MCF10a, MCF-7, T47D-1, ZR-75-1, HCC1428, BT-474, MDA-MB-361, were purchased by ATCC (USA), while EFM192C were purchased by DSMZ (Braunschweig, Germany) and maintained according to the manufacturer’s instructions. 17β-estradiol (E2), DMEM (with and without phenol red), and fetal calf serum were purchased from Sigma-Aldrich (St. Louis, MO). Bradford protein assay kit as well as anti-mouse and anti-rabbit secondary antibodies were obtained from Bio-Rad (Hercules, CA). Antibodies against ERα (F-10, mouse), pS2 (FL-84, rabbit), Bcl-2 (C2 mouse), cyclin D1 (H-295 rabbit), p62^SQSTM (D-3 mouse), and cathepsin D (H75 rabbit) were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA); anti-GART antibody (ab187716) was purchased by Abcam (Cambridge, UK). Anti-phospho ERα (Ser118, mouse) antibody was obtained from Cell Signaling; anti-vinculin (mouse) and anti-LC3 (mouse) antibodies were purchased from Sigma-Aldrich (St. Louis, MO, USA). Chemiluminescence reagent for Western blot was obtained from BioRad Laboratories (Hercules, CA, USA). Fulvestrant (i.e., ICI182,780) was purchased by Tocris (USA); 4OH-Tamoxifen, cycloheximide (CHX), lometrexol (LMX) and esiRNA library were purchased from Sigma-Aldrich (St. Louis, MO, USA). Palbociclib and abemaciclib were purchased by Selleck Chemicals (USA). PolarScreen™ ERα Competitor Assay Kit, Green (A15882) was acquired from Thermo Scientific. Tissue arrays were obtained by Biotechne (Minneapolis, MN, USA). All the other products were from Sigma-Aldrich. Analytical- or reagent-grade products were used without further purification. The identities of all the used cell lines were verified by STR analysis (BMR Genomics, Italy).

A fluorescence polarization (FP) assay was used to measure the binding affinity of lometrexol (LMX) and 17β-estradiol (E2) for recombinant ERα in vitro. The FP assay was performed using a PolarScreen™ ERα Competitor Assay Kit, Green (A15882, Thermo Scientific) as previously reported (24).

In-cell Western blot was used to measure ERα levels in MCF-7, and Y537S cell lines as previously described (17). The cells were treated with esiRNA targeting different metabolic proteins according to the reverse protocol detailed in (25) in quadruplicate for 48 hours. Fulvestrant (i.e., ICI – 100 nM) was used as the control for ERα degradation.

In-cell PI staining was used to measure DNA content in MCF10a, MCF-7, and Y537S cell lines. The experiments were carried out using the protocol previously described (17). The cells were treated with esiRNA targeting different metabolic proteins according to the reverse protocol detailed in (25) in quadruplicate for 48 hours. Taxol (1 µM) was used as the control for cell proliferation.

MCF-7 and Y537S cells were stably transfected with a plasmid containing an ERE-nanoluciferase (NLuc)-PEST reporter gene and measurement of NLuc-PEST expression (i.e., ERα transcriptional activity) was performed after 24 hours of compound administration as described (16, 26).

Cells were grown in DMEM with phenol red plus 10% fetal calf serum for 24 hours and then treated with the different compounds at the indicated doses for the indicated periods. Before E2 stimulation, cells were grown in DMEM without phenol red plus 10% charcoal-stripped fetal calf serum for 24 hours; lometrexol was added 48 hours before E2 administration. After treatment, cells were lysed in Yoss Yarden (YY) buffer (50 mM Hepes (pH 7.5), 10% glycerol, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA and 1 mM EGTA) plus protease and phosphatase inhibitors. Western blot analysis was performed by loading 20–30 µg of protein on SDS-gels. Gels were run, and the proteins were transferred to nitrocellulose membranes with a Turbo-Blot semidry transfer apparatus from Bio-Rad (Hercules, CA, USA). Immunoblotting was carried out by incubating the membranes with 5% milk or bovine serum albumin (60 min), followed by incubation overnight (o.n.) with the indicated antibodies. Secondary antibody incubation was continued for an additional 60 min. Bands were detected using a Chemidoc apparatus from Bio-Rad (Hercules, CA, USA).

