Tumors from kidney of 11 patients (7 males, 4 females) with an average age of 62.2±2.5 years were tested on changes in the gene expression profile of the energy-producing metabolism. Part of the unaffected kidney from the same patient was taken as a corresponding healthy control for the gene profiling. All samples were stored in RNA later. Of these patients, 9 patients were suffering from clear cell renal cell carcinoma, 1 patient suffered from renal oncocytoma and 1 from renal cell B-lymphoma. From ccRCCs, 1 belongs to the nuclear grade I, 1 was nuclear grade III and 1 was nuclear grade IV. All others belong to the nuclear grade II. The ethics committee of the IMPG SAS approved the present study and verbal consent was obtained from these participants prior to surgery.

Tumor and healthy tissues 1–1.5 cm² in size were removed from RNA later, cut into smaller pieces and frozen in liquid nitrogen for 24 h. Tissue was mechanically disrupted using mortar and pestle in liquid nitrogen. Total RNA was extracted from homogenized tissues using GeneJET™ RNA Purification kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions. RNA quality was evaluated using Experion automated electrophoresis system for RNA analysis (Bio-Rad Laboratories) and RNA quantity was measured using NanoDrop ND 2000 (NanoDrop Technologies, LLC, Wilmington, DE, USA). Total RNA degradation/quality was determined based on 18S and 28S ratio. RNA samples, where RQI number (calculated based on 18S/28S ratio) from both, healthy and tumor tissues from the same patient was above 7.5, were selected for gene expression analysis.

Total RNA (500 ng) was transcribed into cDNA, both strands of cDNA were synthesized using (dT) T7-primer. Subsequently labeling reaction was performed using Cy3-dCTP (healthy tissue samples) and Cy5-dCTP (tumor samples) to obtained cRNA. For this purpose, Quick Amp Labeling kit (Agilent Technologies, Santa Clara, CA, USA) was used. After labeling, samples were purified using GeneJE™ RNA Purification kit (Thermo Fisher Scientific) to remove non-incorporated nucleotides. Yield of amplification process and specific activity were determined and only samples with specific activity above 8 proceeded to the hybridization step. A total of 300 ng of appropriate labeled healthy tissue samples and tumor samples were mixed together and cRNA were fragmented by incubation for 30 min at 60°C using components from Gene Expression Hybridization kit (Agilent Technologies). Samples were immediately applied onto SurePrint G3 Human Gene Expression 8×60K v2 Microarray Slide (Agilent Technologies) and hybridized for 17 h at 65°C by rotating slide at 10 rpm in Hybridization Oven (Agilent Technologies). After hybridization two wash steps were performed (Gene Expression Wash Buffer kit; Agilent Technologies) and slide was scanned at 2 μm using NimbleGen MS 200 microarray scanner.

TIFF multiscan images from NimbleGen MS 200 scanner were converted using Feature Extraction software 11.5 (Agilent Technologies), the image processing was performed, and acquired files with spot intensities for every microarray field (corresponding to one patient) were imported into GeneSpring 12.6 GX software to analyze gene expression. Significant differences (fold of expression change ≥2.0) in gene expression were evaluated for each patient separately (non-averaged) as well as averaged from all patients. Pathway analysis was performed to revealed molecular pathways significantly altered in our experiment (P≤0.05). Cluster analysis was performed using GeneSpring 12.6 GX for a selected set of genes.

Dissected tissues were embedded in paraffin according to the standard histological procedure. Sections (4 μm) were placed on polylysine-coated slides, dewaxed and rehydrated. For CA IX detection, the immunostaining procedure was performed on an automated immunostainer (Dako Autostainer) using the DakoCytomation EnVision^®+ System-HRP (DAB) for use with mouse primary antibodies according to the manufacturer’s instructions: a) peroxidase and protein block (10 min each); b) incubation for 1 h with primary antibody (M75 hybridoma medium diluted 1:100 in antibody diluent) or PBS (negative control); c) incubation for 30 min with secondary antibody. Staining was visualized with DAB solution for 1 min with 3,3′-diaminobenzidine as a chromogenic substrate. The slides were washed in PBS with 0.1% Tween-20 for 10 min after step a, twice for 10 min after steps b and c, and three times in distilled water after visualization with DAB.

For HK2 and LDHA immunostaining, antigen retrieval was carried out with citrate buffer, pH 6.0, for 5 min at 125°C using Pascal pressure chamber (Dako). Deparaffinized sections were stained with DakoCytomation EnVision^®+ System-HRP (DAB) for use with rabbit primary antibodies. Primary antibody specific for HXII (1:50; Cell Signaling Technology, Inc., Beverly, MA, USA) and LDHA (1:400; Cell Signaling Technology) was diluted in antibody diluent and incubated overnight at 4°C. Staining was visualized with DAB solution.

