The present study represents a pilot (19), prospective/cross-sectional research approved by the Ethics and Research Committees of the Instituto Nacional de Cancerología (INCAN), México. The study from 2008 to 2013 included the evaluation of 102 women (age from 30–86, mean 51±12 years) diagnosed with infiltrating ductal breast carcinoma (IDC) also called invasive ductal carcinoma at the INCAN, México. According to the Cochran’s categorical formula for determining sample sizes in pilot and prospective studies, 100 was the appropriate sample size based on i) the α level = 0.05; and ii) an acceptable margin of error <10% (20).

Normal tissue was surgically withdrawn from selected areas of normal breast tissue of 31 cancer patients (age from 27 to 89 years, mean 58±17 years) and stored in liquid nitrogen in the INCAN Tumor Bank for 12–24 months (21). Internal control of sample stability revealed that the contents of several proteins (HIF-1α, GLUT, LDH and ATPase) from fresh samples were similar to those stored in the INCAN Tumor Bank (data not shown). Normal samples were further validated as non-tumorigenic by assessing immunohistochemistry (IHC) negativity towards HER2 and hormone receptor. Patients diagnosed with IDC and no previous clinical treatments were used as the inclusion criterion for the present study, whereas insufficient material for biochemical assays was used as the exclusion criterion.

Tumor samples were obtained following the medical proceedings for tissue handling and patient care approved by the Institutional Ethics Committee supported by a patient’s informed consent according to the Declaration of Helsinki. Patients were punctured with a tru-cut biopsy needle in the absence of local anesthesia. Histopathology was performed on hematoxylin and eosin stained biopsy slides. Each specimen was analyzed by visual inspection using standard light microscope by experimented pathologists as described (22) without knowledge of IHC results.

Human tissue was fixed in 10% neutral buffered formalin for 24 h. The sample was embedded in paraffin, cut with a microtome at 3 μm thick and placed on microscope slide. Immunostaining was performed with a BenchMark Ultra automated immunostainer (Ventana Medical Systems, Tucson, AZ, USA); estrogen receptor, progesterone receptor, HER2 and Ki67 protein antibodies were used at 1:100–250 dilutions. All antibodies were from Ventana Medical Systems, except for Ki67 (Bio SB Inc., Santa Barbara, CA, USA).

For estrogen and progesterone receptor identification, H-SCORE method was employed as described (23). For HER2 identification, negative score indicated <10% of stained cells; positive score indicated >30% of cellular staining.

The IHC analysis revealed that from 102 IDC biopsies, 23 corresponded to triple positive; 57 to HER2⁻/HR⁺; 8 to HER2⁺/HR⁻; and 14 to TN.

Samples from biopsies (0.4–1.8 mg total cellular protein, n=102) were placed in liquid nitrogen and kept at −70°C until used. For western blotting processing, frozen tumor and normal samples were powdered, re-suspended and homogenized in 0.6 ml 25 mM Tris-HCl buffer, pH 7.4, plus 1 mM PMSF (phenylmethanesulfonyl fluoride), 1 mM EDTA and 5 mM DTT, and centrifuged at 2,000 × g for 30 min at 4°C. Supernatants were recollected and protein content was determined by using the Lowry assay. IDC and normal breast samples (50 μg cellular protein) were re-suspended in loading buffer (10% glycerol; 2% SDS and 5% β-mercaptoethanol) and loaded onto 12.5% SDS-PAGE under denaturalizing conditions. The proteins were blotted to PVDF membranes (Bio-Rad Laboratories, Hercules, CA, USA) and protein identification was performed by overnight incubation with anti-HIF-1α, -GLUT1, -HKI, -HKII, -LDHA, -COXIV, -ATPase, -ANT, -GA, -2OGDH, -HER2, -c-MYC, -Ki67, -α-tubulin and -HRAS (Santa Cruz Biotechnology, Santa Cruz, CA, USA) specific monoclonal antibodies (1:500–1:1,000 dilution). Detection of the hybridization bands was performed with the horseradish peroxidase reaction in photographic plates as previously described. Densitometry analysis was carried out using Scion Image software (Scion Corp., Walkersville, MD, USA). Double normalization of tumor sample signal was first performed against its respective load control (α-tubulin) and then considering the normal tissue as 100% (24).

