Research ArticleBiochemistry

Novel p47phox-Related Organizers Regulate Localized NADPH Oxidase 1 (Nox1) Activity

See allHide authors and affiliations

Science Signaling  15 Sep 2009:
Vol. 2, Issue 88, pp. ra54
DOI: 10.1126/scisignal.2000370

Abstract

The mechanisms that determine localized formation of reactive oxygen species (ROS) through NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase (Nox) family members in nonphagocytic cells are unknown. We show that the c-Src substrate proteins Tks4 (tyrosine kinase substrate with four SH3 domains) and Tks5 are functional members of a p47phox-related organizer superfamily. Tks proteins selectively support Nox1 and Nox3 (and not Nox2 and Nox4) activity in reconstituted cellular systems and interact with the NoxA1 activator protein through an Src homology 3 domain–mediated interaction. Endogenous Tks4 is required for Rac guanosine triphosphatase– and Nox1-dependent ROS production by DLD1 colon cancer cells. Our results are consistent with the Tks-mediated recruitment of Nox1 to invadopodia that form in DLD1 cells in a Tks- and Nox-dependent fashion. We propose that Tks organizers represent previously unrecognized members of an organizer superfamily that link Nox to localized ROS formation.

Introduction

The NADPH oxidase (Nox) family, which consists of the homologous enzymes Nox1 to Nox4 and the more distantly related Nox5, Duox1, and Duox2, catalyzes the regulated formation of reactive oxygen species (ROS) (14). These homologs of the phagocyte gp91 cytochrome b subunit are found in virtually all tissues and have been implicated in numerous biological processes, including cell growth, apoptosis and cancer, angiogenesis and blood pressure regulation, innate immunity and inflammation, cell signaling, cell motility, and transcription. ROS generated through Nox also contribute to a number of diseases, including atherosclerosis, hypertension, arthritis, Alzheimer’s disease and other neurological disorders, stroke, respiratory syndromes, cancer, and inflammation (5, 6). Many of these activities appear to require the compartmentalized or spatially regulated formation of ROS (7). Very little is known about how the localized formation of ROS by Nox proteins is regulated under normal, much less pathological, conditions.

A subfamily of Nox proteins, Nox1 to 3, are regulated by cytosolic cofactors, including the Rac1 or Rac2 guanosine triphosphatases (GTPases) (811) and an “activator” protein (either p67phox or NoxA1) (1, 12). In addition, although not absolutely required for Nox activity, various “organizer” proteins have been identified, including p47phox, p40phox, and NoxO1 (1, 12). These organizer proteins make up a structurally similar “p47phox organizer superfamily,” characterized by an N-terminal phox (PX) domain that binds anionic phospholipids, and Src homology 3 (SH3) and proline-rich region (PRR) protein-protein interaction domains (1316). The organizers serve as regulatory response elements for Nox assembly in response to various signaling pathways. In the phagocyte Nox2 system, p47phox is phosphorylated in response to inflammatory stimuli, thereby promoting translocation of the cytosolic oxidase components to the plasma membrane and assembly of the functional Nox system (1316). Similarly, both p40phox and p47phox participate in the recruitment and assembly of the Nox2 system on phagocytic vesicles during internalization of pathogens and other particulate stimuli (1719). Such examples suggest that organizer subunits could contribute to the differential recruitment of functional Nox enzymes to specific subcellular compartments. These Nox regulators have distinct PX domains that bind to specific acidic phospholipids present in distinct membranous compartments (20). In addition, they contain protein interaction domains (SH3 and PRR) that may facilitate recruitment to various compartments. Accumulating evidence indicates that organizer proteins act after assembly of the NADPH complex to promote full Nox activity (2123).

Tks5 (tyrosine kinase substrate with five SH3 domains) was identified in a complementary DNA (cDNA) library screen for c-Src substrates (24). Subsequently, a close ortholog with four SH3 domains, called Tks4, was described (25, 26). The Tks proteins, which have a similar domain architecture and composition as that of members of the p47 organizer superfamily (26, 27), are widely distributed in mouse and human tissues, with the notable exception of low abundance in neutrophils. In Src-transformed cells, Tks4 and Tks5 localize to invadopodia (28, 29), dynamic phosphotyrosine-rich structures with an actin core, and abundant actin regulatory proteins (such as cortactin) capable of proteolytically degrading the extracellular matrix (ECM). Invadopodia are found in many metastatic cancer cells. Some invasive nonmalignant cells, such as macrophages, osteoclasts, endothelial, and vascular smooth muscle cells, contain related structures called podosomes (30, 31). c-Src has been implicated in invadopodia and podosome formation (32, 33) and many obligate invadopodia and podosome components, including cortactin, Tks4, and Tks5 (28, 29, 34) are Src substrates. Consistent with this, Tks4 and Tks5 are required for the formation of invadopodia and promote cancer cell invasion (29, 35, 36). However, the signals regulating the formation of invadopodia remain unclear.

The accompanying paper by Diaz et al. (37) shows that Nox formation of Ros is critical to the formation and stability of Tks-dependent invadopodia. Here, we show that Tks4 and Tks5 are previously unrecognized members of the p47phox superfamily of Nox organizers. Furthermore, our results suggest that the interaction of Nox with distinct organizers might be a key element in localizing ROS formation to different subcellular compartments to selectively affect cellular function.

Results

Tks4 and Tks5 are Nox organizers

The c-Src substrates Tks4 and Tks5 have a domain architecture and composition similar to that of known Nox organizer proteins (Fig. 1A). In addition to containing two N-terminal SH3 domains structurally similar to those of p47phox (up to 47% identity) and arranged in the same orientation, Tks proteins contain a phosphoinositol lipid-binding PX domain that is present and highly conserved among the members of the p47 organizer superfamily (26, 27). p47phox, p40phox, and NoxO1 serve as regulatory organizer subunits for the assembly and full activity of various Nox family members, including Nox2 (p47phox and p40phox) (16, 38, 39) and Nox1 and Nox3 (NoxO1 and p47phox) (1, 12, 40, 41). To assess the ability of Tks4 and Tks5 to serve as organizer subunits in Nox1-mediated ROS formation, we transfected human embryonic kidney (HEK) 293 cells with expression vectors for various organizer subunits along with other components (Nox1, NoxA1, RacQL = constitutively active Rac1Q61L, in which glutamine 61 of Rac1 is mutated to leucine) required for Nox1-dependent ROS generation. After 24 hours, ROS formation was measured by luminol-based chemiluminescence (CL) assay (42). Both Tks4 and Tks5 supported more Nox1-dependent ROS production than that found in mock-transfected cells or cells transfected with Nox1 system components in the absence of organizer subunit (Fig. 1B). All transfected proteins were present in similar abundance (fig. S1A).

