Tumor Biol. (2012) 33:1727–1732 DOI 10.1007/s13277-012-0431-2

RESEARCH ARTICLE

Acquired xanthine dehydrogenase expression shortens survival in patients with resected adenocarcinoma of lung Hayato Konno & Yoshihiro Minamiya & Hajime Saito & Kazuhiro Imai & Yasushi Kawaharada & Satoru Motoyama & Jun-ichi Ogawa

Received: 23 March 2012 / Accepted: 28 May 2012 / Published online: 8 June 2012 # International Society of Oncology and BioMarkers (ISOBM) 2012

Abstract Xanthine dehydrogenase (XDH), also known as xanthine oxidoreductase (XOR), has long been recognized as the key enzyme in the catabolism of purines, oxidizing hypoxanthine into xanthine and then xanthine into uric acid. In addition, levels of XDH expression are reportedly related to the prognosis of patients with malignant tumors, though the relationship between the clinicopathological features of lung cancer and XDH is not fully understood. We therefore used semiquantitative real-time reverse transcription polymerase chain reaction to assess expression of XDH mRNA in tumor samples from 88 patients with adenocarcinoma of the lung. We then correlated XDH mRNA levels with known clinicopathological factors. We found that the 5-year overall survival rate among patients strongly expressing XDH was significantly poorer than among those expressing lower levels of XDH (P<0.001; log-rank test). Normal lung tissue does not express XDH. Multivariate Cox proportional hazard analyses revealed that being male (hazard ratio, 3.14; 95 % confidence interval (CI), 1.45–7.07; P00.004), nodal metastasis positivity (hazard ratio, 5.74; 95 % CI, 1.94–19.3; P00.001), and high XDH expression (hazard ratio, 2.33; 95 % CI, 1.11–5.02; P00.026) were all independent factors affecting 5-year disease-free survival. In conclusion, high tumoral XDH expression is an independent predictor of a poor prognosis in patients with adenocarcinoma of the lung. Keywords Adenocarcinoma . Lung cancer . Prognosis . Xanthine dehydrogenase . Xanthine oxidase H. Konno : Y. Minamiya (*) : H. Saito : K. Imai : Y. Kawaharada : S. Motoyama : J.-i. Ogawa Division of Chest Surgery, Akita University Hospital, Akita University Graduate School of Medicine, 1-1-1 Hondo, Akita City 010-8543, Japan e-mail: [email protected]

Introduction Xanthine dehydrogenase (XDH), also known as xanthine oxidoreductase (XOR), has long been recognized as the key enzyme in the catabolism of purines, oxidizing hypoxanthine to xanthine and then xanthine to uric acid [1]. XDH is one of two forms of XOR; the other is xanthine oxidase (XO), which is derived from XDH through posttranslational modification [1]. In humans, XO is normally present as XDH and completes the metabolic pathway by utilizing NAD+ as an electron acceptor. Under pathological conditions, such as ischemia and inflammation, XDH is converted into XO and utilizes oxygen as an electron acceptor [2–4]. The release of XO from injured tissues may lead to further damage of the epithelium, vascular endothelium, and other near sites [5]. The level of XO activity is increased during pathological conditions as compared to normal healthy individuals [4, 6]. The oxidation of hypoxanthine/xanthine and reduction of nitrate/nitrite by XOR leads to the production of reactive oxygen species (ROS) [1], which are reportedly involved in over 150 human disorders [7]. Samra et al. [8] reported that an elevated level of XO activity was observed in cancer patients. ROS can damage a wide variety of biomolecules [9] and are known to attack DNA bases and deoxyribose residues, causing DNA damage and/or genetic mutations that have the potential to contribute to the pathogenesis of cancer [10, 11]. Indeed, it was previously shown that XDH-induced production of ROS leads to oxidative stress that is associated with an increased risk for breast cancer [12, 13] and nonsmall cell lung cancer (NSCLC) [14, 15]. On the other hand, Linder et al. [16–19] found that reducing the normal expression of XDH in mammary gland or epithelial cells in the colon or stomach leads to a poor prognosis in breast, colorectal, and gastric cancer. Furthermore, Kim et al. [20] recently showed that weak XOR expression is associated with

1728

shortened survival in patients with NSCLC. Thus the role XDH in the pathogenesis and progression of cancer remains unclear. To further investigate the actions of XDH in lung cancer, in the present study, we investigated the relationship between tumoral XDH mRNA expression and the clinicopathological characteristics of patients with adenocarcinoma of the lung.

