Identification of a Novel Link between the Protein Kinase NDR1 and TGFβ Signaling in Epithelial Cells

Transforming growth factor-beta (TGFβ) is a secreted polypeptide that plays essential roles in cellular development and homeostasis. Although mechanisms of TGFβ-induced responses have been characterized, our understanding of TGFβ signaling remains incomplete. Here, we uncover a novel function for the protein kinase NDR1 (nuclear Dbf2-related 1) in TGFβ responses. Using an immunopurification approach, we find that NDR1 associates with SnoN, a key component of TGFβ signaling. Knockdown of NDR1 by RNA interference promotes the ability of TGFβ to induce transcription and cell cycle arrest in NMuMG mammary epithelial cells. Conversely, expression of NDR1 represses TGFβ-induced transcription and inhibits the ability of TGFβ to induce cell cycle arrest in NMuMG cells. Mechanistically, we find that NDR1 acts in a kinase-dependent manner to suppress the ability of TGFβ to induce the phosphorylation and consequent nuclear accumulation of Smad2, which is critical for TGFβ-induced transcription and responses. Strikingly, we also find that TGFβ reciprocally regulates NDR1, whereby TGFβ triggers the degradation of NDR1 protein. Collectively, our findings define a novel and intimate link between the protein kinase NDR1 and TGFβ signaling. NDR1 suppresses TGFβ-induced transcription and cell cycle arrest, and counteracting NDR1's negative regulation, TGFβ signaling induces the downregulation of NDR1 protein. These findings advance our understanding of TGFβ signaling, with important implications in development and tumorigenesis.


Introduction
The transforming growth factor beta (TGFb) family of cytokines regulates a wide array of biological responses that are critical for proper development and homeostasis [1,2,3,4]. Deregulation of TGFb-mediated responses contributes to the pathogenesis of diverse disease processes from pulmonary and renal fibrosis to cancer [5,6,7,8,9]. A widely studied and key biological effect of TGFb is the inhibition of hematopoietic and epithelial cell proliferation [10,11,12,13], which has important consequences in cancer biology. Several types of carcinomas acquire resistance to TGFb-induced cell cycle arrest, leading to uncontrolled cell proliferation [10,11,12,13,14].
TGFb ligands form heteromeric complexes with type I and II transmembrane TGFb receptors, which have intrinsic serine/ threonine kinase activities [15,16,17,18,19]. The type II kinase transphosphorylates the type I receptor in a glycine-serine rich motif, thereby stimulating the type I kinase activity [20,21,22]. The Smad family of intracellular signaling proteins is critical for transducing TGFb signals from the cell surface to the nucleus to regulate gene expression and consequent cellular processes [7,23,24]. In particular, the TGFb-stimulated type I receptors associate and phosphorylate the receptor-regulated Smad (R-Smad) proteins Smad2 and Smad3 on the C-terminal two serine residues in the SSXS motif [23,24,25,26]. The phosphorylated R-Smads then form a heteromeric complex with the common partner Smad4, and the R-Smad/Smad4 complex accumulates in the nucleus and binds to specific binding elements within promoters of TGFb responsive genes [26,27,28]. The R-Smad/ Smad4 complex acts together with other proteins to induce or repress transcription of responsive genes [29,30,31].
NDR1 is a member of the evolutionary conserved NDR (nuclear Dbf2-related) family of serine-threonine kinases that form a subgroup of AGC kinases [41]. NDR1 and the closely related family member NDR2 regulate critical cellular processes including cell proliferation, apoptosis and differentiation [42,43,44,45,46]. The expression of NDR kinases is deregulated in carcinomas including breast, lung and prostate cancer [47,48]. Interestingly, NDR kinases have been proposed to harbor positive or negative roles in tumorigenesis [47,48]. Whether these kinases regulate specific signaling pathways has remained largely unexplored [48].
Here, we identify NDR1 as a novel SnoN-interacting protein. We find that NDR1 inhibits TGFb-induced transcription and cell cycle arrest. NDR1 inhibits Smad2 phosphorylation, providing the basis for NDR1 regulation of TGFb responses. Remarkably, TGFb reciprocally promotes the degradation of NDR1, thereby providing a counterbalance to NDR1-inhibition of TGFb signaling. Collectively, our findings point to a novel and intimate link between the protein kinase NDR1 and TGFb signaling, with profound effects on the regulation of gene expression and cell proliferation.

NDR1 Associates with the TGFb Signaling Protein SnoN
To gain new insights into the signaling mechanisms that control TGFb responses, we focused on identifying proteins that interact with SnoN, a key component in TGFb signaling. We used a tandem affinity purification (TAP) approach to immunopurify SnoN in human HaCaT keratinocytes, in which we stably expressed the double epitope-tagged version of SnoN (FLAG, HA-SnoN). To identify true SnoN associated proteins, we used cells expressing epitope-tagged SnoN at levels equivalent to those of endogenous SnoN ( Figure 1A). Interestingly, stable expression of SnoN reduced the level of endogenous SnoN in these cells, further normalizing the level of SnoN between SnoN-expressing and control vector-transfected cells ( Figure 1A, compare endogenous SnoN in lane 1 and exogenous SnoN in lane 3). Exposure of HaCaT cells to TGFb led to the downregulation of endogenous as well as stably expressed SnoN, suggesting that TGFb signaling behaves normally in HaCaT cells expressing epitope-tagged SnoN [37,38,49,50]. We performed tandem affinity purification (TAP) by sequential FLAG and HA immunoprecipitation of lysates of epitope-tagged SnoN-expressing HaCaT cells and control HaCaT cells followed by mass spectrometry of immunocomplexes [51,52]. SnoN and known SnoN-interacting proteins including Ski and Smad4 were immunopurified from SnoN-expressing cells, confirming the validity of the purification procedure (Table 1). We also identified novel SnoN-interacting proteins (Table 1). Among these proteins, we focused on the protein kinase NDR1 (also known as STK38).
