Mice

The Rosa26-RRAS2fl/fl knock-in mouse line was generated by homologous recombination as described [19], and it contains the WT human RRAS2 sequence with an HA-tag driven by a CAG promoter, followed by an EGFP sequence downstream of an IRES sequence, and with a LoxP-flanked stop codon at the 5’ end of the construct (Rosa26-RRAS2fl/fl). Rosa26-RRAS2fl/fl mice were crossed with Cre recombinase lines to induce conditional overexpression of this construct by removing the stop codon. We first studied systemic RRAS2 overexpression using the Sox2-Cre transgenic mice kindly provided by Dr César Cobaleda (Centro de Biología Molecular Severo Ochoa-CBMSO, Madrid). As Sox2 is an embryonic stem cell transcription factor, the LoxP-flanked sequence is deleted in all tissues in these mice. By crossing with the Wap-Cre and MMTV-Cre lines purchased from the NCI Mouse Repository (Frederick, MD), we generated mammary epithelium specific RRAS2-overexpressing mouse lines. In the Rosa26-RRAS2fl/fl x Wap-Cre strain, the Cre recombinase is expressed under the control of the Wap (whey acidic protein) promoter, specifically in the secretory epithelium of the mammary gland during first pregnancy. Rosa26-RRAS2fl/fl x MMTV-Cre mice overexpress RRAS2 in virgin and lactating mammary gland, and Cre recombinase has also been detected in other tissues [20]. All the mice were maintained under SPF conditions at the animal facility of the CBMSO, in strict accordance with national and European guidelines. All the procedures were approved by the ethical committee at the CBMSO and by the Regional Government of Madrid (authorization numbers PROEX 384/15 and PROEX 296.7/21).

Human tissue samples

Samples and data from the patients included in this study were provided by Biobanks integrated into the Spanish National Biobanks Network, including the Aragon Health System Biobank (AHSB) (PT17/0015/0039) and the Fundación Instituto Valenciano de Oncología (FIVO) Biobank (PT17/0015/0051). Samples were processed following standard operating procedures with the approval of the appropriate Ethics and Scientific Committees. In total, 200 blood samples and 449 tumor samples from breast cancer patients were analyzed. A cohort of 234 human capillary blood samples was obtained from healthy volunteers from CBMSO who had provided their informed consent for genomic DNA extraction and MARBiobanc from IMIM (Institut Hospital del Mar d’Investigacions Mèdiques, Barcelona, Spain) (ISO 9001:2015). The study was carried out in compliance with EU guidelines and following the ethical principles of the Helsinki Declaration. All sample donors were of Caucasian race.

Immunohistochemistry and immunocytochemistry

Extracted organs and tumors were fixed in 4% PFA (paraformaldehyde) at 4 °C O/N (overnight). Subsequent processing for paraffin block preparation and sectioning was carried out at the Histology Facility of CNB (Spanish National Centre for Biotechnology, Madrid, Spain), as well as hematoxylin–eosin staining and Masson’s trichrome stain. For immunohistochemical staining, two incubations in xylene for 5 min each were conducted, followed by two times in 100%, 95%, 75%, and 50% ethanol, each for 5 min. Subsequently, samples were treated with methanol + 0.3% H2O2 for 20 min and washed twice for 5 min with distilled water. Antigen retrieval was carried out by incubating with 1 × citrate buffer (Sodium Citrate 1 × pH 6.0, 0.1 M Citric acid) at boiling temperature for 10 min, followed by two washes with PBS (Phosphate-buffered Saline) for 5 min each after allowing the samples to cool down. Tissue sections were then blocked by incubating with PBS + 1% BSA (Bovine Seroalbumin) at 4 °C for 30 min. Staining with the corresponding primary antibodies was performed in PBS + 3% BSA at 4 °C O/N. The primary antibodies used for immunohistochemistry were Erbb2 (#2165, Cell Signaling, 1:200); PR (Polyclonal, Proteintech, 25,871–1-AP 1:200); ERα (Clon E115, Abcam, ab32063 1:500), HA influenza hemagglutinin epitope (12CA5, Sigma, 1:500) and Ki-67 proliferation marker (Polyclonal, concentration of use 5 µg/ml). Next, incubation with the corresponding secondary antibody was carried out at 4 °C for 30 min (Polyclonal Biotinylated anti-Rabbit, DAKO E0432, 1:200 and Polyclonal Biotinylated anti-Mouse, Vector Laboratories BA-2000, 1:200). Detection of the antigen–antibody complexes was carried out using the DAB chromogen (3,3-Diaminobenzidine), incubating all the staining samples for the same time. A counterstaining with hematoxylin was performed to stain the nuclei for 10 s, followed by washes in water to remove excess, and rinsing with PBS. Samples were dehydrated by sequential incubations with distilled water for 5 min, followed by two washes of 50%, 75%, 95%, and 100% ethanol for 2 min each, and three 3-min washes with xylene. Samples were mounted using DPX mounting medium. Microscopy was performed using a Vertical microscope AxioImager M1 (Zeiss) coupled with DMC6200 camera (Leica) and LAS X software (Leica).

