PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Cancer Sci. Author manuscript; available in PMC 2013 March 1.
Published in final edited form as:
PMCID: PMC3294166
NIHMSID: NIHMS353986

Pericellular pH homeostasis is a primary function of the Warburg effect; Inversion of metabolic systems to control lactate steady state in tumor cells

2.0 Summary

The Warburg effect describes a heightened propensity of tumor cells to produce lactic acid in the presence or absence of O2. Currently, a generally held notion is that the Warburg effect is related to energy. Using whole-genome, proteomic MALDI-TOF-MS and metabolite analysis, we investigate the Warburg effect in malignant N2a cells. The findings show that the Warburg effect serves a functional role in regulating acidic pericellular pH (pHe), which is mediated by metabolic inversion or a fluctuating dominance between glycolytic-rate-substrate level phosphorylation (SLP) and mitochondrial (mt) oxidative phosphorylation (OXPHOS) to control lactic acid. Alkaline pHe elevated SLP/OXPHOS ratio (approximately 98% SLP/OXPHOS); vs. neutral (approximately 56% SLP/OXPHOS) or acidic (approximately 93 % OXPHOS/SLP). Acidic pHe paralleled greater expression of mitochondrial biogenesis and OXPHOS genes, such as complex III–V (Uqcr10, Atp5, and Cox7c), mt Fmc1, Romo1, Tmem 173, Tomm6, aldehyde dehydrogenase, mt Sod2 adjunct to loss of mt fission (Mff) and mt biogenesis component PPAR-γ co-activator 1. Moreover, acidic pHe corresponded to metabolic efficiency evidenced by a rise in mTOR nutrient sensor GßL, its downstream target (Eif4ebp1), insulin modulators (Trib3,Fetub) and loss of catabolic (Hadhb,Bdh1,Pygl) / glycolytic processes (aldolase C, pyruvate kinase, Nampt and aldose-reductase). In contrast, alkaline pHe initiated loss of mitofusin 2, complex II-IV (Sdhaf1,Uqcrq,Cox4i2,Aldh1l2), aconitase, mitochondrial carrier triple repeat 1 and mt biosynthetic (Coq2,Coq5,Coq9). In conclusion, the Warburg effect may serve as negative feed-back loop which regulates the pHe toward a broad acidic range by altering lactic acid production through inversion of metabolic systems. These effects were independent of change in O2 concentration or glucose supply.

Keywords: Warburg, acidity, tumor, lactic acid, hypoxia, pH

3.0 Introduction

Malignant tumor cells display an accelerated rate of glucose consumption with dominant reliance on substrate level phosphorylation (SLP) rather than oxidative phosphorylation (OXPHOS) to produce ATP. Heightened glycolytic activity is typically indicative of a metabolic response to lack of oxygen (O2), but in cancer cells this pattern occurs in the presence or absence of O2 with accumulation of lactic acid (Warburg effect). This aberrant metabolic pattern differs from normal tissue and thereby serves useful for diagnostics purposes as in the case of monitoring uptake of 18F-fluorodeoxy-glucose1 Both accelerated rate of glycolysis and greater expression of the lactic acid dehydrogenase (LDH) are characteristic of aggressive tumor cells 2, therefore, research aimed at unveiling the underlying purpose of the Warburg effect could be beneficial.

To date, the Warburg effect continues to remain somewhat of a mystery. A recent review by Mentis and Kararizou, 2010 suggests that diverse held concepts currently exist about its etiology, some of which include the involvement of oncogenes (Akt, Myc and Ras), mutations of the tricarboxylic acid cycle (TCA), alternative splicing of pyruvate kinase pre-mRNA and epigenetic changes.3 Other studies suggest that the Warburg effect is involved with aberrant mitochondrial replication 4, post-translational modification of glycolytic enzymes (ie 6-phosphofructo-1-kinase)5, biomass or cell crowding 6, 7 mammalian target of rapamycin (RTK/PI3K/AKT/mTOR) signaling 8 and lactate production as a potential utilizable substrate for adjacent cells.9 However, by and large the general conclusion still remains that the Warburg effect is ultimately related to energy, an inefficient metabolic process,10 reliant on diverse ergogenic substrates to produce energy under hypoxia or glucose deprivation.11 While the Warburg effect may be related to energy, it is also clear that accumulation of lactate maintains a presiding influence over the acidic pericellular pH (pHe) circumscribing aggressive tumors. In turn, the presence of lactate is known to trigger aggressive forms of malignancy, augment metastases, chemoresistance and correlate to low survival rates. 12, 13 Therefore it is unclear if the Warburg effect plays a primary role in energy or maintaining an acidic pHe of the tumor microenvironment.

Physiological pH homeostasis is intricately regulated to processes involving both cellular metabolism and the ppO2 /CO2 14. Therefore, rigid experimental controls should be set in place to eliminate interfering variables that indirectly alter pHe when studying the effect of hypoxia on tumor cells or the Warburg effect. Changes in pHe are introduced by factors such as the gas composition itself (O2, N2, CO2), metabolic response to the change in gas (lactate or CO2 released in a closed system), selected volume/depth of media, cell plating density, inherent metabolic rates or choice of buffering agents in the media, which could mask true effects 15. Either studies should utilize a means to control the pH to determine the true effect of gases, or should employ a means to hold the gases constant in order to study the effect of pH change.

The purpose of this study is to investigate metabolic patterns in response to experimentally induced pHe shifts extending from acidic to basic under a uniform gas condition. The findings from this study show that the Warburg effect (heightened glycolysis and lactic acid production under adequate O2 concentration) appears to serve as dynamic flexible negative feedback loop which regulates the pH of the microenvironment toward a broad acidic range in N2a cells. This effect is mediated by inversion of anaerobic - aerobic respiratory metabolic systems, which indirectly control lactic acid. Moreover, the findings in this study show that metabolic buffering by N2a cells involves molecular changes within the mitochondria, mTOR signaling and glycolysis.

4.0 Materials and Methods

Nitrocellulose/ PVDF membranes, 10x Tris/ glycine buffer, electrophoresis supplies (Biorad Hercules, CA), 1° antibodies raised in rabbit or mouse, 2° IgG - HRP conjugated antibodies were purchased from Ab-cam (Cambridge, MA) and Assay Biotech, (Sunnyvale, CA) Dulbecco’s modified Eagle medium (DMEM), L-glutamine, fetal bovine serum (FBS) heat-inactivated, phosphate buffered saline (PBS), Hank’s balanced salt solution (HBSS), penicillin/streptomycin, and all other materials were purchased from Sigma Chemical (St. Louis, MO, USA).

