The TGFβ Receptor-interacting Protein km23-1/DYNLRB1 Plays an Adaptor Role in TGFβ1 Autoinduction via Its Association with Ras*

Background: The TGFβ receptor-interacting protein km23-1 plays an important role in TGFβ signal transduction in TGFβ-sensitive epithelial cells. Results: The role of km23-1 in TGFβ activation of Ras/JNK/ERK, as well as TGFβ1 autoinduction, was determined. Conclusion: km23-1 is required for TGFβ1 autoinduction through a Smad2-independent Ras/ERK/JNK pathway. Significance: km23-1 functions as a critical adaptor coupling TGFβ receptors to Ras activation after TGFβ treatment. We have previously elucidated the signaling events that are required for TGFβ1 autoinduction (Yue, J., and Mulder, K. M. (2000) J. Biol. Chem. 275, 30765–30773). Further, we have reported that the TGFβ receptor (TβR)-interacting protein km23-1 plays an important role in TGFβ signal transduction (Jin, Q., Ding, W., and Mulder, K. M. (2007) J. Biol. Chem. 282, 19122–19132). Here we examined the role of km23-1 in TGFβ1 autoinduction in TGFβ-sensitive epithelial cells. siRNA blockade of km23-1 reduced TGFβ1 mRNA expression, as well as DNA binding and transcriptional activation of the relevant activator protein-1 site in the human TGFβ1 promoter. Further, knockdown of km23-1 inhibited TGFβ-mediated activation of ERK and JNK, phosphorylation of c-Jun, and transactivation of the c-Jun promoter. Sucrose gradient analyses indicate that km23-1 was present in lipid rafts together with Ras and TβRII after TGFβ treatment. Immunoprecipitation/blot analyses revealed the formation of a TGFβ-inducible complex between Ras and km23-1 in vivo within minutes of TGFβ addition. Moreover, we demonstrate for the first time that km23-1 is required for Ras activation by TGFβ. Our results indicate that km23-1 is required for TGFβ1 autoinduction through Smad2-independent Ras/ERK/JNK pathways. More importantly, our findings demonstrate that km23-1 functions as a critical adaptor coupling TβR activation to activation of Ras effector pathways downstream.

Smads only appeared to play an indirect role (11). Thus, although the TGF␤1 promoter complex contained JunD, Fra-2, c-Jun, and FosB, neither Smad3 nor Smad4 were detectable at the relevant TGF␤1-inducible promoter site (11). Further, because dominant-negative Smad3 was able to reduce TGF␤1 promoter transactivation, the role of Smad3 in this process was likely mediated by cross-talk with Ras/MAPKs, known to exist at several levels (2,12,22). Further adding to the complexity, the mechanisms underlying TGF␤1 autoregulation are distinct from those for TGF␤3, which were shown to require activation of JNK and p38, among other components (21). Overall, however, because of the critical role that TGF␤ plays in human cancer development and progression, the signaling pathways mediating TGF␤ autoregulation are important to explore. Any novel regulators of these pathways may represent new targets for therapeutic intervention, if they can be leveraged to block TGF␤ production in human cancers that have lost the TGF␤ autocrine inhibitory loop.
Real Time PCR-HaCaT cells were transfected with EV, NC siRNA, km23-1 siRNA, or km23-2 siRNA and treated as for luciferase assays. RNA isolation and complementary DNA synthesis were performed as described previously (38). For real time PCR, experiments were conducted using an ABI-Prism 7700 Thermal Cycler and TaqMan Universal PCR Master Mix (Applied Biosystems, Foster City, CA). Relative expression was calculated for human km23-1 using primer/probe sets (Applied Biosystems). Primers and probes for endogenous control GAPDH and TGF␤1 were purchased from Applied Biosystems. For data analysis, raw C t was first normalized to the housekeeping gene for each sample to obtain ⌬C t . The normalized ⌬C t was then calibrated to control cell samples to obtain ⌬⌬C t .
Electrophoretic Mobility Shift Assays-The cells were transfected with either NC siRNA or km23-1 siRNA. 48 h after transfection, the cells were cultured in SF medium for 1 h. 2 h after TGF␤1 (5 ng/ml) treatment, EMSAs were performed as described previously (20,21). The oligonucleotides used for probe labeling were described previously (20).
Ras Activity Assays-were carried out using a Ras activation assay kit (Millipore) according to the manufacturer's instructions. Briefly, active Ras was pulled down with purified GST-Raf-RBD agarose beads by incubating cell lysates with GST-Raf-RBD prebound to glutathione-Sepharose. Bound proteins were then subjected to 12% NuPAGE gel and blotted with an anti-Ras Ab.