Fig. 1

Tks4 and Tks5 are members of the p47 organizer superfamily and support Nox1-dependent ROS generation in a reconstituted HEK293 cell system. (A) Schematic diagram of the members of the p47phox organizer superfamily. White squares indicate PX domains; white circles indicate SH3 domains. (B) Time course of ROS generation showing that Tks4 and Tks5 support Nox1-dependent ROS formation in a reconstituted HEK293 system. HEK293 cells were transfected as indicated with expression vectors for Nox1, NoxA1, and RacQL, and with different organizers, including NoxO1, Tks4, and Tks5, or empty vector. After 24 hours, ROS production (measured as relative light units, RLU) was determined with a luminol-based CL assay continuously for 30 min. One representative experiment of three is shown. All transfected proteins were present in similar abundance (fig. S1).

Tks4 and Tks5 support Nox1- and Nox3-dependent ROS generation

We next investigated whether Tks4 and Tks5 could support ROS generation by other members of the Nox family and compared their ability to do so to that of known organizer subunits (NoxO1, p47phox, and p40phox). We transfected HEK293 cells with expression vectors encoding different organizer subunits and with all the known components required for Nox1-, Nox2-, Nox3-, and Nox4-dependent ROS generation. All transfected proteins were present in similar abundance (fig. S1, A to D). Tks4 and Tks5 selectively supported Nox1- and Nox3-dependent ROS generation (Fig. 2A) to a degree similar to that seen with NoxO1 or p47phox. Tks4 and Tks5 did not promote Nox2-dependent ROS formation and, like NoxO1 and p47phox, did not enhance Nox4-dependent ROS production. In contrast, Nox2-dependent ROS production was supported by p47phox and NoxO1. Nox4-dependent ROS production was already high; indeed, this Nox is constitutively active in the absence of regulatory subunits (1).

Fig. 2

Tks4 and Tks5 selectively support Nox1- and Nox3-dependent ROS generation in a Rac GTPase–dependent manner. (A) CL assay to determine the ability of Tks4 and Tks5 to support ROS generation by different members of the Nox family. HEK293 cells were transfected with expression vectors for the indicated organizer subunits and with the known components required for Nox1 (Nox1, NoxA1, RacQL)–, Nox2 (Nox2, p67phox, RacQL)–, Nox3 (Nox3, NoxA1, RacQL)–, and Nox4 (Nox4)-dependent ROS generation. After 24 hours, ROS formation was monitored by CL assay, as described in Materials and Methods. One representative experiment of three is shown, and results are given as mean of triplicates ±SD. *P < 0.001 compared to the condition with no organizers. (B) Treatment with the flavoenzyme and Nox inhibitor DPI blocks Tks4- and Tks5-mediated ROS generation by Nox1 in the HEK293 reconstituted system. HEK293 cells were transfected with expression vectors for different Myc-tagged organizer subunits as indicated, and with all protein components required for Nox1 activity. Twenty-four hours later, cells were treated with 10 μM DPI or DMSO control for 30 min and ROS production was measured by CL assay (upper panel). Comparable abundance of Myc-tagged proteins was verified by Western blot (lower panel). One representative experiment of three is shown, and results of CL assay are given as mean of triplicates ±SD. *P < 0.001. (C) Tks4- and Tks5-mediated ROS formation by Nox1 is Rac dependent in the reconstituted HEK293 cell system, as shown with dominant-negative Rac1T17N, the non–Rac-binding R103E NoxA1 mutant, and the non–Rac-binding Nox1 K421A, Y425A, K426E triple mutant (Nox1TM). HEK293 cells were transfected as indicated and after 24 hours ROS production was monitored by CL assay (upper panels). The expression of Myc-tagged proteins was verified by Western blot (lower panels). One representative experiment of three is shown, and results of CL assays are given as mean of triplicates ±SD. *P < 0.001 compared to the paired conditions.

Treatment with the Nox (and flavoenzyme) inhibitor (43) diphenyleneiodonium (DPI) efficiently blocked ROS generation induced by Tks4 and Tks5 (as well as NoxO1) (Fig. 2B, upper panel) without affecting the abundance of the transfected proteins (lower panel), indicating that the Tks protein–mediated ROS production observed was likely due to Nox activity. Rac1 is required along with NoxO1 and NoxA1 for full Nox1 activity (9). To test whether Tks4- and Tks5-mediated ROS generation by Nox1 was Rac dependent, we transfected HEK293 cells with expression vectors encoding Tks4, Tks5 or NoxO1, constitutively active Rac1 (RacQL) or dominant-negative RacN17 (= Rac1-T17N, in which Rac 1 threonine 17 is mutated to asparagine), and the Nox1 system components Nox1 and NoxA1. We found that Tks4- and Tks5-mediated ROS formation was effectively blocked by dominant-negative RacN17 (Fig. 2C, upper left panel). We also monitored ROS generation in HEK293 cells reconstituted with Tks4, Tks5, or NoxO1, with Nox1 and Rac1-Q61L, and with NoxA1 wild type or NoxA1 R103E, a mutant form of NoxA1 unable to bind Rac (44). We found that NoxA1 R103E abrogated Tks4- or Tks5-dependent induction of Nox1-dependent superoxide formation (Fig. 2C, upper middle panel). Finally, we transfected HEK293 cells with expression vectors encoding various organizer subunits, NoxA1 and Rac1-Q61L, and Nox1 wild type or Nox1 TM, a mutant form of Nox1 that is unresponsive to Rac1 because of a triple mutation in its Rac binding site (45), and observed that Nox1 TM abolished Tks4- and Tks5-mediated ROS generation (Fig. 2C, upper right panel). We could detect no effect on the abundance of the other transfected proteins with any of these experimental manipulations (Fig. 2C, lower panels).

