Mitochondrial Dok-4 recruits Src kinase and regulates NF-kappaB activation in endothelial cells.

The downstream of kinase (Dok) family of adapter proteins consists of at least five members structurally characterized by an NH2-terminal tandem of conserved pleckstrin homology and phosphotyrosine binding domains linked to a unique COOH-terminal region. To determine the role of the novel adapter protein Dok-4 in endothelial cells, we first investigated the cell localization of Dok-4. Most surprisingly, immunofluorescence microscopy, cell fractionation studies, and studies with enhanced green fluorescent protein chimeras showed that wild type Dok-4 (Dok-4-WT) specifically localized in mitochondria. An NH2-terminal deletion mutant of Dok-4 (Dok-4-(deltaN11-29)), which lacks the mitochondrial targeting sequence, could not accumulate in mitochondria. Co-immunoprecipitation revealed an interaction of c-Src with Dok-4-WT in endothelial cells. Most interestingly, overexpression of Dok-4-WT, but not Dok-4-(deltaN1-99), increased mitochondrial c-Src expression, whereas knock-down of endogenous Dok-4 with a small interfering RNA vector greatly inhibited mitochondrial localization of c-Src, suggesting a unique function for Dok-4 as an anchoring protein for c-Src in mitochondria. Dok-4-WT significantly decreased 39-kDa subunit complex I expression. PP2, a specific Src kinase inhibitor, prevented the Dok-4-mediated complex I decrease, suggesting the involvement of Src kinase in regulation of complex I expression. Dok-4-WT enhanced tumor necrosis factor-alpha (TNF-alpha)-mediated reactive oxygen species (ROS) production, supporting the functional relevance of a Dok-4-Src-complex I/ROS signaling pathway in mitochondria. Finally, Dok-4 enhanced TNF-alpha-mediated NF-kappaB activation, whereas this was inhibited by transfection with Dok-4 small interfering RNA. In addition, Dok-4-induced NF-kappaB activation was also inhibited by transfection of a dominant negative form of c-Src. These data suggest a role for mitochondrial Dok-4 as an anchoring molecule for the tyrosine kinase c-Src, and in turn as a regulator of TNF-alpha-mediated ROS production and NF-kappaB activation.

The intracellular signaling network triggered by binding of growth factors and cytokines to cellular surface receptors is rigorously regulated in cells and often involves recruitment of adapter molecules to transmembrane receptors (1,2). Downstream of kinase (Dok) 1 proteins are a recently discovered family of adapter molecules (including Dok-1, -2, and -3), which have emerged as an expanding group of insulin receptor substrates (IRSs)-related signaling molecules, consisting of an NH 2 -terminal tandem of PH and PTB domains. The previously identified Dok members, p62 Dok (Dok-1), Dok-2 (Dok-R), and Dok-3, are predominantly expressed in hematopoietic cells (3). Although these three Doks have similar domains to IRSs, they can be distinguished from the IRS family based on sequence homology and functional interactions. Hematopoietic Dok molecules are prominent substrates of the Src and Abl tyrosine kinases. Upon tyrosine phosphorylation, they acquire the ability to bind SH2 domain-containing molecules such as RasGAP, Csk, and Nck (3)(4)(5). More recently, Grimm et al. (6) identified two novel Dok-like molecules, Dok-4 and Dok-5, as partners and substrates for the receptor tyrosine kinase Ret. Whereas Dok-5 appears concentrated in the central nervous system, Dok-4 seems most highly expressed in epithelial and endothelial cells (4,6). We have reported (4) that a significant proportion of Dok-4 is constitutively associated with the cell membrane and that it can serve as a substrate for tyrosine kinases of the Src family. However, because Dok-4 lacks consensus binding sites for known SH2 domains, its exact function has been difficult to define. We found that Dok-4 inhibited activation of the Erk substrate Elk-1 by Ret and by the Src kinase Fyn in epithelial cells. On the other hand, by using different approaches and cellular systems, Grimm et al. (6) and Cai et al. (7) have suggested a role for Dok-4 in enhanced Erk activation. Although it has been reported that Dok-4 is strongly expressed in vascular endothelium (6), its function within this context remains unexplored.
The Src family of nonreceptor protein-tyrosine kinases plays critical roles in a variety of cellular signal transduction pathways, regulating such diverse processes as cell division, motility, adhesion, angiogenesis, and survival. Although c-Src protein kinase does not possess any mitochondrial targeting sequence, several recent studies (8,9) have shown that c-Src exists not only in the cytoplasm but also in mitochondria. Miyazaki et al. (9) have reported that c-Src associates with cytochrome c oxidase (complex IV subunit II (Ccox)) and activates Ccox activity in osteoclasts. However, the physiological role of mitochondrial c-Src in endothelial cells remains unclear.