Whole mount staining and mammary structures counting

The mammary glands from the left fourth pair were extracted from female mice for whole- mount staining with Carmine Alum. The isolated mammary glands were air dried for 10–15 min on a clean glass slide, followed by fixation in Carnoy’s solution (60% ethanol, 30% chloroform, 10% glacial acetic acid) for 4 h at RT (room temperature). The slides were sequentially washed in 70%, 35% and 15% ethanol for 15 min each, and rinsed in distilled water for 5 min. Samples were stained RT O/N with Carmine Alum dye, followed by sequential dehydration in 70%, 95% and 100% ethanol. Glands were cleared in xylene as required, mounted with Permount mounting media (Fisher Scientific), and covered with coverslips.

Images of the whole mounts were acquired using an Olympus digital camera under constant lighting conditions and with a Vertical microscope AxioImager M1 coupled with a DMC6200 camera and equipped with LAS X software. According to defined guidelines, Terminal End Buds (TEBs) were identified as protrusions located at the end of ducts. Alveolar buds were recognized as small, round structures that were not yet organized into lobules. Each structure was classified as one single alveolar bud. Alveoli were observed as discrete units, with each lobule counted as one unit [21]. The experiment was conducted analyzing three independent animals with a minimum of four mammary glands per group and development stage. Four randomly chosen regions of 4 mm2 per mammary gland were selected for counting. The number of structures was manually counted using ImageJ software (Wayne Rasband, National Institutes of Health, Bethesda, Maryland, USA) under magnification, plotting the total number of observed structures per section.

RNA extraction

Mouse breast tumors were isolated and passed through 40 μm Cell Strainers (BD Pharmingen); healthy mammary gland tissue was finely minced using a 22 scalpel. The minced tissue was then digested in DMEM supplemented with 5% FBS and 2 mg/ml Collagenase from Clostridium histolyticum (Sigma) for 1 h at 37ºC under moderate shaking. To separate mammary organoids from fat, fibroblasts, and red cells, a series of centrifugation steps were performed at 600 g for 10 min, 5 min, and 4 rounds of 2 s each.

Human breast cancer samples were supplied in CNB (core needle biopsy) and OCT (optimal cutting temperature compound) format. CNB tissue was homogenized with RNAse free Lysing Matrix Tubes in a MagNa Lyser Rotor (Roche Instrumentation). OCT samples were homogenized using 20 G needles.

TRIzol (Invitrogen) was added to the different homogenates. Subsequently, chloroform (Merck) was added, and the mixture was vigorously vortexed. After a 15-min centrifugation at 12,000 g, the aqueous phase was carefully transferred to another tube and cold isopropanol (Merck) was added and incubated for 10 min at RT. After another centrifugation at 12,000 g for 10 min, the pellet was washed with 70% ethanol and centrifuged for 5 min at 10,000 g. The ultimate pellet was diluted in RNAse-free water and stored at -80 °C for preservation.

gDNA extraction

Tissue homogenates and cell line pellets were resuspended in 500 μL of Lysis Buffer (50 mM Tris–HCl ph = 8.0, 200 mM NaCl, 10 mM EDTA, 1% SDS) containing 0.2 mg/ml Proteinase K (Sigma) and incubated O/N at 55ºC. For mouse genotyping, gDNA was extracted from a small fragment of the tail cut when the animal was 3- 4 weeks old. In the case of peripheral blood, red blood cells were eliminated by incubating the samples with 1 ml of ACK Lysis Buffer (0.15 M NH4Cl, 10 mM KHCO3, 0.1 mM EDTA, pH 7.2- 7.4) for 5 min at RT. Then, samples were centrifuged at 1,500 g for 5 min and the remaining cells were resuspended in Lysis Buffer with Proteinase K as described. On the following day, 500 μL Phenol–Chloroform (Sigma) was added to each sample, and then centrifuged at 9,500 g at RT for 10 min. Chloroform was added to the aqueous phase and samples were centrifuged again at 9,500 g at RT for another 10 min. Afterwards, 1 mL of absolute ethanol along with 1 μL Pellet Paint (Merck) were added to the samples. Then, they were centrifuged at 9,500 g at 4ºC for 10 min. The resulting pellet was washed with 70% ethanol and the samples were once again centrifuged at 9,500 g and 4ºC for 10 min. Finally, gDNA was eluted in 8 mM NaOH, and the pH was adjusted with HEPES. After an O/N incubation at 4ºC, all gDNA samples were sonicated before measuring their concentration.