Tks4 and Tks5 bind NoxA1 through their SH3 domains in a Rac-independent manner

NoxO1 is required for full Nox1 and Nox3 oxidase activity at least partially because of its role in the plasma membrane recruitment of the NoxA1 activator protein. To determine whether the ability of Tks4 and Tks5 to promote Nox1-dependent ROS formation involved interaction with NoxA1, we performed coimmunoprecipitation experiments in either HEK293 cells reconstituted with the other components of the Nox1 system (Nox1, NoxA1, and RacQL) or in human DLD1 colon cancer cells that endogenously express Nox1, NoxA1, Rac1, and Tks4, but not NoxO1 (fig. S2, A to C). In reconstituted HEK293 cells, Tks5 bound NoxA1 to a similar extent as did NoxO1 (Fig. 3A, left panel). We also detected an interaction between NoxA1 and Tks4, although to a lesser extent. We confirmed interaction of NoxA1 with endogenous Tks4 by coimmunoprecipitation analysis of human DLD1 cells (Fig. 3A, right panel).

Fig. 3

Support of Nox1-dependent ROS generation by Tks4 and Tks5 requires the SH3 domain–mediated binding of NoxA1 and is independent of Rac. (A) Coimmunoprecipitation analysis indicates that Tks4 and Tks5 interact with NoxA1 in reconstituted HEK293 cells and with endogenous NoxA1 in DLD1 cells. In the left panel, HEK293 cells were transfected as indicated with Myc-tagged organizer subunits and Flag-tagged NoxA1. After 24 hours, cells were lysed and immunoprecipitation (IP) was carried out (see Materials and Methods) with antibody against Myc. Interaction of Myc-tagged adaptors with NoxA1 was analyzed with NoxA1-specific antibody, whereas protein abundance and equal loading of proteins on gels was verified by reblotting the membranes with antibodies against Myc and actin, respectively. One representative experiment of three is shown. In the right panel, DLD1 cell lysates were immunoprecipitated with NoxA1-specific antibody or preimmune serum. An additional control with protein G–Sepharose beads was performed. Specific interaction between endogenous Tks4 and NoxA1 was confirmed by Tks4-specific antibody, whereas NoxA1 in cell lysates was verified by reblotting (IB) the membrane with NoxA1-specific antibody. One representative experiment of three is shown. (B) Coimmunoprecipitation analysis indicates that interaction between NoxA1 and Tks5 does not depend on the GTP-bound state of Rac1. HEK293 cells were transfected as indicated with Myc-tagged Tks5, Flag-tagged NoxA1, and alternatively with GFP-tagged active Rac1-Q61L or inactive Rac1-T17N. After 24 hours, cells were lysed, and immunoprecipitation was carried out with an antibody against Flag. Interaction of Tks5 with NoxA1 was analyzed with a Tks5-specific antibody, whereas protein abundance in cell lysates and similar amount of immunoprecipitated NoxA1 protein was verified by reblotting the membranes with antibodies against NoxA1 and GFP. One representative experiment of three is shown. (C) Tks5 binds to NoxA1 through its SH3 domains. HEK293 cells were transfected with expression vectors for Nox1, NoxA1, and RacQL and with the Myc-tagged organizer subunits NoxO1, Tks5 wild-type (Tks5WT), or Tks5 M1M5 mutant in which point mutations were made to eliminate binding capacity of all of the SH3 domains. After 24 hours, ROS generation was monitored by CL assay (left panel) and abundance of transfected proteins by Western blot with Tks5 antibody (right panel). Inability of the Tks5 M1M5 mutant to bind NoxA1 was shown by coimmunoprecipitation analysis by Flag antibody for immunoprecipitation and Tks5 antibody to detect specific interaction (lower panel). One representative experiment of three is shown, and results of CL assays are given as mean of triplicates ±SD. *P < 0.001. (D) Deletion of the PX domain of Tks5 prevents activation of Nox1-mediated ROS formation but does not abrogate Tks5 binding to NoxA1. HEK293 cells were transfected with expression vectors for the Myc-tagged organizer subunits NoxO1, Tks5WT, and Tks5ΔPX, along with Flag-tagged NoxA1. After 24 hours, ROS generation was measured by CL assay (left panel), whereas expression of Myc-tagged organizers was verified by Western blot (right panel). The ability of Tks5ΔPX to bind NoxA1 was determined by coimmunoprecipitation analysis (lower panel) with Flag antibody for immunoprecipitation and Tks5 antibody to detect interaction with NoxA1. One representative experiment of three is shown, and results of CL assays are given as mean of triplicates ±SD. *P < 0.001.

The binding of other organizer subunits to activator proteins is independent of the guanine nucleotide state of Rac GTPase (44). Coimmunoprecipitation experiments in HEK293 cells reconstituted with all of the components of the Nox1 pathway and with constitutively active RacQ61L or with inactive RacT17N revealed that neither RacQ61L nor RacT17N affected the interaction between NoxA1 and Tks5 (Fig. 3B).

To identify the structural basis for the interaction of Tks proteins with NoxA1, we generated a Myc-tagged Tks5 expression construct, Tks5M1M5, in which point mutations were introduced into all five SH3 domains to disrupt their function, and Tks5ΔPX, in which the PX domain was deleted. These Tks5 mutants were assayed for their ability to support Nox1-mediated ROS generation in HEK293 cells reconstituted with all components of Nox1 pathway. Tks5M1M5 failed to support ROS production by Nox1 (Fig. 3C, upper panel) and also failed to coimmunoprecipitate with NoxA1 (lower panel). This indicates that the interaction between NoxA1 and Tks5 is mediated through one or more of the Tks5 SH3 domains. We were unable to identify a single SH3 domain of Tks5 responsible for interaction with NoxA1: Single SH3 domain point mutants (Tks5M1 to M5) and a double mutant (Tks5M1M2) supported ROS generation and interacted with NoxA1 as well as Tks5 wild type (fig. S3, A to C).

