OXER1 and RACK1-associated pathway: a promising drug target for breast cancer progression

Recent data indicate that receptor for activated C kinase 1 (RACK1) is a putative prognostic marker and drug target in breast cancer (BC). High RACK1 expression is negatively associated with overall survival, as it seems to promote BC progression. In tumors, RACK1 expression is controlled by a complex balance between glucocorticoids and androgens. Given the fact that androgens and androgenic derivatives can inhibit BC cell proliferation and migration, the role of androgen signaling in regulating RACK1 transcription in mammary tumors is of pivotal interest. Here, we provide evidence that nandrolone (19-nortosterone) inhibits BC cell proliferation and migration by antagonizing the PI3K/Akt/NF-κB signaling pathway, which eventually results in RACK1 downregulation. We also show that nandrolone impairs the PI3K/Akt/NF-κB signaling pathway and decreases RACK1 expression via binding to the membrane-bound receptor, oxoeicosanoid receptor 1 (OXER1). High levels of OXER1 are observed in several BC cell lines and correlate with RACK1 expression and poor prognosis. Our data provide evidence on the role played by the OXER1-dependent intracellular pathway in BC progression and shed light on the mechanisms underlying membrane-dependent androgen effects on RACK1 regulation. Besides the mechanistic relevance, the results of the study are of interest from a translational prospective. In fact, they identify a new and actionable pathway to be used for the design of innovative and rational therapeutic strategies in the context of the personalized treatment of BC. In addition, they draw attention on nandrolone-based compounds that lack hormonal activity as potential anti-tumor agents.


Introduction
Breast cancer (BC) is a heterogeneous disease due to variable histological subtypes and differences in response to therapy and clinical outcome 1 . In human BC cells, the receptor for activated C kinase 1 (RACK1) has been identified as a possible prognostic marker and drug target 2 due to its critical role in cancer cell migration and invasion [3][4][5][6] . Changes in RACK1 levels have been found to subvert physiological functions, leading to the development and maintenance of several BC hallmarks [6][7][8][9] . Indeed, data mining analyses (Chin, 130; Bertucci 266; Booser, 508) revealed that elevated RACK1 expression negatively correlates with overall survival (OS) 10 , thus indicating that RACK1 over-expression associates with poor clinical outcome 2 .
A complex hormonal balance between glucocorticoids and androgens has been demonstrated to control RACK1 expression in both the immune [11][12][13][14][15][16] and the cancer context 5,9 . Consequently, based on the hormone-related nature of many BC types, a deeper understanding of RACK1 transcriptional regulation is of pivotal interest 17 . Accumulating evidence suggests a role for androgen signaling in BC 18 . In this regard, literature data report that the adrenal hormone dehydroepiandrosterone (DHEA) and its endogenous androgenic derivative 7β-Hydroxyepiandrosterone (7β-OH-EpiA) exert a protective role against cancer by inhibiting cell proliferation and migration [19][20][21] . Notably, different testosterone derivatives have been demonstrated to inhibit cell proliferation, thus suggesting that the development of testosterone-based anticancer drugs, lacking for hormonal activity, may represent a major challenge in steroid medicinal chemistry 22,23 . Increasing evidence indicates that androstanes and their structural analogs display a versatile antiproliferative action against a broad variety of cancer cell lines, including prostate, breast, cervix, ovarian, leukemia, melanoma, colon, and gastric cancers [23][24][25][26][27][28][29] . As DHEA regulates RACK1 expression through an androgenic mechanism in the immune context [13][14][15][16] , we investigated whether the synthetic anabolic-androgenic steroid, nandrolone (19-nortestosterone) was able to modulate RACK1 expression in BC cells. Here we report that RACK1 is significantly downregulated by synthetic steroids that bind oxoeicosanoid receptor 1 (OXER1), a member of the G-protein-coupled receptors (GPCR) family involved in the biological action of the arachidonic acid metabolite 5-oxo-eicosatetraenoic acid (5-oxoETE) 30 . OXER1 is involved in both inflammation and oncogenesis, but its role and its significance in BC progression are just emerging 31 . Indeed, immunohistochemical analyses revealed that membrane staining for OXER1 is higher in tumor compared to non-cancerous tissues 32 . In accordance, BC TCGA data analysis revealed that high OXER1 expression correlated with estrogen-(ER) and progesterone receptors (PR) status, suggesting that OXER1 is a new and attractive target for pharmacological intervention. Here, we demonstrate that OXER1 silencing significantly inhibits BC cell proliferation and migration by disrupting PI3K/Akt/NF-κB pathway, thereby leading to RACK1 downregulation. Therefore, RACK1 transcriptional regulation and OXER1 function in BC cell proliferation and migration highlight the existence of a novel molecular mechanism that could provide new and relevant drug targets for BC treatment.
Compounds used for treatments were dissolved in DMSO (Sigma Aldrich Italia, Cas N°67-68-5, purity 99.9%) at concentration of 50 mM and frozen at −20°C in stock aliquots. Stocks were diluted at final concentrations in culture media at the time of use (final concentration of DMSO in the culture medium <0.1%). Control cells were treated with the same amount of DMSO.

