TI - Results and Discussion . AB - To investigate a role of NDRG1 in stabilizing the adherence junction , NDRG1 gene expression was knocked down in the prostate cancer cell lines DU-145 and LNCaP using a U6 based promoter shRNA construct . This led to a decrease in E-cadherin protein levels but not other proteins of the E-cadherin complex investigated ( Figure 1A ) . DU-145 cells transfected with NDRG1 or Flag-tagged NDRG1 constructs when treated with cycloheximide , a protein translation inhibitor , showed an increase in E-cadherin levels as compared to mock controlsNNS (Figure 1B) . This indicates that NDRG1 directly or indirectly stabilizes the E-cadherin protein increasing its longevity in the cell . To investigate whether there is a correlation between NDRG1 and E-cadherin expression in prostate cancer tissues , both proteins were analyzed immunohistochemically on a prostate cancer tissue array by our published methods [18] . A total of thirty two prostate cancer tumors were evaluated . NDRG1 was found to be primarily cytoplasmic with some membrane and nuclear localization . When individual tumors were analyzed for co-expression of the two proteins , a highly significant positive correlation was revealed ( r2 = 08448 , Figure 1C ) . This finding suggests that presence of NDRG1 may be necessary for the stability of the E-cadherin protein in prostate tumors . Intrigued by such a remarkable correlation between the expression of the two proteins we sought to understand the functional relation , if any , between NDRG1 and E-cadherin . Interaction of E-cadherin with the catenins leads to clustering of E-cadherin thereby strengthening adhesion [6] . To understand the possible role of NDRG1 in stabilizing E-cadherin , we first investigated whether the stabilization was due to association of NDRG1 to the E-cadherin complex . None of the proteins of the E-cadherin complex co-immunoprecipitated with flag-tagged NDRG1 in DU-145 and CWR22R prostate cancer cells , suggesting other mechanisms involving NDRG1 may play a role in the stability of E-cadherin ( Figure 2D and data not shown ) . E-cadherin has a short half life of 5-10h . The assembly and turnover of the E-cadherin molecule involves its PHOSphorylation , ubiquitinylation , internalization by endosomes , and subsequent lysosomal or proteasomal degradation or recycle back to the cell surface [19] . This vesicular to and fro movement of E-cadherin from the cell surface to the interior and back is central to the dynamics and stability of the adhesion complex [20] , [21] . Given the fact that NDRG1 is induced by calcium ionophores [22] and E-cadherin turnover is calcium dependent , the location of NDRG1 and E-cadherin in DU-145 cells was investigated after calcium chelation and subsequent recovery in calcium supplemented media by immunoflourescence . Our immunofluorescent data revealed that NDRG1 strongly localizes with E-cadherin during the recovery/recycling phase ( Supplementary Figure S1 ) . However , due to scanty cytoplasm and large nucleus of DU-145 cells it was hard to discern , even with very thin Z-sections , whether the co-localization seen was apparent or real . To confirm whether NDRG1 indeed colocalizes with recycling E-cadherin and also test whether this is true in other cells , live CWR22R prostate cancer cells were labeled with mouse monoclonal antibody that recognizes the extracellular domain of E-cadherin . Cells were labeled on ice to prevent endosomal internalization of the antibody and then immediately chelated with EDTA before placing them back to 37degC . Cells were then restored to calcium supplemented media and immunoprobed at different timepoints with antibody against NDRG1 and secondary anti-mouse antibody to detect endocytosed recycling E-cadherin . Punctate staining for NDRG1 was observed in the cytoplasm . Intense staining for both the proteins was seen near the perinuclear region where the two proteins colocalized . As the cells spread , structures positive for both the proteins resolved into more tubular-vesicular morphology and NDRG1 was seen to colocalize with recycling E-cadherin (Figure 2A) . Vesicular transport in the cell is guided by small ras-like GTPases or RabGTPases of ~20-30 KD that shuttle between the cytoplasm and vesicle membranes . Together the different Rab proteins act as molecular switches that spatially and temporally regulate protein trafficking , recycling and degradation [23] . Endocytosed surface molecules and receptors pass through vesicles decorated by different Rab proteins and effectors before they are degraded or recycled back to the cell surface [24] . Rab11 is known to be associated with recycling E-cadherin [25] . Does NDRG1 play a role in vesicular transport was the next question addressed . A discontinuous sucrose density ultracentrifugation was employed to reveal the localization of NDRG1 