Egr-1 Mediates Hypoxia-Inducible Transcription of the NDRG1 Gene through an Overlapping Egr-1/Sp1 Binding Site in the Promoter

Oxygen homeostasis is essential for eukaryotic cells. Normal cells use both oxidative phosphorylation and glycolysis pathways but rely predominantly on the former, switching to the latter at times of oxygen deprivation (1). Hypoxia are hallmarks of solid tumors (1). Compared with normal cells, cancer cells use glycolysis for energy production (Warburg effect) because of hypoxic conditions (2). Hypoxia triggers hypoxia-inducible 1α (HIF-1α), early growth response 1 (Egr-1), and other transcription factors that activate tumor angiogenesis and promote cell invasion (3, 4). Expression of the N-myc down-regulated gene 1 (NDRG1) correlated with hypoxia in the microenvironment of both normal and tumor cells (5, 6). Understanding how hypoxia regulates transcription of the NDRG1 gene will increase our knowledge of the cellular responses of normal and cancer cells to low oxygen tension.

NDRG1 is a gene known to be down-regulated by the MYC family of proteins (7) and its expression is altered by diverse physiologic and pathologic conditions. NDRG1 (also known as Drg1, RTP, Rit42, PROXY-1, and Cap43) was identified as a gene up-regulated during cellular differentiation (8–10), as well as a reducing agent homocysteine- and tunicamycin-responsive gene (11). The expression of NDRG1 is enhanced by multiple signals, including hypoxia (12), DNA damage (13), and various differentiating agents and chemicals, including synthetic ligands of the peroxisome proliferator activated receptor γ (PPARγ)/retinoid X receptor transcriptional pathway (14), androgens (15), Ni²⁺ compounds, calcium ionophore, and okadaic acid (12, 16).

NDRG1 belongs to a gene family consisting of four members (NDRG1–NDRG4). Sequence analysis revealed that the human NDRG2, NDRG3, and NDRG4 proteins showed 53% to 62% identity to the human NDRG1 protein (394 amino acids; ref. 17). Motif and Prosite analysis revealed an α/β hydrolase fold of ∼220 amino acids and an esterase/lipase/thioesterase active site serine present in all four proteins. This catalytic domain is common to several hydrolytic enzymes with various catalytic functions. The amino acid sequences of the α/β hydrolase fold region show the highest homology among the four proteins, suggesting that the α/β hydrolase fold might be very important to the function of the NDRG family proteins. NDRG1 is unique among the NDRG family in that it contains a motif of three decapeptide tandem repeats (GTRSRSHTSE) near the COOH terminus (17). NDRG1 has been identified as a physiologic substrate for serum- and glucocorticoid-induced kinase 1 (SGK1) and phosphorylation of NDRG1 by SGK1 primes it for phosphorylation by glycogen synthase kinase 3 (18). However, the exact function(s) of NDRG1 remains unknown.

NDRG1 has pleiotropic functions, including cell differentiation, antimetastasis, apoptosis resistance, and spindle checkpoint (9, 10, 12, 13, 19–21). A knockout animal model revealed that the NDRG1 gene was involved in the proper functioning of Schwann cells (22), a type of glial cells of the peripheral nervous system that mediates myelin sheath synthesis. The NDRG1 gene knockout mice showed symptoms similar to the human peripheral degenerative disease Charcot-Mary-Tooth (22). Additionally, N-myc/c-myc can inhibit the expression of the NDRG1 gene during embryogenesis, indicating that NDRG1 may play a role in embryonic development (14). In a recent study, both the basal and hypoxia-inducible expression of NDRG1 was investigated in pancreatic tumors. Compared with the tumors with moderate differentiation, tumor cells with less differentiation showed lower NDRG1 expression, whereas NDRG1 mRNA remained unchanged in poorly differentiated hypoxic tumor cells. In addition, NDRG1 expression was shown to be correlated with the metastatic status of tumors. It was shown that colon tumor cells with NDRG1 overexpression exhibited lower metastatic potential compared with tumor cells with low NDRG1 expression using an in vitro invasion assay and an in vivo liver metastases assay (14). A recent pathologic study revealed that NDRG1 expression correlated with a less aggressive form of colorectal liver metastases (23). On the other hand, Wang et al. (24) reported that NDRG1 might be a tumor metastasis promoter gene and plays an important role in colorectal carcinogenesis. Although hypoxia has been shown to be critical for tumor angiogenesis, invasion, and possibly metastasis (25, 26), the exact role NDRG1 plays in these processes and the mechanism underlying the induction of NDRG1 gene expression by hypoxia remains to be determined.