Lipid Raft Fractionation-HaCaT cells were grown to 80% confluence in 100-mm dishes and were rinsed and incubated in SF medium for 1 h prior to treatment with vehicle or TGF␤1 (5 ng/ml) for 15 min. The cells were harvested for lipid raft fractionation analyses as described previously (39). Twelve 1-ml fractions were collected from the top of the tube, and aliquots of the first 10 fractions were analyzed by immunoblotting.

Blockade of Endogenous km23-1 Inhibits TGF␤ Induction of TGF␤1 Gene Expression in HaCaT
Cells-An important property of TGF␤ is its ability to activate its own mRNA expression and thereby increase its own secretion (19). We have previously shown that Ras and ERK/JNK MAPK pathways are essential for TGF␤1 autoinduction (11). Furthermore, our previous results have shown that forced expression of km23-1 induces specific TGF␤ responses, including an activation of JNK and a phosphorylation of c-Jun, suggesting that km23-1 might play an important role in the TGF␤/JNK pathway (27,32). Thus, it is conceivable that the T␤R-interacting km23-1 might be required for TGF␤1 production, which is an important biological response of TGF␤. HaCaT cells are seen as a human model for the response of TGF␤-sensitive epithelial cells to autocrine and paracrine TGF␤ (40,41). Accordingly, we performed real time RT-PCR analyses of TGF␤1 mRNA expression in human HaCaT cells after transiently transfecting with either EV, km23-1 siRNA, or NC siRNA. Knockdown efficiency of km23-1 was confirmed by real time RT-PCR analyses (Fig. 1A). More importantly, as shown in Fig. 1B, in the km23-1 siRNAtransfected cells, TGF␤ autoinduction of TGF␤1 mRNA expression was significantly decreased. In contrast, in both EV and NC siRNA-transfected cells, TGF␤1 stimulated induction of TGF␤1 mRNA expression, confirming the presence of an active TGF␤1 auto-loop. Thus, our results indicate that the T␤R-interacting protein km23-1 is required for TGF␤ autoinduction of TGF␤1 mRNA expression.
km23-1, but Not Smad2, Is Required TGF␤ Induction of TGF␤1 Promoter Activity, Which Can Be Rescued by siRNA-resistant km23-1-Although we have previously demonstrated that TGF␤ activation of Ras/MAPK pathways is critical for TGF␤ induction of the proximal AP-1 site in the human TGF␤1 promoter (11), the role of km23-1 in the regulation of this promoter site has not been explored. Because our results here have shown that knockdown of km23-1 significantly inhibited TGF␤ autoinduction of TGF␤1 mRNA expression, we determined whether blockade of km23-1 had any effect on TGF␤ induction of TGF␤1 promoter activity. The IEC4-1 and Mv1Lu cells were used for these studies because they represent good models for TGF␤ regulation in mammalian cells because of their high level of TGF␤ sensitivity and their well characterized TGF␤ pathways (6,15,42). In addition, the km23-1 siRNAs employed in these studies were previously shown to specifically knock down km23-1 expression in these cell lines (27,32).
Here, we performed luciferase reporter assays after transiently transfecting Mv1Lu cells with phTG5-luc and the indicated amounts of either NC siRNA or km23-1 siRNA. The TGF␤1 promoter reporter phTG5-luc contains a 450-bp fragment encompassing the proximal AP-1 site of the human TGF␤1 promoter (11). As shown in Fig. 2A, in Mv1Lu cells, TGF␤ induced a 5-7-fold increase in phTG5-luc activity when cells were transfected with EV and NC siRNAs, which is quite similar to that previously reported (11,36). More importantly, the ability of TGF␤ to activate the phTG5-luc reporter was reduced to only 3-fold when low and medium concentrations of km23-1 siRNAs (62.5 and 125 ng/cm 2 ) were used. Transcriptional activity was further decreased to 2-fold after the high dose of km23-1 siRNA (250 ng/cm 2 ) was applied. In contrast, although both basal and TGF␤-induced luciferase values were decreased when the concentration of NC siRNAs was increased to 125 ng/cm 2 , the fold induction by TGF␤ was maintained at similar levels. Similarly, TGF␤ induction of phTG5-luc reporter transactivation was inhibited in IEC4-1 cells transfected with km23-1 siRNAs, but there was no such effect in NC siRNA-transfected cells (Fig. 2B). Thus, our results demonstrate that km23-1 is required for TGF␤1 induction of phTG5luc promoter reporter activity.