The supernatants from the frozen-thawed biopsy samples, prepared as described above, were stored at −20°C in the presence of 10% (v/v) glycerol until determination of enzyme activities. Hexokinase (HK) and lactate dehydrogenase (LDH) activities were spectrophotometrically determined at 340 nm and 37°C as described elsewhere (25). Briefly, HK activity was assayed in 50 mM MOPS buffer, pH 7.0 plus 2 U glucose-6-phosphate dehydrogenase, 1 mM NADP⁺, 15 mM MgCl₂, 10 mM ATP and 20–60 μg cell extract protein/ml. The reaction was started by adding 3 mM glucose after 3-min pre-incubation and generation of NADPH was measured at 340 nm. Lactate dehydrogenase (LDH) was assayed in 50 mM MOPS, pH 7.0, 0.15 mM NADH and 10–20 μg cell extract protein/ml; after 3-min pre-incubation; the reaction was started with 1 mM pyruvate and NADH consumption was registered at 340 nm.

To identify significant differences in protein contents between non-tumor and tumor samples as well as among the different tumor subtypes, parametric and non-parametric statistical analyses were performed. The Kolmogorov-Smirnov and Levene tests (26,27) were applied to demonstrate the protein normal distribution and homogeneity of variance of samples. Analysis of Kolmogorov-Smirnov and Levene tests data indicated that non-parametric analysis (NPA) should be used for appropriate statistical assessment between non-tumor and tumor samples. The NPA analysis and graphical data were carried out by using the Microsoft SPSS v.20 (SPSS Inc., Chicago, IL, USA) and Microsoft OriginPro 8 (OriginLab Corp., Northampton, MA, USA) software, respectively. To validate the results obtained with NPA test, samples were re-analyzed by the Mann-Withney U test with a P<0.01 (28).

To assess differences of the analyzed proteins among the tumor subtypes HER2⁺/HR⁻; HER2⁻/HR⁺; TP and TN the NPA Kruskal-Wallis test and the parametric analysis ANOVA were applied. To validate the results of the Kruskal-Wallis and ANOVA tests, the Mann-Whitney U test (28) corrected by the Holm-Bonferroni method and Scheffé post hoc test was used. Receiver operative characteristic (ROC) curves were also built for further validation as well as identification of the cut-off values for each protein. All statistical tests were performed at significance level of at least 0.05 as reported for the majority of human biopsy studies (29).

Infiltrating ductal breast cancer (IDC) is the most dominant type of breast carcinoma found in the Mexican female population (30). From a total of 116 cases analyzed, 88% (102 biopsies) corresponded to IDC. IDC incidence was identified by standard IHC analyses inspecting hematoxylineosin stains for apparent cellular abnormalities (atypias) and cellular mitosis (Fig. 1A). For HER2 positive subtypes, all samples showed HER2 intensity on tumor cell plasma membrane with 3+ score (i.e., >30% of stained cells) (Fig. 1B), whereas the score of all HER2 negative samples was 0 and 1+ (i.e., 100% non-stained cells) (31).

The clinical characteristics of the IDC subtypes are described in Table I. The highest number of IDC patients (57 samples or 56%) corresponded to HER2⁻/HR⁺ followed by 23 samples (22%) of HER2⁺/HR⁺ (triple-positive, TP), 14 samples (14%) of HER2⁻/HR⁻ (triple-negative, TN) and 8 samples (8%) of HER2⁺/HR⁻. The percentages found in the present study for all IDC subtypes correlated with those reported in other clinical studies (32). The TN patients arguably have survival advantages as a result of the absence of HER2 overexpression. However, they lack the benefit of both routinely available targeted therapy and specific biomarkers. It is worth noting that the present prospective study analyzes a small number of IDC samples, as it has been reported for pilot clinical studies (19), and in contrast to other non-pilot epidemiological studies the IDC sample number is higher (>450 patients) (32); in consequence, in the present study the number of samples analyzed for each IDC subtype was smaller. Therefore, to avoid erroneous interpretations derived from a relative small number of samples, the analyses of the metabolic protein contents and activities were performed without considering the patient’s age or menopausal status. Future investigations will be oriented to increase the sample size to further validate the results found in the present study with other analytical methods such as RT-PCR and microarray approaches.

