figs 2, 3, 4, 6 and 8), as compared to GFP-ERa WT, are more related to a lower efficiency of mRNA translation than to an increased degradation of the protein. Next, we performed limited digestion experiments to assess for putative differing conformation of ERa WT and SYN-opt proteins more directly. For that purpose, increasing amounts of chymotrypsin or trypsin were incubated with the same quantity of cell lysate from GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones. The resulting protein extracts from of these limited digestions were then resolved by Western blotting using C-terminal ERa and GFP antibodies. As shown within Fig. 9B, the digestions of GFP-ERa WT and GFP-ERa SYN-opt proteins exhibited some differences in terms of digestion kinetics and weight of digested peptides. Indeed, the degradation kinetics of GFP-ERa SYN-opt by both digesting enzymes appeared indeed slower than that of GFP-ERa WT, in agreement with the conclusion rose from the previous experiment (Fig. 9A). This is particularly visible with the anti-GFP antibody which targets the N-terminal part of the GFP-fused ERa proteins. Additionally, several digested products of different weights were detected when comparing GFP-ERa WT to GFP-ERa SYN-opt MCF7 cell lysates (indicated by red arrows on Fig. 9B). Altogether, these results obtained by following the degradation kinetics of WT and SYN-opt ERa proteins and their profiles of limited digestions strongly suggest that these proteins have different conformations.

Discussion
The proliferation of ERa-positive breast cancer cells can be inhibited by using antioestrogens. Unfortunately, endocrine resistance often occurs, resulting in the oestrogen-independent development of cancerous cells. Many works aimed to understand the mechanisms leading to this hormonal escape and identified notably the role played by the imbalance of coregulators and growth factor signalling pathways. It has indeed been shown that the relative expressions of P160 family coactivators and NCoR1 and SMRT corepressors modulate the activity of the antioestrogen tamoxifen toward an ERa agonism or antagonism. This, for instance, explains why tamoxifen exerts an agonist effect in the uterus where the level of coactivators is increased, as compared to breast cancer cells where tamoxifen acts as an ERa antagonist [44]. The overexpression of coactivators such as SRC1 and the repression of corepressors such as NCoR1 in breast cancer cells have then been associated with tamoxifen resistance [48]. As intracellular kinases are able to activate ERa by phosphorylation, the overactivation of components of the MAPK or PI3K pathways plays a role in hormonal escape, by promoting the ligandindependent activation of ERa. This dysregulated activation results from mutations that are frequent in ERa-positive breast cancers, or the overexpression of growth factor receptors that activate such pathways, such as the EGF receptor (EGFR), which has been linked to endocrine resistance as well [15]. In healthy breast tissue, ERa is expressed in a small number of luminal cells, a subset of differentiated epithelial cells that are not proliferating, as assessed by a lack of co-staining with the proliferating marker Ki67 [49,50]. In this context, ERa exerts its oestrogeninduced mitogen activity by a paracrine mechanism, favouring the proliferation of surrounding ERanegative cells only [51]. In ERa-positive breast cancers, however, an increased proportion of ERa-expressing cells appears to be proliferating, due to a shift from a paracrine to an autocrine regulation of proliferation by ERa [49,52]. This change from quiescent to proliferating ERa-expressing cells during tumour transformation results in different modifications in cell activity that can lead to changes in functional properties of ERa explaining endocrine resistance.
As previous works highlighted tRNA pool variations between differentiated and proliferating cells [24], we hypothesized that this could occur in proliferative ERapositive breast cancer cells compared to non-cancerous ERa-expressing cells that remain quiescent. These modifications of tRNA pools could then alter the translation speed of ERa, resulting in an alteration of its cotranslational folding. This would ultimately lead to a modification of its conformation, explaining the dysregulation of ERa functions in breast cancer cells, participating to hormonal escape. This tRNA pool variation was correlated to a different codon usage bias in genes specifically expressed in differentiated or proliferating Fig. 9. Limited digestions of ERa WT and ERa SYN-opt proteins suggest conformational differences. (A) Time course of ERa WT and ERa SYN-opt degradation. GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones were treated with tetracycline for 48 h. Cells were then harvested at different times (0, 0.5, 1, 2, 3.5 and 5.5 h) after treatment with 20 lgÁmL À1 cycloheximide. ERa WT and ERa SYN-opt protein expression was analysed by western blot. b actin expression was used as a control. Histograms represent the mean AE SEM of ERa WT and ERa SYN-opt protein expressions normalized to b actin expression from four separate experiments. Results were expressed as a percentage of ERa expression at time 0 h. A Student's t-test was performed at each time point between ERa WT and ERa SYN-opt expression (*P-value < 0.05 and **P-value < 0.01). (B) Limited digestions of ERa WT and ERa SYN-opt proteins. Control, GFP-ERa WT and GFP-ERa SYN-opt MCF7 subclones were treated for 48 h with tetracycline. Limited digestions of cell lysates by increasing concentrations of trypsin or chymotrypsin (0.2-1.6 ngÁlL À1 ) were analysed by western blot using antibodies raised against ERa C-terminal region and GFP. Ponceau red staining was used to validate the homogeneity of protein amounts between each sample. * indicates non-specific band. Red arrows indicate several digested products of different weight between GFP-ERa WT and GFP-ERa SYN-opt MCF7 cell lysates. Blue arrows indicate several digested products with different degradation kinetic between GFP-ERa WT and GFP-ERa SYN-opt MCF7 cell lysates. The experiment was repeated twice with similar results. cells, suggesting an adequacy between codon usage and tRNA pool to enable the correct production of proteins in specific cell states [24].
