A collection of anonymized frozen tissue sections obtained from surgical specimens of 93 patients who had an operation for invasive breast carcinoma at the Hospital Universitario La Paz (HULP) between 1991 and 2000 was interrogated in previous studies for the expression of proteins of energy metabolism (33, 34). The expression of IF1 in these breast biopsies and the study of its correlation with patients’ survival have been partially described previously (21). The Cancer Survey Tissue Microarray (OriGene, Rockville, MD, USA) containing sections of formalin-fixed normal and tumor specimens of the breast were immunostained using the monoclonal anti-IF1 (1:200) antibody (22). Sections were counterstained with hematoxylin. Patient’s medical records were reviewed and identifiers coded to protect patient confidentiality.

The breast cancer BT549-luc cell line was kindly provided by Dr. Murakami from the Japanese Collection of Research Bioresources Cell Bank (#JCRB1373). The BT549-luc cell line is a triple-negative breast cancer cell that is estrogen receptor negative, progesterone receptor negative, and lacks epidermal growth factor receptor 2 (Her2/neu) overexpression. Cells were cultured in RPMI-1640 medium supplemented with 15% FBS and 1% penicillin/streptomycin in a 5% CO₂ incubator. The pCDH-CMV-MCS-EF1-RFP + Puro cDNA Cloning and Expression Vector (SBI#CD616B-2) (SBI System Biosciences) was used to establish cells overexpressing the human ATPase IF1. The plasmid also expresses RFP for selection purposes. The human IF1 insert was obtained from the pCMV-Sport6-IF1 plasmid (22) by digestion with the NotI and EcoRI restriction sites and ligated with the linearized plasmid to generate the pCDH-CMV-MCS-EF1-RFP + PURO-IF1 plasmid. All the constructs were checked by sequencing. Stable cell lines derived by transfection with the pCDH-CMV-MCS-EF1-RFP + PURO empty vector (control, CRL) and the pCDH-CMV-MCS-EF1-RFP + PURO-IF1 were generated using lentivirus. For lentiviral transfection, viral particles were produced in HEK293T cells. They were cultured in DMEM 10% FBS. After the transfection of BT549 cells, stable transfectants were selected by adding 6 μg/ml puromycine (Invitrogen; Thermo Fisher Scientific, Inc.) to the growth medium.

Cell lysis was performed with RLN-T buffer (RLN buffer plus 0.5% Triton X-100 and the complete protease inhibitors cocktail EDTA-free; Roche) at 20 × 10⁶ cells/ml for 5 min on ice. Extracts were centrifuged at 11,000 × g for 15 min at 4°C. The primary antibodies used were as follows: anti-IF1 (1:200) (22), e-cadherin (1:250, BD Biosciences), β1 integrin (1:2, kindly provided by Carlos Cabañas, CBMSO), β-catenin (1:500, BD Biosciences), vimentin (1:1,000, Cell Signaling), NF-κB p65 (1:1,000, Abcam), pIKBα (1:500, Cell Signaling), IKBα (1:1,000, Cell Signaling), α-tubulin (1:3,000, Sigma-Aldrich), β-F1-ATPase (1:25,000) and Hsp60 (1:2,000) from Ref. (35), SDH-B (1:500, Invitrogen),α-F1-ATPase (1:1,000, Molecular Probes), Core 2 of complex III (1:500, Abcam), MTCO2 (1:500, Abcam), VDAC (1:500, Abcam), and PYGM (1:1,000, Abcam).

Mitochondrial isolation was performed according to Acin-Perez et al. (36). In brief, cells were washed with PBS. Cellular pellets were homogenized in a glass–teflon homogenizer with seven volumes of hypotonic buffer (83 mM sucrose, 10 mM MOPS pH 7.2). After homogenization, the same volume of hypertonic buffer (250 mM sucrose, 30 mM MOPS pH 7.2) was added and nuclei and unbroken cells were removed by centrifugation at 1,000 × g. Mitochondria were obtained by centrifugation at 12,000 × g and washed in buffer A (320 mM sucrose, 1 mM EDTA, 10 mM Tris–HCl pH 7.4). Cytosolic fraction was collected in the supernatant. Mitochondria were further purified by centrifugation through a 0.8 M sucrose cushion in 10 mM Tris–HCl pH 7.5 0.1% BSA. Cellular, mitochondrial, and cytosolic proteins were loaded on SDS-PAGE, transferred onto nitrocellulose membranes, and immunoblotted against diverse antibodies.