It has been reported that the PH domains of IRS and Dok proteins bind phospholipids and as such serve as membrane targeting domains. In addition, since Tamir et al. (10) reported that the binding of Dok-1 to plasma membrane is due to the association of the PTB domain of Dok-1 with tyrosine-phosphorylated SH2-containing 5-inositol phosphatase, both the PH and PTB domains of Dok family proteins may have crucial roles in determining cellular localization. In fact, we recently reported that in COS, HEK293, and epithelial cells, overexpressed Dok-4 localized to the cell membrane in a PH and PTB domain-dependent manner (4). In addition, we noted that Dok-4 localized in an as yet undefined punctate cytoplasmic compartment. In the current study, while examining the localization of Dok-4 in endothelial cells, we found that the punctate distribution of Dok-4 in cytosol represented mitochondrial localization. In addition, we found that the NH 2 -terminal region of Dok-4 contains a putative novel mitochondrial targeting sequence within the previously described PH domain, and that both the NH 2 -terminal region and PTB domain are critical for mitochondrial localization. Furthermore, overexpression of Dok-4 recruits Src to mitochondria and increases TNF-␣-induced ROS generation and NF-B activation. Because the Src kinase family does not possess a mitochondrial targeting sequence, this suggests that Dok-4 serves as the anchoring protein of Src kinase family in mitochondria. Overexpression of Dok-4 enhanced TNF-␣-mediated ROS production as well as NF-B activation in endothelial cells. In addition, a knockdown of Dok-4 by siRNA significantly inhibited this NF-B activation, suggesting a role for endogenous Dok-4 expression on mitochondria-mediated inflammatory responses in endothelial cells.

High Transfection Efficiency of siRNA, Dok-4 WT, and Mutants in BAECs and BLMECs and Low Transfection Efficiency of Dok-4 WT and Dok-4-(⌬N1-99) in BAECs-Subconfluent
BAECs and BLMECs in 35-mm dishes were washed twice with serum-free OPTI medium (Invitrogen). For high efficiency transfection in BAECs and BLMECs, we performed two consecutive transfections with Lipofectamine 2000 (Invitrogen). 1-4 l of Lipofectamine 2000 was mixed with 500 l of serum-free OPTI medium. After 20 min, 0.5-2 g of Dok-4 siRNA or control siRNA was added and incubated for 30 min at room temperature. BLMECs in 2 ml of serum-free OPTI medium were treated with this mixture. After 3 h of incubation, the medium was changed to MCDB-131 with 10% FBS, and the cells were cultured for 24 h. Then we transfected BLMECs again with the same serum-free OPTI medium containing siRNA/Lipofectamine 2000 mixture for 3 h followed again by MCDB-131 medium with 10% FBS for another 24 h. For transfection of Dok-4 WT and mutant constructs, 4 l of Lipofectamine 2000 was mixed with 500 l of serum-free OPTI medium. After 20 min, 3 g of Dok-4 WT or mutant constructs were added and incubated for 30 min at room temperature. BAECs or BLMECs in 2 ml of serum-free OPTI medium were treated with this mixture. After 3 h of incubation, the medium was changed to MCDB-131 media with 10% FBS, and the cells were cultured for 6 h. Then we transfected BAECs or BLMECs again with same serum-free OPTI medium containing Dok-4 WT or mutants/ Lipofectamine 2000 mixture for 1.5 h, and after 1.5 h of incubation, the medium was changed to MCDB-131 media with 10% FBS, and cells were cultured for 24 h. Under these conditions (high transfection efficiency), 80 -90% of the BLMECs were transfected (data not shown), and the Dok-4 expression of the culture was decreased by 70% in BLMECs (Fig. 3E). In Fig. 5 we intentionally transfected Dok-4 WT and Dok-4-(⌬N1-99) at low transfection efficiency (30 -40%) to compare directly the Dok-4-transfected BAECs with nontransfected cells as we described previously (13). BAECs were washed and placed in 1 ml of serum-free OPTI medium as above. They were then treated with Dok-4 WT and mutant constructs (1 g/ml) in the OPTI mixture for 3 h, and after 3 h of incubation, the medium was changed to M199 media with 10% FBS, and cells were cultured for 24 h.
RNA Interference-RNA interference construct was created using pSHAG (kindly provided by Greg Hannon, Cold Spring Harbor Laboratories). Briefly, oligonucleotides carrying short RNA hairpins targeted to conserved regions of human (GenBank TM accession number AF466369) and mouse (BC004705) Dok-4 were annealed and cloned into BseRI-BamHI-cut pSHAG just downstream of the U6 promoter. The sequences of the oligonucleotides used for the Dok-4 siRNA construction were as follows: oligonucleotides A (5Ј-TGG ATG TCC  CAG AGG TAG ATG TTC TCG TGA AGC TTG ATG AGA ATA TCT  ATC TCT GGG ACA TTC ACA ATT TTT T-3Ј) and B (5Ј-GAT CAA AAA ATT GTG AAT GTC CCA GAG ATA GAT ATT CTC ATC AAG CTT CAC GAG AAC ATC TAC CTC TGG GAC ATC CAC G-3Ј). The sequences of the oligonucleotides used for GFP (control) siRNA construction were as follows: shGFP A (5Ј-TTGTACTCCAGCTTGTGCCCCAG-GATGTGAAGCTTGACATCCTGGGGCGCAGGCTGGAGTGCAACTA-TTTTTT-3Ј) and shGFP B (5Ј-GATCAAAAAATAGTTGCACTCCAGCC-TGCGCCCCAGGATGTCAAGCTTCACATCCTGGGGCACAAGCTGG-AGTACAACG3Ј). We used GFP siRNA for a control as described previously (14,15), and GFP siRNA construct was kindly provided by Drs. Streb and Miano (15).
