The NDRG family has been recently discovered and contains 4 paralogs: NDRG1, -2, -3 and -4. The NDRG family shares a conserved Ndr domain, and the domain similar to the α/β-hydrolase superfamiy.1 Phylogenetic analysis revealed that NDRG1 and -3 belong to one subfamily, whereas NDRG2 and -4 belong to another.2 NDRG1 is distributed primarily in the cytoplasm, but also in the cell membrane and in the nucleus.3 It appears to play a role in growth arrest and cell differentiation, possibly as a signaling protein shuttling between the cytoplasm and the nucleus.4 Various conditions including hypoxia and androgens activate NDRG1 expression.5–9 Overexpression of NDRG1 has been widely found to reduce proliferation rate, enhance differentiation, and suppress the metastatic potency of various types of cancer cells.10, 11 Decreased expression of NDRG1 was correlated with shorter survival in esophageal squamous cell carcinoma.12 The results suggested a role of NDRG1 as a candidate tumor-suppressor gene.13NDRG2 is highly expressed in adult skeletal muscle and brain. NDRG2 up-regulation seemed associated with Alzheimer's disease pathogenesis.14 The expression of NDRG2 in human glioblastoma tissues is significantly lower than that in the normal brain. Transfection of human glioblastoma U373 cells with NDRG2 markedly reduced proliferation of the glioblastoma cells.15 However, little is known regarding the function of NDRG4 and its relationship with tumors.
NDRG3 is expressed in many organs but the highest expression levels were found in testis and prostate in mice.16 It may play a role in mouse spermatogenesis as it was shown to be localized to the outer layers of the seminiferous epithelium.16 However, the function of NDRG3 and its relationship to cancer remains unknown. Because of the accumulation of evidence for the roles of NDRG1 and NDRG2 in tumor progression and the high expression of NDRG3 in human prostate, we investigated the roles of NDRG3 in prostate cancers. In this manuscript, we show that NDRG3 is an androgen-regulated gene and is expressed most abundantly in prostate and testis tissues. We further demonstrate using in vitro studies that overexpression of NDRG3 increases the growth rate and enhances migratory and invasive properties of PC-3 cells. In in-vivo studies, we observed that exogenous expression of NDRG3 in PC-3 cells enhances tumor growth in a xenograft tumor model. We observed that NDRG3 expression is detected in 58.6% (41/70) of PCa tissues, compared to only 13% in benign prostatic hyperplasia (BPH) tissue, suggesting that NDRG3 might play a role in prostate cancer. Finally, DNA-microarray analysis of the effects of NDRG3 overexpression suggested that it up-regulates the expression of many angiogenic chemokines including CXCL1 (chemokine ligand 1), CXCL3 (chemokine ligand 3) and CXCL5 (chemokine ligand 5), which may increase angiogenesis of tumors.
LNCaP, PC-3, DU145, CL-1 and the immortalized prostate stromal cell line WPMY-1 were obtained from American Type Culture Collection (ATCC, Manassas, VA). All the cells except LNCaP cells were maintained and propagated at 37°C with 5% CO2 in RPMI 1640 containing 2 mM L-glutamine, 10 mM HEPES, 1 mM sodium pyruvate, 4.5 g/l glucose and 1.5 g/l sodium bicarbonate and supplemented with 10% FBS (HyClone, Logan, UT) and 100 unit/ml penicillin and 100 units/ml streptomycin. The conditions were the same for LNCaP-cell growth; however, IMDM's medium was used.