We and other groups have reported that hypoxia and hypoxia-mimicking agents could strongly induce expression of the NDRG1 gene (12, 27). HIF-1α has been shown to be required for NDRG1 induction (12) because NDRG1 was not induced in HIF-1α knockout mouse embryonic fibroblast (MEF) cells (HIF-1α^−/− MEF) by hypoxia as well as agents that activate hypoxia signaling under normal oxygen tension such as Ni²⁺ compounds, dimethyloxaloylglycine, or deferoxamine (12, 19). Additionally, in 786-O cells that are deficient in the HIF-1α ubiquitin ligase gene, human von Hippel-Lindau (VHL), NDRG1 exhibited a high basal but uninducible expression compared with cells that had normal VHL that could degrade HIF-1α (28). The existence of a putative HIF-1α binding site (HRE) in the promoter region NDRG1 (12), as well as the ability of deferoxamine, Co, Ni, and hypoxia to induce this gene, supports the hypothesis that HIF-1α played a role in activating the transcription of this gene (12). However, the HRE element in the NDRG1 promoter contains only four (ACGT) of the nine bases of the consensus HRE motif (3, 29). Additionally, there are reports that both HIF-1α–dependent and HIF-1α–independent mechanisms were involved in hypoxia-inducible NDRG1 expression. However, the effects of the HIF-1α–independent induction of NDRG-1 were small in magnitude compared with those that were HIF-1α dependent (19).

Besides HIF-1α, other transcription factors have been shown to be involved in hypoxia signaling. The Egr-1 transcription factor has been shown to be induced by a variety of stimuli such as 12-O-tetradecanoylphorbol-13-acetate (TPA; ref. 30) and hypoxia (12, 31). The Egr-1 transcription factor belongs to a larger family of early growth response genes that have nearly identical zinc finger structures and recognize the GCGGGGGCG binding motif (30, 32, 33). A canonical overlapping binding motif for Egr-1 and transcription factor Sp1/Sp3 was identified in the promoter of the tissue factor (TF), and platelet-derived growth factor B and C chains (PDGF-B/C; refs. 34, 35). Egr-1 was recently shown to mediate hypoxia-inducible expression of TF through the Egr-1/Sp1 binding motifs located at the proximal promoter (36).

In this study, we have mapped a functional overlapping binding site for Egr-1 and Sp1 in the human NDRG1 promoter. Hypoxia and deferoxamine activated NDRG1 promoter constructs that contained this element, and hypoxia induced a significant increase in the level of Egr-1 protein in RAW 264.7 murine macrophage cells. The binding of Egr-1 to this element correlated with transcriptional activation of the NDRG1 promoter and induction of NDRG1 gene expression. Furthermore, the induction of NDRG1 expression by hypoxia and deferoxamine was abrogated in homozygous Egr-1^−/− MEFs compared with heterozygous Egr-1^+/− MEFs. Knockdown of Egr-1 expression with RNA interference showed that Egr-1 was required for induction of NDRG1 expression by hypoxia and deferoxamine. The important role of this overlapping Egr-1/Sp1 site in the positive regulation of NDRG1 expression is further supported by the conservation of this motif among several species.

Cell culture and reagents. Murine macrophage RAW264.7 cells were maintained in RPMI 1640 containing 10% fetal bovine serum (FBS). Human lung carcinoma A549 cells, HeLa cells, and Hep3B cells were grown in DMEM obtained from Invitrogen and supplemented with 10% FBS, 1% penicillin, and streptomycin (Invitrogen). Egr-1^+/− and Egr-1^−/− MEFs were kindly provided by Dr. Jeffrey Milbrandt (Washington University). All cells were maintained at 37°C in a humidified atmosphere with 5% CO₂ and 95% air. Unless otherwise specified, all reagents were purchased from Sigma. Hypoxia conditions were achieved as follows: cells were grown to 60% to 70% confluence in complete medium with 10% FBS; on the day of hypoxia treatment, growth medium was switched to fresh medium (10% FBS), which had been preequilibrated in the hypoxia chamber with a continuous flow of a hypoxic gas mixture with 1% O₂. Cells were then incubated in the hypoxia chamber for the indicated time. Cell viability was assessed by trypan blue exclusion and quantitating the percentage of cells attached before and after hypoxia treatment; both methods showed that hypoxia condition did not affect cell viability.