The Tks5 PX domain is required for Tks5 localization to invadopodia in Src-transformed fibroblasts because of its ability to bind PI(3,4)P2 (phosphatidylinositol 3,4-bisphosphate) (28, 46). Tks5ΔPX failed to support ROS generation by Nox1 in reconstituted HEK293 cells when present in abundance comparable to that of wild-type Tks5 (Fig. 3D, right panel). However, Tks5ΔPX did coimmunoprecipitate with NoxA1 (Fig. 3D, lower panel). Thus, the Tks PX domain may contribute to the membrane recruitment of the functional Nox1 complex.

Human DLD1 colon cancer cells produce ROS and degrade the ECM in a Tks4-dependent manner

Human DLD1 colon cancer cells have increased Src activity compared to other colon cancer cells, and this correlates with their enhanced ability to produce ROS (42). DLD1 cells only express Nox1, NoxA1, Tks4, and small amounts of Tks5, with no other known organizer proteins (fig. S2, A to C); thus, these cells provide a suitable system for studying Tks-dependent, Nox1-mediated ROS generation. The DLD1 cells were transfected with control siRNA (small interfering RNA) or with increasing concentrations of a Tks4-specific siRNA mixture (SmartPool, Dharmacon) for 72 hours, a time sufficient to decrease Tks4 protein abundance (fig. S4A), and ROS production was determined by CL assay. Tks4-specific siRNA efficiently reduced Tks4 protein abundance (Fig. 4A, upper right panel) and decreased ROS generation in a concentration-dependent manner (upper left panel). The loss of ROS formation observed with siRNA-mediated depletion of Tks4 in DLD1 cells was rescued by overexpression of the siRNA-resistant mouse Tks4 (mTks4) (Fig. 4A, lower panels). To confirm that depletion of Tks4 protein blocks Nox1-mediated, Rac1-dependent ROS generation in DLD1 cells, we transfected cells with Rac1-Q61L and with control or Tks4-specific siRNA. Rac1-Q61L caused a modest, but consistent, increase in ROS production over baseline (Fig. 4B, left panel). Tks4-specific siRNA reduced the Rac1-Q61L–induced ROS generation to that of the mock-transfected control, a reduction that could be partially rescued by expression of mTks4. Western analysis (Fig. 4B, right panel) verified Tks4 protein knockdown and mTks4 overexpression and confirmed the similar abundance of the green fluorescent protein (GFP)–tagged Rac1-Q61L in each condition. Additionally, overexpression of either Tks4 or Tks5 (as well as of NoxO1) increased ROS production to a similar degree (twice control, Fig. 4C, left) when the overexpressed proteins were present in similar abundance (right panel).

Fig. 4

Human DLD1 colon cancer cells produce ROS and degrade the ECM in a Tks4-dependent manner. (A) Tks4 knockdown by siRNA reduces ROS generation in a concentration-dependent fashion and this effect can be rescued by a siRNA-insensitive mouse Tks4. DLD1 cells were transfected with different concentrations of Tks4-specific siRNA mixture or control siRNA, and after 72 hours, ROS generation was determined by CL assay (upper left panel), whereas Tks4 protein knockdown was checked by Western blot with Tks4 antibody (upper right panel). DLD1 cells were transfected with 20 nM Tks4-specific siRNA mixture or control siRNA and with mTks4 expression vector or empty vector (mock). After 72 hours, ROS generation was measured by CL assay (lower left panel), whereas Tks4 knockdown and mTks4 expression were checked by Western blot with antibody against Tks4 (lower right panel). One representative experiment of three is shown and results of CL assays are given as mean of triplicates ±SD. *P < 0.001 compared to control siRNA-transfected cells. (B) Tks4 knockdown by siRNA reduces Rac1-induced ROS generation and this effect can be rescued by mTks4. DLD1 cells were transfected with Tks4-specific siRNA or control siRNA and with expression vector for GFP-tagged RacQL or empty vector. After 72 hours, ROS generation was monitored by CL assay (left panel). Tks4 protein knockdown and mTks4 expression, as well as comparable abundance of GFP-tagged RacQL, were shown by Western blot with Tks4- or GFP-specific antibody, respectively (right panel). One representative experiment of three is shown and results of CL assays are given as mean of triplicates ±SD. *P < 0.01. (C) Tks4 and Tks5 overexpression in DLD1 cells induces ROS generation to a similar extent as NoxO1 overexpression (~twice control ROS production). DLD1 cells were transfected with expression vectors for Myc-tagged organizer subunits (NoxO1, Tks4, or Tks5) or with empty vector. After 24 hours, ROS formation was measured by CL assay (left panel), whereas abundance of the Myc-tagged adaptors was verified by Western blot with antibody against Myc (right panel). One representative experiment of three is shown and results of CL assays are given as mean of triplicates ±SD. *P < 0.001 compared to mock. (D) Tks4 knockdown reduces the ability of DLD1 cells to degrade the ECM. DLD1 cells were transfected with Tks4-specific or control siRNA and, after 24 hours, plated on FITC-labeled, gelatin-coated coverslips. After 20 hours, cells were fixed in 4% PFA, stained with Alexa Fluor 568 phalloidin as described in Materials and Methods, and visualized by epifluorescence microscopy (40×). White arrows indicate areas in which F-actin–positive structures (red) colocalize with areas of ECM degradation. Scale bars, 45 μm. One representative picture from two separate experiments is shown.

Tks4 has been reported to be required for cancer cells to degrade the ECM (29). We first established the ability of DLD1 cells plated on fluorescein isothiocyanate (FITC)-labeled gelatin-coated coverslips to degrade the ECM. The siRNA-mediated knockdown of Tks4 reduced the ability of the DLD1 cells to induce pericellular proteolysis compared to control siRNA-transfected cells (Fig. 4D; quantified in fig. S4B), indicating that Tks4 plays a key role in enabling DLD1 cells to degrade ECM.