Plasmid construction and detection by RT-PCR
pGexT2-NDRG3 is the full-length cDNA of human NDRG3 (NM_022477) cloned into pGexT2 (GE Healthcare Bio-Sciences Corp., Piscataway, NJ), kindly provided by Xie Yi (Fudan University, Shanghai, China). The expression vector pcDNA3.1/myc-His(−)B (Invitrogen, Carlsbad, CA) was used for the expression of full-length NDRG3 cDNA. For the construction of pcDNA/myc-His(−)B-NDRG3, the coding sequence of human NDRG3 cDNA was amplified from pGexT2-NDRG3. The following primers designed from the NDRG3 gene were used for the PCR: 5′-GATCGGATCCGCCACCATGGAT GAACTTCAGG-3′ as forward, and 5′-TAGAAAGCTTGGG CAGGACACCTCCATG-3′ as reverse. The underlined sequence indicates BamH I and Hind III sites in the forward and reverse primers, respectively. The bold sequence shows the start site in the forward primer. A Kozak sequence (the italicized) was introduced to ensure proper initiation of translation. The PCR product was first validated by DNA sequencing and then was digested with BamH I and Hind III for cloning into pcDNA3.1/myc-His(−)B.
Total RNA was extracted using Trizol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. For RT-PCR, first-strand cDNA was synthesized from 2 μg of total RNA using Quant Reverse Transcriptase (Promega, Mannheim, Germany). The NDRG3 RT-PCR primers are 5′-CGCCCATTATAGGAACCCAGGAAG-3′ and 5′-GGCGAATTGTCCCCTACCACCAGT-3′. The β-actin (control) primers are 5′-AGAAA ATCTGGCACCACACC-3′and 5′-CTCCTTAATGTCACGCA CGA-3′. The PCR conditions for the NDRG3 gene were as follows: 94°C for 5 min, followed by 28-cycles of 94°C for 30 sec, 55°C for 30 sec and 72°C for 1 min. The PCR conditions for the β-actin gene were: 94°C for 5 min, followed by 25-cycles of 94°C for 30 sec, 55°C for 30 sec, and 72°C for 1 min. All reactions were done in triplicate. A no-reverse-transcriptase control and a no-template-DNA control were included as negatives.
Cells were lysated with RIPA buffer supplemented with a protease-inhibitor cocktail (Sigma-Aldrich, St. Louis, MO). Typically, 50 μg of protein per lane was loaded on 12% SDS-PAGE gels. The samples were electrotransferred to nitrocellulose membranes and nonspecific binding was blocked in TBST buffer [0.5 mmol/l Tris-HCl, 45 mmol/l NaCl, 0.05% Tween 20 (pH 7.4)] containing 5% nonfat milk. Membranes were incubated overnight with either a 1:200 dilution of anti-NDRG3 goat polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) or a 1:1,000 dilution of anti-β-actin mouse monoclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Membranes were washed and incubated with horseradish peroxidase (HRP) labeled rabbit anti-goat IgG or goat anti-mouse IgG secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA) for NDRG3 and β-actin, respectively. The antigen–antibody reaction was visualized using an ECL assay kit and exposed to ECL film (Fijifilm, Tokyo, Japan). The experiments were conducted in triplicate.
Transfection of cells with overexpression constructs and siRNA
The NDRG3 expression construct or the empty vector alone was transfected into cells using LipofectAMINE 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. PC-3 cells were seeded at 106 per well in 6-well plates. Forty hours after transfection, fresh complete medium containing geneticin (Sigma-Aldrich, St. Louis, MO) at 400 μg/ml was added to the cells for the selection of geneticin-resistant clones.
CL-1 cells were transfected with a mixture of 4 NDRG3-siRNA oligos (Dharmacon, Lafayette, CO). Their sequences are GA UCAAACCACUUCUAAAU, AAACAGACCAGUUAUACUA, GCUAAAGGCUGGAUUGACU and CGCCUGAACCCUAUAA AUA. The Green-fluorescent-protein siRNAs, CGCUGACCCU GAAGUUCATUU and GAGACCUGCUGAACACAAUU17 were used as negative controls. Transfection with siRNA was completed using the TransMessenger Transfection Reagent (Qiagen, Hilden, Germany) according to the manufacturer's protocol. Seventy-two hours after transfection, the cells were lysed for immunoblot analysis using the antibody against NDRG3 or trypsinized and subjected to cell counting using a Coulter counter.