Subcellular fractionation
In brief, 3.5 × 10 6 MCF7 were seeded in 100 mm dishes and treated for 24 h with 100 nM Nandrolone or Testosterone-BSA-FITC and cellular fractionation was conducted as described in detail in ref. 35 .

Western blot analysis
The expression of OXER1, AR, Akt, p-Akt, IkB-α, p-IkB-α, Lamin A/C, c-Rel, RACK1, β-actin, and β-tubulin in cell homogenates was assessed by Western blot analysis as described in detail in ref. 5 . After Western blot acquisition, bands optical analysis was performed with the ImageJ program (W. Rasband, Research Service Branch, National Institute of Mental Health, National Institutes of Health, Bethesda, MD and Laboratory for Optical and Computational Instrumentation, University of Wisconsin). Bands relative densities were expressed as arbitrary units and normalized over control sample run under the same conditions.
Immunofluorescence 5000 MDA-MB-231 or 10.000 MCF7 cells per well were seeded on sterile slides coated with poly-L-Lysine in 24 multi-well plates. Cells were fixed and stained following protocol from ref. 36 . Acquisition was performed with confocal laser scanning SP5 by "Centro Grandi Strumenti" -Pavia.

Molecular modeling and docking
The three-dimensional (3D) structure of OXER1 was obtained using the he GPCR-Sequence-Structure-Feature-Extractor (SSFE) web server (https://doi.org/ 10.1093/nar/gkx399). The GPCR-SSFE 2.0 allows a multi-template approach for the modeling of both transmembrane (TM) helices and extracellular and intracellular loops. The templates that were selected for the modeling of the TM helices of OXER1 are reported in supplementary Table 1.
Before running the molecular docking the OXER1 structural model was regularized with 5000 steps of steepest descent energy minimization using the AMBER 14 software (University of California, San Francisco, 2014).
The structures of nandrolone and testosterone have been prepared for the docking using LigPrep (LigPrep, Schrödinger, LLC, New York, NY, 2015), thus generating all the possible tautomers and protonation states in the pH range 6.0-8.0. The molecular docking search area was placed at the centre of the transmembrane helices facing the extracellular side. The grid has been generated using the grid generation tool of Glide (Glide, Schrödinger, LLC, New York, NY, 2015) with default settings. The Glide SP scoring function was used to run docking calculations and to score the predicted docking poses (Glide, Schrödinger, LLC, New York, NY, 2015).

Small Interference RNA and shRNA
Commercially available MISSION ® esiRNAs for AR (EHU025951) and OXER1 (EHU142961) were obtained from Sigma Aldrich and transfected with Lipofectamine ® RNAiMAX for knock-down experiments. In brief, 2 × 10 5 MDA-MB-231 cells or 3 × 10 5 MCF7 cells were seeded in 6 multi-well plates and, 48 h after transfection, were treated as described in figure legends.
The stable RACK1-silenced cell line (shRACK1) was obtained by lentiviral infection. The lentiviral particles were produced as described elsewhere 37 in Hek293T cells by a second-generation packaging plasmid system containing a GIPZ_GFP human RACK1 Lentiviral shRNA plasmid or a scramble vector (RHS4531-EG10399; GE-Dharmacon). In brief, Hek293T cells were transfected by Lipofectamine ® with GIPZ (scramble or shRACK), pPAX and pMDG vectors. After 48 h, cell medium was collected, filtered and centrifuged for 2 h at 70,000 × g. Viral particles were re-suspended and used to infect MCF7 cells, after virus titration. Infected cells were flow sorted for high positive GFP fluorescence (S3e Cell Sorter BIO-RAD). Silencing was monitored by Western blot analysis.

Scratch wound healing assay
Cells were seeded in a 6-well plate, grown to confluence and incubated in 5% DCC-FBS medium containing nandrolone (100 nM) or testosterone-BSA-FITC (100 nM). Scratch wound healing assays and image analysis were performed as described in refs. 5,19 .

Viability and cell proliferation
Cells were seeded in 96-multiwell culture plates at a concentration of 5000 cells/well in a final volume of 100 μL. Viability and proliferation assay was assessed by Cell Proliferation Kit I (MTT) (Sigma Aldrich) a colorimetric assay based on the reduction of MTT [3-(4,5dimethylthiazol-2yl)-2,5-diphenyl-tetrazolium bromide] and readings were made through Synergy HT microplate reader (Bio Tek Instruments, Milan, IT).
Cells were seeded and after 24 h (t = 0) they were treated according to concentrations and timing reported in detail in figure legends.