Nox1 localizes to invadopodia in DLD1 cells

Several types of invasive human cancer cells, including breast cancers and melanomas, form invadopodia. We have shown that human DLD1 colon cancer cells degrade the ECM in a Tks4-dependent fashion (Fig. 4D). The accompanying paper by Diaz et al. (37) shows that the formation of stable invadopodia and ECM degradation requires ROS production. We have found that Nox1-specific siRNAs (SmartPool, Dharmacon) reduce ROS generation, invadopodia formation, and ECM degradation in DLD1 cells and that this effect can be partially rescued by the overexpression of mNox1 (fig. S5, A to C). Thus, we hypothesized that Tks4 might localize Nox1 to invadopodia in DLD1 cells. Given the lack of good working antibody for the detection of endogenous Nox1, we tested this hypothesis by determining the localization of transfected GFP-tagged Nox1 [or GFP-empty vector control (GFP-EV)] in DLD1 cells by confocal microscopy. In comparing the localization of Nox1-GFP (Fig. 5, A and B, lower left panel) with that of known markers for invadopodia such as F-actin (Fig. 5A, middle panels) and cortactin (Fig. 5B, middle panels), we found that most of the Nox1-GFP localized to invadopodia, where it colocalized with F-actin (Fig. 5A, merge) and cortactin (Fig. 5B, merge). In contrast, no colocalization was observed in GFP-EV–transfected cells. Z-stack images with 0.2-μm steps of the merged images in Fig. 5A and B showed that both Nox1 and F-actin–positive and Nox1 and cortactin-positive structures are ventral organelles, as previously reported for invadopodia (34) (fig. S6A).

Fig. 5

Nox1 localizes to ECM-degrading invadopodia in DLD1 cells. (A and B) Nox1 localizes to F-actin– and cortactin-rich structures in DLD1 cells. DLD1 cells were plated on glass coverslips and after 24 hours cells were transfected with SrcYF and GFP-Nox1 or GFP empty vector (GFP-EV). After 48 hours, cells were fixed in 4% PFA and stained with Alexa Fluor 568 phalloidin (A) or cortactin antibody, followed by Alexa Fluor 568–conjugated secondary antibody (B) and visualized by confocal microscopy (100×). White arrows indicate areas in which Nox1 (lower left panels) and F-actin in (A) or cortactin in (B) (middle panels) colocalize in structures identified as invadopodia (merge in yellow). Scale bars, 5 μm. One representative picture of three separate experiments is shown. (C and D) Nox1 localizes to cortactin-rich structures capable of degrading the ECM in DLD1 cells. DLD1 cells were transfected with SrcYF and with RFP-tagged Nox1 or RFP empty vector in (C) and with GFP-tagged Nox1 or GFP empty vector in (D). After 24 hours, cells were trypsinized and plated on FITC-labeled gelatin-coated coverslips. Forty-eight hours later, cells were fixed in 4% PFA in (C) or methanol in (D) and stained with mouse antibody against cortactin, followed by anti-mouse Alexa Fluor 647–conjugated secondary antibody in (C) and (D) and with rabbit polyclonal antibody against GFP, followed by anti-rabbit Alexa Fluor 568–conjugated secondary antibody in (D). The cells were visualized by confocal microscopy (60×). White arrows indicate areas in which Nox1 (red) and cortactin (green) colocalize in invadopodia (appear yellow because of overlap in the merge) capable of degrading the ECM (in blue). Scale bars, 4 μm. One representative image from three separate experiments is shown.

To confirm Nox1 localization to ECM-degrading invadopodia in DLD1 cells, we examined the localization of transfected red fluorescent protein (RFP)-tagged Nox1 [or RFP–empty vector control (RFP-EV)] in constitutively active SrcYF-overexpressing DLD1 cells plated on FITC-labeled, gelatin-coated coverslips. We found that Nox1-RFP colocalized with cortactin to form Nox1-cortactin structures, which efficiently degraded the ECM (Fig. 5C), a hallmark of invadopodia function. Similar results were obtained when SrcYF-overexpressing DLD1 cells were transfected with Nox1-GFP and plated on FITC-labeled, gelatin-coated coverslips (Fig. 5D). Staining for cortactin revealed partial colocalization of Nox1-GFP with cortactin in Nox1-GFP-cortactin structures that degraded ECM.

NoxO1 overexpression in DLD1 cells reduces invadopodia formation and ECM degradation

We have established that Nox1 localizes to ECM-degrading invadopodia that form in a Tks4-dependent manner in DLD1 cells, and that Tks4 recruits NoxA1 to this complex. We hypothesized that a consequence of the NoxA1-Tks4 interaction is the local generation of ROS by Nox1 to support the formation of invadopodia. We reasoned that a different organizer subunit, such as NoxO1, might compete with endogenous Tks4, displacing Nox1 from the invadopodia and thereby reducing their ROS-dependent formation. We therefore examined the effect of NoxO1 overexpression on Nox1 localization, invadopodia formation, and localized ROS production. We transfected SrcYF-overexpressing DLD1 cells with Nox1-GFP (onefold) and increasing amounts of NoxO1 (one- to threefold) or equal amounts of empty vector. After 48 hours, we fixed the cells and stained them for invadopodia markers, including F-actin (Fig. 6A, middle panels) and cortactin (Fig. 6B, middle panels), and compared their localization with that of Nox1GFP (Fig. 6, A and B, left panels). We found that NoxO1 overexpression reduced Nox1 colocalization with invadopodia and the overall formation of invadopodia (Fig. 6, A, B, and E). Concomitant with this, NoxO1-GFP no longer localized to invadopodia, and invadopodia formation was reduced in NoxO1-GFP–positive cells (fig. S7A).

