Enhanced migration of breast and lung cancer cells deficient for cN-II and CD73 via COX-2/PGE2/AKT axis regulation

Abstract

Purpose

Purine metabolism involves various intracellular and extracellular enzymes, including cN-II and CD73 that dephosphorylate intracellular and extracellular nucleoside monophosphates into their corresponding nucleosides. We conducted a study to better understand the biological roles of these enzymes in breast and lung cancer cells.

Methods

We modified cN-II and/or CD73 expression in human breast cancer cells (MDA-MB-231), human lung cancer cells (NCI-H292) and murine breast cancer cells (4T1) using the CRISPR/Cas9 technique, and evaluated their impact on various cellular parameters such as proliferation, migration, invasion, intracellular nucleotide pools and nucleotide metabolism-related gene expression under extracellular nucleotide stress conditions.

Results

Intracellular nucleotide contents were found to be altered in the modified cancer cell models both at their basal levels and after exposure to adenosine or AMP. Altered cN-II and CD73 levels were also found to be associated with cell migration and invasion alterations, involving TIMP-2, MMP-2 and MMP-9 expression, as well as alterations in the COX-2/PGE2/AKT pathway.

Conclusion

Our results highlight new cell-specific roles of cN-II and CD73 in cancer cell biology and provide insight into their interactions with different intracellular pathways.

Electronic supplementary material

The online version of this article (10.1007/s13402-020-00558-w) contains supplementary material, which is available to authorized users.

Introduction

The purine nucleoside adenosine and its phosphorylated metabolites play major roles in normal human physiological and pathological processes. Within a tumor, ATP can stimulate the anti-tumoral response of the infiltrated immune system as well as inhibit the growth of cancer cells, whereas adenosine can inhibit the immune system and induce apoptotic death of cancer cells at higher concentrations [1]. These effects are tissue- and cell type-specific and depend on the expression of ATP (P2) and adenosine (P1, ADORA) receptors, as well as the associated intracellular signal transduction machinery. Adenosine can also enter the cell and exert its biological effects after phosphorylation to AMP or ATP [2].

As the effects of adenosine are dependent on its concentration, the expression and activities of purine metabolism enzymes within tumor cells are expected to play important roles in various biological processes. In the extracellular compartment, adenosine is produced through the degradation of ATP by CD39 and CD73 [3]. It can, thereafter, either interact with its receptors, enter the cell through nucleoside transporters or be transformed to inosine by adenosine deaminase (ADA) [3]. Intracellularly, adenosine, that may originate either from the extracellular compartment or the hydrolysis of intracellular AMP or S-adenosyl-methionine, becomes rapidly phosphorylated to AMP by adenosine kinase (ADK) and, thereafter, to ADP and ATP. Intracellular purine metabolism involves a number of enzymes including ADA and several nucleotide degrading enzymes such as cN-I and cN-II [2].

Several of the aforementioned proteins have gained increased attention over the last decade, and we are particularly interested in the cytosolic 5′-nucleotidases CD73 (NT5E) and cN-II (NT5C2). The latter is an IMP/GMP-preferring enzyme for which phosphotransferase activity has also been described [4], and both enzymes have demonstrated roles in cancer cell biology. Indeed, the proliferation of astrocytoma cells (ADF) has been found to be increased in case of enhanced cN-II expression [5], whereas its down-regulation did not affect the proliferation of other cells of various origins stably expressing cN-II-targeting shRNA [6, 7]. However, in the breast cancer cell line MDA-MB-231, downregulation of cN-II by stable shRNA transfection was found to be associated with an increased adaptability to glucose starvation [8], and in the lung carcinoma cell line A549 its downregulation was found to induce increased p53 phosphorylation [9], indicating an important role in cancer cell biology. cN-II is also known to be involved in responses to cancer treatment [5, 7, 10]. Concerning CD73, modulation of its activity by enzymatic inhibitors or of its expression level has revealed an involvement in cancer cell proliferation [11–13], migration [14, 15] and sensitivity to radiation-based or targeted treatments [16, 17]. Both cN-II [18] and CD73 [19–21] are currently considered as potential therapeutic targets.

In the current work, we studied the implications of CD73 and cN-II expression in cancer cell biology using MDA-MB-231, NCI-H292 and 4T1 cell models expressing both CD73 and cN-II, only CD73 or cN-II or none of these two. In addition to studying these models under classical cell culture conditions, we evaluated their responses to high concentrations of AMP and adenosine corresponding to those that can be encountered during extracellular nucleotide stress. Part of this work has previously been presented in a thesis [22].

Material and methods

Cell culture and transfection

Human triple negative breast cancer cell line MDA-MB-231, human adenocarcinoma cell line NCI-H292 and murine breast cancer cell line 4T1 were obtained from the ATCC and cultured in complete Roswell Park Memorial Institute (RPMI-1640, Gibco) medium supplemented with 10% (v/v) fetal bovine serum (FBS, Thermofisher Scientific- Courtaboeuf, FRANCE), fungizone (2 μg/ml), penicillin (100 U/ml) and streptomycin (100 mg/ml) in collagen-coated flasks in a humidified atmosphere containing 5% CO₂ at 37 °C. Cells were routinely tested for Mycoplasma infection every two weeks. CD73 and/or cN-II knockout cells were generated using CRISPR/Cas9 technology. To this end, oligonucleotides were inserted into pLentiCRISPRv2-blast or pLentiCRISPRv2-puro plasmids (Addgene, Cambridge, MA, USA) using BsmBI (ThermoFisher-Fermentas). Viruses were produced using HEK 293 T cells during a 24h incubation and used for the infection of cells, after which stable cell models were selected using puromycin and/or blasticidine. Target RNA sequences for the CD73 and cN-II plasmids were 5’-CCACTAGCATCTCAAATATC-3′ and 5’-CTTGTCTTTGACACACTGTA-3′, respectively, in human MDA-MB-231 and NCI-H292 cells and 5’-ACCGTCGAGAAGCCTATCAC-3′ for cN-II in murine 4T1 cells.