Cell culture

CCL-131™ Neuro-2a (N2a) malignant neuroblastoma cells were obtained from ATCC (Manassas, VA). N2a cells were grown in DMEM containing phenol red, 10% FBS, 4 mM L-glutamine, 20μM sodium pyruvate and penicillin/streptomycin (100 U /0.1 mg/ml). The cells were maintained at 37°C in 5% CO2/atmosphere. Every 2–5 days, the medium was replaced and the cells were sub-cultured. The experimental plating media consisted of DMEM, 1.8% FBS, penicillin/streptomycin (100 U/0.1 mg/ml), / 2 mM sodium pyruvate and 3 mM L-glutamine. For experiments, cells were plated in 96-well plates or 75 cm3 flasks at a density of ~0.5×106 cells/ml.

Experimental Design

In order to evaluate cellular toxicity under either 100% N2 or 100% O2, gases were delivered into a closed system where media was pre-equilibrated and pH was neutralized to 7.2, prior to addition of cells. Cells were then incubated in a closed chamber for 24 hours at 37°C with pH maintained at 7.2. In order to investigate the influence of pHe on N2a cells under uniform gas conditions (5% CO2/Atm), treatments were suited to accommodate 48 micro-incremental pHe shifts using (0.0002-0.02N HCl / 0.004-0.08 N NaOH = final working concentrations) in 96 well plates. Media blanks for each concentration were established, and pH (e) plate readings were taken after addition of the cells and at 24 hours to assess differences in pHe (initial) and pHe (final). Experiments requiring large amount of mRNA or proteins for molecular analysis were carried out in 75 cm3 flasks, where pHe was sustained in order to ensure cells were viable, pHe acidic (~6.8), neutral (~7.3) and alkaline (~7.55).

Cell Viability

Cell viability was assessed using resazurin oxidoreduction indicator dye. A working solution of resazurin was prepared in sterile HBSS minus phenol red (0.5 mg/ml)16. Reduction of the dye by viable cells reduces the amount of oxidized form and increases the amount of its bright red fluorescent intermediate. Quantitative analysis of dye conversion was measured on a microplate fluorometer—Model 7620-version 5.02 (Cambridge Technologies Inc, Watertown, MA, USA) set at 550/580 (excitation/emission). The data were expressed as % live control.

Determination of H+ concentration

Extra-cellular pH was determined using a standard phenol red indicator dye and confirmed using a HANNA handheld portable pH checker (Sigma Chemical, MO, USA). The speed at which pH was determined was critical due to both composition of air and temperature affecting results, therefore each plate was tested immediately upon exit of the incubator. Briefly, a phenol red stock solution (0.3 mg/ml water) was added (15% v/v) to the cell supernatants after 24 h. The change in pH was immediately assessed at 550 nm using a Spectra 190-MAX UV spectrophotometric detector (Molecular devices, Sunnydale, CA, USA).

HPLC Quantification of Lactic Acid and Glucose

Determination of lactate and glucose concentrations were acquired using a Shimadzu HPLC system equipped with an SPD-20A UV detector (set at 210 nm), a RID-10A 120V refractive index detector, a workstation containing EZSTART version 7.4 software and an SS420X instrument interface docked to a Waters Autosampler Model 717 Plus (Shimadzu Scientific Instruments, Inc. US; Waters Corp., Milford, MA). The flow rate was isocratic, being controlled by a Waters Model 510 pump at 0.6 ml/min. The mobile phase consisted of 5 mM sulfuric acid, the column; aminex HPX-87H 300 × 7.8 mm, carbohydrate analysis column, 9 μm particle size (Biorad Hercules, CA), run time was 16 min and injection volume 25 μl. Samples were prepared by placing 35 μl cell supernatant into 200ul of 5mM sulfuric acid, immediately stored at −80C. Prior to analysis, samples were thawed and 125 μl was added to 275 μl of 5mM sulfuric acid. Glucose and lactate standard curves were established from preparations in distilled water and matrix blank controls and spikes were run for every experimental treatment condition tested.

Whole genome expression profiling

Whole genome expression profiling was carried out on Agilent Mouse or 4 × 44 k arrays (Beckman Coulter Genomics, Morrisville, N.C.) from total RNA isolated from each sample. Briefly, the quantity of total RNA was determined by spectrophotometry [A260/280 ratio] and the size distribution was assessed by electropherogram using an Agilent Bioanalyzer. 200 ng of total RNA was converted into labeled cRNA with nucleotides coupled to fluorescent Cy3 dye using a Low Input Quick Amp Kit (Agilent Technologies, Palo Alto, CA) following manufacturer’s protocol. Cy3-labeled cRNA (1.65 μg) from each sample was hybridized to an Agilent Mouse Genome 4×44 k array. The hybridized array was then washed and scanned and data were extracted from the scanned image using Feature Extraction version 10.7 software (Agilent Technologies).

The data was analyzed by both Gene Sifter and manual analysis. Manual analysis was achieved by normalizing the raw data gProcessed signal to the average signal/sample hybridization for biological triplicates. Subsequently a filtering of noise was established by omitting any gene below a threshold limit of gProcessed >500 so to omit false positives that could arise from noise/low expression abundance. The ratios for each group were calculated, p-values determined by a students t-test and data analyzed by manual 1) examination of individual genes accounting for potential duplicates in the array using diverse primers for same gene, or by QC replicates 2) sorting and analysis by greatest difference in intensity (expression dominance) with significance p<0.05 and 3) then fold change. Next, the combined fold-changes, p-values for each set of hybridizations were classified manually by literature review and differential statistically relevant genes were entered into the “database for annotation, visualization and integrated discovery” (DAVID) where patterns by enrichment scores averaging less than p<0.01 into the discussion were focused on.17, 18

O2 Consumption—Clark Electrode

A Hanna HI 9142 O2 meter was used to measure dissolved oxygen as an indicator of cellular and mitochondrial respiration. The electrode was calibrated with both air saturated de-ionized water and de-ionized water with sodium dithionite. Briefly, 800 μl of cell supernatant vs. blank controls were directly loaded into a small chamber. After rate equilibration, a 30 s reading was taken for each sample. The temperature was maintained at 37°C.