NDR1 and its close relative NDR2 regulate several biological processes including cell proliferation, apoptosis and differentiation [42,43,44,45,46]. However, whether these kinases regulate specific growth factor signaling pathways has remained incompletely understood. Therefore, we further characterized the interaction of NDR1 with the TGFb signaling protein SnoN. In co-immunoprecipitation assays, we confirmed that SnoN and NDR1 formed a complex ( Figure 1B, C). We next used the fusion of NDR1 or SnoN with the Renilla luciferase (Rluc) protein to assess the interaction of endogenous SnoN or endogenous NDR1, respectively, using Renilla luciferase activity as the readout [53]. We found that endogenous NDR1 robustly interacted with Rluc-SnoN ( Figure S1A). Likewise, endogenous SnoN strongly interacted with Rluc-NDR1 ( Figure S1B). Consistent with these results, endogenous NDR1 formed a complex with endogenous SnoN in the absence or presence of TGFb in 293T cells ( Figure S1C). Together, these data suggest that NDR1 associates with SnoN in epithelial cells.
To gain further insight and evidence for the specificity of the SnoN-NDR1 association, we mapped the structural determinants of SnoN that are required for its interaction with NDR1 ( Figure 1D). We used a series of Rluc-SnoN mutants in coimmunoprecipitation assays to enable quantitative assessment of the effect of mutations of SnoN on its interaction with expressed NDR1 in 293T cells. Removal of the N-terminal 96 amino acids did not affect SnoN's association with NDR1 ( Figure 1E). Consistent with these results, the N-terminal 96 amino acids of SnoN failed to associate with NDR1. Interestingly, we found that deletion of amino acid residues 366-684 or 1-366 decreased significantly the ability of SnoN to interact with NDR1 ( Figure 1E).
We also identified the regions in NDR1 that specify its association with SnoN ( Figure 1F). We used the Rluc-NDR1 fusion and its mutants in these experiments. Deletion of the Nterminal 82 amino acid residues reduced the ability of NDR1 to associate with SnoN ( Figure 1G). Interestingly, deletion of the Cterminal regulatory region alone or together with the N-terminal domain dramatically increased the ability of NDR1 to coimmunoprecipitate with SnoN ( Figure 1G). Together, these data suggest that a region within the kinase domain specifies the association of NDR1 with SnoN, and the C-terminal regulatory region may interfere with the NDR1-SnoN interaction.

NDR1 Regulates TGFb-dependent Transcription
The finding that NDR1 associates with SnoN raised the important question of whether NDR1 regulates TGFb signaling. The plasminogen activator inhibitor 1 (PAI-1) is a TGFbresponsive immediate early gene that has been linked to the control of cell proliferation [40,54,55,56]. We characterized the role of NDR1 in TGFb-induced transcription employing the widely used 3TP-luciferase reporter gene containing TGFbresponsive promoter elements of the PAI-1 gene [56]. We first determined the effect of inhibition of endogenous NDR1 on TGFb-induced transcription. We used RNA interference (RNAi) to induce knockdown of endogenous NDR1 in epithelial cells. Two short hairpin RNAs (shRNAs) targeting distinct regions of NDR1 mRNA induced efficient knockdown of NDR1 protein in 293T cells ( Figure S2A). Importantly, in reporter assays, knockdown of NDR1 using the two shRNAs singly or in combination significantly enhanced the ability of TGFb to induce expression of the 3TP-luciferase reporter gene in HaCaT keratinocytes ( Figure 2A). Knockdown of NDR1 in NMuMG mammary epithelial cells with expression of NDR1 shRNAs also increased TGFb-induced 3TP-luciferase-reporter gene expression ( Figure 2B). In complementary reporter assays, expression of NDR1 reduced in a dose-dependent manner the ability of TGFb to induce expression of the 3TP-luciferase gene in NMuMG cells ( Figure S2B and Figure 2C). Similarly, NDR1 repressed the ability of TGFb to induce the expression of the 3TP-luciferase reporter gene in HaCaT cells ( Figure 2D). Thus, based on knockdown and gain of function analyses, we conclude that NDR1 inhibits the ability of TGFb to induce transcription.