Fig. 6

The overexpression of NoxO1 in DLD1 cells reduces invadopodia formation and ECM degradation. (A and B) NoxO1 overexpression reduces invadopodia formation in a concentration-dependent manner in DLD1 cells. DLD1 cells were transfected with SrcYF, GFP-tagged Nox1, and with increasing amounts of NoxO1 (one- to threefold) or with equal amount of empty vector (mock). After 48 hours, cells were fixed in 4% PFA, stained with Alexa Fluor 568 phalloidin in (A) or antibody followed by Alexa Fluor 568–conjugated secondary antibody in (B) and visualized by confocal microscopy (100×). In the upper panels, the white arrows indicate areas in which Nox1 (left panels, red in merge image) and F-actin in (A) or cortactin in (B) (middle panels, green in merge image) colocalize in invadopodia (merge in yellow). Scale bars, 5 μm. One representative image from three separate experiments is shown. (C) Overexpression of NoxO1 in DLD1 cells reduces ECM degradation. DLD1 cells were transfected with SrcYF, RFP-tagged Nox1, and NoxO1 (threefold) or with equal amount of empty vector (mock). Twenty-four hours later, cells were trypsinized and plated on FITC-labeled, gelatin-coated coverslips, and after 48 hours the cells were fixed in 4% PFA and visualized by epifluorescence microscopy (40×). In the left panel, the white arrows indicate areas in which RFP-tagged Nox1-expressing cells (in red) degrade the ECM (in green). Scale bars, 45 μm. One representative image from three separate experiments is shown. (D) Overexpression of NoxO1 in DLD1 cells reduces the number of ROS-positive invadopodia-like structures. DLD1 cells were transfected with SrcYF and with NoxO1 or empty vector (mock). Seventy-two hours later, cells were incubated with 2.5 μM ROS-sensitive probe PY1-AM in HBSS for 30 min at 37°C (see Materials and Methods) and visualized by confocal microscopy (100×). One representative image from three separate experiments is shown. Scale bars, 5 μm. (E) Quantifications of experiments illustrated in (A) to (D). In the first two panels, quantification from three independent biological experiments shown in (A) and (B) is given: The number of Nox1/phalloidin-positive structures in (A) or Nox1/cortactin-positive structures in (B) was counted and averaged from 25 cells for each experiment. Error bars represent SEM. *P < 0.008; **P < 0.002. In the third panel, quantification from three independent biological experiments as shown in (C): for each experiment, the total degradation area was obtained as sum of degradation areas calculated with Metamorph software from 25 images and reported as percentage (mock set as 100%). In the graph, error bars represent SEM. *P < 0.05 (Mann-Whitney U test). In the right panel, quantification from three independent biological experiments, as shown in (D): the number of ROS-positive structures was counted and averaged from 10 pictures for each experiment. Error bars represent SEM. *P < 0.005.

NoxO1 overexpression was accompanied by loss of the ability of the Nox1-RFP expressing cells to degrade the ECM (Fig. 6, C and E, third panel). Consistent with an antagonistic function of NoxO1 on ROS-dependent formation of invadopodia, the presence of NoxO1 in SrcYF-overexpressing DLD1 cells incubated with the ROS-specific probe PY1-AM (see fig. S7B and Materials and Methods) reduced the formation of ROS-positive invadopodia-like structures (Fig. 6, D and E, rightmost panel). These experiments strengthen the hypothesis that localized Tks-mediated ROS generation by Nox1 in DLD1 cells contributes to invadopodia formation and function.

Discussion

Tks4 and Tks5 are members of a p47 organizer superfamily that direct localized ROS generation by Nox1

The Nox family, consisting of Nox1 to 5 and Duox1 and 2, catalyzes the regulated formation of ROS. Although excessive amounts of ROS are toxic, ROS also function physiologically as signaling molecules to mediate various responses (1). The signaling properties of ROS result largely from their ability to catalyze the reversible oxidation of redox-sensitive target proteins. For example, the activities of several protein tyrosine phosphatases and receptor tyrosine kinases (47) and the cofilin phosphatase Slingshot (48) are modulated by a reactive cysteine at the active site that is susceptible to reversible oxidation by hydrogen peroxide.

Because ROS are diffusible and short-lived, the localization of ROS production at precise subcellular compartments may be essential for redox-mediated signaling events (49). In particular, emerging evidence that ROS regulate the dynamic behavior of the actin cytoskeleton during such highly spatially and temporally regulated processes as cell motility (5055), angiogenesis (56), and neuronal extension (57) suggests that the localized formation of ROS is of critical importance. The Nox enzymes exist in specific subcellular localizations, making them ideal candidates for localized production of ROS, with consequent activation of specific redox signaling pathways (49, 58). Much evidence has shown that Nox family members are activated within discrete cellular compartments, including caveolae and lipid rafts (59), focal adhesions (60), cell-cell contacts (61), phagosomes (39), lamellipodia and leading edges (62), membrane ruffles (63), endosomes (64), and the nucleus (65). We propose that one mechanism by which Nox enzymes are spatially regulated in specific subcellular compartments is through interaction with organizer subunits.

We show that Tks4 and Tks5 are functional members of the p47 organizer superfamily that selectively support Nox1- and Nox3-dependent ROS generation (Fig. 1B). Tks4 and Tks5 were unable to act as organizer subunits for Nox2, or to enhance Nox4-dependent ROS formation (Fig. 2A). The former may represent an inability of Tks to functionally recruit p67phox. The latter contrasts with the Tks-dependent regulation of Nox4 activity (37), raising the possibility that Nox4 activity may be organizer-dependent under certain conditions. Nox1 activation mediated through Tks4 and Tks5 was blocked by the Nox flavoenzyme inhibitor DPI (Fig. 2B) and was Rac dependent (Fig. 2C). The interaction between Tks organizers and NoxA1 is not affected by the GTP-binding state of Rac (Fig. 3B). Moreover, this interaction, as well as Tks-mediated ROS generation, can be prevented by the disruption of all five Tks SH3 domains (Fig. 3C). We did not observe any change in the ability of Tks5 single- or double–SH3 domain mutants to support ROS generation and to bind NoxA1 (fig. S3, A to C), possibly because of their overexpression. However, the SH3-mediated interaction between Tks5 and N-Wasp has also been reported to involve multiple Tks5 SH3 domains (46). The interaction between Tks5 and NoxA1 differs from that of the p47phox organizer with p67phox, which occurs through binding of a C-terminal proline-rich sequence on p47phox to the C-terminal SH3 domain of p67phox (44). Tks proteins also appear to be capable of binding the p22phox subunit usually associated with the Nox1 to Nox4 proteins through the two N-terminal SH3 domains of Tks (37). Finally, we showed that the PX domain of Tks5 is required for Nox1 activation, although not for binding to NoxA1. These observations presumably reflect the need for multiple protein-lipid interactions for proper membrane assembly of the Nox1 complex, analogous to those documented for Nox2 assembly by the p47phox organizer in neutrophils (14).