Cell proliferation and growth assays
Cellular proliferation was assessed by counting the numbers of viable cells using a Coulter counter. Each assay was performed in triplicate for at least 3 independent experiments. For the in vitro growth assay, the NDRG3-expressing clone, the clone with the vector alone and the untransfected cell line were seeded at an initial concentration of 3 × 103 cells per well in a 96-well plate. At regular intervals, cells were trypsinized and resuspended. Cell numbers were determined using a Coulter cell counter. Three replicate experiments were performed.
Wound-healing migration assay
About 1.2 million cells from the NDRG3-expressing clone, the clone with the vector alone and the untransfected cell line as control were seeded, respectively, into 6-well plates in complete RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% FBS (HyClone, Logan, UT), 100 μg /ml penicillin and 100 μg/ml streptomycin, and incubated at 37°C with 5% CO2 until confluent. Wounding was introduced to the monolayer of cells by scraping with a sterile pipette tip. The healing process was examined dynamically and recorded with a Canon S 80 digital camera with a microscope adapter. The experiments were done in triplicate.
The NDRG3-expressing PC-3 cells (PN3 and PN10) and their negative controls (PNK and parental PC-3 cells) were harvested while in exponential growth, washed, resuspended in serum-free RPMI 1640. 1 × 106 cells in a 100 μl volume was subcutaneously injected into the flanks of 5-week-old male athymic mice. The trypan-blue exclusion assay was used to ensure cell viability (>99%) before injection. For direct comparison, each mouse was injected with 2 different cell clones, the NDRG3-expressing clones into one side of the flank and the cells with vector alone into the other side of the flank. Tumor sizes were measured in 2 dimensions with calipers every other day. Tumor volumes (mm3) were calculated using the following formula, V = (length × width2)/2.
Tissue-microarray slides (Cybrdi, Xian, China) containing 70 prostate cancer specimens and 38 BPH specimens were obtained from the hospitals localized in Xian, China. The ranges of Gleason scores were between 2 and 6 in these cancer specimens. Specimens were fixed in formalin immediately after operation and embedded in paraffin. Microarrays were built using a manual tissue arrayer (Beecher Instruments). All the prostate cancer specimens used in microarray consisted of central parts of tumors. Microarray paraffin block were treated at 37°C for 50 min, were cut into 4 μm sections and mounted onto silicon slides and the slides were baked at 37°C.
Sections were deparaffinized in xylene and hydrated in graded ethanol solutions and distilled water. Antigen retrieval was performed by microwave processing at 50% power (800 W) for 18 min in 10 mM citrate buffer, pH 6.0. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide in methanol for 20 min followed by washing in PBS pH 7.4 3 times. The sections were blocked with normal rabbit-serum containing PBS buffer for 20 min at room temperature, then incubated with the anti-NDRG3 goat polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, CA, diluted 1:50) overnight at 4°C. For the negative control, normal rabbit serum without primary antibody was used. Slides were washed 3 times with PBS followed by incubation with biotinylated rabbit anti-goat secondary antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) at 37°C for 15 min. Slides were treated with avidin–biotin-peroxidase-complex solution (Santa Cruz Biotechnology, Santa Cruz, CA) at 37°C for 15 min and developed using the DAB substrate kit (Sigma-Aldrich). All sections were counterstained with hematoxylin.
Microarray studies and gene expression validated by RT-PCR
We used the Genome Oligo Set Version 3.0 (Qiagen, Hilden, Germany). It contains 34, 580 70mer probes representing 24,650 genes and 37,123 gene transcripts. Microarray hybridization and data analysis were performed as we described previously.18 Only those genes identified in the 2 replicate experiments were considered truly differentially expressed. Genes that showed more than a 3-fold change were considered differentially expressed.