We next characterized the role of NDR1 in the regulation of TGFb-induced expression of the endogenous PAI-1 gene. As expected, TGFb stimulation of control-transfected NMuMG cells increased the abundance of endogenous PAI-1 mRNA as assessed by quantitative real time PCR. Knockdown of endogenous NDR1 in NMuMG cells significantly enhanced the ability of TGFb to increase the abundance of PAI-1 mRNA ( Figure 3A). We also measured the ability of TGFb to induce PAI-1 expression in NMuMG cells stably expressing wild type NDR1 (WT) or a kinase-inactive version of NDR1 in which Lysine 118 was mutated to arginine (KR) (Fig. 3B). The expression of wild type NDR1 blocked TGFb-induced PAI-1 mRNA expression in NMuMG cells ( Figure 3C). In contrast, the kinase-inactive NDR1 enhanced the ability of TGFb to increase the abundance of PAI-1 mRNA ( Figure 3C). These data suggest that the kinase-inactive NDR1 enhanced TGFb-induced PAI-1 gene expression by acting in a dominant negative fashion to block the ability of endogenous NDR1 to antagonize TGFb-induced transcription. Together, Lysates of 293T expressing MYC-SnoN alone or together with HA-NDR1 were subjected to immunoprecipitation with the HA antibody followed by immunoblotting with the SnoN or HA antibody. Total lysates were also subjected to immunoblotting with the SnoN or actin antibody, the latter to serve as a loading control. C. Lysates of 293T cells expressing HA-NDR1 alone or together with MYC-SnoN were subjected to immunoprecipitation with the SnoN antibody followed by immunoblotting with the HA or SnoN antibody. Lysates were also immunoblotted with the HA or actin antibody. NDR1 formed a complex with SnoN. D. A schematic diagram showing the wild type (amino acid (aa) 1-684) and four deletion mutants of SnoN. The dotted area represents the ski/sno/dac (DACH) domain, the shaded area the SAND domain, and the striped areas the helical dimerization domains [33]. E. Lysates of 293T cells expressing Rluc in fusion with wild type or a series of SnoN mutants, as shown in D, alone or together with HA-NDR1 were subjected to immunoprecipitation with the HA antibody followed by luciferase assays to determine the levels of Rluc-SnoN fusion proteins in the NDR1 immunoprecipitates. Aliquots of cell lysates were also assayed for luciferase activity as a measure of Rluc-SnoN expression. The expression of HA-NDR1 in aliquots of immunoprecipitates (10%) and cell lysates was confirmed by immunoblotting using the HA antibody (data not shown). NDR1-associated Rluc-SnoN luciferase activity was normalized to Rluc-SnoN and NDR1 expression. Data are presented as the mean+SEM (n = 4) of NDR1-associated Rluc activity relative to Rluc activity associated with NDR1 in the case of the wild type Rluc-SnoN fusion protein. F. A schematic diagram showing the wild type (aa1-465) and three deletion mutants of NDR1. The dotted area represents the N-terminal regulatory domain, the shaded area the kinase domain, and the striped area the C-terminal regulatory domain. G. Lysates of 293T cells expressing Rluc in fusion with wild type or a series of NDR1 mutants, as in F, alone or together with MYC-SnoN, were subjected to immunoprecipitation using the MYC or SnoN antibody. Immunoprecipitates and cell lysates were subjected to luciferase assays, SnoN immunoblotting (Data not shown), and data analyses as described in E. Data are presented as the mean+SEM (n = 7) of SnoN-associated Rluc activity expressed relative to SnoN-associated Rluc activity in the case of wild type NDR1-Rluc. *, or *** indicates significant difference as compared to wild type SnoN-Rluc (E) or NDR1-Rluc (G) at p,0.05 or p,0.001, respectively (ANOVA). doi:10.1371/journal.pone.0067178.g001 these data suggest that NDR1 acts in a kinase-dependent manner to negatively regulate TGFb-induced transcription.

NDR1 Antagonizes TGFb-induced Cell Cycle Arrest
The finding that NDR1 inhibits TGFb-induced gene expression led us to ask whether NDR1 might also impact biological processes regulated by TGFb. NMuMG cells undergo growth arrest in response to TGFb stimulation [57]. We analyzed the growth rate curves of NMuMG cells stably expressing wild type NDR1 or kinase-inactive NDR1 protein (KR). Cells were plated and left untreated or incubated with TGFb and counted after one, two or three days ( Figure 4A, B). As expected, TGFb reduced the population growth of control-transfected cells ( Figure 4A, B). However, expression of wild type but not the kinase-inactive NDR1 significantly inhibited the ability of TGFb to suppress the population growth of NMuMG cells ( Figure 4A, B). These data suggest that NDR1 impairs TGFb-induced cell cycle arrest in a kinase-dependent manner. NDR1 similarly opposed TGFb suppression of population growth of NMuMG cells when we employed automated cell counts using the Cellomics KSR instrument whereby NMuMG cells were labeled with the DNA dye bisbenzimide (Hoechst) ( Figure S3A). To determine if NDR1 regulates the ability of TGFb to control cell proliferation, we performed BrdU incorporation assays ( Figure 4C). As expected, TGFb markedly reduced the ratio of BrdU-labeled cells, suggesting that TGFb induces cell cycle arrest in NMuMG cells ( Figure 4D). The expression of wild type NDR1, but not kinaseinactive NDR1, opposed TGFb-suppression of BrdU incorporation in NMuMG cells ( Figure 4C, D).
In complementary studies, we found that knockdown of NDR1 enhanced the potency of TGFb to induce 50% reduction in the population growth of NMuMG cells (EC50) ( Figure 4E, F and Figure S3B, C). Interestingly, knockdown of NDR1 together with knockdown of the closely related protein NDR2 further enhanced the potency of TGFb to induce cell cycle arrest in NMuMG cells ( Figure 4E, F, Figure S3C). Collectively, our data suggest that NDR antagonizes the ability of TGFb to inhibit cell proliferation.