Western Blot

Briefly, after treatment, cells were washed, centrifuged and the supernatant discarded using ice cold sterile PBS at 4° C. The pellet was re-suspended and homogenized/ sonicated in a lysis buffer containing protease inhibitors. Samples were placed on ice for 30 min, and centrifuged at 10,000 × g for 10 minutes at 4°C. The supernatant was assessed for protein concentration and added to 1:1 of Laemmli Sample Buffer (Biorad #161-0737) + fresh β-ME and boiled for 5 minutes. Approximately 50 μg of protein was loaded / lane and separated using 5%-15% SDS-PAGE gels, running buffer, 25 mM Tris, 192 mM glycine, pH 8.3 (Biorad #161-0734) and applying 200 constant V constant ~ 35 min. The proteins were transferred to polyvinylidene fluoride membranes (100V for 30-60 minutes), in ice cold transfer buffer containing 25mM Tris, 192mM glycine and 20% methanol. The membranes were placed in a blocking buffer consisting of 5% bovine serum albumin Fraction V (BSA) w/v in Tween / PBS buffer (TTBS) 0.01 M phosphate buffer, 0.0027 M KCl, 0.14 M NaCl, 0.05% Tween, pH 7.4. The membranes were washed and placed in 1° antibody (1:500-3000) 1% BSA in TTBS and 0.2% sodium azide at 4°C for 24 h. The membranes washed in TTBS and incubated in 2° anti-mouse/ or rabbit IgG (Fc specific) peroxidase conjugate (1:4000) in 2% non fat dried milk on TTBS for 1 h at room temperature. After a final wash, peroxidase was detected with Sigma FAST™ DAB (3, 3′-diaminobenzidine tetrahydrochloride) with a metal enhancer cobalt chloride. Images were scanned using an Epson Stylus CX-8400.

High-Resolution Two-Dimensional Gel Electrophoresis

Cell lysates (150 μg of protein) were solubilized for 30 min with 2D rehydration/ sample buffer (7 M urea, 2 M thiourea, 1% ASB-14, 40 mM Tris) and 2% immobilized pH gradient (IPG) buffer pH 3–10 (Bio-Rad). Proteins were absorbed and separated by charge overnight using ReadyStrip™ IPG Strips, pH 3-10 NL and focused by isoelectric point with the PROTEAN IEF Cell with [System settings: 250 V rapid voltage ramping - 30 min, 10,000 V slow voltage ramping - 60 min, and 10,000 V rapid voltage ramping - 50 kV hours]. The strips were incubated in equilibration buffer I with 6 M urea, 20% glycerol, 2% SDS, 2% DTT, and 0.375 M Tris, pH 8.8, for 10 min at room temperature, then in equilibration buffer II with 6 M urea, 20% glycerol, 2% SDS, 2% iodoacetamide, and 0.375 M Tris, pH 8.8. The strips were then loaded onto 6-18% SDS-PAGE gels on 13.3 × 8.7 cm and run at 50 V overnight. ReadyPrep overlay agarose was added on top of the strip to secure it and included bromophenol blue tracking dye. A molecular standard was used to estimate relative mass (Mr). Gels were pre-rinsed with water, stained overnight with Bio-Safe Coomassie, destained in water, and scanned with the Versadoc Model 1000 system (Bio-Rad, CA). Gel image analyses were performed with PD Quest software (Bio-Rad) version 7.4.0. Individual spot volumes for each gel were normalized relative to the total spot volume of that gel. Normalized spot volume data from each experimental set were analyzed the Student’s t-test (P < 0.05 was regarded as significant). Spots 2-fold higher/lower were considered to be differentially regulated. All experiments were performed in biological and technical triplicates starting with the collection of cells.

Protein identification by Peptide Mass Fingerprinting (PMF)

Protein spots were excised and tryptic digests were analyzed by LC-MS/MS. Briefly, protein spots from 2D gel were de-stained with 50% acetonitrile in 50 mM ammonium carbonate. In-gel tryptic digest was performed using reductively methylated trypsin (Promega, Madison, WI). Before digestion, samples were reduced with DTT (10 mM in 50 mM ammonium carbonate for 60 min at 56°C) and subsequently alkylated with iodoacetamide (55 mM in 50 mM ammonium carbonate for 45 min in the dark at room temperature). The digestion reaction was performed overnight at 37°C. Digestion products were extracted from the gel followed by evaporation using an APD SpeedVac (ThermoSavant). The dried tryptic digest samples were cleaned with ZipTip (CB18B; Millipore) before analysis by tandem mass spectrometry for protein identification. The digested sample was resuspended in 10 μl of 60% acetic acid, injected via autosampler (Surveyor; ThermoFinnigan, San Jose, CA) and subjected to reverse phase liquid chromatography using Thermo Finnigan Surveyor MS Pump in conjunction with a BioBasic-18 100 × 0.18 mm reverse-phase capillary column (Thermo Finnigan). Mass analysis was done using a Thermo Finnigan LCQ Deca XP Plusion trap mass spectrometer equipped with a nanospray ion source using a 4.5 cm long metal needle (Hamilton; 950– 00954) in a data-dependent acquisition mode. Electrical contact and voltage application to the probe tip took place via the nanoprobe assembly. Spray voltage of the mass spectrometer was set to 2.9 kV and heated capillary temperature at 190°C. The column was equilibrated for 5 min at 1.5 μl/min with 95% solution A and 5% solution B (A, 0.1% formic acid in water; B, 0.1% formic acid in acetonitrile) before sample injection. A linear gradient was initiated 5 min after sample injection ramping to 35% A and 65% B after 50 min and 20% A and 80% B after 60 min. Mass spectra were acquired in the mass-to-charge ratio 400–1800 range. Protein identification was performed with the MS/MS search software Mascot 1.9 (Matrix Science, Boston, MA) with confirmatory or complementary analyses with TurboSequest as implemented in the Bioworks Browser 3.2, build 41 (Thermo Finnigan).

Data analysis

Statistical analysis was performed using Graph Pad Prism (version 3.0; Graph Pad Software Inc. San Diego, CA, USA) with significance of difference between the groups assessed using a one-way ANOVA, followed by Tukey post hoc means comparison test, a two way ANOVA or Student’s t test.

