Materials and methods

Plasmid constructs, generation of stable cell clones, cell culture and small interfering RNA (siRNA)

To generate an inducible H1 promoter, synthetic oligonucleotides (Biomers, Ulm, Germany) coding for an H1 promoter with two TetO₂ sequences flanking the TATA box were cloned into the ClaI and Acc65I sites of a psiRNA-hH1.GFPzeo plasmid (Invivogen, Toulouse, France). The insert targeting CD147 was designed and cloned into the resulting construct following the instructions of the manufacturer of the original plasmid using the Acc65I/HindIII strategy with an adjusted linker sequence. The insert sequences were 5′ GTAC AN(21) TCAAGAG N(21)T TTTTTGGAAA 3′ and 5′ AGCTTTTCCAAAAAA AN(21) CTCTTGA N(21)T 3′, respectively. The sequence targeting CD147 used was AGTCGTCAGAACACATCAACG, and the scrambled control sequence was AATTCTCCGAACGTGTCACGT. Correct integration of all sequences was verified by sequencing (GATC, Konstanz, Germany). Cells were co-transfected with the constructs and pcDNA6/TR using FuGene6 (Roche, Mannheim, Germany) and subsequently cultured in the presence of blasticidin/zeocin for clonal selection. Expression of short hairpin RNA (shRNA) was induced in the presence of 1 μg/ml doxycycline. Two Stealth-Select siRNA reagents (Invitrogen, Carlsbad, California, USA) with independent sequences targeting CD147 (BSG1 and BSG2) were used for transient silencing.

Cell quantification

Cell growth was assessed by fluorophotometric quantification of DNA using bisbenzimide as described previously4 or by the tetrazolium salt-based cell proliferation assay CCK-8 (Dojindo Laboratories, Kumamoto, Japan).

Bromodeoxyuridine (BrdU) incorporation

Subconfluent MiaPaCa2 cells were cultured in the presence or absence of doxycycline in 24-well plates. Cells were made quiescent by serum withdrawal for 48 h before 10% fetal calf serum (FCS) was added and cells were pulse labelled with BrdU, fixed and stained as previously described.26 The labelling index was calculated for three random view fields.

CD147 immunoassay, lactate and glucose quantification

CD147 immunoassay was performed as previously described.4 Glucose and lactate were quantified on a Dade Dimension clinical chemistry analyser (Siemens, Erlangen, Germany). Intracellular lactate was assessed according to previously described methods.27 Briefly, lysates of cells generated by osmotic cytolysis were cleared by centrifugation for 10 min at 10 000 g before determination of the lactate concentration.

Assessment of the lactate transport rate

The lactate transport rate was determined by ratiometric intensity measurement of the fluorescent pH indicator 2′,7′-bis (carboxyethyl)-5(6)-carboxyfluorescein acetomethyl ester (BCECF-AM, Invitrogen) as described previously with minor modifications.14 Briefly, MiaPaCa2 cells were seeded in 24-well plates and transfected with the indicated siRNA 48 h before loading with BCECF-AM. The ratiometric emission intensity of BCECF-AM at 536 nm was assessed on an Olympus IX71 inverted microscope using the excitation wavelengths 470 nm and 436 nm, respectively, where the latter represents the isosbestic wavelength. Image analysis was performed using ImageJ.28 The lactate uptake is reflected by intracellular acidification after addition of lactate that results in a decrease in the 470 nm/436 nm ratio. The rate of change of the intracellular pH was assessed by the slope of a linear trend line fitted to the ratiometric intensity decrease of BCECF-AM 0–10 s after addition of 20 mM lactate to the supernatant.

Western blot

Detergent-soluble proteins were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) using an 8% gel and transblotted on polyvinylidene fluoride (PVDF) membranes. The detection antibodies used were rabbit antihuman MCT1 (Chemicon, Hampshire, UK), rabbit antihuman MCT4 (kindly supplied by Dr Philp12) and mouse antihuman CD147 clone MEM-M6/1 (Serotec, Oxford, UK).