Protein expression analysis using Western blotting and flow cytometry

Cells (0.5–1.10⁶ per flask) were seeded and allowed to adhere before being exposed to experimental conditions for the indicated times. Next, the cells were rinsed with PBS and lysed with RIPA buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 1 M DTT, 1 M NaF, protease inhibitor cocktail, phosphatase inhibitor buffer and 100 mM sodium orthovanadate). After centrifugation (15 min, 12,000 g, 4 °C), supernatants were collected and proteins (60 μg per condition) were separated by electrophoresis and transferred to nitrocellulose membranes using iBlot™2. TBS or PBS Odyssey Blocking buffer were used to block the membranes and dilute the primary antibodies, and PBS or TBS to rinse the membranes and dilute the secondary antibodies. The following antibodies were used: anti-cN-II (H00022978-M02, Novus Biologicals, Lille, FRANCE: 1/500), anti-pAkt (4060, Cell Signaling, Saint Quentin Yvelines, FRANCE: 1/500), anti-Akt (5239, Cell Signaling: 1/500), anti-COX-2 (12282S, Cell Signaling: 1/500), anti-actin (A5441, Sigma, Saint-Quentin Fallavier, FRANCE: 1/5000), anti-murine IgG antibody (IRDye^® 800CW, 1/5000; LI-COR Biosciences, Bad Homburg, GERMANY) or anti-rabbit IgG antibody (IRDye^® 680, 1/5000; LI-COR Biosciences). Protein expression was visualized using an Odyssey infrared system (LI-COR Biosciences).

Surface CD73 and CD44 expression patterns were evaluated by flow cytometry using anti-CD73 FITC-labeled (561,254, BD Biosciences - Le Pont de Claix, FRANCE: 1:100) and anti-CD44 APC-labeled (A10193, BD Biosciences: 1/100) antibodies. Cells were harvested, pelleted and washed in PBS before 30 min staining with antibodies or control isotype antibodies (IgG1 FITC-labeled: 555748, BD Biosciences: 1:100 and IgG1 APC-labeled: 555751, BD Biosciences: 1:100).

Quantitative real-time PCR analysis

mRNA extraction was performed on cell pellets (10⁶ cells) using a Qiagen column extraction kit (Les Ulis, FRANCE), following the manufacturer’s protocol. One microgram of mRNA was used for reverse transcription with M-MLV reverse transcriptase (InVitrogen, Cergy Pontoise, FRANCE). The resulting cDNA was diluted, and relative gene expression was determined by PCR in a final volume of 5 μl using a Takyon NO ROX SYBR MMix dTTP blue mix (Eurogentec, Angers, FRANCE). Runs were performed on a Lightcycler (LC480, Roche Life Science). Quantification was performed by the ΔΔCT method using 28S mRNA expression as a housekeeping gene. Primers used for each gene are listed in Table S3.

Intracellular nucleotide quantification

Cells (2.10⁶ per flask in 25 mm² flasks) were seeded, allowed to adhere for 24 h and next incubated 1 h in the presence or absence of 400 or 1600 μM adenosine (Sigma) or AMP (Adenosine 5′-monophosphate sodium salt, Sigma). Subsequently, the cells were rinsed three times with cold PBS after which nucleotides were extracted using a cold mixture of methanol/water (70/30, v/v). The extracted nucleosides and nucleotides were quantified using a validated on-line extraction coupled with LC-MS/MS as described elsewhere [52]. The nucleotide content of each sample was calculated as the peak area of the compound of interest divided by the peak area of the corresponding internal standard, further divided by the number of cells as determined in a flask containing cells cultured under the same conditions. For each nucleoside and nucleotide, internal standards were correspondingly labeled nucleotides, except for IMP for which we used GTP.

Cell proliferation assay

Cells were trypsinized, rinsed with PBS and labeled with a solution containing 10 μM Carboxyfluorescein Diacetate Succinimidyl Ester (CFSE) in PBS-0.1% BSA. After labeling, the cells were rinsed with culture medium and seeded in 6-well plates (200,000 cells/well) and allowed to adhere before being exposed to adenosine or AMP. Cells from individual wells were scratched for every time point, to evaluat CFSE fluorescence by flow cytometry on the BD FACSCalibur.

Cell survival assay

Cells (50,000 per well in 24 well-plates) were cultured in the presence or absence of the indicated compounds. At the indicated times, cells were trypsinized, washed and stained with an anti-Annexin V-FITC labeled antibody and propidium iodide (PI) using an Annexin-V FLUOS kit (11,988,549,001, Roche; Manheim, GERMANY). The resulting fluorescence was measured using flow cytometry on a BD LSR-II Flow Cytometer. Annexin-V and/or PI-positive cells were considered as dead or dying cells.

Cell migration assay

For IncuCyte migration analyses, cells (50,000 per well for MDA-MB-231 and 70,000 per well for 4T1) were seeded in an ImageLock 96-well plate (Essen BioScience, Welwyn Garden City, Hertfordshire, UK) and cultured till confluence. Next, a wound was generated using a Woundmaker 96 (Essen BioSCience), after which the cells were rinsed with PBS and incubated under the indicated conditions. Wound closure was monitored every 2 h using an IncuCyte® device. For AKT siRNA experiments, cells (100,000 per well in 50 μl) were seeded in similar plates in wells containing 50 μl of a mix with control (Qiagen, reference 1,027,281) or AKT (5′-AAGGTTTAAATTTGTTATTGT-3′, Qiagen, reference SI00073192) siRNA (10 pmol per well), 0.25 μl Lipofectamine 2000 (Life technologies) and OptiMEM, and incubated for 6 h before the medium was changed to complete cell culture medium. After another 48 h incubation, wounds were generated and analyzed as indicated above.