Mitochondrial Preparation-Mitochondria fraction in BAECs was purified with ApoAlert cell fractionation kit (Clontech) according to the manufacturer's instructions.
PathDetect in Vivo Signal Transduction Pathway Reporting System-NF-B and AP-1 activity was assayed by using the PathDetect Signal transduction Pathway trans-Reporting Systems (Stratagene) as described previously (18). Elk-1 activity was assayed by using the PathDetect Signal Transduction Pathway trans-Reporting Systems (Stratagene). The pRL-TK Renilla luciferase vector was used for normalization of transfection. Cells were co-transfected with pNF-B-Luc reporter plasmid and pRL-TK with other plasmids as indicated in the figures.
Immunofluorescence and Confocal Microscopy Analysis-Cells were fixed with 3.7% formaldehyde for 10 min followed by permeabilization with 0.05% Triton X-100 for 15 min. After fixation and permeabilization of the cells, they were incubated with 10% normal goat serum (Vector Laboratories) and treated with anti-Dok-4 antibody or anti-Myc antibody (Invitrogen) for 45 min at room temperature. Secondary antibodies labeled with Alexa Fluor 546 dye against rabbit IgG or 488 dye against mouse IgG 1 were purchased from Molecular Probes. To detect mitochondria distribution, we transfected the cells with mito-EYFP vector (Clontech) (containing targeting sequence from subunit VIII of cytochrome c oxidase). Images were observed using a confocal laser-scanning microscope (Fluoview; Olympus) with a LUMPlanFl60ϫ lens or a fluorescence microscope (Axiophot; Carl Zeiss MicroImaging, Inc.) equipped with a CCD camera and an Acroplan Water 60ϫW lens. Quantification of the overlay image was done using Photoshop 7.0 program.
Mitochondria-reactive Oxygen Species (ROS) Detection, CM-H 2 XRos Staining-ROS production by mitochondria was monitored by using the Reduced MitoTracker Red CM-H 2 XRos staining (Molecular Probes). The reduced version of MitoTracker Red CM-H 2 XRos does not fluoresce until entering actively respiring cells, where it is oxidized by ROS to a red fluorescent compound, which is sequestered in the mitochondria (19). After 24 h of transfection with EGFP-tagged wild type (Dok-4 WT) and NH 2 -terminal domain deletion mutant (Dok-4⌬-(N1-99)) in BAECs, cells were stimulated with TNF-␣ for 1 h. The medium was changed to pre-warmed culture medium containing 100 nM of Reduced Mito-Tracker Red (CM-H 2 XRos, purchased from Molecular Probes) for 20 min. After 20 min, the medium was carefully removed and replaced with pre-warmed 3.7% formaldehyde containing serum.
ROS Measurement by Dihydroethidium (DHE)-Intracellular ROS were also measured using the redox-sensitive dye DHE (Molecular Probes). DHE enters the cell and is oxidized by ROS, particularly superoxide (O 2 . ), to become ethidium that emits red fluorescence.
Ethidium intercalates with the DNA of the cell, further amplifying fluorescence. Therefore, the fluorescence in the nuclear region is indicative of ROS generation and was selected as the region of interest for data analysis (20). Cultured endothelial cells were loaded with 10 M dihydroethidium in HEPES buffer (10 mM HEPES, 10 mM glucose, 140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 5 mM NaHCO 3 , 0.6 mM Na 2 HPO 4 , 1.2 mM Na 2 SO 4 , pH 7.4) at 37°C for 30 min. After incubation, cells were washed three times with indicator-free HEPES buffer and left in the last washing buffer for imaging studies. Singlecell imaging was taken by a fluorescent microscope (TILL Photonics LLC) using Nikon TE2000s inverted microscope with a 40ϫ oil objective. The cells was illuminated at 515 nm, and the emitted fluorescence was collected between 580 and 630 nm. Fluorescent images from 12-15 cells were taken every minute in a 60-min time course in each set of experiments. At the end of each experiment, mitochondrial respiratory chain complex I and III inhibitors rotenone and antimycin A (Sigma) were added to induce maximal ROS production resulting in maximal DHE fluorescence. Ccox Activity-Ccox activity of cell lysates was measured as cyanidesensitive oxidation of reduced cytochrome c, spectrophotometrically at 550 nM as described previously (21).
Statistical Analysis-Data are reported as mean Ϯ S.D. Statistical analysis was performed with the StatView 4.0 package (ABACUS Concepts, Berkeley, CA). Differences were analyzed with a one-way or a two-way repeated measure analysis of variance as appropriate, followed by Schéffe's correction for multiple comparisons. p values less than 0.05 are indicated by * and less than 0.01 by ** (Figs. 3-5, 7, and 8).

