Abstract
Although the causal relationship between chronic inflammation and carcinogenesis has long been discussed, the molecular basis of the relation is poorly understood. In the present study, we focused on reactive oxygen species (ROS) and their signals under inflammatory conditions leading to the carcinogenesis of epithelial cells and found that repeated treatment with a low dose of H2O2 (0.2 mmol/L) for periods of 2 to 4 days caused a phenotypic conversion of mouse NMuMG mammary epithelial cells from epithelial to fibroblast-like as in malignant transformation. The phenotypic conversion included the dissolution of cell-cell contacts, redistribution of E-cadherin in the cytoplasm, and up-regulation of a set of integrin family members (integrin α2, α6, and β3) and matrix metalloproteinases (MMPs; MMP-3, -10, and -13), as analyzed using Northern blot analysis and quantitative reverse transcription-PCR. Gelatin zymography indicated post-transcriptional activation of gelatinases, including MMP-2 and -9. In parallel, p38 mitogen-activated protein kinase and extracellular signal-regulated kinase 1/2 were activated, which contributed to the induction of MMP-13, and a glutathione S-transferase pull-down assay showed the activation of a small GTPase, Rac1. Surprisingly, the prolonged oxidative treatment was sufficient to induce all of the aforementioned events. Most importantly, depending on the MMP activities, the epithelial cells exposed to oxidative conditions eventually acquired invasiveness in a reconstituted model system with a Matrigel invasion chamber containing normal fibroblasts at the bottom, providing the first substantial evidence supporting the direct role of ROS signals in the malignant transformation of epithelial cells.
INTRODUCTION
The causal relationship between inflammation and carcinogenesis has been discussed ever since the possibility was raised in the late 19th century that cancer originated from sites of chronic inflammation. The most compelling evidence in support of such a relation was provided by cases of colon carcinogenesis originating from chronic ulcerative colitis, Crohn’s disease, and hepatitis C infection. Furthermore, it is believed that approximately one-third of cancers are caused by the effects of chronic inflammation. Thus, it is now widely accepted that inflammation is a critical factor in tumor progression (1). However, the molecular basis of the relationship has remained largely unclear. The complexity of the process of inflammation, involving several different types of cells and a wide variety of chemicals and growth factors composing networks of signals, is a major obstacle to detailed analyses at a molecular level.
Among the molecules involved in inflammation, reactive oxygen species (ROS) such as superoxide anion and hydrogen peroxide, which are released from leukocytes and other phagocytic cells that are accumulated at sites of infection and injury, are the most likely to play a role in linking inflammation to carcinogenesis. Numerous studies have established a deleterious effect of ROS on DNA that results in permanent genomic alterations such as point mutations, deletions, or rearrangements of genes involving proto-oncogenes and tumor suppressor genes (1). Thus, chronic inflammation tends to be causally related to neoplasms through DNA damage caused by ROS (1).
Apart from the genotoxic effects, the ROS generated at low levels from enzymatic sources such as NADPH oxidase in response to a variety of extracellular stimuli now are assigned a distinctive role as signal messengers required for the optimal activation of signaling pathways, mediating a wide range of cellular responses such as adhesion/migration, proliferation, differentiation, apoptosis, and senescence under pathophysiologic and physiologic conditions (2, 3, 4, 5). Furthermore, in recent years, a host of molecules have been reportedly identified as direct or indirect targets of ROS, potentially constituting the molecular basis underlying ROS signaling (2, 3, 4). Of these, accumulating evidence has highlighted the role of protein tyrosine phosphatases (PTPs), whose activity is susceptible to the cellular redox state (6, 7, 8). Being primarily modified by ROS, PTPs have emerged as a kind of a receptor for ROS signaling (9, 10, 11, 12, 13, 14, 15).
Although less well defined than PTPs, protein tyrosine kinases such as src family members also have been noted as targets of ROS, whose activation initiates a flow of downstream signal transduction in which intermediate roles often are carried out by mitogen-activated protein (MAP) kinases (16, 17). At the endpoint of the flow of these pathways, the signals generally are coupled to transcriptional activity in the nucleus. In some cases, the signals ultimately promote cellular proliferation, following the induction of proto-oncogene such as c-fos and c-myc (18). Direct or indirect redox regulation of transcription factors also has been well studied (19, 20).
In summary, depending on the dose, species, and situation, ROS modify distinct cellular molecules as their targets, thereby exhibiting pleiotropic, including epigenetic and genetic, effects on cells. Thus, the epigenetic effects of ROS also are conceivably associated with cellular transformation under chronic inflammation, leading to eventual malignant conversion. Despite a large number of provocative observations implicating ROS in the cellular transformation, the epigenetic effects of ROS thus far are not well described in relation to metastasis and pathogenesis and far from comprehensively understood. Most of the previous results also were obtained using fibroblastic cells. As for the malignant conversion of normal epithelial cells, from which a majority of human neoplasms originate, little effort has been made to elucidate the role of the ROS signals.
In this study, we investigated epigenetic effects of ROS on cellular phenotypes, particularly those of epithelial cells, aiming at comprehensive understanding of the role of ROS signals in malignant transformation. We examined the changes in morphology and in gene expression of mouse NMuMG mammary epithelial cells on long-term exposure to H2O2, mimicking chronic inflammation, and showed that the oxidative conditions induced a cellular phenotypic conversion with striking similarities to malignant transformation, accompanied by the induction of genes associated with cell adhesion and migratory behavior together with activation of the small G protein Rac1 and MAP kinases. Importantly, following these changes, the epithelial cells ultimately acquired the potential to invade a reconstituted basement membrane in the presence of normal fibroblasts.
MATERIALS AND METHODS
Cell Culture and Chemicals.