Immunocytology and immunohistology

Immunostaining of chorioallantoic membrane (CAM) samples for cytokeratin and CD147 was performed as previously described.29 and immunoreactivity was visualised by an APAAP (alkaline phosphatase–antialkaline phosphatase) conjugate using Fast Red as chromogen. CD147 immunocytology was essentially performed as described previously4 using Alexa Fluor 594 for visualisation. Immunostaining of tumour samples was performed essentially as described.29 Briefly, after deparaffinisation and rehydration of tissue sections following standard protocols, antigen retrieval was performed by heating in citrate buffer (0.01 M, pH 6.0). Specimens were incubated with anti-CD147 (1:100), anti-MCT1 (1:100), anit-MCT4 (1:200) or pancytokeratin antibody (1:100) at 4°C overnight. Immunoreactivity was visualised using the VECTASTAIN ABC Kit (Vector Laboratories, Burlingame, California, USA).

PCR

cDNA was synthesized using Superscript reverse transcriptase (Invitrogen) and PCR was performed on a light-cycler platform using the Roche Light-Cycler Fast-Start Sybr-Green DNA Kit. The primer sequences for CD147 were 5′-CCGAGGACGTCCTGGAT-3′ and 5′-CGGGCCACCTGCCTCA-3′. Quantitec Primer Assays (Qiagen, Hilden, Germany) were used for detection of MCT1 and MCT4. Negative controls comprised replacement of either cDNA or reverse transcriptase with water. XS13 was used as housekeeping gene. Mean normalized expression was calculated as previously described.30

CAM assay, animals and tumours

The CAM assay was performed as previously described.29 Xenografts were initiated by subcutaneous injection of 10⁶ MiaPaCa2 cells into six nude mice on the right side or 10⁶ MiaPaCa2 clones expressing shRNA targeting CD147 on the left side. The tumour size was measured using a caliper, and the volume was calculated according to the formula [(width×height×depth)/2]. All experiments were performed in accordance with institutional guidelines for the care and use of experimental animals.

Statistical analysis

Linear regression, multivariate analysis of variance (ANOVA) and Kruskal–Wallis ANOVA with multiple comparisons on ranks was calculated using Statistica 6.1 for Windows.

Results

Inhibition of CD147 inhibits cancer cell growth in vitro

To understand better the role of CD147 in pancreatic cancer we established an inducible shRNA expression model to silence CD147 in MiaPaCa2 cells. As demonstrated by quantitative reverse transcription-PCR (qRT-PCR), fluorescence activated cell sorting (FACS) analysis and immunofluorescence microscopy, induction of shRNA targeting CD147 by doxycycline inhibited expression of CD147 (fig 1A). A time course analysis of CD147 protein expression over a period of 7 days revealed that inducible shRNA expression allows for a long-term silencing of CD147 (fig 1B). CD147 has been suggested to induce MMP in tumour-associated mesenchymal cells, yet silencing of CD147 did not elicit a relevant reduction in the ability of MiaPaCa2 to induce MMP by soluble factors under our experimental conditions (Supplementary fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Inducible CD147 silencing by RNA interference (RNAi) reduces the growth rate of MiaPaCa2 cells (A) Silencing of CD147 was triggered in MiaPaCa2 cells after stable transfection with constructs coding for short hairpin RNA (shRNA) targeting CD147 (BSG shRNA) using 1 μg/ml doxycycline (Dox.) for 2 days. Subsequently, CD147 expression was assessed by real-time PCR (upper left), flow cytometry (FACS, upper right) and immunofluorescence microscopy (IFM, bottom). For real-time-PCR and FACS analysis, values are expressed as means (SE) of three independent experiments. For immunofluorescence CD147 was stained by Alexa 594 (red) and the nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI (blue)). Scrambled (SCR) shRNA was used as control. (B) Western blot demonstrating the time course of CD147 silencing in MiaPaCa2 cells after induction of shRNA targeting CD147 or scrambled shRNA with doxycycline (LG, low glycosylated CD147; HG, high glycosylated CD147). Total extracellular signal-regulated kinase (Erk) was used as an internal control. (C) Effect of CD147 silencing on the growth curve and bromodeoxyuridine (BrdU) incorporation (insets) of MiaPaCa2 cells. The growth curve was assessed by quantification of DNA on each day of the observation period of 7 days in the presence of 10% fetal calf serum (FCS). The insets demonstrate the DNA synthesis of the indicated MiaPaCa2 clones as assessed by BrdU incorporation assay with or without induction of CD147 silencing.

