The metastasis suppressor NDRG1 down-regulates the epidermal growth factor receptor via a lysosomal mechanism by up-regulating mitogen-inducible gene 6

The metastasis suppressor, N-Myc downstream-regulated gene-1 (NDRG1) inhibits a plethora of oncogenic signaling pathways by down-regulating the epidermal growth factor receptor (EGFR). Herein, we examined the mechanism involved in NDRG1-mediated EGFR down-regulation. NDRG1 overexpression potently increased the levels of mitogen-inducible gene 6 (MIG6), which inhibits EGFR and facilitates its lysosomal processing and degradation. Conversely, silencing NDRG1 in multiple human cancer cell types decreased MIG6 expression, demonstrating the regulatory role of NDRG1. Further, NDRG1 overexpression facilitated MIG6–EGFR association in the cytoplasm, possibly explaining the significantly (p <0.001) increased half-life of MIG6 from 1.6 ± 0.2 h under control conditions to 7.9 ± 0.4 h after NDRG1 overexpression. The increased MIG6 levels enhanced EGFR co-localization with the late endosome/lysosomal marker, lysosomal-associated membrane protein 2 (LAMP2). An increase in EGFR levels after MIG6 silencing was particularly apparent when NDRG1 was overexpressed, suggesting a role for MIG6 in NDRG1-mediated down-regulation of EGFR. Silencing phosphatase and tensin homolog (PTEN), which facilitates early to late endosome maturation, decreased MIG6, and also increased EGFR levels in both the presence and absence of NDRG1 overexpression. These results suggest a role for PTEN in regulating MIG6 expression. Anti-tumor drugs of the di-2-pyridylketone thiosemicarbazone class that activate NDRG1 expression also potently increased MIG6 and induced its cytosolic co-localization with NDRG1. This was accompanied by a decrease in activated and total EGFR levels and its redistribution to late endosomes/lysosomes. In conclusion, NDRG1 promotes EGFR down-regulation through the EGFR inhibitor MIG6, which leads to late endosomal/lysosomal processing of EGFR.

The EGFR is a membrane-bound tyrosine kinase that plays a key role in critical cellular programs, including survival, proliferation, and metastasis, with spurious EGFR activation being involved in cellular transformation (23). EGFR activation is prevented by self-inhibitory constraints imposed on the extracellular ligand-binding domain (24) and its intracellular catalytic domain (25). These constraints are liberated by epidermal growth factor (EGF) binding that drives dimerization, allosteric activation of the kinase, EGFR autophosphorylation and downstream signaling (24,26). Interestingly, EGFR signaling is negatively controlled by 1) multiple inducible inhibitors (27,28) and 2) receptor-mediated endocytosis, leading to its internalization and degradation by the lysosomal compartment (29).
The mitogen-inducible gene 6 (MIG6), also known as the receptor-associated late transducer (RALT), or ERBB receptor feedback inhibitor 1 (ERRFI), is a transcriptionally induced EGFR inhibitor that is also a tumor suppressor (30,31). MIG6 is a cytoplasmic protein (32) that binds to the EGFR dimer interface, preventing the formation of asymmetric catalytic dimers, locking it into a catalytically inactive conformation (26). Significantly, MIG6 can also induce internalization and degradation of EGFR via a lysosomal mechanism, which integrates its ability to act to inhibit EGFR catalytic activity and down-regulate its levels (33).
A recently described group of anti-cancer agents of the di-2pyridylketone thiosemicarbazone (DpT) class potently inhibit tumor growth and metastasis at least in part by their ability to up-regulate NDRG1 through a mechanism involving intracellular iron binding (10, 16, 22, 34 -37). The first lead agent of this class of agents, di-2-pyridylketone 4,4-dimethyl-3-thiosemicarbazone (Dp44mT), leads to the marked up-regulation of NDRG1 in many tumor cell types (10,35). An analog of Dp44mT, namely di-2-pyridylketone 4-cyclohexyl-4-methyl-3-thiosemicarbazone (DpC) (36), also potently up-regulates NDRG1 (35). This agent possesses marked anti-tumor activity against a variety of belligerent tumors in vitro and in vivo (35,36,38,39) and has entered Phase I clinical trials for the treatment of advanced and resistant cancer (40). Of interest, agents that bind intracellular iron, such as desferrioxamine (DFO), can also up-regulate MIG6 (41), which could be mediated through an iron-responsive increase in HIF-1␣ levels, which is known to transcriptionally up-regulate MIG6 (42).
Herein, we demonstrate that the mechanism involved in EGFR down-regulation mediated by NDRG1 occurs through it potently increasing MIG6 levels, which both directly inhibits EGFR and facilitates its lysosomal degradation. NDRG1 was shown to increase EGFR internalization and co-localization with the late endosome/lysosomal marker, lysosome-associated membrane protein 2 (LAMP2). Moreover, MIG6 and NDRG1 were demonstrated to be associated and co-localized in the cytoplasm with NDRG1 overexpression increasing MIG6 half-life. Increased EGFR levels after MIG6 silencing was particularly apparent with NDRG1 overexpression, suggesting a role for MIG6 in the NDRG1-mediated down-regulation of EGFR. As an innovative therapeutic targeting strategy, novel anti-cancer drugs of the DpT class that up-regulate NDRG1 expression (10,35) increased MIG6 and induced its co-localization with NDRG1 in the cytoplasm. This led to a decrease in activated and total EGFR levels as well as the redistribution of EGFR to LAMP2-stained late endosomes/lysosomes. In conclusion, NDRG1 promotes MIG6-mediated EGFR down-regu-lation by lysosomal processing, and this could be important for understanding the broad ability of NDRG1 expression to inhibit a plethora of oncogenic signaling pathways.

NDRG1 overexpression up-regulates MIG6 and decreases EGFR total levels and EGFR phosphorylation at Tyr-1068
Our laboratory previously demonstrated that NDRG1 inhibited a plethora of oncogenic signaling pathways by down-regulating a key upstream regulator, EGFR (21). The aim of the current study was to dissect the mechanism by which NDRG1 decreased EGFR expression. Therefore, considering the integral role of MIG6 in this process (33), it was deemed important to investigate the potential effect of NDRG1 on MIG6 expression and its ability to down-regulate EGFR.
Initially, PANC-1 vector control (VC) cells and cells stably transfected to overexpress NDRG1 ( Fig. 1A; N1) were used as they have been well-characterized in our laboratory regarding the ability of NDRG1 to inhibit multiple oncogenic signaling pathways (12) by down-regulating EGFR (21). First, PANC-1 VC and N1 cells were incubated in the presence and absence of the ligand EGF (10 ng/ml) for 2, 5, and 10 min (Fig. 1A). Examining NDRG1 expression in VC cells, two endogenous NDRG1 bands were apparent at 41 and 46 kDa that did not significantly alter upon incubation with EGF relative to the control. The densitometry of NDRG1 shown in Fig. 1A represents the total of all bands. These bands are consistent with the different isoforms of NDRG1 reported previously that may be the result of NDRG1 processing, cleavage, or differential phosphorylation (43,44). The overexpression of NDRG1 in the N1 clone resulted in a third exogenous band at 48 kDa that was above the endogenous 41-and 46-kDa bands (Fig. 1A), which corresponds to NDRG1 with its FLAG-tag (12). The expression of total NDRG1 in the N1 clone was significantly (p Ͻ0.001) greater than that of VC cells at all incubation times (Fig. 1A).
Total EGFR levels in PANC-1 VC cells were slightly decreased by the addition of EGF as a function of incubation time relative to the control (Fig. 1A). Overexpression of NDRG1 in control PANC-1 N1 cells resulted in a significant (p Ͻ0.01) decrease in EGFR compared with the VC control. This decrease in EGFR expression in N1 cells relative to its VC counterparts became more apparent as a function of incubation time with EGF, in agreement with our previous studies (21). In contrast, EGFR mRNA levels were not affected by NDRG1 overexpression in the absence and presence of EGF, as demonstrated using two different NDRG1-overexpressing clones, namely N1 and N2 cells (Fig. S1B). Relative to the effect observed with NDRG1 overexpression, silencing this molecule resulted in up-regulation of EGFR in PANC-1 cells (Fig. S1C), as we also demonstrated in CFPAC-1 cells (21). Furthermore, silencing NDRG1 in HT-29 colon cancer cells also resulted in a similar up-regulation of EGFR (Fig. S1C).
In terms of the effect of NDRG1 expression on EGFR activation, treatment of VC cells with EGF caused pronounced and significant (p Ͻ0.001) increase in pEGFR levels at Tyr-1068 at incubations from 2 to 10 min relative to the 0 min time point (Fig. 1A). Of note, the pEGFR phosphorylation at Tyr-1068 has NDRG1 up-regulates MIG6 to inhibit EGFR expression been demonstrated to activate the Ras/mitogen activated protein kinase signaling pathway, which is implicated in uncontrolled cancer growth (45).