NDR1 Suppresses TGFb-dependent Smad2 Phosphorylation
The ability of NDR1 to oppose TGFb-dependent transcription and cell cycle arrest raised the question of the mechanism by which NDR1 exerts this effect. TGFb-dependent phosphorylation and consequent nuclear accumulation of the receptor-regulated Smad proteins mediate TGFb-induced transcription and biological responses. We, therefore, determined the effect of NDR1 on the phosphorylation and nuclear accumulation of Smad2 in NMuMG cells upon exposure to TGFb. As expected, TGFb robustly increased the phosphorylation of Smad2 ( Figure 5A, B). Strikingly, we found that wild type NDR1 substantially reduced the ability of TGFb to induce Smad2 phosphorylation in NMuMG cells ( Figure 5). Consistent with these results, wild type NDR1 suppressed TGFb-induced accumulation of Smad2 in the nucleus in NMuMG cells ( Figure S4). By contrast to wild type NDR1, expression of the kinase-inactive NDR1 (KR) enhanced TGFb-induced phosphorylation and nuclear accumulation of Smad2 in NMuMG cells ( Fig. 5 and Fig. S4). Together, these data suggest that NDR1 inhibits the ability of TGFb to trigger the phosphorylation and consequent nuclear accumulation of Smad2 and thereby impairs TGFb-induced transcription.

TGFb Signaling Induces NDR1 Degradation
The identification of NDR1 as a negative regulator of TGFbinduced transcription and cell cycle arrest raised the important question of whether TGFb signaling might in turn influence NDR1. We characterized the effect of TGFb on the abundance of endogenous NDR1 in NMuMG cells. Remarkably, we found that TGFb stimulation reduced the steady-state levels of NDR1 ( Figure 6A, B). Incubation of cells with the TGFb-type I kinase inhibitor SB431542 (TbRI-KI) restored the abundance of NDR1 protein in TGFb-treated cells ( Figure 6A, B). Using quantitative real-time RT-PCR analyses, we found that TGFb did not reduce and instead increased the abundance of NDR1 mRNA, suggesting that TGFb-induced downregulation of NDR1 protein is not due to changes in NDR1 gene expression ( Figure 6C). We next considered the possibility that the downregulation in NDR1 protein in response to TGFb might result from the increased turnover of NDR1 protein. We measured the rate of NDR1 protein turnover in NMuMG cells treated for different times with the protein synthesis inhibitor cycloheximide in the absence (2) or presence (+) of TGFb ( Figure 6D, E). We found that the half-life of NDR1 in control cells was greater than 16 h, suggesting that NDR1 is a relatively stable protein. Stimulation of cells with TGFb reduced the half-life of NDR1 to approximately 8 h, suggesting that TGFb increased the turnover of NDR1 ( Figure 6D, E). To further explore the decrease in steady-state levels of NDR1 by long-term activation of TGFb signaling ( Figure 6A, B), we assessed NDR1 turnover rates in cells that were left untreated or pretreated with TGFb for 24 h prior to the time-course treatment of cycloheximide. Interestingly, we found that pretreatment of cells with TGFb led to eight-fold reduction in the half-life of NDR1, suggesting that prolonged TGFb treatment induced substantial degradation of NDR1 ( Figure 6F, G). Using in vivo ubiquitination assays, we found that NDR1 was conjugated with ubiquitin in 293T cells ( Fig. 6H and Figure S5) [49]. Importantly, expression of a constitutively active TGFb type I receptor, which activates the Smad signaling pathway in the absence of TGFb addition [21], robustly stimulated the ubiquitination of NDR1 in cells ( Figure 6H and Figure S5A). We also found that exposure of 293T cells to the proteasome inhibitor MG132 suppressed the ability of TGFb to reduce the abundance of NDR1 ( Figure 6I and Figure S5B). Together, these data suggest that TGFb signaling induces NDR1 ubiquitination and its  (2), or NDR1 RNAi NDR1i-1, NDR1i-2 plasmid alone or together, were subjected to dual luciferase assays. Data are presented as the mean+SEM (n = 4) of normalized-3TP-luciferase activity expressed relative to that of the untreated control. B. Lysates of NMuMG cells transfected with reporters as in A together with the control RNAi plasmid or the NDR1i-2 plasmid, and analyzed as in A. Data are presented as the mean+SEM (n = 3) of luciferase activity expressed relative to the untreated control. C. Lysates of untreated or TGFb-treated NMuMG cells transfected with reporters as in A, together with a control vector (2) or increasing concentrations of an NDR1 expression plasmid, were subjected to dual luciferase assays and data analysis as in A. D. Lysates of untreated or TGFb-treated HaCaT cells transfected as described for NMuMG cells in C except for using the CMV-b-galactosidase expression plasmid as a transfection efficiency reporter, were subjected to luciferase and b-galactosidase assays. For each experiment, luciferase activity was normalized as in A. Data in C and D are presented as the mean+SEM (n = 5) of 3TP-luciferase activity expressed relative to the untreated control. *, **, or *** indicates significant difference from the TGFb-treated control at p,0.05, p,0.01, or p,0.001, respectively (ANOVA). doi:10.1371/journal.pone.0067178.g002 consequent degradation involving the 26S proteasome ( Figure 6H, I and Figure S5A, B).
Collectively, our study identifies an important functional and regulatory link between NDR1 and the TGFb signaling pathway. NDR1 suppresses TGFb-induced transcription and cell cycle arrest, and to overcome this effect, TGFb promotes the ubiquitination and turnover of NDR1.

Discussion
In this study, we have discovered a critical role for the protein kinase NDR1 in the regulation of TGFb signaling in proliferating cells. We have identified NDR1 as a novel interacting protein with SnoN, a key component of the TGFb signaling pathway. Loss and gain of function analyses reveal that NDR1 suppresses TGFbinduced transcription and cell cycle arrest in epithelial cells. NDR1 inhibits the ability of TGFb to induce the phosphorylation and consequent nuclear accumulation of Smad2, providing the mechanistic basis for NDR1 regulation of TGFb-induced transcription and cellular responses. Remarkably, we have also found that TGFb reciprocally regulates NDR1, triggering the degradation of NDR1. These findings define an intimate link between NDR1 and TGFb signaling, whereby NDR1 inhibits TGFbinduced transcription and cell cycle arrest, and to counteract this effect, TGFb enhances the turnover of NDR1 protein (Figure 7).