Immunohistochemistry of tissue slides was performed as described in (27). Sections were incubated with the anti-GART antibody described above.

MCF-7 and Y537S cells were transfected with esiRNA against GART and the procedure was carried out using Lipofectamine RNAi Max (Thermo Fisher) as previously reported (25).

For growth curves, the xCELLigence DP system ACEA Biosciences, Inc. (San Diego, CA) Multi-E-Plate station was used to measure the time-dependent response to the indicated drugs by real-time cell analysis (RTCA), as previously reported (13, 16, 24, 28). Synergy studies and Tam growth curves were done using Crystal Violet staining as reported in (29). Next, the synergy index was calculated with Combenefit freeware software (24). Alginate-based and tumor spheroid cultures were performed and quantitated as previously reported (24).

The sequences for gene-specific forward and reverse primers were designed using the OligoPerfect Designer software program (Invitrogen, Carlsbad, CA, USA). The primers used for human ERα were 5’-GTGCCTGGCTAGAGATCCTG-3’ (forward) and 5’-AGAGACTTCAGGGTGCTGGA-3’ (reverse), and for human GAPDH were 5’-CGAGATCCCTCCAAAATCAA-3’ (forward) and 5’-TGTGGTCATGAGTCCTTCCA-3’ (reverse). Total RNA was extracted from the cells using TRIzol Reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s instructions. To determine gene expression levels, cDNA synthesis and qPCR were performed using the GoTaq 2-step RT-qPCR system (Promega, Madison, MA, USA), with an ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA), according to the manufacturer’s instructions. Each sample was tested in triplicates, the experiment was repeated twice, and gene expression was normalized to GAPDH mRNA levels.

Bromodeoxyuridine (BrdU) was added to the medium in the last 30 min of growth, and the cells were then fixed and permeabilized. Histones were dissociated with 2 M HCl as previously described (30). BrdU-positive cells were detected with anti-BrdU primary antibody diluted 1:100 (DAKO; Santa Clara, CA, USA) and Alexa488-conjugated anti-mouse antibody diluted 1:100 (Thermo Fisher Scientific; Waltham, MA, USA). Both antibodies were incubated with the cells for 1 h at room temperature in the dark. BrdU fluorescence was measured using a CytoFlex flow cytometer, and S-phase analysis was performed with CytExpert v 2.3 software (Beckman Coulter, Brea, CA, USA). All samples were counterstained with propidium iodide (PI) for DNA/BrdU bi-parametric analysis.

Statistical analysis was performed using the InStat version 8 software system (GraphPad Software Inc., San Diego, CA). Densitometric analyses were performed using the freeware software Image J by quantifying the band intensity of the protein of interest with respect to the relative loading control band (i.e., vinculin) intensity. The p-values and the used statistical test (i.e., either Student t-test or ANOVA Test) are given in figure captions.