For xCELLigence migration analyses, cells (30,000 per well for MDA-MB-231 and 40,000 per well for 4T1) were seeded in serum-free media in a CIM-Plate 16 in which complete medium was added in the lower chambers following the manufacturer’s instruction (ACEA Biosciences, San Diego, CA, USA). Cell indexes were recorded for at least 24 h using a xCELLigence device.

For transwell migration analyses, cells (50,000 per well) were seeded in 300 ml serum-free medium in transwell chambers (8.0 μm pore size, Thincerts Greiner) inserted into 24-well plates containing 800 μl complete medium (with 10% serum). After 15 h of culture, remaining cells in the upper chamber were removed with a cotton swab, and the cells on the lower surface were fixed and stained with Coomassie bleu solution. Cells were imaged by microscopy (10x) and quantified using ImageJ software.

PGE2 quantification assay

Cells were seeded in 24-well plates (500,000 per well) and allowed to adhere before being exposed to fresh culture medium with or without 15 μM arachidonic acid (Abcam; Paris, FRANCE). Subsequent PGE2 quantification was performed on cell supernatants after a 24-h incubation, using a Parameter™ Prostaglandin E₂ assay from R&D Systems®, according to the manufacturer’s protocol.

Statistical analysis

Statistical analyses were performed using unpaired Student’s t test using Microsoft Excel. A p value < 0.05 was considered statistically significant. No correction for multiple testing was performed.

Results

Generation and characterization of cancer cell line models

Using the CRISPR/Cas9 technique, we abolished cN-II and/or CD73 expression in human MDA-MB-231 and NCI-H292 cells, and cN-II expression in murine 4T1 cells. These cells were validated for cN-II and CD73 protein expression and hereafter referred to as cN-II⁺/CD73⁺ cells (expressing both cN-II and CD73), cN-II⁺/CD73⁻ cells (expressing cN-II but not CD73), cN-II⁻/CD73⁺ cells (deficient for cN-II but not for CD73) and cN-II⁻/CD73⁻ cells (deficient for both cN-II and CD73) (Fig. 1). We next determined the expression profiles of selected genes coding for adenosine receptors or proteins involved in purine metabolism in these human cancer cell models (Table S1). Among the notable differences observed in MDA-MB-231 cells, we observed a 6–8-fold decrease in the expression of adenosine receptor A1 in cN-II negative cells, an 11-fold increase of ADSL in cN-II⁺/CD73⁻ cells as well as a 1.5–1.7-fold increase in SAMHD1 in cN-II negative cells compared to their respective controls. cN-II⁻/CD73^- NCI-H292 cells showed a slight decrease in adenosine receptor A2 (1.4-fold) and RRM1 (2.3-fold) expression.

Fig. 1.

Characterization of cell models. a CD73 and/or cN-II knockout in MDA-MB-231, NCI-H292 and 4T1 cells obtained using CRISPR/Cas9 technology. b cN-II expression in the different cell models determined by Western blotting. Representative images of 3 similar Western blots are shown. c CD73 expression in MDA-MB-231 and NCI-H292 cell models determined by flow cytometry. The images shown are representative of 4 analyses

Extracellular adenosine affects cancer cell proliferation and survival, irrespective of cN-II or CD73 expression status

Considering the involvement of nucleotide pools in cellular processes, we set out to investigate whether cN-II and/or CD73 knock out may affect cancer cell proliferation and/or viability. Under normal cell culture conditions, these knockouts did not affect the proliferation of MDA-MB-231, NCI-H292 or 4T1 cells. The CSFE intensities decreased to 23.5–25.4% (MDA-MB-231 cells), 15.0–17.6% (NCI-H292 cells) and 7.1–8.6% (4T1 cells) of the initial values after 3 days (Fig. 2a). In addition, no changes in viability were observed for the different MDA-MB-231 cells after 72 h in culture (16.2 ± 7.9% Annexin-V positive cells for cN-II⁺/CD73⁺, versus 17.6 ± 3.5%, 15.7 ± 8.0% and 16.7 ± 8.3% for cN-II⁻/CD73⁺, cN-II⁺/CD73⁻ and cN-II⁻/CD73⁻ cells, respectively) (Fig. 2b). These results indicate that cN-II and CD73 are not involved in the proliferation and/or survival of these cells under the applied conditions.

Fig. 2.

Proliferation and survival of cell models. a MDA-MB-231 (left panel), NCI-H292 (middle panel) and 4T1 (right panel) cells were stained with CFSE to monitor cell proliferation over 3 days by flow cytometry in the absence (continuous line) or presence of 1600 μM adenosine (·····) or AMP (- - -). The graphs show mean values of 4 independent experiments and error bars as standard deviations. * p < 0.05 and *** p < 0.005 using Students t-test for comparisons between adenosine and control conditions. b Cell death was evaluated in the MDA-MB-231 cell models under control conditions (white bars) or in the presence of adenosine (1600 μM, light grey bars), adenosine (1600 μM) + A2A antagonist ZM 241385 (100 nM) + A2B antagonist PSB 1115 (10 μM) (dark grey bars), AMP (1600 μM, black bars), AMP (1600 μM) + APCP (100 μM) (vertical lines) or APCP (100 μM) (horizontal lines). The graphs show mean values of 5 independent experiments and error bars as standard deviations. * p < 0.05 and ** p < 0.01 using Students t-test in comparison with the control condition