Dok-4 Is Localized in Mitochondria through Its NH 2 -terminal Region and PTB Domain in Endothelial
Cells-To determine the intracellular localization of Dok-4 in endothelial cells, we performed immunostaining analysis by using anti-DOK-4 antibody in BAECs. Strikingly, we found a clear pattern of cytosolic punctate Dok-4 staining, reminiscent of mitochondrial localization (Fig. 1A). No signal was detected when only secondary antibody was used (data not shown).
To confirm whether Dok-4 localizes in mitochondria, we utilized a mitochondrion-specific localizing vector, mito-EYFP (containing targeting sequence from subunit VIII of cytochrome c oxidase), which exhibits bright green, fluorescein-like fluorescence as described under "Material and Methods." As shown in Fig. 1B, punctate staining observed after anti-Dok-4 immunostaining co-localized with mitochondrial signals by confocal microscopy. Of note, there was no signal detected in non-mito-EYFP-transfected cells, suggesting that the red fluorescence signal from Dok-4 did not leak into the green fluorescence of mito-EYFP (data not shown). Similar co-localization of Dok-4 and mito-EYFP was confirmed by confocal microscopy following anti-Myc immunostaining of cells transfected with a Myc-tagged Dok-4 ( Fig. 1C). To ascertain that this was indeed mitochondrial localization, we transfected Myc-tagged Dok-4 and then isolated the mitochondrial fraction from the BAECs. As shown in Fig. 1D, we found that transfected Myc-tagged Dok-4 and endogenous Dok-4 existed in mitochondria but was undetectable in extra-mitochondrial fractions. To determine the purity of the mitochondrial preparation, we also performed Western blot analysis for expression of Ccox, MEK-1, caveolin, PCNA, GM130, Bip, and ␣-tubulin as markers of mitochondrial, cytosolic, membrane, nucleus, Golgi, endoplasmic reticulum, and cytoskeleton localization, respectively. We found that the immunoreactions were negative for MEK-1, caveolin, PCNA, GM130, Bip, and tubulin expression in the mitochondrial preparations. In contrast, Ccox was only positive in the mitochondrial preparation (Fig. 1D).
Dok-4 Associates and Increases Mitochondria Src-Previously, we found that Dok-4 can be a good substrate for Src family kinases (4), but the possibility of an interaction between Src family kinases and Dok-4 remained unexplored. Because it had been reported that Src family kinases not only exist in cell membranes but also in mitochondria (9), we investigated the role of Dok-4 in mitochondrial Src localization. First, we determined whether Dok-4 could associate with c-Src in endothelial cells. We transfected Myc-tagged Dok-4 WT in BAECs, immunoprecipitated with anti-c-Src antibody and immunoblotted with anti-Myc antibody. We found that Dok-4 WT could coimmunoprecipitate with c-Src (Fig. 3A, upper). The results were identical when we conversely immunoprecipitated with anti-Myc antibody, and immunoblotted with anti-c-Src antibody (Fig. 3B, upper).
To ensure that a small amount of plasma membrane-bound Dok-4 was not responsible for association with Src, we transfected Myc-tagged Dok-4-(⌬N1-99) or Dok-4 WT, and we isolated mitochondrial and extra-mitochondrial fractions. As shown in Figs. 1D and 3C, we detected myc-Dok-4 WT only in the mitochondrial fraction and not in the extra-mitochondrial fraction. In contrast, myc-Dok-4-(⌬N1-99) was observed in the extra-mitochondrial fraction but not in the mitochondrial fraction. After purifying mitochondrial fractions, we used them to perform co-immunoprecipitation experiments with anti-Myc antibody. As shown in Fig. 3C, we found that c-Src co-immunoprecipitated with myc-Dok-4 WT in the mitochondrial fraction. In contrast, we could not see any association of myc-Dok-4 WT with c-Src in the extra-mitochondrial fraction. Most interestingly, myc-Dok-4-(⌬N1-99) was expressed entirely in the extra-mitochondrial fraction, and it did not associate. Because caveolin was undetectable in our mitochondrial fraction, it is unlikely that the association of c-Src with Dok-4 was because of contamination of the mitochondrial fraction by the membrane c-Src.
Next, we determined whether Dok-4 could regulate mitochondrial Src expression. Because c-Src exists not only in mitochondria, but also mostly in the cytosol and at the membrane, it is difficult to discern its mitochondrial pool by confocal microscopy. Therefore, we utilized two alternative approaches to determine whether mitochondrial localization of c-Src might be regulated by Dok-4. First, by cell fractionation studies, we found a significant increase of c-Src in mitochondria after Dok-4 overexpression as shown in Fig. 3D. To determine the role of endogenous Dok-4 for mitochondrial Src expression, we utilized Dok-4 siRNA. As shown in Fig. 3E, treatment of Dok-4 siRNA, but not control siRNA, dose-dependently inhibited Dok-4 expression in endothelial cells, confirming the validity of this knock-down approach. More importantly, we found that mitochondrial Src was significantly decreased by Dok-4 siRNA treatment but not by control siRNA (Fig. 3F). Most interestingly, we also observed that extra-mitochondrial Src expression was slightly higher in Dok-4 siRNA-treated cells. Taken together, these results suggest that mitochondrial Dok-4 is critical for recruiting extra-mitochondrial Src into mitochondria.