Colony assays
Cells were seeded in 6 multi-well plates at a concentration of 2000 cells/well and colony assay was performed as described in ref. 38 .

Cytofluorimetry
MCF7 cells were seeded at a concentration of 10 6 cells/ dish grown in complete DMEM and subsequently were starved for 24 h in DMEM without FBS. After 24 h of starvation, MCF7 cells were grown for 24 h in complete DMEM medium and then treated with 100 nM nandrolone for 24 h. Pellets were washed in ice-cold PBS, resuspended in ice-cold ethanol 70% and stored at −20°C overnight. Pellets were resuspended in ice-cold PBS with the RNase (100 µg/mL) addition and incubated at 37°C for 1 h. EDTA 5 mM and propidium iodide (100 µg/mL) were added to the suspension and the solution was incubated on ice and protected from light for 30 min before FACS acquisition (protocol adapted from ref. 19 ). FACS analysis was performed with CyFlow Space ® cytofluorometer (Sysmex Partec Italia S.R.L., Cornaredo, IT) and data were analyzed with the FlowMax ® software (Partec Inter AG, Freienbach, CH).

Statistical analysis
Statistical analyses were conducted with GraphPad Prism version 7 (GraphPad Software, San Diego, CA, USA). Statistical differences were determined by analysis of variance (ANOVA) followed, when significant, by an appropriate post hoc test, as indicated in the figure legends. In all reported statistical analyses, effects were designated as significant if the p value was < 0.05.

Androgens negatively regulate RACK1 expression by an independent AR mechanism
Since BC is mostly a hormone-related tumor 18 and RACK1 expression regulation is under control of a complex hormonal balance 12 , we investigated RACK1 involvement in BC progression and its expression following androgen treatment. To this purpose, we used the MCF7 cell line, a well-established model to study hormone response and resistance pathways, given its androgen receptor (AR), PR, and glucocorticoid receptor (GR) positive profile 39 . In accordance with the tight correlation demonstrated between RACK1 expression as well as cell proliferation and migration in BC [3][4][5]40 , cell growth is significantly increased in MCF7 cells stably overexpressing RACK1 (RWT) and completely abolished in RACK1-silenced cells (shRACK1), compared to controls (Suppl. Fig. S1A, B). In addition, RACK1 depletion significantly affects MCF7 migration (Suppl. Fig. S1C). Consistent with the literature data reporting DHEAmediated regulation of RACK1 expression and inhibition of cell proliferation/migration 19,20 , MCF7 cells treated with increasing concentrations of DHEA showed a significant reduction in both cell proliferation (Suppl. Fig.  S2A-B) and RACK1 expression (Suppl. Fig. S2C, D).
Considering that DHEA effects on RACK1 regulation and cell proliferation/migration are mediated by its endogenous androgenic derivatives 21 , MCF7 cells were treated with nandrolone, which is able to regulate RACK1 expression in different cellular contexts 9,12,15 and is not subjected to aromatase-mediated intracrinological androgen-to-estrogen conversion, unlike other androgenic derivatives 41 .
Nandrolone-mediated RACK1 transcriptional regulation was investigated in terms of RACK1 promoter activity, mRNA and protein levels, showing that 100 nM nandrolone strongly reduced RACK1 expression (Suppl. Fig. S2E-G).
To evaluate AR involvement in RACK1 expression, MCF7 cells were pre-treated for 1 h with flutamide 42 , a nonsteroidal antiandrogen antagonist for intracellular AR, in order to abolish nandrolone-reduced RACK1 expression. Our data showed that flutamide was not able to counteract nandrolone-induced RACK1 down-regulation as demonstrated by Δ1 luciferase activity, mRNA and protein results ( Fig. 1A-C), suggesting that nandrolone effect on RACK1 regulation is AR-independent. The ARindependent effect of nandrolone was confirmed through AR knockdown, suggesting the involvement of an extracellular membrane-bound AR (mAR) (Fig. 1D, E). Indeed, a large amount of evidence highlighted an alternative, membrane-initiated androgen mechanism that involves rapid signaling via specific kinases and induces a decrease in cell growth and migration by dynamic modulation of the cytoskeleton 32,43-47 .
Nandrolone downregulates RACK1 expression by antagonizing the PI3K/Akt/NF-κB pathway MCF7 cells treated with increasing concentration of wortmannin 48 and BAY 11-7085 49 showed that RACK1 was significantly downregulated with both treatments in a dose-concentration dependent manner. This demonstrates that RACK1 expression is regulated by the PI3K/ Akt/NF-κB pathway (Suppl. Fig. S3A, B), in line with our previous observations 34 .
To assess whether nandrolone-induced RACK1 downregulation was related to its PI3K/Akt/NF-κB antagonism and its effect was mAR-mediated, MCF7 cells were treated with 100 nM of testosterone, testosterone-BSA-FITC or nandrolone 50 . Only nandrolone and testosterone-BSA-FITC induced a comparable and significant decrease in Akt phosphorylation ( Fig. 2A), which correlated with a lower NF-κB activation as demonstrated by the reduction in IκBα phosphorylation (Fig. 2B) and c-Rel nuclear translocation (Fig. 2C). For this reason, we transiently transfected MCF7 cells with deletion mutants of a luciferase reporter construct (Δ1, Δ2, Δ7, and Δ11) involving three distinct c-Rel binding sites located inside the RACK1 promoter 34 . Δ7 had a basal activity comparable to Δ1, indicating that this c-Rel binding site could be pivotal for RACK1 promoter regulation (Fig. 2D). Accordingly, MCF7 transfected with Δ7 and subsequently treated with 10 µM BAY 11-7085 resulted in a significant reduction of luciferase activity (Fig. 2E). Similarly, nandrolone and testosterone-BSA-FITC significantly decreased Δ7 luciferase activity. This supports the idea that the previously demonstrated impairment of c-Rel nuclear translocation (Fig. 2C) is related to RACK1 down-regulation, as also confirmed by protein expression (Fig. 2F).
Nandrolone and testosterone-BSA-FITC mAR-mediated effects on RACK1 expression were confirmed in MDA-MB-231 cells (Fig. 2G-I; Suppl. Fig. S3C), a metastatic and hormone-independent BC cell line lacking ARs, which is widely used as a TNBC model in vitro 49 . Once again, the PI3K/Akt/NF-κB pathway is involved in RACK1 expression as demonstrated by treatment with PI3K and IκBα inhibitors (Suppl. Fig. S3D, E).
Altogether, these results indicate that nandrolone exerts a strong negative regulation on RACK1 expression by antagonizing the PI3K/Akt/NF-κB pathway.