All incubations and washings were carried out at room temperature. Finally, the sections were counterstained with Mayer’s hematoxylin, washed for 5 min and were mounted in Aquamont (Merck, Darmstadt, Germany). The stained sections were examined with Leica DM4500 B microscope and photographed with Leica DFC480 camera.

Uncontrolled cell proliferation within a developing tumor often outstrips its blood supply. Consequently, oxygen availability declines, but this physiological constrain is not sufficient to deprive cells of glucose. Prolonged hypoxic state leads to the stabilization of the hypoxia-inducible transcription factor 1 (HIF-1α), which helps cells to adapt to stressful environment by transactivating a broad spectrum of hypoxia-related proteins (8,9). The major molecular event leading to the development of ccRCC is an inactivation or loss of the VHL gene (10–12). VHL protein is part of the E3-ubiquitin ligase complex that binds to the hypoxia-inducible factor subunit α (HIF-α) under normoxic conditions and directs it to proteasomal degradation. Increased levels of the HIF-1 resulting in upregulation of its target proteins equip ccRCC cells with a specific set of enzymatic machinery that favors aerobic glycolysis over oxidative phosphorylation (13). These factors support a shift toward a more glycolytic metabolism by stabilization of the HIF1α and inhibition of prolyl hydroxylases. In order to distinguish, whether tumors of our patients were hypoxic and/or pseudohypoxic, we evaluated changes in the marker of hypoxia carbonic anhydrase IX (CA IX) (14). Expression of the CA IX coding gene was increased in 9 of 11 patients and the average increase was 12.6±1.3-fold (Fig. 1). The 2 patients with no increase in CA IX expression suffered from renal oncocytoma (dark gray column) and B-lymphoma (light gray column), respectively. Moreover, immunohistochemistry proved increased CAIX staining in ccRCC tumor and no signal in renal oncocytoma and B-lymphoma (Fig. 3). Increased transcription of the gene coding for CA IX in ccRCC tumors would suggest that HIF1α is stabilized and active. Also, this observation is in a clear agreement with the work of many other authors who suggested that CA IX is one of the best available markers for the clear cell RCC (15,16).

It is well known that acidification of the tumor microenvironment often develops due to hypoxia-triggered oncogenic metabolism, which leads to the extensive production of lactate, protons and carbon dioxide. Mechanisms of pH regulation in tumor cells are very complex and intertwined with other cancer-related processes (17). Acidosis in tumor microenvironment is compensated by the alkalization of an intracellular pH in cancer cells through the lactate and proton export and the bicarbonate import (17). CA IX is a catalytic component of the bicarbonate import arm, in which it cooperates with bicarbonate transporters (mainly with sodium-bicarbonate cotransporter NBCe1) and regulates pH in response to hypoxia or during cell migration-invasion (18,19). Notably, ccRCC tumors displayed considerable decrease in expression of the genes encoding Cl⁻/HCO₃⁻ anion exchangers AE1, AE3 and AE4, which are considered acid loaders (the most pronounced decrease was observed in AE1 that is normally expressed in intercalated cells of the renal collecting ducts (20). This is probably because cancer cells prefer acid extrusion to acid loading in order to preserve the accumulation of an intracellular acid). No significant change was found in the expression of AE2, in contrast to findings of Karumanchi et al (21), who proposed AE2 as potential VHL target. Furthermore, sodium-bicarbonate co-transporters revealed less marked changes. Expression of NBCe1 was mostly decreased (and this is possibly compensated by the strongly elevated CA IX), while levels of NBCe1 and NBCn1 were increased (the latter being previously associated with breast cancer) (Fig. 1). These data indicate existence of complicated crosstalk between the components of the pH-regulating bicarbonate transport machinery, which is clearly deregulated in ccRCC.

Besides bicarbonate import, intracellular and extracellular pH is dependent on the lactate and proton export. Lactate is transported from the cells by monocarboxylic acid transporters (MCT; SLC16A), transmembrane proteins, which can transport lactate anion across the plasma membrane of tumor cells in association with proton (22). MCT1 (SLC16A1) and MCT4 (SLC16A3) are two isoforms most relevant for cancer physiology. From our cohort, all patients with diagnosed ccRCCs have increased expression of the MCT4 (10.0±1.7-fold), but only 4 of them possess increased MCT1 (Fig. 1). Expression of the MCT4 was observed in many tumors, e.g. triple-negative breast (23), prostate cancer (24) and is always associated with poor prognosis.