To confirm that km23-1 siRNA knockdown of TGF␤1 autoinduction was specifically mediated by the km23-1 siRNA, we performed rescue experiments using the phTG5-luc reporter after transiently transfecting Mv1Lu cells with the indicated forms of either NC siRNA or km23-1 siRNA, along with either km23-1-FLAG or siRNA-resistant ⌬km23-1-FLAG. The results in Fig. 2D demonstrate that the inhibition of TGF␤1-dependent TGF␤1 transactivation by km23-1 knockdown could be rescued by the siRNA-resistant pCMV5-⌬km23-1-FLAG, but not by pCMV5-km23-1-FLAG. Therefore, the inhibition of TGF␤1 autoinduction was specifically mediated by the km23-1 siRNA and not by off target effects.
The Smad pathway is crucial to many aspects of the biological actions of TGF␤ (16,44). Although we have shown that the Ras/MAPKs are essential for TGF␤1 autoinduction, Smads3/4 appeared to play an indirect role because they were not present in the relevant TGF␤1 promoter complex (11). However, we had not examined the role of Smad2 in TGF␤ autoinduction in our previous report. Because we have now shown that km23-1 is required for Smad2-depedent TGF␤ signaling (32), here we examined whether Smad2 was required for TGF␤ induction of the human TGF␤1 promoter. To this end, we performed luciferase reporter assays after transiently transfecting Mv1Lu cells with phTG5-luc and the indicated amount of either EV, NC siRNA, or Smad2 siRNA. The results in Fig. 2E demonstrate that blockade of Smad2 had no effect on TGF␤ induction of the human TGF␤1 promoter. Thus, TGF␤ regulation of the proximal AP-1 site in the human TGF␤1 promoter, which requires km23-1, occurs through Smad2-independent pathways. Collectively, our results demonstrate that km23-1 plays an important role in TGF␤1 autoinduction through Smad2-independent pathways.
Blockade of Endogenous km23-1 Inhibits TGF␤1-inducible AP-1 Binding to the Proximal AP-1 Site in the Human TGF␤1 Promoter-Because blockade of km23-1 decreased TGF␤1 induction of phTG5-luc reporter activity, we examined whether blockade of km23-1 had any effect on AP-1 binding to the proximal AP-1 site in the human TGF␤1 promoter. Thus, we performed EMSAs using an oligonucleotide (Ϫ372 to Ϫ345), which spans the proximal AP-1 site in the TGF␤1 promoter as a probe. As shown in Fig. 3A, in the NC siRNA-transfected cells, exogenous TGF␤ induced a significant increase in AP-1 binding activity (lane 3 versus lane 4). However, in the km23-1 siRNA-transfected cells, TGF␤-induced AP-1 binding activity was significantly decreased (lane 5 versus FIGURE 2. km23-1, but not Smad2, is required for TGF␤ induction of TGF␤1 promoter activation, which could be rescued by siRNA-resistant km23-1. A and B, km23-1 is required for TGF␤ induction of TGF␤1 promoter activation in Mv1Lu cells (A) and in IEC4-1 cells (B). The cells were transfected with the indicated amounts of either NC siRNA or km23-1 siRNA along with the phTG5-luc reporter and were treated with TGF␤1 as described under "Materials and Methods." The data plotted are the means Ϯ S.E. of triplicate samples from three independent experiments. C, Western blot analysis to verify that expression of pFLAG-⌬km23-1 was not affected by km23-1 siRNA. 293T cells were transfected with the indicated plasmids, and Western blotting was performed with the indicated Abs. D, inhibition of TGF␤-dependent phTG5 promoter reporter activity by km23-1 knockdown could be rescued by siRNA-resistant km23-1. Mv1Lu cells were transfected with the indicated forms of either NC siRNA or km23-1-siRNA along with either pFLAG-km23-1 or pFLAG-⌬km23-1 and were treated with TGF␤1 as described under "Materials and Methods." Luciferase reporter assays were performed as described under "Materials and Methods." The data plotted are the means Ϯ S.E. of triplicate samples from three independent experiments. E, blockade of Smad2 has no effect on TGF␤ induction of TGF␤1 promoter activation. Mv1Lu cells were transfected with the indicated amounts of either NC siRNA or Smad2 siRNA along with the phTG5 reporter, and luciferase reporter assays were performed as described under "Materials and Methods." The data plotted are the means Ϯ S.E. of triplicate samples from three independent experiments. lane 6). The unlabeled and mutant AP-1 probes were used as the controls (lanes 1 and 2). We have also performed similar EMSAs in IEC4-1 cells. As shown in Fig. 3B, the binding to the proximal AP-1 site in TGF␤1 promoter was strongly induced by TGF␤ in the NC siRNA-transfected cells (lane 3 versus lane 4). In contrast, the AP-1 binding activity induced by exogenous TGF␤ was markedly reduced in the km23-1 siRNA transfected cells (lane 5 versus lane 6). This TGF␤-induced AP-1-DNA binding complex was completely blocked when unlabeled and mutant AP-1 probes were used as controls (lanes 1 and 2), demonstrating that the AP-1-DNA binding is specific. Thus, our results demonstrate that km23-1 is required for TGF␤-inducible AP-1-DNA complex formation at the TGF␤1 promoter.