5.0 Results

Tumor cells display resistance to hypoxia or mitochondrial toxins, provided ample glucose supply is sustained to drive SLP production of ATP. 19 In contrast, high levels of O2 reportedly augment the effects of chemotherapy and attenuate tumor growth in rodents. 20 Our studies show that tumor cells appear capable of surviving in diverse ambient gases including high levels of O2, where two vulnerable limitations were observed: 1) a rise of pHe toward alkaline (induced from the gases themselves) or 2) the depletion of glucose that arises from hypoxic accelerated glycolysis. In this study, gas injection of either 100 % O2 or 100 % N2 into a closed chamber rendered alkaline pHe shifts in the media alone that reached pHe in excess of 7.8. If pHe was not pre-neutralized prior to addition of N2a cells, we saw a loss of viability (data not shown). Here, we demonstrate that N2a cells continue to remain viable when cultured under hypoxic or hyperoxic conditions for 24 hours, provided that the media remained continually buffered ~ pHe (7.2) and ample glucose supply was maintained– Figure 1. While level of O2 is not so much a factor to tumor survival, alkaline pHe is. A toxicity profile (Figure 2) establishes tolerance of N2a cells to broad acidic pHe range, where a steep loss of viability occurs as pHe exceeds 7.6.

Figure 1
Toxicity profile of N2a cells in response to hyperoxia or hypoxia at a sustained pHe 7.2
Figure 2
Toxicity profile of N2a cells in response to variation in pHe after 24 hr incubation

Given the loss of viability with minuscule elevation in pHe above neutral, we examined for metabolic patterns across a broad spectrum pHe (Figure 3). The use of an aminex HPX-87H carbohydrate analysis column and dual detection systems in HPLC application, allowed for simultaneous quantification of a range of metabolites including organic acids, carbohydrates and fermentation products within the same sample. The data show a pHe dependent rise in lactic acid occurred at alkaline pHe in a linear fashion, even as the cells were approaching death. An equal but opposite pattern emerged under acidic conditions (Figure 3) also presented as raw data chart readings (RI) and (UV-210 nm) (Figures 4a, 4b). These findings suggest that the metabolism of N2a cells and potentially other tumor cells, could fluctuate from aerobic (acidic pHe) to anaerobic (alkaline pHe), not in response to change of energy demands or O2 supply, but rather to neutralize pHe toward a slight acidic by altering lactic acid production.

Figure 3
Effects of pHe on glucose to lactate conversion in N2a cells
Figure 4Figure 4
a. Glycolytic lactate production in response to pHe shift from acidic to alkaline in N2a cells at 24 hrs. The data represent the raw data chart recording traces from the SPD-20A UV detector (210 nm) quantifying lactic acid produced.

In order to evaluate if the control of pHe on glucose / lactate conversion corresponds to a proclivity to buffer pHe, we compare profiles between media blank controls containing HCl to NaOH, versus cell supernatant acquired at 24 hrs (Figure 5a, b). The data show that N2a cells will buffer externally introduced pHe shifts in both directions (acidic pHe or neutral/alkaline toward slight acidic). Moreover, Figure 5b (right hand panel) shows a gradually lessened capability of N2a cells to buffer alkaline pHe, likely due to cell death and or exhaustion of energy supplies. Clearly, cross points within the buffering curves occurring under acidic and alkaline pHe (5a, b right hand panels) suggest other molecular contributions are involved with buffering to complement that described. Further investigation will be required to elucidate these systems.

Figure 5Figure 5
The effect of acidic pHe (A) or alkaline pHe (B) on buffering capability of N2a at 24 hrs

In order to understand the molecular events associated with pHe mediated metabolic inversion, expression profiles for proteins of interest were analyzed by western blot (Figure 6). The data obtained show a slight positive change in the expression of LDH from acidic toward basic pHe. However, this change is minuscule and not likely to account for extensive metabolic inversion profiles as shown in Figures 3--4.4. Similarly, the data also show slight changes from acidic pHe toward basic for glycogen synthase kinase-3 alpha (which inactivates glycogen synthase rendering greater glucose availability) and fructose-bisphosphate aldolase A (which breaks down fructose 1,6-bisphosphate into glyceraldehyde 3-phosphate and dihydroxyacetone phosphate in glycolysis). While these changes may contribute to pHe controlled metabolic inversion, the changes are meager suggesting additional processes are likely to be involved. Of the proteins evaluated, acidic pHe rendered significant loss of UDP glucose P2Y purinoceptor 14. While further research will be required to investigate a role for this receptor, it is believed that UDP-glucose is released upon stress or cell damage and acts on P2Y14 to initiate signaling 21 which results in inhibition of adenylyl cyclase, reduced conversion of ATP to cAMP 22 and accumulation of inositol phosphates, effects which may influence the rate of glycolysis in diverse cell systems. 22, 23

Figure 6
Relative expression for basic proteins of interest in N2a cells cultured under diverse pHe acidic (~6.8), neutral (~7.3) and alkaline (~7.55) for 24 hrs.

Cell lysates were further subjected to 2-D gel electrophoresis LC-MS/MS proteomic analysis. Because it was not feasible to sequence all resolved proteins, the selection criteria for sequencing was based on gel image quantitative analysis of the 2-D analysis (Table 1). Of over 300 protein spots, several met the criterion for a 2-fold or greater significance, p < 0.05. LC-MS/MS and Mascot database matching resulted in successful identification of proteins with spot densities over 3 gels per condition (p < 0.05) are listed in Table 1. The data show acidic pHe mediated down-regulation of aldolase C and pyruvate kinase (rate limiting control in glycolysis) (Figure 7) and a rise of the acidic mitochondrial matrix protein p32 (complement component C1q), which is believed to play a role in mitochondrial biogenesis and OXPHOS.24 These events correspond to reduction of lactic acid as shown in the previous Figures 3, 4a, 4b.

Figure 7
Multichannel [Left] and 3=D view [Right] of aldolase C [A] (red arrows) and pyruvate kinase [B] (black arrows) exhibiting differential expression in N2a cells under pHe acidic (~6.8) vs. neutral Controls (~7.3).
Table 1
Identification of proteins by peptide mass fingerprinting

Whole genome expression profiling was carried out on labeled cRNA coupled to fluorescent Cy3 hybridized to an Agilent Mouse Genome 4×44 k array (Beckman Coulter Genomics, Morrisville, NC). Differential profiles with statistical significance <0.05 in groups acidic pHe vs. alkaline pHe were entered into the “database for annotation, visualization and integrated discovery” (DAVID) where patterns by enrichment scores are presented Table 2. 17, 18 Although not the focus of this study, the most striking observation was an acidic to alkaline pHe change in epigenetic reprogramming. These events signify adaptive change requirements occur as the pHe shifts toward alkaline, which involved in excess of 100 genes involved with chromatin reorganization. In terms of metabolic gene patterns, influences on the mitochondria were identified where alkaline pHe elevated 52 genes with an enrichment score in excess of 2.0 and a p-value of 1.6 E-08. Of these, genes involved with greater OXPHOS under acidic pHe vs. alkaline pHe are shown in Figure 8 overlapped on a KEGG diagram. 25-27 Respectively, 80 gene shifts occurred in the opposite direction for the mitochondria (p-value 3.4E-04), however these were not related to OXPHOS but rather processes involved with carbohydrate, fatty acid, mitochondrial or related metabolic activity (Table S1, Table S2). In summary, these findings show preliminary evidence that alkaline pHe evokes reduction of processes related to mitochondrial respiration or stability, and a rise in processes associated with catabolic breakdown of fatty acids or glycogen to provide energy. In contrast, acidic pHe vs. control or neutral was associated with greater mitochondrial biogenesis, fusion and OXPHOS function. The pHe control over inversion of metabolic systems correlated to direct neutralizing effects through lactic acid production.