Finally, because it has been reported that c-Src associates with mitochondrial Ccox (9), we determined the impact of Dok-4 overexpression on the association of Src with Ccox by immunoprecipitation with an anti-Ccox antibody. As shown in Fig. 4, A and B, we found that Dok-4 WT significantly increased the association of Src with Ccox. In contrast, Dok-4-(⌬N1-99) did not increase the association of c-Src with Ccox. Because Ccox only exists in mitochondria, the increase of Src/Ccox interaction induced by Dok-4 supports the interpretation that Dok-4 WT is responsible for the localization of Src in mitochondria and its association with Ccox.
Dok-4 and Mitochondrial Respiratory Complexes-Because we determined that Dok-4 increased Src/Ccox interaction, we next investigated whether Dok-4 induction affects Ccox expression and activity. We did not observe any significant change of Ccox expression (Fig. 4C, lower) or activity (89 Ϯ 12% compared with empty vector transfection) by induction of Dok-4 WT. Because it has been reported that nitric oxide decreases complex I expression in endothelial cells (21) and nitric oxide can activate Src kinase activity (22, 23), we investigated whether Dok-4 regulates complex I expression. Most interest-ingly, we found that complex I subunit 39-kDa expression was significantly decreased by Dok-4 WT but not Dok-4-(⌬N1-99) induction (Fig. 4C, upper). To determine the involvement of Src kinase activity in the Dok-4 WT-mediated complex I decrease, we used the Src-specific inhibitor PP2. As shown in Fig. 4D, incubation of PP2 inhibited the Dok-4 WT-mediated complex I decrease without affecting Dok-4 WT induction and Ccox expression. These data suggest that the decrease of complex I expression by Dok-4 is mediated by Src kinase activation.
Dok-4 Enhanced TNF-␣-mediated ROS Production-Several lines of evidence indicate that mitochondria-mediated ROS generation is a major source of oxidative stress in the cell (24 -26). It has been suggested that complex I is the primary source of ROS in a variety of pathological scenarios ranging from aging to Parkinson disease (24,27). Therefore, to determine the role of Dok-4 in this process in endothelial cells, we transfected EGFP-tagged Dok-4 WT in BAEC, and we measured ROS production using Reduced Mito-Tracker Red probe (CM-H 2 XRos). CM-H 2 XRos does not fluoresce until it enters an actively respiring cell, where it is oxidized predominantly by reactions involving hydrogen peroxide production, and sequestrated into the mitochondria by virtue of its cationic state (19,28). For these studies, we intentionally transfected EGFPtagged Dok-4 WT at reduced transfection efficiency (30 -40%) in order to facilitate direct comparison of transfected and nontransfected cells within individual fields, as we had performed previously (11,13). As shown in Fig. 5, A-D, in the absence of stimulation, we could not detect any differences of CM-H 2 XRosmediated signals (middle) between the cells overexpressing Dok-4 WT and nontransfected cells (compare top and bottom panels).
Because TNF-␣ has been reported to enhance ROS generation and induce several ROS-sensitive genes (29), we used this cytokine to examine a possible role of Dok-4 in ligand-induced ROS production. Strikingly, we found that TNF-␣-mediated ROS production was significantly increased in Dok-4-transfected cells compared with non-Dok-4-transfected cells in the same optical field (Fig. 5B). ROS production was about a 4 -5-fold increased, compared with nontransfected BAECs (Fig. 5E). In contrast, we did not observe any difference in the TNF-␣-induced CM-H 2 XRos probe signal between cells transfected with Dok-4-(⌬N1-99)-EGFP and nontransfected cells (Fig. 5, C-F). These data suggested that expression of Dok-4 in endothelial mitochondria enhanced TNF-␣-induced ROS production. Of note, because of the difficulties inherent with comparing fluorescence signal in cells from different dishes (as opposed to cells within the same field), we could not reliably quantify the amount of ROS induction by TNF-␣ by using this technique.
To confirm these results, we also performed experiments under high transfection efficiency conditions as we described under "Materials and Methods." Under the conditions used here, over 80 -90% of the cells were transfected (data not shown). Because the mitochondrial respiratory chain is a major source of O 2 . , and DHE is much more sensitive for O 2 . than for H 2 O 2 , OH 2 . , or ONOO Ϫ , we selected this probe to detect mitochondrial Dok-4-mediated ROS production. After loading with 10 M DHE, fluorescence from 15 to 25 cells was simultaneously monitored. In our cell systems (BLMECs), we did not find a significant increase of ROS production induced by TNF-␣. However, we observed that rotenone and antimycin A could increase DHE fluorescence in the same cells, and the capacity for ROS production in these cells was preserved. We transfected BLMECs with Dok-4 WT or Dok-4-(⌬N11-29) as a control. After 24 h of transfection, we loaded the cells with the DHE probe, and we detected TNF-␣-mediated ROS production. As shown in Fig. 6, A and B, we did not find any significant difference among Dok-4 WT, Dok-4-(⌬N11-29), and nontransfected cells in the basal state. However, DHE fluorescence induced by TNF-␣ was significantly increased in Dok-4 WTtransfected cells. In contrast, no significant increase of DHE fluorescence was observed in the nontransfected and Dok-4-(⌬N11-29)-transfected cells. These data further substantiate the measurements using CM-H2XRos probe to detect TNF-␣mediated mitochondrial ROS production in endothelial cells.