The content of the transcription factor HIF-1α and its glycolytic targets GLUT1, HK (I and II) and LDH-A of tumor and non-tumor biopsies were normalized against α-tubulin (Fig. 2). The HIF-1α contents in tumor samples were significantly higher than those in non-tumor biopsies following the Mann-Whitney U test (Table II) and ROC analyses (Fig. 3A). The last test also revealed a cut-off point of 27 (i.e., 27% of the HIF1-α band intensity respect to the α-tubulin signal) for tumor vs. non tumor samples, where the highest sensibility (>80%) and specificity (>90%) were attained (Table II). These data indicated that, at least, a HIF-1α band intensity of a 27% is required to ensure that, in the IDC biopsy, <20% of false negatives and 10% of false positives can be found.

The increased HIF-1α content detected in tumor samples was accompanied by a slight, but non-significant increase in the protein content of its glycolytic target GLUT1. On the contrary, HK (I and II) and LDH-A contents were without change compared to normal breast tissue (Fig. 2). The functional determination of HK and LDH supported the observed western blot analysis. Activity of HK and LDH from tumor biopsies (22±5 mU/mg protein for total HK; n=23; and 307±78 mU/mg protein for total LDH; n=23) was similar to those from breast normal tissue (16±5 mU/mg protein for total HK; n=5; and 474±196 mU/mg protein for total LDH; n=5).

The contents of HIF-1α, GLUT1, HKI and LDH-A were similar among the different IDC subtypes (Fig. 4). However, for HKII a significant change, indicated by ANOVA and Scheffé post hoc tests, was determined in HER2⁺/HR⁻ vs. TP, vs. HER2⁻/HR⁺ and vs. TN (Fig. 4; Table III). The ROC for HKII indicated a cut-off of 73.5 with sensitivity and specificity of 86 and 31%, respectively.

For mitochondrial proteins, no apparent changes were observed in the contents of 2OGDH, GA isoforms K and L, and ATP synthase between tumor breast biopsies vs. non-tumor tissue (Fig. 2). On the contrary, COXIV and adenine nucleotide translocase (ANT) levels were significantly higher and lower, respectively, in tumor vs. non-tumor samples by using Mann-Withney U test (Fig. 2; Table II). However, the ROC tests (Fig. 3A) revealed that only COXIV, but not ANT, was significantly different vs. non-tumor tissue. For this case, the cut-off point was 15 with 100% specificity but low (38%) sensitivity, indicating that COXIV could be considered as a good tumor biomarker but with high probability (>50%) to detect also false positives (Table III). The content of 2OGDH was similar between TP and HER2⁻/HR⁺. On the contrary, in TN samples, the levels of 2OGDH were significantly higher vs. the pooled subpopulation of the other IDC subtypes (Fig. 4 and Table IV), whereas for HER2⁺/HR⁻ the content of 2OGDH was lower although non-significantly different vs. TP and HER2⁻/HR⁺ subtypes, by applying ANOVA and Scheffé post hoc tests (Fig. 4). For 2OGDH, data of cut-off value, sensitivity and specificity (Table V) showed that it is required at least 70% of band intensity (respect to that of tubulin as the loading control) in the biopsy to diminish to 20 and 40% the probability to have false positives or negatives, respectively. Also, a substantial but not significant increase in the ANT content was observed in TN compared to pooled subpopulation of the other IDC subtypes (Fig. 4), in which the ANT content was similar.