HIF1α is known to regulate expression of several enzymes of the glycolytic pathway. To increase the glucose uptake as a way to compensate low yield of ATP from glycolysis, HIF1α promotes the overexpression of the glucose transporter of type 1 (Glut1; SLC2A1) and type 3 (Glut3; SLC2A3). In ccRCCs we observed increase in Glut1 in 8 patients (12.7±2.2-times; Fig. 2) and no change in 1 patient. The Glut1 mRNA was also upregulated in the patient with the renal oncocytoma (Fig. 2, dark gray column) but no change was observed in the patient suffering from cell lymphoma. The Glut3 mRNA (SLC2A3) was also increased in ccRCCs in 7 patients (fold of increase 8.0±1.5; Fig. 2); four tumors displayed no change compared to healthy tissues. Both of these transporters are associated with the tumorigenesis, however, the mechanism, as to which transporter is preferred, and why, is not known. Nevertheless, overexpression of the Glut1 and Glut3 in different tumor types correlates with specific clinical pathological characteristics, malignant potential and poor prognosis (25). Some authors have observed the most prominent Glut1 expression around the necrotic areas or the hypoxic regions of tumors (26,27). Although cancer glucose uptake is thought to be primarily driven by Glut1, recently it was shown that Glut3, but not Glut1, correlates with poor survival of patients with brain tumors and other cancers (28). Notably, mRNA of some other glucose transporters were changed in tumors of our patients, e.g. Glut2 was markedly downregulated in 10 tumors (fold of decrease 9.0±1.5; Fig. 2).

Cancer cells metabolize glucose mostly via glycolysis, even in the presence of sufficient oxygen (5). Therefore, in our patients we focused primarily on pathways producing energy. We observed that glycolytic pathway is affected very significantly in ccRCCs (P=2.5×10⁸; Table I). Hexokinase 2 (HK2) and 3 (HK3) were upregulated in ccRCCs. From 11 patients, gene expression of the HK2 in tumors was increased in 9 patients (fold of increase 34.2±7.5; Fig. 2), in the patient with renal oncocytoma it was downregulated and in 1 ccRCC patient it was not changed. Immunohistochemistry also proved the robust HK2 signal in ccRCC tumors, weak signal in B-lymphoma and almost no signal in renal oncocytoma (Fig. 3). HK3 mRNA was elevated in tumors from 10 patients (fold of increase 5.9±0.9; Fig. 2). HK2, a pivotal glycolytic enzyme is often overexpressed in tumor cells and contributes to glycolysis. Recently, it was found that HK2 protein is increased in cancer-associated fibroblasts (29). Also, current studies demonstrated that HK2 is overexpressed and promotes glycolysis in tumor cells, but not in normal cells (30,31). Moreover, HK2 overexpression is associated with a short progression-free survival, which could be associated with chemoresistance of the epithelial ovarian cancer (32). Conversion of the fructose 6-phosphate to fructose-1,6-bisphosphate is catalyzed by phosphofructokinase (PFK). We observed significant increase in P isoform of the PFK (PFKP), but not in L or M isoforms (Fig. 2). PFKP was shown to play a critical role in cell proliferation in breast cancer cells. It is suggested that a transcription factor KLF4 plays a role in the maintenance of high glycolytic metabolism by transcriptional activation of the PFKP gene in breast cancer cells (33). Phosphoglycerate mutase (PGAM) of type 1 was >2-fold increased, but that of type 2 (PGAM2) was decreased 7.2±1.3-fold in 7 out of 11 patients (Fig. 2). Recently, it was shown that gluconeogenic enzyme fructose-1,6-bisphosphatase 1 (FBP1) was uniformly depleted in over 600 ccRCC tumors examined (34). The human FBP1 locus resides on chromosome 9q22, the loss of which is associated with poor prognosis for ccRCC patients. In our patients, we observed strong decrease of the FBP1 in the ccRCC tumors (13.9±2.5-fold in 9 from 11 patients), which is in line with the results of Li and co-workers (34). While the glycolytic pathway in these tumors is upregulated, gene expression of proteins involved in the gluconeogenesis is suppressed, except of aldolase C (ALDOC), which was significantly increased in 8 out of 9 ccRCC patients (Fig. 2). Pyruvate is the end product of cytosolic glycolysis and has a variety of possible fates. In tumors, majority of the pyruvate is converted to lactate by lactate dehydrogenase (LDH) of type A (LDHA), which was increased 5.7±0.8-fold in 10 tumors. Immunohistochemical staining revealed strong LDHA signal in ccRCC tumor, weak signal in renal oncocytoma and no signal in B-lymphoma (Fig. 3). LDHA upregulation was observed also in the study of Girgis and co-workers (35) on the 170 ccRCC samples. Thus, upregulation of the LDHA could be a predictor of poor prognosis in clear cell renal cell carcinoma. On the other hand, gene expression of the LDHB was significantly suppressed in 8 tumors compared to corresponding healthy tissues (3.5±0.4-fold). Notably, the patient with renal oncocytoma had significantly increased expression of LDHB (Fig. 2, dark column). In human cancers, overexpressed LDHA was associated with the increased aggressiveness (36).