Real-time PCR

From the total RNA isolated, cDNA was synthesized with NZY First Strand cDNA Synthesis kit (NZYtech) along with Oligo-dT primers. Quantitative Real-Time Polymerase Chain Reaction (RT-qPCR) was then performed in triplicate using 50 ng cDNA per well as template. The reaction was performed with the NZYSupreme qPCR Green Master Mix (NzyTech) and gene- specific primers in a CFX384 Touch Real-Time PCR Detection System (BioRad). All oligonucleotide sequences are provided in their 5’-3’ orientation. A set of primers was used to measure mRNA expression of human RRAS2 exclusively or a combination of human and mouse RRAS2/Rras2. (human RRAS2: GCAGGACAAGAAGAGTTTGGA and TCATTGGGAACTCATCACGA; human and mouse RRAS2/Rras2: GAGTTTGGAGCCATGAGAGA and CCTTTACTCTGAGAATCTGTCTTTGA).

For mouse samples, the normalizer gene used was C-Terminal Binding Protein 1 (Ctbp1; GTGCCCTGATGTACCATACCA, GCCAATTCGGACGATGATTCTA), as it has been validated as a reliable, tissue- specific reference gene for mouse mammary gland studies [22]. For human samples, the normalizer gene was Pumilio Homolog 1 (PUM1, AGTGGGGGACTAGGCGTTAG, GTTTTCATCACTGTCTGCATCC). PUM1 has been previously recognized as one of the most stable reference genes for characterizing human tumors [23]. Each plate used in the RT-qPCR assays for human BC samples included a consistent set of 10 tumor-adjacent normal tissue samples to ensure precise normalization.

Quantitative PCR has also been widely utilized for quantification of copy number variation. The analysis involves amplification of a test locus with unknown copy number and a reference locus with known copy number. This method has been used for example for the determination of HER-2/neu amplification in human breast carcinoma and has been defined as useful alternative to FISH in breast cancer patients [24, 25]. We characterized amplifications in the RRAS2 gene in human breast cancer samples specifically targeting the exon 2 region of RRAS2 (CGGGCTGCTCTGTCATCTATC, CCATGCCTGGCCATGAATTTTA) in order to prevent any non-specific amplification of RAS genes due to their substantial sequence similarity. For normalization, COX8A (ATCATGTCCGTCCTGACGCC, CCGTTCCTCACCATGATCCC) was employed.

RNAseq of mouse tumors and differential gene expression analysis

Six breast tumor samples from Rosa26-RRAS2fl/fl x Wap-Cre mice, five samples from Rosa26-RRAS2fl/fl x MMTV-Cre mice and four samples from Rosa26-RRAS2fl/fl x Sox2-Cre mice were sequenced, along with normal mammary gland samples (2 wild-type nulliparous, 3 wild-type breeders, 2 Cre- Rosa26-RRAS2fl/fl nulliparous females and 3 Cre- Rosa26-RRAS2fl/fl breeder mice). The quality control, library preparation and data processing were carried out at the Centre for Genomic Regulation (CRG, Barcelona, Spain). Samples were sequenced with 50 base pair single-end read using a HiSeq 2000 platform (Illumina) after quality control of the samples on a Bioanalyzer Instrument (Agilent). The quality of sequencing data was assessed using FastQC software. The generated libraries were then aligned to the genome reference sequence of Mus musculus GRCm39 from Ensembl. The alignment was performed using the Spliced Transcripts Alignment to a Reference (STAR) software. Read counting was carried out with the FeatureCounts tool, and differential gene expression analysis between sample groups was performed in an R environment (version 4.2.2.) with Bioconductor using the DESeq2 package (V1.38.2) [26]. For differential gene expression analysis, the “apeglm” method was used to obtain unbiased logFC estimates, specifically targeting accurate identification of significant differences in the dataset. Genes with adjusted p-values < 0.05 after Benjamini–Hochberg correction were identified as differentially expressed genes. Fast Gene Set Enrichment Analysis was performed with the fgsea package (V1.2.4.0). Ranked gene lists derived from DESeq2 test statistics were compared to predefined gene sets from the MSigDB database [27] of Gene Ontology, Hallmarks, Reactome and community contributors. For the analysis, 1000 permutations were utilised.