NDRG1 up-regulates MIG6 to inhibit EGFR expression
significantly (p Ͻ0.01-0.05) less marked relative to that observed in VC cells at all time points. In addition, the ratio of pEGFR to total EGFR was markedly and significantly (p Ͻ0.01) increased with increasing EGF incubation time relative to the 0 min control for both VC and N1 cells (Fig. 1A). However, there was no (p Ͼ0.05) significant difference in these ratios at each incubation time when comparing VC to N1 cells. This is in agreement with our previous investigation (21), whereby NDRG1 was able to decrease pEGFR by reducing total EGFR levels, rather than specifically decreasing EGFR phosphorylation.
Examining MIG6 expression in VC cells, incubation with EGF had no significant (p Ͼ0.05) effect relative to the control (Fig. 1A). Interestingly, NDRG1 expression in N1 cells resulted in significantly (p Ͻ0.01-0.05) higher levels of MIG6 relative to the VC at all time points, with EGF having no appreciable effect relative to the 0 min control. Examination of mRNA levels of NDRG1 and MIG6, demonstrated that in contrast to the marked up-regulation of NDRG1 mRNA in N1 cells relative to VC cells, there was no significant (p Ͼ0.05) alteration in MIG6 mRNA levels in the N1 cell type (Fig. S1D). This observation suggests that the effect of NDRG1 on MIG6 was mediated by posttranscriptional regulation. Collectively, the ability of NDRG1 to down-regulate EGFR and decrease pEGFR (Tyr-1068) could be because of its effect on up-regulating protein levels of the EGFR inhibitor MIG6.
To confirm the role of NDRG1 in regulating MIG6 protein levels (Fig. 1A), the effect of NDRG1 silencing (siNDRG1) relative to its negative control siRNA (siControl) was assessed in three different pancreatic cancer cell lines, namely PANC-1, CFPAC-1, and AsPC-1 (Fig. 1B). In all cell types, effective NDRG1 silencing was observed, resulting in marked and significant (p Ͻ0.001) down-regulation of the NDRG1 46-kDa band, but not the 41-kDa band (Fig. 1B). This is in good agreement with previous studies (12,14,43), and indicates the presence of NDRG1 isoforms, with the lower band being consistent with a long-lived cleaved isoform (43,44). The upper NDRG1 band has been suggested to be the active form of the protein for metastasis suppression, as it is potently up-regulated by drugs that inhibit tumor growth and block metastasis (16, 35,43). Moreover, silencing of the top NDRG1 band prevents the ability of NDRG1 to act on downstream effectors (e.g. E-cadherin and ␤-catenin) (14) involved in preventing the epithelial mesenchymal transition and metastasis. The down-regulation of the upper NDRG1 band by siRNA is important to achieve experimentally in terms of preventing its downstream anti-oncogenic functions. In fact, silencing NDRG1 resulted in a significant (p Ͻ0.001-0.05) decrease in MIG6 expression in all three cell types (Fig. 1B). An additional siRNA of NDRG1 was also examined in PANC-1 cells, resulting in a marked decrease in MIG6 expression relative to the control (Fig. S1E). Hence, the regulation of MIG6 occurs through the differential expression of NDRG1.

Lysosomotropic agents that prevent lysosomal acidification perturb pEGFR (Tyr-1068) and MIG6 levels
It has been established that a major process by which EGFR undergoes degradation is through trafficking to lysosomes (29,33,46,47). Considering this, it was important to investigate whether NDRG1 promoted EGFR processing through this mechanism. In these studies, PANC-1 cells were treated in the absence and presence of the well-characterized lysosomotropic agents ammonium chloride (NH 4 Cl; 15 mM) or methylamine (CH 3 NH 2 ; 15 mM) that inhibit lysosomal acidification and function (48) for 24 h/37°C, followed by treatment with EGF (10 ng/ml) for the last 10 min/37°C of the incubation (Fig. 1C).
For VC or N1 cells, incubation with NH 4 Cl or CH 3 NH 2 in the presence or absence of EGF had no significant (p Ͼ0.05) effect on NDRG1 expression relative to the control. However, as observed for Fig. 1A, N1 cells had significantly (p Ͻ0.001-0.01) higher levels of NDRG1 compared with VC cells (Fig. 1C). Examining total EGFR expression, in VC cells, NH 4 Cl or CH 3 NH 2 had no significant (p Ͼ0.05) effect in the presence or absence of EGF. Under all incubation conditions in N1 cells, EGFR expression was significantly (p Ͻ0.05) decreased in the presence and absence of EGF. Treatment of N1 cells with NH 4 Cl or CH 3 NH 2 had no significant (p Ͼ0.05) effect on EGFR levels in the presence of EGF relative to the N1 control without EGF (Fig. 1C).
When assessing the levels of pEGFR (Tyr-1068) in VC cells, the addition of EGF to VC cells caused a significant (p Ͻ0.001-0.05) increase in the levels pEGFR at Tyr-1068 relative to the control under all conditions (Fig. 1C), which is expected in this cell type (21). This response was markedly enhanced when EGF was added to VC cells incubated with NH 4 Cl or CH 3 NH 2 , there being a pronounced and significant (p Ͻ0.01-0.05) increase in pEGFR (Tyr-1068) relative to the VC control (Fig. 1C). The increase in pEGFR (Tyr-1068) levels after incubation with these lysosomotropic agents is probably because of inhibition of the degradation of the phosphorylated receptor, as demonstrated for these agents by others (49). However, examining N1 cells, pEGFR (Tyr-1068) levels were lower under all conditions relative to their VC counterparts (Fig. 1C). Upon incubation of N1 cells with EGF, there was a significant (p Ͻ0.001-0.05) increase in pEGFR (Tyr-1068) under all conditions relative to N1 cells in the absence of EGF. As observed for VC cells, incubation of N1 cells with NH 4 Cl or CH 3 NH 2 significantly (p Ͻ0.05) increased pEGFR (Tyr-1068) in the presence of EGF relative to the EGF control (Fig. 1C). Again, this could be because of the inhibited degradation of the phosphorylated receptor by these agents leading to higher levels (49).
Furthermore, the ratios of pEGFR/EGFR were calculated in both VC and N1 cells with and without treatment with NH 4 Cl or CH 3 NH 2 (Fig. 1C). It was demonstrated that in the absence of EGF, there was a slight increase in the pEGFR/EGFR ratio upon incubation of VC cells with the lysosomotropic agents. Upon the addition of EGF to VC cells, the pEGFR/EGFR ratio was increased compared with their untreated counterparts (Fig.  1C). Further, the addition of lysosomotropic agents to VC cells in the presence of EGF caused a marked and significant (p Ͻ0.05) increase in the ratio relative to the EGF-treated control. Similar to VC cells, the effect of the lysosomotropic agents in the presence of EGF caused a significant (p Ͻ0.05) increase in the pEGFR/EGFR ratio relative to the control in N1 cells (Fig.  1C). These results demonstrate that inhibiting lysosomal degradation increases pEGFR, which is able to accumulate because of its inability to be degraded via the lysosome.