NMuMG (mouse mammary gland epithelial) cells were obtained from American Type Culture Collection (Manassas, VA) and cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 4.5 g/L of glucose, 10 μg/mL of insulin (Sigma, St. Louis, MO), and 50 μg/L of kanamycin at 37°C in an atmosphere of 5% CO2 in air and passaged every third day following treatment with 0.05% trypsin. Human TIG-7 normal fibroblasts were obtained and cultured as described previously (21).
Anisomycin, SB203580, PD98059, and Galargin were purchased from Sigma.
Immunocytochemistry and Immunoblot Analysis.
Immunocytochemistry and immunoblot analysis were performed as reported previously (22). For immunocytochemistry, a monoclonal antibody to E-cadherin (Transduction Laboratories, Lexington, KY) and FITC-conjugated antimouse IgG (Dako, Copenhagen, Denmark) were used as the primary and secondary antibodies, respectively. F-actin was stained with tetramethylrhodamine isothiocyanate-conjugated phalloidin (Sigma). Fluorescence microscopy was carried out using an Axioskope microscope (Carl Zeiss, Oberkochen, Germany) equipped with a high-speed cooled digital charge-coupled device camera fluorescence imaging system (HiSCA; Argus, Inverness, IL).
For immunoblot analysis, the following were used as primary antibodies: E-cadherin, monoclonal (Transduction Laboratories); matrix metalloproteinase-13 (MMP-13), polyclonal (Oncogene Research Products, Cambridge, MA); active–extracellular signal-regulated kinase (ERK) 1/2, -p38, and –c-Jun NH2-terminal kinase (JNK; phospho-p44/42 MAP, -p38 MAP, and –stress-activated protein kinase/JNK kinase), polyclonal (New England Biolabs, Inc., Beverly, MA); and pan-ERK1/2 (Zymed Laboratories, Inc., San Francisco, CA). The secondary antibody was horseradish peroxidase-conjugated antimouse IgG antibody from Amersham Biosciences (Piscataway, NJ).
Northern Blot Analysis and Quantitative Reverse Transcription-PCR.
The procedure used for Northern blot analysis was essentially that described previously (23). A mouse MMP-2 cDNA fragment provided by Dr. Seiki (Institute of Medical Science, University of Tokyo, Tokyo, Japan; ref. 24) and fragments of MMP-9 and MMP-13 from Dr. Miyaura (Tokyo University of Pharmacy and Life Science, Tokyo, Japan; ref. 25) were used as probes. The probe for glyceraldehydes-3-phosphate dehydrogenase (GAPDH) was described previously (23).
For quantitative reverse transcription-PCR, 1 μg of total RNA extracted with TRIzol reagent (Life Technologies, Inc., Rockville, MD) was reverse-transcribed with random hexamer (Takara Shuzoh Co., Kyoto, Japan) and SuperScript II (Invitrogen, Carlsbad, CA). A fragment of MMP cDNA subsequently was amplified by PCR with 40 cycles of denaturing (90°C, 20 seconds), annealing (55°C, 20 seconds), and extension (72°C, 30 seconds) using SYBR Green PCR master mix (Applied Biosystems, Foster City, CA). The monitoring and quantitative analysis of PCR products were performed with a GeneAmp 5700 (Applied Biosystems), and the amount of PCR product derived from each mRNA was normalized to that from GAPDH in the same sample whose expression was essentially stable with deviation within 10% from sample to sample. All of the PCR primers were designed using PrimerExpress 1.0 (Applied Biosystems).
Gelatin Zymography.
NMuMG cells were treated with H2O2 (0.2 mmol/L) for 2 or 4 days under normal condition, and the medium then was changed to serum-free DMEM. After 24 hours, the conditioned medium was collected, and equal aliquots were concentrated using Molcut II (Millipore Corporation, Bedford, MA), diluted in the sample buffer [50 mmol/L Tris (pH 6.8), 0.5% SDS, 10% glycerol, and 0.2% bromphenol blue], and separated by electrophoresis on a 7.5% polyacrylamide gel containing 1 mg/mL of gelatin as substrate. Gels were washed in 2.5% Triton X-100 for 1 hour at room temperature, incubated overnight in a reaction buffer [50 mmol/L Tris (pH 7.6), 150 mmol/L NaCl, 2.5 mmol/L CaCl2, and 0.02% sodium azide] and stained in Coomassie blue solution (0.25% Coomassie blue R250, 50% methanol, and 7.5% acetic acid).
Pull-Down Assay.
The active forms of Rac1 and Cdc42 were detected with glutathione S-transferase (GST) fused to the CRIB domain (residues 29 to 90) of PAK (GST-CRIB; ref. 26) and that of RhoA with a GST-fused form of Rhotekin, a Rho effecter protein (GST-Rhotekin), provided by Dr. Narumiya (Kyoto University, Kyoto, Japan; ref. 27).
Cells were lysed on ice in lysis buffer [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1% Triton X-100, 0.5% deoxycholate, and proteinase inhibitor mixture; Wako Pure Chemical Industries, Ltd, Osaka, Japan], and the lysate was mixed, rotated with GST-CRIB or GST-Rhotekin precoupled to Sepharose-glutathione beads (Amersham Biosciences) for 1 hour at 4°C, and washed three times in wash buffer [20 mmol/L Tris (pH 7.5), 150 mmol/L NaCl, 1% Triton X-100, and protease inhibitor mixture]. The bound Rac1, Cdc42, and RhoA were eluted for 1 hour at 4°C in elution buffer [50 mmol/L Tris (pH 8.0), 200 mmol/L KCl, 20 mmol/L glutathione, proteinase inhibitor mixture], and subjected to Western blot analysis. Anti-Rac1 and -Cdc42 were obtained from Transduction Laboratories, and anti-RhoA was from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA).
Invasion Assay.
TIG-7 cells were grown in the bottom of a BD BioCoat Matrigel Invasion Chamber (Discovery Labware, Bedford, MA) until confluent and treated with H2O2 for 48 hours. NMuMG cells (2.0 × 105) then were suspended in normal culture medium, seeded in the upper layer of a well, and incubated with or without H2O2 for 72 hours. Following the removal of the noninvaded cells from the upper surface of the Matrigel with a cotton swab, the invaded cells were fixed, stained with 0.5% crystal violet, and then lysed using 10% SDS. The absorbance at 595 nm representing the number of cells was measured.