The ATP synthetic activity was determined in digitonin-permeabilized cells following the detailed protocol in Ref. (37). Inhibition of the synthase activity of the H⁺-ATP synthase was accomplished by the addition of 30 μM oligomycin (OL). Cellular ATP concentrations were determined using the ATP Bioluminescence Assay Kit CLS II (Roche) following the manufacturer’s instructions.

The initial rates of lactate production were determined as an index of glycolysis by enzymatic determination of lactate concentrations in the culture medium (38). Oxygen consumption rates were determined in an XF24 Extracellular Flux Analyzer (Seahorse Bioscience) (22). Cells were seeded in the microplates, and incubated at 37°C and 5% CO₂ for 24 h. The final concentration and order of injected substances was 6 μM OL, 0.75 mM DNP (2,4-dinitrophenol), 1 μM rotenone, and 1 μM antimycin. The intracellular production of hydrogen peroxide was monitored by flow cytometry using 5 μM 6-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (DCFH2-DA) (Molecular Probes, Eugene, OR, USA) and incubated 30 min at 37°C (27). Cells were analyzed in a BD FACScan. For each analysis, 10,000 events were recorded.

Total RNA was extracted from BT549 cells using Trizol (Invitrogen, Carlsbad, CA, USA) followed by the Qiagen RNeasy (Qiagen, Valencia, CA, USA). Each RNA preparation was tested for degradation using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Total RNA (200 ng) was amplified using One-Color Low Input Quick Amp Labeling Kit (Agilent Technologies) and purified with RNeasy Mini Kit (Qiagen). Preparation of probes and hybridization was performed as described in One-Color Microarray Based Gene Expression Analysis Manual Ver. 6.5, Agilent Technologies. Briefly, for each hybridization, 600 ng of Cy3 probes were mixed and added to 5 μl of 10× blocking agent, 1 μl of 25× fragmentation buffer, and nuclease-free water in a 25-μl reaction, incubated at 60°C for 30 min to fragment RNA, and stopped with 25 μl of 2× hybridization buffer. The samples were placed on ice and quickly loaded onto arrays, hybridized at 65°C for 17 h in a hybridization oven, and then washed in GE Wash Buffer 1 at room temperature (1 min) and in GE Wash Buffer 2 at 37°C (1 min). Arrays were dried by centrifugation at 2,000 rpm for 2 min. Slides were Sure Print G3 Agilent 8x60K Human (G4852A-028004). Images were captured with an Agilent Microarray Scanner and spots quantified using Feature Extraction Software (Agilent Technologies). Background correction and normalization of expression data were performed using LIMMA (39, 40). Linear model methods were used for determining differentially expressed genes. Each probe was tested for changes in expression over replicates by using an empirical Bayes moderated t-statistic (39). To control the false discovery rate p-values were corrected by using the method of Benjamini and Hochberg (41). The expected false discovery rate was controlled to be <5%. Hybridizations and statistical analysis were performed by the Genomics Facility at Centro Nacional de Biotecnologia (Madrid, Spain).

A gene list was made by selecting genes with a fold change ≥1.5 or ≤−1.5 between control and IF1-overexpressing cells. A corrected p-value ≤0.05 was considered statistically significant. GeneCodis3 tool was used to perform gene set enrichment analysis with GeneCodis (42) to infer the main biological functions associated with IF1 overexpression using KEGG and Panther pathways enrichments. Gene set enrichment by Ingenuity Pathway Analysis (Qiagen) was also used to determine the specific canonical signaling pathways and related diseases and functions affected. A Fisher’s exact right-tailed test identified significantly enriched pathways, and a z score was computed to determine whether the pathway was activated or inhibited at each stage.

Reverse transcription reactions were performed using 1 μg of total RNA and the High Capacity Reverse Transcription Kit (Applied Biosystems; Thermo Fisher Scientific, Inc.). Real-time PCR was performed with an ABI PRISM 7900HT SDS (Applied Biosystem) and Power SYBR Green PCR Master Mix (Applied Biosystems). The following primers were used: ACTA2 (NM_001141945), DCN (NM_001920), ITGB1 (NM_002211), POSTN (NM_001135934), TGFB3 (NM_003239), SERPINB7 (NM_001040147), SERPINB11 (NM_080475), VCAN (NM_001126336), and WNT2B (NM_004185). All primers provided by Sigma-Aldrich.

Cellular proliferation was determined by counting the number of cells/well after 24, 48, and 72 h of culture. For cell death assays, 30,000 cells/well were seeded and treated with 1 μM staurosporine (STS) or 120 μM hydrogen peroxide during 24 h. Cell death was determined by flow cytometry after staining with annexin V (ApoScreen FITC; Southern #10010-02) (27).