Integration of mouse and human RNAseq data and clinical traits from TCGA dataset

The Cancer Genome Atlas (TCGA) consortium has collected tumors from 161 tissue source sites across the world, acquiring tumors from 11,160 patients with 33 different cancer types [28]. In this study, the Breast Invasive Carcinoma dataset was studied, including 1,116 samples of breast invasive carcinomas as well as 112 samples of normal mammary tissue. Raw counts of bulk RNAseq data, coupled with the associated clinical attributes, were obtained with the GDCquery function found in the TCGAbiolinks package, integrated within the R/Bioconductor environment. Count data from 1,228 samples were filtered to remove lowly-expressed genes. For Principal Component Analysis, the data underwent preprocessing via variance stabilization using the vsd function to normalize gene expression data, in combination with RRAS2-overexpressing mouse breast tumors RNAseq data. Raw count data from mouse breast tumors developed under RRAS2 overexpression and from breast cancer patients were merged by combining human and murine gene annotations using the BioMart tool from Ensembl (https://academic.oup.com/nar/article/51/D1/D933/6786199?login=true). Subsequently, counts were filtered to remove lowly-expressed genes and transformed for normalization using the vst() function, in order to stabilize count data variances. Principal Component Analysis (PCA) was then conducted using the pca() function. For this analysis, samples were classified according to their murine or human origin and, in the latter case, according to their previously defined molecular subtype. In addition, hierarchical clustering was performed. For this cluster analysis, genes were sorted by variance, selecting those above the cutoff value of 0.995, representing the top 0.5% of genes with the highest variances. A total of seventy genes met these criteria. To build the heatmap, 80 random samples from each subtype of breast cancer patients were selected, along with the 15 mouse breast tumors. The heatmap was generated using the pheatmap() function using “ward.D2” as hierarchical clustering method.

Triple-negative breast cancer molecular subtyping

By combining microarray expression data from 21 breast cancer datasets, which included 587 TNBC cases, Lehmann et al. identified by cluster analysis six TNBC subtypes with unique gene expression profiles: two basal-like (BL1 and BL2), an immunomodulatory (IM), a mesenchymal (M), a mesenchymal stem-like (MSL), and a luminal androgen receptor (LAR) subtype [29]. We analyzed the gene expression data from our mouse breast tumors generated under wild-type RRAS2 overexpression to determine which of the defined subtypes they resemble. For this, we used the web-based subtyping tool TNBCtype, which was developed for the characterization of TNBC samples based on Lehmann’s gene expression data and classification methods [30]. This analysis allowed us to obtain the corresponding correlation coefficient and permutation P-value for each TNBC subtype.

Western blot assays

Mouse mammary tumors and normal mammary tissue were minced, digested, and processed to separate mammary organoids as previously described. Cell aggregates were resuspended in RIPA lysis buffer (10mM Tris–HCl, pH 8.0, 1mM EDTA, 0.5mM EGTA, 1% Triton X-100, 0.1% Sodium Deoxycholate, 0.1% SDS, 140mM NaCl) containing protease and phosphatase inhibitors, mechanically disaggregated using a Polytron® tissue homogenizer, and lysed for 2 h under rotation at 4°C. Cell lysates were separated by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were then blocked for 1 h in TBS-Tween with 5% BSA and incubated overnight at 4°C with primary antibodies diluted in blocking buffer. The primary antibodies used included anti-phospho-AKT S473 (#4060), phospho-FoxO1 (Thr24)/FoxO3a (Thr32) (#9464), phospho-MEK1/2 (Ser217/221) (#9121), phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) (#4370), phospho-S6 Ribosomal Protein (Ser240/244) (#2217), and β-Actin (#4967) from Cell Signaling, and anti-HA from Sigma. Primary antibodies were used at 1:1000 dilution, except β-Actin at 1:20,000. The next day, membranes were washed twice with TBS-Tween and incubated for 45 min at RT with secondary antibody (1:30,000, Jackson Immunoresearch). After at least four washes, antibody binding was detected using standard chemiluminescence with a Kodak X-OMAT 2000 Processor.