As cN-II and CD73 are involved in nucleotide metabolism, we next investigated to what extent these enzymes may modify cellular responses to extracellular nucleotide stress using the MDA-MB-231 cell models. To do so, we exposed the cells to high concentrations of AMP or adenosine (1600 μM) and evaluated their survival after 48 h. We found that AMP exposure resulted in a strong increase of Annexin-V positive cells in CD73⁺ cells (from 16.2 ± 7.9% to 41.6 ± 12.2% for cN-II⁺/CD73⁺, and from 15.7 ± 8.0% to 32.2 ± 10.4% for cN-II⁻/CD73⁺ cells). This AMP-induced cell death was completely inhibited by the CD73 inhibitor APCP (adenosine 5′-(α,β-methylene)diphosphate), indicating that CD73 activity, and thus adenosine generation, is required for this effect in these cancer cell models. Interestingly, cN-II⁻/CD73⁻ cells also showed a slight sensitivity to AMP (27.8 ± 4.2% vs 16.7 ± 8.3% for cN-II⁺/CD73⁺, p < 0.05), but no sensitivity to APCP. Overall, we found that exposure to adenosine induced cell death in all cancer cell models tested. This effect was not significantly different with respect to cN-II or CD73 expression, suggesting that these 5′-nucleotidases do not modulate cancer cell survival in the presence of high concentrations of adenosine. When the cells were incubated with adenosine and A2A and A2B antagonists, we did not observe any rescue. Thus, A2A and A2B receptors do not appear to mediate adenosine-induced cell death under these conditions.

Similarly, we evaluated cancer cell proliferation in the presence of high initial concentrations of AMP or adenosine (Fig. 2a). We found that both purines slowed the proliferation down of MDA-MB-231 cells, but that adenosine had a stronger effect than AMP (p < 0.05 for adenosine, p > 0.05 for AMP). Indeed, in presence of this nucleoside, the CFSE intensity after 3 days reached 35.3–50.1% of the initial value, versus 23.5–25.4% without adenosine. Here again, no significant difference was observed with respect to cN-II and CD73 expression between the cell lines. 4T1 cells were also sensitive to adenosine (20.4–23.7% at day 3 versus 7.1–8.6% under control conditions), but not to AMP. Finally, NCI-H292 cells showed no alterations in proliferation in the presence of high concentrations of either adenosine or AMP (1600 μM), and this was the case for all four models (Fig. 2a). Together, these results indicate that extracellular nucleotide stress can affect cell proliferation and cell survival in a cell type-specific manner, irrespective of cN-II or CD73 expression.

Determination of intracellular nucleotide pools in cancer cell models

As indicated above, both cN-II and CD73 are expected to regulate intracellular nucleotide pools. Therefore, we measured the pools, with a particular interest in purines, in the MDA-MB-231 cell models both under baseline conditions and after 1 h exposure to 400 or 1600 μM adenosine or AMP (Fig. 3 and Table S2). Exposure to 2-deoxyglucose was used as a control condition inducing major modifications in nucleoside triphosphate (NTP) content. We found that both ATP and AMP were more abundant in cN-II⁻/CD73⁺ and cN-II⁻/CD73⁻ cells (1.4–1.6-fold) and unaltered in cN-II⁺/CD73⁻ cells compared to control cells, whereas the adenosine and inosine levels were decreased in both cell models lacking CD73 (1.4–3.2-fold) and increased in cN-II⁻/CD73⁺ cells (4.6- and 2.1-fold, respectively). Finally, IMP was found to be decreased in cN-II⁻/CD73⁺ cells (3.4-fold) and increased in cN-II⁻/CD73⁻ cells (5.7-fold). After exposure to adenosine or AMP, control cells exhibited increased levels of ATP, AMP, inosine and adenosine. In the modified cells, the most striking modifications observed were increases in AMP and adenosine in cN-II⁻/CD73⁺ cells and of IMP in cN-II⁻/CD73⁻ cells. The lack of precision in samples exposed to adenosine or AMP resides in the matrix effect during LC-MS/MS analysis.

Fig. 3.

Relative contents of intracellular adenosine, inosine, AMP and IMP. In the studied MDA-MB-231 cell models, cN-II⁺/CD73⁺ is indicated in white, cN-II⁺/CD73⁻ in bright grey, cN-II⁻/CD73⁺ in dark grey and cN-II⁻/CD73⁻ in black. Nucleotides and nucleosides were quantified as indicated in Material and methods and are expressed as mean values of the ratios of the surface of the compound of interest / surface of the internal standard. The results where further normalized for one million cells and expressed as relative content compared to unexposed cN-II⁺/CD73⁺ cells in each experiment. The values shown are from three independent experiments. For complete data, see Table S2