RESULTS
Fibroblastic Phenotypes Induced by Prolonged Exposure to H2O2.
We first examined microscopically the changes in cell morphology and actin fibers that were induced in mouse NMuMG mammal epithelial cells on prolonged exposure to H2O2. In this study, we used a relatively low dose of H2O2 (0.2 mmol/L) that was sublethal to the cells 1 day after the treatment. A single exposure of the cells to H2O2 did not cause a significant change in morphology (data not shown), but multiple treatments resulted in remarkable change (Fig. 1,A). With daily exposure to H2O2, the cells gradually became enlarged and elongated, eventually more than four times bigger than untreated cells in diameter after 4 days (Fig. 1,A). This morphologic change was persistent after withdrawal of H2O2, suggesting its irreversibility. A similar morphologic change was induced in fibroblastic cells by reiterated treatment with H2O2, although to a lesser extent than in the epithelial cells (data not shown). When actin-based cytoskeletons were stained with phalloidin, the daily exposure to H2O2 induced small projections resembling prickles at the cell periphery and punctuate spots in the cytoplasm, in addition to the cortical filaments that were prominent in the original NMuMG cells (Fig. 1 B).
In addition to the changes in individual cells, of note was a scattered mesenchymal morphology with gaps between cells, implying disruption in the integrity of cell-cell contacts (Fig. 1,A and B). On the basis of the well-established fact that cell-cell adhesion systems are maintained by adhesion molecules such as E-cadherin constituting a junction structure, we carried out the indirect immunofluorescence labeling of E-cadherin to examine whether cell-cell adhesion was affected under oxidative conditions. In line with the change noted previously, the subcellular distribution of E-cadherin was considerably disturbed by the oxidative conditions (Fig. 1,C). E-cadherin labeled with a specific antibody was almost exclusively concentrated at the cell-cell boundary before the treatment, but after daily treatment with H2O2 for 4 days, the signals dispersed mostly throughout the cytoplasm as small dots, suggesting vesicular localization, with residuals at cell edges (Fig. 1,C). The expression level of the protein was unaltered under the conditions (Fig. 1 D).
These morphologic changes under prolonged oxidative stress, including the dissolution of cell-cell contacts, suggested a conversion of the cell adhesive mode from epithelial to fibroblastic in which cell–extracellular matrix (ECM) interactions predominate over cell-cell adhesions. To gain more insight into this point, we next investigated the expression of integrin family members, which are key molecules in mediating cell-matrix interactions as a receptor for the ECM and in transmitting microenvironmental cues to inside the cells. Total RNA was extracted from the cells exposed to H2O2 for 2 and 4 days, and the expression of a subset of integrins was examined by quantitative reverse transcription-PCR. As shown in Fig. 2, oxidative conditions notably affected the expression patterns, and the effects were enhanced by repetition of the treatment. The expression was up-regulated to some extent except in the case of integrin α1, whose expression was down-regulated. Most prominent was the sixfold to sevenfold induction of integrins α2, α6, and β3. Integrins α2 and α6 constitute a collagen and laminin receptor and a major receptor for laminin, respectively. Integrin β3 serves as a receptor for vitronectin and fibronectin together with αv or αIII. Interestingly, recent studies suggested that α6β4 functions in cell migration and α2β1 functions in the dispersion of epithelial cells (28, 29). The change in the expression patterns of these integrin family members under oxidative conditions was possibly one of the molecular events supporting the morphologic changes.
Increased Expression and Activity of MMPs under Prolonged Oxidative Stress.
We further characterized the phenotype induced in the epithelial cells by the prolonged exposure to oxidative conditions by searching for genes whose expression was affected on reiterated exposure to H2O2 and found that the expression of a subset of MMPs was remarkably increased. Representative of these MMPs was MMP-13 (a functional mouse analog of human MMP-1). Northern blot analysis clearly showed that its expression was induced by daily treatment with H2O2 for 2 days and further augmented for another 2 days (Fig. 3,A). The expression levels of MMP-2 and -9 in NMuMG cells were undetectable by Northern blot analysis. Quantification of the mRNA by reverse transcription-PCR showed that MMP-13 expression increased >20-fold after the repetition of treatment for 4 days (Fig. 3,A). In parallel with the increase in mRNA, Western blot analysis showed an increase in the amount of MMP-13 protein released from the cells into the medium with a slight shift in mobility for reasons as yet unknown (Fig. 3 D). To our knowledge, this is the first indication that MMP-13 expression was sensitive to the cellular redox state and induced directly by H2O2 in cells of epithelial and fibroblastic origin (see below).
In Fig. 3,B, we examined the significance of the reiteration of H2O2 treatment on the MMP-13 induction. In this experiment, NMuMG cells were treated once with the oxidant at specific doses on the first day, and total RNA was extracted after 2 and 4 days. Northern blot analysis indicated that MMP-13 mRNA was clearly induced 2 days after the single treatment at doses >0.2 mmol/L and then declined (0.2 mmol/L) or was continuously expressed (0.4, 0.6 mmol/L) at the fourth day (Fig. 3 B). Thus, repetition of the treatment appeared to be necessary for the sustained expression, especially with low doses of the oxidant, but not for the induction itself.
The investigation with quantitative reverse transcription-PCR was extended to other MMPs. Fig. 3 A shows that besides MMP-13, MMP-3 (stromelysin-1) and -10 (stromelysin-2) were induced under the conditions. In sharp contrast to these three MMPs, MMP-2, -9, -7 (matrilysin), and -11 (stromelysin-3) showed only a marginal increase in their mRNA levels. This result pointed out the specificity of the response to the oxidative conditions among MMPs, making it unlikely that the general detrimental effects of the oxidative conditions caused the induction.