To identify a potential novel target within metabolic proteins that could reduce ERα stability and cell proliferation in BC cells, we reasoned that such protein should be lethal for tumor cells while its inhibition must not significantly affect the proliferation of non-transformed cells. Therefore, we manually collected a library of siRNA directed against about 140 metabolic proteins belonging to different metabolic pathways (i.e., lipid, nucleotide, and amino acid metabolism) and tested it in non-transformed breast cells (i.e., MCF10a cells) (20). Notably, we decided not to include in the library those proteins of the glucose metabolism, as the Warburg effect in the context of cell proliferation has been deeply analyzed (1–4). We measured the ability of siRNA oligonucleotides to reduce cell proliferation by using in-cell propidium iodide (PI) staining (14, 17) in MCF10a cells. We calculated the robust Z score (Z*) (31, 32) in two different experiments and identified a group of 70 metabolic proteins, which reduction in expression levels did not modify the proliferation of MCF10a cells in both experiments ( Figure 1A ). Next, we transfected these 70 siRNA oligonucleotides in both MFC-7 cells and in MCF-7 cells expressing the Y537S ERα mutant (Y537S) hyperactive receptor variant, which confers resistance to ET (23, 33). The in-cell PI staining experiment was performed three times and those siRNA oligonucleotides with a calculated Z* > 0.5 were considered as a positive hit only if such a result was scored in at least 2 out of the 3 experiments. By using these limits, we shortlisted 31 and 32 siRNA oligonucleotides that reduced cell proliferation in MCF-7 and Y537S cells, respectively ( Figures 1B, C ). Next, we applied these 31 and 32 siRNA oligonucleotides in MCF-7 and Y537S cells and measured ERα intracellular levels in in-cell Western blot (WB) assays (14, 17). Also in this case we repeated the experiment three times and set a Z* > 0.5 threshold to identify those siRNA oligonucleotides reducing receptor levels. Treatments, which were fished out in at least 2 out of the 3 experiments, were considered positive hits ( Figures 1D, E ). Overall, we identified that the reduction in the expression levels of 8 and 11 metabolic proteins in MCF-7 and Y537S cells, respectively ( Supplementary Table 1 ), were potentially able to reduce ERα levels and cell proliferation without affecting the basal growth rate of the non-transformed breast MCF10a cells. Remarkably, 2 of these metabolic proteins were in common between MCF-7 and Y537S cells ( Figure 1F ; Supplementary Table 1 ).

In parallel, we inspected the DepMap portal (https://depmap.org/portal) to understand if the amount of some metabolites was preferentially increased in ERα-positive BC cell lines. The dataset contains the measurement of 225 metabolites in 18 ERα-positive and 30 ERα-negative BC cell lines. For each metabolite, we calculated the mean value in ERα-positive and ERα-negative BC cell lines, then we extrapolated the difference in the metabolite amount between ERα-positive and ERα-negative BC cell lines and finally we generated a p-value (using Student t-test) corresponding to the differences in metabolite abundance between ERα-positive and ERα-negative BC cell lines ( Supplementary Table 2 ). The volcano plot in Figure 1G shows the -Log₂ of the p-value as a function of the compound abundance expressed as the difference between ERα-positive and ERα-negative BC cell lines. We considered those metabolites with a difference > 0 and a p-value < 0.05 (green dots in Figure 1G ; Supplementary Table 2 ) as significantly enriched in ERα-positive versus ERα-negative BC cell lines.

Next, we performed Reactome (https://reactome.org/PathwayBrowser/#/) pathway enrichment analysis both for the metabolic proteins identified in either MCF-7 or Y537S cells ( Figure 1F ; Supplementary Table 1 ) and for the metabolites that are more abundant in ERα-positive than ERα-negative BC cell lines (green dots in Figure 1G ; Supplementary Table 2 ) and found the de novo purine nucleotide biosynthesis as the only common pathway to the 3 datasets ( Figure 1H ).

Accordingly, we found that the abundance of glutamine ( Figure 1I ) (34), a required substrate for de novo purine biosynthesis, was higher in ERα-positive than ERα-negative BC cell lines and phosphoribosylglycinamide formyltransferase, phosphoribosylglycinamide synthetase, phosphoribosylaminoimidazole synthetase (GART) ( Figure 1I ) (34), the tripartite enzyme required for the second enzymatic step in the chain of reactions for purine production, was one of the 2 common metabolic proteins, which siRNA-based depletion reduces ERα levels and prevents the proliferation of MCF-7 and Y537S cells.

