The regulation of mitochondrial DNA copy number in glioblastoma cells

As stem cells undergo differentiation, mitochondrial DNA (mtDNA) copy number is strictly regulated in order that specialized cells can generate appropriate levels of adenosine triphosphate (ATP) through oxidative phosphorylation (OXPHOS) to undertake their specific functions. It is not understood whether tumor-initiating cells regulate their mtDNA in a similar manner or whether mtDNA is essential for tumorigenesis. We show that human neural stem cells (hNSCs) increased their mtDNA content during differentiation in a process that was mediated by a synergistic relationship between the nuclear and mitochondrial genomes and results in increased respiratory capacity. Differentiating multipotent glioblastoma cells failed to match the expansion in mtDNA copy number, patterns of gene expression and increased respiratory capacity observed in hNSCs. Partial depletion of glioblastoma cell mtDNA rescued mtDNA replication events and enhanced cell differentiation. However, prolonged depletion resulted in impaired mtDNA replication, reduced proliferation and induced the expression of early developmental and pro-survival markers including POU class 5 homeobox 1 (OCT4) and sonic hedgehog (SHH). The transfer of glioblastoma cells depleted to varying degrees of their mtDNA content into immunocompromised mice resulted in tumors requiring significantly longer to form compared with non-depleted cells. The number of tumors formed and the time to tumor formation was relative to the degree of mtDNA depletion. The tumors derived from mtDNA depleted glioblastoma cells recovered their mtDNA copy number as part of the tumor formation process. These outcomes demonstrate the importance of mtDNA to the initiation and maintenance of tumorigenesis in glioblastoma multiforme.

The circular, double-stranded human mitochondrial genome (mitochondrial DNA, mtDNA) is 16 569 bp in size and encodes 13 subunits of the electron transfer chain (ETC), 1 which is the major generator of cellular adenosine triphosphate (ATP) through oxidative phosphorylation (OXPHOS). 2 It also possesses 22 transfer RNAs and 2 ribosomal RNAs, and one non-coding region, the D-loop, 1 which is the site of interaction for the nuclear-encoded mtDNA replication factors. 3 MtDNA replication is initiated by mitochondrial transcription factor A (TFAM), 4 which generates the primer used by the catalytic subunit of the mtDNA-specific DNA polymerase, polymerase g A (POLGA), to copy mtDNA. Replication is supported by POLGA's accessory subunit, polymerase g B (POLGB), the mtDNA-specific helicase, TWINKLE, and the mtDNA single-stranded-binding protein (MTSSB). 5,6 Regulation of mtDNA copy number is essential for maintaining cellular energy requirements. High-energy requiring cells, such as muscle and neurons, require large quantities of ATP and maintain high numbers of mtDNA copy while lowenergy requiring cells, spleen and endothelial cells, maintain fewer copies. 7 MtDNA replication and transcription are tightly coupled such that the expression of the mtDNA genes, and hence the generation of ATP through OXPHOS, requires continuous replication of mtDNA. 8 Metabolism in tumor-initiating cells is described by the Warburg effect. Tumors utilize aerobic glycolysis even under normoxic conditions, which normally promotes OXPHOS. 9 This promotes self-renewal and the highly proliferative nature of tumor cells, 10 enabling them to generate sufficient energy and pools of metabolic intermediates. This is similar to embryonic stem cells (ESCs), which are highly proliferative, undergo selfrenewal 11 and maintain few copies of mtDNA. 12 This establishes the mtDNA set point, whereby ESCs accumulate mtDNA in a cell-specific manner to meet the functional requirements of specialized cells during differentiation. 12,13 This requires synchrony between the nuclear and mtDNA genomes to ensure that mtDNA replication and differentiation take place concurrently.