NDRG1 promoter constructs and transient transfection. A promoter construct encompassing a region of −3,057 to +510 bp relative to the transcription start site of the human NDRG1 gene was amplified from human genomic DNA using a High Fidelity PCR kit (Invitrogen) according to the manufacturer's instructions. The sequence of the upstream primer was 5′-CAATAAATCGCCGCAATGG-3′; whereas that of the downstream primer was 5′-CACTGTGACACACCCTGGAC-3′. A ∼3.7 kb fragment was successfully amplified, gel-purified, and cloned into the PCR 2.1 vector using the TA cloning kit (Invitrogen). The sequence was verified by restriction enzyme cutting and sequencing. The 3.7 kb fragment was digested with KpnI and XhoI, purified by gel electrophoresis and cloned into the pGL-3 Basic plasmid (Promega). Unidirectional 5′ deletions of the parental plasmid were carried out by using the unique restriction enzyme cutting sites NheI, XbaI, NcoI, NdeI, BglI, and ApaI with KpnI and the shortened promoter constructs were then religated. When necessary, the parental plasmid was digested and blunt-ended using T4 polymerase (Roche) in the presence of 1 mmol/L deoxyribonucleotide triphosphates. The corresponding deletion fragments were cleaved by XhoI, gel purified and cloned into the SmaI/XhoI sites of the pGL-3 Basic plasmid. To obtain unidirectional 3′ deletions, an exonuclease III method was used (35). Briefly, plasmid −257/+1,100 was opened at the PstI and EcoRI sites, purified by gel electrophoresis, and digested with exonuclease III. The desired lengths of the promoter were obtained by careful controlling of the time interval of incubation. The resulting promoter fragments were digested with S1 nuclease to eliminate single-stranded regions and the ends of the fragments were repaired with T4 polymerase (Roche). Following these treatments, the plasmids were ligated, purified, and sequenced.

NDRG1 promoter constructs were transfected into RAW264.7 and A549 cells by a calcium phosphate method (37). To control for transfection efficiency, cells were cotransfected with the pCMVβ-galactosidase plasmid (kindly provided by Dr. T-C. Suen, Tuxedo, NY). Twenty-four hours after transfection, cells were incubated in the hypoxia chamber for 6 h or treated with 200 μmol/L deferoxamine for 16 h. Cells were collected, lysed in 1× reporter lysis buffer, and luciferase or β-galactosidase activity was measured (37).

Electrophoretic mobility shift assay. A fragment encompassing the NDRG1 gene promoter −80 to −50 bp was synthesized by annealing two complementary DNA primers, and the double-stranded oligonucleotide probe was labeled at the 3′ end with Klenow fragment (Roche) and purified using a G-25 column (Roche). Nuclear extracts were obtained as previously described (38). The electrophoretic mobility shift assay (EMSA) reactions were carried out as described (30), with the following minor modifications. For the antibody supershift assay, the probe and nuclear extracts were incubated for 20 min and then the desired antibody was added into the reaction mixture for 15 min and incubated at 37°C before loading the mixture into a 5% 0.5 × Tris-borate EDTA (TBE) gel. Sp1 and Egr-1 antibodies were purchased from Santa Cruz Biotechnology. Sp3 antibody was obtained from e-Biosciences.

Western blot. Nuclear extracts were prepared and proteins were separated using SDS-PAGE (7.5%). Proteins in the gel were transferred electrophoretically to polyvinylidene difluoride membranes (Bio-Rad), and immunoblotted with rabbit anti–Egr-1 IgG (Santa Cruz Biotechnology) or anti-Sp1 IgG (Santa Cruz Biotechnology) using the “blotto” technique (39). Bound primary antibody was reacted with horseradish peroxidase–conjugated goat anti-rabbit IgG (Santa Cruz Biotechnology) and detected with the enhanced chemiluminescent Western blotting system (Amersham Biosciences).