The finding that NDR1 antagonizes TGFb-induced cell cycle arrest in epithelial suggests that cancer cells may employ an NDR1-dependent mechanisms to evade the tumor suppressive effect of TGFb. Consistent with this possibility, we have found that knockdown of NDR1 restores the ability of TGFb to induce cell cycle arrest in the human breast MDA-MB-231 carcinoma cells, which are resistant to the TGFb-induced cell cycle arrest ( Figure  S6A, B, C). Thus, deregulation of NDR1 control of TGFb signaling may be relevant in cancer pathogenesis.
The identification of NDR1 as a novel regulator of TGFbinduced transcription advances our understanding of the mechanisms that control TGFb responses. We have found that NDR1 markedly inhibits TGFb-induced cell cycle arrest. In future studies, it will be interesting to determine whether NDR1 modulates other TGFb responses including epithelial-mesenchymal transition, extracellular remodeling, and cell migration, or whether NDR1 specifically regulates cell proliferation.
How does NDR1 inhibit TGFb-induced transcription and cell cycle arrest? We have found that NDR1 strongly inhibits the phosphorylation and the nuclear accumulation of Smad2. The inhibition of Smad2 phosphorylation provides a basis for NDR1inhibition of TGFb-induced transcription and cell cycle arrest. Recent studies suggest that the protein kinase lats, which is related to NDR1, restricts the nuclear accumulation of Smad2 without affecting its phosphorylation [58]. Thus, NDR1 and lats employ distinct mechanisms to regulate TGFb signaling. How NDR1 inhibits Smad2 phosphorylation remains to be characterized. Since the kinase activity of NDR1 is required for its ability to inhibit Smad2 signaling, it will be critical in future studies to identify substrates of NDR1 that lead to the inhibition of Smad2 phosphorylation.
The finding that TGFb triggers the degradation of NDR1 proteins suggests that reciprocal negative feedback regulation of NDR1 and TGFb signaling provides balance in their mutually opposing effects. Intriguingly, TGFb induces the degradation of SnoN, which as we have found in this study interacts with NDR1. The E3 ubiquitin ligases Cdh1-APC, Smurf2, and Arkadia mediate the TGFb-induced ubiquitination and consequent degradation of SnoN [49,50,59,60,61]. In future studies, it will be interesting to determine if these E3 ubiquitin ligases or others induce the ubiquitination of NDR1 in cells upon exposure to TGFb. where GAPDH mRNA was used as an internal control. Data are presented as the mean+SEM (n = 5) of GAPDH-normalized PAI-1 mRNA abundance relative to untreated control. B. Lysates of NMuMG cells expressing FLAG-tagged wild type NDR1 (WT) or kinase-inactive NDR1 in which Lysine 118 is mutated to arginine (KR), or vector control were subjected to immunoblotting using the NDR1 or actin antibody, with the latter serving as a loading control. C. RNA extracts from untreated or TGFb-treated NMuMG cells expressing wild type or kinase-inactive NDR1 or the vector control were subjected to quantitative RT-PCR analysis of PAI-1 and GAPDH mRNA as described in A. Data are presented as the mean+SEM (n = 3) of relative GAPDH-normalized PAI-1 mRNA abundance as in A. *, **, or *** indicates significant difference from the TGFb-treated control at p,0.05, p,0.01, or p,0.001, respectively (ANOVA). doi:10.1371/journal.pone.0067178.g003 Although we have focused our studies on the identification of NDR1 as a novel regulator of TGFb signaling in proliferating cells, our findings may have broader implications for both TGFb signaling and NDR1. In the developing mammalian nervous system, TGFb-Smad2 signaling has been implicated in the control of axon development, whereby Smad2 inhibits axon growth in granule neurons of the rat cerebellar cortex [62]. In view of our finding demonstrating that NDR1 inhibits Smad2 signaling, it will be interesting to determine whether NDR1 promotes axon growth in mammalian neurons. Conversely, recent studies have revealed that NDR1 controls the development of dendrites and synapses in mouse hippocampal neurons [46]. Our finding that TGFb induces the degradation of NDR1 raises the interesting question of whether TGFb might influence these aspects of neuronal morphogenesis.
Our findings have implications beyond cellular development and homeostasis. Since loss of responsiveness to TGFb-induced cell cycle arrest contributes to tumorigenesis [10,11,12,13,14], the identification of a novel role for NDR1 in TGFb signaling suggests that NDR1 may also influence tumor initiation. Notably, NDR1 is upregulated in lung and mammary carcinomas [47,48], raising the possibility that NDR1 might contribute to loss of TGFb responsiveness in these tumors. Thus, our study raises the potential for NDR1 as a target for drug discovery in cancer biology.