NDRG1 up-regulates MIG6 to inhibit EGFR expression
Examination of MIG6 levels demonstrated that VC cells incubated without or with lysosomotropic agents in the absence or presence of EGF were not significantly (p Ͼ0.05) altered relative to the controls (Fig. 1C). On the other hand, in N1 cells in the absence of EGF, MIG6 levels were significantly (p Ͻ0.001-0.01) enhanced by NDRG1 under all conditions relative to the respective VC controls. Although NH 4 Cl and CH 3 NH 2 had no significant effect on MIG6 expression in the absence of EGF, these agents significantly (p Ͻ0.01-0.05) decreased MIG6 in the presence of EGF in N1 cells (Fig. 1C). Together, these data indicate that the lysosomotropic agents are perturbing pEGFR (Tyr-1068) and MIG6 levels, suggesting potential lysosomal processing that was then investigated further below.

NDRG1 overexpression increases EGFR internalization and co-localization with the early endosomal marker EEA1 and the late endosome/lysosomal marker LAMP2
Considering the lysosomal process that down-regulates EGFR (50, 51), the results above in Fig. 1C suggest lysosomal involvement in the role of NDRG1 in decreasing EGFR expression and activation. To examine this further, PANC-1 VC and N1 cells were incubated in the presence and absence of EGF (10 ng/ml) for 10 min/37°C ( Fig. 2A). Cells were then assessed by immunofluorescence using confocal microscopy and stained for EGFR (red), the well-characterized early endosomal marker, EEA1 (green) (52) and DAPI (blue) for the nucleus.
Considering first nuclear DAPI staining, some slight, but not significant, differences in nuclear size were apparent upon close examination of the images taken. This is probably because nuclear size does naturally vary depending on the stage of the cell cycle (53). As expected in VC control cells, EGFR showed strong red staining on the membrane with some intracellular staining as well ( Fig. 2A). In VC cells incubated with EGF, EGFR expression was again evident on the plasma membrane and intracellularly (Fig. 2Ai). In contrast, in control N1 cells overexpressing NDRG1 in the presence and absence of EGF, EGFR levels were overall significantly (p Ͻ0.01) lower. These findings are in agreement with the Western blot studies in Fig. 1, A and C. In addition, N1 cells incubated in the presence or absence of EGF resulted in greater intracellular clustering of EGFR relative to VC cells ( Fig. 2A). Assessing EEA1 expression in both VC and N1 cells under all conditions, there was no significant (p Ͼ0.05) alteration in the intensity of this marker, which was uniformly distributed in the cytoplasm ( Fig. 2A).
Quantification of co-localization demonstrated increased co-localization of EGFR and EEA1 being evident in yellow puncta in N1 cells treated with EGF ( Fig. 2A). This co-localization was significantly (p Ͻ0.001) greater in the presence of EGF relative to the VC counterpart treated with EGF (Fig. 2Aii). Furthermore, examining the extent of this association using Pearson's correlation coefficient (r), higher co-localization was evident in N1 EGF-treated cells (r ϭ 0.699) compared with EGF-treated VC cells (r ϭ 0.346; Fig. 2A), suggesting a role for NDRG1 in trafficking and sorting of EGFR into the early endosomal compartment.
After being sorted to the endosome, EGFR can be transported to the lysosome for degradation (51). In order to further explore the association of NDRG1 with EGFR and the lysosome, the localization of these proteins was also examined (Fig.  2B). Similarly to Fig. 2A, PANC-1 VC and N1 cells were incubated with EGF (10 ng/ml)/10 min at 37°C, and then examined for the distribution of EGFR, DAPI, and the late endosome/ lysosomal marker LAMP2 (shown in green) (54). Examining EGFR, similar effects were observed in terms of NDRG1 expression to that observed in Fig. 2A, with EGFR intensity being significantly (p Ͻ0.05) decreased in N1 cells relative to VC cells (Fig. 2Bi). Furthermore, intracellular clustering of EGFR was evident in N1 cells relative to their VC counterparts (Fig. 2B). In contrast to EEA1 expression ( Fig. 2A), LAMP2 distribution became markedly more clustered in the N1 cells regardless of EGF and appeared perinuclear (Fig. 2B). However, the overall intensity of LAMP2 in N1 cells was not significantly (p Ͼ0.05) changed relative to its VC counterparts (Fig. 2Bi), suggesting altered distribution rather than expression.
In VC cells in the presence and absence of EGF, minimal co-localization was observed between LAMP2 and EGFR (r ϭ 0.436 -0.445) (Fig. 2Bii). However, examining N1 cells, there was an increase in intracellular EGFR and the clustering of LAMP2-stained late endosomes/lysosomes, as well as co-localization of EGFR and LAMP2 (Fig. 2B). Quantification analysis showed that in N1 cells, co-localization of EGFR and LAMP2 occurred in the presence and absence of EGF and was significantly (p Ͻ0.01) higher than that observed in VC cells (Fig.  2Bii). This was further documented by the increased Pearson's correlation coefficients calculated in N1 cells (r ϭ 0.768 -0.813) compared with VC cells (r ϭ 0.436 -0.445). Together, these findings in Fig. 2, A and B, indicate the role of NDRG1 in fostering the trafficking and sorting of EGFR to the early endosome, followed by the lysosome. Of note, in contrast to the marked co-localization between LAMP2 and EGFR observed in

Inhibition of the proteasome by MG132, but not lactacystin, increases levels of pEGFR (Tyr-1068)
Although the involvement of the proteasome in the degradation of EGFR is not completely understood, there is evidence to suggest its role in inhibiting EGFR signaling. To therefore assess the possible involvement of proteasomal degradation in the NDRG1-mediated inhibition of EGFR expression and signaling, the well-characterized proteasomal inhibitors, MG132 or lactacystin, were used. PANC-1 cells were incubated with MG132 (2.5 M) or lactacystin (5 M) for 24 h/37°C, with EGF (10 ng/ml) being added in the last 10 min of the incubation (Fig. 3).
Incubation with MG132 resulted in the detection of a 47-kDa NDRG1 band in VC cells which is slightly above the 46-kDa NDRG1 band observed in VC cells in the absence of MG132 (Fig. 3A). As reported above (Fig. 1A), the 48-kDa NDRG1 band was prominent in N1 cells (Fig. 3A). Upon incubation of N1 cells with MG132, there was marked and significant (p Ͻ0.05) potentiation of this 48-kDa NDRG1