Subcutaneous administration of β-estradiol and progesterone

To study the effects of β-estradiol and progesterone on Rras2 levels in the mammary gland, 12-week-old wild-type C57BL6/J females were housed for 14 days prior to hormone administration to synchronize their estrous cycles by the Lee-Boot effect [31]. To confirm estrous cycle synchronization, vaginal cytological evaluation was conducted through vaginal lavage followed by crystal violet staining [32]. Mice were administered subcutaneous 17β-estradiol (1 μg in sesame oil) and progesterone (1 mg in sesame oil) or control sesame oil daily for 5 days. This hormone injection protocol has previously been shown to promote mammary gland development [33, 34]. Subsequently, the females were sacrificed, and mammary glands were isolated and processed for RNA extraction as previously described, followed by RT-qPCR assays to assess Rras2 levels.

Luciferase assays

For luciferase assays, the pGL3-Control Firefly Luciferase Vector was utilized to clone either the canonical 3’-UTR of RRAS2 mRNA or the 3’-UTR containing the rs8570 SNP. Simultaneously, the intrinsic SV40 polyadenylation signal within the vector was removed through Gibson Assembly. The pGL3-RRAS2 3’-UTR WT or G124C were transiently co-transfected with Renilla Luciferase pRL-SV40 vector into CAL-51 cells using the jetPEI reagent (Polyplus) in p24-well plates. After 48 h, luciferase levels were assessed through the Dual- Luciferase® Reporter Assay System (Promega). Cell lysates were prepared utilizing 100 μL of Passive Lysis Buffer from the Dual-Luciferase Reporter Assay System. Non transfected cells were used as negative control, in order to determine the background signal from the measurement of total luminescence. Then, 20 μL of cell extracts were dispensed per well of a p96-well plate and loaded into a FLUostar Optima microplate reader equipped with syringe injectors. In addition to quantifying the luminescence of the luciferase signals, the stability of the mRNA of the luc gene, containing the different 3’UTR regions within the different BC cell lines, was quantified by RT-qPCR.

Sequencing strategy for patient samples

The technology used for the characterization of the rs8570 SNP involves a Fluorescent dual-labelled probe system, which was pre-developed by Applied Biosystems (Foster City, CA, USA). In summary, the TaqMan® MGB (minor groove binder) probes consist of specific oligonucleotides for the target SNP that carry a reporter VIC™ fluorescent dye at the 5 ́ end of the probe for allele 1 (124G) or a reporter FAM™ fluorescent dye for allele 2 (124C). Both probes are equipped with a non-fluorescent quencher (NFQ) dye at their 3 ́ ends. The minor groove binder stabilises the target-probe duplex. During the PCR reaction, the AmpliTaq Gold™ DNA polymerase used in the assay specifically cleaves the probes that are 100% complementary to the DNA template. This will lead to the release of the fluorescent reporters and fluorescence emission through the removal of the NFQ. Reactions were performed with 20 ng of template and an initial cycle of 10 min at 95ºC followed by 40 cycles of 15 s at 95ºC and 1 min at 60ºC.

When analyzing paired samples of tumor tissue and adjacent healthy margins, due to the limited availability of non-tumoral tissue, we implemented an alternative strategy. In this subset of samples, RNA was extracted and the characterization of the rs8570 SNP in RRAS2 3’UTR was achieved through nested PCR utilizing cDNA, as detailed in previous studies [19]. For the nested PCR, two reactions were conducted, each with 30 cycles of 45 s at 95ºC, 45ºC at 60ºC and 2 min at 72ºC. The oligonucleotides used for each PCR reaction are GCAGGACAAGAAGAGTTTGGA and TGAAGCAGCCTTAGTGTTTCCTT for the first reaction, TCCATGAACTTGTCCGGGTT and TGAAGCAGCCTTAGTGTTTCCTT for the second PCR. This PCR product was sent to sequence by the Sanger method (Eurofins Genomics).

Statistical analysis

Statistical parameters, including the exact n value, and the mean ± S.D. or ± S.E.M. are described in the figures and figure legends. Non-parametric Wilcoxon, Mann–Whitney, Kolmogorov–Smirnov tests, and parametric Student’s T-tests and One or Two-way ANOVA tests were used to assess the significance of the mean differences or cumulative distributions as indicated. All the data were analyzed using the GraphPad Prism 9.5.1 software. Experiments were independently repeated as described in the figure legends. The number of mice used for comparison was calculated with the aim of generating significant data to give an alpha = 0.05 and a standard deviation of about 0.3 when statistical tests were conducted with the minimum number of animals. The different deviation of the control and test groups suggested the use of different numbers of each animal for the definitive experiments.