In healthy cells majority of the pyruvate is used for the mitochondrial oxidation. One of the most consistent hallmarks of cancer biology is the preference of tumor cells to derive energy through glycolysis as opposed to the more efficient process of oxidative phosphorylation (OXPHOS). By controlling the mitochondrial flow of pyruvate, a cancer cell can tune its physiology to meet the demands of rapid growth (37). Mitochondrial pyruvate carrier of type 1 (MPC1), but not 2 (MPC2) was slightly decreased in tumors, compared to the matched control tissue. Since pyruvate transport is a rate-limiting step in pyruvate oxidation, downregulation of the MPC1 can dampen its further utilization, e.g. electron transport chain. Indeed, the gene expression of several proteins involved in electron transport chain was significantly decreased (P<0.0044; Table I).

In ccRCC tumors, we observed decrease in the gene expression of proteins in complex I, II, III, IV and V (Fig. 4). Mayr et al (38) showed that loss of respiratory chain complex I (NADH/ubiquinone oxidoreductase) is associated with renal oncocytoma. In ccRCC tumors, we observed rapid decrease in NADH dehydrogenase (ND) type 1, 3, 4, 5 and 4L mRNA compared to healthy part of the tissue (Fig. 4). Cytochrome c oxidase (COX) is a crucial enzyme of the complex IV. We observed decrease in the mRNA levels of COX1, COX2 and COX3, with the highest decrease in the COX3 levels (3.9±0.9-fold; n=7 ccRCC tumors; Fig. 4).

In complex V, ATP5A1 was significantly downregulated in ccRCC tumors. This observation is in agreement with the work of Yusenko and co-workers (39), who observed downregulation of ATP5A1, the α subunit of complex V in chromophobe RCCs. In majority of our patients with ccRCC, ATP5A1 was also significantly decreased (fold-decrease 2.5±0.7; n=6 ccRCC patients; Fig. 4).

Cluster analysis based on the volume difference of tumor/healthy tissue pairs was performed using selected set of genes consisting of all significantly changed genes involved in glycolysis and hypoxia-related metabolic pathways (Fig. 5). Most of the ccRCC samples were clustered together with a highly similar gene expression pattern, except one ccRCC sample, which was more variable. However, oncocytoma sample differs from other samples forming a separate subgroup in the cluster tree. Gene clustering into two main subgroups is based on their upregulation or downregulation in individual samples: all genes of the electron transport chain were clustered together due to downregulation of the whole pathway.

Cancer cells reprogram their metabolism in order to satisfy their bioenergetic and biosynthetic requirements. These cells display reduced ability to use mitochondrial oxidation and favor the conversion of pyruvate into lactate, despite the availability of oxygen. Overproduction of the lactate activates pH regulating transport systems, including CAIX as its catalytic component. Beroukhim and co-workers (11) suggested that the assessment of appropriate markers of the VHL pathway dysregulation (e.g., expression of gene coding for CA IX) may be more useful in predicting response to therapy than assessment of the VHL inactivation status.

To compensate the inefficient extraction of energy from glucose and to maintain the biomass production, malignant cells have an at least a 20- to 30-fold higher rate of glycolysis than normal cells (25). Mitochondrial impairment increases from the less aggressive to the most aggressive RCCs, and correlates with a considerably decreased content of OXPHOS complexes (complexes II, III, and IV of the respiratory chain, and ATPase/ATP synthase) rather than to the mitochondrial content (citrate synthase and mitochondrial (mt)DNA) (40). The present study is among the first showing complex changes in the glycolytic pathway and mitochondrial respiration in tumors of patients suffering from ccRCCs. Moreover, our results nicely correlate with other studies showing strong upregulation of CAIX (41), Glut1 (42), LDHA (35,43) and downregulation of AE1, FBP1 (34) and all complexes of the mitochondrial respiratory chain in ccRCC. Also, these results correlate with changes observed in animal studies and/or cell cultures, e.g., distorted tubules in TRACK (transgenic model of cancer of the kidney) mice exhibit higher levels of CA IX, Glut1, and VEGF than tubules in non-transgenic control mice (44). Specific genes and signaling molecules involved in the tumor-forced glycolysis and related phenomena, such as pH regulation, may represent potential therapeutic targets of agents that specifically interact with the key factors of tumor phenotype. Although some groups (45,46), dispute the role of CA IX as a prognostic marker of the ccRCC, we believe that together with Glut1, Glut3, HK2, FBP1, ALDOC and LDHA, CA IX is an important biomarker for ccRCC diagnosis. Moreover, targeting these proteins might be of therapeutic importance.