NDRG1 up-regulates MIG6 to inhibit EGFR expression
band in the presence and absence of EGF. Total EGFR levels were again significantly (p Ͻ0.05) decreased by NDRG1 expression relative to the VC counterparts (Fig. 3A). Addi-tionally, EGFR was also slightly decreased by MG132 in both VC and N1 cells in the presence or absence of EGF relative to the respective controls. A and B, PANC-1 VC or N1 cells were incubated with control media or media containing EGF (10 ng/ml; 10 min/37°C) and examined via (A) co-immunoprecipitation or (B) immunofluorescence. A, NDRG1 was immunoprecipitated and then Western blotting performed to detect NDRG1 and MIG6 expression. Input lysates of the total protein were included for comparison with immunoprecipitated lysates for each sample. Western blots are typical of three independent experiments, with densitometric analysis representing mean Ϯ S.D. (n ϭ 3). Relative to untreated vector control cells, *, p Ͻ0.05 in the respective condition. B, under the same conditions, immunofluorescence images show staining for MIG6 (red), NDRG1 (green), and DAPI for nuclei (blue). All images were taken with a 63ϫ objective and at the same exposure time using AxioVision™ software. Images are representative from three experiments performed. Pixel intensity of (Bi) MIG6 and NDRG1 and (Bii) MIG6/NDRG1 co-localization was calculated using ImageJ Software, whereby ***, p Ͻ0.001 denotes significance comparing each condition to control cells. Pixel intensity and co-localization analysis utilized a total of 30 -60 cells over three experiments. B, the scale bar in the bottom right-hand corner of the first image represents 10 m and is the same across all images, except the close-up panel, where the scale bar is 5 m.

NDRG1 up-regulates MIG6 to inhibit EGFR expression
When examining pEGFR at Tyr-1068, EGF was able to markedly and significantly (p Ͻ0.001-0.01) increase its levels relative to the respective controls without EGF in both VC and N1 cells, as expected. In both VC and N1 cells, MG132 was able to significantly (p Ͻ0.05) increase pEGFR (Tyr-1068) levels in the presence of EGF relative to the control incubated with EGF alone (Fig. 3A). This was despite the ability of NDRG1 expression to significantly (p Ͻ0.05) reduce pEGFR (Tyr-1068) levels relative to the respective VC control treated with MG132 in the presence of EGF (Fig. 3A). Additionally, the ratios of pEGFR to EGFR were calculated, with MG132 and EGF treatment significantly (p Ͻ0.01) increasing the ratio relative to EGF alone in VC cells, whereas in N1 cells there was an increase (p Ͼ0.05) with MG132 and EGF (Fig. 3A). Together, these observations suggest that inhibiting proteasomal activity plays a role in preventing the decrease in pEGFR (Tyr-1068) levels in the presence and absence of NDRG1 overexpression.
The effect of lactacystin on NDRG1 expression (Fig. 3B) was similar to that observed for MG132 (Fig. 3A). lactacystin also had similar effects to MG132 on EGFR (Fig. 3B). However, in contrast to MG132, incubation with lactacystin only slightly decreased pEGFR (Tyr-1068) in the presence of EGF in both VC and N1 cells (Fig. 3B). Additionally, the ratios of pEGFR to EGFR were calculated, with lactacystin and EGF treatment not significantly (p Ͼ0.05) increasing the ratio relative to EGF alone in VC or N1 cells (Fig. 3B). As lactacystin has a different proteasomal inhibitory profile to MG132, this may explain the difference observed between these studies.
Of note, it has been shown that polyubiquitination of EGFR by the proteasome is needed to allow for lysosomal sorting of activated EGFR, suggesting the proteasome may, in part, assist trafficking and/or processing of EGFR. Furthermore, similarly to our current studies, MG132 has been demonstrated to increase pEGFR levels in the presence of EGF. Overall, there is some evidence from this investigation and previous studies with MG132 to suggest involvement of proteasomal processing in EGFR trafficking/processing.

Association of MIG6 and NDRG1 as demonstrated by co-immunoprecipitation and confocal microscopy
Taking into account the effects of NDRG1 on inhibiting EGFR expression (see Figs. 1, A and C, and 2 and 3), and the role of MIG6 in promoting EGFR processing by the lysosome (33), it was important to understand how exactly these proteins affect each other. In these studies, PANC-1 lysates were immunoprecipitated with NDRG1 antibody, and then NDRG1 or MIG6 levels examined by Western blotting (Fig. 3A). When NDRG1 was immunoprecipitated and then assessed for NDRG1 expression by Western blotting as a control, significantly higher (p Ͻ0.05) levels of NDRG1 were observed in the N1 cells relative to the VC cells in the presence and absence of EGF (Fig. 3A). Of note, when using the co-immunoprecipitation protocol, the 46-kDa NDRG1 band was not present or faintly present in either the input (Western only) or the IP for the N1 overexpression clone (Fig. 3A). In contrast, for VC cells, the 46-kDa band is present in both the input and the IP. This was a highly consistent finding throughout many experiments. A similar effect was also sometimes observed for the Westerns (e.g. Fig. 1, A and C), where the 46-kDa band is present in VC cells, but less visible in the N1 clones with high exogenous NDRG1 expression at 48 kDa. Considering this, we speculate that it may be because of three possible reasons: 1) the high expression of the exogenous NDRG1 48-kDa band sometimes overshadows the 46-kDa NDRG1 band, making it less visible; 2) alternatively, there appears to be some association between the two bands in N1 clones, leading to the appearance of the significantly larger 48-kDa NDRG1 band (Fig. 3A); and/or 3) the high levels of exogenous NDRG1 somehow influences endogenous 46 kDa NDRG1. Irrespective of the cause in the shift of the 46-kDa band, the higher total NDRG1 expression up-regulates MIG6.
Importantly, the NDRG1 immunoprecipitate was also found to contain substantial MIG6 levels, which was significantly (p Ͻ0.05) increased in N1 cells when compared with VC cells regardless of EGF treatment (Fig. 3A). These observations suggest a possible association between MIG6 and NDRG1.
To further examine the potential association of NDRG1 with MIG6, co-localization between these proteins was assessed using confocal microscopy (Fig. 3B). As this technique images cells on a single lateral focal plane, only molecules in the same location are co-localized (55). In the VC control and EGFtreated samples, MIG6 was mainly distributed in the cytosol as very fine puncta, whereas NDRG1 was present in both the cytosol and nucleus, again as fine puncta, with little co-localization being evident (r ϭ 0.549 -0.646). This distribution agrees with previous studies assessing the subcellular localization of MIG6 (32) and NDRG1 (15,43).
Upon NDRG1 overexpression in N1 cells, the overall intensity of both NDRG1 and MIG6 markedly and significantly (p Ͻ0.001) increased, resulting in intense clustering of MIG6 staining predominantly in the cytosol (Fig. 3B). On the other hand, for NDRG1 expression, in the N1 control and EGFtreated cells, its distribution was both cytoplasmic and nuclear. This distribution of NDRG1 and MIG6 upon NDRG1 overexpression in the presence and absence of EGF, led to a marked and significant (p Ͻ0.001) increase in co-localization in the cytoplasm (r ϭ 0.873-0.881), suggesting increased association between these proteins. These results in terms of NDRG1 and MIG6 expression are in agreement with the Western blotting data (Fig. 1A), whereas the co-localization is concordant with the co-immunoprecipitation results (Fig. 3A). Together, these data in Fig. 3, A and B, suggest an association of MIG6 and NDRG1 in the cytosol of cells that overexpress NDRG1, and prompted further dissection into the mechanisms involved in NDRG1's attenuation of EGFR signaling mediated via MIG6.