Nandrolone decreased-RACK1 expression inhibits BC cell proliferation and migration
Testosterone-BSA-FITC or nandrolone treatments significantly inhibited MCF7 cell growth (Fig. 3A), in line with the RACK1 expression data (Fig. 2F) and the role played by the protein in cell proliferation (Suppl. Fig. S1A). Consistently, in MCF7 RWT cells, RACK1 over-expression rescued testosterone-BSA-FITC and particularly nandrolone treatments (Fig. 3A), confirming that cell growth inhibition is mediated by RACK1 downregulation. Nandrolone significantly increased the fraction of MCF7 cells in the G1 phase with a consequent decrease of cells in the G2/ M phase (Fig. 3B), demonstrating that its anti-proliferative effect is associated with cell cycle alterations. Moreover, nandrolone or testosterone-BSA-FITC caused a significant decrease in BC cell migration, as shown by the reduction of wound area closure compared to control cells (Fig. 3C, D).

OXER1 mediates the rapid antagonistic effect of androgens
Accumulating evidence indicates that androgens exert membrane-initiated actions leading to the modulation of cancer cell growth and metastasis. Three different G protein-coupled receptors (GPCRs) have been reported to mediate the membrane-initiated androgen effects: the G protein-coupled receptor family C group 6 member A (GPRC6A) 51-53 , the zinc transporter member 9 protein (ZIP9/SLC39A9) 54 , and the G protein-coupled OXER1 32,47 . Interestingly, testosterone acts agonistically on the first two receptors whereas it antagonizes OXER1-mediated actions 32,55 , which activate the PI3K/Akt and the focal adhesion kinase (FAK) signaling pathways 47 , promoting cell survival, adhesion and migration. Since OXER1 is expressed at high levels in MCF7 cells 32 and our data demonstrate that nandrolone and testosterone-BSA FITC exert an antagonistic effect on survival and migration, we investigated OXER1 involvement in the two processes. As expected, testosterone-BSA FITC binds to OXER1 as demonstrated by in silico molecular docking data (Fig. 4) according to literature immunofluorescence data 44 .