According to recent reports, oxidative stress up-regulated the enzymatic activity of MMP-2 or -9 in cancer cells and cardiac fibroblasts (30, 31, 32). We then performed a gelatin zymographic assay of the conditioned medium from the NMuMG cells exposed daily to H2O2 to detect the change in the activity of gelatinases including MMP-2 and -9, although their mRNA was only slightly affected under the oxidative conditions (Fig. 3,A). The assay detected increased activity at the molecular weights corresponding to pro-MMP-9, MMP-9, and pro-MMP-2 (Fig. 3 D). The increase was moderate but reproducible after the treatment for 4 days, suggesting that in this cell line, MMP-2 and -9 were activated by the prolonged oxidative stress.
In conclusion, the results above (Figs. 1, 2, and 3) suggested that under prolonged oxidative stress, the epithelial cells underwent dramatic phenotypic changes characterized by the dissolution of cell-cell contacts, relocalization of E-cadherin, and up-regulation of integrins and MMPs, which were regulated at the transcriptional and/or post-transcriptional level, collectively resulting in an overall down-regulation of epithelial phenotypes and up-regulation of more motile and invasive fibroblastic phenotypes.
For comparison, we investigated the expression and activity of MMPs in fibroblastic cells on exposure to H2O2. Here we used human TIG-7 cells, normal diploid fibroblasts. As shown in Fig. 4, reverse transcription-PCR analysis indicated that as in NMuMG, MMP-13, stromelysin-1, and stromelysin-2 also were induced in this cell line by prolonged exposure to the oxidant, although between the two cell lines, a difference was noted in the magnitude of induction among the MMPs. In TIG-7 cells, the induction of MMP-3 (stromelysin-1) was outstanding with levels reaching >100 times the basal value (Fig. 4,A). Gelatin zymography showed an increase in activity at the positions corresponding to activated MMP-9 and MMP-2 with additional activity at a lower molecular weight that was absent in NMuMG cells (Fig. 4 B). Thus, the up-regulation of the MMP members appeared to be common to both types of cells under oxidative conditions.
Mitogen-Activated Protein Kinases Activated under Prolonged Oxidative Stress and Involved in MMP-13 Expression.
We next studied the signal transduction pathways activated under the oxidative conditions. We focused on those pathways that lead to MMP-13 expression, whose sensitivity to oxidative conditions is consistently high, and tested the involvement of MAP kinases for which important roles have been established in the signal transduction pathway downstream of growth factors and cytokines, leading to the activation of gene expression in the nucleus (33, 34, 35). We first examined the effects of PD98059 and SB203580, selective inhibitors for MAP kinase kinase 1/2 and p38 MAP kinases, respectively, on the induction of MMP-13 expression. Fig. 5,A and B shows that both inhibitors almost completely abolished the induction, suggesting that the two kinase activities were essential to induce MMP-13 expression on treatment with H2O2. LY294002, an inhibitor for phosphatidylinositol 3′-kinase, did not affect the induction (data not shown), excluding the involvement of phosphatidylinositol 3′-kinase. The activation of ERK1/2, p38, and JNK of the MAP kinase family subsequently was examined by Western blot analysis to detect the activated forms of each kinase using phospho-specific antibodies. As seen in Fig. 5 C and D, the phosphorylated forms of ERK1/2 and p38 increased on daily treatment with H2O2 for 2 and 4 days. The level of activated ERK1/2 appeared to reach a plateau by the second day, but that of activated p38 continued to increase, reaching a comparable level to that evoked by anisomycin, a strong inducer for p38 and JNK kinases, after the 4 days. Unlike the two kinases, the phosphorylation of JNK was not evident under the oxidative conditions despite its efficient phosphorylation by anisomycin. The results suggested that under oxidative conditions, ERK1/2 and p38 kinases but not JNK were activated and mediated the expression of MMP-13 in NMuMG cells.
Up-regulation of the Small GTPase Rac1 and Acquisition of Invasive Potential under Prolonged Oxidative Stress.
As above, the cells underwent marked alterations of phenotype under long-term oxidative stress changing from epithelial- to fibroblast-like cells, which might have required the activation of multiple paths of signaling and their coordinated regulation. Accordingly, it was conceivable that the prolonged oxidative treatment activated key regulators that were able to trigger a vast array of downstream signals including MAP kinases. In a search for such regulators, we focused on the Rho-like small GTPases, RhoA, Rac1, and Cdc42Hs, and examined their activities under prolonged oxidative stress. In recent years, these three have been established as upstream regulators critical for the actin reorganization and adhesive structural changes associated with cellular size, shape, and motility (36). We used the GST–pull-down assay to assess their activity levels and found that the activity of Rac1 specifically increased after the treatment (Fig. 6 A). Rac1 activity increased nearly fourfold in several independent experiments. This is the first study to suggest the regulation of Rac1 activity by ROS. The increase in the active form of Rac1 was not clear by the fourth day, suggesting that it required continual exposure to the oxidant, paralleling the changes in morphology, gene expression, and activation of p38 MAP kinase that were augmented by reiteration of the treatment. In contrast, the amount of active Cdc42 remained constant even in the presence of the oxidant, and that of active RhoA was undetectable under the experimental conditions.
Of note, activation of Cdc42 and Rac1 was shown to disrupt the polarization of mammary epithelial cell and promote motility and invasion (37). The activation of Rac1 together with the basal activity of Cdc42 and induction of MMPs in NMuMG cells under conditions of prolonged oxidative stress strongly implied that even cells of normal origin acquired invasive potential under the circumstances. We finally tested this possibility in a reconstituted in vitro model system with a Matrigel invasion chamber. When seeded on the upper layer coated with Matrigel and exposed daily to H2O2 for 2 days, NMuMG cells did not show a marked potential to invade the gel (data not shown). However, under coculture with the normal diploid fibroblasts that were grown at the bottom of the chamber and exposed to the oxidant for 2 days beforehand, a significant population of the NMuMG cells migrated to the lower layer in response to the oxidant (Fig. 6,B–D). The pretreatment of the fibroblasts with the oxidant resulted in a marked increase in the number of migratory cells, implying the requirement of the accumulation of some activity such as MMP to promote the migration (see below). In the conditions, cell growth was repressed because of the cytostatic effect of the oxidant; thus, the density of NMuMG cells on the upper layer was evidently less than that of the untreated control as seen in Fig. 1 A.