It is not known whether precursor cells giving rise to tumors can mimic the mtDNA replication events of stem cells as they differentiate into somatic cells and whether they require mtDNA to promote tumorigenesis. Glioblastoma multiforme (GBM) is the most common subgroup of highly malignant astrocytomas with a median survival time of 12 to 16 months. 14 Cell lines derived from GBM tumors, which are dependent on aerobic glycolysis, 15,16 are excellent models to understand the role that mtDNA has in tumor formation and self-renewal. The best characterized of the GBM cell lines is HSR-GBM1. It expresses neural stem cell (NSC) makers, such as NESTIN, MUSASHI1 and prominin 1 (CD133) and their levels of expression correlate with patient prognosis. 17 They differentiate into neurons and astrocytes and express the astrocyte end point marker, glial fibrillary acidic protein (GFAP). 18 We have determined whether GBM cells can modulate mtDNA copy number and chromosomal gene expression during differentiation when compared with human (h)NSCs. Furthermore, we have determined whether mtDNA is essential for the differentiation of GBM cells, their survival and the initiation of tumorigenesis by depleting them of mtDNA and allowing them to recover in vitro and in vivo. We demonstrate that mtDNA is essential to tumorigenesis.

Results
GBM cells do not expand mtDNA copy number during differentiation. To determine whether GBM cells can modulate mtDNA copy number during differentiation, we induced three human GBM (HSR-GBM1, GBM-L1 and L2) and one human neural stem cell (hNSC) lines to undergo astrocyte differentiation (Figure 1a). hNSCs increased mtDNA copy number progressively resulting in a 3.23-fold increase by day 28 (Po0.001). Although HSR-GBM1 cells significantly increased (1.25-fold) mtDNA copy number on day 7 (Po0.001), this only increased to 1.37-fold by day 28 (Po0.001). A similar pattern was observed for GBM-L2 cells, except they had significantly more copies than HSR-GBM1 and GBM-L1 cells on day 28 (Po0.001). Although GBM-L1 cells displayed significant increases on days 7 and 14, levels were reduced by day 28. Nevertheless, cells from all three GBM lines had significantly fewer copies of mtDNA than the hNSCs by day 28 (Po0.001).
GBM cells do not maintain elevated levels of GFAP expression during differentiation. We determined whether GBM cells concordantly regulate the expression of  (Figure 1c) expression increased on day 28 in HSR-GBM1 cells, which were lower than for hNSCs (Po0.01), only POLGB ( Figure 1d) had similar levels of expression to hNSCs on day 28 but not before this. Although expression of MTSSB in HSR-GBM1 cells was similar to hNSCs on day 14, it was significantly higher than for hNSCs on day 28 ( Figure 1f). Consequently, GBM cells do not concordantly regulate expression of the nuclearencoded mtDNA replication factors and markers of early and late differentiation, resulting in their failure to expand mtDNA copy number.
GBM cells do not increase their respiratory capacity during differentiation. To determine whether the failure of GBM cells to increase mtDNA copy number during differentiation altered oxygen (O 2 ) consumption rates, ATP content and lactate production, we compared HSR-GBM1 cells and hNSCs (Supplementary Table SI). Undifferentiated HSR-GBM1 cells consumed significantly more O 2 than undifferentiated hNSCs (Po0.001). Uncoupling of the ETC revealed that undifferentiated hNSCs and HSR-GBM1 cells were respiring at maximal rate based on limited reserve capacities of 0.9 and 1.1, respectively. During differentiation, hNSCs increased O 2 consumption rates by 2.46-fold (Po0.001) and ETC reserve capacity (Po0.001). However, HSR-GBM1 cells marginally increased O 2 consumption rates and ETC reserve capacity. The increased respiratory capacity of differentiated hNSCs represented a fourfold increase in ATP content relative to undifferentiated hNSCs (Po0.001), which was less profound in HSR-GBM1 (Po0.001). Undifferentiated HSR-GBM1 cells secreted significantly more lactate than hNSCs (Po0.001), which was reduced following differentiation (Po0.001). Consequently, increased O 2 consumption and OXPHOSgenerated ATP are dependent on increased mtDNA copy number.