Plasmids
A pCaGiP vector was used to generate FLAG, HA double epitope-tagged SnoN or FLAG epitope-tagged NDR1 stable expression constructs, where a bicistronic transcript containing an internal ribosomal entry site (IRES) encoded the puromycin resistance marker and the protein of interest [39,63,64]. FLAG, HA-tagged SnoN containing nucleotides to express FLAG, Tobacco Etched Virus (TEV) protease site (ENLYFQG), and HA peptides upstream of the SnoN cDNA was generated using a polymerase chain reaction (PCR)-based cloning approach. Expression vectors to express fusion proteins of Renillla luciferase (Rluc) with wild type or deletion mutant SnoN were generated by PCR-based amplification and subcloning of the Rluc cDNA upstream of SnoN cDNA in CMV-based (pCMV5B) SnoNexpression vectors [37,49,50,65]. The NDR1 cDNA product of PCR amplification of epithelial cell-derived polyA-cDNA using NDR1gene-specific oligonucleotides was used to generate HAand FLAG-tagged NDR1 expression vectors (pCMV5B and pCaGiP). Constructs expressing Rluc in fusion with wild type or deletion mutant NDR1 were generated as described for Rluc-SnoN. NDR1 and NDR2 RNA interference (RNAi) plasmids were constructed using the pU6/CMV/enhanced green fluorescent protein (GFP) expression control vector, with NDR1 or NDR2 shorthairpin RNAs (shRNAs), and GFP under the control of U6 and CMV promoters, respectively [40]. Two shRNAs-expressing constructs were generated to target distinct regions in each of NDR1 and NDR2 mRNAs as follows: NDR1i-1, 59GCAACCT-TATCGCTCAACAT39, NDR1i-2, 59GGCAGA-population growth by TGFb. C. Representative fluorescence microscopy images of untreated or 48h-TGFb-treated NMuMG cells expressing wild type or kinase-inactive NDR1 or the vector control that were incubated for 1 h with bromodeoxyuridine (BrdU) and subjected to indirect immunofluorescence using the BrdU antibody and a Cy2-conjugated secondary antibody (green), and labeling with the DNA Hoechst dye (blue) D.  Lysates of untreated or 48h-TGFb-incubated NMuMG cells transfected with an expression plasmid encoding wild type or kinase-inactive NDR1, or transfected with the vector control were subjected to immunoblotting using an antibody that recognizes Smad2 when phosphorylated specifically at the TGFb-induced sites (pSmad2) or an antibody that recognizes Smad2 regardless of its phosphorylation status (Smad2) or with an actin antibody, the latter serving as a loading control. B. Actinnormalized TGFb-phosphorylated Smad2 was expressed relative to the actin-normalized total Smad2. Data are presented as the mean+SEM (n = 3) ratio of TGFb-phosphorylated Smad2 to total Smad2. Statistical significance between TGFb-induced phosphorylation of Smad2 in the vector control cells and each of the wild type and kinase-inactive NDR1expressing cells is indicated (ANOVA). doi:10.1371/journal.pone.0067178.g005

Luciferase Reporter Assays
HaCaT and NMuMG cells were seeded at 2.5 to 3.5610 4 cells/ well in 24-well plates. Cells were co-transfected with the 3TP-Firefly luciferase reporter constructs, the CMV-b-galactosidase or pR-TK Renilla luciferase internal control reporter constructs, together with a control vector or one encoding an NDR1 protein, or with a control or NDR1 RNAi vector, incubated for 16 to 18 h in 0.2% fetal bovine serum-containing medium in the absence or presence of 100 pM TGFb (R & D Systems, Minneapolis, MN), lysed and subjected to single or dual luciferase activity assays [39,40,63,72,75], with each experimental condition carried out in triplicates. Each replicate's arbitrary Firefly luciferase activity, in relative light units, was normalized to its b-galactosidase or Renillaluciferase activity, to control for variations in transfection efficiency.
differences are indicated in B and C as determined by unpaired, two-tailed t-test. D. Lysates of NMuMG cells left untreated or incubated with 10 mg/ ml cycloheximide for different times, alone or together with 100 pM TGFb, were subjected to immunoblotting using the NDR1 or actin antibody. E. Protein abundance of NDR1 in immunoblots, including the one shown in Figure 6D, were quantified and normalized to actin. Data are presented as the mean6SEM (n = 3) of normalized protein abundance of NDR1 expressed relative to that at time 0 for the respective minus or plus TGFb group. Data interpolation indicated that NDR1's half-life was greater than 16 h. TGFb reduced NDR1's half-life to approximately 9 h. F. Lysates of untreated or 24 h-TGFb-preincubated NMuMG cells followed by exposure to cycloheximide for different time points, were subjected to immunoblotting using NDR1 or actin antibody. G. Protein abundance of NDR1 in immunoblots as described and including the one shown in Figure 6F was quantified as described in E. Data are presented as the mean6SEM (n = 4) of relative NDR1 levels. TGFb reduced the half-life of NDR1 from greater than 16 h to approximately 2.5 h. H. TGFb signaling enhances the ubiquitination of NDR1. Lysates of 293T cells expressing FLAG-NDR1, HA-ubiquitin, and constitutively active TGFb type I receptor, were subjected to immunoprecipitation using the FLAG antibody, followed by immunoblotting with the HA or NDR1 antibody. Cell lysates were also immunoblotted with the FLAG or actin antibody. HC refers to the heavy chain of the FLAG antibody. I. Lysates of 293T cells transfected with FLAG-NDR1 alone or together with constitutively active TGFb type I receptor and treated without or with 0.5 mM MG132 (Sigma) for 7 hours were subjected to immunoblotting with the FLAG or actin antibody. *, **, or *** in E and G indicates significant difference from respective control at P,0.05, p,0.01, or p,0.001, respectively (ANOVA). # indicates significant difference from control (p,0.05, unpaired, one tailed t-test). doi:10.1371/journal.pone.0067178.g006 Quantitative Real Time PCR DNase-treated TRIzol (Gibco)-extracted RNA from NMuMG cells cultured in the absence or presence TGFb was reverse transcribed using SuperScript II transcriptase (Invitrogen) and oligo-(dT) [12][13][14][15][16][17][18] (Amersham Biosciences) [40,63,75,76]. The polyA-cDNAs were subjected to quantitative PCR using gene-specific primers for Plasminogen Activator Inhibitor 1 (forward-59TCTCAGAGGTGGAAAGAGCCAG39, reverse-59TGAAG-TAGAGGGCATTCACCAGC39), NDR1 (forward-59ATTTGGTGAGGTACGGCTTG39 and reverse-59CAGG-CAGGAACTCCA TGATT39), and the house-keeping gene glyceraldehyde-3-phosphate dehydrogenase (forward-59TCAA-CAGCAACTCCCACTCTTCCA39 and reverse-59AC-CCTGTTGCTGTAGCCGTATTC A39) using a 2X Sybr Green Mix (BioRad) and Rotor-Gene Thermocycler (Corbett Research). The specificity of the products was confirmed using the melt curve method. Data were analyzed and expressed as described [75].