Among the changes in cellular phenotype, induction and/or activation of MMPs were the most likely events contributing to the acquisition of invasive potential, and it was possible that the MMP activities produced by NMuMG and TIG-7 cells quantitatively and qualitatively collaborated to facilitate the movement of the NMuMG cells across the in vitro reconstituted basement membrane. Not mutually exclusive but less likely was the possibility that molecules other than MMPs produced by the fibroblasts stimulated the motility or invasiveness of NMuMG cells in a oxidant-dependent manner. Support for the critical contribution of MMP activities was obtained from the experiment using an inhibitor with a broad specificity for MMPs, Galargin. The inhibitor markedly prevented migration (Fig. 6 B and C).
DISCUSSION
During transformation into invasive carcinoma, epithelial cells undergo profound alterations in morphology and adhesive mode, resulting in a loss of normal epithelial polarization and differentiation, and a switch to a more motile, invasive phenotype. In the present study, we found that oxidative conditions caused a phenotypic conversion of mammary epithelial cells, from epithelial to fibroblast-like, similar to the malignant transformation process (Fig. 1). Repeated treatment with H2O2 for periods of 2 to 4 days induced the disruption of cell-cell adhesions, redistribution of E-cadherin (Fig. 1,C), induction of sets of integrin family members (Fig. 2) and MMPs (Fig. 3), and activation of the small GTPase Rac1 (Fig. 6), p38 MAP kinase, and ERK1/2 (Fig. 5) in the epithelial cells. Surprisingly, the oxidative stress was sufficient to induce all of the aforementioned events along with the eventual acquisition of invasiveness in an in vitro reconstituted model system including normal fibroblasts.
Similar loss of epithelial phenotype has been observed during epithelial-mesenchymal transition that is induced by ECM components and soluble factors, such as transforming growth factor β1 (38), whose signaling was partly mediated by H2O2 (39). We then compared the cellular responses to the oxidative stress with those to transforming growth factor β1 and noticed that some but not all of the responses were commonly induced. For example, relocalization of E-cadherin occurred in both cases, whereas the members of MMPs and integrins induced during the two processes were different (data not shown). Among each family member, MMP-9 and integrin α5 were the most inducible by transforming growth factor β1 in this cell line.1 Thus, although partly overlapped, distinct sets of signaling pathways appeared to be activated as the responses to stress or physiologic stimulus.
We currently know little about the role of the epigenetic effects of ROS in the malignant transformation of cells. In fibroblastic cells, the sustained production of H2O2 recently was shown to activate MMP-2 and to increase cell invasion (31). The present study provides the first substantial evidence in support of a direct role for ROS signals in the malignant transformation of epithelial cells. To our knowledge, this also is the first study showing the direct and concomitant induction of subsets of MMPs and integrin subunits at the mRNA level under oxidative conditions. A growing list of genes now have been shown to depend on ROS signals for their expression (2, 3), but only a limited number, such as c-fos, c-jun, c-myc, egr-1, KC, and JE, are directly induced by ROS (39, 40, 41). Collectively, the present study has shed light on the long-sought linkage between chronic inflammation and tumorigenesis at a molecular level and also provided an experimental model to study further the molecular mechanisms underlying the epigenetic effects of ROS on cellular phenotype and malignant transformation.
Activation of the Small GTPase Rac1 and Morphologic Changes.
In the present study, we first addressed the possibility that one of the small Rho GTPases, Rac1, was a target activated by ROS and mediated the signals. Among small GTPases, Ras, Gαi, and Gαo reportedly were activated by oxidants (42, 43), whereas the redox regulation of other members is unknown. The results shown in Fig. 6 A indicated that Rac1 was up-regulated in H2O2-treated NMuMG cells. Because Rac1 is known as a component of NADPH oxidases required for their activation, the result suggested the presence of a positive feedback loop between the activation of Rac1 and production of ROS. This may explain the high level of ROS production in neoplastic cells (44).
An increasing body of evidence has critically implicated small Rho GTPase (RhoA, Rac1, and Cdc42), key regulators of the actin cytoskeleton and adhesive structures, in epithelial tumor progression (37). In particular, Rac1 has been revealed to play a critical role in the motility and/or invasiveness of cells together with Cdc42 or other signaling molecules (45, 46, 47, 48). Thus, the activation of Rac1 by prolonged oxidative stress was noteworthy in terms of relating ROS signals to the malignant conversion of epithelial cells at a molecular level. To obtain more detailed information in this respect, we tried to find a causal relation between Rac1 activity and the observed events by introducing dominant negative forms of small GTPases into the cells and found that the Rac1 activity appeared to be unrelated at least to the expression of MMPs, including MMP-13 (data not shown). Rather, plenty of evidence as mentioned previously suggested that the increased activity of Rac1 observed in this study (Fig. 6 A) was responsible for the observed morphologic change, disruption of cell-cell adhesion, and enlargement of the cells induced by the oxidant along with the basal activity of Cdc42.
With regard to downstream signaling, the activity of Rac1 was reportedly coupled to that of JNK (49, 50). In the experimental conditions of our study, however, JNK was not activated. Instead, ERK1/2 and p38 MAP kinases were activated (Fig. 5,C and D), and their activities contributed to the expression of MMP-13 (Fig. 5 A and B). From these findings, we speculate at this stage that ROS activated two independent signaling pathways, one of which was downstream of Rac1 mediating morphologic changes and the other of which led to the gene expression through the activities of p38 and ERK1/2 kinase. The precise mechanisms underlying the signaling pathways leading to each outcome, including the downstream targets of Rac1, await additional study.