Northern blot and RNA interference inhibition. NDRG1 Northern blot was done as previously described (12, 16). Ten micrograms of total RNA from each treatment was isolated with TRIzol reagent (Invitrogen) and fragmented on a 1% and 1× MOPS agarose gel containing formaldehyde. Gel was soaked in alkaline buffer for denaturation and transferred to a positively charged nylon membrane using a vacuum apparatus (Amersham Biosciences). Hybridization was carried out in 68°C for 16 h. The membrane was washed in stringent buffer containing 0.1% SDS and 0.1× SSC for 30 min. Northern blot probes for human NDRG1 and Egr-1 gene were PCR amplified using human NDRG1 and Egr-1 cDNA as templates. RNA interference (RNAi) ON-TARGET smart pool Egr-1 RNAi and GFP RNAi were purchased from Dharmacon. Transfection was done with DharmaFect 1 transfection reagent and the final concentration of small interfering RNA in medium was 50 nmol/L. Twenty-four hours after transfection, cells were treated with either hypoxia or deferoxamine as described above.

Identification of cis-elements in the NDRG1 promoter that mediate inducibility by hypoxia, deferoxamine, and okadaic acid. To study the basal and hypoxia-inducible promoter activity of the human NDRG1 gene, the 5′ flanking region encompassing −3,057 to +510 bp relative to the transcriptional start site of the human NDRG1 gene was cloned upstream of a luciferase reporter gene (−3,057/+510 construct) and various deletions from both the 5′ and 3′ ends were generated (see Fig. 1A). These NDRG1 promoter constructs were cotransfected with a cytomegalovirus (CMV)–β-galactosidase vector into the murine RAW264.7 macrophage cells, which have been previously used for hypoxia-inducible transcriptional studies (40) and then incubated with either hypoxia or deferoxamine for 16 h. Because we have previously shown that okadaic acid strongly induced NDRG1 expression (16), the inducibility of these constructs after treatment with okadaic acid in A549 cells was also investigated (Fig. 1D). Luciferase activity was normalized to β-galactosidase activity to control for transfection efficiency. This −3,057/+510 luciferase construct exhibited a low basal activity in normoxia and treatment with hypoxia or deferoxamine significantly induced NDRG1 promoter activity (Fig. 1C). 5′ Deletions of the promoter from −3,057 to −257 bp did not significantly affect basal activity and inducibility by hypoxia or deferoxamine as shown in the −350/+510 and −257/+510 luciferase constructs. Similar results were obtained with okadaic acid treatment (Fig. 1D). These results were consistent with a previous report showing that 5′ deletions up to −250 bp in the mouse NDRG1 promoter were not associated with any appreciable changes in basal promoter activity (7). However, when the promoter was further truncated at the 3′ end from +510 to +355 bp in the −257/+355, −150/+355, and −80/+355 constructs, both the basal and the inducible promoter activities were significantly increased compared with −350/+510 and −257/+510 promoter constructs, suggesting the presence of a negative regulatory element within +355/+510 bp. Further 5′ deletions beyond −80 in −45/+355, −20/+355, and 0/+355 constructs abrogated both basal and inducible promoter activities, thereby mapping a putative cis-element within −80 and −45 bp relative to the transcriptional start site. The loss of induction of promoter activity by hypoxia, deferoxamine, and okadaic acid in the −45/+355 construct suggested the existence of a positive regulatory element responsible for both basal and inducible promoter activity. Similar results were obtained using the human Hep3B hepatocellular carcinoma cells (data not shown).

A binding site for the transcription factor Sp1 had previously been reported within −90 to −45 bp relative to the transcriptional start site and further examination of the nucleotide sequence revealed a transcription factor Egr-1 binding site overlapping with the Sp1 site (Fig. 1B). A functional and overlapping binding site for Egr-1 and Sp1 had been reported in the promoter region for several genes, including TF (40), PDGF-B (34), and PDGF-C (35). Promoter activity of the latter genes is likely the result of competition between Egr-1 and Sp1 for binding to this overlapping binding site. Hence, in the uninduced condition, a low level of Egr-1 expression and activity will favor the binding of the abundant Sp1 transcription factor. However, in the presence of environmental stressors and modifiers such as hypoxia, okadaic acid, and TPA, Egr-1 was induced, which competed out the binding of Sp1 to the overlapping binding site (34, 35).