in DU-145 cells after calcium chelation using EDTA . Fractions collected and subjected to western blotting revealed that NDRG1 strongly localizes to a membrane organelle in the presence of EDTA ( Figure 2B and 2C ) . These fractions were positive for E-cadherin (Figure 2C) . However , the presence of E-cadherin was evident only when an increased amount of each fraction was loaded indicating all organelles positive for NDRG1 were not positive for E-cadherin . That this is a common feature of all cells and not specific for a particular cell type was confirmed when the same findings held true in adenovirus transformed human embryonic kidney ( HEK293 ) cells (Figure 2D) . To decipher the identity of the organelle to which NDRG1 localized NDRG1 positive fractions were probed with different organelle markers . NDRG1 positive fractions were positive for , Rab4 and Rab11 , ( markers for recycling/sorting endosomes ) and Rab7 ( late endosomal marker ) while negative for Rab5a ( early endosomal marker ) , LAMP1 ( lysosomal marker ) , GRP78 ( endoplasmic reticulum marker ) , and CoxI ( mitochondrial marker ) , ( Figure 2C and D ) . This indicated that NDRG1 containing organelles co-fractionates with recycling/sorting and late endosomes . Also , the distribution of NDRG1 and Rab4a was remarkably similar . In order to investigate whether NDRG1 physically associates with Rab4a or other RabGTPases , HEK293 cells were transfected with NDRG1 Flag constructs . NDRG1 specifically immunoprecipitated with Rab4a and this interaction was sensitive to TritonX100 ( Figure 3A and 3C ) . A reciprocal immunoprecipitation using Rab4a antibody under similar condition when performed and probed for flag-tagged NDRG1 revealed a positive interaction between the two proteins confirming our finding ( Figure 3B ) . None of the other RabGTPases ( Rab5 , Rab11 and Rab7 ) which are located on distinct endosomal membranes immunoprecipitated with NDRG1 (Figure 3A) . We hypothesized that the interaction between NDRG1 and Rab4a may occur on the surface of endosomal membranes rich in disordered lipids that are sensitive to TritonX100 solubilization [26] . To ascertain this , recombinant flag-tagged NDRG1 protein were generated in S2 Drosophila insect cells and purified using a combination of anion-exchange and affinity chromatography . Rab4a was produced as a GST tagged protein in BL21 E.coli cells ( Figure S2A and B ) . RabGTPases shuttle between cytoplasm and endosomal membranes , they are converted to the GTP-bound form by a GTP/GDP - exchange factor and localizes to the membranes of endosomes recruiting Rab effector proteins on the surface of the vesicle . Therefore , recombinant Rab4a was loaded with GTPgammaS and GDP and incubated with M2-agarose bound purified NDRG1flag protein . NDRG1 specifically interacted with GTPgammaS-bound Rab4a and not GDP-bound Rab4a indicating that NDRG1 indeed has an affinity for membrane-localized Rab4a (Figure 3D) . This suggested that NDRG1 is a candidate Rab4a effector protein . To investigate whether this is true in vivo we employed constitutive active and inactive mutants of flag tagged Rab4a . The Rab4aQ67L is GTPase deficient and is constitutively active whereas the Rab4aS22N is GDP-bound and inactive [27] . Immunoprecipitation using these mutants revealed NDRG1 specifically binds to wild type and GTP-bound Rab4aQ67L but does not bind to the GDP-bound Rab4aS22N mutant corroborating our in vitro pull-down experiments ( Figure 4A ) . We further validated these findings by transfecting EGFP fusion constructs of wild type and Rab4a mutants in HEK293 cells stably transfected with NDRG1DsRed2 fusion constructs . Stable expression of NDRG1 as a DsRed2 fusion protein showed NDRG1DsRed2 to be primarily a cytoplasmic protein recruiting to discrete vesicles ranging from 30-300nm in size with an average size of 50nm that had an asymmetric distribution in the perinuclear region ( Figure 4B , Movie S1 ) . NDRG1DsRed2 also localized to membrane ruffles , a structure whose maintenance is closely associated with endosome function ( Figure 4B ) [28] . Vesicular wild type Rab4aEGFP was seen to colocalize with NDRG1DsRed2 at the perinuclear region while dominant negative inactive S22N mutant EGFP protein did not colocalize with NDRG1 protein (Figure 4C) . The constitutively active Q67L mutant EGFP protein on the other hand showed marked colocalization with NDRG1DsRed2 at the perinuclear region (Figure 5A) . For a protein to qualify as a RabGTPase effector it should recognize and interact with GTP bound RabGTPase [23] . Our in vitro and in vivo data indicates that NDRG1 does qualify as a Rab4a effector protein . Rab4a is known to recruit many effector molecules on the surface of endosomes [23] . Our pull-down experiment suggested the likelihood of Rab4a recruiting NDRG1 onto the vesicles . However the fact that the dominant negative S22N mutant failed to solubilize vesicular NDRG1 suggested NDRG1 recruitment onto endosomal membranes might be independent of Rab4a protein . To understand the recruitment of NDRG1 on recycling/sorting vesicles , recycling/sorting vesicles were purified by immunoisolation using Rab11 antibody bound to ProteinA magnetic beads from a population of early/recycling endosomes , purified using flotation gradient ultracentrifugation . 