Induction and Activation of MMPs and Invasive Potential.
Invasion is a central feature of malignancy and supported by the combined actions of several proteolytic enzyme systems. It is believed that the MMPs play the central role, and their increased expression reportedly is associated with the invasion and metastasis of malignant tumors of different histogenetic origins (51). In this study, we found that MMP-13, -3, and -10 were remarkably up-regulated by the oxidant directly (Fig. 3,A), and consistent with the above premise, their activities were critically implicated in the invasive potential induced in NMuMG cells in the reconstituted model (Fig. 6 B and C).
MMP-13, together with MMP-1 and -8, has the ability to degrade native fibrillar collagens, thereby playing an initial role in the remodeling of collagenous ECM. Importantly, MMP-13 displays an exceptionally broad substrate specificity. In addition to fibrillar collagens, its substrates include components of the basement membrane such as type IV collagen, laminin, and fibronectin, and it also shows >40 times stronger gelatinase activity than MMP-1 and -8 (51). Such wide substrate specificity strongly implicates it in the metastasis of neoplastic cells as a potent proteolytic weapon of invasion. The expression of MMP-13 has been detected in breast carcinomas, squamous cell carcinomas of the head, neck, and vulva, cutaneous basal cell carcinomas, and chondrosarcomas (51). MMP-3 and -10, classified as stromelysins, also mainly degrade basement membrane components, such as type IV collagen, and are implicated in neoplastic conversion (52). Another subgroup of MMPs, gelatinases (MMP-2 and -9), which are key enzymes for degrading type IV collagen and thought to play a critical role in tumor invasion and metastasis (51), were found to be activated post-transcriptionally by prolonged oxidative treatment in the present study (Figs. 3,D and 4 B). Thus, like Rac1, MMP-3, -10, and -13 along with MMP-2 and -9 were expected to be effector molecules induced or activated under prolonged oxidative stress and relating chronic inflammation to malignant transformation, in particular to the invasive potential of cells, at a molecular level.
MMP gene expression is primarily regulated at the transcriptional level. Given that a variety of MMPs like MMP-1 and -3 are subjected to similar transcriptional regulation because of their promoter similarities (51), the concomitant induction of MMP-3, -10, and -13 by ROS would not be surprising. The induction of human MMP-1 (53, 54) and MMP-3 (55) reportedly was related to ROS signaling, whereas the direct role of naturally occurring ROS has not been addressed in previous studies. Unlike these MMPs, it was suggested that MMP-2 and -9 were regulated at the post-transcriptional level under oxidative conditions (Figs. 3,D and 4 B), consistent with previous reports (30, 31, 32), although the activating mechanisms are as yet undefined.
The diverse array of events occurring under oxidative conditions presumably relies on the pleiotropic character of ROS signals. Accumulating evidence has shown that ROS transmit signals by modifying redox-sensitive molecules, and the category of such putative targets has expanded in recent years to include kinases, phosphatases, small GTPases, transcription factors, and signal adaptors (56) as mentioned previously. This diversity of targets potentially provides an opportunity for the concomitant stimulation of multiple signaling pathways, which would meet the requirements for malignant transformation, which is predicted to involve multiple interconnected mechanisms. The reversible and specific oxidation of the low pKa sulfhydryl group of the cysteine of proteins has emerged as one of the initial events in ROS signaling (9, 11, 12, 13). For the chemoprevention of cancer, it appears critical to identify the primary targets of ROS and to elucidate their roles in phenotypic conversion during long-term oxidative stress such as chronic inflammation.
Morphologic changes in NMuMG cells exposed to H2O2. A, NMuMG cells were either left untreated (left) or treated daily with 0.2 mmol/L H2O2 for 2 (middle) and 4 (right) days and observed under a phase-contrast microscope. B, The cells were treated as in A, and actin fibers stained with tetramethylrhodamine isothiocyanate-conjugated phalloidin were observed under a fluorescent microscope; bars, 100 μm. C, NMuMG cells were left untreated (left) or treated daily with 0.2 mmol/L H2O2 for 4 days (right), and the subcellular localization of E-cadherin was examined by indirect immunofluorescence labeling. D, The cells treated as in C for 2 and 4 days were lysed and subjected to immunoblot analysis with the antibody to E-cadherin.
Morphologic changes in NMuMG cells exposed to H2O2. A, NMuMG cells were either left untreated (left) or treated daily with 0.2 mmol/L H2O2 for 2 (middle) and 4 (right) days and observed under a phase-contrast microscope. B, The cells were treated as in A, and actin fibers stained with tetramethylrhodamine isothiocyanate-conjugated phalloidin were observed under a fluorescent microscope; bars, 100 μm. C, NMuMG cells were left untreated (left) or treated daily with 0.2 mmol/L H2O2 for 4 days (right), and the subcellular localization of E-cadherin was examined by indirect immunofluorescence labeling. D, The cells treated as in C for 2 and 4 days were lysed and subjected to immunoblot analysis with the antibody to E-cadherin.
Induction of a set of integrin family members under long-term exposure to H2O2. Total RNA was extracted from NMuMG cells left untreated (open bar) or treated with 0.2 mmol/L H2O2 for 2 (shaded bar) and 4 days (filled bar) and analyzed by the quantitative reverse transcription-PCR method as described in Materials and Methods. The amount of PCR product derived from each mRNA was normalized to that from GAPDH in the same sample and showed as a ratio to the untreated control. Bars and error bars represent the mean ± SD obtained from at least three independent experiments.