5’-Nucleotidase expression and extracellular adenosine modulate cell migration

Cell migration was determined using a wound healing assay and monitored using an IncuCyte^® device. Under normal cell culture conditions, reduced cN-II expression was found to be associated with enhanced migration of MDA-MB-231 cells (Fig. 4a and S1), i.e., 10 h after injury, wound closure reached 70 to 77% for cN-II⁻/CD73⁺ and cN-II⁻/CD73⁻ cells, whereas it reached only 49 to 50% for cN-II⁺/CD73⁻ and cN-II⁺/CD73⁺ cells. For NCI-H292 cells, CD73 expression deficiency was found to be associated with a decreased migration, whereas cN-II modulation did not affect this parameter, i.e., 10 h after injury, wound closure reached 72 to 79% for cN-II⁺/CD73⁺ and cN-II⁻/CD73⁺ cells, whereas it reached only 43 to 55% for cN-II⁺/CD73⁻ and cN-II⁻/CD73⁻ cells. In 4T1 cells, cN-II deficiency did not alter its migration (Fig. 4a). We also used a transwell assay to assess the migration of MDA-MB-231 cells, and observed a 1.7-fold decrease in migration for cN-II⁺/CD73⁻ cells (p = 0.01) and a 1.9-fold increase for cN-II⁻/CD73⁻ cells (p = 0.006) compared to cN-II⁺/CD73⁺ cells (Fig. 4b). Thus, the effect of 5′-nucleotidases on cell migration was confirmed using this method. Even though these migration results were repeatedly obtained using the Incucyte device and the transwell assay, the xCELLigence device with CIM plates did not give the same results for MDA-MB-231 and 4T1 cells. Instead, we found that cN-II⁻/CD73⁻ cells migrated faster and cN-II⁺/CD73⁻ cells slower than control MDA-MB-231 cells, and that cN-II⁻/CD73⁺ cells migrated faster than control 4T1 cells (Fig. S2). This discrepancy between these two methods may be attributed to the fact that the xCELLigence device takes the strength of adherence of the cells into account in addition to the number of cells, suggesting a difference in adherence between our models. We also monitored MDA-MB-231 cell migration in the presence of high initial concentrations of AMP. In cN-II⁺/CD73⁺ and cN-II⁻/CD73⁺ cells, we found that AMP delayed the migration and that this effect could be reversed by co-incubation with APCP, suggesting that its conversion into adenosine may be necessary to affect this process (data not shown). Again, we found that AMP could also affect the migration of cN-II⁻/CD73⁻ cells, suggesting a CD73-independent AMP effect.

Fig. 4.

Cell migration in cancer cell models. Cell migration assays were performed as indicated in Material and methods. a Quantification of confluence at 10 h after generating wounds in MDA-MB-231 (left), NCI-H292 (middle) and 4T1 (right) cells. Graphs represent mean values of 4 independent experiments performed in triplicate, and error bars indicate standard deviation. * p < 0.05 using Students t-test in comparison with cN-II⁺/CD73⁺ cells; ^# p < 0.05 using Students t-test in comparison with the corresponding cN-II-positive cells, ^$ p < 0.05 using Students t-test in comparison with corresponding CD73-positive cells. b Quantification of MDA-MB-231 cell migration during a 15 h transwell assay period. Graphs show mean values of 12 images of four wells from two independent experiments, and error bars indicate standard deviation. * p < 0.05 and ** p < 0.01 using Students t-test in comparison with cN-II⁺/CD73⁺ cells. c Effect of adenosine (bright curves) compared to unexposed cells (dark curves) over 16 h after generating wounds in MDA-MB-231 (right), NCI-H292 (middle) and 4T1 (right) cells. Graphs show mean values of 4 independent experiments performed in triplicate. d Quantification of the effect of adenosine on cell migration. Δ wound closure calculated 10 h after wounding. For each cell line we used Δ wound closure = |% wound confluence _control - % wound confluence _adenosine|. Graphs show mean values of 4 independent experiments performed in triplicate, and error bars represent standard deviation. * p < 0.05 and ** p < 0.01 using Students t-test in comparison with corresponding cN-II-proficient cells, ^# p < 0.05 using Students t-test in comparison with corresponding cN-II⁺/CD73⁺ cells

When we performed this assay in the presence of high initial concentrations of adenosine, the nucleoside reduced cell migration in all our cell models, confirming the importance of adenosine production in affecting migration (Fig. 4c). Interestingly, we found that the migration was more affected by adenosine in cN-II-deficient MDA-MB-231 cells compared to their cN-II-proficient counterparts. This was also true at some time-points for NCI-H292 cN-II⁻/CD73⁺ cells, but not for the 4T1 model. Figure 4d illustrates this notion through differences in wound closure between unexposed and adenosine-exposed cells at 10 h. The observed effects were not due to adenosine or AMP-induced cell death, as these purines do not significantly affect cell survival at early times (Fig. S3).

cN-II knock out affects migration-related molecular alterations

In order to find a molecular explanation for the differences observed in migration between cN-II-deficient and cN-II-proficient cells, we set out to assess the expression of the migration-related genes TIMP-1, TIMP-2, MMP-2 and MMP-9. We found that, at the mRNA level, cN-II-proficient MDA-MB-231 cells expressed high levels of TIMP-2, but not TIMP-1, compared to their cN-II-deficient counterparts (Table 1), i.e., cN-II knock out was accompanied by 34.0% and 47.9% reductions in TIMP-2 mRNA expression for cN-II⁻/CD73⁺ and cN-II⁻/CD73⁻ cells, respectively. MMP-2 and MMP-9 were more abundantly expressed in the CD73-deficient cells, but did not vary according to cN-II expression status. This MMP upregulation was not associated with any significant modification in cell migration. In addition, we observed a slight increase in TIMP-1 expression (1.4-fold) in cN-II⁻/CD73⁺ NCI-H292 cells compared to control cells, and in MMP-2 expression (1.7-fold) in cN-II⁻/CD73⁻ cells compared to cN-II⁻/CD73⁺ NCI-H292 cells.