Methods Patients Eighty-eight patients with adenocarcinoma of the lung participated in this study after providing informed consent and signing a human subject institutional review board consent form. All had undergone surgery at the Division of Chest Surgery, Department of Surgery, University Hospital of Akita University School of Medicine between March 1997 and August 2004. In addition, patients with nodal metastasis received platinum-based chemotherapy followed by oral tegafur-uracil for 2 years. Semiquantitative real-time PCR assays Fresh tumor samples were taken from regions macroscopically judged to be neoplastic and immediately stored in liquid nitrogen until use. For semiquantitative real-time PCR, total RNA was isolated from the samples using a FastRNA Kit Green (Qbiogene, Carlsbad, CA, USA) according to manufacturer's instructions. After quantifying the isolated RNA using a spectrophotometer, 1 μg aliquots were reverse transcribed by incubation with Superscript II reverse transcriptase (Gibco BRL, Gaithersburg, MD, USA) and 0.5 μg oligo(dT) 12–18 for 50 min at 42 °C and then for 25 min at 70 °C. We amplified XDH mRNA (GenBank accession no. AF025375) using primers 5′-AACCATCTCAGCCCT CAAGA-3′ (left) and 5′-AGCTCCTCCTTCCAGAGCTT-3′ (right) and Universal ProbeLibrary #34 (Roche Diagnostics, Mannheim, Germany). PCR was carried out in a LightCycler 480 (Roche Diagnostics) using a LightCycler 480 Kit (Roche Diagnostics). Thermocycling was done in a final volume of 20 μl containing 2 μl of sample cDNA (or standard), 0.4 μl of each primer (final concentration 0.2 μM), 0.2 μl of Universal Probe #60 (Roche Diagnostics), 7 μl of water, and 10 μl of LightCycler 480 mix. After a 5-min initial denaturation at 95 °C, the cycling protocol consisted of 45 cycles of denaturation at 95 °C for 10 s, annealing at 60 °C for 30 s, and elongation at 72 °C for 1 s. As an internal control, we also amplified the mRNA for β-actin using primers 5′C C A A C C G C G A G A A G AT G A - 3 ′ ( l e f t ) a n d 5 ′ CCAGAGGCGTACAGGGATAG-3′ (right) and Universal

Tumor Biol. (2012) 33:1727–1732

Probe Library #64 (Roche Diagnostics). The reaction mixture and the cycling conditions was the same as for XDH mRNA. The LightCycler apparatus measured the fluorescence of each sample during every cycle at the end of the annealing step. After proportional background adjustment, the fit point method was used to determine the cycle in which the log-linear signal was distinguishable from the background, and that cycle number was used as the crossing point value. The software then produced a standard curve by measuring the crossing point for each standard sample and plotting them against the logarithmic values of the concentrations. Levels of XDH mRNA were then normalized to those of β-actin. However, because the (XDH mRNA)/(β-actin mRNA) ratios were not normally distributed, they were log transformed to log10 [(XDH mRNA)/(β-actin mRNA)], after which the normality of the log10 [(XDH mRNA)/(β-actin mRNA)] distribution was confirmed using the Shapiro–Wilk test (P 0 0.1358). The log10 [(XDH mRNA)/(β-actin mRNA)] value was defined as the XDH level. Western blot Samples of tumor and normal lung tissue were lysed in lysis buffer (100 mM Tris–HCl (pH 7.4), 1 % Triton X-100, 150 mM NaCl, with Complete Mini protease inhibitors [Roche Biochemicals] per 10 ml), then separated by 12 % SDS-PAGE, and transferred to nitrocellulose membranes. Immunoblots were processed using a rabbit polyclonal anti-XDH antibody (H-110) sc20991 (Santa Cruz Biotechnology, Inc., CA, USA), followed by detection using enhanced chemiluminescence kit (ECL Western blotting reagents, Amersham Biosciences). Immunohistochemical staining Specimens of resected lung cancer tissue obtained at surgery were fixed in formalin, embedded in paraffin, and cut into 3μm-thick sections. The sections were then deparaffinized in xylene and ethanol, placed in 0.1 mol/L citrate buffer (pH 6.0), and irradiated with microwaves (750 W) for 15 min. The cut section was immunostained using a rabbit polyclonal antibody against human XDH (H-110) sc20991. Briefly, the sections were incubated for 1 h at room temperature with the rabbit anti-XDH antibody (1:50 dilution), rinsed five times with PBS, and then incubated with peroxidase-conjugated antimouse/rabbit immunoglobulin (DAKO EnVisionTM System; Dako Corporation, Carpinteria, CA). The sections were then developed with 3,3′-diaminobenzidine tetrahydrochloride and counterstained with hematoxylin. The immunostained specimens were separately evaluated by two investigators blinded to the clinicopathological characteristics of the patients. It is very good cooperative studies by two investigators, which strengthen the practical work