We then determined whether mtDNA depletion induced changes in expression of NESTIN, MUSASHI1, CD133 and GFAP. There were differential patterns of expression for NESTIN ( Figure 2b) and MUSASHI1 (Figure 2c) after 7 and 14 days of depletion. However, expression of both genes was upregulated by day 21 and downregulated by days 25 and 50. CD133 expression was significantly reduced after 7 and 14 days (Po0.001; Figure 2d), returned to basal levels by day 21, and was undetectable by day 50 (Po0.001). There were no significant changes in GFAP expression during the first 14 days. However, from day 21 onward, expression progressively decreased ( Figure 2e).
As there were significant decreases in multipotent gene and GFAP expression during mtDNA depletion, we analyzed expression profiles of non-depleted and cells depleted for 25 and 50 days using a neural real-time PCR array (see Supplementary Table SII). Days 25 and 50 depleted cells showed significant and differential expression of 26/80 genes relative to non-depleted cells. Genes upregulated in both depleted groups are associated with growth factor signaling, anti-apoptosis and cell adhesion, namely fibroblast growth factor 13 (FGF13), glial-derived neurotrophic factor (GDNF) and semaphorin-4D (SEMA4D). The cell pro-proliferation factor, vascular endothelial growth factor A (VEGFA), was upregulated on day 25 and the anti-proliferation factor, anaplastic lymphoma kinase, was upregulated by day 50. By day 50, proliferation rates for depleted cells were reduced ( Figure 2f). These outcomes suggest that HSR-GBM1 cells require sufficient copies of mtDNA to support cell proliferation.
The early neural patterning factor, sonic hedgehog (SHH), was upregulated in day 50 depleted cells while levels of expression for acetylcholinesterase (ACHE), dopamine receptor D2 (DRD2) and neuronal pentraxin 1 (NPTX) were upregulated by day 25 but decreased by day 50. Similarly, the regulators of cell fate and differentiation, apolipoprotein E (APOE), Achaete-scute homolog 1 (ASCL1) and delta-like 1 (DLL1) were downregulated while hairy/enhancer-of-split related with YRPW motif (HEY1), a transcriptional repressor involved in neurogenesis, was upregulated in days 25 and 50 depleted cells. Consequently, mtDNA depletion leads to increased expression of genes associated with early developmental processes.
As there were significant increases in expression of early developmental markers, we analyzed POU class 5 homeobox 1 (OCT4), Nanog homeobox (NANOG), sex determining region Y-box 2 (SOX2), V-Myc myelocytomatosis viral oncogene homolog (Avian) (c-MYC) and human telomerase reverse transcriptase (hTERT), which are associated with pluripotency, cell proliferation and self-renewal. Although the expression of OCT4 (Figure 3a Figures S3B-E). These outcomes further emphasize the asynchronous relationship between the nuclear and mitochondrial genomes in HSR-GBM1 cells. Furthermore, depletion had no adverse effects on optic atrophy 1 (OPA1) processing indicating that mitochondrial networks were not disrupted through microtubule-associated protein 1 light chain 3 beta (LC3B), though autophagy was upregulated during depletion and reduced during recovery (Supplementary Figure S3F).
To determine whether modulation of mtDNA copy number promotes cellular differentiation, we induced HSR-GBM1 cells depleted for 7, 14 and 21 days to differentiate into astrocytes for 14 days. Day 7 depleted cells replenished copy number by day 14 of differentiation to levels 1.46-fold higher than observed in day 0 cells (Po0.001; Figure 5a). Similarly, day 14 depleted cells reached levels 1.32-fold greater by day 14 of differentiation ( Figure 5b). However, surviving cells depleted for 21 days increased copy number to 44-fold lower than day 0 (Po0.001; Figure 5c). were generated from mtDNA 20 cells, 6/12 were derived from mtDNA 3 cells and 3 regressed and 2/12 were generated from mtDNA 0.2 cells. Tumor formation to 500 mm 3 was least in the mtDNA 0.2 and greatest in the mtDNA 100 cohorts (Figure 6b). These data demonstrate that increased mtDNA depletion reduces the frequency of tumor formation.