However, CD147 silencing inhibited the proliferation of cultured MiaPaCa2 cells as demonstrated by a growth curve (fig 1C) and a BrdU incorporation assay (fig 1C insets). The effect of CD147 knock-down on the growth rate was confirmed by transient siRNA-mediated silencing in additional cancer cell lines (Supplementary fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Pancreatic cancer cells express monocarboxylate transporter 1 (MCT1) and MCT4. MCT1 (SLC16A1) and MCT4 (SLC16A3) were assessed by real-time PCR in mRNA from (A) cultured cell lines as indicated or (B) human pancreatic ductal adenocarcinoma (PDAC) tissue samples. Results are shown as mean normalised expression (SE) using XS13 as a housekeeping gene calculated as described in the Materials and methods section. (C) Immunohistology of pancreatic cancer. Serial sections of human PDAC were stained for cytokeratin, MCT1, MCT4 and CD147 using the antibodies as described in the Materials and methods section. Immunoreactivity was visualised by horseradish peroxidase using 3,3'-diaminobenzidine as a substrate. Original magnification ×400.

CD147 silencing inhibits MCT1 and MCT4 expression and function

The growth-inhibiting effect of CD147 silencing suggests that CD147 expression confers a hitherto unrecognised direct growth advantage for pancreatic cancer cells that is independent of MMP induction in tumour-associated stroma cells. To elucidate the underlying mechanism of this effect we investigated whether the natural transporters of lactate MCT1 and MCT4 may be involved.

As demonstrated by qRT-PCR and immunohistology, cancer cells express MCT1 and MCT4 in vivo and in vitro (fig 2). MCT1 and MCT4 are expressed at varying levels in cancer cells of different origin such as PDAC (Panc1, MiaPaCa2), carcinoid (BON), gastric adenocarcinoma (AGS) and breast cancer (MDA-MB-436) (fig 2A). In these cells, the level of expression of MCT1 mRNA correlated to that of MCT4 (r² = 0.8, p<0.02).

MCT1 and MCT4 mRNA was demonstrated at varying levels by qRT-PCR in human PDAC samples (fig 2B). Differences between mRNA expression of MCT1 and MCT4 were usually less than twofold and did not reach statistical significance when averaged over all samples.

In addition, co-expression of MCT1 and MCT4 was demonstrated in tissue sections of human pancreatic cancer by immunohistology (fig 2C). However, in contrast to the results from qRT-PCR we noted a weaker signal for MCT4 as compared with MCT1, which might be related to a lower immunoreactivity of the MCT4 antibody in formalin-fixed and paraffin-embedded tissue sections (fig 2C).

We then tested whether CD147 silencing inhibits the expression and function of lactate transporters (fig 3A). CD147 silencing resulted in a downregulation of CD147 as well as MCT1 and MCT4 protein in MiaPaCa2 and Panc1 cells (fig 3A). Similar results were found for additional cell lines (Supplementary fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Silencing of CD147 inhibits monocarboxylate transporter 1 (MCT1) and MCT4 expression as well as the lactate transport rate. (A) Western blot of CD147 as well as MCT1 and MCT4 after silencing of the ancillary protein CD147. MiaPaCa2 and Panc1 cells were transfected with small interfering RNA (siRNA) oligonucleotide BSG1 48 h before western blot analysis. Cells transfected with a scrambled oligonucleotide (SCR) were used as a control. β-Actin (β-act.) was used as an internal control. (B) Rate of lactate transport in MiaPaCa2 cells with (BSG1) and without (SCR) CD147 silencing. MiaPaCa2 cells were transfected with siRNA BSG1 targeting CD147 48 h before assessment of intracellular acidification as described in the Materials and methods section. The rate of lactate transport was assessed for at least 10 individual cells and two independent experiments and expressed as mean (SD) *p<0.05. HG, high glycosylated CD147; LG, low glycosylated CD147.

To confirm that CD147 silencing and the consecutive reduction of MCT1 and MCT4 expression also inhibits the function of these transporters we assessed the rate of lactate transport in MiaPaCa2 cells with and without CD147 silencing (fig 3B). Ratiometry using BCECF-AM as a pH sensor revealed that CD147 silencing decreased the kinetics of intracellular acidification after application of exogenous lactate. This demonstrates that the CD147 silencing-induced decrease in MCT1 and MCT4 expression is associated with a reduction in lactate transport kinetics (fig 3B).