EGFR does not directly associate with NDRG1
Overall, the studies above in Fig. 3 indicate MIG6 and NDRG1 associate to form a potential complex. As an additional confirmation of this observation, PANC-1 lysates were immunoprecipitated with MIG6 antibody, and then MIG6, EGFR, or NDRG1 levels examined by Western blotting (Fig. 4A). When MIG6 was immunoprecipitated and then assessed for MIG6 expression by Western blotting as a control, markedly and significantly (p Ͻ0.001) higher levels of MIG6 were observed in N1 cells relative to the VC cells in the presence and absence of EGF (Fig. 4A). Assessing EGFR levels in the MIG6 immunoprecipi-

NDRG1 up-regulates MIG6 to inhibit EGFR expression
tate, it was demonstrated to be significantly (p Ͻ0.01-0.05) up-regulated in N1 cells, both with and without EGF treatment (Fig. 4A), suggesting an association between these proteins, particularly in the presence of NDRG1 overexpression. Further, the MIG6 immunoprecipitate contained substantial NDRG1 levels, which was significantly (p Ͻ0.05) increased in N1 cells (Fig. 5A). Notably, the MIG6 immunoprecipitate contained predominantly the 46-kDa isoform of NDRG1, which again is consistent with the active form of NDRG1 that has been shown to inhibit metastasis (16, 35,43). Hence, this was further evidence supporting the MIG6 and NDRG1 interaction observed in Fig. 3A.
Because MIG6 is known to directly bind to EGFR (26,56), it can be hypothesized that NDRG1 may also bind to EGFR. To examine this, PANC-1 lysates were immunoprecipitated with NDRG1 antibody and NDRG1 and EGFR levels then examined by Western blotting (Fig. 4B). When NDRG1 was pulled down, EGFR was not present in the immunoprecipitate. This was despite the fact that the procedure was successful in immunoprecipitating MIG6 and also NDRG1 itself (Fig. 3A). Furthermore, immunoprecipitation with the EGFR antibody also resulted in no detection of NDRG1 (Fig. S4B). Collectively, these results in Figs. 3, A and B, and 4, A and B, and Fig. S4B suggest that NDRG1 directly associates with MIG6, but not EGFR.
To further examine any possible association of NDRG1 with EGFR, co-localization between these proteins was assessed using confocal microscopy (Fig. S3). Again, EGFR expression was significantly (p Ͻ0.001-0.01) decreased by NDRG1 expression in N1 cells (Fig. S3i). However, little co-localization was evident between NDRG1 and EGFR (r ϭ 0.428 -0.487). Unlike the studies examining MIG6 and NDRG1 association (Fig. 3B), quantification of co-localization demonstrated no significant increase in the merged images in N1 cells relative to VC cells (Fig. S3ii).
Collectively, these studies in Fig. S3 suggest that there was no marked association between NDRG1 and EGFR. Considering this and the results in Figs. 3 and 4B, it can be hypothesized that NDRG1 may stabilize MIG6 to increase its half-life. Thus, as a theoretical model, the increased half-life of MIG6 would enable its binding to EGFR, with NDRG1 then dissociating, resulting in its not being detected with the EGFR.

NDRG1 overexpression increases MIG6 protein half-life
To assess the hypothesis that NDRG1 stabilizes MIG6 to increase its half-life, PANC-1 VC and N1 cells were preincubated with cycloheximide (10 g/ml) for 1 h/37°C, with EGF being added in the last 10 min and the incubation then being continued for 1-8 h/37°C, as per standard protein stability assessment protocols (57,58). From Fig. 4C, Western blotting demonstrated the half-life of MIG6 in VC cells was calculated as 1.6 Ϯ 0.2 h (3), which was very similar to previous reports using another cell type (i.e. 90 -120 min) (59). Interestingly, compared with VC cells, the half-life of MIG6 was significantly (p Ͻ0.001) increased to 7.9 Ϯ 0.4 h (3), which could explain 1) the significantly higher levels of MIG6 in N1 cells, 2) the posttranscriptional regulation of MIG6 (Fig. S1D), and 3) the ability of MIG6 to more markedly down-regulate EGFR expression in N1 relative to VC cells.

MIG6 is involved in the NDRG1-mediated down-regulation of EGFR
We have demonstrated previously (21) and herein (Figs.  1-3), that NDRG1 is able to down-regulate EGFR. Further, in this investigation, we have shown that NDRG1 up-regulates MIG6 protein levels (Figs. 1A and 3B), and intriguingly there is an association between NDRG1 and MIG6, but not NDRG1 and EGFR (Figs. 3 and 4 and Fig. S3). To dissect the potential role of MIG6 in the ability of NDRG1 to down-regulate EGFR, PANC-1 cells were transiently transfected with MIG6 siRNA (siMIG6) or nonspecific control siRNA (siControl). Then MIG6, EGFR, pEGFR (Tyr-1068), and NDRG1 levels were examined in the presence and absence of EGF by Western blotting (Fig. 5A).
In these studies, effective silencing of MIG6 was achieved, with a significant (p Ͻ0.01) decrease of MIG6 levels in VC and N1 cells. Additionally, similar to Fig. 1A, MIG6 was significantly (p Ͻ0.05) increased in N1 compared with VC siControl cells. When examining EGFR levels, NDRG1 expression in N1 cells resulted in a significant (p Ͻ0.01-0.05) decrease in its levels, in the presence and absence of EGF. Interestingly, a slight, but significant increase (p Ͻ0.05) of EGFR was observed in VC cells when MIG6 was silenced (siMIG6) in the presence and absence of EGF (Fig. 5A). Notably, the effect of siMIG6 was markedly more pronounced at significantly (p Ͻ0.05) increasing EGFR levels in N1 cells, irrespective of EGF treatment when compared with siControl N1 cells. This finding demonstrates the importance of MIG6 in the NDRG1-mediated decrease of EGFR expression.
Assessing pEGFR (Tyr-1068), its levels were also significantly (p Ͻ0.05) enhanced upon MIG6 silencing in both VC and N1 cells (Fig. 5A). However, this effect was less marked in N1 cells, with NDRG1 overexpression markedly decreasing the overall pEGFR levels relative to VC cells. This latter finding suggests the NDRG1-mediated decrease of pEGFR (Tyr-1068) is occurring, at least in part, via MIG6.