To determine whether tumor formation was dependent on the recovery of mtDNA in mtDNA depleted cells, we analyzed the proliferative potential of the mtDNA non-depleted and depleted tumors. Tumors derived from mtDNA 0.2 cells (Figures 6c and j) exhibited significantly lower proliferative potential compared with mtDNA 100 counterparts (Figures 6d-f  and j). We analyzed mtDNA copy number in all tumors using human-specific primers. All tumors derived from depleted cells restored mtDNA copy number during tumorigenesis to levels at or near to in vitro grown HSR-GBM1 cells. Although there was no difference in mtDNA copy number between mtDNA 100 , mtDNA 50 , mtDNA 3 and mtDNA 0.2 tumors, there was significantly lower copy number in mtDNA 20 tumors (Po0.05; Figure 6k).

Discussion
We have compared GBM cells with hNSCs. We selected ESC-derived hNSCs as they express NSC markers, such as NESTIN, MUSASHI1, CD133 and those listed in Supplementary Table SII. Furthermore, their regulation of mtDNA copy number is similar to primary murine NSCs. 20 We observed differential expression of the multipotent genes, NESTIN, MUSASHI1 and CD133 between hNSC and GBM cells during differentiation. However, HSR-GBM1 cells could not maintain elevated levels of GFAP expression at later stages of differentiation. Furthermore, HSR-GBM1 cells failed to expand mtDNA copy number during differentiation, which was reflected in altered patterns of expression for the nuclearencoded mtDNA replication factors. GBM-L2 cells also failed to increase mtDNA copy number and GFAP expression, whereas GBM-L1 cells increased GFAP expression but failed to expand mtDNA copy number synchronously.
Since the generation of ATP through OXPHOS is coupled to the continuous replication of mtDNA, 8 GBM cells do not have the capacity to utilize OXPHOS effectively, probably due to mtDNA mutations, which have been identified in GBM patient tumors. 21 Tumor cell metabolism is normally defined by glycolysis, with increased glucose uptake and lactate production. 9 Indeed, undifferentiated HSR-GBM1 cells failed to increase O 2 consumption rates and copy number and had high levels of lactate production, which were downregulated during differentiation, unlike their hNSC counterparts. An enhanced glycolytic state provides multiple benefits to tumor cells including sufficient ATP and biosynthetic intermediates to support cell division and growth 15,16,22 and the generation of NADPH, which is involved in redox control. 23 Similarly, ESCs are rapidly proliferating cells that promote self-renewal and pluripotency by maintaining low mtDNA copy number and primarily use glycolysis to  12,13,[24][25][26] This establishes the mtDNA set point, 12,13 which ensures that, during differentiation, heart, neural and muscle cells acquire high numbers of mtDNA copy to utilize OXPHOS, 7 whereas endothelial cells possess few copies and rely on glycolysis. 27 GBM is derived from multiple origins, such as glial cells 28 and/or NSCs 29 that have undergone neoplastic transformation. The abnormal regulation of mtDNA copy number in HSR-GBM1 cells is possibly indicative of a transformed glial cell. During transformation, the acquisition of aberrant oncogenic signaling 30 and reactivation of pluripotent and multipotent regulators, such as OCT4, NANOG, 31 SOX2, 32 c-MYC 33 and NESTIN, 17 likely leads to abnormal regulation of mtDNA copy number. Pluripotent stem cells can be derived by reprogramming somatic cells through the forced expression of pluripotent factors. 34 However, these cells often fail to re-establish the mtDNA set point and do not accumulate sufficient numbers of tDNA copy during differentiation. 25 It appears that HSR-GBM1 cells regulate mtDNA in a similar manner to poorly induced pluripotent stem cells.