MiaPaCa2 cells exhibit the Warburg effect

To test whether MiaPaCa2 cells exhibit the Warburg effect we assessed glucose consumption in these cells (fig 4A). During lactic acid fermentation 1 mol of glucose is converted to 2 mol of lactate and, indeed, lactate was detected to be generated at approximately double the rate (0.30 (nM/(h×μg DNA)) R = 0.999, p<0.0001) of glucose consumption (0.14 (nM/(h×μg DNA)) R = 0.99, p<0.0001) (fig 4A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

MiaPaCa2 cells require CD147 for lactate transport. (A) The concentration of lactic acid and glucose was assessed in the supernatant of MiaPaCa2 cells over a period of 12 h at the time points indicated. The rate of lactic acid production or glucose consumption was assessed as the slope of a linear regression trend line. (B, C) Inhibition of oxidative phosphorylation by sodium azide (NaN₃) increases glucose consumption, lactic acid production (B) and monocarboxylate transporter 1 (MCT1) and MCT4 expression (C) in MiaPaCa2 cells. MiaPaCa2 cells were treated with NaN₃ at the concentrations indicated 48 h before (B) the glucose and lactate concentrations in the supernatant were determined or (C) MCT1, MCT4 and CD147 (HG, high glycosylated; LG, low glycosylated) were detected in lysed cells by western blot. (D) Intracellular lactate in MiaPaCa2 cells after CD147 silencing. MiaPaCa2 cells were transfected with small interfering RNA (siRNA) oligonucleotide BSG1 targeting CD147 for 48 h before assessment of the intracellular lactic acid. Cells transfected with a scrambled oligonucleotide (SCR) were used as a control. (E) Growth of MiaPaCa2 cells in the presence of increasing concentrations of lactic acid. Lactic acid was added to the supernatant of MiaPaCa2 cells at the concentrations indicated 48 h before assessment of DNA content of the respective wells. (**p<0.01; ***p<0.001; results are expressed as mean (SEM).). β-act., β-actin (used as an internal control).

To test whether the increased production of lactate that can be observed in cancer cells exhibiting the Warburg effect may contribute to upregulation of MCT1 and MCT4 expression, we induced an increase in glycolysis using the mitochondrial respiratory complex IV inhibitor sodium azide (fig 4B, C). Sodium azide treatment resulted in an increase in lactate production (fig 4B) as well as MCT1 and MCT4 expression (fig 4C). Interestingly, azide treatment also resulted in a greater expression of the high glycosylated form of CD147 (fig 4C).

To address the question of whether CD147 expression is functionally relevant for lactate transport in cancer cells exhibiting a glycolytic phenotype, we examined the effect of CD147 silencing on the intracellular concentration of lactate in MiaPaCa2 cells (fig 4D). CD147 silencing resulted in a 2.2-fold increase in intracellular lactate concentration as compared with control (fig 4D).

To elucidate whether elevated concentrations of lactate may influence cell growth we investigated the effect of exogenous lactate on DNA of MiaPaCa2 cells. After 48 h incubation, lactic acid reduced the number of MiaPaCa2 cells, assessed by DNA quantification, in a dose-dependent fashion (Fig. 4E).

CD147 silencing inhibits invasion and tumourigenicity of MiaPaCa2 cells in in vivo models

To test whether CD147 silencing also reduces cell growth in vivo, the effect of CD147 silencing on the invasion of MiaPaCa2 cells through the CAM of chick embryos as well as the tumourigenicity of xenografts in nude mice was investigated (fig 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

CD147 RNA interference inhibits invasion and tumour growth in vivo. (A) CD147 immunostaining of tumours induced by incubating 10⁶ MiaPaCa2 cells embedded in Matrigel for 4 days to induce tumours on the chorioallantoic membrane (CAM) of chick embryos demonstrating successful CD147 silencing. Cancer cells were identified by cytokeratin staining. Immunoreactivity was visualised by horseradish peroxidase staining using Fast Red as chromogen; nuclei were counterstained with H&E. (B) The number of induced tumours that invaded through the CAM and formed subepithelial tumours. Without doxycycline (–Dox.), an invasive tumour was observed on 10 (90%) of the 11 chick embryos inoculated with MiaPaCa2 cells. Silencing of CD147 by induction of short hairpin (shRNA) expression (+ Dox.) reduced the number of invasive tumours to 50% (6 of 12). (C) Xenografts were induced in nude mice by subcutaneous injection of 10⁶ MiaPaCa2 wild-type cells as a reference on the right side and the same number of clones stably transfected with shRNA expression vector targeting CD147 (clone) cells on the left side of the mice. Six mice were fed drinking water containing doxycycline (+ Dox.) and another six mice served as a control group (− Dox.). Tumours became apparent 2 weeks after induction. The volume of the induced tumours was measured every week over a period of 9 weeks beginning from the second week after subcutaneous injection. A statistically highly significant difference (p<0.001) was reached after 9 weeks. Data are expressed as means ± 95% CI.