The ERK Pathway Is Required for TGF␤ Autoinduction and Can Be Blocked by km23-1 Depletion-We have previously reported that TGF␤ can induce AP-1 binding at the proximal AP-1 site of the TGF␤1 promoter to increase TGF␤1 expression through Ras/ERK/JNK pathways (11). To further confirm that the ERK pathway is required for TGF␤ induction of TGF␤1 promoter activity, we employed PD98059 as a selective MEK inhibitor (Fig. 4A). As shown in Fig. 4A, PD98059 effectively suppressed TGF␤ induction of TGF␤1 promoter activity in a dose-dependent manner, demonstrating that the MEK1/ERK pathway is indeed required for TGF␤ induction of TGF␤1 promoter activity.
Because MEK/ERK activation is upstream in the pathway for TGF␤-inducible regulation of TGF␤1 transcription and DNA binding, it was of interest to determine whether blockade of km23-1 affected TGF␤ induction of ERK1/2 activity. Accordingly, we performed Western blot analyses to examine phospho-ERK1/2 expression induced by TGF␤ in HaCaT cells after siRNA knockdown of km23-1. In addition, for comparison and to indicate specificity, we examined the effects of depletion of another isoform of km23, termed km23-2 (24,38). As shown in Fig. 4B, in the NC siRNA-transfected cells, TGF␤ induced a rapid increase in ERK1/2 activation (lane 2, top panel). The levels at 15-30 min after TGF␤ treatment began to decrease (lanes 3 and 4, top panel). Similar results were obtained in the km23-2 siRNA-transfected cells, indicating that km23-2 was not absolutely essential for TGF␤ activation of ERK1/2. In contrast, in the km23-1 siRNA-transfected cells, phospho-ERK levels were significantly decreased at all time points after TGF␤ treatment ( lanes 5-8, top panel). Total ERK1/2 expression was confirmed by Western blot analyses, as shown in the middle panel. The expression of DIC was used as a loading control. Knockdown efficiency of km23-1 and km23-2 in these cells was confirmed by real time RT-PCR (Fig. 4C). Thus, our results demonstrate that km23-1, but not km23-2, is specifically required for TGF␤ activation of ERK in TGF␤-sensitive cells.
The JNK/Jun Pathway Is Required for TGF␤ Stimulation of TGF␤1 Promoter Activity and Can Be Inhibited by km23-1 Depletion-Our previous results have shown that overexpression of dominant-negative MAPK kinase 4 significantly inhibited TGF␤3 induction of TGF␤1 mRNA expression in IEC4-1 cells, suggesting that TGF␤ activation of JNK is required for TGF␤ autoinduction (11). To confirm whether the JNK pathway was required for TGF␤1 induction of TGF␤1 promoter activity in Mv1Lu cells, we employed the pharmacological JNKselective inhibitor SP600125 (Fig. 5A). As shown in Fig. 5A, SP600125 effectively suppressed TGF␤1 induction of TGF␤1 promoter activity in a dose-dependent manner, demonstrating that JNK activity is also required for TGF␤1 induction of TGF␤1 promoter activity.
Next, we performed Western blot analyses to examine phospho-JNK1/2 expression levels induced by TGF␤ in Mv1Lu cells after siRNA knockdown of km23-1. As expected, in the NC siRNA-transfected cells, TGF␤ induced a rapid increase in phospho-JNK1/2 expression (Fig. 5B, top panel, lane 2). However, in the km23-1 siRNA-transfected cells, TGF␤ induction of phospho-JNK1/2 was significantly decreased (lanes 5 and 6, top panel). Total JNK1/2 expression was confirmed by Western blot analyses, as shown in the middle panel. DIC was used as a loading control in the bottom panel as described previously (45). Similar results were also obtained in HaCaT cells (data not shown). Thus, km23-1 is required for TGF␤ activation of JNK1/2 in TGF␤-sensitive epithelial cells.