Figure 8
KEGG pathway mapping overlay for acidic pHe upregulated genes pertinent to OXPHOS
Table 2
Gene-annotation enrichment analysis, functional annotation clustering on statistically differentially expressed genes in N2a cells cultured under pHe acidic (~6.8) vs alkaline (~7.55) for 24 hrs

6.0 Discussion

The data from this study provide initial evidence that a prime function of the Warburg effect may extend beyond inefficient or deviant energy metabolism. These findings demonstrate that micro-environmental pHe exerts control over the extent of glucose/lactate conversion (SLP); through changing mitochondrial respiration (OXPHOS) capacity in a manner alters production of lactic acid as a means to control pH. Moreover, these effects were not mediated by availability of glucose, change in O2 concentration and did not correlate to detectable changes in either protein or gene expression of hypoxia inducible factor - 1 α. These patterns are similar to respiratory rate control effects on physiological buffering systems, where metabolic acidosis/ alkalosis correspond to changes in respiration as a compensatory buffering response.

In this study, a number of observations give the impression that a slight shift towards acidic pHe is associated with attenuated glucose use, less production of lactate and diverse molecular changes that reflect greater energy efficiency. These changes include the attenuated protein expression of LDH, pyruvate kinase (rate limiting control); aldolase C and glycolytic genes; enolase (Eno1), galactokinase 2 (Galk2), lactate dehydrogenase A (Ldha), phosphoglycerate mutase 1, triosephosphate isomerase 1 (Tpi1) aldose reductase (Gm8531, Akr1b3, Akr1b10), and nicotinate phosphoribosyltransferase domain containing 1 (Naprt1) tantamount to elevation of pyruvate - alanine transaminase (Gpt2). Of these glycolytic metabolic processes, Gpt2 has the greatest capacity for competitive substrate docking to pyruvate in a reversible enzyme reaction which converts pyruvate to L-alanine and α-ketoglutarate. This metabolic pathway occurs in diverse tumor cell lines 28 where C14 tracing studies show that this enzyme converts pyruvate to alanine, perhaps to a greater extent of pyruvate to lactate 29 suggesting a functional role which may serve useful under diverse environmental conditions. High levels of alanine are known to inhibit pyruvate kinase activity, in particular in aggressive stage tumor cells30. Although future research will be required, it is possible that alanine may play a role in negative feedback on pyruvate kinase under acidic pHe.

Symbiotic effects of acidic pHe appeared to involve positive changes within the mitochondria which could yield greater use of pyruvate through TCA cycle for aerobic energy production. The data show acidic pHe corresponded to a rise in mitochondrial reactive oxygen species modulator 1 (Romo1), which is reportedly involved with tumor cell proliferation31, “the formation of mitochondrial complexes 1 homolog” (Fmc1), mitochondrial aldehyde dehydrogenase (Aldh18a1), ATP synthase (Atp5g1), cytochrome c oxidase (Cox7c) and transmembrane protein 173 (Tmem173). These genes encode for a number of proteins involved with the function and synthesis of mitochondrial components. In contrast, acidic pHe initiated loss of gene expression for the mitochondrial fission factor (Mff), an effect that could render greater stability and OXPHOS capacity. Mff plays a negative role on mitochondria by altering size, shape and morphology in a process which is opposite to fusion.32 A loss of Mff as in the case of acidity - would prevent fission mediated inactivation through mitochondrial fragmentation and separation 33 due to lost tethering by dynamin proteins. 34, 35 Conversely, alkaline pHe initiated a loss of mitofusin 2 (Mfn2), which would lessen the adhesive tethering of mitochondria, rending inadequate exchange of proteins, mtDNA and required elements to maintain a strong constitution.36 Under alkaline culture conditions, the mitochondria appear vulnerable or unstable, evidence by diverse forces including the loss of HCLS1 associated X-1 (Hax1). The HAX1 gene encodes for a protein called HS-1-associated protein X-1 (HAX-1) located in nuclear matrix and the mitochondria, which blocks apoptosis through countering high concentrations of Ca+2, within the mitochondrial matrix 37 and sarcoplasmic reticulum 38 and inhibiting caspase-3.39 A loss of stability within mitochondria due to subsequent Ca 2+- overload under alkaline pH would lead to greater likelihood of permeability transition pore opening and thus cell death.40 The data in this study demonstrate onset of cell death as pHe rises above 7.4.

These findings show a potential ergonomic molecular advantage under an acidic pHe. As example, the data show elevation of SCAN domain-containing 1 (Scand1), which encodes a SCAN box domain-containing protein also known as peroxisome proliferator-activated receptor-γ coactivator 1-α (PGC-1α). This transcriptional co-activator is a functional relative of PRC believed to contribute to mitochondrial biogenesis, yielding higher OXPHOS capacity. 41 PRC binds to and activates nuclear respiratory factor 1 (NRF-1) which if silenced results in ablated OXPHOS capacity. 42, 43 The activation of PGC-1α may have a role in augmenting the expression of the insulin nutrient sensor; tribbles homolog 3 (trb3) under acidic condition44. TRB3, protein levels are influenced by high glucose concentrations, often observed in association with type 2 diabetic patients, streptozotocin -diabetic rats, Zucker fatty rats, and a loss of insulin-stimulated glucose transport. 45 A number of metabolic shifts under acidity indicate change toward reduced insulin sensitivity, greater glucose availability and maximized nutrient efficiency.