The tumours induced on the CAM by cancer cells, which express shRNA, targeting CD147, exhibited a much lower CD147 expression as detected by immunohistology (fig 5A). Of 11 tumours induced without CD147 silencing, 10 (91%) were invasive whereas out of 12 tumours induced by cancer cells with CD147 silencing, only six (50%) invaded through the CAM (fig 5B). These results demonstrate that CD147 silencing reduces the number of invasive tumours by CD147 silencing in the CAM assay by almost 50% (fig 5B).

In addition, CD147 silencing decreased the tumourigenicity of pancreatic cancer in a xenograft model in nude mice (fig 5C). In mice supplemented with doxycycline, xenografts induced by wild-type MiaPaCa2 cells grew progressively up to a mean size of 89.7 mm³ whereas clones targeting CD147 grew only up to a mean size of 39.7 mm³, which represents a 55.7% reduction in tumour size (fig 4C). In the control group of mice that did not receive doxycycline, no significant difference in tumour size was observed (fig 5C).

Discussion

CD147 is a multifunctional glycoprotein14 31 32 33 which is enriched on PDAC cells.3 4 High expression levels of CD147 have been associated with poor prognosis in many malignancies; the molecular mechanisms involved and the role of CD147 in pancreatic cancer, however, remain poorly understood. We demonstrate here that inhibition of CD147 by RNA interference directly inhibits the growth of cancer cells in vitro, indicating that CD147 expression confers a direct growth advantage for cancer cells.

The present results show that silencing of CD147 resulted in a clear reduction of MCT1 and MCT4 expression, supporting the concept that CD147 is an ancillary protein required for the expression of these MCTs.14 Several lines of evidence suggest that MCT1 and MCT4 are heteromeric transporters composed of an MCT and a CD147 subunit.11 13 14

Heteromeric membrane proteins are assembled in the endoplasmic reticulum prior to further transport to the membrane via the Golgi complex34 and thus also MCT1 and MCT4 have been demonstrated to be required for glycosylation and cell surface expression of CD147.11 13 Consequently, MCT4 has been suggested to be coupled to CD147 overexpression in metastatic cancer cells.11

Only recently, MCT4 has been demonstrated to be upregulated by hypoxia-inducible factor α (HIF1α),35 which is overexpressed in multiple malignancies36 and crucial for the glycolytic phenotype of cancer cells.37 Though MCT1 expression is not responsive to HIF1α,35 there are multiple pathways that regulate expression of genes associated with the glycolytic phenotype of cancer cells. We demonstrate here that MCT1 expression in cancer cells exhibits a high correlation with MCT4 expression and that MCT1 is induced in cancer cells by increased glycolysis. Thus, our results suggest the CD147 cell surface expression in cancer cells that exhibit a glycolytic phenotype is associated not only with MCT411 35 but also with MCT1 expression.

Inhibition of MCT1 and MCT4 by CD147 silencing resulted in a reduced lactate transport capacity in cancer cells. MCT1 and MCT4 are the natural lactate transporters that channel lactate from muscle cells to the blood, and vice versa.17 PDACs exhibit excessive glucose consumption, as demonstrated by the trapping of fluorodeoxyglucose utilised for positron emission tomography.38 Cumulating evidence suggests that high rates of glucose utilisation by cancer cells are a result of one of the earliest hallmarks of cancer, the Warburg effect, which is characterised by excessive production of lactate in cancer cells.19 Though the Warburg effect is usually not associated with mitochondrial defects,21 the lactate production due to the Warburg effect is, in contrast to the Pasteur effect, not reduced by the presence of oxygen. In the light of present knowledge, the Warburg effect is, in contrast to the initial claim of Warburg,19 not the cause of cancer,37 but recent evidence suggests that the Warburg effect confers a significant growth advantage for cancer cells. For example, it has been demonstrated that inhibition of the Warburg effect by targeting enzymes critically involved in glycolysis such as hexokinase-II,23 pyruvate kinase,22 lactate dehydrogenase21 and pyruvate dehydrogenase20 represents a novel promising avenue to inhibit cancer growth.