Because JNKs are the main upstream kinases for Jun phosphorylation (46), we next tested whether depletion of km23-1 had any effect on c-Jun phosphorylation after TGF␤ activation. As shown in Fig. 5C, in the NC siRNA-transfected cells, TGF␤ induced a rapid increase in phospho-c-Jun expression (lane 2). In contrast, in the km23-1 siRNA-transfected cells, TGF␤ induction of phospho-c-Jun was significantly decreased (lane 4). Thus, km23-1 is required for phosphorylation of c-Jun after TGF␤ stimulation in Mv1Lu cells. Because this TGF␤-sensitive cell model employed transiently transfected km23-1 siRNAs, to provide further evidence of a km23-1 requirement for AP-1-dependent TGF␤ autoinduction, we performed phosphoblotting for c-Jun in a model system stably expressing km23-1-siRNAs (Fig. 5D). The RKO cells used displayed constitutive phTG5-luc reporter transactivation and produced high levels of TGF␤1, making exogenous TGF␤ unnecessary (20). As shown in Fig.  5D, the EV and NC siRNA stably transfected cells displayed constitutive phosphorylation of c-Jun (lanes 1 and 2). In contrast, in RKO cells stably transfected with km23-1 siRNA, the phosphorylation of c-Jun was significantly suppressed (lanes 3  and 4). Equal loading was confirmed using DIC, and stable km23-1 knockdown was confirmed using a km23-1-specific Ab (27) (bottom panel) as used in Fig. 5C. Thus, our results further confirm that km23-1 is required for TGF␤ activation of AP-1dependent events previously shown to be essential for TGF␤1 production.
Previous reports have suggested synergies among AP-1 components and Smad proteins for mediating various TGF␤-inducible responses (37). Although we have ruled out an effect of Smad2 in TGF␤1 autoinduction (Fig. 2E), we had previously found an indirect role for Smad3 in mediating TGF␤3 regulation of TGF␤1 expression, presumably through cross-talk among MAPKs and/or AP-1 components (11). For example, Smad3 has been shown to synergize with Jun family members in regulating c-Jun transcription (37). Because we showed that km23-1 is required for TGF␤ activation of JNK/c-Jun, we chose to use a portion of the c-Jun promoter (Ϫ79/ϩ170), previously shown to contain both AP-1 and Smad-binding element motifs that are regulated by TGF␤-inducible effects on Jun and Smad3 (37). As shown in Fig. 5E, in both the EV and NC siRNA-transfected cells, (Ϫ79/ϩ170)-c-Jun luciferase activity was induced by TGF␤ by ϳ6and 9-fold, respectively. In contrast, in the km23-1 siRNA-transfected cells, TGF␤ induction of (Ϫ79/ϩ170)-c-Jun reporter activity was significantly decreased compared with that in the NC siRNA-transfected cells (to levels of only 3-fold). Thus, km23-1 is required for TGF␤ induction of a Smad3-/Jun-dependent promoter known to be involved in TGF␤1 autoinduction. More importantly, our results demonstrate that km23-1 depletion can block even TGF␤-inducible events requiring Smad3.
TGF␤ Regulates Complex Formation between Ras, km23-1, and RII, and km23-1 Depletion Inhibits Ras Activation by TGF␤-It is well documented that T␤Rs are endocytosed via both clathrin-coated vesicles and cholesterol-rich membrane microdomain lipid rafts/caveolae vesicles (47). In addition, a previous report has shown that lipid rafts are required for TGF␤-mediated MAPK activation (39). Here our results have suggested that the T␤R-interacting protein km23-1 regulates both JNK and ERK pathways, suggesting that TGF␤-inducible events upstream of both of these cascades might also require km23-1. Because Ras is one such component and is known to be localized to lipid rafts (48), it was of interest to examine whether endogenous km23-1 might be co-localized with Ras in lipid rafts in the presence of TGF␤. Accordingly, we performed sucrose gradients to separate lipid raft from non-raft membrane compartments, followed by Western blotting with km23-1 antiserum, Ras, and T␤RII. Immunoblotting for endogenous caveolin-1, a marker of membrane rafts, and endogenous EEA1, a marker of non-raft membrane compartments, was performed to identify the correct fractions corresponding to the raft versus non-raft compartments (marked as R or NR in Fig.  6A). Membrane rafts were concentrated in fractions 1-5, and non-raft membrane compartments were concentrated in fractions 6 -10. As shown in Fig. 6A, in the absence of TGF␤ (left panel), the majority of T␤RII (top panel) was present in the non-raft fractions as expected (fractions 6 -9). However, TGF␤ treatment caused notable receptor redistribution by shifting T␤RII to lipid rafts (fractions 1-5) (right panel), consistent with a previous report (39). In terms of km23-1 localization (middle panel), in the absence of TGF␤ (left panel), the majority of km23-1 accumulated in non-raft membranes (fractions 6 -9). However, as early as 15 min after TGF␤ addition (right panel), a portion of km23-1 was present in lipid raft fractions (fractions [1][2][3][4][5]. Similarly, in the absence of TGF␤ (left panel), the majority of Ras was localized in nonlipid raft fractions (fractions 6 -9), whereas TGF␤ treatment caused a portion of Ras to shift to the lipid rafts. Thus, both endogenous km23-1 and Ras are present in raft membranes together with T␤RII in the presence of TGF␤ in TGF␤-sensitive cells.