35S-labeled in-vitro translated NDRG1 was incubated with purified recycling endosomes in the presence of GTPgammaS-bound Rab4a . HEK293 cytosol was included to investigate the role of other cytosolic proteins in the recruitment . NDRG1 recruited onto the surface of endosomes independent of Rab4a or other cytosolic proteins (Figure 5B) . This suggests that NDRG1 recruits to recycling vesicles and interacts directly with Rab4a after recruitment . It is also possible that NDRG1 recruits on vesicles after binding to endogenous membrane-bound Rab4a and does not need any exogenous GTPgammaS-bound Rab4a protein for membrane recruitment . We speculated that recruitment of NDRG1 onto endosomal membranes may occur by its interaction with lipids on endosomal membranes . Rab5 has been known to recruit phosphatidylinositol kinases that modify lipids and create membrane domains for recruitment of effector proteins [29] . To investigate this possibility purified NDRG1Flag protein from insect cells was used in a lipid protein overlay analysis . NDRG1 strongly interacted with phosphatidylinositol 4-phosphate (Figure 5C) , a lipid concentrated and maintained by ARF1 in the Trans Golgi Network , and is involved with vesicle formation , transportation and sorting cargo proteins to endosomes [30] . However , NDRG1 did bind weakly to cardiolipin , a lipid found in the inner membrane of the mitochondria and its oxidation is involved in induction of apoptosis [31] . The biological consequence of NDRG1 binding to cardiolipin remains unknown . Thus , this data indicates that NDRG1 recruits onto vesicles by binding to phosphatidylinositol 4-phosphate and interacts with membrane bound Rab4a . Further , localization of NDRG1 in live NDRG1DsRed2-HEK293 cells was studied by live cell confocal microscopy . NDRG1DsRed containing vesicles was seen to be motile undergoing both fission and homotypic fusion ( Movie S2 ) . NDRG1 vesicles also assume long tubular shapes ( 5-20 um in length ) and rapidly move from the perinuclear space to the peripheral region near the plasma membrane , a feature that is characteristic of perinuclear recycling/sorting compartment ( Movie S2 ) [32] , [33] . To confirm the involvement of NDRG1 in recycling , NDRG1DsRed2-HEK293 cells were pulsed with Alexa-fluor-488 conjugated transferrin for 5min to load the early endosome and 60min to load the recycling endosomes and followed by live cell confocal microscopy . Vesicular NDRG1DsRed2 specifically interacted with recycling transferrin and there was a spatial difference between the transferrin positive early endosomal vesicles that were localized near the plasma membrane and NDRG1 containing vesicles that were localized in the perinuclear region ( Figure 6A , Movie S3 ) . NDRG1 vesicles positive for transferrin were seen to recycle transferrin back to the cell surface . To understand the role of NDRG1 in the recycling process we employed transferrin recycling assays on NDRG1 knockdown and NDRG1 overexpressing HEK293 cells . Serum starved cells were loaded with biotinylated transferrin for 1h to load the endosomal recycling compartment and recycling was initiated with excess of transferrin ( 1 mg/ml ) . Biotinylated transferrin within the endosomal recycling compartment had a slower recycling rate in NDRG1 knockdown cells as compared to control shRNA vector transfected cells (Figure 6B) . This data was also confirmed when recycled transferrin was compared between NDRG1 knockdown and control transfected cells ( Figure 6B graph ) . However , a difference in recycling rates was evident only at early time points ( 5min and 15min ) and NDRG1 knockdown cells were able to recycle most of the endocytosed transferrin after 30 min . Overexpression of NDRG1 in HEK293 cells increased the rate of transferrin clearance from the endosomal recycling compartment as compared to vector transfected control cells (Figure 6C) . This was also demonstrated by an increased rate of recycled transferrin in NDRG1 overexpressing cells as compared to vector transfected control cells ( Figure 6C graph ) . Thus both our knockdown and overexpression data demonstrates a functional role of NDRG1 in the recycling pathway . A delay in transferrin recycling has been noted after knockdown or knockouts of a number of protein involved with vesicular transport , specifically proteins belonging to the EHD family that localizes to tubular vesicular regions of recycling