Nandrolone and testosterone binding mode on OXER1
The results so far discussed indicate a significant difference in the antagonist effects exerted by nandrolone and testosterone on OXER1. Indeed, despite the structural similarity of the two compounds, which differ only for a methyl group, nandrolone was endowed with a stronger activity. Thus, in order to provide a rationale for this difference in antagonist activity, we docked the two compounds in the binding pocket following modeling of the OXER1 three-dimensional (3D) structure (Fig. 4). The docking was performed using the induced fit protocol of the software Glide that confers flexibility to the lateral chains of the amino acids composing the binding pocket. In this way, we can obtain a more accurate prediction of the protein-ligand interaction mode.
Visual inspection of the ten docking poses obtained for each molecule (i.e., nandrolone e testosterone) clearly shows that nandrolone is always characterized by the same binding mode in all the best poses (Fig. 4B), while testosterone displays a larger variability in the binding mode (Fig. 4C). Furthermore, the absence of a methyl group on nandrolone C 10 slightly decreases its steric hindrance compared to testosterone, allowing a deeper accessibility to the OXER1 binding site (Fig. 4). However, the network of interactions established with OXER1 is conserved between the two compounds, with a main difference regarding the interaction with residue D288, which is located on the extracellular loop three (ELIII) (Fig. 4B, C) and connects the two transmembrane (TM) helices 6 and 7. The ability of nandrolone to enter the cavity more deeply allows it to interact with ELIII, while this does not happen in the case of testosterone.
Finally, the docking score, which represents an indirect measure of the binding energy of the three best poses ranked for each compound, shows that nandrolone has a better mean score than testosterone, −10 ± 0.4 vs −8.5 ± 0.5 (Fig. 4D). These data further confirm the experimental evidence reported above.
OXER1 role in cancer cell proliferation and migration through RACK1 expression regulation OXER1 role in BC progression was confirmed by esiRNA, a mixture of siRNA oligos for the silencing approach, as OXER1 knockdown induced a significant down-regulation of the PI3K/Akt/NF-κB pathway and a substantial decrease in RACK1 expression (Fig. 5A-E). In addition, MCF7 cells silenced for OXER1 showed a (see figure on previous page) Fig. 2 Nandrolone effect on PI3K/Akt/NF-κB pathway and RACK1-related expression. A-F Nandrolone reduces PI3K/Akt/NF-κB signaling pathway and RACK1-related expression. A-B MCF7 cells were treated for 24 h with 100 nM of testosterone (T), testosterone-BSA-FITC (T-BSA) or nandrolone (N). Vehicle controls were treated with DMSO < 0.1% (CTRL). The images are representative Western Blots. Results are shown as p-Akt/Akt (A), p-IκBα/IκBα (B) ratios ± SEM of four independent experiments. C Analysis of c-Rel expression in nucleus fraction of MCF7 cells treated for 24 h with 100 nM of testosterone-BSA-FITC (T-BSA) or nandrolone (N). Vehicle controls were treated with DMSO < 0.1% (CTRL). The image is a representative Western blot and results are shown as c-Rel/Lamin A/C ratio. Each value in the graph represents the mean±SEM of four independent experiments. A-C Statistical analysis was performed with Dunnett's multiple comparison test with *p < 0.05, **p < 0.01, ***p < 0.001. D MCF7 cells were transiently transfected with Δ1, Δ2, Δ7, and Δ11 RACK1 promoter constructs. Δ1 is the longest construct available and contains