Prompted by these results, we next evaluated whether different levels of GART mRNA expression could impact the survival of women carrying ERα-negative or ERα-positive BCs. Kaplan-Meier curves were done using the Kaplan-Meier Plotter database (https://kmplot.com/analysis/) (35) and showed that women with ERα-positive BCs expressing low levels of GART have a significantly higher relapse-free survival (RFS) probability with respect to those patients expressing high GART mRNA levels while no changes in RFS rate has been observed in women with ERα-negative BCs ( Figures 2A, B ; Supplementary Table 4 ).

Because these data suggest that GART expression could have a role in ERα-positive BC progression, we next evaluated its mRNA expression in breast tumors versus normal breast epithelium. The data retrieved by the TNM plot (https://tnmplot.com/analysis/) (36) show that breast tumors express higher levels of GART mRNA with respect to their normal counterpart ( Figure 2C ; Supplementary Table 5 ). However, the analysis of the Metabrick datasets extrapolated by the cBioPortal database (https://www.cbioportal.org/) (37, 38) revealed that GART mRNA was slightly but significantly reduced in ERα-positive (i.e., all the luminal tumors) than ERα-negative (i.e., all the other tumors) BCs ( Figure 2D ; Supplementary Table 6 ). To solve the contradiction for which low GART mRNA expression increases the RFS of women carrying ERα-positive tumors but GART mRNA levels are lower in ERα-positive than ERα-negative tumors, we decided to stratify the Metabrick data according to the histological type of the tumor by comparing the invasive ductal carcinoma (IDC - the most frequent one) with all the other types of tumors. As shown in Figure 2E ; Supplementary Table 7 , GART mRNA is significantly increased in IDC. Prompted by these results, we next performed immunohistochemical analysis for GART in a set of 74 breast tumors spotted on a tissue array. Quantitation of the amount of GART protein expression in those tumors classified as IDC and stratified as ERα-positive and ERα-negative revealed that ERα-positive IDCs express higher levels of GART than ERα-negative IDCs ( Figures 2F, F’ ). Altogether these data indicate that low GART levels are correlated with a higher survival rate in patients with ERα-positive tumors and that GART expression is higher in the IDCs expressing the ERα.

In parallel, because glutamine is one of the upstream substrates of GART in the de novo purine biosynthetic pathway (34), we next evaluated if glutamine content was upregulated in the same tumor types by inspecting its abundance both in the breast tumors analyzed in the Breast Cancer proteome, proteogenomic and metabolomics landscape (https://www.breastcancerlandscape.org/) (7) and in the BC cell lines profiled in DepMap portal (https://depmap.org/portal). Glutamine amount was increased in ERα-positive with respect to ERα-negative breast tumors ( Figure 2G ; Supplementary Table 8 ) and BC cell lines ( Figure 2H ; Supplementary Table 8 ). Moreover, the classification of the BC cell lines performed according to the histological type as reported both in Neve et al. (21), and in Dai et al. (22), revealed that ERα-positive IDC cell lines contain an increased level of glutamine than the ERα-positive Not-IDC cell lines while no difference in Not-IDC and IDC cell lines has been detected in glutamine levels among ERα-negative cells ( Figure 2I ; Supplementary Table 8 ).

Overall, these data indicate that both glutamine and GART levels are increased in ERα-positive IDC, thus further suggesting that targeting the de novo purine biosynthetic pathway by reducing GART activity could be a valuable strategy to prevent ERα-positive BC progression.

Next, we evaluated the effect of both the GART siRNA-dependent downregulation and the GART inhibition by lometrexol (LMX) (39) in a panel of 7 ERα-positive BC cell lines characterized by different histological types and different clinical surrogates ( Supplementary Table 9 ). In these 7 cell lines, we compared the siRNA-mediated effect on cell proliferation as retrieved by the DepMap portal database with the inhibitory concentration 50 (IC₅₀) at 5 days for the LMX-induced reduction in cell proliferation that we calculated after performing growth curve analyses in the cell lines treated in the presence of different doses of LMX. As shown in Figures 3A, B ; Supplementary Table 9 , a significant linear correlation between the antiproliferative effect of GART siRNA and the LMX effect has been evidenced only in IDC cell lines (r = 0.9647 p=0.0353) and not in Not-IDC cells (r = -0.5431 p=0.6345). Interestingly, data also indicate that IDC cell lines belonging to the LumA class of the clinical surrogates are more sensitive to LMX- and GART siRNA-dependent anti-proliferative effect (green dots in Figure 3B ).