For gene functional classification, we used the EASE program (http://david.abcc.ncifcrf.gov/). For Gene Ontology terms enrichment analysis, we used Gominer (http://discover.nci.nih.gov/gominer/). The background list for Gominer analysis is the filtered (removing lowly expressed genes) list from our microarrays. Pathway analysis was performed using the Ingenuity Pathway Analysis (IPA) software (www.ingenuity.com).
Gene expression in NDRG3 overexpressing PC-3 cells (PN10) and its controls (PNK) and parental PC-3 cells was also studied by RT-PCR. The sence primers for FGFBP1, CXCL1, CXCL3 and CXCL5 are GCAAACCAGAGGAAGACTGC, ATGGCCCGCGCTGCTCTCTCC, CACTGTTAGGGTAAGGGAATG, CTGT GTTGAGAGAGCTGCGTTGC, respectively. The antisence primers of the 4 genes are GCAGGAAACAGCCTCTGAAC, CTTAACTATGGGGGATGCAGG, CAAGGGAAAGAGAAAC GC GTTTTCCTTGTTTCCACCGTCC, respectively. The PCR conditions for the 4 genes were as follows: 94°C for 5 min, followed by 28-cycles of 94°C for 30 sec, 58°C for 30 sec and 72°C for 1 min. The PCR conditions for the β-actin were stated earlier.
Cell numbers were represented as the mean ± SD. Tumor volumes were represented as the mean ± SD. The student's t-test with a p value of <0.05 was used for the statistical analysis.
NDRG3 is an androgen-regulated gene and is expressed abundantly in prostate and testis tissue
To study the distribution of NDRG3 expression in different human organs/tissues, we decided to query the 32 normal human-tissue massively parallel signature sequencing (MPSS) data available at The National Center for Biotechnology Information (NCBI)'s GEO site.19 As the NCBI's 32 tissues do not include liver, we also conducted a comparison with our own private liver MPSS data generated using the same protocol as we described.20 We found that NDRG3 (MPSS tag, GATCACGAACCCACTCA) was expressed at significant levels only in prostate [30 transcript per million (tpm)] and testis tissue (92 tpm) (Fig. 1a).
To further study, the expression of NDRG3 in human prostate cells, we examined the NDRG3 expression in 4 prostate cancer cell lines (LNCaP, CL-1, DU145 and PC-3) and the immortalized prostatic stromal cell line WPMY-1 by RT-PCR. As shown in Figure 1b, NDRG3 was expressed in the LNCaP, CL-1, DU145 and PC-3 cell lines, as well as in the WPMY-1 cell line (all with 28-cycles of PCR). We next examined levels of NDRG3 protein in these prostate cell lines by western blot using a goat anti-human-NDRG3 polyclonal antibody (Santa Cruz, #sc-19471) that recognizes the N-terminal region of NDRG3. We detected a single specific band of approximately 41 kDa. The same result was obtained with a different goat anti-human-NDRG3 polyclonal antibody (Santa Cruz, #sc-19470) that binds the internal region of human NDRG3 (data not shown). The NDRG3 protein is present in 4 prostate cancer epithelial cell lines as well as in the stromal WPMY-1 cell line. Please note in the above culture conditions, the serum contains physical concentration of androgens.
As NDRG1 is an androgen-regulated gene,7, 21 we wondered whether expression of NDRG3 was also regulated by androgen. LNCaP cells were incubated in charcoal-stripped FBS for 48 hr and then treated with 10 nM of R1881, a synthetic androgen. The cells were then collected for RT-PCR and western blotting. As show in Figure 1c, both the mRNA levels and the protein levels of NDRG3 increased significantly in R1881-treated LNCaP cells when compared to ethanol-treated cells (negative control). The increase in NDRG3 expression continued from 24 to 48 hr (Fig. 1c).