Table 1.

mRNA expression of cell migration-related and PGE2-related genes in MDA-MB-231 and NCI-H292 cell models

	MDA-MB-231				NCI-H292
Gene	cN-II+/CD73+	cN-II+/CD73−	cN-II−/CD73+	cN-II−/CD73−	cN-II+/CD73+	cN-II+/CD73−	cN-II−/CD73+	cN-II−/CD73−
TIMP1	0.65 ± 0.25	1.48 ± 0.68	0.89 ± 0.50	1.18 ± 0.63	0.34 ± 0.09	0.51 ± 0.30	0.49 ± 0.08*	0.44 ± 0.05
TIMP2	1.12 ± 0.15	1.19 ± 0.07	0.69 ± 0.21*	0.72 ± 0.46**	0.25 ± 0.11	0.41 ± 0.49	0.19 ± 0.02	0.16 ± 0.03
MMP2	0.72 ± 0.26	1.20 ± 0.05#	0.78 ± 0.26	1.15 ± 0.22	1.08 ± 0.24	1.01 ± 0.58	0.93 ± 0.003	1.62 ± 0.44#
MMP9	0.88 ± 0.11	1.62 ± 0.26##	0.81 ± 0.31	1.68 ± 0.26	1.49 ± 0.54	1.76 ± 0.57	2.30 ± 1.69	1.55 ± 0.43
COX2	0.87 ± 0.16	1.18 ± 0.42	21.9 ± 6.8***	55.2 ± 33.1*	1049 ± 396	373 ± 428	1577 ± 1271	798 ± 707
PLA2G4A	1.10 ± 0.40	79.1 ± 9.8***	3.14 ± 1.18	111 ± 10***	1.37 ± 0.30	1.42 ± 0.99	1.29 ± 0.15	1.44 ± 0.48

Values are means ± standard deviation of four independent experiments, performed in triplicate. * p < 0.05, ** p < 0.01 and *** p < 0.001 using Students t-test in comparison with corresponding cN-II-proficient cells. ^# p < 0.05 and ^## p < 0.01 using Students t-test in comparison with corresponding CD73-proficient cells. See legend of Table S3 for gene names

CD44 is a glycoprotein that is known to promote migration in cancer cells in association with MMPs [23]. Thus, we set out to determine its expression at the surface of our cell models and found that cN-II-deficient MDA-MB-231 cells expressed less CD44 than their cN-II-proficient counterparts, whereas CD73-deficient NCI-H292 cells expressed less CD44 than their CD73-proficient counterpart (Fig. S4). This result suggests a link between CD44, cN-II and CD73 expression and cell migration, which appears to be cancer cell type-specific.

cN-II modulates the COX-2/PGE2/AKT signaling axis in cancer cells

Previously, it has been reported that the COX-2/PGE2/AKT signaling axis can regulate TIMP-2 expression as well as cell migration [24–27]. Indeed, cyclooxygenase 2 (COX-2) participates in prostaglandin E2 (PGE2) production from arachidonic acid in cells. The latter can bind surface G protein-coupled receptors (GPCRs), leading to the activating AKT phosphorylation, thereby promoting cell migration and regulating gene expression. Therefore, we next set out to investigate the involvement of this axis in our cancer cell models. We found that cN-II deficiency was associated with a 25–47-fold increase in COX-2 expression and a 35–72-fold increase in PGLA2 expression at the mRNA level (Table 1) in MDA-MB-231 cells, and an overt increase in COX-2 protein expression both in MDA-MB-231 and NCI-H292 cells (Fig. 5a). This increase was accompanied by a 1.6–2.1-fold higher PGE2 secretion in the supernatants of MDA-MB-231 cells (Fig. 5b). Interestingly, CD73 silencing alone was not associated with any significant changes, but when combined with cN-II silencing in MDA-MB-231 cN-II⁻/CD73⁻ cells, it enhanced COX-2 expression and PGE2 production. COX-2 expression was not altered in 4T1 cN-II⁻/CD73⁺ cells.

Fig. 5.

COX-2/PGE2/AKT axis analysis in cancer cell models. a COX-2 protein expression in MDA-MB-231, NCI-H292 and 4T1 cells determined as indicated in Material and methods under control conditions or after one-hour exposure to adenosine (1600 μM). The Western blots shown are representative of 3 experiments. b PGE2 secretion quantified in supernatants of MDA-MB-231 cells after 24-h stimulation with arachidonic acid (15 μM). Graphs represent mean values of three independent experiments performed in triplicate and error bars represent standard deviation. * p < 0.05 using Students t-test in comparison with corresponding cN-II⁺/CD73⁺ cells. c Wound healing curves of MDA-MB-231 cells exposed or not to celecoxib (− -, 60 μM) or arachidonic acid (− -, 30 μM). Graphs represent mean values of three independent experiments performed in triplicate and error bars represent standard deviation. d Quantification of the effect of celecoxib and arachidonic acid on MDA-MB-231 cell migration 10 h after wound generation. For each cell line and each condition, we used Δ wound closure = |% wound confluence _control - % wound confluence _{celecoxib or arachidonic acid}|. Graphs represent mean values of four independent experiments performed in triplicate and error bars represent standard deviation * p < 0.05 using Students t-test in comparison with corresponding cN-II-proficient cells. ^# p < 0.05 using Students t-test in comparison with cN-II⁺/CD73⁺ cells. e AKT expression and phosphorylation in cells exposed or not to adenosine (1600 μM, 1 h)

In the presence of the COX-2 inhibitor celecoxib, cell migration tended to slow down in the four cell lines tested (cN-II⁺/CD73⁺ cells were decreased to 19.5 ± 9.6% wound closure after 10 h, cN-II⁺/CD73⁻ to 21.1 ± 11.7% wound closure, cN-II⁻/CD73⁺ to 30.6 ± 8.7% wound closure and cN-II⁻/CD73⁻ to 31.2 ± 14.5% wound closure while exposed to 60 μM celecoxib) (Fig. 5c). This decrease was, however, not observed with the COX-2 inhibitors rofecoxib and valdecoxib (Fig. S5A). Conversely, we found that arachidonic acid-induced PGE2 production slightly enhanced cell migration in cN-II-proficient MDA-MB-231 cells, thus confirming that COX-2 activity is involved in this process in these models. Interestingly and similar to adenosine, we found that celecoxib was able to more efficiently reduce the migration of cN-II-deficient MDA-MB-231 cells, particularly when this was combined with CD73-deficiency (Fig. 5d). Consistently, arachidonic acid enhanced migration less efficiently in cN-II deficient MDA-MB-231 cells, suggesting that cN-II may be involved in COX-2/PGE2 pathway-mediated modulation of cancer cell migration.

