Oncogene (2013) 32, 724 --735 & 2013 Macmillan Publishers Limited All rights reserved 0950-9232/13 www.nature.com/onc

ORIGINAL ARTICLE 2 specifically regulates the oxidative and metabolic stress response of human metastatic cells in lungs

V Stresing1,2,9, E Baltziskueta1,9, N Rubio3, J Blanco3,MaC Arriba1, J Valls4,5, M Janier6, P Cle´ zardin2, R Sanz-Pamplona7, C Nieva1, M Marro8, P Dmitri8 and A Sierra1

Little is known about metastatic pathways that are specific to the lung rather than other organs. We previously showed that such as were specifically upregulated in lung metastatic breast cancer cells. We hypothesize that cancer cells that live under aerobic conditions, as might be the case in lungs, protect themselves against the damage caused by reactive oxygen species (ROS). To examine this hypothesis, we studied the role of peroxiredoxin-2 (PRDX2) in lung vs bone metastasis formation. A metastatic variant of MDA-MB-435 breast cancer cells that specifically metastasize to lungs (435-L3) was transduced with short hairpin RNAs to specifically silence PRDX2. Conversely, a bone metastatic variant of MDA-MB-231 cells (BO2) was stably transfected to overexpress PRDX2. The 435-L3 cells silenced for PRDX2 were significantly more sensitive to H2O2-induced than the parental and scrambled transfected cells. BO2/PRDX2 cells produced less ROS than BO2/green fluorescent control cells under oxidative stress. Moreover, PRDX2 knockdown inhibited the growth of 435-L3 cells in the lungs, whereas lymph node metastasis remained unaffected. In contrast, PRDX2 overexpression in bone metastatic BO2 breast cancer cells led to drastic inhibition of the skeletal tumor burden and reduction of bone destruction. Furthermore, PRDX2 expression in breast cancer cells was associated with a glucose-dependent phenotype, different from bone metastatic cells. Overall, our results strongly suggest that PRDX2 is a targetable ‘metabolic adaptor’ driver protein implicated in the selective growth of metastatic cells in the lungs by protecting them against oxidative stress.

Oncogene (2013) 32, 724--735; doi:10.1038/onc.2012.93; published online 19 March 2012 Keywords: angiogenesis; lung metastasis; metabolic stress; oxidative stress; peroxiredoxins

INTRODUCTION Different tropisms to bone and lungs have recently been The development of metastatic capacity is due to the acquisition associated with discrete variations in overall expression of specific characteristics that accompany tumor cells during patterns. Thus, specialized gene sets have been defined that malignant progression in the local environment and at mediate metastasis to these target organs.4,7,9 Moreover, many distant sites.1 As migratory tumor cells are confronted by key oncogenic signaling pathways converge to adapt tumor cell different microenvironments, which they must survive and metabolism to support their growth and survival.10 eventually counteract, ‘cross-talk’ between metastatic cells By proteomic comparison of metastatic breast cancer cell lines, and these microenvironments is needed for metastasis devel- we found several antioxidant proteins, such as peroxiredoxins, opment.2 Indeed, the interactions between early disseminated that were specifically upregulated in lung metastatic cells cancer cells and the surrounding tissues lead to the selection or compared with other organ-specific metastatic variants.11 Indeed, adaptation of the cells in an early metastatic niche previous to cells living under aerobic conditions, as might be the case in lung metastasis outgrowth.3 Additional genetic or epigenetic metastasis, protect themselves from the damage caused by events and/or release from dormancy are critical for the reactive oxygen species (ROS), which arises from the incomplete productive metastatic growth of early disseminated cancer reduction of oxygen or exposure to external agents.12 --16 cells.4 Moreover, primary breast tumors might contain ‘cancer During a recent protein--protein interaction network analysis of stem cells’ that can metastasize and drive the formation and breast cancer cells that metastasize in lungs, we found that growth of new tumors.5 peroxiredoxin-2 (PRDX2) interacted with 105 different proteins.17 After bone, lung is the second main target of breast cancer Thus, we hypothesized that PRDX2 is a key phenotype with an metastasis. Transcriptomic analysis of a variety of cell lines has important role in antagonizing oxidative stress in breast cancer identified the that mediate metastasis to bone or lungs.6--8 cells that metastasize in lungs.