Tumor Biol. (2012) 33:1727–1732

and idea. The percentage of cells, staining positive for XDH expression, was designated as the frequency of XDH expression and was scored as follows on a scale of 0 to 3: 0, <10 % staining; 1, 10–25 % staining; 2, 26–50 % staining; or 3, >51 % staining. In general, XDH positivity was defined as 10 % or more cells staining positive for XDH. In addition to the XDH frequency, XDH intensity was graded on a scale of 0 to 2: 0, no staining; 1, weak staining; and 2, strong staining (Fig. 2). The overall scores of XDH frequency×XDH intensity (XDH FXI score) were compared with the levels of XDH mRNA. The FXI scores of 0–2 were defined as negative staining and higher scores as positive staining. Statistics Values are expressed as means±SD. Categorical data were compared using the chi-squared and Fisher's exact tests. Survival rates were estimated using the Kaplan–Meier method, and statistical analysis was carried out using the log-rank test for equality of the survival curves (JMP 9 statistical software, SAS Institute). Groups were compared using oneway ANOVA in combination with Dunnette's methods. For multivariate analysis, independent prognostic factors were determined using a Cox proportional hazards model (JMP 9). Values of P<0.05 were considered significant.

Results The clinical characteristics of the patients are summarized in Table 1. All patients were staged pathologically according to the seventh edition of the UICC-TNM staging system [21]. To estimate levels of XDH mRNA, a conventional receiveroperating characteristic (ROC) curve was used in order to determine the cutoff value that yielded the highest combined sensitivity and specificity with respect to distinguishing disease-free 5-year survivors from nonsurvivors (Fig. 1). The threshold that gave the maximal sensitivity and specificity for XDH mRNA was 3.11 log10 [(XDH mRNA)/ (β-actin mRNA)]. It was defined that levels of XDH mRNA of >3.11 as being “high” and those of≤3.11 as being “low.” There were no differences between patients expressing high levels of XDH mRNA and those expressing low levels with respect to sex or age (Table 1). On the other hand, XDH levels were significantly lower in patients with no lymph node metastasis than in those with nodal involvement. The patterns of XDH expression revealed by immunohistochemical analysis are illustrated in Fig. 2. Cancer cells showed cytoplasmic staining for XDH (Fig. 2b), whereas normal lung tissue did not stain for XDH (Fig. 2d). To assess the relationship between the expression levels of XDH protein and mRNA, we first scored expression of

1729 Table 1 Clinical characteristics of patients and their XDH levels Total

XDH High (>3.11)

Patients Age Sex Male Female Histological grade Well Moderate Poor Tumor size ≥30 <30 Nodal status Negative Positive Pathological stage I II–IV Adjuvant therapy Positive Negative

88

Low (≤3.11)

34 61.4±12.1

54 65.1±8.5

43 45

17 17

28 26

0.866

33 30 25

9 14 11

24 16 14

0.225

48 40

15 19

33 21

0.119

64 24

19 15

45 9

0.005

60 28

17 17

43 11

0.004

27 55

13 18

14 37

0.179

0.174

XDH protein from the pattern of XDH immunohistochemical staining (see the “Methods” section for details), after which the tumors were defined as negative staining or positive staining based on those scores. It was found that expression of XDH mRNA in the negative-staining tumors was significantly weaker than in the positive-staining tumors (Fig. 3).