all three c-Rel sites. Δ2 luciferase reporter construct contains only the two cis c-Rel sites whereas Δ11 and Δ7 contain one of the two cis c-Rel sites as described in ref. 34 . The basal activity of these luciferase reporter constructs was evaluated through luciferase assay. Luciferase activity was expressed as RLU% and compared to Δ1 values assumed at 100%. Each bar represents the mean ± SEM of three independent experiments, in quadruplicate. Statistical analysis was performed with Dunnett's test, with **p < 0.01, ****p < 0.0001. E MCF7 cells were transiently transfected with Δ7 and subsequently were treated for 16 h with 100 nM of testosterone-BSA-FITC (T-BSA), nandrolone (N) or with 10 µM BAY 11-7085. Vehicle controls were treated with DMSO < 0.1% (CTRL). Luciferase activity was expressed as RLU% and compared to CTRL values assumed at 100%. Each bar represents the mean±SEM of three independent experiments, in triplicate. Statistical analysis was performed with Tukey's test, with **p < 0.01, ****p < 0.0001 vs CTRL and with § § p < 0.01 vs T-BSA. significant decline of cell growth and migration (Fig.  5F-H), suggesting a functional link between RACK1 and OXER1.
Similarly, OXER1 silencing in MDA-MB-231, a BC cell line that significantly expresses OXER1 32 , causes a significant down-regulation of RACK1 expression due to impairment of the PI3K/Akt/NF-κB pathway (Fig. 5I-M). In accordance with our previous data 5 , decreased expression of RACK1 reduced cell proliferation and migration ( Fig. 5N-P).
Altogether, these results demonstrate OXER1 involvement in MCF7 and MDA-MB-231 cells proliferation/ Fig. 3 Nandrolone significantly reduces cell proliferation and migration. A MTT assay on MCF7 wild-type cells or over-expressing RACK1 (RWT) was performed to evaluate cell proliferation following 100 nM of testosterone-BSA-FITC (T-BSA) or nandrolone treatment for different timings (1, 2, and 3 days). Cells were treated at t = 0 and vehicle controls with DMSO < 0.1% (CTRL). Value bar in the graph represents the mean ± SEM of four independent experiments. The analysis was performed by two-way ANOVA with Tukey's multiple comparisons test with *p < 0.05, **p < 0.01, ***p < 0.001 vs t = 0, # p < 0.05 vs T-BSA at t = 2, § § § p < 0.001 vs T-BSA at t = 3, ++p < 0.01 vs RWT at t = 1. B Effect of 100 nM Nandrolone treatment on the phases of the cell cycle. MCF7 cells were grown in complete medium and starved for 24 h. After 24 h of starvation, MCF7 cells were grown for 24 h in complete medium and then treated with DMSO < 0.1% (CTRL) or with 100 nM Nandrolone (N100 nM) for 24 h. The percentage of cells in each phase of the cell cycle was determined by flow cytometry after 24 h. The results are shown as the percentage of cells in each phase of the cell cycle. Data were expressed as mean ± SEM. from three independent experiments. The analysis was performed with Bonferroni's multiple comparisons test, with **p < 0.01 vs CTRL G1, § § p < 0.01 vs CTRL G2. C-D Effect of testosterone-BSA-FITC or nandrolone on MCF7 cells migration by scratch wound healing assay, performed as described in materials and methods. MCF7 cells were cultured with DMSO < 0.1% (CTRL) or with 100 nM of T-BSA or nandrolone and subsequently migration was analyzed at 24 h. C The image is a representative result. D Value bars in the graph represents the mean ± SEM of three independent experiments, in duplicate. The analysis was performed by two-way ANOVA with Tukey's multiple comparisons test with *p < 0.05, ***p < 0.01 vs CTRL. migration and provide insights into the underlying signaling pathway.