endosomes [34] . Interestingly , during the revision of this manuscript a report by Taketomi et al. , demonstrated impaired exocytosis and maturation of mast cells in NDRG1 knockout mice . Mast cells from NDRG1 knockout mice displayed 50% less exocytosis and exhibited fewer , smaller and irregular secretory granules as compared to wild type controls [15] . Forced expression of NDRG1 in mast cells by the same group had demonstrated an increase in exocytosis and degranulation [35] . Arguing that the exocytosis process and the endosome recycling pathway shares common protein components and are functionally related , our data demonstrating a delayed kinetics of transferrin recycling in NDRG1 knockdown cells and increase rate of transferrin recycling in NDRG1 overexpressing cells is consistent with the findings of Taketomi et al. , [15] , [36] . As indicated above NDRG1 containing vesicles ranged in size from 30-300 nm . Large vesicle size ( 300 nm ) suggests that besides recycling , NDRG1 may also have secretory function [37] . After extensive literature search and surveying proteomic data of prostasome , a vesicular body secreted by the prostate that helps in sperm motility , NDRG1 was found to be one of the protein components of prostasome besides other proteins involved in vesicular transport [38] . Recently , by confocal microscopy and BRET analysis , NDRG1 was shown to colocalize with APO AI and AII and may be involved with secretion or transport of these lipoproteins [39] . Interestingly , a plant homolog of NDRG1 is also expressed in secretory cells of reproductive tissues [13] . Although this report does not investigate the role of NDRG1 in secretory pathways , it may not be surprising if NDRG1 also has secretory functions as many proteins of the endocytic and exocytic pathways overlap [15] , [36] . Having established that NDRG1 is involved with recycling/sorting endosomes , involvement of NDRG1 with E-cadherin recycling was studied using live cell confocal microscopy . For this purpose an E-cadherinEGFP construct that is known to be functional and which also interacts with cytoplasmic catenins was employed [7] . Transient transfection of E-cadherinEGFP construct in NDRG1DsRed2-HEK293 cells was followed by calcium chelation and recovery in calcium-supplemented media . Live cell confocal images show endocytosed E-cadherinEGFP to be vesicular and are enriched at the perinuclear space before being recycled back to the cell surface . NDRG1DsRed2 vesicles both near the perinuclear space and close to the membrane fuse with E-cadherinEGFP vesicles as they are trafficked back to the cell surface confirming its involvement with recycling E-cadherin ( Figure 7 and Movie S4 ) . Although NDRG1 has been reported to be downregulated in a variety of cancers which includes the cancers of prostate , breast , colon and oesophagus there are also reports that NDRG1 is upregulated in hepatic , pancreatic and kidney cancers [10] , [11] , [40] -[43] . Induction of NDRG1 in these tumors is speculated to be in response to tumor stress or hypoxia and NDRG1 is proposed as a marker of tumor hypoxia [44] . However , in pancreatic cancer , cellular differentiation and not hypoxia was demonstrated to be the determining factor for NDRG1 expression [42] . In renal cancer , induction of NDRG1 in the tumor tissue was restricted to infiltrating macrophages and not cancer cells [43] . Also none of the reports demonstrate experimental overexpression of NDRG1 leading to increased proliferation or invasion in cellular or animal models of these cancers . Mutation status of NDRG1 in tumors overexpressing NDRG1 is lacking , it is also likely that NDRG1 in these tumors may be non functional . Apart from an alpha-beta hydrolase motif , the functionality of which remains questionable [45] , NDRG1 has no known protein motifs that would impart it a function . Germline mutations in NDRG1 causes Charcot-Marie Tooth Disease type 4D , a demyelinating disorder [14] . Myelin biogenesis involves coordinated activities of both the endocytic and the exocytic pathways [46] . NDRG1 mutation may lead to perturbation in one or both the pathways leading to the observed phenotype . In summary , this report provides evidence defining NDRG1 function in vesicular transport . The data presented here indicates that NDRG1 is a Rab4a effector that can localize to the recycling/sorting endosomes . Being a Rab4a effector it is very likely that NDRG1 may also be involved with vesicular transport of other cargo molecules that may account for its metastasis suppressor function . Our report suggests the involvement of NDRG1 in recycling of atleast one metastasis suppressor , the E-cadherin molecule . Also , the fact that NDRG1 and E-cadherin protein levels correlate significantly in tumors from prostate cancer patients renders clinical significance to our in vitro findings .