Dok-4 Induced NF-B Activation in Endothelial
Cells-Because ROS are now recognized as critical regulators of intracellular signaling cascades, including NF-B activation, we wondered whether Dok-4 could regulate NF-B activation in endothelial cells. To determine this, we co-transfected Dok-4 WT or empty vector with an NF-B reporter gene and measured its activation by luciferase assay as described previously (18). As shown in Fig. 7A, stimulation with TNF-␣ induced NF-B activation. Most interestingly, overexpression of Dok-4 itself increased NF-B activation in a dose-dependent manner (Fig. 7A). Dok-4 also significantly enhanced TNF-␣ induced NF-B activation (Fig. 7A).
To determine the role of mitochondrial localization of Dok-4 in NF-B activation, we also transfected Dok-4-(⌬N1-99), -(⌬N11-29), -(⌬C234 -325), or -(⌬100 -233) in endothelial cells. We found that Dok-4-(⌬N1-99), -(⌬N11-29), and -(⌬100 -233) could not increase NF-B activation, suggesting the important roles of the NH 2 -terminal mitochondrial targeting sequence and PTB domain-mediated mitochondrial localization of Dok-4 in NF-B activation (Fig. 7A). However, Dok-4-(⌬C234 -325), although intact in its ability to localize in mitochondria, was virtually unable to enhance NF-B activation, suggesting a role for the COOH-terminal portion of the molecule in this function (Fig. 7A). We found that Dok-4-(⌬N11-29) did not increase but slightly decreased TNF-␣-mediated NF-B activation, and we could not detect ROS production in Dok-4-(⌬N11-29)-transfected cells, sugimmunoprecipitates. No difference in the amount of expression of c-Src in c-Src immunoprecipitates (middle) was observed, and we could not detect any band of myc-Dok-4 WT in control rabbit IgG immunoprecipitates (top). *, nonspecific band. B, BAECs were transfected with pcDNA myc-Dok-4 WT or empty vector (pcDNA3.1-myc) (middle), and Myc-tagged Dok-4 proteins were immunoprecipitated with monoclonal anti-Myc antibody reversely. Immunoblotting with anti-c-Src (top) antibody was performed to demonstrate the occurrence of co-immunoprecipitation with Src. No difference in the amount of c-Src was observed by Western blot analysis with anti-c-Src antibody (bottom). C, BAECs were transfected with pcDNA myc-Dok-4 WT, Dok-4-(⌬N1-99), or empty vector. After 24 h of transfection, extra-and mitochondrial fractions were isolated, and each fraction was immunoprecipitated (IP) by anti-Myc antibody. Immunoblotting with anti-c-Src (top) antibody was performed to demonstrate the occurrence of co-immunoprecipitation with Src in the mitochondrial fraction but not in the extra-mitochondrial fraction. D, BAECs were transfected with pcDNA myc-Dok-4 WT construct (ϩ) or empty vector (Ϫ), and Western blot analysis of mitochondria and extra-mitochondrial fraction with anti-Src was performed (upper). We used the same cell fraction as we described in Fig. 1D (lower). Densitometric analysis of c-Src expression in mitochondria is shown. Results were normalized by arbitrarily setting the densitometry of one of the empty vector transfected cells to 1.0 (shown is mean Ϯ S.D., n ϭ 3). E, because the transfection efficiency of plasmid siRNA in BAECs was not sufficient to perform inhibitory studies, we transfected Dok-4 siRNA in BLMECs, which showed over 80 -90% transfection efficiency as described previously (12). gesting that mitochondrial localization of Dok-4, which requires the mitochondrial targeting sequence, is critical for Dok-4-mediated ROS production and NF-B activation.
To determine the importance of TNF-␣ and Dok-4-mediated ROS production in NF-B activation, we examined the effect of Mn(III)tetrakis(1-methyl-4-pyridyl)porphyrin (MnTMPyP), a su- peroxide dismutase/catalase mimetic, on the TNF-␣ and/or Dok-4-induced NF-B activation in BAEC. For these experiments, we avoided using the compounds N-acetyl-L-cysteine and pyrrolidine dithiocarbamate because they have been reported to inhibit TNF-␣-mediated NF-B activation independently of their antioxidant effect (30). As shown in Fig. 7B, MnTMPyP could not inhibit TNF-␣-mediated NF-B activation, consistent with data reported previously (30). MnTMPyP did not show any inhibitory effect on Dok-4-mediated NF-B activation. In contrast, we found a significant inhibitory effect of MnTMPyP on NF-B activation in- duced by a combination of TNF-␣ stimulation and Dok-4 overexpression (Fig. 7B). These data were consistent with those presented in Fig. 5 in that Dok-4 could enhance TNF-␣-mediated ROS production and NF-B activation, but did not do so in the absence of TNF-␣ stimulation.