Following depletion for 7 days, HSR-GBM1 cells recovered and increased mtDNA copy number to levels higher than non-depleted cells. Furthermore, they exhibited increased expression of NSC markers and GFAP. This is similar to human neural progenitors, which survive short-term mtDNA depletion without losing multipotency. 35 This also demonstrates that the short-term modulation of mtDNA copy number alters multipotent gene expression and triggers differentiation. However, days 7 to 14 of depletion represent the dividing line for HSR-GBM1 cells to restore mtDNA copy number whereby day 14 depleted cells do not completely restore mtDNA copy number while day 7 depleted cells increase mtDNA copy number in a similar manner to hNSCs. Nevertheless, cells recovered in conditioned media rescued mtDNA copy number but gene expression profiles were aberrantly regulated suggesting that other gene expression pathways were likely enacted. The tumorigenicity of tumor cells depleted of mtDNA has been an unresolved issue with reports of increased 36 and decreased 37 tumorigenic potential in multiple tumor cell lines. The depletion of HSR-GBM1 cells to B90% (14 days of depletion) of their original mtDNA copy number did not reflect the significant changes in gene expression observed by day 50. By day 50, population-doubling times were increased, which could be explained by mtDNA depletion disrupting the synthesis of deoxyribose nucleoside triphosphates by mitochondria. 38 We observed extensive changes in gene expression with reductions in NESTIN, MUSASHI1 and GFAP, whereas CD133, which is a marker of bioenergetic stress during the very early stages of mtDNA depletion, 39 was lost.
We depleted GBM cells of their mtDNA with ddC, which, at low concentrations, acts by directly inhibiting POLGA. 40 At higher concentrations, it has been associated with cellular toxicity inducing neuropathies, liver disorders and lactic acidosis. 41 Chemically induced mtDNA depletion has been associated with increased production of reactive O 2 species, changes to mitochondrial ultrastructure and mitochondrial networks, 42,43 which may alter chromosomal gene expression. 42 Nevertheless, our results indicate that mitochondrial stress responses were not activated, as demonstrated by L-OPA1 stability, which also suggests that cristae structures were not compromised. Furthermore, the unwanted organelle material would be degraded through the lysosomal machinery during depletion because of increased expression of LC3B. It is, therefore, likely that the changes in mtDNA levels are responsible for altered gene expression patterns.
Extended depletion of mtDNA did not induce cell death in HSR-GBM1 cells, an outcome that is likely to be driven by disruption of the mitochondrial apoptotic pathway caused by depletion of mtDNA 36 and the increased expression we observed in anti-apoptotic genes. MtDNA depleted tumor cells have been reported to increase their anchorageindependent growth properties. 44 We observed that HSR-GBM1 cells grew as tightly packed neurospheres and showed increased expression of cell adhesion-associated genes. Although GBM tumors have exhibited altered patterns of expression for the key regulators of pluripotency, OCT4, NANOG 31 and SOX2, 32 and SHH signaling, 45 depleted HSR-GBM1 cells only exhibited elevated levels of OCT4 and SHH.
Depleted HSR-GBM1 cells formed tumors in immunocompromised mice. However, the frequency was significantly hindered by mtDNA depletion. For those tumors that formed, mtDNA copy number recovered levels similar to in vitro cultured HSR-GBM1 cells demonstrating the necessity for tumor-initiating cells to establish a mtDNA set point. This is best exemplified by the mtDNA 0.2 cells that possessedo1 copy of mtDNA per cell.

Conclusion.
We have established clear differences in the modulation of mtDNA copy number during differentiation between multipotent GBM cells and hNSCs. During differentiation, GBM cells failed to expand their mtDNA copy number and increase their respiratory capacity, which is underpinned by uncoordinated expression of nuclearencoded mtDNA replication factors and lineage-specific markers. It is likely that a sub-population of cells exist with very few copies of mtDNA, which represents a tumorsustaining population that, in vivo, has the capacity to re-establish its mtDNA set point after a prolonged period and generate tumors. This clearly demonstrates the importance of mtDNA to the establishment and maintenance of tumorigenesis.