For soft agar assay, 2,000 cells were added to 750 μl of complete growth media with 0.5% agarose and layered onto a 750 μl bed of complete growth media plus 0.7% of agar on a 12-well plate. After 5 weeks, viable colony numbers were stained with 0.5% crystal violet in 4% paraformaldehyde and counted using ImageJ software. For wound healing assays, confluent cell monolayers were mechanically disrupted with a sterile pipette tip to produce a clean uniform scratch. The wells were photographed every 30 min. Gap distances were analyzed with ImageJ software. Corning BioCoat Matrigel Invasion Chambers were used to quantify the cellular invasive capacity. A total of 2 × 10⁴ cells were seeded in 1% FBS and chemoattraction perform during 72 h in 20% FBS.

The results shown are the means ± SEM of four to six biological replicates. Statistical analysis was performed by Student’s t-test, Mann–Whitney U test, and/or Kruskal–Wallis as appropriate. Tests were calculated using the SPSS 13.0 software package (IBM, Chicago, IL, USA). Survival curves were derived from Kaplan–Meier estimates and compared by log-rank test. Statistical tests were two sided at the 5% level of significance.

We have previously documented in a large cohort of breast cancer patients (33) that IF1 is highly overexpressed in breast carcinomas (21) [for details, see supplemental Table 1 in Ref. (21)]. These results have been further confirmed in another large cohort of breast carcinomas (43). Figure 1A provides an illustrative example of the changes taking place in IF1 expression in breast by oncogenesis. The normal breast tissue shows negligible expression of IF1, whereas a very high expression level of the protein is observed in the carcinomas. Figure 1B shows that the tumor expression level of IF1 has no relevant influence on the 5-year overall survival rate in breast cancer patients. However, when the content of IF1 in the carcinoma is high (21), it significantly associates with an increase in disease-free survival of the patients (Figure 1B). The association of high IF1 levels with a lesser chance to develop metastatic disease is magnified in the poor prognosis group of triple-negative breast carcinomas (Figure 1C) (21). Overall, these findings suggest that high IF1 levels in breast carcinomas in some way correlate with a decrease of metastasis despite supporting a pro-oncogenic metabolic phenotype (21, 22, 27).

In order to investigate the molecular basis of this paradox, we developed stable IF1-overexpressing breast cancer cells (IF1-cells). For this purpose, we used the triple-negative breast cancer cell line BT549-luc. After transfection and selection, the overexpression of IF1 was confirmed both by immunoblotting (Figure 2A) and immunofluorescence microscopy (Figure 2B). The overexpressed transgene was localized in mitochondria (Figures 2B,C). Consistent with previous findings in transient expression experiments (21), the lentivirus-driven overexpression of IF1 significantly diminished the rates of cellular respiration (Figure 2D). Moreover, and when compared to controls, the rates of glycolysis were significantly augmented in IF1-cells (Figure 2E), further supporting that IF1 is interfering with the mitochondrial production of ATP by the ATP synthase (21, 22, 27). In fact, the expression of IF1 partially inhibited the ATP synthase activity as determined in permeabilized cells (Figure 2F) resulting in a diminished cellular content of ATP (Figure 2G). However, it should be noted that the rates of aerobic glycolysis in IF1-cells can still be overstimulated by incubation with OL (Figure 2E), indicating that the overexpression of IF1 is not enough to inhibit all the ATP synthase present in the cell. In agreement with previous reports (21, 27), the IF1-mediated inhibition of the ATP synthase resulted in a significant increase in basal cellular ROS levels (Figure 2H) without affecting the expression of other mitochondrial proteins (Figure 2I). Altogether, these results confirmed the biological activity of IF1 as a regulator of cellular energy metabolism in agreement with previous results.