Because km23-1 is co-localized with RII and endogenous Ras in the raft membrane fractions after TGF␤ endocytosis, we examined whether km23-1 would interact with Ras after TGF␤ treatment. Accordingly, we performed IP/blot analyses, using FIGURE 4. The ERK pathway is required for TGF␤ stimulation of TGF␤1 promoter activity, which can be inhibited by km23-1 depletion. A, the ERK pathway is required for TGF␤ stimulation of TGF␤1 promoter activity. Mv1Lu cells were transfected with phTG5-Luc. 24 h after transfection, the cells were pretreated with the MEK inhibitor PD98059 at the concentrations indicated for 30 min and then incubated in the absence or presence of TGF␤1 for an additional 24 h. The cells were harvested, and luciferase assays were performed as described under "Materials and Methods." B, blockade of km23-1 partially inhibits TGF␤ activation of ERK in HaCaT cells. HaCaT cells were transiently transfected with NC, km23-1, or km23-2 siRNAs and were treated with TGF␤1 (5 ng/ml) for the indicated times. Western blotting was performed using the indicated Abs. C, km23-1 and km23-2 siRNAs selectively and specifically knockdown endogenous km23-1 and km23-2, respectively. The efficiency of km23-1 and km23-2 knockdown was confirmed by real time RT-PCR. The data are representative of three independent experiments.
anti-Ras as the IP Ab and anti-FLAG as the blotting Ab, after transiently transfecting Mv1Lu cells with human km23-1-FLAG in the absence and presence of TGF␤. As shown in Fig.  6B, TGF␤ stimulated a rapid interaction of km23-1 with Ras, occurring as early as 2 min after TGF␤ treatment (lanes 3 and 6). Expression of EV only (lane 1) and the IgG control (lane 7) indicated that the interaction noted is specific for km23-1. Equal loading and expression of endogenous Ras was confirmed by reprobing with an anti-Ras Ab (middle panel). Equal expression of km23-1-FLAG was confirmed by Western blot analyses with an anti-FLAG Ab (bottom panel). Thus, our results demonstrate that km23-1 interacts with Ras in a TGF␤and time-dependent manner in TGF␤-sensitive epithelial cells.
To ensure that the interaction was not the result of overexpression of km23-1, we examined whether endogenous km23-1 and endogenous Ras were present in the same complex after TGF␤ stimulation. To assess this, we performed IP/blot analyses in the absence or presence of TGF␤ without primary km23-1-FLAG transfection. As shown in Fig. 6C, there was minimal interaction between km23-1 and Ras in the absence of TGF␤ (lane 1). However, TGF␤ induced a rapid interaction of km23-1 with Ras at 2-20 min after TGF␤ treatment (lanes 3-7, top  panel). In contrast, the association between km23-1 and Ras was significantly decreased at 30 min after TGF␤ addition (lane 8). The IgG control was negative (lane 2, top panel). Equal load-ing and expression of endogenous Ras and endogenous km23-1 were confirmed by Western blotting as shown in the middle panels. Collectively, our results indicate for the first time that TGF␤ regulates the interaction of km23-1 with Ras in vivo in a time-dependent manner, suggesting that km23-1 plays an important role in TGF␤-dependent Ras/MAPK signaling events.
Because TGF␤ regulated the interaction of km23-1 with Ras in vivo in a time-dependent manner, we determined whether blockade of km23-1 would have any effect on TGF␤-dependent Ras activation. Accordingly, we transiently transfected Mv1Lu cells with either km23-1 siRNA or NC siRNA and then performed Ras activation assays in the absence and presence of TGF␤. As expected for the NC siRNA-transfected cells, TGF␤ rapidly stimulated Ras activation (Fig. 6D, lanes 1-5). However, in the km23-1 siRNA-transfected cells, Ras activation was decreased at all time points after TGF␤ treatment (Fig. 6D,  lanes 6 -10), with respect to those for the NC siRNA. Equal loading and expression of total Ras was confirmed by Western blotting in the middle panel. The results were scanned by densitometry and are expressed graphically in the bottom panel. Thus, we show for the first time that km23-1 is required for Ras activation by TGF␤.