Cells cultured under acidic pHe also showed differential elevated protein expression for nutrient sensor GßL (G protein, beta protein subunit-like) a component of mTOR (mammalian target of rapamycin) PI3K /Akt signaling and its down stream eIF4E tumor promoting target. 46 These changes are indicative that feedback sensors are receiving information to switch from anabolic to catabolic. 47 The high levels of glucose present in the acidic media or a rise in cytosolic ATP/AMP would inactivate AMP-activated protein kinase and activate mTOR signaling. 48 Alkaline shift associated with a low nutrient signal, would accompany higher levels of catabolic processes in order to provide adequate fuel in the form of ATP through SLP, so that lactate production can neutralize the rise in any alkaline pHe challenge. The equilibrium that controls mTOR signaling is maintained by adenylate kinase within the mitochondrial intermembrane space, and corresponding pH dependent shifts were also observed. These findings, while broad, also show that acidic pHe lessens expression of a large number of genes associated with catabolic energy yielding processes; hydroxyacyl-Coenzyme A dehydrogenase/3-ketoacyl-Coenzyme A thiolase/enoyl-Coenzyme A hydratase (trifunctional protein), beta subunit (Hadhb), carnitine palmitoyltransferase and 3-hydroxybutyrate dehydrogenase, type 1 (Bdh1) in addition to glycogen phosphorylase (Pygl). Reduction of glycogen phosphorylase is again indicative of abundant energy supplies (in acidic media), where a slower degradation of glycogen breakdown would be required to supply glucose-1-phosphate to glycolysis. Tumor cells in general are wasteful (or do so to maintain lactate); however this means a higher reliance on fuel storages such as glycogen. Inhibition of these energy sources, as the case with CP-320626, will reduce cancer growth. 49 Oddly, acidity was associated with a consistent loss of lipid biosynthetic processes across the spectrum, including that for stearoyl-Coenzyme A desaturase 2 (Scd2), fatty acid desaturase 1 (Fads1) as well as Gm8163 farnesyl pyrophosphate synthase (FPS). It is difficult at this time to ascertain why this response occurred.

Major effects of alkaline pHe on genes than encode proteins involved with mitochondrial function involved losses of ubiquinol-cytochrome c reductase, complex III subunit VII (Uqcrq), cytochrome c oxidase subunit IV isoform 2 (Cox4i2), cytochrome c oxidase, subunit VIIc (Cox7c), ubiquinol-cytochrome c reductase, complex III subunit X (Uqcr10), succinate dehydrogenase complex assembly factor 1 (Sdhaf1), flavin adenine dinucleotide synthetase, homolog (yeast) (Flad1) and mitochondrial coenzyme Q homologs (yeast). These effects accompanied reduction of aconitase 2, mitochondrial (Aco2) and mitochondrial glutaminase 2 (Gls2) which regulate the first series of events and transaminations that branch from the TCA cycle. Other mitochondrial enzymes adversely affected by alkalinity include aldehyde dehydrogenase 18 family, member A1 (Aldh18a1) ALDH. ALDH is regulated by TGF-β signaling and if silenced will lead to loss of CD44(+) stem/progenitor migration and invasion properties 50, where elevation is associated with tumor growth.51 A pHe linear response was observed also for cytosolic (aldose reductase) (AR) (Akr1b8) (Akr1b3) which is likely to play a role in glycolysis. AR inhibitors show potential therapeutic value against solid hypoxic tumors, demonstrating block of metastatic invasion mediated by lessening signaling transduction of NF-kB-AP1. 52 However, AR is component of the polyol pathway, which catalyzes the NADPH-mediated reduction of glucose to sorbitol for re-entrance into glycolysis where it conserves energy and may account for up to 25–30% of the glucose metabolism.53 And, while AR inhibitors or gene silencing show anti-angiogenic properties and a loss of adhesion molecules ICAM, and VCAM.24, this enzyme is likely to play a role in glycolysis, although future research will be required to evaluate this. The data also show that acidic pHe lead to rise of the acidic mitochondrial matrix protein p32, (complement component C1q.). While little is know about this protein, it is harbored within the mitochondrial matrix with potential role in both OXPHOS and nuclear trafficking.24 It is also likely that its prime function may extend beyond metabolism and alter cytoskeletal organization or migration properties of human cancers.54

In addition, a few anomalies were noted of which includes the effects of pH on peroxiredoxin-thioredoxins. These findings stand out. Although these proteins are known to play an important role in reactive oxidative stress, they are also required for iron-sulfur cluster biosynthesis assembling [2Fe-2S], iron homeostasis and iron-responsive element–binding (IRE-binding) activity of iron regulatory protein 1 IRP1. Similarly, acid can lead to loss of iron-sulfur cluster assembly 1 homolog (S. cerevisiae) (Isca1). Future research will be required to examine if these proteins have anything to do oxidative reduction reactions required to fuel SLP /anaerobic metabolism or potentially the de novo synthesis of nucleotides. Other noted trends included an alkaline induced ± acidic suppressive effect on epigenetic chromatin remodeling genes; (Hmgn2, Hmgb2, Hmgn1, H2afv, H2afz, H3A, and Phf5a), nuclear pore trafficking (Kpna2, Nutf2, nucleoporins 35, 37, 54, 62,107,133), and nucleophosmin 1. While outside the scope of this paper, it is believed that rapid epigenetic adaptation may be involved with greater ability of cancer cells to adapt to diverse conditions. Further, a number of epigenetic modifying drugs, such as 5-azacytidine (DNA methyltransferase inhibitors) (i.e. vidaza, decitabine) deazaneplanocin A (histone methyltransferase inhibitor) or belinostat (-HDAC) are now being evaluated for treatment for diseases including cancer. 54, 55 The epigenetic activity associated with rise in pHe toward alkaline were consistently evidenced at the genomic level in this study (Table 2).

In conclusion, this work provides preliminary basis to suggest that the Warburg effect may expand beyond its role in energy metabolism and could be a primary means by which tumor cells can regulate the pace of glycolysis (glucose / lactate conversion) to neutralize the microenvironment. The negative feedback system involves rapid switch between acid mediated mitochondrial OXPHOS (lactate production is halted) or excessive lactate produced by SLP with a rise in alkalinity. These effects are independent of energy requirements and /or levels of O2. Future research will be required to assess if similar observations are consistent amongst tumor cells of diverse species and phenotype.

Supplementary Material

Supp Table S1-S2

Table S1. Whole genome analysis of differentially expressed genes in N2a cells under pHe acidic (~6.8) or alkaline (~7.55) vs neutral (~7.3) for 24 hrs.