MtDNA copy number. Total DNA was extracted from cell samples using the DNeasy Blood and Tissue kit (Qiagen, Valencia, CA, USA) with proteinase K and RNase treatment, according to the manufacturer's instructions. To quantify mtDNA copy number, real-time PCR was performed using a 72-well Rotorgene-3000 (Corbett Research, Cambridge, UK) against external standards for mtDNA and b-globin, as previously described 47 using primers listed in Supplementary Table SIII.
Gene expression analysis. Total RNA was extracted using the RNeasy Mini Kit (Qiagen) with a DNase treatment step, according to the manufacturer's instructions. cDNA was synthesized using the Bioscript system (Bioline, London, UK) and PCR products were generated using RT-PCR and analyzed, as described in Facucho-Oliveira et al. 12 Purified PCR products were prepared for each primer pair and used as standards in a series of 10-fold dilutions of known concentration for each gene to determine real-time PCR reaction efficiencies. For gene expression analysis, real-time PCR reactions contained 20 ng of total cDNA and primer sequences are described in Supplementary Table SIII. Data acquisition and melt curve analyses were performed, as described previously. 12 Reactions were performed in triplicate twice. Relative gene expression was calculated using the Pfaffl method (mean ± S.E.M.) and b-actin was selected as the housekeeping gene. 12 Cellular respiration analysis. O 2 consumption rates for hNSCs and GBM cells were determined by high-resolution respirometry (Oroboros Oxygraph-2K, Innsbruck, Austria). Briefly, cells were dissociated using Accutase and cell numbers determined. Cells were resuspended in 50 ml Hanks balanced salt solution and transferred into chambers containing 2 ml Hanks balanced salt solution maintained at 37 1C and continuously stirred at 750 r.p.m. O 2 consumption was measured using the integrated software package Datlab Cancer cells and mitochondrial DNA A Dickinson et al (Version 3.1; Oroboros, Innsbruck, Austria), which presented respiration as O 2 flux, pmol O 2 per 10 6 cells per second. Initial resting measurements ('basal') were recorded for 5 min following O 2 flux stabilization, after which a series of respiratory chain inhibitors were added at 10-min intervals to manipulate cellular respiration. In all, 5 mg/ml of the ATP synthase inhibitor oligomycin (Sigma) was used to assess mitochondrial coupling and the amount of non-phosphorylating respiration ('non-phos'). Maximal uncoupled ETC respiratory capacity ('uncoupled') was measured using 50-100 nM carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP; Sigma). In total, 5 mM of the complex III inhibitor Antimycin A (Sigma) was used to determine background O 2 consumption and was subtracted from all calculated values.
Total ATP and cellular lactate measurements. Cellular ATP content was determined using the ATPlite Luminescence ATP Detection Assay (PerkinElmer Life Sciences, Zaventem, Belgium), according to the manufacturer's instructions. Cellular lactate production was determined using the Lactate Assay Kit II (Biovision, San Francisco, CA, USA), according to the manufacturer's instructions. Cells were cultured under routine conditions and, before analysis, culture media completely removed and replaced with fresh media. After 24 h, a sample of media was removed for analysis. Luminescence (ATP) and absorbance (lactate) were measured using an optical plate reader (BMG Labtech, Allmendgrün, Ortenberg). ATP and lactate concentrations were extrapolated from standard curves. Each experimental sample was measured in triplicate and the experiment repeated three times.
RT 2 PCR array analysis. Total RNA was extracted from HSR-GBM1 cells and hNSCs using the RNeasy Kit (Qiagen), according to the manufacturer's protocol. cDNA was synthesized using the RT 2 First Strand Kit (SABiosciences, Frederick, MD, USA), as previously described. 48 Undifferentiated and day 25 and 50 mtDNA depleted HSR-GBM1 cells were analyzed using the Neurogenesis and Neural Stem Cell RT 2 Profiler PCR Array (SABiosciences).