The effect of both the GART siRNA-dependent downregulation and the GART inhibition by LMX on ERα content in the same 7 ERα-positive BC cell lines was further tested. siRNA-mediated experiments and LMX dose-response curves were performed twice in each cell line at 48 hours after siRNA or LMX administration and quantitation of the treatment effect on receptor level [i.e., percentage of reduction for siRNA experiments and effective concentration 50 (EC₅₀) for LMX] was -Log₂ transformed to obtain the sensitivity for each cell line. IDC cell lines are significantly more sensitive to GART siRNA-mediated depletion and LMX in terms of ERα degradation than Not-IDC cells ( Figures 3C, D , respectively, Supplementary Table 9 ). Remarkably, also in this case, IDC cell lines belonging to the LumA class of the clinical surrogates appear more sensitive to GART inhibition-dependent receptor degradation (green dots in Figures 3C, D ).

On this basis, we next evaluated the survival rate of women carrying LumA and LumB BC as a function of GART mRNA expression. According to the in vitro data, the RFS probability is significantly increased in women carrying a LumA breast tumor (i.e., ERα-positive, PR-positive/negative, HER2-negative) expressing low GART mRNA levels ( Figure 3E ; Supplementary Table 4 ) while no significant differences in the survival rate depending on GART expression have been evidenced for women with LumB tumors (i.e., ERα-positive, PR-positive/negative, HER2-positive) ( Figure 3F ; Supplementary Table 4 )

Therefore, these data indicate the LumA ERα-expressing IDCs as the subclass of BC, which is more sensitive to GART inhibition.

To validate these observations, we next employed 3 ERα-expressing LumA BC cell lines, 2 of them belonging to the IDC histological type (i.e., MCF-7 and ZR-75-1 cells) and one of them being classified as Not-IDC (i.e., HCC1428) (21, 22). As shown in Figures 4A, B siRNA-mediated depletion of GART reduced both the total amount of ERα and the proliferation rate in MCF-7 cells. Accordingly, GART inhibition by LMX reduced both in a time and dose-dependent manner the ERα intracellular levels ( Figures 4C, D ; Supplementary Figures 1A, B ) and the cell proliferation in MCF-7 cells ( Figure 4E ). Similar effects of LMX on ERα content and cell proliferation was scored also in the IDC ZR-75-1 cells ( Figures 4F, H ; Supplementary Figure 2A ) but not in the Not-IDC HCC1428 cells ( Figures 4G, I ; Supplementary Figure 2B ).

ERα reduction in BC cells can be triggered by the direct ability of a ligand to bind the receptor and induce its degradation (40). Therefore, we performed ERα binding assays in the presence of different doses of LMX, and E2, to understand if LMX could bind the receptor in vitro. While E2 displaced the fluorescent E2 tracer with an IC₅₀ (i.e., K_d) value of approximately 3.3 nM, as previously reported (13), LMX did not affect the ability of the recombinant ERα to bind the fluorescent E2 tracer at any of the tested LMX concentrations ( Figure 5A ).

LMX effect on ERα protein turnover rate was further tested in MCF-7 cells pre-treated with different doses of the protein synthesis inhibitor cycloheximide (CHX) for 6 hours before 48 hours of LMX administration. As expected, LMX and CHX reduced ERα intracellular content and LMX further diminished CHX-dependent reduction in the receptor intracellular amount ( Figure 5B ). Accordingly, time course analysis of CHX administration in MCF-7 cells shows that this protein synthesis inhibitor reduces ERα levels both in MCF-7 control cells and in MCF-7 cells where GART was transiently depleted ( Figure 5C ). Next, the ERα mRNA levels in MCF-7 cells, treated with LMX for 24 hours were measured, and found not to be modified by LMX ( Figure 5D ).