Table S2. Whole genome analysis of differentially expressed genes in N2a cells under pHe acidic (~6.8) vs alkaline (~7.55) for 24 hrs.

Acknowledgments

This research was supported by a grant from NIH NCRR RCMI program (G12RR 03020) and NIH-RCMI Biomedical Proteomics Facility Grant 2G12RR03035. The authors wish to thank the mass spectrometry facility at the University of Texas Medical Branch, Galveston, Texas, for the mass spectrometry service.

Footnotes

Disclosure Statement The authors have no conflict of interest.

References

1. Champion L, Brain E, Giraudet AL, et al. Breast cancer recurrence diagnosis suspected on tumor marker rising: value of whole-body 18FDG-PET/CT imaging and impact on patient management. Cancer. 2011;117:1621–9. [PubMed]
2. Serganova I, Rizwan A, Ni X, Thakur SB, Vider J, Russell J, Blasberg R, Koutcher JA. Metabolic imaging: a link between lactate dehydrogenase A, lactate, and tumor phenotype. Clin Cancer Res. 2011 Oct 1;17(19):6250–61. [PubMed]
3. Mentis AF, Kararizou E. Metabolism and cancer: an up-to-date review of a mutual connection. Asian Pacific journal of cancer prevention. APJCP. 2010;11:1437–44. [PubMed]
4. Shapovalov Y, Hoffman D, Zuch D, de Mesy Bentley KL, Eliseev RA. Mitochondrial dysfunction in cancer cells due to aberrant mitochondrial replication. J Biol Chem. 2011;286:22331–8. [PubMed]
5. Smerc A, Sodja E, Legisa M. Posttranslational modification of 6-phosphofructo-1-kinase as an important feature of cancer metabolism. PloS one. 2011;6:e19645. [PMC free article] [PubMed]
6. Shlomi T, Benyamini T, Gottlieb E, Sharan R, Ruppin E. Genome-scale metabolic modeling elucidates the role of proliferative adaptation in causing the Warburg effect. PLoS computational biology. 2011;7:e1002018. [PMC free article] [PubMed]
7. Vazquez A, Oltvai ZN. Molecular crowding defines a common origin for the Warburg effect in proliferating cells and the lactate threshold in muscle physiology. PloS one. 2011;6:e19538. [PMC free article] [PubMed]
8. Sun Q, Chen X, Ma J, et al. Mammalian target of rapamycin up-regulation of pyruvate kinase isoenzyme type M2 is critical for aerobic glycolysis and tumor growth. Proc Natl Acad Sci U S A. 2011;108:4129–34. [PubMed]
9. Whitaker-Menezes D, Martinez-Outschoorn UE, Lin Z, et al. Evidence for a stromal-epithelial “lactate shuttle” in human tumors: MCT4 is a marker of oxidative stress in cancer-associated fibroblasts. Cell Cycle. 2011;10:1772–83. [PMC free article] [PubMed]
10. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science. 2009;324:1029–33. [PMC free article] [PubMed]
11. Zaugg K, Yao Y, Reilly PT, et al. Carnitine palmitoyltransferase 1C promotes cell survival and tumor growth under conditions of metabolic stress. Genes Dev. 2011;25:1041–51. [PubMed]
12. Onozuka H, Tsuchihara K, Esumi H. Hypoglycemic/hypoxic condition in vitro mimicking the tumor microenvironment markedly reduced the efficacy of anticancer drugs. Cancer science. 2011;102:975–82. [PubMed]
13. Hirschhaeuser F, Sattler UG, Mueller-Klieser W. Lactate: a metabolic key player in cancer. Cancer Res. 2011;71:6921–5. [PubMed]
14. Grant BJ. Influence of Bohr-Haldane effect on steady-state gas exchange. J Appl Physiol. 1982;52:1330–7. [PubMed]
15. Shiomi S, Kawabe J. Clinical applications of positron emission tomography in hepatic tumors. Hepatology research: the official journal of the Japan Society of Hepatology. 2011;41:611–7. [PubMed]
16. Evans SM, Casartelli A, Herreros E, et al. Development of a high throughput in vitro toxicity screen predictive of high acute in vivo toxic potential. Toxicology in vitro: an international journal published in association with BIBRA. 2001;15:579–84. [PubMed]
17. Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009;37:1–13. [PMC free article] [PubMed]
18. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature protocols. 2009;4:44–57. [PubMed]
19. Mazzio E, Soliman KF. D-(+)-glucose rescue against 1-methyl-4-phenylpyridinium toxicity through anaerobic glycolysis in neuroblastoma cells. Brain Res. 2003;962:48–60. [PubMed]
20. Stuhr LE, Iversen VV, Straume O, Maehle BO, Reed RK. Hyperbaric oxygen alone or combined with 5-FU attenuates growth of DMBA-induced rat mammary tumors. Cancer Lett. 2004;210:35–40. [PubMed]
21. Harden TK, Sesma JI, Fricks IP, Lazarowski ER. Signalling and pharmacological properties of the P2Y receptor. Acta Physiol (Oxf) 2010;199:149–60. [PMC free article] [PubMed]
22. Jacobson KA, Deflorian F, Mishra S, Costanzi S. Pharmacochemistry of the platelet purinergic receptors. Purinergic signalling. 2011;7:305–24. [PMC free article] [PubMed]
23. Bruni P, Vasta V, Berti L, Avila MA, Farnararo M, Varela-Nieto I. An inositol phosphoglycan stimulates glycolysis in human platelets. Biochem Biophys Res Commun. 1991;180:1041–7. [PubMed]
24. Matthews DA, Russell WC. Adenovirus core protein V interacts with p32--a protein which is associated with both the mitochondria and the nucleus. J Gen Virol. 1998;79(Pt 7):1677–85. [PubMed]
25. Kanehisa M, Goto S, Furumichi M, Tanabe M, Hirakawa M. KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res. 2010;38:D355–60. [PMC free article] [PubMed]
26. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27–30. [PMC free article] [PubMed]
27. Kanehisa M, Goto S, Hattori M, et al. From genomics to chemical genomics: new developments in KEGG. Nucleic Acids Res. 2006;34:D354–7. [PMC free article] [PubMed]