Binding of transcription factors Egr-1 and Sp1 to an overlapping Egr-1/Sp1 binding motif in the NDRG1 promoter. Overlapping binding sites for Egr-1/Sp1/Sp3 in the promoters have been shown to be involved in the control of the transcription of several genes (34, 35). To investigate the binding of transcription factor(s) to this putative overlapping binding site within the NDRG1 promoter, EMSA assays were done using a ³²P-labeled oligonucleotide probe spanning the −80 to −50 bp region. Two major DNA protein complexes were detected when the probe was incubated with nuclear extract from cells maintained at 21% O₂ (Fig. 2A,, lanes 2, bands I and III). A third DNA-protein complex (CII complex) was detected in nuclear extract from RAW264.7 cells incubated in 1% O₂ for 6 h (Fig. 2A,, band II lane 3, compared with lane 2). Similar DNA-protein complexes were found in the Egr-1/Sp1 binding motif of the promoter of the Egr-1 gene itself and the PDGF-C gene (34, 35). The specificity of each complex formed was further examined by cold competition; the formation of the CII complex was abrogated in the presence of cold 25- and 50-molar excess Egr-1 oligonucleotide (Fig. 2A,, compare lane 3 with lanes 4 and 5), whereas both CI and CIII complex formation could be competed away with 25- and 50-molar excess of unlabeled Sp1 oligonucleotide (Fig. 2A,, compare lane 3 with lanes 6 and 7). Furthermore, 25- and 50-molar excesses of unlabeled −80 to −50 bp oligonucleotide probe (FP) competed against the formation of all three CI, CII, and CIII complexes (Fig. 2A,, compare lane 3 with lanes 8 and 9). Okadaic acid and deferoxamine also induced the formation of the CII complex similar to that induced by hypoxia (Fig. 2B).

To further characterize the specific binding of Egr-1, nuclear extracts were incubated with labeled probe in the presence of antibodies specific for Egr-1, Sp1, and Sp3. A new, slower-migrating supershifted (SS) complex was formed upon incubation of the EMSA reaction with Sp1 antibody (Fig. 2C,, lanes 4 and 5, labeled as Sp1 SS). Incubation with the Sp3 antibody did not induce any SS band nor affect the formation of the CII complex (Fig. 2C,, lane 3). The CII complex was diminished after incubation with an anti–Egr-1 antibody, concomitant with the formation of a SS band with reduced intensity (Fig. 2C , lane 2, labeled as Egr-1 SS). These results further showed the presence of Sp1 and Egr-1 proteins in the CI and CII complexes, respectively. Interestingly, although the CIII complex could be competed effectively with cold Sp1 oligonucleotide probe, the Sp1 antibody did not affect the intensity nor supershift the CIII complex. It is conceivable that the epitope recognized by the anti-Sp1 antibody was masked after formation of the CIII complex.

Mutations of Egr-1 binding site abrogated Egr-1 binding and promoter activity. To confirm the role of Egr-1 in the positive regulation of NDRG1 promoter activity, mutations in the Egr-1 binding site were generated in the −80/+355 minimal promoter luciferase construct shown to confer inducibility by hypoxia and deferoxamine. RAW264.7 cells were transfected with the wild-type (WT) or Egr-1 mutant construct (M1) treated with hypoxia and deferoxamine. Mutations of the Egr-1 binding site in the M1 construct did not affect the basal promoter activity compared with the WT promoter construct −80/+355 (Fig. 3A). In contrast, Egr-1 binding site mutations abrogated promoter inducibility by hypoxia and deferoxamine (Fig. 3A), further demonstrating the role of the Egr-1 binding motif in the latter. To show specificity of the Egr-1 and Sp1 binding sites in the CI-CIII complexes, we did EMSA using labeled WT probe and cold competition with unlabeled probes containing mutations in the Egr-1 binding site (M1) or Sp1 binding site (M2). As shown in Fig. 3B, the CII complex (Fig. 3B,, lane 3) was not competed by increasing concentrations of unlabeled M1 probe containing mutations in the Egr-1 binding site (Fig. 3B,, lanes 4 and 5) but was abrogated by unlabeled M2 probe containing mutations in the Sp1 binding site (Fig. 3B,, lanes 6 and 7). On the other hand, the CI and CIII complexes were competed by unlabeled M1 probe (Fig. 3B,, lanes 4 and 5) but not by M2 probe (Fig. 3B , lanes 6 and 7). These results showed the role of the Egr-1 binding site in inducibility by hypoxia and deferoxamine and the specificity of Egr-1 and Sp1 sites in the formation of the CI-CIII complexes in EMSA.