1Biological Clues of the Invasive and Metastatic Phenotype Group, Bellvitge Biomedical Research Institute (IDIBELL), L’Hospitalet de Llobregat, Barcelona, Spain; 2INSERM, UMR 664, IFR62, Laennec School of Medicine, Lyon, France; 3Cardiovascular Research Center (CSIC-ICCC), CIBER de Bioingenierı´a, Biomateriales y Nanomedicina (CIBER-BBN), Hospital de la Santa Creu i Sant Pau, Barcelona, Spain; 4Institut de Recerca Biome`dica de Lleida (IRBLLEIDA), Lleida, Spain; 5Servei d’Epidemiologia, IDIBELL-Institut Catala d’Oncologia, L’Hospitalet de Llobregat, Barcelona, Spain; 6Edouard Herriot Hospital, Lyon, France; 7Unit of Biomarkers and Susceptibility, and CIBERESP, IDIBELL-Institut Catala d’Oncologia, L’Hospitalet de Llobregat, Barcelona, Spain and 8ICFO-Institut de Cie`ncies Foto` niques, Parc Mediterrani de Tecnologia, Castelldefels, E-Barcelona, Spain. 9These authors contributed equally to the paper. Correspondence: Dr A Sierra, Biological Clues of the Invasive and Metastatic Phenotype Group, Bellvitge Biomedical Research Institute (IDIBELL), Avda. Gran Via de l’Hospitalet de Llobregat, 199, L’Hospitalet de Llobregat, Barcelona E-08908, Spain. E-mail: [email protected] Received 1 September 2011; revised 8 February 2012; accepted 12 February 2012; published online 19 March 2012 PRDX2 promotes lung metastasis V Stresing et al 725 Here, we show that PRDX2 expression in breast cancer cells Downregulation of PRDX2 sensitizes lung metastatic MDA-MB-435 buffers metastatic cells against oxidants, as a virulent phenotype breast cancer cells against ROS specifically selected by the lung microenvironment. This suggests To investigate whether peroxiredoxins provide lung metastatic a putative role of PRDX2 in stratifying patients at first diagnosis for cells with more effective protection against oxidative stress, we a more specific therapy that might be based on ROS-inducing transiently depleted 435-L3 cells of PRDX2 or PRDX3 using small drugs. interfering RNA oligos (Supplementary Figure S2A). Knockdown of PRDX2 or PRDX3 sensitized cells against H2O2 and induced more stress than in nontransfected 435-L cells (Supplementary Figure S2B). RESULTS 3 Moreover, intracellular ROS levels increased in 435-L3 cells that Expression patterns and subcellular distribution of peroxiredoxins were transiently depleted of PRDX2 and PRDX3 when cells were in lung metastatic MDA-MB-435 breast cancer cells exposed to H2O2 (Supplementary Figure S2C). Interestingly, cells We first examined human peroxiredoxin expression (PRDXs 1--6) depleted of PRDX2 had significantly higher basal levels of ROS in parental 435-P cells and their metastatic variants to validate than untransfected 435-L3 cells, 435-L3 cells transfected with a their differential expression, according to previously reported 11 control oligo, and siPRDX3-transfected cells (Supplementary data. We found that PRDX2 and PRDX3 levels were higher in Figure S2C). This indicates that the ROS scavenger function of lung metastatic variants (435-L2, 435-L3 and 435-L2/5) than in 435-P PRDX2 in lung metastatic cells was greater than that of PRDX3. cells and lymph node (435-N) or bone (435-B) metastatic variants We therefore decided to stably downregulate PRDX2 expression (Figure 1a, left panel). The more aggressive 435-L3 cell line showed in 435-L3-eGFP-CMV/Luc cells to further investigate the con- the highest overall level of peroxiredoxin expression (12 times sequences of elevated PRDX2 in lung metastatic breast cancer more) compared with parental cells (Supplementary Table S1). cells in vivo. PRDX2 expression was verified by western blot Furthermore, we used confocal laser scanning microscopy to analysis and quantitative PCR (Figure 2a and Supplementary investigate the compartmental subcellular distribution of PRDX2 Figure S3A). Plots of the amount of light produced by and PRDX3. PRDX3 was localized in the mitochondria, whereas predetermined numbers of 435-L3/scbl cells vs the number of PRDX2 was cytosolic (Figure 1a, right panel). Moreover, PRDX2 cells were linear throughout the range of cells tested (R2 ¼ 0.9928). translocated from the to the nucleus in response to an This shows that light measurements can be used to estimate cell oxidant challenge with its substrate hydrogen peroxide (3 mM), numbers (Supplementary Figure S3B). We chose a stably down- while PRDX3 remained localized in the mitochondrial membrane regulated clone (shPRDX2, clone #7) with low protein expression, (Figure 1a, right panel). PRDX2 protein translocation under 3 as well as a pool of control clones transfected with the scrambled oxidative stress was similar in 435-P and 435-L metastatic cells. version of PRDX2 (scbl) for further characterization (Figure 2a). Hence, PRDX2 and PRDX3 may exert their antioxidant activity in shPRDX2 cells showed significantly higher sensitivity to H2O2- different cellular compartments without redundancy to control induced oxidative stress (Figure 2b) than 435-L3 and scbl control intracellular ROS. cells (Po0.0005 for both cell lines). Production of internal ROS correlated well with PRDX2 levels in cells under basal and Lung metastatic MDA-MB-435 breast cancer cells have increased oxidative stress conditions (Figure 2c). 435-P cells had the highest oxidative stress resistance ROS in basal conditions, which increased in the presence of H2O2. We analyzed the oxidative stress response in 435-P and lung Moreover, basal ROS that was not modified in shPRDX2 cells, metastatic variants 435-L2 and 435-L3 to investigate the relation- showed significantly increase in stress conditions. No significant ship between PRDX expression and the cell’s ability to resist differences in cell cycle characteristics were found under basal or oxidative stress. The cells showed dose-dependent resistance to oxidative stress conditions, as revealed by flow cytometry oxidative stress (Figure 1b). We found that parental cells were (Figure 2d). more susceptible to oxidative stress than their lung metastatic variants (Figure 1b). This was evident at a concentration of 2 mM H2O2 and even more marked at a concentration of 3 mM H2O2,at PRDX2 has a cause--effect role in lung metastasis which the viability of 435-L3 cells was twice as high as parental To examine whether downregulation of PRDX2 in lung metastatic cells (60% vs 30%, Po0.01). At this concentration, the viability cells affects their metastatic capacity and their propensity to grow of the less aggressive 435-L2 cells was also still higher than that in the lungs, we induced intra-mammary fat pad orthotopic of the parental cells (40% vs 30%, Po0.01). This correlated well tumors by inoculating nude mice (n ¼ 7 per group) with a PRDX2 with the overall levels of peroxiredoxins detected in these cells. knockdown clone (shPRDX2, clone #7) or with control cells (scbl). We further determined the internal generation of ROS in 435 cells, Tumorigenesis and metastatic progression in animals was 0 0 using 2 ,7 -dichlorofluorescein diacetate (DCFH2-DA) as a ROS monitored periodically (Figures 3a and b). Statistical analysis scavenger (Figure 1c). 435-L3 cells produced lower amounts of using a linear model algorithm to compare PRDX2 knockdown ROS than 435-P cells in physiological conditions, and this effect cells and the scbl control cells showed that, 40 days after cell was maintained under oxidative stress (Po0.05 for both condi- inoculation, the tumors in mice inoculated with shPRDX2 cells tions). The ratios of ROS production in basal vs oxidative stress tended to be smaller (82.8±47.0 mm3) than the scbl control 3 conditions were similar in both cell lines (435-P, 0.46; 435-L3, 0.37), tumors (187.9±97.5 m ). However, these differences in tumor which indicates a similar increase in cellular stress levels. volume were not significant (P ¼ 0.165). Moreover, the metastatic Furthermore, we examined the oxidative state of PRDXs in basal progression to lungs after tumor induction, as quantified from the conditions and under oxidative stress by measuring PRDX-SO3, normalized photon flux of cancer cells, differed between groups which is the irreversibly overoxidized form of PRDX. Figure 1d (Supplementary Figure S4A). By the end of the protocol at day 82, shows that under severe oxidative stress, inactive PRDX-SO3 the normalized photon flux differences were significant between accumulated more in parental than in lung metastatic cells. This scbl and shPRDX2 groups (Po0.005). Indeed, although all of the suggests that there was a higher degree of depletion of active mice developed primary tumors in both groups, the metastasis PRDX in 435-P cells via inactivation of the function and incidence in mice bearing shPRDX2 tumors was substantially subsequent PRDX degradation.18 different from that observed in control animals (incidence: 14% vs The used of a histone deacetylase inhibitor valproic acid, which 86%, respectively) (Figure 3b). To avoid bias in metastatic causes an accumulation of ROS in transformed cells,19 confirmed evolution because of differences in primary tumor size, we that the ROS-generating agents counteracted PRDXs and pre- evaluated the global incidence of lung metastasis in each group vented breast cancer lung metastasis (Supplementary Figure S1). relative to the primary tumor size at the time of tumor exeresis by

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735 PRDX2 promotes lung metastasis V Stresing et al 726 calculating the normalized photon flux ratios of metastasis/tumor between cell proliferation and cell death in shPRDX2 net tumor for each animal (Supplementary Figure S4B). The differences in growth as a consequence of PRDX2 downregulation (Figure 3c), metastatic capacity were significant at the end of the experiment which favored cell death. (Po0.005). Moreover, we examined the expression of endothelial marker We further examined the ex vivo expression of PRDX2 in tumor CD31 as an indicator of neovasculature in breast tumors samples from shPRDX2 tumor-bearing mice and control animals (Figure 3d). Although similar numbers of vessels were produced by immunohistochemistry analysis. As expected, stable knock- in all groups, a parametric Wilcoxon test revealed that tumors down of PRDX2 persisted in breast tumors and in metastatic from shPRDX2 mice had a significantly lower value for vascular tissues (Figure 3c). Furthermore, there were no differences among lumen than control tumors (Po0.05). With regard to blood vessel tumors in proliferating cell nuclear antigen staining as a marker of structure in lung metastases, no vasculature was formed in the proliferation; expression was high in all breast tumors at the time single case of shPRDX2 lung metastases, unlike metastases from of exeresis (Figure 3c). In contrast, increased cell death as control mice (Table 1). indicated by caspase-3 activation was evident in shPRDX2 tumors We analyzed PRDXs profiling across a series of relative to controls. Hence, there may be a negative imbalance breast tumors, GSE2603.6 Although we did not find significant

basal H2O2 (3mM) PRDX2

PRDX2 22 kDa PRDX3 28 kDa

PRDX4 30 kDa

PRDX5 22 kDa

PRDX6 25 kDa

Actin 42 kDa Mitocondria PRDX1 22 kDa Actin 42 kDa Merge

435-P 435-L2 435-L3 120 * * * 100 * 125000

80 100000 basal * H2O2 (3 mM) 60 75000 * g/µL protein]  % viability 40 50000 -DA/[ 2 20 25000

0 DCFH 0.4 123 0 435-P 435-L3 mM H2O2

H2O2 (3 mM) - +-+ H2O2 (3 mM) - +-+ PRDX2 PRDX-SO3 Tubulin Actin Rel. expr. 1 13.2 1.6 6.1 Rel. expr. 1.10.4 1.9 1

435-P 435-L3 435-P 435-L3

Oncogene (2013) 724 --735 & 2013 Macmillan Publishers Limited PRDX2 promotes lung metastasis V Stresing et al 727 shPRDX2 clones association between PRDX2 gene expression and lung metastasis #1 #6 #7 #9 free survival, the total of PRDXs was significantly associated with lung metastatic progression (P ¼ 0.0055) but not with the rest of PRDX2 metastases (P ¼ 0.077), suggesting that PRDX2 function might be regulated at the protein level (Supplementary Table S2). Moreover, Tubulin PRXD2 expression in a breast cancer tissue array from 104 patients showed that high expression of PRDX2 in breast carcinomas trend to *** lung metastasis progression, Po0.053 (Supplementary Figure S5). 100 *** H2O2 (3 mM) 75