We first evidenced that ERa encoding mRNA contains codons whose usage frequency is associated with differentiation programme. We, therefore, reasoned that ERa production and/or conformation would likely differ between dedifferentiated and proliferating cells. We stressed this hypothesis by generating a synonymous version of ERa by adapting the frequency of its codon usage to that observed in genes specifically expressed in proliferative cells, allowing us to investigate if this sequence optimization affects ERa functional properties in cancerous cells. Because this adaptation results in significant changes in the coding nucleotide sequence that can impact transcriptional and translation efficiency, we first verified that the synonymous version of ERa (ERa SYN-opt) was correctly produced both qualitatively and quantitatively. While the expression levels of both mRNAs were identical, we observed a slightly attenuated production of the ERa SYN-opt protein as compared to ERa WT, probably due to a reduced translational efficiency. Nevertheless, ERa SYN-opt remains abundant and functional. Interestingly, several of our results clearly demonstrate that codon adaptation to expression in proliferating cell indeed restores functional properties of ERa normally observed in differentiated cells: First, in MCF7, endogenous ERa nuclear expression is high in quiescent cells and is downregulated during cell cycle [53], resulting in heterogeneous expression of ERa in asynchronous cells. But the adaptation of ERa codons to those used in proliferating cells, as occurs in ERa SYN-opt, led to the homogenization of ERa expression in MCF7 cells with no downregulation in mitotic cells. Second, in differentiated cells, ERa exerts a genomic activity mainly driven by its AF1 function, especially its box 1 structural unit [11,12]. But optimizing the codon usage of ERa for proliferating cells improved the transcriptional activity mediated by AF1 box 1 even in dedifferentiated cells. Third, hormonedependent breast cancers are usually more differentiated and less invasive than the other types of breast cancers and generally respond well to endocrine therapy [7]. In particular, ERa transcriptional activity in breast cancer cells responding to endocrine therapy is repressed by the 4-hydroxytamoxifen through corepressor recruitment by ERa [43,54]. As mentioned previously, one mechanism of endocrine resistance is an imbalance in ERa recruitment between P160 family coactivators and NCoR1 and SMRT corepressors in favour of coactivators. We clearly show that the ability of OHT to repress transcription was increased after ERa codon adaptation, through improved NCoR1 and SMRT corepressor recruitment, which was further validated by PLA in MCF7 cells. Finally, nongenomic activity of ERa, which leads to the phosphorylation of intracellular kinases involved in MAPK or AKT signalling pathways, is mediated by the formation of complexes between ERa and kinases such as Src or PI3K. These complexes are increased in aggressive breast cancer cells, playing a role in endocrine resistance [17,55]. We demonstrate that codon optimization of ERa reduces the interaction of ERa with Src and PI3K kinases and represses MAPK and AKT signalling pathways when the protein is stably expressed in MCF7. All these changes in ERa activity are likely related to changes directly affecting ERa protein, because only the codon sequence of ERa has been modified. Since protein degradation kinetics and limited digestions with trypsin or chymotrypsin showed differences between WT and codon optimized forms of ERa, the changes in ERa activity are likely based on differences in ERa conformation.
The impact of ERa SYN-opt expression on MCF7 cell fate was also investigated in this study. While cells expressing this ERa SYN-opt appear to enter S-phase normally after oestrogenic stimulation, we found that these cells exhibited a lower E2-induced proliferation rate than those expressing ERa WT, due to the abolition of the anti-apoptotic effect of E2. There is no clear information regarding the apoptotic activity of ERa-positive epithelial cells in healthy breast tissue, but it is noteworthy that ERa-positive cells are few in number and represent about 10% of the total luminal epithelial cell population in human, which may suggest some sensitivity to apoptosis.
The observation of changes in ERa activity as a result of codon adaptation clearly reinforces the notion of distinct translation programmes. With specialized ribosomes [56], variation in the tRNA pool is one of the most effective ways to modify translation programmes [26,29]. It is known that tRNA function or expression is affected under stress conditions [57], through cell cycle [24,58], by cell-fate [24,59] or during tumorigenesis and cancer progression [25,60]. We do not know whether the cell lines used in this study have different tRNA repertoires or not. However, using YAMAT-seq method, Shigematsu et al. showed variations in the expression levels of mature tRNAs between three different subtypes of breast cancer cell lines, MCF7 (luminal A subtype), SK-BR-3 (HER2positive subtype) and BT-20 (triple negative subtype) [61]. We recently demonstrated that the activation and nuclear accumulation of MRTFA, a master regulator of EMT, in the luminal breast cancer cell line MCF7 induces a basal-like phenotype and remodels ERa functions by shifting its activity from nuclear genome regulation to extra-nuclear non-genomic signalling [17,18]. Further studies are now needed to verify whether this change in ER activity is accompanied by a change in the tRNA repertoire.

Conclusion
The results obtained in this study attest that modifying the codons used to produce ERa can lead to functional modifications, resulting from alterations of its conformation. ERa SYN-opt displays indeed a more repressive phenotype than ERa WT, meaning that adapting its codons in agreement with the frequencies observed in genes specifically expressed in proliferating cells enhanced the phenotype of ERa in differentiated epithelial cells. These data highlight a translational regulation of ERa conformation linked to its nucleotide sequence and cell state, likely due to the role played by codon usage in the translational process regulation. Such an observation could have important implications in understanding the biology of ERa from stem cells to differentiated cells and from healthy to cancerous tissues and will require further investigation.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Codon usage optimization of ERa coding sequence by synonymous mutations. Fig. S2. Regulatory elements, promoter/enhancer and secondary mRNA structure predictions from ERa WT and ERa SYN opt coding sequences. Fig. S3. GFP-ERa mRNA and protein expression in control, GFP-ERa WT and GFP-ERa SYN-op MCF7 subclones. Table S1. List of primers used to detect target genes by RT-qPCR.