Hypoxia-induced Egr-1 protein correlated with increased NDRG1 expression. To further examine the role of Egr-1 in NDRG1 expression, we determined if hypoxia up-regulated Egr-1 expression. Compared with normoxia, RAW264.7 cells treated with hypoxia for 1 and 6 h had significantly higher levels of Egr-1 protein (Fig. 4A,, top). On the other hand, the levels of Sp1 protein were not affected by hypoxia treatment (Fig. 4A,, bottom). The kinetics of induction Egr-1 protein by hypoxia was consistent with the induction of binding of Egr-1 to the overlapping Egr-1/Sp1 binding motif in the NDRG1 proximal promoter, detected as the CII complex in EMSA Fig. 2A, and correlated with promoter inducibility by hypoxia and deferoxamine (Fig. 3A). Additionally, compared with the early induction of Egr-1 protein at 1 h, hypoxia treatment increased the levels of NDRG1 mRNA and protein beginning at 4 h and continued to increase over 16 to 24 h as determined by Northern blot (Fig. 4B) and Western blot (Fig. 4C), respectively. To ascertain the direct role of Egr-1 expression in the up-regulation of NDRG1 expression, HeLa cells were transfected with Egr-1 expression vector and NDRG1 expression was examined by Northern blot analysis. As shown in Fig. 4D, forced expression of Egr-1 resulted in strong induction of NDRG1 mRNA compared with vector control (Vec) or mock transfection (−).

Requirement of Egr-1 for induction of NDRG1 expression. To investigate whether Egr-1 was required for induction of NDRG1 gene expression, two approaches were used: RNAi gene knockdown and Egr-1 gene knockout cells. First, we showed that hypoxia exposure could induce Egr-1 expression in HeLa cells transfected with control RNAi against green florescence protein (GFP RNAi) whereas transfection of Egr-1 RNAi could efficiently knockdown hypoxia-induced Egr-1 expression (Fig. 5A). Next, the effects of Egr-1 RNAi knockdown on the induction of NDRG1 expression by hypoxia and deferoxamine were examined. As shown in Fig. 5B, compared with untreated control, the levels of NDRG1 mRNA were more strongly induced by deferoxamine than hypoxia in HeLa cells transfected with control GFP RNAi (Fig. 5B,, compare lanes 2 and 3 with lane 1). Compared with the latter, transfection with Egr-1 RNAi significantly reduced the levels of NDRG1 mRNA (Fig. 5B,, compare lane 5 with lane 2, and lane 6 with lane 3). In the second approach, the requirement of Egr-1 in the induction of NDRG1 expression was examined using Egr-1 heterozygous (Egr-1^+/−) and homozygous (Egr-1^−/−) MEFs. NDRG1 gene expression was strongly induced by hypoxia and deferoxamine in Egr-1^+/− MEFs but significantly reduced in Egr-1^−/− MEFs (Fig. 5C , compare lane 2 with lane 5, and lane 3 with lane 6). Collectively, these results showed the requirement of Egr-1 in the positive regulation of NDRG1 gene expression in response to hypoxia signaling.

Functionality of the overlapping Egr-1/Sp1 binding site in the murine NDRG1 promoter. To investigate whether the Egr-1/Sp1 binding site located at the human NDRG1 proximal promoter is conserved, we compared sequences of the proximal promoters of the NDRG1 gene among several species. As depicted in Fig. 6A, in human and chimpanzee, the first two bases “GC” in the Egr-1/Sp1 overlapping binding motif are substituted by “CT” in the mouse and rat motif. To determine the functionality of the two-nucleotide substitution in the murine Egr-1/Sp1 binding site, we did EMSA using a labeled probe encompassing this putative Egr-1/Sp1 binding site present in the murine NDRG1 promoter (shown in Fig. 6B). We observed CI and CIII complex formation in both normoxia- and hypoxia-treated cells (Fig. 6C,, lanes 2 and 3), whereas the CII complex was induced only in nuclear extract from hypoxia-treated cells (Fig. 6C,, compare lane 3 with lane 2). Interestingly, compared with the Egr-1/Sp1 motif in other species, the bovine motif contained an insertion of one “G” in the sequence as GCGGGGGGCG (Fig. 6A). The significance of the G insertion in mediating the binding of Egr-1 protein remains to be determined. Note that the relative position of the Egr-1/Sp1 overlapping binding motif to the TATA box is conserved and are within 100 bp from exon I (Fig. 6A). The relative distance of this Egr-1/Sp1 binding motif to TATA box of several genes, including TF and PDGFs, were compared previously. Interestingly, in all cases, the Egr-1/Sp1 binding motif was located within 100 bp from TATA box. To examine whether the CTGGGGGCG Egr-1 binding motif in the murine NDRG1 promoter could render hypoxia inducibility to a heterologous promoter, two copies of this sequence were inserted upstream of a minimal thymidine kinase (TK) promoter (−80/+30) luciferase construct (mouse Egr1-TK; Fig. 6D). Because we have previously shown that a four-nucleotide replacement could abolish Egr-1 binding and promoter inducibility (Fig. 3A–D), a similar four-nucleotide replacement was engineered in the core Egr-1 binding sequence in the mutant mouse Egr1-TK luciferase construct (Fig. 6D). Hypoxia dramatically increased promoter activity of mouse Egr1-TK construct containing two copies of the Egr-1 binding sites, whereas no hypoxia inducibility was observed when the Egr-1 binding sites were mutated in the mutant mouse Egr1-TK construct (Fig. 6D). In summary, the Egr-1/Sp1 binding site in the NDRG1 promoter is conserved among several species and the Egr-1 binding motif in the human and murine NDRG1 promoters confer inducibility of the NDRG1 promoter and the heterologous TK promoter to hypoxia signaling, respectively.