In tumor breast biopsies, the proliferation protein Ki67, the oncogene c-MYC and the growth factor HER2 (in TP and HER2⁺/HR⁻ subtypes) were fully apparent, whereas no presence of these proteins was detected in normal breast tissue (Fig. 5), in agreement with previous reports (33). The Mann-Withney U test (Table II) showed that Ki67, c-MYC and HER2 were significantly different in TP and HER2⁺/HR⁻ vs. normal biopsies. ROC analysis showed that HER2 and c-MYC exhibited higher area under curve (AUC) than Ki67 (Table III) indicating that HER2 and c-MYC were indeed significantly different in TP and HER2⁺/HR⁻ vs. normal biopsies (Fig. 3A). For these two proteins, similar cut-off points, sensitivity and specificity were determined (Table III), indicating that only 26–30% of protein signal (respect to the tubulin content) in TP and HER2⁺/HR⁻ samples is required to achieve: i) high probability (80–100%) to discard negative diagnosis (i.e., values lower than 26–30% indicate the absolute certainty of absence of disease), although ii) moderate probability (55–60%) for true positive identification (vs. false positive, i.e., the sample derived from a non-cancer patient having some other health problem).

HER2 presence (TP, HER2⁺/HR⁻) or absence (TN, HER2⁻/HR⁺), as well as HR positivity or negativity, was confirmed for the majority of the samples by western blotting (Fig. 6), further validating the standard IHC clinical approach (Fig. 1B; Table I) currently used at the Instituto Nacional de Cancerología de México. However, 12 out of 71 HER2⁻ samples (17%), and 3 out of 22 HR⁻ samples (14%), as indicated by the IHC assay, yielded positive signal by western blot assay (data not shown). These results clearly indicated that additional support for breast cancer subtype diagnosis such as western blotting should be routinely implemented to decrease the emergence of false negatives and hence to establish the appropriate therapy.

Surprisingly, H- and K-RAS proteins were also found in normal tissue (Fig. 5). This result could not be confirmed by literature data regarding the presence or absence of H- or K-RAS in normal cells. Assessment of Ki67, oncogenes and transcription factor contents among the different IDC subtypes by the Krustal-Wallis test only revealed significant difference for the HER2 content in TP vs. HER2⁺/HR⁻. However, the Holm-Bonferroni/Mann-Whitney U corrected test showed no significant difference for such datasets.

Proteins involved in the metastatic response of tumor cells such as vimentin and E-cadherin were determined in tumor vs. non-tumor samples to assess the migration and invasion profiles of tumor biopsies. Both proteins were not statistically different between assayed groups (Fig. 5; Table II). Although, both BNIP3 (a key regulator of hypoxia-induced autophagy) as well as LAMP1 (lysosome biogenesis-induced autophagy) were significantly higher in IDC samples vs. non-tumor samples (Fig. 5), according to the Mann-Whitney test, ROC analysis showed no significant differences. Among IDC subtypes, metastasis and autophagy protein contents were similar, except for TN whose E-cadherin content was significantly higher (Fig. 3B; Table IV) to those determined for HER2⁻/HR⁺ and TP. In this regard, an E-cadherin cut-off value of 3–37 was determined (Table V), indicating that this is the required protein expression (respect to the tubulin signal) to reach a high probability (80–100%) to identify true positive samples and discard false positive and false negative results.

Identification of significant differences among the contents of all proteins assayed (metabolic proteins, oncogenes, transcription factors, as well as metastatic and autophagic proteins) between tumor and non-tumor tissue was determined by using the stringent non-parametric statistical Mann-Whitney U test validated by the ROC graphs (Fig. 3A; Tables II, III and V). With these rigorous statistical analyses, significant differences between non-tumor and tumor samples were observed for HIF-1α > HER2 > c-MYC > COXIV (Fig. 3A; Table II), whereas non-significant changes were detected for the rest of the mitochondrial and glycolytic proteins analyzed. In contrast, for the identification of significant differences on the proteins assayed among the different IDC subtypes, non-parametric statistical Kruskal-Wallis test and the parametric ANOVA validated by the ROC graphs (Figs. 1B and 3B; Tables IV and V) were used. These last analyses showed significant changes in 2OGDH and E-cadherin protein contents in TN vs. TP, HER2⁺/HR⁻ and HER2⁻/HR⁺, thus, providing a differentiated and selective panel of biomarkers constituted by HIF-1α > c-MYC > 2OGDH > E-cadherin for TN subtype IDC biopsies.