50 % viability 25 Figure 2. Stable downregulation of PRDX2 expression in MDA-MB- 435-L3 lung metastatic cells. (a) Stable downregulation of PRDX2 in 0 435-L cells infected with GFP/LUC. PRDX2 expression levels as de- 435-P435-L scbl shPRDX2 3 3 termined by immunoblot analysis are shown for several shPRDX2 clones, 435-L3 cells, mock-transfected (scbl) control cells and a pool 30000 * of shPRDX2 clones (pool A). Further analyses were done with lung * basal metastatic cells stably downregulated for PRDX2 (shPRDX2, clone #7) and scramble control cells (scbl) (b) Tumor cells were treated H2O2 (3 mM) 20000 with 3 mM H2O2 for 24 h. The cell viability of tumor cells under

l protein] oxidative stress conditions was calculated as a percentage of treated  cells with regard to untreated cells. Downregulation of PRDX2 in g/  10000 lung metastatic cells (shPRDX2, clone #7) made them more sus- ceptible to oxidative stress than the scbl control cells and parental

DHE/[ 435-L3 cells. ***Po0.0005, using the Student’s t-test. (c) Internal ROS production in stably downregulated shPRDX2 cells in basal condi- 0 435-P435-L scblsh PRDX2 tions or treated with 3 mM H2O2 for 24 h. Under oxidative stress, 3 shPRDX2 (clone #7) showed significantly higher ROS production than the scbl control and 435-L3 cells. *Po0.05 using the Student’s t- 75 test. (d) Cell cycle response to oxidative stress. We cocultured the 435-P following cells with or without H2O2 (3 mM) for 24 h: 435-P, 435-L3, stably downregulated shPRDX2 and scbl control cells. Ethanol-fixed 435-L3 50 scbl cells were incubated with 5 ml RNase (10 mg/ml) and 50 ml propidium shPRDX2 iodide (PI) in phosphate-buffered saline (0.5 mg/ml, Sigma) for 30 min at 37 1C, and PI-stained cells were analyzed for DNA content by flow cytometry (FACSCalibur, BD Biosciences, Franklin Lakes, NJ, 25 USA). At least 20 000 events were collected in each sample and evaluated using the ModFit LT (Verity Software, BD Biosciences)

% cells in each phase program. The percentages of cells present in each cell phase are shown. No significant differences in cell cycle characteristics were 0 found between shPRDX2 and mock-transfected scbl or un- - +-+-+-+ -+-+-+-+ -+-+-+-+ 3 mM H2O2 transfected 435-L3 cells, respectively, under basal or oxidative stress G0 –G1 S G2 –M conditions. The mean±s.d. of three experiments are shown.

Figure 1. Characterization of the MDA-MB-435 lung metastatic phenotype. (a) Left-hand panel: western blots of peroxiredoxin (PRDX) isoforms 1--6 in 435 parental cells and metastatic variants. Total protein extracts (50 ml per lane) were fractioned by sodium dodecyl sulfate--poly- acrylamide gel electrophoresis and blotted onto polyvinylidine fluoride membranes. b-Actin or a-tubulin expression was measured as the loading control. The expression of PRDXs in lung metastatic variants (435-L2 and 435-L3) was assessed with regard to parental 435-P cells. Lymph nodes (435-N) and bone (435-B) metastatic variants are included for comparison. Right-hand panel: subcellular localization of PRDX2 and PRDX3. The 435-P and 435-L3 cells on coverslips were washed, fixed and stained individually with anti-PRDX2 and anti-PRDX3 antibodies, and mitochondria were stained with MitoTracker. Confocal microscopy images of 435-L3 cells show that PRDX2 is localized in the cytosol, while PRDX3 is mitochondrial in basal conditions (images for 435-P cells not shown). Under oxidative stress, PRDX2 translocates to the nucleus. The subcellular distribution was similar in both cell types. (b) Viability of 435-P cells and lung metastatic variants stimulated with 0.4-- 3mM H2O2 for 48 h at 5 Â 103 per well after starving them for 24 h. 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide at 5 mg/ml was added before measuring the optical density at 540 nm. Results are expressed as the percentage of cell viability of untreated control cells (absorbance of stimulated cells/absorbance of control cells) and are the mean of three independent experiments. 435-L2 and 435-L3 cells show higher cell viability than parental cells under oxidative stress conditions. *Po0.01 for 435-P vs 435-L3 and 435-L2, respectively, using one- way analysis of variance followed by Dunnett’s multiple comparison test. (c) Internal ROS production. Cells were incubated with 7 mM DCFH2- DA in Hanks’s balanced salt solution (Invitrogen) for 40 min or with 2 mM dihydroethidium (DHE) in Hanks’s balanced salt solution for 30 min at 37 1C, according to the manufacturer’s instructions, with or without H2O2 (0.4--4 mM). Fluorescence was measured in a fluorescence microplate reader (FLUOstar Optima, Biogen, Spain) at 485--520 or 530--562 nm (DCF and ethidium, respectively) to assess the fluorescence intensity relative to protein concentration. The results are the mean of three independent experiments. 435-L3 cells produced less ROS than parental cells under basal conditions, and when cells were treated with 3 mM H2O2 *Po0.05 for 435-L3 vs 435-P in both conditions using the Student’s t-test. (d) Western blot images of PRDX-SO3 expression in 435-P and 435-L3 cells (left panel). Relative expression of PRDX-SO3 was assessed by band densitometry. The overoxidized, inactive form of PRDX, PRDX-SO3, accumulated less in lung metastatic cells than in parental cells under oxidative stress conditions. The comparative expression of PRXD2 is showed at the right panel.

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735 PRDX2 promotes lung metastasis V Stresing et al 728 300 Tumor Lung Metastasis

) scbl 3 Day 40 Day 40 Day 82 250 shPRDX2 200 150 P = 0.165 scbl

100 187.9 ± 97.5 6/7 6/7 50 Tumor volume (mm 0 0 7 14 21 28 35 42 shPRDX2 days after i.m.f.p. injection 82.8 ± 47.0 1/7 1/7

Breast Tumor Lung Metastasis scbl shPRDX2 scbl shPRDX2 H&E PRDX2

CD31 scbl shPRDX2 PCNA Breast Tumor Casp-3 Lung Metastasis Figure 3. PRDX2 is a cause--effect player in lung metastasis. (a) Tumor growth curves from athymic Nude/Balb-c female mice (n ¼ 7 animals per group) after intra-mammary fat pad inoculation (1 Â 106 cells) of scbl control or PRDX2 knockdown (shPRDX2, clone #7) cells. Tumor volumes were calculated at the indicated times after cell injection using the formula (volume (mm3) ¼ L2 Â W2/2), where L and W are the major and minor diameters in millimeters, respectively. Tumor growth did not differ statistically significantly between both groups (P ¼ 0.165 using the Mann--Whitney test). (b) Representative images of primary breast tumors and spontaneous lung metastases on day 40 and day 82 after intra-mammary fat pad implantation. The standard rainbow color scale was used to depict relative light intensities (red ¼ highest; blue ¼ lowest). Tumor volume and lung metastasis incidence are shown in the insets. (c) Immunohistochemistry analysis of paraffin-embedded histological sections of tumors and lung metastases stained for PRDX2, proliferating cell nuclear antigen, Casp-3 ( Â 20 magnification) and hematoxylin and eosin stained ( Â 10) in light microscopy. Representative images of tumors (at day 40) and respective metastasis (at day 82) from intra-mammary fat pad implantation are shown. (d) The in vivo angiogenic effect of PRDX2 downregulation in 435 lung tumors. Representative images of CD31 stained frozen tissues sections of tumors and lungs from mice injected with shPRDX2 (clone #7) and scbl control cells ( Â 10) are shown (upper panel).