Hypoxia signaling plays an important role in mammalian development and several pathologic conditions, such as cardiovascular disease and cancer (41). Given the central role of oxygen in the production of ATP through oxidative phosphorylation, it is critical for cells and tissues to respond rapidly to hypoxia (1). The primary response to hypoxia within the cell is the up-regulation of proteins and pathways such as glycolytic enzymes and angiogenic factors that lead to alternative routes of ATP generation and increased oxygen availability. Normal tissues maintain a balance between cellular proliferation and oxygen supply; however, this balance is altered in solid human tumors, resulting in focal hypoxic regions (1, 25, 26). The cells in these regions either adapt to hypoxic stress or die but adaptation to a low-oxygen environment has serious consequences. Hypoxic tumors cells have a higher resistance to radiotherapy and certain chemotherapies (25, 26). Hypoxia can promote a higher mutation rate and select for a more metastatic and malignant phenotype (25, 26).

Environmental stresses, including hypoxia and the iron chelator deferoxamine, have been shown to induce NDRG1 gene expression (27). Recently, iron chelators with high antitumor activity were found to up-regulate NDRG1 expression by HIF-1α–dependent and HIF-1α–independent mechanisms, linking NDRG1 to iron metabolism and tumor cell proliferation (42). Human umbilical vein endothelial cells incubated with homocysteine also showed a significant increase in NDRG1 mRNA expression (11). Elevated levels of homocysteine can damage cells and are associated with atherosclerosis and thrombosis. Hence, NDRG1 seems to be involved in a number of biological systems and may play different roles in various cells and tissues.

We provide the first evidence that Egr-1 mediates hypoxia-inducible expression of the NDRG1 gene through an overlapping Egr-1/Sp1 binding motif present in the promoter of NDRG1. By using various promoter deletion constructs, we were able to define a −80 to −45 bp region responsible for both basal and hypoxia- and deferoxamine-inducible promoter activity. This region contained an overlapping Egr-1/Sp1 GC-rich motif for the binding of the zinc finger transcription factors Egr-1 and Sp1. Similar Egr-1/Sp1 overlapping binding sites have been shown to play a critical role in the expression of other stress-related genes, such as TF and PDGF-A (34, 35). Egr-1 has recently been shown to be critical for the hypoxic up-regulation of TF independent of HIF-1α in glioma cells (36).