Microscopy and Cell Proliferation Assays
For fluorescence microscopy experiments, NMuMG cells left untreated or incubated with TGFb for two days were formaldehyde-fixed, and incubated with DNA dye bisbenzimide (Hoechst) (Invitrogen). For indirect Smad2/3 immunofluorescence, untreated or TGFb-treated NMuMG cells were incubated with a mouse Smad2/3 antibody, and a Cy3-labelled anti-mouse antibody (Jackson Laboratories) in the presence of the DNA dye Hoechst. Images of cells were captured using a Kinetic Scan Reader (KSR) (Cellomics, Inc., Pittsburgh, PA) equipped with a Carl Zeiss Axiom x microscope and a charge-coupled device (CCD) digital camera [39,63]. Bromodeoxyuridine (BrdU) assays were carried out using a BrdU assay kit (Roche), fluorescence images were captured using fluorescence microscopy, and data were generated using the Target Activation Bio-Application of the Cellomics KSR [39]. For cell growth rates analyses, cells were grown for one, two or three days in the presence or absence of 100 pM TGFb, and counted using a haemocytometer prior to and at each day during treatment. Alternatively, population cell growth was determined based on nuclei counts in fixed cells stained with the DNA dye Hoechst. Cell counts were normalized to cell numbers before treatment, and replicate values were averaged. For RNAi assays, NMuMG or MDA-MB-231 cells were transfected with GFPexpressing plasmids containing an RNAi control vector, or the NDR1 shRNAs alone or together with NDR2 shRNA expressing vectors as described in the legends of Figure 4, Figure S3, and Figure S6. Untreated or TGFb-incubated NMuMG or MDA-MB-231 cells were fixed and incubated with the DNA dye Hoechst 36 h after ligand treatment. Cells were identified by GFP (green) and nuclei (Hoechst) (blue) fluorescence signals, and the number of GFP-positive cells was assayed using the Target Activation Bio-Application of the Cellomics KSR instrument. Data were analyzed as described in figure legends.

Statistical Analysis
Data were subjected to Analysis of Variance (ANOV) or student t-test as indicated in the figure legends with significant difference set at p,0.05.

Supporting Information
Figure S1 Related to Figure 1. A. Lysates of 293T cells expressing Renilla luciferase (Rluc), alone, or as fusion with SnoN (Rluc-SnoN) were subjected to immunoprecipitation using the NDR1 antibody or IgG immunoglobulins, as a negative control, followed by analysis of immunoprecipitates by luciferase assays (90%) or immunoblotting (10%) with NDR1 antibody (data not shown). Cell lysates were also analyzed by luciferase assays and immunoblotting using NDR1 or actin antibody (data not shown). Endogenous NDR1-associated Rluc or Rluc-SnoN luciferase (IgG-subtracted) were normalized to Rluc or Rluc-SnoN, respectively, and endogenous NDR1 expression. The data are presented as the mean +SEM (n = 3) of NDR1-associated Rluc activity relative to Rluc activity associated with NDR1 in the case of the Rluc control. Rluc-SnoN associated robustly with endogenous NDR1. B. Lysates of 293T cells expressing Rluc or Rluc-NDR1 were subjected to immunoprecipitation using a SnoN antibody or IgG immunoglobulins, as a negative control, followed by analysis of the immunoprecipitates by luciferase assays (90%) or immunoblotting (10%) with SnoN antibody (data not shown). Cell lysates were also subjected to luciferase assays or immunoblotting with SnoN or actin antibody (data not shown). Endogenous SnoNassociated Rluc or Rluc-NDR1 activity was determined as in A. Data are presented as the mean+SEM (n = 4) of SnoN-associated Rluc activity relative to Rluc activity associated with SnoN in the case of the Rluc control. Rluc-NDR1 interacted strongly with endogenous SnoN. C. Lysates of untreated or TGFb-treated 293T cells were subjected to immunoprecipitation using NDR1 antibody or IgG immunoglobulins, as a negative control, followed by immunoblotting with the SnoN or NDR1 antibody. Cell lysates were also subjected to immunoblotting with the SnoN, NDR1 or actin antibody with the latter serving as a loading control. **** in A and B indicates significant difference from the control (p,0.0001, t-test). (TIF) Figure S2 Related to Figure 2. A. Lysates of 293T cells expressing HA-NDR1 in the presence of the control RNAi vector, or NDR1 RNAi NDR1i-1 or NDR1i-2 plasmid were subjected to immunoblotting using the HA or actin antibody, with the latter to serve as a loading control. NDR1i-1 or NDR1i-2 induced 80 to 90 percent knockdown of NDR1. B. Lysates of NMuMG cells transfected with increasing concentrations of a plasmid expressing HA-NDR1 together with the TGFb-responsive 3TP-luciferase reporter and a transfection efficiency vector as described in Figure 2C, were subjected to immunoblotting using the HA or actin antibody. Images in A and B are representative blots from experiments that were repeated at least two independent times. (TIF) Figure S3 Related to Figure 4. A. Population growth of NMuMG cells expressing wild type (WT) or