OXER1 and RACK1 mRNA expression in mammary tumors
Our data support the idea that the OXER1/RACK1 pathway plays an oncogenic role in BC and they demonstrate that androgens downregulate the two proteins in BC cell lines. Thus, we evaluated the expression levels of OXER1 and RACK1 transcripts in estrogenreceptor-positive (ER + ) and estrogen-receptor-negative (ER − ) mammary tumors using the RNA-sequencing data available in the TCGA database. The data obtained were correlated with the copy number variation (CNV) and the mutation status of the two corresponding genes. The levels of OXER1 mRNA are significantly lower in ER + compared to ERmammary tumors (Fig. 6A). This is likely to be due to the fact that the genome of ER + tumors shows a high frequency of shallow deletions of the OXER1 gene. In contrast, the RACK1 mRNA expression levels do not show statistically significant differences in ER + and ER − cases (Fig. 6B). We performed further quantitative correlation studies between the expression levels of OXER1/RACK1 mRNAs and the following parameters: age of diagnosis, ER-positivity, PR-positivity, HER2-positivity, incidence of relapse (new tumor event) and tumor stage (Fig. 6C). As expected, OXER1 mRNA levels show a significant inverse correlation with ER-positivity and PRpositivity. More importantly, the transcript amounts are inversely correlated with the age of diagnosis. In other words, OXER1 mRNA is over-expressed in BC diagnosed at relatively young age, which is consistent with the idea that a large fraction of these cases are characterized by a triple-negative and ER − phenotype. Once again, similar correlations are not observed in the case of RACK1 (Fig.  6D). Finally, the amounts of OXER1 and RACK1 mRNAs expressed in single tumors show a low but significant r correlation-value, indicating the existence of a weak coregulation (Suppl. Fig. S4).
Given the potential role of OXER1 and RACK1 in BC progression, we evaluated the possibility that expression of the two corresponding mRNAs represent a useful predictive marker of survival for specific subtypes of mammary tumors. Hence, we determined possible correlations between OS and the levels of OXER1/RACK1 mRNAs in the ER + , HER2 + and triple-negative (TN) BC subgroups (Suppl. Fig S5). To conduct these analyses we divided the TCGA breast cancers in OXER1-high or RACK1-high and OXER1-low or RACK1-low tumors according to a threshold value represented by the median expression levels of the two transcripts (Suppl. Fig. S5). The results of this analysis demonstrate that there is no significant difference in the OS of OXER1-high and OXER1-low tumors as well as RACK1-high and RACK1low tumors, regardless of the mammary tumor phenotype. Statistical analysis was performed with Student's t-test, with *p < 0.05, **p < 0.01. F-H Cellular proliferation and migration in MCF7 OXER1 silenced and unsilenced cells was performed as described in "Materials and methods". F MTT assay on MCF7 silenced and unsilenced cells was performed to evaluate cell proliferation for different timings (1, 3 and 5 days). Value bar in the graph represents the mean ± SEM of three independent experiments. Statistical analysis was performed by two-way ANOVA with Tukey's multiple comparisons test with **p < 0.01, ***p < 0.001 vs CTRL at t = 1, ## p < 0.01 vs CTRL at t = 3, § § § § p < 0.0001 vs CTRL at t = 5. G The image is a representative result. H Value bar in the graph represents the mean ± SEM of three independent experiments, in duplicate of wound healing area in OXER1 silenced and unsilenced MCF7 cells. The analysis was performed by two-way ANOVA with Tukey's multiple comparisons test with ***p < 0.001 vs CTRL at t = 0, § § § p < 0.001 vs CTRL at t = 24 h. I-M MDA-MB-231 cells were silenced for 48 h with OXER1 siRNA to investigate OXER1-related PI3K/Akt pathway. I The image is a representative Western Blot. J-M Results are shown as OXER1/β-actin (J), p-Akt/Akt (K), p-IκBα/IκBα (L) and RACK1/β-tubulin (M) ratios ± SEM of three independent experiments. Statistical analysis was performed with Student's t-test, with *p < 0.05, **p < 0.01. N-P Cellular proliferation and migration in MDA-MB-231 OXER1 silenced and unsilenced cells was performed as described in materials and methods. N MTT assay on MDA-MB-231 silenced and unsilenced cells was performed to evaluate cell proliferation for different timings (1, 3, and 5 days). Value bars in the graph represents the mean ± SEM of three independent experiments. Statistical analysis was performed with by two-way ANOVA with Tukey's multiple comparisons test with **p < 0.01, ****p < 0.0001 vs CTRL at t = 1, ## p < 0.01 vs CTRL at t = 3, § § § § p < 0.0001 vs CTRL at t = 5. O The image is a representative result. P Value bar in the graph represents the mean ± SEM of three independent experiments. The analysis was performed by two-way ANOVA with Tukey's multiple comparisons test with ****p < 0.0001 vs CTRL at t = 0, § § § p < 0.001 vs CTRL at t = 24 h.
The results are validated by comparison of the OS observed in tumors belonging to the first and last quartile in terms of both OXER1 and RACK1 mRNA expression. Thus, neither OXER1 nor RACK1 mRNAs seem to be useful predictive biomarkers for BC progression and lethality. However, a wider panel of patients needs to be analyzed to confirm this preliminary observation, especially in TNBC, and immunohistochemical analysis need also to be carried out. Indeed, a correlation with RACK1 expression, clinical state and histological grade was found in primary breast carcinomas [2][3][4] , providing evidence that RACK1 positive staining is related to a worse clinical outcome. RACK1 was also found to significantly discriminate between healthy controls and early stage BC patients 56-58 thereby suggesting RACK1 as a possible superior predictor of BC prognosis and an independent prognosis related factor [2][3][4] . However, studies to evaluate RACK1 as a BC biomarker are still in their infancy and much validation work remains to be done 6 .