AKT activation occurs downstream of prostaglandin receptor activation and is known to promote cell migration [28–31]. Therefore, we next evaluated its phosphorylation status and, by doing so, found that cN-II-negative cells, particularly cN-II⁻/CD73⁻ MDA-MB-231 cells, showed a relatively stronger basal activity of AKT (Fig. 5e). In addition, we found that incubation of the cells with 1600 μM adenosine for 1 h reduced both COX-2 expression (Fig. 5a) and AKT phosphorylation (Fig. 5e), which could explain its impact on cell migration. Using cell migration assays in presence of the AKT inhibitor VIII at 10 μM, which indeed did inhibit AKT phosphorylation (data not shown), no alterations in migration patterns in any of the MDA-MB-231 cell models were noted (Fig. S5B). This was also observed using an AKT-targeting siRNA (Fig. S6). Taken together, these results suggest that the COX-2/PGE2/AKT axis is reinforced when cN-II is silenced in MDA-MB-231 cells, and that adenosine can inhibit this axis. Validation studies using the 4T1 cancer cell model revealed that celecoxib (Fig. S7A) and rofecoxib, and to a lesser extend valdecoxib (Fig. S7B), decreased migration in both cell lines, whereas arachidonic acid (Fig. S7A) and AKT inhibitor VIII (Fig. S7C) did not alter this capacity.

Discussion

We established new cell models that allow a better understanding of the role of 5′-nucleotidases in cancer cell biology. We found that in MDA-MB-231, NCI-H292 or 4T1 cancer cells, cN-II and CD73 do not affect their proliferation and/or survival under optimal or nucleotide stress conditions. We additionally found, however, that adenosine does affect cancer cell biology in these new models. In the tumor microenvironment, extracellular adenosine concentrations may increase due to inflammatory conditions and/or the presence of damaged cells. Under these conditions, the concentrations can reach the micromolar range whereas it is in the nanomolar range under normal physiological conditions [32]. Next, adenosine is rapidly degraded by ADA or internalized by nucleoside transporters. Because of its very short half-life time, we used high initial concentrations of adenosine (1600 μM), but these conditions do not allow us to exactly delineate which concentrations remain in the medium at the time points studied. Although considered as pro-tumoral due to its immunomodulatory effects, we found that adenosine also affects cancer cells themselves by slowing down their proliferation and/or migration and triggering cancer cell death when present at very high concentrations, as suggested before [33, 34]. cN-II and CD73 do not seem to be involved in these processes.

Currently, little is known about the intracellular nucleoside and nucleotide pools in cells with altered cN-II and CD73 expression levels. In non-cancerous cells, overexpression of cN-II has been shown to only slightly decrease nucleoside triphosphate (NTP) levels [35–37]. No major differences were observed in cells stably transfected with shRNA targeting cN-II in four different models [6] and, to our knowledge, no such data exist for CD73-modified cancer cells. In our models, ATP was increased when either or both cN-II and CD73 were knocked out. We also observed increased IMP and decreased inosine levels in cN-II⁻/CD73⁻ cells, consistent with the fact that these cells will degrade less IMP into inosine. Whether these differences in nucleotide pools play a role in the phenotypic differences observed between our cancer cell models remains to be established.

Previous studies have suggested that CD73 may be involved in cancer cell migration through adenosine-dependent and adenosine-independent mechanisms. Our results support the predominant importance of this nucleoside in MDA-MB-231 and NCI-H292 cells, as adenosine could affect cell migration independent from CD73 expression. In cN-II⁻/CD73⁻ cells migration and survival were also sensitive to AMP, independent from its conversion to adenosine. This could be the result of a direct effect of the nucleotide on these cells. Indeed, it has been reported that the adenosine receptor A1 can be activated by AMP [38] and can mediate cell death [39]. In our MDA-MB-231 cell models, the adenosine receptor A1 was found to be downregulated in cN-II⁻/CD73⁻ cells at the mRNA level and, therefore, may not be responsible for this effect. Other enzymes such as prostatic acid phosphatase can hydrolyze AMP independent from CD73 and mediate cell death by generating adenosine [3]. Thus, their expression might be studied in cN-II⁻/CD73⁻ cells compared to the other cancer cell models.

Metalloproteinases 2 and 9 are gelatinases that are highly expressed in MDA-MB-231 cells in which they can regulate migration. Their activity depends on the inhibitors TIMP-1 and TIMP-2. High TIMP expression is associated with low MMP activity and reduced migration, whereas low TIMP expression is associated with enhanced MMP activity and migration [27]. We observed higher MMP-2 and MMP-9 mRNA expression levels after CD73 knockout in MDA-MB-231 cells. This phenomenon is possibly related to purine-dependent signaling. Indeed, by degrading AMP, CD73 promotes high concentrations of adenosine which, subsequently, reduce AKT activation. Downstream of AKT, the expression of several genes, including those encoding MMPs, is enhanced [40–42]. Thus, in the absence of CD73, we expected reduced pools of adenosine, leading to a better AKT phosphorylation and a higher expression of MMPs, consistent with previous studies that revealed an inhibiting effect of adenosine on MMP expression [43, 44]. However, we did not observe any detectable difference in AKT activation or migration in the absence of CD73 alone in the breast cancer cells studied. In association with cN-II deficiency, however, AKT activation was more striking. This suggests that MMP-2 and MMP-9 expression modulation upon CD73 silencing is not sufficient to significantly affect cancer cell migration, but that these alterations interact with cN-II-related modifications to accentuate the pro-migratory phenotype induced by cN-II silencing.