Analysis of the transcriptome of control (empty vector) and IF1-overexpressing breast cancer cells rendered 2,661 genes differentially expressed (p ≤ 0.05; fold change ≥1.5; LIMMA analysis) (Figure 3A; Table S1 in Supplementary Material). From the global list of significant genes (Table S1 in Supplementary Material), 1,427 were upregulated and 1,234 were downregulated (Figure 3A). A volcano plot that combines the statistical significance with the magnitude of change illustrates the distribution of the IF1-regulated genes further emphasizing some of the genes that display large magnitude of significant change (Figure 3B). Consistently, the ATPIF1 gene was found significantly overexpressed in the transcriptome of IF1-cells (Figure 3B). A pathway enrichment analysis using the GeneCodis tool (42) against KEGG and Panther pathways databases highlighted that most of the affected pathways on IF1-cells were related to cancer (Figure 3C). In the case of KEGG database, the affected pathways revealed association with the ECM (focal adhesion, ECM–receptor interaction, and regulation of actin cytoskeleton), intercellular communication (cytokine–cytokine receptor interaction and MAPK signaling, calcium signaling, and chemokine signaling pathways), and cell cycle (Figure 3C). In the case of Panther database, similar groups appeared with the addition of angiogenesis and apoptosis (Figure 3C). Moreover, we performed an additional enrichment analysis with IPA ingenuity tool in order to complement the information with a prediction of the activation/repression status of the affected pathways. The results obtained pointed out that the majority of the canonical pathways affected in IF1-cells were related to cancer (Figure 3D) and, more specifically, to cellular mobility/metastasis and tumorigenesis in agreement with the previous enrichment analysis. Remarkably, these functions were predicted to be inhibited in IF1-cells when compared to controls (Figure 3D). Moreover, the IPA analysis of diseases and functions highlighted that migration and invasion of breast cancer cells were specifically inhibited in IF1-overexpressing cells (Figure 3E). Overall, these results suggest that IF1 overexpression could favor the generation of a cellular phenotype with less mobility and invasiveness, which could underpin the lower metastasis and better prognosis of breast cancer patients bearing tumors with high expression level of the protein.

Tables S2 and S3 in Supplementary Material list the 52 upregulated and 79 downregulated genes in IF1-cells when compared to controls that meet the more restrictive multiple correction test of Bonferroni (Figure 4A). Pathway enrichment analysis using the GeneCodis tool with the set of 138 genes also confirmed that cellular movement and cell-to-cell signaling and interaction are affected in IF1-overexpressing breast cancer cells.

The results of the transcriptomic analysis were validated by the quantification of the expression of some of the genes involved in ECM and its signaling by quantitative PCR (Figure 4B). Consistently, SERPINB7 and SERPINB11, which are members of the superfamily of inhibitors of extracellular proteases, were highly up regulated in IF1-cells (Figure 4B). On the other hand, a significant diminished expression of ACTA2, encoding a member of the actin protein family important for cell movement and contraction; POSTN which encodes periostin—a component of the ECM involved in the organization of collagens; and VCAN—a member of the versican proteoglycan family which is involved in cell adhesion, were found downregulated in IF1-overexpressing cells (Figure 4B). Likewise, the expression of signaling molecules such as the cytokine TGFB3, which is involved in the regulation of cell adhesion, of the ECM, and of WNT2B (formerly WNT13), encoding a glycoprotein of the wingless secreted signaling factors that promote epithelial–mesenchymal transition (EMT), was significantly downregulated in IF1-overexpressing cells (Figure 4B). Overall, these results support that a low expression of IF1 in breast cancer cells favors remodeling of the ECM to facilitate cellular migration and metastasis.

Next, we studied by immunoblotting the expression of some proteins commonly associated with the modification of the ECM, the induction of EMT, and cell migration. The results revealed no differences in vimentin or β-catenin levels between control and IF1-overexpressing cells (Figure 4C). Likewise, the NF-κB and Snail pathways—two pathways frequently activated in cancer—were not significantly affected by the overexpression of IF1 in breast cancer cells (Figure 4C). However, the levels of E-cadherin, an adhesion molecule involved in the maintenance of cellular and epithelial tissue integrity, was significantly upregulated in IF1-overexpressing cells (Figure 4C). Furthermore, integrin-β1—an adhesion molecule related to migration, extravasation, and metastasis (44)—levels were found significantly upregulated in control when compared to IF1-overexpressing cells (Figure 4C). Overall, the results reveal that the overexpression of IF1 in breast cancer cells supports the maintenance of ECM and tissue integrity.

Assessment of the rates of cellular proliferation (Figure 5A) and of cell death after hydrogen peroxide or STS treatment (Figure 5B) revealed no relevant differences between control and IF1-expressing cells. Interestingly, soft agar colony-formation assays showed that IF1-cells had a significant less capacity to grow and form colonies in the anchorage-independent assay (Figure 5C), suggesting a lower tumorigenic potential. To verify the migration ability of the cells, wound healing assays were carried out (Figure 5D). The results revealed that control cells started filling and fully occupied the scratched area earlier than IF1-cells (Figure 5D; see Video S1 in Supplementary Material), indicating that IF1-overexpressing cells had less migration ability than control cells. Similarly, matrigel invasion assays also revealed that control cells had a higher invasive capacity than IF1-cells (Figure 5E). Overall, these results suggest that IF1 overexpression in breast cancer cells induces a less aggressive phenotype by diminishing the migration and invasive capacities of the cells. This finding agrees with the fact that breast cancer patients with elevated tumor levels of IF1 had less metastatic disease.