Discussion
BC is a hormone-dependent disease that encompasses biologically and clinically different tumor types 59 . The role of androgens in BC development and progression is still a matter of debate 1 , although androstanes and their structural analogs are endowed with significant antiproliferative activity against different BC cells 23,60 . In this context, we demonstrated that nandrolone strongly impairs MCF7 and MDA-MB-231 cell proliferation and migration through an AR-independent mechanism since, in presence of the well-known AR inhibitor flutamide 42 , it is still able to downregulate RACK1 expression. This is in accordance with emerging evidence that mAR belonging to GPCR family are reported and proposed to mediate membrane-initiated androgen effects 32,51-55 . Our data provide the first evidence that nandrolone effect is due to an AR-independent mechanism, which involves downregulation of RACK1, a protein controlling BC progression and positively regulated by the PI3K/Akt/NF-κB pathway [3][4][5] . We provided evidence that testosterone-BSA-FITC and nandrolone antagonize the PI3K/Akt/NF-κB pathway leading to a significant decrease of c-Rel nuclear translocation through a strong decline in RACK1 expression, which inhibits cell proliferation and migration (Fig. 7). Although NF-κB plays an important role in normal mammary gland morphogenesis, its increased levels are commonly observed in BCs as a consequence of the constitutive expression of the NF-κB subunits (e.g., c-Rel, p65 and p50) [61][62][63] . Indeed, stimulation of the NF-κB signaling pathway promotes angiogenic neovascularization, epithelial-to-mesenchymal transition (EMT) and increases cancer cell stemness, leading to chemoresistance, radioresistance and endocrine resistance. By converse, NF-κB inhibition increases the sensitivity of cancer cells to chemo-and radio-therapy and consequently disease-free survival in BC patients 64 .
AR-independent nandrolone activity was confirmed in the MDA-MB-231 cell line, where androgens have been reported to trigger membrane-initiated steroid signaling (MISS) and gene transcription 50 . Many of the genes regulated by testosterone-BSA control the activity of DNA binding inhibitors, exert dominant negative effects on the helix-loop-helix proteins pathways and related transcription factors, preventing cancer cell-growth/survival and angiogenesis 50,65 . In addition, testosterone-BSA decreases the levels of cytoskeletal proteins, cytokines, inflammation-related molecules and their correlated-pathways, including several interleukins (IL2, IL4, IL6, and IL12) and TNF-related molecules. All this suggests modulation of a variety of cytokines, chemokines and growth factors, which are produced in the local tumor environment and promote tumor progression. In addition, testosterone-BSA treatment induces pro-apoptotic and anti-proliferative effects in both hormone-responsive or resistant BC cells 50 . Finally, literature indicates that testosterone-BSA and testosterone modify the expression of largely different sets of genes, in line with the idea that the two compounds act via an AR-independent and an AR-dependent mechanism, respectively 50 . This is exactly what we observe in MCF7 cells where the anti-tumor Fig. 7 Schematic representation of nandrolone-antagonizing effects on OXER1-initiated pathwaw in BC cells. The binding of nandrolone to OXER1 antagonizes the PI3K/Akt/NF-κB pathway leading to a significant decrease of c-Rel nuclear translocation through a strong decline in RACK1expression, which inhibits cell proliferation and mi-initiated PI3K/Akt/ NF-κB pathway and RACK1 expression in BC cells (see text for more details). action of testosterone-BSA is OXER1 and RACK1 dependent.
Many of the non-classical and cell surface dependent actions exerted by androgens are mediated by novel mARs, i.e., GPCRC6A, ZIP9/SLC39A9 and OXER1. Testosterone acts as an agonist of ZIP9/SLC39A9 and GPRC6A, although the latter protein is a non-specific membrane AR 32,53 . On the contrary, testosterone is an OXER1 antagonist. As OXER1 inhibits some of the major signaling pathways (FAK, PI3K/Akt, p38α) controlling cell growth, is likely to represent a new mediator of the anti-proliferative and anti-migratory actions exerted by androgens in cancer cells 47,50 . Indeed, our data demonstrate that antagonistic effects on the OXER1 receptor mediate RACK1 downregulation induced by testosterone-BSA and nandrolone. This results in a significant reduction of BC cell proliferation and migration, which is known to be promoted by the interaction between FAK and RACK1 4,6,38,66 . Hence, we provide evidence that OXER1 is a determinant of BC cell proliferation and migration in both luminal BC subtype and TNBC models via activation of the PI3K/ Akt/NF-κB pathway. In particular, our findings provide new insights on RACK1 transcriptional regulation in our TNBC cell model. In accordance with our literature data reporting in RACK1 promoter region the presence of a GRE (glucocorticoid response element) site close to the c-Rel site here investigated 5,11 and the possible synergism between GR and NF-κB in BC 67 , we hypothesize that in MDA-MB-231 cells NF-κB and GR could work synergistically for RACK1 transcriptional regulation, considering that we previously demonstrated GR involvement in RACK1 regulation in this same TNBC model 5 .
Interestingly, NF-κB/GR cross-talk promotes BC development/progression and it has been suggested as a potential target for BC therapy 67 .
The data provided in the present study indicate that OXER1 represents a novel link between androgens and their AR-independent action. In addition, they suggest that androgenic molecules lacking a classic AR-dependent profile and characterized by effects mediated by membrane ARs (e.g., nandrolone) can be exploited to reduce BC cell proliferation and migration. In fact, our molecular docking data reveal that nandrolone is a strong OXER1 antagonist, which suggests that OXER1 is a promising molecular target of androgen derivatives. With respect to this, it is worthwhile mentioning that the literature data indicate significant OXER1 expression differences between non-cancerous and cancerous tissues 32 . Interestingly, our study shows a significant inverse correlation between OXER1 mRNA levels and ER-positivity, PR-positivity or age of diagnosis in BC. These last observations are consistent with the idea that a large fraction of the cases defined by high OXER1 expression is characterized by a triple-negative and ER − phenotype. However, in prostate cancer (PC), OXER1 transcripts are lower in tumor than in normal tissue, pointing out that epigenetic elements may also need to be considered in the expression of OXER1 in PC 47 .
In conclusion, our data support the idea that androgen derivatives tailored to antagonize OXER1 activation pathway may represent a promising and rational agents for the personalized treatment of TNBC.