Similarly, COX-2 expression can be promoted by AKT activation [45, 46] and be regulated by adenine nucleotides/nucleosides. Lin et al. showed that ATP promotes COX-2 expression through NADHP oxidase activity and an increase in ROS production [45, 47]. The triphosphate nucleotide and adenosine tend to have opposite effects on cells and, in conformity with this notion, we found that the latter triggered a lower AKT activation and a lower COX-2 expression in our cancer cell models.

Furthermore, cN-II repression was found to be associated with an enhanced constitutive AKT activation that can mediate COX-2 expression in our MDA-MB-231 cell model. As mentioned above, cN-II can affect intracellular nucleotide pools and its knockout can lead to an accumulation of nucleotides and a decrease of adenosine levels in the cytoplasm. As adenosine is transferred to the extracellular space through nucleoside transporters, this decrease can also affect extracellular adenosine pools, thus resulting in a weaker activation of adenosine A (ADORA) receptors, which intuitively is not in accordance with the observed increase in AKT activation in our models. Nevertheless, we should also consider the expression and activation state of the four prostaglandin E2 (PGE2) receptors (EP1, EP2, EP3 and EP4 receptors) in our models. Indeed, these receptors also belong to the G protein-coupled receptor (GPCR) family and are expressed in MDA-MB-231 cells [48]. These receptors share intracellular mediators with ADORA signaling involving cAMP generation and AKT activation. Simultaneous activation of these signaling pathways can thus lead to different cellular responses according to the panel of receptors expressed on the cells. In addition, we previously showed that stable shRNA-mediated cN-II downregulation results in lower ROS contents [8]. If a complete knockout of this 5′-nucleotidase affects the cellular ROS contents, this should also act on COX-2 downregulation. Adenosine also has dual roles on phospholipase A2 expression and activity, which releases arachidonic acid, the PGE2 precursor [49, 50], thus revealing another link between nucleotide metabolism and the COX-2/PGE2 axis. Further studies are needed to decipher whether COX-2 is directly affected by cN-II and nucleotide metabolism. A recent study has shown that cN-II may interact with cytoplasmic proteins, opening the possibility of enzymatic activity-independent effects of the cytosolic 5′-nuceotidase. Indeed, cN-II has been found to interact with the inflammasome protein Ipaf through its leucin-rich region (LRR) [51]. Absence of this interaction could result in Ipaf oligomerization and, thus, inflammation. Other members of the inflammasome such as NLRP3 also contain a LLR region and thus may interact with cN-II, and NLRP3 has been found to regulate COX-2 expression and PGE2 production. With a similar mechanism as for Ipaf, interaction between cN-II and NLRP3 could affect COX-2 expression as observed in our models.

We showed that the COX-2/PGE2/AKT axis is involved in MDA-MB-231 cell migration and that this axis is reinforced when cN-II is absent. Indeed, the COX-2 inhibitor celecoxib reduced cell migration and, inversely, stimulating PGE2 production with arachidonic acid resulted in enhanced cell migration. Nevertheless, we found that cN-II-deficient cells were more sensitive to these effects than their cN-II-proficient counterparts. These observations suggest that cN-II may play a role in migration regulation. The mechanism through which cN-II is involved in the sensitivity of cancer cells to agents that can affect their migration remains to be elucidated. This effect was unfortunately not confirmed using other COX-2 inhibitors, which may be explained by either a difference in the effectiveness of COX-2 inhibition in our cells using the different compounds, or by putative off-target effects of celecoxib or the other two compounds. The fact that we did not see any difference between cells transfected with control siRNA and AKT siRNA can be explained by technical issues such as an actual decrease in AKT1 level at the relevant time point. Indeed, we validated the AKT siRNA sequences on MDA-MB-231 cells cultured under different conditions (i.e., flasks instead of 96-well plates) than in the wound healing assays, and the actual knock down of AKT as well as the ratio of pAKT/AKT at the relevant time point in the latter experiment remain unknown. Also, this result could indicate that the differences in migration are dependent on COX-2/PGE2 as shown with arachidonic acid and celecoxib, but not on AKT even though this protein is upregulated in cN-II-deficient cells.

We used three different techniques for the assessment of cell migration (wound healing on an Incucyte device, CIM-plates on a xCELLigence RTCA DP instrument and transwell analysis). The results we obtained with the Incucyte device were not in conformity with those obtained with the other two platforms (Fig. 4 and Fig. S2). There is, to our knowledge, no study comparing directly these three techniques, and we conclude that the molecular mechanisms involved in the migration patterns observed in the different platforms are different. Indeed, one is based on the recovery of a two dimensional surface that has already been populated by the cells, the second is based on the migration of cells from one side to the other on a plate and the signal takes into account both the number of cells and the strength of adherence, and the third is based on the migration of a cell from one side to the other on a plate, but the analysis is only taking into account the space occupied by the cells that migrated. Our conclusions on cell migration are, therefore, confined to the two dimensional recovery and migration capacities, not to adherence strength.

Taken together, we conclude that the generation of cN-II and/or CD73 knockout cell models provides new tools to better understand nucleotide metabolism in cancer cells, more specifically, possible interactions between intra- and extracellular compartments of purines. Indeed, until recently, these compartments have been studied independently and in various pathologies, whereas both direct (nucleoside and nucleotide transports) and indirect (transcriptional regulation through receptors) interplays are known to exist. Our model characterization provides further arguments to consider these 5′-nucleotidases as targets to disturb cancer cell biology and reduce their aggressiveness. The differences in behavior between our three cell models (MDA-MB-231, NCI-H292 and 4T1) show that the cN-II and CD73-mediated phenotypes are largely cell type specific, and that these may depend on the overall nucleotide metabolism in these cells.

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.