Silencing PTEN decreases MIG6 and results in up-regulation of total EGFR and increased pEGFR (Tyr-1068) levels in the presence and absence of NDRG1 overexpression
Previous studies have shown that the tumor suppressor PTEN is able to attenuate EGFR signaling by promoting the maturation of early to late endosomes (60). To decipher the involvement of PTEN in the ability of NDRG1 to down-regulate EGFR via MIG6, PANC-1 cells were transiently transfected with PTEN siRNA (siPTEN) or nonspecific control siRNA (siControl), and PTEN, EGFR, pEGFR (Tyr-1068), MIG6, and NDRG1 levels were examined in the presence and absence of EGF (Fig. 5B).
Efficient down-regulation of PTEN expression was achieved by siPTEN, with a significant (p Ͻ0.01) decrease of PTEN levels in both VC and N1 cells. When assessing EGFR levels, NDRG1 expression in N1 cells again significantly (p Ͻ0.01-0.05) decreased EGFR in the siControl treatment group. However, in NDRG1 up-regulates MIG6 to inhibit EGFR expression

NDRG1 up-regulates MIG6 to inhibit EGFR expression
VC cells, siPTEN caused a slight and significant (p Ͻ0.05) increase of EGFR in the presence and absence EGF (Fig. 5B).
Similarly, siPTEN also slightly, but significantly (p Ͻ0.01) increased EGFR in N1 cells, but only with EGF-treatment. This is consistent with the role of PTEN in enhancing EGFR degradation (60) and siPTEN decreasing this. Examining pEGFR (Tyr-1068) levels, siPTEN resulted in a significant (p Ͻ0.05) 2.0-to 2.1-fold increase in its levels only in the presence of EGF in both VC and N1 cells (Fig. 5B). Although siPTEN caused a similar -fold increase in both VC and N1 cells, the overall levels of pEGFR (Tyr-1068) were markedly and significantly (p Ͻ0.05) decreased with NDRG1 expression in N1 cells. Together, these findings suggest PTEN is also involved, at least in part, in the NDRG1-mediated degradation of EGFR.
Of interest, siPTEN resulted in a significant (p Ͻ0.01-0.05) decrease in MIG6 expression in both VC and N1 cells relative to their respective siControls in the presence and absence of EGF (Fig. 5B). This result indicates that the ability of PTEN to facilitate the down-regulation of EGFR (60) could be mediated through the up-regulation of MIG6. Collectively, these results suggest a role for PTEN in regulating MIG6.

The novel thiosemicarbazones Dp44mT and DpC up-regulate NDRG1 and MIG6 and promote their co-localization
As shown herein, NDRG1 is able to promote the degradation of a master regulator of oncogenic signaling, EGFR, via the upregulation of MIG6. Given the demonstrated ability of NDRG1 to suppress metastasis, this makes it a viable therapeutic target for cancer therapy (10). NDRG1 is potently up-regulated by the novel DpT class of anti-cancer agents that includes Dp44mT and DpC (10, 35) (Fig. S1A), the latter of which has entered clinical trials (40). In this study, the two lead agents, Dp44mT and DpC, were used in addition to the positive control, DFO (Fig. S1A), which is a well-known iron chelator used to treat iron overload disease (61). A negative control, 2-benzoylpyridine 2-methyl-3-thiosemicarbazone (Bp2mT), was also used, as it is an analog of Dp44mT and DpC, which was specifically designed not to bind metal ions or show anti-tumor activity (62). Although the lipophilic and permeable Bp2mT, Dp44mT, and DpC (10,35) were examined at 10 M, DFO was used at 250 M, because of its greater hydrophilicity, which leads to poor membrane permeabilization (34,63).
As demonstrated in previous studies (10,35), DFO, Dp44mT, and DpC were all able to markedly and significantly (p Ͻ0.001) up-regulate NDRG1 in PANC-1 cells (Fig. 6A). Both Dp44mT and DpC also significantly (p Ͻ0.001-0.01) decreased EGFR expression. Additionally, DpC significantly (p Ͻ0.05) decreased pEGFR (Tyr-1068), whereas Dp44mT was not as effective (p Ͼ0.05) in reducing its levels (Fig. 6A). As we observed previously (21), the pEGFR (Tyr-1068)/total EGFR was not significantly altered relative to the control by these agents, suggesting the decrease in total EGFR resulted in reduction in its phosphorylated levels. These agents also potently and significantly (p Ͻ0.001-0.05) increased HIF-1␣ expression, which is important, as this transcription factor targets MIG6 (42).
In good agreement with the increased HIF-1␣ levels, MIG6 was also markedly and significantly (p Ͻ0.001-0.01) up-reg-ulated by these agents, with DFO being less effective than either Dp44mT or DpC (Fig. 6A). Collectively, these agents up-regulate MIG6 probably via HIF-1␣ that induces NDRG1 expression (10), which then decreases EGFR expression and activation. This is notable, as MIG6 has been demonstrated to suppress tumor formation (64), and its ability to be harnessed by a therapeutic agent has not been investigated previously.
To further investigate these promising results, the effect of our novel agents was then assessed using confocal microscopy to examine the intracellular distribution and co-localization of MIG6 and NDRG1 in PANC-1 cells incubated for 24 h/37°C with our agents (Fig. 6B), under the conditions described above in Fig. 6A. The expression of MIG6 was predominantly cytoplasmic, whereas NDRG1 was both cytoplasmic and nuclear. The levels of MIG6 and NDRG1 were significantly (p Ͻ0.01-0.001) increased after incubation with DFO, and especially Dp44mT and DpC (Fig. 6Bi). This result was in good agreement with the Western blotting data (Fig. 6A). Quantification analysis demonstrated significantly (p Ͻ0.001) increased co-localization of MIG6 and NDRG1 (i.e. yellow fluorescence) in cytoplasmic puncta particularly after incubation with Dp44mT and DpC, which was accompanied by an increase in the Pearson's correlation coefficient from r ϭ 0.516 in the control to r ϭ 0.814 -0.913 (Fig. 6Bii). It is of interest to note that DFO demonstrated less efficacy than both Dp44mT and DpC at increasing MIG6 expression (Fig. 6A), which led to less co-localization with NDRG1 (Fig. 6B). This decreased efficacy of DFO is probably related to its greater hydrophilicity and poorer membrane permeability relative to the lipophilic and highly membranepermeable Dp44mT and DpC (34,36). Together, these data in Fig. 6 indicate that the anti-cancer agents Dp44mT and DpC markedly up-regulate NDRG1 and MIG6, which co-localize intracellularly.
To demonstrate the mechanism by which the thiosemicarbazones up-regulate MIG6, in Fig. S4A we have examined the effect of the agents used in Fig. 6, A and B, on MIG6 expression upon NDRG1 silencing after a 24-h/37°C incubation. In siControl cells, Dp44mT and DpC robustly increased MIG6 expression. In contrast, after NDRG1 silencing, both agents were markedly less effective at up-regulating MIG6 (Fig. S4A). These results indicate that the up-regulation of MIG6 by Dp44mT and DpC depends on NDRG1 expression.