Fig. 1 ROC curve used to define the cutoff value between the higher and lower expression levels of XDH mRNA. We analyzed the relative levels of XDH mRNA expression using a conventional ROC curve to determine the cutoff value that yielded the highest combination of sensitivity and specificity with respect to distinguishing 5-year disease free survivors from nonsurvivors. The threshold that gave the maximal sensitivity and specificity for XDH expression was log10 [(XDH mRNA)/(β-actin mRNA)]03.11. Area under the curve00.662. At this threshold, the sensitivity was 33.3 %, and the specificity was 75.0 % (P00.004)

1730

Fig. 2 XDH expression. Tumor tissues were stained with an anti-XDH antibody. Panels depict adenocarcinomas exhibiting strong (a) normal lung (b) strong (c) weak (d) no. Within tumor cells, XDH immunoreactivity was mainly localized in the cytoplasm. Normal lung did not express XDH

Although normal lung tissue did not stain for XDH protein (Fig. 2d), to further confirm the absence of XDH expression in normal lung tissue, the levels of XDH mRNA in tumors were compared with the levels in normal lung tissue. XDH mRNA levels in normal lung tissue were similar to those in tumors expressing low levels of XDH mRNA, but were significantly lower than in tumors expressing high levels of the transcript (Fig. 4a). The western blotting was used to assess the levels of XDH protein in tumors and normal lung tissue (Fig. 4b). Whereas XDH levels varied among the tumors tested, none of the samples of normal lung tissue exhibited XDH expression, which is consistent to the earlier report [16] that no cells express XDH within lung tissue. Figure 5 summarizes the follow-up data. The 5-year overall and disease-free survival rates among patients

Fig. 3 Relationship between the expression of XDH protein and mRNA. XDH protein expression was scored and classified into positive- and negative-staining tumors. XDH mRNA levels in the two tumor groups were then compared. *P00.031

Tumor Biol. (2012) 33:1727–1732

Fig. 4 XDH expression of normal lung and tumors. a XDH mRNA expression is depicted. XDH mRNA levels in normal lung are similar to those in tumors in the low XDH group, but lower than the levels in tumors in high XDH group. Note the value of XDH mRNA was expressed as log10. XDH mRNA levels in normal lung is nearly ten times lower than in the high XDH group. b XDH protein expression of tumors (T1–T4) and normal lung (L1–L5) is depicted

with tumors expressing higher levels of XDH were significantly less than the survival rates among patients with tumors expressing lower levels of XDH. Therefore, a multivariate Cox proportional hazard analysis was carried out to assess the factors affecting 5-year disease-free survival. The factors considered were sex, age, differentiation grade, tumor size, lymph node metastasis, lymphatic invasion, vessels invasion, and XDH. Among those, sex, lymph node metastasis, and XDH were

Fig. 5 Effect of XDH expression on survival rates among patients with adenocarcinoma of the lung. Five-year overall survival (a) and disease-free survival (b) rates were significantly poorer among patients with tumors expressing high levels of XDH mRNA than among those with tumors expressing low levels

Tumor Biol. (2012) 33:1727–1732

1731

Table 2 XDH levels and Cox hazard ratios for disease-free 5-year survival Factors

Number XDH XDH

XDH High Low Sex Male Female Age ≥70 70> Grade Poor Well, moderate Tumor size ≥30 <30 Nodal metastasis Positive Negative Lymphatic invasion Positive Negative Vascular invasion Positive Negative

Multivariate P value

Hazard ratio (95 % CI)

P value

34 54

3.54±0.29 <0.001 2.33 (1.11–5.02) 0.026 2.60±0.46

45 43

2.96±0.62 2.96±0.61

0.790 3.14 (1.45–7.07) 0.004

33 55

2.91±0.63 2.99±0.61

0.701 1.94 (0.86–4.42) 0.111

25 63

2.90±0.72 2.99±0.57

0.778 0.53 (0.22–1.15) 0.111

40 48

2.99±0.72 2.95±0.51

0.513 2.05 (0.87–4.93) 0.100

24 64

3.16±0.52 2.89±0.63

0.054 5.74 (1.94–19.3) 0.001

49 37

30.1±0.62 2.89±0.62

0.458 1.53 (0.52–4.37) 0.433

37 50

3.04±0.63 2.90±0.61

0.257 0.71 (0.27–1.96) 0.503

independent factors affecting the 5-year disease-free survival rate (Table 2).