28. Gallagher FA, Kettunen MI, Day SE, et al. Detection of tumor glutamate metabolism in vivo using (13)C magnetic resonance spectroscopy and hyperpolarized [1-(13)C]glutamate. Magn Reson Med. 2011;66:18–23. [PubMed]
29. Darpolor MM, Yen YF, Chua MS, et al. In vivo MRSI of hyperpolarized [1-(13) C] pyruvate metabolism in rat hepatocellular carcinoma. NMR Biomed. 2010 [PMC free article] [PubMed]
30. Javalkar VK, Vinod KY, Sharada S, Chandramouli BA, Subhash MN, Kolluri VR. Study of pyruvate kinase activity in human astrocytomas - alanine-inhibition test revisted. Neurol India. 2009;57:140–2. [PubMed]
31. Chung JS, Lee SB, Park SH, et al. Mitochondrial reactive oxygen species originating from Romo1 exert an important role in normal cell cycle progression by regulating p27(Kip1) expression. Free Radic Res. 2009;43:729–37. [PubMed]
32. Gomes LC, Di Benedetto G, Scorrano L. During autophagy mitochondria elongate, are spared from degradation and sustain cell viability. Nature cell biology. 2011;13:589–98. [PMC free article] [PubMed]
33. Huang P, Galloway CA, Yoon Y. Control of mitochondrial morphology through differential interactions of mitochondrial fusion and fission proteins. PloS one. 2011;6:e20655. [PMC free article] [PubMed]
34. Merrill RA, Dagda RK, Dickey AS, et al. Mechanism of neuroprotective mitochondrial remodeling by PKA/AKAP1. PLoS biology. 2011;9:e1000612. [PMC free article] [PubMed]
35. Palmer CS, Osellame LD, Laine D, Koutsopoulos OS, Frazier AE, Ryan MT. MiD49 and MiD51, new components of the mitochondrial fission machinery. EMBO reports. 2011;12:565–73. [PubMed]
36. Kowald A, Kirkwood TB. Evolution of the mitochondrial fusion-fission cycle and its role in aging. Proc Natl Acad Sci U S A. 2011;108:10237–42. [PubMed]
37. Vafiadaki E, Papalouka V, Arvanitis DA, Kranias EG, Sanoudou D. The role of SERCA2a/PLN complex, Ca (2+) homeostasis, and anti-apoptotic proteins in determining cell fate. Pflugers Arch. 2009;457:687–700. [PubMed]
38. Zhao W, Waggoner JR, Zhang ZG, et al. The anti-apoptotic protein HAX-1 is a regulator of cardiac function. Proc Natl Acad Sci U S A. 2009;106:20776–81. [PubMed]
39. Li WB, Feng J, Geng SM, et al. Induction of apoptosis by Hax-1 siRNA in melanoma cells. Cell Biol Int. 2009;33:548–54. [PubMed]
40. Ichas F, Mazat JP. From calcium signaling to cell death: two conformations for the mitochondrial permeability transition pore. Switching from low- to high-conductance state. Biochim Biophys Acta. 1998;1366:33–50. [PubMed]
41. D’Errico I, Salvatore L, Murzilli S, et al. Peroxisome proliferator-activated receptor-gamma coactivator 1-alpha (PGC1alpha) is a metabolic regulator of intestinal epithelial cell fate. Proc Natl Acad Sci U S A. 2011;108:6603–8. [PubMed]
42. Scarpulla RC. Nuclear control of respiratory chain expression by nuclear respiratory factors and PGC-1-related coactivator. Ann N Y Acad Sci. 2008;1147:321–34. [PMC free article] [PubMed]
43. Andersson U, Scarpulla RC. Pgc-1-related coactivator, a novel, serum-inducible coactivator of nuclear respiratory factor 1-dependent transcription in mammalian cells. Mol Cell Biol. 2001;21:3738–49. [PMC free article] [PubMed]
44. Koo SH, Satoh H, Herzig S, et al. PGC-1 promotes insulin resistance in liver through PPAR-alpha-dependent induction of TRB-3. Nat Med. 2004;10:530–4. [PubMed]
45. Liu J, Wu X, Franklin JL, et al. Mammalian Tribbles homolog 3 impairs insulin action in skeletal muscle: role in glucose-induced insulin resistance. American journal of physiology Endocrinology and metabolism. 2010;298:E565–76. [PubMed]
46. Rouschop KM, Dubois L, Schaaf MB, et al. Deregulation of cap-dependent mRNA translation increases tumour radiosensitivity through reduction of the hypoxic fraction. Radiother Oncol. 2011;99:385–91. [PubMed]
47. Kimball SR, Jefferson LS. Signaling pathways and molecular mechanisms through which branched-chain amino acids mediate translational control of protein synthesis. J Nutr. 2006;136:227S–31S. [PubMed]
48. Roe CR, Mochel F. Anaplerotic diet therapy in inherited metabolic disease: therapeutic potential. J Inherit Metab Dis. 2006;29:332–40. [PubMed]
49. Lee WN, Guo P, Lim S, et al. Metabolic sensitivity of pancreatic tumour cell apoptosis to glycogen phosphorylase inhibitor treatment. Br J Cancer. 2004;91:2094–100. [PMC free article] [PubMed]
50. van den Hoogen C, van der Horst G, Cheung H, Buijs JT, Pelger RC, van der Pluijm G. The aldehyde dehydrogenase enzyme 7A1 is functionally involved in prostate cancer bone metastasis. Clin Exp Metastasis. 2011;28:615–25. [PMC free article] [PubMed]
51. Kim MP, Fleming JB, Wang H, et al. ALDH activity selectively defines an enhanced tumor-initiating cell population relative to CD133 expression in human pancreatic adenocarcinoma. PloS one. 2011;6:e20636. [PMC free article] [PubMed]
52. Ramana KV. ALDOSE REDUCTASE: New Insights for an Old Enzyme. Biomolecular concepts. 2011;2:103–14. [PMC free article] [PubMed]
53. Tammali R, Reddy AB, Srivastava SK, Ramana KV. Inhibition of aldose reductase prevents angiogenesis in vitro and in vivo. Angiogenesis. 2011;14:209–21. [PMC free article] [PubMed]
54. Stresemann C, Bokelmann I, Mahlknecht U, Lyko F. Azacytidine causes complex DNA methylation responses in myeloid leukemia. Molecular cancer therapeutics. 2008;7:2998–3005. [PubMed]
55. Hayden A, Johnson PW, Packham G, Crabb SJ. S-adenosylhomocysteine hydrolase inhibition by 3-deazaneplanocin an analogues induces anti-cancer effects in breast cancer cell lines and synergy with both histone deacetylase and HER2 inhibition. Breast Cancer Res Treat. 2011;127:109–19. [PubMed]