Downregulation of PRDX2 specifically inhibits the growth of shPRDX2 cells (Figure 4a). Whole-body in vivo imaging of animals metastatic cells in lungs confirmed that the lung metastasis burden was substantially lower We then intravenously injected shPRDX2 and scbl control cells and in mice inoculated with shPRDX2 cells than in control mice, monitored the homing of cells to the lungs over time. There was a whereas the burden of scbl and shPRDX2 cells in lymph nodes was time-dependent increase in the shPRDX2 and scbl cell burden in similar (Figure 4b). Furthermore, hematoxylin and eosin staining lungs. The number of scbl cells retained there was statistically and optical microscopy showed higher lung colonization in significantly higher until the end of the experiment at day 55 control mice (Figure 4c) than in shPRDX2 mice. An analysis of (P ¼ 0.056) than that observed in lungs of mice injected with green fluorescent protein (GFP)-positive cells in ex vivo lung

Oncogene (2013) 724 --735 & 2013 Macmillan Publishers Limited PRDX2 promotes lung metastasis V Stresing et al 729 Table 1. CD31 staining of breast tumor and lung metastatic tissue Next, we performed experiments to visualize in vivo the samplesa metabolic features of metastatic bone lesions using radiopharma- ceuticals that target tumors cells with a high glycolytic metabo- 18 18 scbl shPRDX2 lism ([ F]FDG) or bind to bone ([ F]NaF). The positron emission tomography (PET) scan of animals did not show any accumulation Breast tumor of FDG in metastatic hind limbs (Figure 5d, middle panel), despite Mean vessels per field 83.13 68.91 the presence of BLI-positive tumor cells (Figure 5d, left panel). In % Branching 41.9±20.2 31.5±16.5 contrast, there was a far greater accumulation of NAF in hind limbs % Lumen 31.0±6.0 12.9±7.1* (Figure 5d, right-hand panel), which confirms the presence of osteolytic lesions. Lung metastases Mean vessels per field 38.75 0* % Branching 43.4±10.9 0* PRDX2 regulates the metabolic stress response of lung metastatic % Lumen 27.1±7.2 0* cells a Forming vessels were counted from four fields of vision ( Â 40) per tumor We used Raman microspectroscopy to study the chemical and are expressed as mean CD31 values (mm2). Differences in vessel composition of metastatic variants to assess the importance of branching and amount of vascular lumen were statistically significant in the metabolic pressure of the target organ to select the metabolic tumors and lung metastases from animals injected with shPRDX2 cells. phenotype of metastatic cells (Supplementary Figure S7). *Po0.05 for shPRDX2n vs scbl, using a parametric Wilcoxon test. The principal component analysis models using the spectral region 600--1700 cmÀ1 differentiated the data set of lung metastatic 435-L3 cells from the bone metastatic 435-B cells, (Supplementary Figures S7A and B). Loadings on PC3 containing the bands: 1003 cmÀ1, samples of mice injected with shPRDX2 and scbl cells confirmed trigonal ring breathing of the benzene ring (present in phenylalanine residue); 1443 cmÀ1, methylene scissoring deformation and the in vivo results. We examined the lymph node PRDX2 À1 expression by immunohistochemistry and found that it was low 1655 cm ,HC¼ CH stretch (cis conformation), compatible with fatty acids and protein; and loadings on PC4, bands in region 729 and similar in all samples (Figure 4c). À1 The evolution of lung metastasis in controls limited the course and 1617 cm , compatible with NADH, were involved (Supplemen- of the experiments. Thus, by the time lung and lymph tary Figure S7B). Under glucose deprivation, 435-L3 and 435-B node metastasis had developed no signal compatible with bone metastatic cells were almost completely separated by the PC4 metastasis was detected. component, which increased in bone metastatic cells. Moreover, If during metastasis formation cells that reach lungs are selected chemical hypoxia shifted each variant in opposite PC3 way: while by their ability to remove ROS and PRXD2 is an exponent of this 435-B increased PC3 value, the 435-L3 cells decreased it (Supple- organe-specific phenotype, we wondered why PRXD2 also mentary Figure S7C). These results indicated that metastatic variants decreased in metastasis from hypoxic tissue like bone (Figure 1a had different metabolic phenotype and can be separated by their and Supplementary Table S1). We stably overexpressed PRDX2 in different metabolic features. the highly bone metastatic BO2-GFP/tTA cell line (selected from To further explore the role of PRDX2 in the selection of cells six in vivo/in vitro passes), which expressed low levels of the according the bioavailability of glucose in the target organ, we protein (Figure 5a, left panel). Two clones (#10 and #88) challenged cells from both breast cancer models, 435-P (lung overexpressing PRDX2 were obtained (Supplementary Figure metastatic tropism) and 231 cells (bone metastatic tropism) and S6A). Clone #88 was selected for further studies as the luciferase their metastatic variants through starving them from glucose expression levels were similar to those expressed by the BO2/GFP (Figure 6a). 435-L3 cells survived significantly more than 435-P control cell line, thereby allowing direct comparison by biolumi- cells (Po0.0005). As 435-P and 435-L3-shPRDX2 had only 50% of nescence imaging in bone metastasis experiments in vivo viability, the data suggested that PRDX2 provided 435-L3 cells with (Supplementary Figure S6B). the ability to the preferential use of glucose. In addition, PRDX2- Overexpression of PRDX2 in BO2 cells (Figure 5a, left panel) was induced BO2 cells more susceptible against glucose deprivation inversely correlated with intracellular ROS production (Figure 5a, with regard to parental cells (P ¼ 0.016), suggesting that PRDX2 right panel); BO2/PRDX2 cells produced less ROS than BO2/GFP disturbed the metabolic switch of the bone glucose-independent control cells in basal conditions (Po0.05) and under oxidative cells. Moreover, 435-B cells (P ¼ 0.001) and BO2 cells (P ¼ 0.003) stress stimuli (Po0.005). were less affected by hypoglycemic conditions, with regard their We next examined the effect of overexpression of PRDX2 on respective control cells. The putative glucose dependence breast cancer bone metastasis formation. BO2/PRDX2 cells (clone metabolism of lung metastatic cells was checked by measuring #88) and BO2/GFP control cells were inoculated intracardiacally the lactate production of 435-L3 and 435-B cells with regard to into nude mice (n ¼ 6 per group). A radiographic analysis of hind protein concentration: 435-L3 had 20.88 mM and 435-B had legs at day 36 and day 40 after tumor cell inoculation showed that 14.57 mM lactate/mg protein. These results suggested the glyco- mice bearing BO2/PRDX2 tumors had significantly smaller lytic phenotype of lung metastatic cells and indicate that the osteolytic lesions than mice injected with BO2/GFP cells overexpression of PRDX2 in lung metastatic cells might subtend (Figure 5b, left panel). The mean area of osteolytic lesions at day the preferential use of glucose (Figure 6b). 40 was 5.1±2.9 mm2 and 15.8±1.6 mm2 for BO2/PRDX2 and BO2/ GFP tumor-bearing animals, respectively (Po0.005). Biolumines- cence imaging of mice revealed that the BO2/PRDX2 tumor DISCUSSION burden in animals was statistically significantly less than that The lung receives many cancer cells transported in the blood observed in control mice injected with BO2/GFP cells (Po0.05). stream, however, the majority of pro-metastatic cells are This indicates that PRDX2 overexpression impaired BO2 cell destroyed in lungs, which act as a filter that involves the activation growth in bones (Figure 5c). of alveolar macrophages.20 Thus, the levels of partial oxygen The decreased in vivo skeletal growth of BO2/PRDX2 cells, in pressure that metastatic cells can suffer in lungs are different from contrast to the effect of PRDX2 in lung metastatic cells, suggested those experienced in other systemic tissues.21,22 We have that PRDX2, which protected lung metastatic cells against demonstrated that PRDX2 overexpression in lung metastatic cells oxidative stress, did not improve bone metastatic cell colonization. effectively removes the intracellular ROS, suggesting that this

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735 PRDX2 promotes lung metastasis V Stresing et al 730 1.0×1009 P = 0.056 Day 55 08 1.0×10 scbl (n = 7) + 1.0×1007

NPF 1000000 shPRDX2 100000 (n = 6)

10000 - 0 10 20 30 40 50 60 days after injection

Day 48 1.0×1009 1.0×1008 + 1.0×1007 scbl 1000000 RLU shPRDX2 100000 10000 1000 - Lung Lymph node