To evaluate the impact of Dok-4 on other TNF-␣-triggered signals in endothelial cells, we determined TNF-␣-mediated AP-1 and Elk-1 activation in empty vector or Dok-4 WTtransfected cells. As shown in Fig. 7C, we detected a small but significant activation of AP-1 in Dok-4-transfected cells, but we could not detect any significant effect of Dok-4 on basal or TNF-␣-stimulated Elk-1 activation. These data suggest that the effect of Dok-4 is limited to specific signaling pathways.

Role of Endogenous Dok-4 in TNF-␣-induced NF-B Activa-
tion-By having determined the effect of overexpressed Dok-4 in endothelial cells, we wanted to confirm the role of endogenous Dok-4 in these cells. In order to do this, we examined whether deletion of Dok-4 by siRNA could inhibit NF-B activation. Because the transfection efficiency of plasmid siRNA in BAECs was not sufficient to perform the inhibitory studies, we transfected Dok-4 siRNA in BLMECs, which showed over 80 -90% transfection efficiency as described previously (12), and we determined the inhibitory effect of Dok-4 siRNA. As shown in Fig. 3E, we found that Dok-4 siRNA, but not control siRNA, inhibited Dok-4 expression. We next examined the impact of Dok-4 siRNA on TNF-␣-mediated NF-B activation. As shown in Fig. 8A, Dok-4 siRNA significantly inhibited activation of that Dok-4 serves as an adapter for recruitment of Src to mitochondria.
We have also investigated the physiological function of the Dok-4-mediated recruitment of c-Src to mitochondria. First, we found that Dok-4-induced recruitment of c-Src to mitochondria leads to decreased complex I, and we enhanced TNF-␣-mediated ROS production in endothelial cells. Second, based on the data from overexpression (Fig. 7A) and siRNA-mediated knockdown (Fig. 8A) of Dok-4 and anti-oxidant (Fig. 7B), Dok-4 enhanced TNF-␣-mediated NF-B activation in endothelial cells in an ROS-dependent manner (Fig. 8C). Finally, dominant negative Src inhibited Dok-4-mediated NF-B activation, suggesting the importance of Src activity induced by Dok-4 to regulate NF-B activation. Taken together, these data suggest that Dok-4-induced recruitment of c-Src to mitochondria contributes to the downstream activation of NF-B through a mechanism that is at least partially ROS-dependent (Fig. 8C).
Of note, although Src activity is largely accepted to regulate

FIG. 8. Dok-4 expression regulates TNF-␣-mediated NF-B activation.
A, Dok-4 siRNA inhibited TNF-␣ (10 ng/ml)-mediated NF-B activation. BLMECs were transfected with pNF-B Luc-plasmid and Dok-4 siRNA, or control siRNA as indicated, and incubated with or without TNF-␣ (10 ng/ml) for 12 h. After 12 h of TNF-␣ stimulation, luciferase NF-B transcriptional activity was assayed as described in Fig. 7. Results are the mean Ϯ S.D. of three independent experiments (**, p Ͻ 0.01). B, DN-Src inhibits Dok-4-mediated NF-B activation. BAECs were transfected with pNF-BLuc-plasmid and pcDNA myc-Dok-4 WT, DN-Src, or empty vector as indicated, and after 48 h of transfection luciferase NF-B transcriptional activity was assayed as described in Fig. 6. Results are the mean Ϯ S.D. of three independent experiments (*, p Ͻ 0.05). C, scheme of the role of mitochondria Dok-4/c-Src in TNF-␣-mediated NF-B activation. Induction of Dok-4 recruits cytosolic c-Src in mitochondria to associate with Dok-4 and inhibits complex I expression via increasing Src kinase activity and then the reduction of complex I accelerates TNF-␣-mediated mitochondria-ROS production and leads to NF-B activation.
ROS production (31), it is impossible to specifically inhibit mitochondrial Src activity without affecting cytosol and membrane Src activity. Therefore, it is difficult to identify the specific role of mitochondrial Src in TNF-␣-mediated ROS production. However, we found that complex I expression was significantly inhibited by Dok-4 induction and that inhibition of Src activity restored this expression, strongly suggesting that Src activity is important in mitochondrial Dok-4-mediated reduction of complex I expression. Because the importance of complex I in regulating mitochondrial ROS production is well established (32)(33)(34), we believe that it is reasonable to speculate that Dok-4/Src regulates complex I expression as well as subsequent ROS production and NF-B activation.