kinase-inactive (KR) NDR1, or control vector (2) after culturing for one, two, or three days in the absence or presence of 100 pM TGFb was determined by subjecting DNA dye (Hoechst)-labeled NMuMG cells to fluorescence microscopy and data analysis using the Cellomics KSR platform and Target Activation algorithm. Percent decrease in population growth of NMuMG cells by TGFb was quantified as described in Figure 4B. Data are presented as the mean+SEM of percent reduction of population growth of NMuMG cells by TGFb from three (day 1 and day 3) or five (day 2) independent experiments. ** or *** indicates significant difference from the respective control within each day at p,0.01, or P,0.001, respectively (ANOVA). B. Representative fluorescence images of NMuMG cells one day post transfection with control RNAi, NDR1i or NDRi plasmids as described Figure 4E, where the DNA dye Hoechst (blue) and GFP (green)-induced signals indicate total NMuMG cells and transfected NMuMG cells, respectively. Analysis of the GFP-labeled cells as compared to total cells using the target activation algorithm indicated approximately 50 percent transfection efficiency for all three sets of transfections. The width of each micrograph corresponds to 330 mm. C. For each experiment including the one shown in Figure 4E, triplicate average of GFP-positive cells at each TGFb concentration was determined. Data are presented as the mean6SEM of relative GFP-positive cell numbers from six (control and NDRi) or five (NDR1i) independent experiments. (TIF) Figure S4 Related to Figure 5. Representative images of untreated or TGFb-treated NMuMG cells expressing wild type or kinase-inactive NDR1 or vector control that were subjected to indirect immunofluorescence using the Smad2 antibody and a Cy3-secondary antibody (red) and labeling with the DNA Hoechst dye (blue), and scanned by fluorescence microscopy. The width of each micrograph corresponds to 330 mm. (TIF) Figure S5 Related to Figure 6. A. Lysates of 293T cells coexpressing FLAG-NDR1 and HA-ubiquitin alone or together with the constitutively active TGFb type I receptor, harboring a mutation in Threonine 204 to aspartate, were subjected to immunoprecipitation using the FLAG antibody followed by immunoblotting with the HA or NDR1 antibody as described in Figure 6H. Ubiquitin-conjugated NDR1 protein species as indicated in and including the protein species in Figure 6H immunoblots were quantified and normalized to NDR1 levels in the immunoprecipitates. Data are presented as the mean+SEM (n = 3) of ubiquitin-conjugated NDR1 species relative to the ubiquitinated NDR1 in cells coexpressing NDR1 and ubiquitin. Significant difference between the two groups was determined using unpaired, two-tailed t-test. B. Lysates of untreated or MG132-treated 293T cells expressing FLAG-NDR1 alone or together with constitutively active receptor were subjected to FLAG and actin immunoblotting as described in Figure 6I. NDR1 protein species as indicated and including the protein species in Figure 6I were quantified and normalized to respective actin. Data are presented as the mean+SEM (n = 6) of NDR1 relative to NDR1 in cells expressing NDR1 alone and left in the absence of MG132. *** indicates significant difference from the control (p,0.001, ANOVA). (TIF) Figure S6 NDR1 knockdown restores the ability of TGFb to inhibit cell proliferation in the breast MDA-MB-231 carcinoma cells. A. Lysates of MDA-MB-231 transfected with a control or NDR1 RNAi plasmids, were subjected to immunoblotting with an NDR1 or actin antibody. Values shown below lanes 1 and 2 represent actin-normalized NDR1 level expressed relative to the actin-normalized NDR1 level in the RNAi control vector transfected cells. B. GFP-expressing and DNA-Hoechst-labeled MDA-MB-231 cells transfected as in A and incubated one day post transfection with 0, 25, 100, or 400 pM TGFb for 72 h were imaged and quantified by fluorescence microscopy and the target activation bio-application, respectively, using the Cellomics KSR as in Figure 4 and Figure  S3. Untreated or TGFb-treated cells were seeded in triplicates or quadruplicates in a 96-well plate, and population growth of GFPpositive cells were averaged. Data are presented as the mean+-SEM of average population growth of GFP-positive MDA-MB-231 cells from five independent experiments expressed relative to the untreated control. C. MDA-MB-231 cells transfected as in A and left untreated or treated with 400 pM TGFb were incubated for the last hour with bromodeoxyuridine, and subjected to immunocytochemistry using a BrdU antibody, fluorescence microscopy and analysis as described in Figure 4. Target activation bioapplication was used to quantify ratio of GFPexpressing BrdU labeled cells, and averages of replicates quantified as in B. Data are presented as the mean+SEM of GFP-expressing BrdU-positive cells from 5 independent experiments. ** or *** indicates significant difference from the control at p,0.01, or p,0.001, respectively (ANOVA). (TIF)