Therefore, LMX does not bind to ERα in vitro and does not change ERα mRNA levels, while it continues to induce receptor degradation in the presence of CHX and CHX has an increased effect in MCF-7 cells depleted of GART. Therefore, these results demonstrate that GART inhibition determines ERα degradation through post-translational mechanism not implying the binding of the inhibitor to the receptor.

It has been shown that purine-dependent starvation induces the activation of autophagy (41), and our laboratory provided evidence that ERα intracellular content in BC cells can be controlled by autophagy in addition to other cellular degradation mechanisms (42). Therefore, to investigate further the mechanistic connection between GART inhibition and the control of ERα intracellular levels, we hypothesized that LMX could induce the activation of autophagy. Consequently, we evaluated in cells treated with LMX for 48 hours both the cellular amount of LC3-II [i.e., LC3-II/(LC3-I+LC3-II)], a marker of autophagosome number, and the levels of p62^SQSTM, also known as sequestrosome, a substrate of autophagy that is considered as a marker of autophagolysosomes degradation activity (43). Figures 5E, E’ show that LMX increased LC3-II cellular content while it decreased the p62^SQSTM intracellular levels. These data suggest that LMX increases both the autophagosome number and the autophagolysosome activity.

The ERα is a transcription factor, which regulates in an E2-dependent manner the expression of those genes that contain or not the estrogen response element (ERE) in their promoter region (40). In turn, we evaluated the impact of GART inhibition on the E2-dependent control of the expression of some ERE-containing (i.e., presenilin2 – pS2 and cathepsin D – Cat D) (16) and some non-ERE containing (i.e., cyclin D1 – CycD1 and Bcl-2) (44) genes in MCF-7 cells treated either with GART siRNA oligonucleotides or with LMX. The complete anti-estrogen fulvestrant (i.e., ICI182,780 - ICI) was also introduced as an internal control.

As shown in Figure 6A , pre-treatment of MCF-7 cells with LMX or ICI determined a reduction in the E2 ability to increase the expression of both pS2, CatD, CycD1, and Bcl-2 ( Figure 6A ; Supplementary Figure 3A–D ). As expected, E2 induced also ERα degradation and ICI induced the same effect on receptor intracellular levels both in the presence or in the absence of E2, LMX was able to reduce the basal amount of ERα without blocking the E2 capability to trigger receptor degradation ( Figure 6A ; Supplementary Figure 3E ). Accordingly, siRNA-mediated depletion of GART reduced the ERα intracellular content in MCF-7 cells but did not change the ability of E2 to further trigger the ERα reduction and prevented the E2-induced increase in pS2, CycD1 and Bcl-2 expression levels ( Figure 6B ; Supplementary Figures 3F–I ). Next, we directly compared the ability of both ICI and LMX to reduce the expression levels of both pS2 and CatD in both MCF-7 and Y537S cells, which express a hyperactive receptor variant constitutively increasing receptor-dependent gene expression (23). As previously reported (23), the basal levels of both pS2 and CatD are increased in Y537S than in MCF-7 cells. Notably, both ICI and LMX reduced the pS2 and CatD protein abundance in Y537S cells ( Figure 6C ; Supplementary Figures 4A, B ). Finally, we tested the ERα activity on a synthetic ERE-containing reporter gene stably transfected in MCF-7 cells (16) in the presence and the absence of both LMX, ICI, and E2. As expected (40), E2 increased the ERα transcriptional activity, which was completely prevented both in the presence and in the absence of ICI ( Figure 6D ). According to the data shown in Figure 5A , the pre-treatment of MCF-7 cells with LMX significantly reduced both the basal and the E2-induced ERα-mediated activation of the ERE-containing synthetic promoter but E2 was still significantly able to increase ERα transcriptional activity in the presence of the GART inhibitor ( Figure 6D ). Because the complete transcriptional activation of the ERα depends on the ability of E2 to trigger the receptor phosphorylation at the S118 site (40), we next treated MCF-7 cells with E2 after the administration of LMX and measured ERα S118 phosphorylation. Figures 6E, E’ shows that LMX administration failed to block the E2-driven increase in the S118 phosphorylated ERα fraction in MCF-7 cells.