Real-time PCR array reactions were performed in triplicate using 384-well (4 Â 96) optical reaction plates (SABiosciences). A PCR master mix was prepared per sample containing 102 ml cDNA, 550 ml RT 2 SYBR Green/ROX Mix (SABiosciences) and 448 ml ddH 2 O. In all, 10 ml PCR reaction mixtures were prepared using a CAS-1200 Robotic Liquid Handling System (Corbett Robotics, Queensland, Australia). Real-time reactions were conducted on an ABI PRISM 7900 HT Fast Real-Time PCR System (Applied Biosystems, Carlsbad, CA, USA) and consisted of an initial denaturation step of 95 1C for 10 min; and 40 cycles of denaturation at 95 1C for 15 s; and an annealing/extension phase at 60 1C for 1 min. Gene expression data were generated in the form of cycle threshold (Ct) values. Relative gene expression was calculated by the DDCt method and normalized against the average Ct values of five housekeeping genes (glucuronidase beta (GUSB), hypoxanthine phosphoribosyltransferase 1 (HPRT1), heat shock protein 90 kDa alpha (cytosolic), class B member 1 (HSP90AB1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and beta cytoskeletal actin (b-ACTIN)), according to the manufacturer's protocol. Analyses were performed using the Web-Based PCR Array Data Analysis software (SABiosciences).
Xenograft models. Mouse experiments were approved by the Animal Ethics Committee, Monash University, Approval Number: MMCA/2011/76. HSR-GBM1 tumor cells (5 Â 10 5 ) in 100 ml of growth medium were inoculated subcutaneously into both flanks of 5-to 6-week-old, female BALB/c nude mice (Animal Research Centre, Perth, Australia). Tumor volume was determined by: (length Â width 2 )/2, where the length was the longest axis and width was measured at right angles to the length. Data are expressed as mean tumor volume ± S.E.M. in cubic millimeters.
Proliferating nuclei were identified using a mouse monoclonal anti-proliferating cell nuclear antigen antibody (1 : 1000; Cell Signaling Technology, Danvers, MA, USA). Formalin-fixed paraffin-embedded sections (5 mm) were dewaxed, rehydrated and microwaved in citrate buffer for antigen retrieval. Once cooled, the sections were incubated with 3% hydrogen peroxide in methanol for 15 min to quench endogenous peroxidase. All sections were then incubated with the DAKO protein blocking solution (Dako Australia, Kingsgrove, Australia) to prevent nonspecific binding. The negative control was performed by deleting the primary antibody. The antibodies were incubated for 1 h at room temperature. The proliferating cell nuclear antigen was visualized with Link Label-Horseradish Peroxidase system by DAKO, according to the manufacturer's protocol (Dako Australia), followed by the chromogen Vector Red for 15 min (Vector Laboratories; Burlingame, CA, USA).
Image analysis was performed using a Leica (Wetzlar, Germany) inverted bright field microscope. Sections were scanned at low magnification to identify areas of high proliferation (hot spots). Images were then captured at Â 40 optical lens. Positive nuclei were counted from four fields of view. Using the Image J analysis software (National Institute of Health, Bethesda, MD, USA), positive nuclei were counted using the cell counter plug in analysis tool and the results were presented as mean values ± S.E.M.
Statistical analysis. Statistical significance for the RT 2 PCR arrays was determined using the Web-Based PCR Array Data Analysis software (SABiosciences), which used a two-tailed Student's t-test. For real-time PCR and tumor assays, statistically significant differences were determined using one-way ANOVA followed by Bonferroni post hoc test using GraphPad v5.0c (GraphPad Software, Inc., San Deigo, CA, USA). Statistical significance is expressed as *Po0.05, **Po0.01 and ***Po0.001.

Conflict of Interest
The authors declare no conflict of interest.