To our knowledge this is the first description of mitochondrial localization for IRS or Dok family members, and it therefore represents an unexpected and important finding. In the case of another Dok family member, Dok-1, it has been reported that translocation to the cell membrane occurs in a phosphatidylinositol 3-kinase-dependent manner (35). In contrast, Noguchi et al. (40) detected constitutive (but adhesion-dependent) membrane association of Dok-1 in serum-starved CHO cells overexpressing Dok-1 protein. In the current study, we used both Myc-and EGFP-tagged Dok-4 protein-expressing endothelial cells and detected localization of Dok-4 by using fluorescence and confocal microscopy with mitochondria-specific constructs (mito-EYFP), and we found exact co-localization of mitochondria and Dok-4. Furthermore, our cell fractionation experiments also showed that the expression of Dok-4 was very high in mitochondria but undetectable in the extra-mitochondrial fraction (Fig. 1D). Taken together, these data strongly support the conclusion that a major pool of Dok-4 resides in mitochondria in endothelial cells.
We found that Dok-4 and -5, but not other Dok family members, including Dok-1, -2, and -3 and IRS-1 and -2, possess a putative mitochondrial targeting sequence in their NH 2 -terminal region. NH 2 -terminal mitochondrial targeting sequence deletion mutants of Dok-4 lose their ability to localize in mitochondria, suggesting the importance of this peptide sequence in subcellular targeting of Dok-4. However, we also found that the PTB domain of Dok-4 was necessary for mitochondrial localization, suggesting that the putative mitochondrial targeting sequence of Dok-4 is not sufficient for proper localization. This is similar to our previously reported findings regarding the requirement for both the PH and PTB domains for membrane localization of Dok-4 in mammalian cells and yeast (4). Notably, evidence of membrane localization in yeast suggested a possible interaction of Dok-4 with membrane phospholipids. However, it remains unclear whether this structural requirement for localization of Dok-4 reflects a need for both domains to engage distinct membrane or mitochondrial determinants or whether it reflects the intramolecular stabilization of one domain by the other. Indeed, the interactions involved in localization of Dok-4 at the membrane and in mitochondria may well be distinct and will require further investigation.
The nonreceptor tyrosine kinase c-Src is a member of a family of nine protein-tyrosine kinases that associate with the cytoplasmic surface of the cellular membrane (36). c-Src is thought to be involved in intracellular signaling pathways mainly as a plasma membrane-associated molecular effector in response to a variety of extracellular stimuli. Recently, it has been reported that c-Src is also located within mitochondria, where it appears to be associated with the inner membrane, indicating that c-Src is imported from the cytosol into mitochondria (9). However, c-Src does not possess a classical mitochondrial import motif (an amphipathic ␣-helix). In this study, we found that Dok-4 contains a putative mitochondrial target-ing sequence in its NH 2 -terminal region that it localizes in mitochondria where it associates with Src. We propose here that Dok-4 is one of the anchoring molecules of mitochondrial Src based on the following data. 1) Overexpression of Dok-4 increases mitochondrial Src, whereas inhibition of Dok-4 expression by siRNA abolishes mitochondrial Src expression. 2) PP2, an Src kinase-specific inhibitor, prevents Dok-4-mediated reduction of complex I expression. 3) DN-Src inhibits Dok-4mediated NF-B activation.
In this study, we also found that Dok-4 is involved in TNF-␣-mediated ROS production in endothelial cells. Of note, Dok-4 overexpression alone (without TNF-␣ stimulation) did not induce a detectable increase in ROS production, but Dok-4 significantly increased ROS production induced by TNF-␣ stimulation (Figs. 5 and 6). The contribution of ROS in NF-B activation is still debatable (30). Because our data suggest that mitochondrial Dok-4-mediated ROS production may have a great impact on NF-B activation (Figs. 7B and Fig. 8B), we postulate that the differences in Dok-4 expression and/or localization between different cell types and conditions may explain this controversy, but this will need further investigation.
The involvement of the mitochondrial electron transport chain in TNF-␣-induced ROS production has been demonstrated in several cell types (37). Two sites of the respiratory chain produce ROS: one is dependent on the auto-oxidation of complex I, whereas the other depends on the auto-oxidation of the unstable complex III (24). In fact we found that overexpression of Dok-4 WT, but not Dok-4-(⌬N1-99), inhibited mitochondrial complex I 39-kDa subunit expression. Complex I is a multisubunit assembly with a characteristic L-shape and contains 46 different subunits with a combined molecular mass of 980 kDa in bovine heart mitochondria (38). It has been reported that complex I inhibitors such as rotenone and rolliniastatin-2 increase growth factors or cytokine-mediated ROS production (24). Genova et al. (34) have reported that the site of ROS production in mitochondrial complex1 is iron-sulfur cluster N2 by using various complex I inhibitors. The 39-kDa subunit is suggested to be important either for the stability of the complex I or for its assembly (38). Because the recruitment of c-Src in mitochondria mediated by Dok-4 induction inhibited complex I 39-kDa subunit expression, it is likely that this changes the function of complex I and enhances TNF-␣-mediated mitochondria-ROS production. Further investigation is necessary to determine the role of mitochondrial Src in TNF-␣/Dok-4-mediated regulation of the mitochondrial electron transport system and mitochondrial ROS production. In particular, the ability of Src to phosphorylate complex I has not been investigated, although phosphorylation of complex I by other kinases has been proposed to regulate its ROS generation (39).