Induction of a set of integrin family members under long-term exposure to H2O2. Total RNA was extracted from NMuMG cells left untreated (open bar) or treated with 0.2 mmol/L H2O2 for 2 (shaded bar) and 4 days (filled bar) and analyzed by the quantitative reverse transcription-PCR method as described in Materials and Methods. The amount of PCR product derived from each mRNA was normalized to that from GAPDH in the same sample and showed as a ratio to the untreated control. Bars and error bars represent the mean ± SD obtained from at least three independent experiments.
Induction of MMPs on exposure to H2O2. A and B, After NMuMG cells were exposed daily to 0.2 mmol/L H2O2 for the indicated period, total RNA was extracted as in Fig. 2 and analyzed by Northern blot analysis (A, left) or quantitative reverse transcription-PCR (A, right). For Northern blot analysis, 20 μg of RNA were resolved on gel, transferred to a membrane, and hybridized with labeled cDNA probes. The membranes were washed and autoradiographed. GAPDH was used to monitor the amount of total RNA in each lane. Quantitative reverse transcription-PCR was performed as in Fig. 2. Open, shaded, and filled bars represent the results for untreated cells and cells treated with 0.2 mmol/L H2O2 for 2 and 4 days, respectively. B, NMuMG cells were treated once with H2O2 at the indicated doses, and 2 or 4 days later, total RNA was extracted and subjected to Northern blot analysis as in A. The level of radioactivity in each lane was measured with a BAS2000 Bioimage Analyzer (Fuji, Tokyo, Japan) and is shown relative to the value of the untreated control after normalization to the intensity of GAPDH. The closed circle, triangle, and rectangle indicate the results of 0.2, 0.4, and 0.6 mmol/L H2O2 treatment, respectively. C and D, NMuMG cells were treated daily with 0.2 mmol/L H2O2 for the indicated period, and at the end of the incubation, the medium was replaced with a serum-free one. Twenty-four hours after the replacement, the conditioned medium then was collected, concentrated, and subjected to Western blot analysis for MMP-13 (C) or gelatin zymographic analysis to detect gelatinase activity (D). The experiments were repeated several times and gave essentially the same results.
Induction of MMPs on exposure to H2O2. A and B, After NMuMG cells were exposed daily to 0.2 mmol/L H2O2 for the indicated period, total RNA was extracted as in Fig. 2 and analyzed by Northern blot analysis (A, left) or quantitative reverse transcription-PCR (A, right). For Northern blot analysis, 20 μg of RNA were resolved on gel, transferred to a membrane, and hybridized with labeled cDNA probes. The membranes were washed and autoradiographed. GAPDH was used to monitor the amount of total RNA in each lane. Quantitative reverse transcription-PCR was performed as in Fig. 2. Open, shaded, and filled bars represent the results for untreated cells and cells treated with 0.2 mmol/L H2O2 for 2 and 4 days, respectively. B, NMuMG cells were treated once with H2O2 at the indicated doses, and 2 or 4 days later, total RNA was extracted and subjected to Northern blot analysis as in A. The level of radioactivity in each lane was measured with a BAS2000 Bioimage Analyzer (Fuji, Tokyo, Japan) and is shown relative to the value of the untreated control after normalization to the intensity of GAPDH. The closed circle, triangle, and rectangle indicate the results of 0.2, 0.4, and 0.6 mmol/L H2O2 treatment, respectively. C and D, NMuMG cells were treated daily with 0.2 mmol/L H2O2 for the indicated period, and at the end of the incubation, the medium was replaced with a serum-free one. Twenty-four hours after the replacement, the conditioned medium then was collected, concentrated, and subjected to Western blot analysis for MMP-13 (C) or gelatin zymographic analysis to detect gelatinase activity (D). The experiments were repeated several times and gave essentially the same results.
Induction of MMPs by long-term exposure to H2O2 in human TIG-7 normal diploid fibroblasts. A, After TIG-7 cells were exposed daily to 0.2 mmol/L H2O2 for the indicated period, total RNA was extracted and analyzed by quantitative reverse transcription-PCR as in Fig. 3 A. Open, shaded, and filled bars represent the results for untreated cells and cells treated with 0.2 mmol/L H2O2 for 2 and 4 days, respectively. B, TIG-7 cells were treated with 0.2 mmol/L H2O2 as in A, and at the end of the incubation, the medium was replaced with a serum-free one. Twenty-four hours after the replacement, the conditioned medium then was collected, concentrated, and subjected to gelatin zymographic analysis to detect gelatinase activity. The experiments were repeated several times, and essentially the same results were obtained.
Induction of MMPs by long-term exposure to H2O2 in human TIG-7 normal diploid fibroblasts. A, After TIG-7 cells were exposed daily to 0.2 mmol/L H2O2 for the indicated period, total RNA was extracted and analyzed by quantitative reverse transcription-PCR as in Fig. 3 A. Open, shaded, and filled bars represent the results for untreated cells and cells treated with 0.2 mmol/L H2O2 for 2 and 4 days, respectively. B, TIG-7 cells were treated with 0.2 mmol/L H2O2 as in A, and at the end of the incubation, the medium was replaced with a serum-free one. Twenty-four hours after the replacement, the conditioned medium then was collected, concentrated, and subjected to gelatin zymographic analysis to detect gelatinase activity. The experiments were repeated several times, and essentially the same results were obtained.
Activation of MAP kinase and its involvement in MMP-13 induction under oxidative conditions. A and B, NMuMG cells were exposed daily to 0.2 mmol/L H2O2 for 2 or 4 days in the presence or absence of PD98059 (20 μmol/L) or SB203580 (10 μmol/L), and total RNA was extracted and analyzed by Northern blot analysis as in Fig. 3 A. The level of radioactivity in each lane was measured with a BAS2000 Bioimage Analyzer (Fuji) and is shown in B relative to the value of the untreated control after normalization to the intensity of GAPDH. Open bar represents the results for untreated sample, and shaded and filled bars note those for samples treated with 0.2 mmol/L H2O2 in the presence of PD98059 and SB203580, respectively. C and D, NMuMG cells were treated with 0.2 mmol/L H2O2 for 2 or 4 days, lysed in lysis buffer supplemented with 10 mmol/L PPI and 0.4 mmol/L sodium orthovanadate, and analyzed by immunoblot analysis with antibodies recognizing the phosphorylated form of ERK1/2 (active ERK1/2), that of p38 MAP kinase (active p38), and that of JNK (active JNK), or that of ERK1/2 protein (pan-ERK). The band intensity of pan-ERK indicates an almost equal loading of the proteins in each lane. The lysate of cells treated with anisomycin (50 μg/mL) for 30 minutes was used as a positive control (aniso) to detect activated p38 MAP kinase and JNK.