Human DLD1 colon cancer cells provide a system for studying the effect of Tks-mediated ROS generation by Nox1

Here, we show that human DLD1 colon cancer cells (i) contain Nox1 (and no other members of the Nox family), NoxA1, Tks4, and small amounts of Tks5 (but no other members of the p47 organizer superfamily), (ii) form invadopodia and degrade the ECM in a Tks4-dependent manner, and (iii) generate ROS through Nox1 by a mechanism that depends on the abundance of Tks4 as a result of its biochemical interaction with endogenous NoxA1. In this cell system, Nox1 colocalizes with F-actin– and cortactin-positive structures that degrade the ECM, a hallmark of invadopodia function. We found that the expression of a different organizer subunit (NoxO1) competes with endogenous Tks4 to reduce Nox1 localization to these structures, thereby decreasing ROS-dependent formation of invadopodia. Indeed, despite overall ROS generation comparable to that observed with Tks4 alone, NoxO1 overexpression reduced invadopodia formation and function. This was most likely a result of the loss of local ROS generation by the displaced Nox1, as suggested by the reduction of ROS-positive invadopodia-like structures observed in NoxO1-overexpressing DLD1 cells. Consistent with this, NoxO1 did not localize to invadopodia, and its presence reduced invadopodia formation. These data suggest how different organizer subunits may localize Nox1 complex activity to distinct subcellular compartments, facilitating spatially confined ROS production near redox-sensitive targets to initiate specific signaling events.

Physiological and pathophysiological implications

Nox1 is abundant in colon epithelial cells (1, 66, 67), and Nox1 overexpression and increased ROS generation have been observed in human colon cancers and colon cell lines (68, 69). ROS generated by Nox1 have been implicated in the migration of colon cancer cells (53). Increased c-Src activity is also characteristic of both premalignant and progressively advanced colon neoplasia and correlates with the conversion of benign polyps into malignant metastatic tumors (70). We have previously shown that there is a correlation between the amounts of c-Src activity in colon cancer lines and their ability to generate ROS; however, the exact relationships between Src activity, ROS formation, carcinogenesis, and metastasis remain poorly defined.

Metastatic cancer cells both migrate to and degrade the ECM, and invasiveness has been correlated with the presence of invadopodia (30, 31), as well as ROS production (71). ROS formation through Nox enzymes is a critical component of the formation of invadopodia in cancer cells (37). Consistent with this, we have found that (i) Nox1-specific siRNAs decrease ROS generation and, consequently, invadopodia formation and ECM degradation, (ii) the formation of invadopodia in DLD1 cells is Tks4 dependent, and (iii) the presence of the NoxO1 organizer subunit reduces formation of invadopodia and ECM degradation in DLD1 cells. Our results strengthen the notion that, rather than overall ROS production, it is the localized Tks-mediated generation of ROS at invadopodia that is responsible for invadopodia formation and function. Thus, members of the p47phox organizer superfamily may be more generally important in regulating other processes requiring the localized formation of ROS. A better understanding of the mechanisms underlying the compartmentalization of redox signaling may suggest approaches to the use of antioxidant and anti-Nox–directed therapies for treatment of various redox stress-dependent diseases and cancer metastasis.

Materials and Methods

DNA constructs, reagents, and antibodies

Expression vectors for all members of the Nox family and p47 organizer family, along with plasmids for SrcYF, RacQL, RacN17, NoxA1, NoxA1-R103E, and Nox1TM have been previously described (29, 42). RFP-tagged and GFP-tagged Nox1 were created by subcloning pcDNA3.1-Nox1 into pDsRed-Monomer-C1 and pEGFP-N1 (Clontech), respectively. Notably, both the Nox1-GFP and the Nox1-RFP plasmids used in these experiments supported ROS production (fig. S6B). GFP-tagged NoxO1 was created by subcloning pRK5m-NoxO1 into pEGFP-N1. The Tks5 SH3 domain mutant constructs were generated by inserting the mutations W188A, W260A, W441A, W827A, and W1056A with the QuikChange site-directed mutagenesis kit (Stratagene) as described (45). Cell culture medium, fetal bovine serum, supplements, and Hanks’ balanced salt solution (HBSS, cat. no. 24020-117) were from Invitrogen. Plasmids for transfection were purified with the Qiagen Qiafilter system. DLD1 colonic adenocarcinoma cells (cat. no. CCL-221) and other carcinoma cell lines were purchased from ATCC. DPI (D2926), horseradish peroxidase (HRP) (77330) and luminol (09253) were purchased from Sigma. The following antibodies were purchased as indicated: rabbit polyclonal antibody against GFP from Molecular Probes; mouse monoclonal antibody against actin (691002) from MD Bioscience; mouse monoclonal antibody M2 against Flag from (Sigma-Aldrich), and mouse monoclonal antibody 4F10 against cortactin from Millipore. Alexa Fluor 568 and 647 against mouse, Alexa Fluor 568 against rabbit, and Alexa Fluor 568 phalloidin were purchased from Molecular Probes. Rabbit polyclonal antibodies against Tks4 and Tks5 have been previously described (24, 29). Rabbit polyclonal antibody against Nox1 was provided by D. Lambeth (Emory). 9E10 antibody against myc and rabbit polyclonal NoxA1 were prepared in-house.