E2 is a complete mitogen for BC cells as it induces DNA synthesis, which results in cell cycle progression, and cell proliferation (40, 45). In turn, we next measured the effect of GART inhibition by LMX on E2-induced DNA synthesis by evaluating the bromodeoxyuridine (BrdU) incorporation in MCF-7 cells. Twenty-four hours of E2 administration increased the percentage of BrdU positive cells ( Figure 6F ) as previously reported (24). Notably, ICI and LMX administration reduced BrdU incorporation both in the absence and in the presence of E2 but did not block the ability of E2 to trigger BrdU incorporation ( Figure 6F ). Accordingly, E2 increased the cell number in a time-dependent manner ( Figure 6G ) and the co-treatment of MCF-7 cells with LMX prevented both the basal and the E2-induced time-dependent increase in cell number ( Figure 6G ).

Therefore, altogether these data indicate that GART inhibition i) decreases receptor transcriptional activity, but it does not impede the E2 functioning to regulate gene expression via the ERα and ii) deregulates the E2 ability to drive DNA synthesis and cell proliferation in MCF-7 cells.

In our initial screen, we identified GART as one of the 2 common proteins, which siRNA-mediated reduction induced ERα degradation and dampened cell proliferation in both MCF-7 and Y537S cells. Therefore, we next tested the impact of GART inhibition in Y537S cells. siRNA-mediated depletion of GART reduced both ERα levels and cell proliferation ( Figures 7A, B , respectively) as well as LMX administration induced receptor degradation ( Figure 7C ; Supplementary Figure 4C ) and an anti-proliferative effect ( Figure 7D ) in a dose-dependent manner.

To further evaluate if GART inhibition affects Y537S ERα transcriptional activity, we treated a Y537S cell line stably transfected with a synthetic ERE-containing reporter gene (26) with different doses of LMX for 48 hours. As shown in Figure 7E , the GART inhibitor significantly reduced in a dose-dependent manner the transcriptional activity of the Y537S mutant ERα.

Because Y537S cells are a well-known cell line modeling MBC that express a transcriptional hyperactive receptor point mutation, which confers resistance to ET drugs (e.g., Tam) (13, 23, 24, 46) and present results indicate that GART inhibition prevents the mutant receptor transcriptional activity and induces its degradation, we next decided to evaluate the possibility that GART inhibition could play a role in preventing the onset of resistance to ET and/or in inhibiting the proliferation of ET resistant cells.

Initial experiments were performed to understand if a correlation exists between the Tam and LMX effect in BC cells. For this purpose, growth curve analyses were done in the 7 above-mentioned BC cell lines ( Supplementary Table 9 ) treated with different doses of Tam for 14 days ( Supplementary Figure 5A ). We calculated the IC₅₀ for Tam in each cell line and then -Log₂ transformed this value to have a measure of the sensitivity of the different BC cell lines to Tam. Comparison of the calculated sensitivity to Tam with that extrapolated by the DepMap portal in the same cell lines revealed a significant linear correlation between the two datasets (r = 0.7453 p=0.05) ( Supplementary Figure 5B ; Supplementary Table 9 ), thus confirming our cell lines to respond to Tam as expected.

Next, we compared the sensitivity to Tam with the sensitivity of the different cell lines to LMX-induced receptor degradation in all the tested cell lines and stratified the results according to both the histological type on the cell lines (i.e., IDC and Not-IDC) and the clinical surrogates of BC (i.e., LumA and Lum B). Interestingly, we found a linear correlation between the sensitivity to Tam and to the ability of LMX to determine a reduction in ERα intracellular content in both Not-IDC and IDC ( Figures 7F, G ; Supplementary Table 9 ) cell lines. However, this correlation was significant only in IDC cell lines (r=0.9922 p=0.0078). Notably, LumA cell lines appear to be the most sensitive to both drugs (green dots in Figure 7G ).