Discussion We have shown that, in patients with adenocarcinoma of the lung, overexpression of XDH in tumor cells is related to a poor prognosis, and the level of XDH expression is an independent predictor of a poor prognosis. These findings are at variance with those of Linder et al. [16–19] who reported that a reduction in normally expressed XDH in mammary gland and epithelial cells in the colon or stomach leads to a poor prognosis in breast, colorectal, and gastric cancer. The difference may be tissue specific, as those investigators and others observed that normal lung epithelial cells do

not express XDH [16, 19]. That finding is consistent with our present observations and suggests that the transformation of normal lung epithelial cells to cancer cells involves the induction of XDH expression. Moreover, it is likely that the mechanism by which XDH production is acquired in lung cancer involves different signaling pathways than the mechanism underlying the loss of XDH production in breast, colon, and gastric cancers. Kim et al. [20] reported that a low XOR expression frequency is associated with shortened survival and a poorer prognosis in patients who received adjuvant chemotherapy for NSCLC. On the other hand, they found no relationship between XOR intensity and survival [20]. Their findings are in contrast to our observations and may reflect the difference in the methods used. All of our samples were from anatomically resected adenocarcinomas of the lung, whereas 21 of the 82 samples used in the earlier study were biopsy specimens taken from unresected NSCLCs. How this difference in methodology contributed to the apparent difference in results remains to be determined. The mechanism by which overexpressed XDH acts as a tumor progression factor is unknown. However, Lin et al. recently demonstrated that a tumor suppressor, scaffold attachment factor B1 (SAFB1), suppresses XDH expression by directly binding to an E-box motif within the XDH promoter and through interaction with DNA-dependent protein kinase and other tumor suppressors [22, 23]. Moreover, the reduction of SAFB1 levels induced overexpression of XDH and tumor progression, which is consistent with the observed shorter survival times among patients with adenocarcinoma of the lung. On the other hand, reductions in XDH are thought to be associated with an absence of cellular retinol-binding proteins and a resultant retinoic acid deficiency. This defect leads to impaired stem cell differentiation and has been implicated in carcinogenesis [24], which could explain why reducing normally expressed XDH leads to tumor progression. We found that there were more patients with lymph node metastasis in the high XDH group than the low XDH group. Although the precise effect of XDH on lymph node metastasis is not known, this finding raises the possibility that XDH and ROS in some way contribute to lymph node metastasis. Consistent with that idea are the earlier findings that serum ROS levels were higher in patients with nodepositive lung adenocarcinoma [25], and that oxidative stress induces cancer cell metastasis [26]. In conclusion, our findings suggest that induction of XDH expression in tumor tissue is an independent predictor of a poor prognosis in patients with adenocarcinoma of the lung, perhaps in part, because XDH contributes to lymph node metastasis.

1732

Tumor Biol. (2012) 33:1727–1732

Conflicts of interest None. 15.

References 16. 1. Boueiz A, Damarla M, Hassoun PM. Xanthine oxidoreductase in respiratory and cardiovascular disorders. Am J Physiol Lung Cell Mol Physiol. 2008;294:L830–40. 2. McCord JM. Oxygen-derived radicals: a link between reperfusion injury and inflammation. Fed Proc. 1987;46(7):2402–6. 3. Parks DA, Granger DN. Xanthine oxidase: biochemistry and physiology. Acta Physiol Scand Suppl. 1986;548:87–99. 4. Ashraf M, Samra ZQ. Subcellular distribution of xanthine oxidase during cardiac ischemia and reperfusion: an immunocytochemical study. J Submicrosc Cytol Pathol. 1993;25:193–201. 5. Yokoyama Y, Beckman JS, Beckman TK, Wheat JK, Cash TG, Freeman BA, Parks DA. Circulating xanthine oxidase: potential mediator of ischemic injury. Am J Physiol. 1990;258:G564–G70. 6. Samra ZQ, Siddique Z. Distribution of xanthine oxidase during calcium paradox conditions of rat heart. Punjab Univ J Zool. 1998;13:89–98. 7. Halliwell B, Gutteridge JM, Cross CE. Free radicals, antioxidants, and human disease: where are we now? J Lab Clin Med. 1992;119:598–620. 8. Samra ZQ, Pervaiz S, Shaheen S, Dar N, Athar MA. Determination of oxygen derived free radicals producer (xanthine oxidase) and scavenger (paraoxonase1) enzymes and lipid parameters in different cancer patients. Clin Lab. 2011;57:741–7. 9. Nishikawa M. Reactive oxygen species in tumor metastasis. Cancer Lett. 2008;266:53–9. 10. Dizdaroglu M, Jaruga P, Birincioglu M, Rodriguez H. Free radicalinduced damage to DNA: mechanisms and measurement. Free Radic Biol Med. 2002;32:1102–15. 11. Klaunig JE, Kamendulis LM. The role of oxidative stress in carcinogenesis. Annu Rev Pharmacol Toxicol. 2004;44:239–67. 12. Wright RM, McManaman JL, Repine JE. Alcohol-induced breast cancer: a proposed mechanism. Free Radic Biol Med. 1999;26:348–54. 13. Castro GD, Delgado de Layño AM, Costantini MH, Castro JA. Cytosolic xanthine oxidoreductase mediated bioactivation of ethanol to acetaldehyde and free radicals in rat breast tissue. Its potential role in alcohol-promoted mammary cancer. Toxicology. 2001;160:11–8. 14. Kaynar H, Meral M, Turhan H, Keles M, Celik G, Akcay F. Glutathione peroxidase, glutathione-S-transferase, catalase, xanthine oxidase, Cu-Zn superoxide dismutase activities, total glutathione, nitric oxide, and malondialdehyde levels in erythrocytes of