Our results demonstrate that PDAC cell lines may also exhibit the Warburg effect. This finding is in accordance with the idea that the Warburg effect is a result of stable genetic and epigenetic changes and thus cultured cancer cells typically maintain their metabolic phenotype.37

Excess intracellular lactate is channelled out of cells by lactate transporters MCT1 and MCT4 together with protons,17 and our results demonstrate that the reduced lactate transport capacity induced by CD147 silencing results in intracellular lactate accumulation. Our results indicate that lactate may reduce cell proliferation, a finding which is also supported by previous reports.39 The increase in lactate concentration may reduce cell growth since lactate has been demonstrated to decrease pyruvate reduction to lactate by inhibition of lactate dehydrogenase (LDH).39 40 Inhibition of LDH has been demonstrated to reduce the Warburg effect and to increase mitochondrial respiration, thereby compromising the hypoxia tolerance and reducing proliferation as well as tumourigenicity of cancer cells.21

These in vitro findings suggest that the chaperone function of CD147 for MCT1 and MCT4 is a relevant molecular mechanism that links CD147 expression to poor prognosis in malignancies that depend on the Warburg effect. Most recently, it has been demonstrated that MCT1 inhibition by a newly characterised small molecule can reduce the proliferation of activating T cells,27 further supporting the relevance of lactate transporters for proliferation of cells that exhibit increased glycolysis.

In addition to the role of CD147-associated lactate transporters for cell proliferation, some of the phenotypes observed in CD147 knockout mice (ie, impaired embryo implantation and spermatogenesis)8 might be explained by inhibition of the lactate transporters MCT1 and MCT4. For example, certain spermatogenic cells depend on l-lactate as an energy supply and also express MCT1 and MCT4.41 Furthermore, cells in the early embryo are exposed to hypoxic conditions in early development stages, rendering them dependent on anaerobic glycolysis for energy supply.42

One of the most remarkable effects of CD147 silencing was the ability to inhibit tumourigenicity and invasion of PDAC cells in in vivo models. Inducible CD147 silencing substantially inhibited the invasion of MiaPaCa2 cells through the CAM of chick embryos and significantly inhibited their tumourigenicity in a xenograft model in nude mice. In addition to the data presented in this study, a role for the association of CD147 in cancer was also shown in a xenograft model of malignant melanoma, where CD147 silencing significantly decreased growth of the tumours.43 44 An antitumour potential of MCT1 inhibition was most recently supported by a reduction of tumour growth in in vivo models of lung carcinoma, colorectal carcinoma and a squamous carcinoma cell line after α-cyano-4-hydroxycinnamate-mediated MCT1 inhibition.45

In addition to the demonstrated in vitro effects, reduction of the lactate export rate by silencing of CD147 in vivo may also lead to conditions in the microenvironment which are less advantageous for tumour growth.11 44 46 Since MCT1 and MCT4 are proton-coupled symporters, they may support acidification of the surrounding microenvironment by glycolytic cancer cells, thereby facilitating cell migration.46 Therefore, inhibition of cell migration after CD147 silencing11 may be due to an increase in extracellular pH. Furthermore, since lactate is supposed to be a signalling molecule that induces vascular endothelial growth factor (VEGF) expression in macrophages and fibroblasts,47 48 lactate transporters could support the process of lactate-induced upregulation of VEGF, thereby promoting angiogenesis in the peritumoural stroma. Indeed, CD147 silencing has been demonstrated to reduce VEGF expression in xenograft models.44 Finally, inhibition of MCT1 has also been suggested to result in glucose deprivation as well as cell death of hypoxic cancer cells and central tumour necrosis by decreasing lactate oxidation in well oxygenated cancer cells with a subsequent increase in glucose consumption for oxidative phosphorylation.45

Taken together, our results demonstrate for the first time that CD147 expression confers a direct significant growth advantage for PDAC cells by a mechanism that involves the function of CD147 as an ancillary protein required for the function and expression of MCT1 as well as MCT4. Furthermore, our findings support the idea that the Warburg effect increases the malignant potential of PDACs and indicates that inhibiting the CD147/MCT transporters represents a potential avenue to inhibit the growth of pancreatic cancer cells.