Lung Lymph node

H&E GFP (10x) GFP (40x) H&E PRDX2 scbl 2 shPRDX

Figure 4. Lung metastasis development after intravenous injection of 435 lung metastatic cells. (a) Metastasis was induced by injecting 106 cells in 150 ml Hanks’s balanced salt solution into the tail vein. Pulmonary tumor burden in mice bearing shPRDX2, clone #7, and scbl tumors, as judged by fluorescence measurement. Left-hand panel: monitoring of tumor growth development in animals from day 0 to day 55 after tumor cell inoculation. Results are expressed as normalized photon flux (NPF). Metastatic progression varied between groups (P ¼ 0.056 using a mixed linear model) and the number of shPRDX2 cells retained in the lungs was higher than in scbl controls. Right-hand panel: representative fluorescent images of mice on day 55 after tumor cell inoculation. (b) Left-hand panel: comparison of light production (NPF) between lungs and lymph nodes on day 48 after tumor cell inoculation. There was a trend toward less light production in the lungs of animals injected with shPRDX2 cells than in those injected with scbl control cells (P ¼ 0.09, Student’s t-test). The light production in lymph nodes was similar between groups (P ¼ 0.59). Right-hand panel: representative images of mice on day 48 after tumor cell inoculation. (c) Representative images of cryostat sections (30 mm) from hematoxylin and eosin (H&E) stained ( Â 10 magnification) and GFP-labeled ( Â 10) lung metastases, as well as lymph node tissues stained with H&E ( Â 10) and labeled with an antibody for PRDX2 expression ( Â 20), respectively.

phenotype is helpful in the lung microenvironment to metastatic lung metastatic cells had the highest PRXD2 level and the lowest cell colonization. The scavenger ability of cells overexpressing hyperoxidation. Preventive therapies with ROS-inductor drugs PRXD2 might counteract the ROS delivered by the preferential use diminished lung metastasis. Therefore, several ROS-generating of glucose, primed in the lung microenvironment, allowing agents like histone deacetylase inhibitors19 and the glycopeptide metastases grow by shortening the dormancy period. antibiotic Bleomycin,26 might perform preventive therapeutic PRDX2 is a highly efficient redox protein that neutralizes strategies overtaking PRXD2 function. Conoiding A is a novel hydrogen peroxide, resulting in protection of cells from oxidative inhibitor of this important class of antioxidant and redox signaling damage and in regulation of peroxide-mediated signal transduc- enzymes27 that binds covalently to the peroxidatic cysteine of tion events.23 Its unique redox properties may account for TgPrxII blocking its enzymatic activity. On the other hand, drugs nonredundant role in defense against oxidative stress. Indeed, such as 6-amino-nicotinamide, which inhibits G6P dehydrogen- metastatic cells that overexpress PRDX2 might mimic normal ase28 might be effective against bone metastatic cells that had erythrocytes and lung cells that are well-buffered against increased the NADH levels. Additional basic and in vivo experi- exogenous oxidants and oxidative stress metabolism.24 ments are needed to answer these questions. PRDX2’s pathogenic function is driven by a mechanism of The mechanism of the high use of glucose under aerobic protein regulation against its irreversible hyperoxidation.25 Indeed, conditions is still debated. The highly glycolytic phenotype

Oncogene (2013) 724 --735 & 2013 Macmillan Publishers Limited PRDX2 promotes lung metastasis V Stresing et al 731 of cancer cells was reported to be induced by activation of features that metastatic cells can acquire in the target organ. the oncogene Akt.29 The clear differences in the metabolic Indeed, the mechanism to cope with the cumulative ROS is different phenotype of lung and bone metastatic cells raise questions in lung than in bone tissues, where the lowest oxygen gradient in about preventive therapies according the mimetic metabolic bone marrow induces fewer metabolic free radical challenges.30

** 25000 *

20000 B02/PRDX2 basal #10 #88 B02/GFP

l prot] 15000 H2O2  g/

PRDX2  10000 Tubulin DHE/ [ dox --+ + - 5000

0 B02/GFP B02/PRDX2

20 B02/GFP ** ) 2 15 B02/PRDX2

10

5 Osteolytic area (mm

0 B02/GFP B02/PRDX2 Day 36 Day 40

B02/GFP B02/PRDX2 1000000 B02/GFP B02/PRDX2 100000

10000 RLU

1000 P < 0.05

100 20 25 30 35 Day 28 days after injection

Figure 5. For caption please see page 732.

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735 PRDX2 promotes lung metastasis V Stresing et al 732 Moreover, PRDX2 is produced by bone marrow stromal cells and which is a critical early step in the signaling pathway.39 regulate the hydrogen peroxide level of this complex niche.31 A proposed model for ROS homeostatic control by PRDXs includes PRDX2 influences oxidative and metabolic stress through c-Myc-dependent regulation.40 multiple mechanisms.23 As PRDX2 knockdown dramatically Furthermore, cytosolic PRDX2 can migrate to the nucleus in decreases lung metastasis formation, we suggest that PRDX2 is stress conditions. In addition to their antioxidant activities, PRDXs a ‘metabolic adaptor’, which functions stabilizing the redox state participate in various biological functions, such as cell prolifera- required for cell survival in an oxidative atmosphere. Indeed, and tion, differentiation, apoptosis, gene expression and intracellular effective anti-oxidant defense system is needed to protect signaling.12,13 PRDX2 positively regulates JNK-dependent DNA metastatic cells from free radicals and ROS.32 These results point repair.41 Loss of PRDX2 seems to accelerate cellular senescence out that, as in well-oxygenated (aerobic) tumor regions,33 PRXD2 in vitro in murine fibroblasts by overexpression of negative cell might be a ‘metabolic adaptor’ with a pathogenic role in lung cycle regulators.42 The higher cell death of shPRDX2 cells, as metastasis acting through its ROS scavenger function of cancer evidenced by caspase-3 activation in shPRDX2 tumor samples and cells, which reach the inhospitable lung microenvironment. controls, illustrates an imbalance in tumor turnover favoring the The low metastatic burden of BO2/PRDX2 cells in bones survival of lung metastatic cells overexpressing PRDX2. suggests that PRDX2 might interfere in the preferential bone In conclusion, we have shown that PRDX2 regulates breast metastatic cells metabolism with deleterious consequences to cancer cell colonization in lungs by acting on the oxidative and carcinoma cells that try to adapt to the bone microenvironment. It metabolic stress responses of metastatic cells. Moreover, PRDX2 has recently been reported that mitochondrial glucose oxidation trend to be expressed in primary tumors, which metastasize in may be incompatible with the survival of some cancer cell.34 lungs suggesting a putative lung metastasis marker that might be Moreover, the balance mechanism of the antioxidant system used to stratify patients at first diagnosis. We believe these might prevent collapse by redirecting the glycolytic flux into the findings will pave the way to the development of new therapeutic pentose phosphate pathway.35 Indeed, BO2/GFP cells displayed strategies to prevent lung metastasis and improve the response to low glucose dependence suggesting a metabolic shift from therapy. glucose oxidation to fatty acid oxidation to resist metabolic insults in a hypoxic and hypoglycemic tissue like bones. Although PRXD2 interfered with BO2 cell bone metastasis progression, we MATERIALS AND METHODS did not observe lung metastasis because the aggressiveness of the Cell lines cells limited the length of the experiment. Thus, we suggest that PRXD2 protects lung metastatic cells against the lung microenvir- MDA-MB-435 cells (435-P) supplied by Dr Fabra (IDIBELL) in 1992 and their onment but is not sufficient to induce lung metastasis. metastatic variants established from primary cultures of lung (435-L2,435-L3 The use of the controversial MBA-MB-435 cell model brings with and 435-L2/5), lymph node (435-N) and bone (435-B) metastases, maintained under standard conditions have been described elsewhere.43 it the limitations of such a poorly differentiated, aggressive breast tumor line, which expresses both epithelial and melanocytic Although controversial, it has recently been demonstrated that MDA-MB- markers.36 As 435-P cells grow efficiently via the intra mammary 435 cells are a useful breast cancer model and that they express both epithelial and melanocytic markers.36 In some experiments, we used breast fat path and after tumor removal clinical lung metastasis appears cancer bone metastatic cell lines MDA-MB-231 originally obtained from the in 470% of the mice, it is a useful standard model that mimics the clinical situation and counteracts the lack of good experimental European Type Culture Collection (ECACC 92020424) in 2007, maintained metastatic breast cancer models. Furthermore, the results are in accordance with ECACC guidelines for o4 months before use in these experiments. BO2 cell line has been established from bone metastases reinforced by those obtained from analyzing the expression of caused by 231-P after six in vivo passages in nude mice using a heart PRXD2 in primary breast tumors, which suggest the association injection model. The characteristics of luciferase-expressing MDA-BO2 cells between PRXD2 and lung metastasis. 44,45 The mechanism of action of PRDX2 that is associated with its have been described elsewhere. Cells were cultured under standard conditions. intrinsic role in ROS scavenging might also include the second messenger function of ROS.37 Molecular mechanisms for ROS- mediated control of nuclear factor-kB activity are now beginning Cell viability assay to be revealed.38 Moreover, peroxiredoxins participate in signal The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay transduction to regulate cytochrome release from mitochondria, was used as described elsewhere.46 The cells were serum starved for