Activation of MAP kinase and its involvement in MMP-13 induction under oxidative conditions. A and B, NMuMG cells were exposed daily to 0.2 mmol/L H2O2 for 2 or 4 days in the presence or absence of PD98059 (20 μmol/L) or SB203580 (10 μmol/L), and total RNA was extracted and analyzed by Northern blot analysis as in Fig. 3 A. The level of radioactivity in each lane was measured with a BAS2000 Bioimage Analyzer (Fuji) and is shown in B relative to the value of the untreated control after normalization to the intensity of GAPDH. Open bar represents the results for untreated sample, and shaded and filled bars note those for samples treated with 0.2 mmol/L H2O2 in the presence of PD98059 and SB203580, respectively. C and D, NMuMG cells were treated with 0.2 mmol/L H2O2 for 2 or 4 days, lysed in lysis buffer supplemented with 10 mmol/L PPI and 0.4 mmol/L sodium orthovanadate, and analyzed by immunoblot analysis with antibodies recognizing the phosphorylated form of ERK1/2 (active ERK1/2), that of p38 MAP kinase (active p38), and that of JNK (active JNK), or that of ERK1/2 protein (pan-ERK). The band intensity of pan-ERK indicates an almost equal loading of the proteins in each lane. The lysate of cells treated with anisomycin (50 μg/mL) for 30 minutes was used as a positive control (aniso) to detect activated p38 MAP kinase and JNK.
Activation of Rac1 and increased potential for invasion under coculture with the fibroblasts during long-term exposure to H2O2. A, NMuMG cells were treated with H2O2 as above. The active forms of Rac1, Cdc42 and RhoA were precipitated with GST-CRIB (Rac1, Cdc42) and GST-Rhotekin (RhoA) and analyzed by immunoblot analysis with antibodies to each small GTPase. In parallel, GST-bound proteins (GST) and whole cell lysate (lysate) were immunoblotted. The intensity of the band was quantified with NIH image, and a ratio of the active form in total cell lysate relative to that of the untreated control was plotted. The values are the average ± SD obtained from three independent experiments. Straight and dotted lines indicate Rac1 and Cdc42 activity, respectively. B–D, TIG-7 cells were grown at the bottom of an invasion chamber to confluence and treated with 0.2 mmol/L H2O2 for 2 days. Then, 2 × 105 NMuMG cells were seeded on the upper layer and incubated with (H2O2 treatment) or without (No treatment) 0.2 mmol/L H2O2 in the presence (+MMP inhibitor) or absence of 10 μmol/L Galargin. After 3 days, following the removal of the noninvaded cells from the upper surface of the Matrigel with a cotton swab, the invaded cells were fixed, stained with crystal violet, and observed microscopically (B). Thereafter, the stained cells were lysed with 10% SDS, and absorbance at 595 nm was measured (C). The values are average ± SD from more than three independent experiments. The significance of the differences was assessed with a t test, and asterisks mean that P values were <0.05.
Activation of Rac1 and increased potential for invasion under coculture with the fibroblasts during long-term exposure to H2O2. A, NMuMG cells were treated with H2O2 as above. The active forms of Rac1, Cdc42 and RhoA were precipitated with GST-CRIB (Rac1, Cdc42) and GST-Rhotekin (RhoA) and analyzed by immunoblot analysis with antibodies to each small GTPase. In parallel, GST-bound proteins (GST) and whole cell lysate (lysate) were immunoblotted. The intensity of the band was quantified with NIH image, and a ratio of the active form in total cell lysate relative to that of the untreated control was plotted. The values are the average ± SD obtained from three independent experiments. Straight and dotted lines indicate Rac1 and Cdc42 activity, respectively. B–D, TIG-7 cells were grown at the bottom of an invasion chamber to confluence and treated with 0.2 mmol/L H2O2 for 2 days. Then, 2 × 105 NMuMG cells were seeded on the upper layer and incubated with (H2O2 treatment) or without (No treatment) 0.2 mmol/L H2O2 in the presence (+MMP inhibitor) or absence of 10 μmol/L Galargin. After 3 days, following the removal of the noninvaded cells from the upper surface of the Matrigel with a cotton swab, the invaded cells were fixed, stained with crystal violet, and observed microscopically (B). Thereafter, the stained cells were lysed with 10% SDS, and absorbance at 595 nm was measured (C). The values are average ± SD from more than three independent experiments. The significance of the differences was assessed with a t test, and asterisks mean that P values were <0.05.
Grant support: Grants-in-aid for Scientific Research, a grant-in-aid for Cancer Research, and the High-Technology Research Center Project from the Ministry for Education, Science, Sports, and Culture of Japan.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Kiyoshi Nose, Department of Microbiology, Showa University School of Pharmaceutical Sciences, Hatanodai 1-5-8, Shinagawa-ku, Tokyo, Japan 142-8555. Phone: 81-3-3784-8208; Fax: 81-3-3784-6850; E-mail: [email protected]
Unpublished data.
Acknowledgments
We thank Dr. Seiki (Institute of Medical Science, University of Tokyo) and Dr. Miyaura (Tokyo University of Pharmacy and Life Science) for donating cDNAs for MMP-2, and MMP-9 and -13, respectively, and Dr. Narumiya (Kyoto University) for GST-Rhotekin construct.