Cell cultures, transfection, and siRNAs

Human HEK293, human DLD1, HT29, SW620 colonic adenocarcinoma, and mouse 3T3 and Src-3T3 fibroblasts were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen) containing 10% heat-inactivated fetal bovine serum (Invitrogen), 2 mM glutamine, and antibiotics [penicillin (100 U/ml) and streptomycin (100 g/ml)] at 37°C in 5% CO2. Cells were transfected by Lipofectamine 2000 (Invitrogen) as described (42), and 24 to 96 hours after transfection, cells were processed for various analyses. ON-TARGETplus SMARTpool Tks4, Nox1, and control siRNA mixture (J-032834-05/08, J-010193-05/08) were purchased from Dharmacon.

Reverse transcription–polymerase chain reaction, Western blot, and immunoprecipitation

RNA extraction was performed with the RNeasy kit (Quiagen) and reverse-transcribed with Superscript II enzyme (Invitrogen) according to the manufacturer’s instructions. Western blot and protein extracts were performed as described (42). For immunoprecipitation, 1- to 2-μl antibodies were incubated with 1 mg protein lysate for 2 hours at 4°C, followed by 30-min incubation with 20 μl of Protein G–Sepharose (Amersham). Immunoprecipitates were washed three times in lysis buffer, and proteins were released by boiling in Laemmli SDS sample buffer then analyzed by Western blot.

Measurements of ROS

ROS were measured with a luminol-based CL assay as described (42). Chemiluminescence was recorded with a 96-well plate luminometer (Berthold) 5 min after the addition of HRP/luminol mixture (with the exception of Fig. 1B in which ROS generation was monitored continuously for 30 min). For real-time visualization of ROS generation illustrated in Fig. 6D, DLD1 cells plated on glass coverslips were incubated with 2.5 μM ROS-sensitive probe PY1-AM (Peroxy-Yellow 1 Acetoxymethyl-ester) in HBSS for 30 min at 37°C, washed with PBS, mounted on slides, and analyzed by confocal microscopy. PY1-AM (provided by B. C. Dickinson and C. J. Chang, University of California, Berkeley), is a selective fluorescent indicator for hydrogen peroxide related to Peroxy Green 1 (72) and Mito Peroxy Yellow 1 (73).

Confocal and epifluorescence microscopy

Twenty-four to 48 hours after transfection, DLD1 cells plated on glass or FITC-labeled, gelatin-coated coverslips were fixed in 4% paraformaldehyde (PFA) at room temperature or in methanol at -20°C for 10 min. Successively, cells were permeabilized in 0.5% Triton for 10 min and blocked in 2% bovine serum albumin in phosphate-buffered saline for 45 min at room temperature. Cells were then immunolabeled as indicated in the figure legends with appropriate primary and Alexa Fluor 568– or 647-conjugated secondary antibodies. F-actin was detected by Alexa Fluor 568–conjugated phalloidin. Cells were mounted on slides with Mowiol mounting medium (Calbiochem) according to the manufacturer’s instructions. Epifluorescence images of fixed cells were acquired on an inverted microscope (Eclipse TE 2000-U, Nikon) equipped with an electronically controlled shutter, filter wheels, and a 14-bit cooled CCD camera (Cool SNAP HQ, Photometrics) controlled by MetaMorph software (Universal Imaging Corp.) with a 60×/1.4 NA Plan Apo DIC or a 40×/1.4 NA Plan Apo Ph3 objective lens (Nikon) (74). Confocal images were acquired on a spinning-disk confocal microscope system (74), equipped with a CoolSnapHQ camera and 100×/1.4 NA Plan Apo or a 60×/1.4 NA Plan Apo objective lens (Nikon).

ECM degradation assays

Fluorescently labeled gelatin-coated coverslips were prepared as previously described (35). Cells were transfected as indicated in figure legends and incubated on labeled coverslips for 24 to 48 hours, then processed accordingly for epifluorescence or confocal microscopy.

Statistical analysis

In this study, overall, representative experiments from at least three independent experiments are shown. Results for each experiment are given as the mean of triplicates ±SD except where otherwise noted. Statistically significant differences between sample groups were determined by two-tailed t tests (Microsoft Excel) except where otherwise noted. P values for differences observed in figures were always <0.001 unless differently indicated by asterisks in the individual figure legends.

Acknowledgments

We acknowledge and profoundly appreciate B. C. Dickinson and C. J. Chang (University of California, Berkeley) for allowing our use of the ROS-specific probe PY1-AM in this study. We thank D. Lambeth (Emory) for Nox1 antibody. We also acknowledge B. Bohl for expert technical assistance and V. Delorme-Walker for helping with confocal microscopy and suggestions in revising this manuscript. Tks4 constructs were provided by P. Bromann (Biomedicum, Helsinki). We thank G. Danuser for access to the confocal microscope used in these studies. This work was supported by NIH grant HL48008 (G.M.B.), The National Cancer Institute (S.C.), and a postdoctoral fellowship from Fondation pour la Recherche Medicale FRM (N.T.). This is manuscript IMM20051 from The Scripps Research Institute.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/2/88/ra54/DC1

Fig. S1. Myc-tagged transfected organizer subunits are expressed in comparable amounts in reconstituted HEK293 cells.

Fig. S2. DLD1 cells only express the Nox1 member of the Nox family, the NoxA1 activator, and Tks4 (and small amounts of Tks5) as members of the p47phox organizer superfamily.

Fig. S3. The presence of Tks5 single SH3 domain inactivating mutations (M1 to M5), or the TksM1M2 double mutation did not affect the ability of Tks5 to support Nox1-dependent ROS generation and to bind NoxA1.

Fig. S4. Tks4-specific siRNA efficiently knocks down Tks4 protein at 72 and 96 hours and significantly reduces the ability of DLD1 cells to degrade ECM.

Fig. S5. Nox1-specific siRNAs efficiently block ROS generation, invadopodia formation, and ECM degradation in DLD1 cells and this effect can be rescued by overexpression of murine Nox1 (mNox1).

Fig. S6. In DLD1 cells, Nox1 and phalloidin- and Nox1 and cortactin-positive structures are ventral organelles that have the characteristics of invadopodia.

Fig. S7. NoxO1-GFP does not localize to invadopodia in DLD1 cells and invadopodia formation in NoxO1-transfected cells is greatly reduced.

References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article