Previously, it was proposed that both HIF-1α–dependent and HIF-1α–independent mechanisms were responsible for hypoxia-inducible expression. HIF-1α, activator protein 1, and HIF-2α had been shown to be involved in hypoxia-induced NDRG1 expression (43). A p53 binding site had also been previously reported to regulate NDRG1 expression (20). However, deletion of promoter region containing this putative p53 binding site in the −257/+355, −150/+355, and −80/+355 luciferase constructs (Fig. 1C) did not affect either basal or hypoxia-inducible promoter activity. We had also investigated a putative HIF-1α binding motif (12) and found that neither mutation nor deletion of this site had any appreciable effects on both basal and hypoxia-inducible promoter activities (data not shown). On the other hand, inducibility of a −90/+50 luciferase construct by deferoxamine and okadaic acid was abrogated in HIF-1α^−/− MEFs compared with HIF-1α^+/+ MEFs, supporting the requirement of HIF-1α for induction of NDRG1 expression (data not shown). It is conceivable that hypoxic induction of the NDRG1 promoter by HIF-1α may not be mediated via the putative HIF-1α binding motif; however, the mechanisms remains to be determined. Here, we showed that an overlapping Egr-1/Sp1 motif mediates the induction of NDRG1 promoter by hypoxia or deferoxamine, and mutations in the Egr-1 but not Sp1 binding sites abrogated promoter inducibility. The direct role of Egr-1 in the positive regulation of NDRG1 gene was further shown by transient overexpression of Egr-1, Egr-1 RNAi gene knockdown, and Egr-1 knockout MEFs. Furthermore, insertion of two copies of the Egr-1 binding site upstream of the heterologous TK promoter rendered the latter inducible by hypoxia. Because we could still detect residual NDRG1 mRNA induction after Egr-1 RNAi knockdown and in Egr-1^−/− MEFs after hypoxia and deferoxamine treatment, it is plausible that other members of the Egr family, such as Egr-2, Egr-3, or Egr-4, may have weak activity in place of Egr-1. Nevertheless, Egr-1 is the major Egr family member that mediates maximal promoter inducibility by hypoxia and its mimetics.

NDRG1 has a number of potential functions and downstream effects, ranging from myelin sheath maintenance, differentiation, metastasis suppression, chemotherapy resistance, and enhanced exocytosis in mast cells (44). NDRG1 expression has been shown to be induced by ligands of nuclear transcription factors involved in cell differentiation such as PPARγ (14) and is up-regulated during differentiation of keratinocyte and U937 myelomonocytic cells (8, 45). There is evidence that NDRG1 is involved in cell cycle regulation. NDRG1 has been shown to be a microtubule-associated protein that localized to centrosomes and participates in the spindle checkpoint in a p53-dependent manner (21). In addition to its role in cell cycle regulation and differentiation, NDRG1 plays an important role in the Charcot-Marie-Tooth disease, an autosomal recessive disease caused by demyelination of peripheral nerves disease (46). NDRG1 is expressed in the cytoplasm of Schwann cells and plays a role in Schwann cell differentiation and signaling necessary for axonal survival. NDRG1 knockout mice exhibited progressive demyelination in peripheral nerves, suggesting that the defect was not in the ability to form myelin sheaths but in the maintenance of myelin sheath in Schwann cells (22).

Although it is clear that NDRG1 plays an important role in human cancers, there is some controversy regarding the relationship of NDRG1 expression in certain cancers and in suppression of metastasis (44). The majority of studies have identified NDRG1 as a gene that is down-regulated in cancers and associated with metastasis suppression. However, NDRG1 has also been shown to be up-regulated in prostate cancer due to its response to hormones such as androgens. Studies examining colorectal cancers have found NDRG1 expression to be higher in more advanced lesions, leading to the speculation that NDRG1 is a metastasis promoter (24).

Recently, it was shown that hypoxic pancreatic tumors with benign behavior feature express high levels of NDRG1 protein versus malignant ones (5). Interestingly, the authors reported that hypoxic tumors tested positive for NDRG1 expression were also fibrotic. This could result from the chronic effect of hypoxia-induced Egr-1 expression and fibrin deposition, which promotes the development of fibrosis (47). These findings suggested that NDRG1 expression may be dependent on tumor cell types, differentiation, and hypoxic states, as well as hormonal and metastatic status of the cancers, which, in turn, are governed by transcription factors such as Egr-1 and HIF-1α. Hence, a better understanding of the transcriptional regulation of NDRG1 gene by hypoxia will provide not only mechanistic insights into promoter regulation but also insights underlying the differences in NDRG1 expression in a variety of human cancers.

Grant support: ES00260 (M. Costa and K-M. Tchou-Wong), ES10344 (M. Costa), and T32-ES07324 (M. Costa) from the National Institutes of Environmental Health Sciences; CA16087 from the National Cancer Institute (M. Costa); and DK63603 and CA101234 from the NIH (K-M. Tchou-Wong).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thanked Dr. T-C. Suen for valuable discussion, Jingxia Li for helping in luciferase assay, Lorraine Merceda and Sheryl for help in sequencing experiment, and Juliana Powell for help in preparation of the manuscript.