17.

18.

19.

20.

21. 22.

23.

24.

25.

26.

patients with small cell and non-small cell lung cancer. Cancer Lett. 2005;227:133–9. Tsao SM, Yin MC, Liu WH. Oxidant stress and B vitamins status in patients with non-small cell lung cancer. Nutr Cancer. 2007;59:8–13. Linder N, Rapola J, Raivio KO. Cellular expression of xanthine oxidoreductase protein in normal human tissues. Lab Investig. 1999;79:967–74. Linder N, Lundin J, Isola J, Lundin M, Raivio KO, Joensuu H. Down-regulated xanthine oxidoreductase is a feature of aggressive breast cancer. Clin Cancer Res. 2005;11:4372–81. Linder N, Martelin E, Lundin M, Louhimo J, Nordling S, Haglund C, Lundin J. Xanthine oxidoreductase—clinical significance in colorectal cancer and in vitro expression of the protein in human colon cancer cells. Eur J Cancer. 2009;45:648–55. Linder N, Haglund C, Lundin M, Nordling S, Ristimäki A, Kokkola A, Mrena J, Wiksten JP, Lundin J. Decreased xanthine oxidoreductase is a predictor of poor prognosis in early-stage gastric cancer. J Clin Pathol. 2006;59:965–71. Kim AW, Batus M, Myint R, Fidler MJ, Basu S, Bonomi P, Faber LP, Wightman SC, Warren WH, McIntire M, Arvanitis LD, Gattuso P, Xu X, Liptay MJ. Prognostic value of xanthine oxidoreductase expression in patients with non-small cell lung cancer. Lung Cancer. 2011;71:186–90. Sobin L, Wittekind CH, editors. TNM classification of malignant tumours. 6th ed. New York: Wiley; 2002. p. 99–103. Oesterreich S. Scaffold attachment factors SAFB1 and SAFB2: innocent bystanders or critical players in breast tumorigenesis? J Cell Biochem. 2003;90:653–61. Lin J, Xu P, LaVallee P, Hoidal JR. Identification of proteins binding to E-Box/Ku86 sites and function of the tumor suppressor SAFB1 in transcriptional regulation of the human xanthine oxidoreductase gene. J Biol Chem. 2008;283:29681–9. Taibi G, Carruba G, Cocciadiferro L, Granata OM, Nicotra CM. Low levels of both xanthine dehydrogenase and cellular retinol binding protein are responsible for retinoic acid deficiency in malignant human mammary epithelial cells. Ann N Y Acad Sci. 2009;1155:268–72. Tsukioka T, Nishiyama N, Iwata T, Nagano K, Tei K, Suehiro S. Preoperative serum oxidative stress marker as a strong indicator of nodal involvement in clinical stage I lung adenocarcinoma. Int J Clin Oncol. 2011. doi:10.1007/s10147-011-0283-6 Mongaret C, Alexandre J, Thomas-Schoemann A, Bermudez E, Chéreau C, Nicco C, Goldwasser F, Weill B, Batteux F, Lemare F. Tumor invasion induced by oxidative stress is dependent on membrane ADAM 9 protein and its secreted form. Int J Cancer. 2011;129:791–8.