Figure 5. Overexpression of PRDX2 in BO2 breast cancer metastatic cells impairs tumor growth in bone. (a) Left-hand panel: WB of PRDX2 expression in MDA-MB-231 breast cancer cells (231) and the highly bone metastatic BO2/GFP cell line, as well as in BO2-GFP/tTA cells overexpressing PRDX2 (BO2/PRDX2). For PRDX2 overexpression, we used the pBIL vector system for protein and luciferase coexpression. Protein expression levels of PRDX2 in BO2/PRDX2 cells (clones #10 and #88), parental BO2-GFP cells and 231 cells are shown. Right-hand panel: internal ROS production in BO2/GFP control and BO2/PRDX2 cells. Under both basal and oxidative stress conditions, BO2/PRDX2 cells produced significantly less ROS than parental BO2/GFP cells. *Po0.05 and **Po0.005 using the Student’s t-test. (b) Effects of overexpression of PRDX2 in BO2 cells on bone metastasis formation. BO2/PRDX2 cells and BO2/GFP control cells were inoculated intracardiacally into nude mice (n ¼ 6 animals per group). Left-hand panel: mice bearing BO2/PRDX2 tumors developed significantly smaller osteolytic lesions than mice injected with BO2/GFP cells on day 36 and day 40 after tumor cell inoculation (**Po0.005 using the Student’s t-test). Right-hand panel: representative radiographic images of hind legs on day 40 after tumor cell inoculation. Arrows indicate osteolytic lesions. (c) Tumor burden in mice bearing BO2/GFP control and BO2/PRDX2 tumors, as judged by bioluminescence measurement. Left-hand panel: monitoring of tumor growth development in animals from day 14 to day 35 after tumor cell inoculation. Results are expressed as relative light units (RLUs) in photons per second. *Po0.05 using a mixed linear model. Right-hand panel: representative bioluminescent images of mice on day 28 after tumor cell inoculation. (d) Left-hand panel: representative bioluminescent image (BLI) of mice bearing BO2/GFP tumors on day 28 after tumor cell inoculation. Middle- and right-hand panels: representative whole-body microPET images of BO2/GFP tumor-bearing mice on day 28 after 18 18 tumor cell inoculation, 1 h after administration of 400 mCi of 2-deoxy-2[ F]fluoro-D-glucose (FDG) or sodium [ F]fluoride ion (NAF) via the tail vein. FDG is used to target tumors cells that have a high glycolytic metabolism, whereas NAF is a radiopharmaceutical that has a high affinity to bone. The PET scan of animals did not show any accumulation of FDG in metastatic hind limbs, despite the presence of BLI-positive tumor cells. In contrast, there was far greater accumulation of NAF, which confirms the presence of osteolytic lesions.

Oncogene (2013) 724 --735 & 2013 Macmillan Publishers Limited PRDX2 promotes lung metastasis V Stresing et al 733 Glc - Pyr + (Sigma-Aldrich, St Louis, MO, USA). These probes react specifically with H2O2 ** to induce the highly fluorescent DCF or ethidium. ** 200 *** 435-P 435-B 160 * ** 435-L3 Cell transfections and constructs 120 shCTROL shPrdx2#7 Retroviral transduction was used to label 435-L3 lung metastatic cells. 80 231-P Vector preparation and packaging of viral particles was performed as B-Gal % Viability 48 40 88 described previously. A cell population that uniformly expressed the highest levels of enhanced GFP (435-L3-eGFP-CMV/Luc) was selected by 0 48 72 96 FACS (MoFlo, Cytomation, Dako, Denmark). Time Short hairpin RNA and corresponding scrambled sequences directed at PRDX2 target sites based on the human transcript (GI:33188450) were designed using the online small interfering RNA Target Designer software QUIESCENCE PRDX2 (Promega Corporation, Madison, WI, USA). Sequence analysis confirmed PRDX2 435-L3 cells the sequence integrity of the PRDX2 short hairpin RNA plasmids. Plasmids ROS were transfected into 435-L peGFP-CMV/Luc cells using Lipofectamine ANGIOGENESIS 3 2000 reagent. MICROENVIRONMENT To overexpress PRDX2 a complementary DNA fragment encoding ADAPTATION PRDX2 (transcript variant 1, GI:33188450) was isolated by reverse PRDX2 transcriptase--PCR. The PCR-amplified fragment was digested and cloned ROS pO2 into the bidirectional pBI-L vector (Clontech, Hampshire, UK). Stable BO2- ANGIOGENESIS GFP/tTA cells were then co-transfected with the pBiL/PRDX2 construct and a plasmid conferring puromycin resistance (Promega Corporation) so that AEROBIC CONDITIONS they would express PRDX2 and luciferase.45

Fatty acids In vivo mouse models, bioluminescence and PET imaging glutamine Six-week-old athymic nude Balb/c female mice were used with the Acetyl-CoA  oxidation approval of the animal care committee. For the tumorigenesis experiments, cells (106 cells in 50 ml serum-free Krebs cycle medium) were inoculated intra-mammary fat pad into anesthetized nude mice.49 Mice were controlled periodically during the experiment until ATP Respiratory chain symptoms of metastasis appeared, and metastasis development was 50,51 Glucose dependence monitored by non-invasive bioluminescence imaging. Osteolytic lesions were identified on radiographs as demarcated radiolucent lesions Figure 6. PRDX2 regulates the metabolic stress response of meta- in the bone and the extent of bone destruction per animal was expressed 44,50 static cells. (a) Viability of MDA-MB-435 and MDA-MB-231 metastatic in square millimeters, as described previously. variants was measured by a 3-(4,5-dimethylthiazol-2-yl)-2,5-di- In vivo optical imaging of nude mice engrafted with 435 cells was phenyltetrazolium bromide assay maintaining cells for 48, 72 or 96 h performed as described previously.52 The quantification and analysis of the in glucose-free culture medium, with no serum or pyruvate. Results photons recorded in the images was performed using Wasabi image are expressed as the percentage of cell viability of untreated control analysis software (Hamamatsu Photonics, Hamamatsu City, Shizuoka cells (absorbance of stimulated cells/absorbance of control cells) Prefecture, Japan). The number of photons was expressed as photon and are the mean of three independent experiments. *P 0.016, ¼ counts and data were expressed as the number of photon counts vs the **P ¼ 0.001 and P ¼ 0.003 and ***Po0.0005, using one-way analysis of variance and Dunnett’s multiple comparison test. (b) Schematic number of grafted cells. For the image quantification, we calculated the representation of metastatic progression in lungs. For metastatic average number of photon counts per pixel of a selected area of interest, cells to grow in lungs, the cancer cells must possess high levels of using Wasabi image analysis software (average photon counts per pixel). redox scavenging molecules that allow adaptation to a micro- PET were performed with the ClearPET commercially available system environment with functional oxidative phenotype. The over- with a spatial resolution at the center of 1.3 mm. Mice were injected with 18 18 18 expression of PRDX2 in lung metastatic cells has pathogenic role 400 mCi of 2-deoxy-2[ F]fluoro-D-glucose ([ F]FDG) or sodium [ F]fluoride acting through its ROS scavenger function. Its unique ability to ion (Na18F) via the tail vein. remove ROS subtends the use of glucose, exerting an effective anti- oxidant defense system needed to protect metastatic cells from free radicals and ROS. As PRDX2 knockdown dramatically decreases lung Protein expression in cells and tissues metastasis formation, PRXD2 might be a ‘metabolic adaptor’, which The specific antibodies used were: anti-PRDX1 to -6, anti-PRDX-SO3 stabilizes the redox state required for cell survival, then allowing (LabFrontier, Seoul, Korea), anti-AOP1, anti-a-tubulin, anti-b-actin (Sigma), metastasis to progress and shortening the metastasis dormancy anti-proliferating cell nuclear antigen (Santa Cruz, Santa Cruz, CA, USA), period. anti-CD-31/PECAM-1 (BD Pharmingen, San Jose, CA, USA) and anti-caspase-3 (Cell Signaling, Boston, MA, USA).