DISCUSSION
TGF␤1 has been reported to autoregulate its own mRNA expression, resulting in increased secretion of the peptide (18, 19). In TGF␤-responsive cells, such autoinduction can amplify the growth inhibitory effects of TGF␤ in an autocrine fashion (2). However, in late stage solid tumors, which have lost growth inhibitory responsiveness to TGF␤, the TGF␤ produced can result in enhanced tumor progression, largely mediated through the paracrine effects of TGF␤ (2). Therefore, any therapeutic approach that can block TGF␤ production should block the pro-oncogenic effects of TGF␤. Along these lines, our previous work in human colon carcinoma cells has shown that c-Fos is a critical target for blocking TGF␤1 secretion, the associated paracrine tumor-enhancing effects, and tumor progression in vivo (20,49). In this report, we have shown that inhibition of km23-1 decreased both TGF␤ induction of the human TGF␤1 promoter and TGF␤1 gene expression. Therefore, targeting km23-1 may represent one approach for blocking tumor progression by reducing TGF␤1 in the tumor microenvironment in vivo.
We have previously reported that the TGF␤ receptor-interacting protein km23-1 plays an important role in TGF␤ signal transduction in TGF␤-sensitive epithelial cells, including its requirement in Smad2-dependent signaling (27,32). However, in these studies the T␤RII-km23-1-Smad2 complexes were found in early endosomes and not in lipid raft compartments as we show here for km23-1 and Ras. Here we report that km23-1 is present in lipid rafts with Ras after T␤RII endocytosis and that knockdown of km23-1 reduces TGF␤ activation of Ras and of both ERK and JNK effector pathways downstream. Further, we demonstrate for the first time that blockade of Smad2 has no effect on TGF␤-inducible regulation of human TGF␤1 promoter transactivation in TGF␤-sensitive cells, indicating that Smad2 is not required for TGF␤1 production. This is consistent with previous work using fibroblasts from Smad2 and Smad3 knock-out mice demonstrating that TGF␤1 autoinduction was Smad3-dependent but did not involve Smad2 (50). Collectively, our results indicate for the first time that the T␤R-interacting protein km23-1 is required for TGF␤1 autoinduction through Ras-and JNK/ERK-dependent pathways that do not involve Smad2.
Cooperative actions of MAPKs and Smad pathways have previously been implicated in mediating TGF␤ responses (2,11,12). In particular, previous results have shown that many levels of cross-talk exist between JNK and Smad3 signaling. For example, JNK-mediated phosphorylation of Smad3 outside the SSXS motif enhances Smad3 nuclear translocation and transcriptional activity in response to TGF␤ (51). In addition to this direct regulation of Smad3 by JNK activity, TGF␤-activated Smad3 can physically interact with and/or cooperate with Jun family transcription factors to activate transcription of TGF␤ target genes (37, 52). Because we now show that blockade of km23-1 inhibits JNK activation by TGF␤, an inhibition of the FIGURE 6. km23-1 is required for Ras activation by TGF␤, with T␤RII, km23-1, and Ras being co-localized in lipid rafts after TGF␤ treatment. A, T␤RII, km23-1, and Ras are co-localized in lipid rafts after TGF␤ treatment. HaCaT cells were incubated with SF medium for 30 min, followed by incubation in the absence (left panel) or presence of TGF␤ (5 ng/ml) for 15 min (right panel). The cell lysates were subjected to sucrose density gradient centrifugation, and endogenous proteins from each sucrose fraction were analyzed by immunoblotting. R indicates lipid raft compartment, and NR indicates non-raft membrane compartment. The results shown are representative of two similar experiments. B, exogenous km23-1 interacts with endogenous Ras in a TGF␤-dependent manner. Mv1lu cells were transiently transfected with either EV (lane 1) or km23-1-FLAG (lanes 2-7) and were treated with TGF␤1 (5 ng/ml) for the indicated times. Top panel, cell lysates were subjected to IP/blot analyses using a Ras Ab or IgG (control) as the IP Ab and a FLAG Ab as the blotting Ab. The same membrane was reblotted with anti-Ras to show equal loading and expression of endogenous Ras (middle panel). Bottom panel, Western blot analysis to demonstrate equal loading and expression of km23-1-FLAG. C, endogenous km23-1 interacts with endogenous Ras in a TGF␤-dependent manner. Studies were performed as for B except that km23-1-FLAG was not transfected and a km23-1 specific Ab was used (27). The data are representative of three independent experiments. D, blockade of km23-1 resulted in an inhibition of Ras activation by TGF␤. Mv1Lu cells were transiently transfected with either NC siRNA or km23-1 siRNA and were treated with TGF␤1 (5 ng/ml) for the indicated times. Top