DpC and Dp44mT promote EGFR and LAMP2 co-localization
Next, considering the results in Fig. 6 above regarding the up-regulation of MIG6 and its co-localization in intracellular vesicles by our agents, studies then assessed if this could result in degradation of EGFR via the induction of lysosomal processing (33). Hence, studies examined the effect of these agents on inducing EGFR redistribution and co-localization with LAMP2-stained late endosomes/lysosomes using confocal microscopy (Fig. 7). In control and Bp2mT-and DFO-treated PANC-1 cells, EGFR (red) showed distinct plasma membrane staining with some fine punctate cytosolic staining, whereas the LAMP2 late endosome/lysosomal marker (green) was finely distributed in puncta within the cytosol. Upon quantification,

NDRG1 up-regulates MIG6 to inhibit EGFR expression
only limited co-localization (r ϭ 0.228 -0.426) of the EGFR and LAMP2 marker was detected in control and Bp2mT-and DFOtreated cells (Fig. 7ii).
In contrast, treatment of PANC-1 cells with Dp44mT and DpC resulted in pronounced clustering of EGFR intracellularly into defined foci. Similarly, LAMP2 staining was also markedly

NDRG1 up-regulates MIG6 to inhibit EGFR expression
altered upon treatment with DpC and Dp44mT, resulting in clustering into perinuclear foci that co-localized with EGFR. However, this redistribution of LAMP did not alter its total levels between conditions, with no significant change in pixel intensity being observed (Fig. 7i). Notably, there was a significant (p Ͻ0.001) increase in the co-localization intensity of EGFR and LAMP2 upon quantification, with a marked increase in Pearson's correlation coefficient from r ϭ 0.426 in the control to r ϭ 0.640 -0.743 in Dp44mT-and DpC-treated cells (Fig. 7). Once again, the efficacy of DFO at reducing EGFR was less marked than that found for Dp44mT and DpC, probably because of its lower membrane permeability (34,36,63). This resulted in decreased induction of MIG6 expression by DFO relative to Dp44mT and DpC (Fig. 6, A and B), and thus, no significant effect on reducing EGFR expression (Figs. 6A and 7). Collectively, these studies in Fig. 7 demonstrate that the novel anti-cancer agents Dp44mT and DpC up-regulate MIG6 and promote EGFR localization to LAMP2-defined late endosomes/lysosomes.

Discussion
Metastasis is the major killer in cancer (65), which remains the "emperor of all maladies." Understanding the mechanisms of action of metastasis suppressor proteins that block metastasis could lead to new therapeutic targets. The metastasis suppressor NDRG1 has demonstrated comprehensive anti-cancer activity via its ability to inhibit a series of critical oncogenic signaling pathways that play roles in metastasis and cancer progression (12,13,15,16,18). As such, a major question in the field was how does NDRG1 inhibit the activity of such a broad range of signaling pathways? Recently, our laboratory has reported that NDRG1 could effectively do this via its ability to down-regulate a master regulator of these pathways, namely  EGFR (20,21). In fact, these studies indicated that NDRG1 expression accelerated the loss of EGFR monomers and dimers relative to control cells in the presence of EGF, suggesting that NDRG1 facilitated its processing (21). The current study investigated the mechanisms involved in this effect.

NDRG1 up-regulates MIG6 to inhibit EGFR expression
The degradation of EGFR is largely controlled by MIG6 (33), which is a cytoplasmic protein (32) that down-regulates EGFR activity by direct inhibition of the EGFR dimer (26) and induces lysosomal processing (33). We demonstrate that NDRG1 expression markedly increased MIG6 levels, whereas silencing NDRG1 in multiple cell types decreased MIG6, demonstrating that NDRG1 is able to regulate this protein. Notably, there is no evidence in the literature that NDRG1 induces its effects by acting as a transcription factor, and the up-regulation of MIG6 observed herein was not because of an increase in MIG6 mRNA levels after up-regulation of NDRG1, suggesting a posttranscriptional mechanism of regulation. Indeed, we demonstrate that NDRG1 markedly extends the protein half-life of MIG6, suggesting that it functions to stabilize this molecule. An important finding was that MIG6 became bound to NDRG1, but not EGFR, in the cytoplasm, as determined using both co-immunoprecipitation and co-localization using confocal microscopy. This direct interaction may be key to how NDRG1 enhances the half-life of MIG6.
The well-characterized lysosomotropic agents NH 4 Cl and CH 3 NH 2 that inhibit endosomal and lysosomal acidification (48, 66,67) were demonstrated to reverse the effects of NDRG1 on decreasing pEGFR (Tyr-1068) and increasing MIG6, suggesting perturbation of the acidic lysosomal pH interferes with EGFR and MIG6 processing. The increased MIG6 levels increased EGFR internalization and co-localization of EGFR with the early endosome marker, EEA1, and the late endosome/ lysosome marker, LAMP2.
From these results, a model of the interaction between these molecules can be proposed, whereby NDRG1 increases MIG6 expression by a posttranscriptional mechanism, potentially by binding to it and stabilizing its protein levels. As such, NDRG1 may also facilitate MIG6 binding to EGFR, resulting in its inhibition and the late endosomal/lysosomal processing of EGFR (Fig. 8). As discussed above, relative to the strong colocalization between LAMP2 and EGFR, no appreciable colocalization was observed between LAMP2 and NDRG1, or LAMP2 and MIG6. This suggests that the effect of NDRG1 and MIG6 on increasing EGFR processing via the lysosome could occur prior to its internalization into the lysosomal compartment (Fig. 8).
Supporting evidence for this model comes from studies where MIG6 silencing reversed the inhibitory effect of NDRG1 on EGFR expression (Fig. 5A), suggesting a role for MIG6 in the NDRG1-mediated down-regulation of the EGFR. Importantly, the direct binding of MIG6 to EGFR was also significantly increased in NDRG1 overexpressing cells, further demonstrating that NDRG1 is enhancing the formation of the MIG6 -EGFR complex. Silencing PTEN, which is known to augment early to late endosome maturation (60), decreased MIG6 in the presence and absence of NDRG1 overexpression, with a concurrent increase in EGFR and pEGFR (Tyr-1068) levels. Hence, the PTEN-mediated trafficking of endosomes to lysosomes (60) may play a role in MIG6-directed processing of EGFR and pEGFR.
Of note, NDRG1 silencing in epithelial cells was previously reported to decrease the uptake of low-density lipoprotein (68), which suggests a potential shared, and more general, role for this metastasis suppressor in receptor internalization and vesicular trafficking. However, the direct interaction of NDRG1 with MIG6 and the formation of an MIG6 -NDRG1 complex probably accounts for the increased levels and stability of MIG6 (Fig. 4C), which facilitates EGFR processing by the lysosome.
Unlike most metastasis suppressors that are yet to be targeted by pharmacological interventions, it has been demonstrated that novel anti-cancer agents of the DpT class, namely Dp44mT and DpC (22,34,36), can up-regulate NDRG1 expression (10,35). This effect of these agents is known to occur through both HIF-1␣-dependent and -independent processes (10). The mechanism of this induction of NDRG1 is mediated by the pronounced iron chelation efficacy of Dp44mT and DpC (34,36), which inhibits the activity of prolyl hydroxylase that requires iron in its active site and is critical for HIF-1␣ degradation (69). The inhibition of prolyl hydroxylase results in increased HIF-1␣ levels (Fig. 6A) that then lead to increased transcription of its downstream targets, which include NDRG1 (10). This effect of HIF-1␣ could be responsible for the marked up-regulation of MIG6 observed after incubation of cells with DpC and Dp44mT (Fig. 6A), as MIG6 is a known transcriptional target of HIF-1␣ (42). Relevant to this, the up-regulation of MIG6 was demonstrated after incubation of cancer cells with several well-characterized iron chelators (41). Furthermore, the concurrent up-regulation of both NDRG1 and MIG6 via HIF-1␣ could then lead to stabilization of MIG6 through its binding to NDRG1, as shown herein. In fact, silencing of NDRG1 prevented the ability of Dp44mT and DpC to markedly up-regulate the EGFR inhibitor MIG6. Hence, the potent effects of Dp44mT and DpC are, at least in part, because of the up-regulation of NDRG1 and its effects on increasing MIG6 levels.
Importantly, the effect of these pharmacological agents on up-regulating both NDRG1 and MIG6 expression was very similar to that observed by genetic overexpression of NDRG1, with the following being observed: NDRG1 co-localization being demonstrated with MIG6 in the cytosol (cf. Figs. 3B and 6B), decreased total and activated EGFR levels (cf. Figs. 1A and  7), and redistribution of EGFR to LAMP2-stained late endosomes/lysosomes (cf. Figs. 2B and 7). Moreover, both Dp44mT and DpC also markedly inhibit PANC-1 pancreatic tumor xenograft growth in vivo, demonstrating their potent activity (10,35). In fact, DpC was demonstrated to show greater activity than a "gold standard" agent for pancreatic cancer treatment, namely gemcitabine, with DpC totally inhibiting PANC-1 tumor xenograft growth (10,35).
In conclusion, the clinical trial DpT group of agents can down-regulate a "master switch" of oncogenic signaling, namely EGFR, via an entirely different molecular mechanism currently employed to target this tyrosine kinase. That is, by up-regulating NDRG1 to increase MIG6 expression via enhancing its half-life. As such, the DpT compounds provide an innovative strategy for treating cancer, with these agents being capable of 1) potently inhibiting the growth and metastasis in NDRG1 up-regulates MIG6 to inhibit EGFR expression vivo of a variety of aggressive cancers (16, 21, 22, 34 -38, 70) and 2) markedly inhibiting multiple oncogenic signaling pathways (12-15, 21, 30, 71-73).