High HIF-1α content and glycolysis are indicative of increased malignancy and poor prognosis (41). Therefore, HIF-1α has been proposed as a biomarker for different metastasic carcinomas (colon, breast, gastric, lung, skin, prostate, ovarian, pancreatic, brain, glioblastoma and renal) (reviewed in refs. 42,43). In all these studies, the HIF-1α content in cancer samples, solely evaluated by IHC, showed an increase of 30–83% vs. normal tissue (43). Our present data using western blot analyses revealed a substantial increment of HIF-1α in all cancer subtypes compared with their respective normal tissue (Fig. 2), indicating that this transcription factor can be a reliable biomarker of breast cancer biopsies. The high sensitivity and specificity of the HIF-1α detection by western blotting (i.e., the antibody selectively detects a target protein in a mixture of several thousand different proteins coming from heterogeneous tissue) compared to the IHC assay makes the former the method of choice for routine clinical use, despite the longer processing time for western blotting (days) than for IHC.

On the other hand, HIF-1α functional status has not been systematically evaluated to establish a correlation between protein content and transcriptional activity (42,44). In the present study the transcriptional functionality of HIF-1α was assessed in all IDC biopsies. The data indicated that although a substantial increment in the HIF-1α levels was observed compared to normal tissue (Fig. 1), all IDC subtypes maintain similar GLUT1, HKII and LDH-A contents and activities to those determined in non-tumor samples. Unfortunately, there are no studies in which the content and activity of these glycolytic proteins have been determined, except for LDH whose activity, determined here, was within the range reported for breast cancer patients (45). Nevertheless, the present data indicated that HIF-1α, but not the glycolytic proteins can be a striking reliable marker of Mexican breast carcinoma as has been already suggested for human squamous cancer cervix epithelium (46). In other studies performed in breast tumor perinecrotic area and cervical cancer a strong correlation between HIF-1α and GLUT1 has been found (44,46).

Regarding mitochondrial OxPhos proteins, only the ANT content was detected to be significantly decreased in the cancer biopsies. In PC12 tumor cells, short-term hypoxia (30 min) downregulates the transcription of genes encoding mitochondrial complex I/NADH dehydrogenase as an adaptive response mechanism for adjusting the OxPhos rate to the low O₂ availability (47). A putative regulatory site in the ANT gene for HIF-1α has not been reported.

Epidemiological studies of breast cancer in the Mexican female population revealed that approximately 20% of patients develop the TN phenotype (36). Unfortunately until now, the treatment with anti-hormone drugs (tamoxifen) or monoclonal anti-HER2 (trastuzumab) has been ineffective in the majority of these diagnosed cases (48). Therefore, it appears relevant to develop research focused on the identification of specific TN biomarkers. Recently it has been documented that the MAG13-AKT3 protein may fulfill such as role (49). However, its content is only overexpressed in a scarce number of TN patients (7% or 5/72), disabling its use for all TN patients (49). In the present study, significant changes in some mitochondrial (2OGDH) and metastatic (E-cadherin) proteins were observed in TN samples vs. other IDC subtypes (Figs. 4 and 6). Therefore, an alternative therapeutic strategy may be the use of combined treatment including the usual first line of treatment (tamoxifen or 5-fluorouracil or doxorubicin or cyclophosphamide or trastuzumab) plus some mitochondrially-targeted inhibitor (casiopeina II-gly and vitamin E analogues) (50,51).

In conclusion, all analyzed breast cancer subtypes exhibited high HIF-1α levels. Therefore, anti-HIF therapy (i.e., echinomycin and bortezomib) combined with canonical drugs (and/or energy metabolism drugs for TN cases) could be a promising alternative treatment against breast cancer. In this regard, it has been documented that HIF-1α overexpression in cancer cells is linked to a substantial augment in EGFR levels. However, mono-therapy with cetuximab, a monoclonal anti-EGFR antibody or gefitinib yields low efficacy (52), suggesting that combinatory therapy with canonical drugs plus anti-HIF1α therapy (plus energy inhibitors for TN cases) may be required for effective tumor abolishment.