24 h and exposed for a further 24 h to 0--4 mM H2O2 (Sigma, St Louis, MO, Statistical analysis USA). For experiments involving cell nutrient deprivation, cells were The unpaired Student’s t-test and one-way analysis of variance were used seeded in complete medium for 24 h, and then cultured for a further 24, to analyze in vitro experiments, and the results were presented as 48, 72 or 96 h in Dulbecco’s modified Eagle medium with no serum, mean±s.e.m. Analyses were performed using Stat View v5.0 software glucose or pyruvate (DMEM/Glc--/Pyr--, Invitrogen, San Diego, CA, USA). (www.programas-gratis.net/b/statview-v5). For the in vivo experiments, parametric and non-parametric tests were applied to assess the observed differences in tumor volume at a specific time. For this purpose, we used Intracellular ROS measurement the Student’s t-test and the Wilcoxon test. Furthermore, to evaluate the ROS generation in cells was assessed as described elsewhere47 using two effect of the differences over time we fitted linear mixed models,53 which probes: DCFH2-DA (Molecular Probes, Alcobendas, Spain) and dihydroethidium took into account the random effects because of within-mice variation,

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735 PRDX2 promotes lung metastasis V Stresing et al 734 given the repeated measures data structure in the experimental design. 18 Chevallet M, Wagner E, Luche S, van Dorsselaer A, Leize-Wagner E, Rabilloud T. Estimated parameters and their variability were used to solve specific Regeneration of peroxiredoxins during recovery after oxidative stress. J Biol Chem contrasts based on likelihood estimation. Thus, we could assess whether 2003; 278: 37146 --37153. the difference was consistent throughout the period, and adjust the effect 19 Ungerstedt JS, Sowa Y, Xu WS, Shao Y, Dokmanovic M, Perez G et al. Role of of time itself in the model. All analyses were performed using the statistical in the response of normal and transformed cells to histone package R.54 Data and code from these analyses are available on request to deacetylase inhibitors. Proc Natl Acad Sci USA 2005; 102: 673 --678. 20 Iles KE, Forman HJ. Macrophage signaling and respiratory burst. Immunol Res the authors. Statistical significance was set at Po0.05. 2002; 26: 95 --105. 21 Ganong WF. Gas transport between the lungs & the tissues. In: Weitz M, Brown RY (eds). Ganong’s Review of Medical Physiology, Chapter 37, 22 edn. LANGE basic CONFLICT OF INTEREST science, McGrow-Hill Companies, USA, 2005. pp 631 --697. The authors declare no conflict of interest. 22 Archer S, Gomberg-Maitland M, Maitland M, Rich S, Garcia J, Weir K. Mitochondrial metabolism, redox signaling, and fusion: a mitochondria-ROS-HIF-1_-Kv1.5 O2-sensing pathway at the intersection of pulmonary hypertension and cancer. ACKNOWLEDGEMENTS Am J Physiol Heart Circ Physiol 2008; 294: H570 --H578. 23 Ola´hova´ M, Taylor SR, Khazaipoul S, Wang J, Morgan BA, Matsumoto K et al. We would like to thank Oriol Casanovas (Angiogenesis Laboratory, IDIBELL) for A redox-sensitive peroxiredoxin that is important for longevity has tissue- and scientific discussions and his expert advice on angiogenesis. We would also like to stress-specific roles in stress resistance. Proc Natl Acad Sci USA 2008; 105: thank Berta Martı´n, Vanessa Herna´ndez and Dı´dac Dominguez for his expert technical 19839 --19844. assistance. We are grateful to Mr R Rycroft for expert language advice and Victor 24 Low FM, Hampton MB, Peskin AV. functions as a noncatalytic Moreno for expert statistical advice (Barcelona University). We thanks Jose Carlos scavenger of low-level hydrogen peroxide in the erythrocyte. Blood 2007; 109: Perales and Andy Me´ndez for the advice and support the metabolic analysis 2611 --2617. (Barcelona University). We acknowledge all the partners of the MetaBre consortium 25 Seo JH, Lim JC, Lee DY. Novel protective mechanism against irreversible for their collaboration and stimulating criticism. This study was supported by grants hyperoxidation of peroxiredoxin: Nalpha-terminal acetylation of human peroxi- from the Spanish Ministry of Health and Consumer Affairs (FIS/PI071245 and FIS/PI10/ redoxin II. J Biol Chem 2009; 284: 13455 --13465. 00057), EC MetaBre (contract no. LSHC-CT-2004-506049), INCA (ONCOIMAGE, n107/ 26 Tao M, Wang L, Wendt-Pienkowski E, George NP, Galm U, Zhang G et al. The 3D1316/PF-108-01/NG-LC), the Ministry of Education and Science (SAF2004-0188-E), tallysomycin biosynthetic gene cluster from Streptoalloteichus hindustanus E465- AECC Scientific Foundation and Private Foundation Cellex Barcelona. 94 ATCC 31158 unveiling new insights into the biosynthesis of the bleomycin family of antitumor antibiotics. Mol Biosyst 2007; 3:60--74. 27 Haraldsen JD, Liu G, Botting CH, Walton JG, Storm J, Phalen TJ et al. Identification REFERENCES of conoidin as a a covalent inhibitor ofperoxiredoxin II. Organic Biomolecular 1 Dong F, Budhu AS, Wang XW. Translating the metastasis paradigm from scientific Chem 2009; 7: 3040 --3048. 28 Budihardjo I, Walker DL, Svingen PA, Buckwalter CA, Desnoyers S, Eckdahl S et al. theory to clinical oncology. Clin Cancer Res 2009; 15: 2588 --2593. 6-Aminonicotinamide sensitizes human tumor cell lines to cisplatin. Clin Cancer 2 Fidler IJ. The organ microenvironment and cancer metastasis. Differentiation 2002; Res 1998; 4: 117 --130. 70: 498 --505. 29 Buzzai M, Bauer DE, Jones RG, DeBerardinis R, Hatzivassiliou G, Elstro RL 3Hu¨semann Y, Geigl JB, Schubert F, Musiani P, Meyer M, Burghart E et al. Systemic et al. The glucose dependence of Akt-transformed cells can be reversed by spread is an early step in breast cancer. Cancer Cell 2008; 13:58--68. pharmacologic activation of fatty acid-b.oxidation. Oncogene 2005; 24: 4 Nguyen DX, Massague´ J. Genetic determinants of cancer metastasis. Nat Rev 4165 --4173. Genet 2007; 8: 341 --352. 30 Jang Y, Sharkis S. A low level of reactive oxygen species selects for primitive 5 Weigelt B, Hu Z, He X, Livasy C, Carey LA, Ewend MG et al. Molecular portraits and hematopoietic stem cells that may reside in the low-oxygenic niche. Blood 2007; 70-gene prognosis signature are preserved throughout the metastatic process of 110: 3056 --3063. breast cancer. Cancer Res 2005; 65: 9155 --9158. 31 Wang W, Gou L, Xie G, Tong A, He F, Lu Z et al. Proteomic analysis of interstitial 6 Minn AJ, Gupta GP, Siegel PM, Bos PD, Shu W, Giri DD et al. Genes that mediate fluid in bone marrow identified that peroxiredoxin 2 regulates H O level of bone breast cancer metastasis to lung. Nature 2005; 28: 518 --524. 2 2 marrow during aging. J Proteome Res 2010; 9: 3812 --3819. 7 Minn AJ, Gupta GP, Padua D, Bos P, Nguyen DX, Nuyten D et al. Lung metastasis 32 Lehtonen ST, Svensk AM, Soini Y, Pa¨a¨kko¨ P, Hirvikoski P, Kang SW et al. genes couple breast tumor size and metastatic spread. Proc Natl Acad Sci USA Peroxiredoxins, a novel protein family in lung cancer. Int J Cancer 2004; 111: 2007; 17: 6740 --6745. 514 --521. 8 Gupta GP, Perk J, Acharyya S, de Candia P, Mittal V, Todorova-Manova K et al. ID 33 Semenza GL. Tumor metabolism: cancer cells give and take lactate. Commentary. genes mediate tumor reinitiation during breast cancer lung metastasis. Proc Natl J Clin Invest 2008; 118: 3835 --3837. Acad Sci USA 2007; 104: 19506 --19511. 34 Bonnet S, Archer SL, Allalunis-Turner J, Haromy A, Beaulieu C, Thompson R 9 Padua D, Zhang XH, Wang Q, Nadal C, Gerald WL, Gomis RR et al. TGFbeta primes et al. A mitochondria-K+ channel axis is suppressed in cancer and its breast tumors for lung metastasis seeding through angiopoietin-like 4. Cell 2008; normalization promotes apoptosis and inhibits cancer growth. Cancer Cell 2007; 133: 66 --77. 11:37--51. 10 Cairns RA, Harris IS, Mak TW. Regulation of cancer metabolism. Nat Rev Cancer 35 Ralser M, Wamelink MM, Latkolik S, Jansen EE, Lehrach H, Jakobs C. Metabolic 2011; 11: 85 --95. reconfiguration precedes transcriptional regulation in the antioxidant response. 11 Espan˜a L, Mart´ın B, Aragu¨e´s R, Chiva C, Oliva B, Andreu D et al. Bcl-x(L)-mediated Nat Biotechnol 2009; 27: 604 --605. changes in metabolic pathways of breast cancer cells: from survival in the blood 36 Chambers AF. MDA-MB-435 and M14 cell lines: identical but not M14 melanoma? stream to organ-specific metastasis. Am J Pathol 2005; 167: 1125 --1137. Cancer Res 2009; 69: 5292 --5293. 12 Kim H, Lee TH, Park ES, Suh JM, Park SJ, Chung HK et al. Role of peroxiredoxins in 37 Neumann CA, Fang Q. Are peroxiredoxins tumor suppressors? Curr Opin regulating intracellular hydrogen peroxide and hydrogen peroxide-induced Pharmacol 2007; 7: 375 --380. apoptosis in thyroid cells. J Biol Chem 2000; 275: 18266 --18270. 38 Bubici C, Papa S, Dean K, Franzoso G. Mutual cross-talk between reactive oxygen 13 Wen ST, Van Etten RA. The PAG gene product, a stress-induced protein with species and nuclear factor-kappa B: molecular basis and biological significance. antioxidant properties, is an Abl SH3-binding protein and a physiological inhibitor Oncogene 2006; 25: 6731 --6748. of c-Abl tyrosine kinase activity. Genes Dev 1997; 11: 2456 --2467. 39 Ueda S, Masutani H, Nakamura H, Tanaka T, Ueno M, Yodoi J. Redox control of cell 14 Rabilloud T, Heller M, Gasnier F, Luche S, Rey C, Aebersold R et al. Proteomics death. Antioxid Redox Signal 2002; 4: 405 --414. analysis of cellular response to oxidative stress. Evidence for in vivo overoxidation 40 Graves JA, Metukuri M, Scott D, Rothermund K, Prochownik EV. Regulation of of peroxiredoxins at their active site. J Biol Chem 2002; 277: 19396 --19401. reactive oxygen species homeostasis by peroxiredoxins and c-Myc. J Biol Chem 15 Karihtala P, Ma¨ntyniemi A, Kang SW, Kinnula VL, Soini Y. Peroxiredoxins in breast 2009; 284: 6520 --6529. carcinoma. Clin Cancer Res 2003; 9: 3418 --3424. 41 Lee KW, Lee DJ, Lee JY, Kang DH, Kwon J, Kang SW. Peroxiredoxin II restrains DNA 16 Fujii J, Ikeda Y. Advances in our understanding of peroxiredoxin, a multifunctional, damage-induced death in cancer cells by positively regulating JNK-dependent mammalian redox protein. Redox Rep 2002; 7: 123 --130. DNA repair. J Biol Chem 2011; 286: 8394 --8404. 17 Mart´ın B, Sanz R, Aragu¨es R, Oliva B, Sierra A. Functional clustering of metastasis 42 Han YH, Kwon JH, Yu DY, Moon EY. Inhibitory effect of peroxiredoxin II (Prx II) on proteins describes plastic adaptation resources of breast-cancer cells to new Ras-ERK-NFkappaB pathway in mouse embryonic fibroblast (MEF) senescence. microenvironments. J Proteome Res 2008; 7: 3242 --3253. Free Radic Res 2006; 40: 1182 --1189.