Cell culture
Human pancreatic cancer cell types, namely PANC-1, CFPAC-1, and AsPC-1, were purchased from the American Type Culture Collection (Manassas, VA). These cell types were authenticated based on viability, recovery, growth, morphology, and also cytogenetic analysis, antigen expression, DNA profile, and iso-enzymology by the provider. The cells were maintained in DMEM with 10% fetal calf serum (Life Technologies) and cultured by standard procedures (14) for less than 3 months after resuscitation. Two PANC-1 clones stably transfected to overexpress NDRG1 (N1 and N2) were compared with PANC1 cells transfected with the empty vector (VC), as described previously (12,14).
siRNA Two specific siRNAs for NDRG1 were used, namely siN-DRG1 (Cat. # s20336; Life Technologies) and siNDRG1II (Cat. # s20334; Life Technologies), and also an siRNA against MIG6 (siMIG6; Cat. # AM16708; Ambion), and these were compared with nontargeting negative control siRNA (siControl; Life Technologies). The siRNA was transiently transfected into PANC-1 cells using Lipofectamine 2000 (Life Technologies) and incubated for 72 h/37°C. Western blotting was then performed. We demonstrated a high positive correlation (r 2 ϭ 0.82-0.83) between NDRG1 down-regulation via siNDRG1 and the decrease in MIG6 expression in all three cell types used (i.e. PANC-1, CFPAC-1, AsPC-1) (data not shown). This dem- Figure 8. Schematic illustrating the role of NDRG1 and MIG6 in facilitating EGFR processing by the lysosomal compartment. EGF binds to EGFR, inducing its internalization via endocytosis into the cell leading to EGFR in early endosomes and lysosomes. The tumor suppressor PTEN has been demonstrated to play a role in the trafficking of early endosomes to lysosomes (60). The novel di-2-pyridylketone thiosemicarbazones Dp44mT and DpC bind and deplete cellular iron (10,34,35), which inhibits prolyl hydroxylase activity and results in up-regulation of NDRG1 via HIF-1␣-dependent and -independent mechanisms (10). Depletion of cellular iron also up-regulates MIG6 potentially via HIF-1␣ (41). Subsequently, MIG6 down-regulates EGFR by directly binding to it and inhibiting the EGFR dimer (26) and inducing EGFR internalization and processing by the lysosome (33). Notably, MIG6 binds to NDRG1 in the cytosol, with increased stabilization of MIG6 being observed in NDRG1 overexpressing cells. This increased half-life of MIG6 probably increases the access of MIG6 to EGFR to ensure down-regulation. However, NDRG1 does not bind to EGFR, with no association of MIG6 or NDRG1 being demonstrated with the late endosome/lysosomal marker, LAMP2. In contrast, NDRG1 expression increases the co-localization of EGFR with EEA1 and LAMP2. In summary, NDRG1 overexpression increases expression of the EGFR inhibitor MIG6 by stabilizing it, resulting in an MIG6-NDRG1 complex that potentiates EGFR down-regulation by enhanced lysosomal processing of EGFR.

NDRG1 up-regulates MIG6 to inhibit EGFR expression
onstrates that there is a dose escalation effect between NDRG1 and MIG6 expression. The siRNA specific for PTEN (siPTEN) was also compared with siControl but was transiently transfected using RNAiMax (Life Technologies) and incubated for 72 h/37°C.

Cell treatments
EGF (Cat. # 8916; Cell Signaling Technology, Danvers, MA) was used at 10 ng/ml diluted in 1% FCS-containing media and incubated with cells for 10 min/37°C. This concentration and time point was implemented based on preliminary studies examining the efficacy of EGF in vitro and following that used by previous studies (21,74). Prior to treatment with EGF, cells were serum-starved in media containing 1% FCS for 24 h/37°C (75). DFO (Fig. S1A) was purchased from Sigma-Aldrich. The thiosemicarbazones Dp44mT and DpC and the negative control compound Bp2mT were synthesized and characterized using standard methods (36,62,76). We utilized concentrations of 10 M for Dp44mT, DpC, and Bp2mT and 250 M for DFO in 10% FBS-supplemented medium. The greater concentration of DFO was implemented because of its limited ability to permeate the cell membrane (63). Both Dp44mT and DpC were utilized at lower concentrations because these agents show far higher membrane permeability and demonstrate marked iron chelation efficacy (34 -36). Bp2mT, Dp44mT, and DpC were freshly dissolved in dimethyl sulfoxide (DMSO) and diluted in culture media (final [DMSO] Յ 0.1%).

Co-immunoprecipitation
Co-immunoprecipitation was performed using Dynabeads Protein G (Life Technologies) following the manufacturer's instructions. Briefly, cells were washed with ice-cold PBS and lysed using the IP Lysis Buffer (Pierce) containing protease inhibitors (Roche Diagnostics). Protein (400 g) was incubated with either monoclonal NDRG1 antibody (1:2000 dilution) (Cat. # ab37897; Abcam), EGFR antibody (Cat. # 2085; Cell Signaling Technology), or MIG6 antibody (Cat. # 2440; Cell Signaling Technology) overnight at 4°C. This mixture was added to 30 l of Dynabeads Protein G and incubated for 3 h at 4°C. The beads were then washed three times with ice-cold PBS, and the protein was eluted using the LDS Loading Buffer. Samples were then heated at 95°C for 5 min and placed on a magnet, and equal amounts of supernatant were loaded and separated on a 10% SDS-PAGE gel. Then NDRG1, MIG6, or EGFR were detected by Western blotting. Notably, control IP experiments using an IgG isotype control antibody (Cat. # 3900; Cell Signaling Technology) were performed under the same conditions. No pulldown was observed when lysates were incubated with the IgG control antibody (Fig. S5). This demonstrates that there was no nonspecific binding to the antibody.

Densitometry
Densitometry was performed using Quantity One software (Bio-Rad) and normalized using the relative ␤-actin loading control for protein.

Statistics
Data are mean Ϯ S.D. (number of experiments) and were compared using Student's t test. Results were considered significant when p Ͻ0.05.