and middle panels, Ras activity assays were performed as described under "Materials and Methods." Bottom panel, plot of densitometry scans of results in top panel. The data are representative of three independent experiments. phosphorylation, nuclear translocation, and subsequent AP-1dependent effects of Smad3 would be expected. Therefore, although cross-talk between Smad3 and JNK/c-Jun activation appears to play a role in TGF␤ autoinduction, km23-1 depletion also appears to inhibit these cooperative effects (Fig. 5, km23-1 was originally described in TGF␤ signaling by its ability to interact with the T␤R complex in TGF␤-sensitive epithelial cells (23). Our previous results have shown that TGF␤ rapidly regulates the interaction of km23-1 with T␤RII (23). However, km23-1 also functions as a dynein light chain that can recruit signaling cargoes for intracellular transport (23,32). In this regard, cytoplasmic dynein is a motor complex that transports membrane vesicles (i.e., endosomes, lysosomes) and diverse motor cargoes along microtubules to the minus ends (53). In addition to binding DIC at distinct regions, dynein light chains have been shown to directly interact with a number of proteins to exert diverse functions (53). Our results here have shown that TGF␤ leads to the recruitment of Ras to km23-1 in a rapid, TGF␤-inducible manner and that km23-1 is essential for TGF␤ activation of Ras. Thus, km23-1 appears to be functioning as a novel adaptor linking T␤Rs to Ras activation after TGF␤ treatment.
Of particular interest with regard to the role of km23-1 as a Ras adaptor are recent reports of a structural homolog of km23-1 in bacteria (MglB) that interacts with a Ras-like small G protein (MglA) (54). km23-1 is actually part of an ancient superfamily that is widely represented in archaea and bacteria (55). This superfamily appears to be involved in regulating NTPase activity. By analogy to MglA/B, km23-1 would be expected to regulate the activity and biological functions of Ras family proteins. In contrast to km23-1, however, MglB functions as a GTPase-activating protein that regulates MglA in bacteria (54,56,57). Given the higher evolutionary level of km23-1, many of its functions would be expected to be far more complex than those of its bacterial counterparts, presumably involving additional positive and negative regulatory factors. Future studies will likely reveal other Ras family members, effectors, and biological processes that are regulated by km23-1.
In addition to the analogous role of km23-1 as a Ras-binding partner with respect to MglA-MglB, other T␤R-interacting proteins have been described and shown to activate Ras/MAPK pathways downstream. For example, the T␤RII-interacting protein Daxx has been shown to act as an adaptor linking activated T␤Rs to JNK activation by TGF␤ (58). In addition, it has been reported that TNF receptor-associated protein 6 and TAK1 interact with T␤Rs and function as adaptors in the TGF␤ activation of p38 and JNK (35,59). Previous results have also shown that ShcA acts as a direct link between TGF␤ stimulation and ERK signaling (60,61). However, in contrast to our results here for km23-1, the adaptor protein ShcA was shown to associate specifically and more efficiently with T␤RI than with T␤RII (60). Because km23-1 preferentially associates with T␤RII (23,27), and our results herein have shown that km23-1 is required for TGF␤ induction of ERK, km23-1 may function upstream of the ShcA-mediated activation of ERK. Overall, however, from the data presented here, km23-1 appears to be a novel linker between T␤R activation and Ras signaling.
In summary, the TGF␤ receptor-interacting protein km23-1 plays a novel role in Ras/MAPK signaling in TGF␤-sensitive epithelial cells via the following mechanism. In the absence of TGF␤, both Ras and km23-1 are present in non-raft membrane compartments. Once T␤Rs are activated and internalized, km23-1 acts as an adaptor linking T␤Rs to Ras. In membrane rafts, TGF␤ activates Ras, followed by activation of both JNK and ERK. These cascades, in turn, stimulate Ras/MAPK-mediated TGF␤ responses, such as AP-1-dependent transcriptional events. Here we have provided evidence of TGF␤-autoinducible regulation of the AP-1 site in the TGF␤1 promoter to induce TGF␤1 gene expression as an example of one TGF␤ response that is regulated by this mechanism. Although Smad2 was not involved in this pathway, the indirect role of Smad3 was also abrogated by the inhibition of km23-1. Overall, then, using TGF␤1 autoinduction as a representative TGF␤-activated Ras/ MAPK response, we have shown for the first time that inhibition of the T␤R-interacting protein km23-1 can be used as a strategy to curtail Ras/MAPK/AP-1-mediated events induced by TGF␤.