Therefore, these data suggest that the combination of a drug inducing ERα degradation (i.e., LMX) with a drug blocking ERα transcriptional activities and functions (i.e., Tam) in LumA IDC could be an effective approach to treat this kind of breast tumor.

To test this hypothesis, we first evaluated the RFS rate as a function of GART mRNA expression of women with LumA BCs (i.e., ERα-positive, PR-positive/negative, HER2-negative) treated with chemotherapy and ET as we reasoned that this cohort of patients could benefit the most by the contemporary inhibition of both GART and ERα. Kaplan-Meier plot retrieved by the Kaplan-Meier Plotter database (https://kmplot.com/analysis/) (35) showed that in this group of patients a reduced expression of GART mRNA significantly prolongs the survival probability ( Figure 8A ; Supplementary Table 4 ). Prompted by these results we next performed proliferation studies in MCF-7 cells treated with different doses of LMX and Tam for 14 days and found significant synergy between these two drugs ( Figures 8B, B’ ), thus sustaining that these inhibitors could be used in the combinatorial treatment of IDC LumA primary BCs.

Next, we sought to determine if GART could be considered also a target for the treatment of MBC. For this purpose, we measured the GART mRNA expression as a function of grade and stage of the tumor (8) as annotated in the Metabrick datasets extrapolated by the cBioPortal database (https://www.cbioportal.org/) (37, 38). In particular, we evaluated the mRNA expression of GART in ERα-positive IDCs. Interestingly, GART mRNA expression is increased both in grade 3 tumors with respect to grade 0-2 neoplasms ( Figure 8C ; Supplementary Table 10 ) and in BCs classified as stage III-IV with respect to stage 0-II tumors ( Figure 8D ; Supplementary Table 10 ). Moreover, CRISPR-CAS9 dropout screening performed in MCF-7 cells treated for 26 days with Tam (47) revealed that GART scored within the first 150 genes that are considered important in the development of Tam resistance in MCF-7 cells ( Figure 8E ; Supplementary Table 10 ). As a reference, the position of ARID1A, which has been demonstrated as a key gene for the onset of Tam resistance in MCF-7 cells (47) has been also shown in Figure 8E . Accordingly, GART mRNA is upregulated in MCF-7 cells resistant to Tam (48).

Because these data indicate that GART is highly expressed in ERα-positive IDCs, which are poorly differentiated and metastatic, and that this metabolic protein could play a role in the development of Tam resistance, GART could be considered also a target in BC cell lines resistant to ET. CDK4/CDK6 inhibitors (i.e., palbociclib - Palbo and abemaciclib - Abe) treatment has been implemented in clinical practice as co-adjuvant drugs for ERα-expressing MBC management (5, 6, 8). Thus, we next treated Y537S cells with either Abe or Palbo in combination with LMX for 7 days. Results show that the proliferation of Y537S cells was synergistically reduced when CDK4/CDK6 inhibitors were co-administered with LMX ( Figures 8F, F’, G, G’ ).

Overall, these data demonstrate that LMX exerts synergistic anti-proliferative activities with drugs being used for the treatment of primary and MBC.

Finally, we studied the anti-proliferative effects of LMX in MCF-7 and Y537S tumor cell spheroids as well as in alginate-based cultures (24) to understand if LMX could maintain its antiproliferative activity in cells grown in 3D structures (45). Tumor spheroids and cells included in alginate-based spheres grew in spheroids and inside alginate spheres within 7 days and LMX significantly prevented MCF-7 and Y537S cell proliferation both as spheroid ( Figures 9A, A’ ) and in alginate-based cultures ( Figures 9B, B’ ).