Oncogene (2013) 724 --735 & 2013 Macmillan Publishers Limited PRDX2 promotes lung metastasis V Stresing et al 735 43 Me´ndez O, Ferna´ndez Y, Peinado MA, Moreno V, Sierra A. Anti-apoptotic proteins 49 Zhang RD, Fidler IJ, Price JE. Relative malignant potential of human breast induce non-random genetic alterations that result in selecting breast cancer carcinoma cell lines established from pleural effusions and a brain metastasis. metastatic cells. Clin Exp Metastasis 2005; 22: 297 --307. Invasion Metastasis 1991; 11: 204 --215. 44 Peyruchaud O, Serre CM, NicAmhlaoibh R, Fournier P, Clezardin P. Angiostatin 50 Peyruchaud O, Winding B, Pe´cheur I, Serre CM, Delmas P, Cle´zardin P. inhibits bone metastasis formation in nude mice through a direct anti-osteoclastic Early detection of bone metastases in a murine model using fluorescent activity. J Biol Chem 2003; 278: 45826 --45832. human breast cancer cells: application to the use of the bisphosphonate 45 Boucharaba A, Serre CM, Gres S, Saulnier-Blache JS, Bordet JC, Guglielmi J et al. zoledronic acid in the treatment of osteolytic lesions. J Bone Miner Res 2001; 16: Platelet-derived lysophosphatidic acid supports the progression of osteolytic 2027 --2034. bone metastases in breast cancer. J Clin Invest 2004; 114: 1714 --1725. 51 Pe´cheur I, Peyruchaud O, Serre CM, Guglielmi J, Voland C, Bourre F et al. Integrin 46 Mosmann T. Rapid colorimetric assay for cellular growth and survival: alpha(v)beta3 expression confers on tumor cells a greater propensity to application to proliferation and cytotoxicity assays. J Immunol Methods 1983; metastasize to bone. FASEB J 2002; 16: 1266 --1268. 65:55--63. 52 El Hilali N, Rubio N, Blanco J. Different effect of paclitaxel on primary 47 Vanden Hoek TL, Li C, Shao Z, Schumacker PT, Becker LB. Significant levels of tumor mass, tumor cell contents, and metastases for four experimental human oxidants are generated by isolated cardiomyocytes during ischemia prior to prostate tumors expressing luciferase. Clin Cancer Res 2005; 11: 1253 --1258. reperfusion. J Mol Cell Cardiol 1997; 29: 2571 --2583. 53 Pinheiro JC, Bates DM. Mixed-Effects Models in S and S-PLUS. Springer, Germany, 48 Roma´n I, Vilalta M, Rodriguez J, Matthies AM, Srouji S, Livne E et al. Analysis of 2000. progenitor cell-scaffold combinations by in vivo non-invasive photonic imaging. 54 R Development Core Team R. A Language and Environment for Statistical Biomaterials 2007; 28: 2718 --2728. Computing. Foundation for Statistical Computing: Vienna, Austria, 2009.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

& 2013 Macmillan Publishers Limited Oncogene (2013) 724 --735