Overexpression of NDRG3 increased the growth rate of PC-3 cells
To investigate the function of NDRG3 in prostate cancer cells, we established 3 stably transfected clone (PN3, PN5 and PN10) by transfecting the NDRG3 expression construct pcDNA3.1/myc/His(−)B-NDRG3 into PC-3 cells and by subsequent screening of the transfected clones. We used PC-3 cells for the experiments, because PC-3 cells have the lowest native expression level of NDRG3 protein in the 4 prostate cancer cell lines that we screened. For use as a negative control, we also generated a stable clone (PNK) that harbors the empty vector. We compared the levels of NDRG3 protein in the parental PC-3 cells, PNK, and 3 clones of PC-3 cells with NDRG3 overexpression by western blot. All of these 5 cell lines expressed endogenous NDRG3 as shown by the 41 kDa band (Fig. 2a). Exogenous Myc/His-NDRG3 fusion protein was observed as a 44 kDa band due to the addition of Myc/His tag, which is approximately 3 kDa (Fig. 2a). The expression levels of the NDRG3 fusion protein were dramatically higher than the endogenous levels of NDRG3 protein in PN3, PN5 and PN10 cells (Fig. 2a).
To examine whether the expression of NDRG3 has an effect on prostate cancer cell growth, the growth rates of PN3, PN5, PN10, PNK and the parental cell line were determined by counting the cells 5 days after seeding of 5,000 cells per clone. At Day 5, the cell numbers were 822.8% ± 0.300% (mean ± SD), 796.7 ± 0.603, 740.0 ± 0.200, 445.3 ± 0.219 and 480.8 ± 0.347, respectively, for PN10, PN5, PN3, PNK and the parental PC-3 cells comparing to those at Day 0. The growth rates of all the clones that overexpress NDRG3 were higher than those of PNK and parental PC-3 cells (p < 0.05) (Fig. 2b). There was no significant difference between the growth rates of PNK and parental PC-3 cells. These data indicate that overexpression of NDRG3 promotes the proliferation of prostate cancer cells.
Overexpression of NDRG3 increased migratory property of PC-3 cells
Wound-healing assays have been applied for the analysis of migration of prostate cancer cells,22 therefore, we performed a wound-healing assay to examine the effect of NDRG3 overexpression on cell migration. Cells at 1 × 106 from each examined group were seeded in 6-well plates until confluent, and wounds were introduced by the use of a pipette tip. Over the course of the next 12 hr, PC-3 cells migration was monitored. We found that the migration of both PN3 and PN10 across the wounded area was faster than the parental and PNK cells (Fig. 2c), indicating that NDRG3 overexpression can promote PC-3 cell migration. However, a small contribution of elevated proliferation rates of NDRG3-overexpressing clones to the increased numbers of cells in the wound areas during the limited period cannot be excluded.
NDRG3 overexpression promoted tumor formation in a nude mouse model
We then explored the in vivo role of NDRG3 in tumor formation using a nude-mouse model. PN3, PN10, PNK and the parental PC-3 cells were implanted into 5 male athymic nude mice, respectively. Four of the five mice in each group developed tumors. The NDRG3-overexpressing PN3 and PN10 inoculated mice developed tumor earlier than the controls PNK and the parental PC-3 cells. By Day 48, tumors formed by NDRG3-overexpressing clones (PN3 and PN10) were bigger than those formed by the controls (PNK and parental PC-3 cells). The biggest difference (1.95 ± 0.079 cm3vs. 0.55 ± 0.091 cm3, p < 0.01) in tumor volumes was observed between the tumors formed by PN10 and PNK cells (Fig. 2d). Interestingly, tumors formed by PN10 cells are significantly bigger than those by PN3, which seems to correlate with much higher levels of NDRG3-expression in PN10 cells comparing to PN3 cells. Finally, we noticed that the volumes of tumors formed by the parental PC-3 cells (0.63 ± 0.208 cm3) are slightly larger than those by PNK cells (0.55 ± 0.091 cm3), but it is not statistically significant. Taken together, the data suggested that NDRG3 overexpression promotes tumor cell growth in vivo. We did not observe any metastasis probably because that we did not keep the mice long enough.
Knockdown of NDRG3 expression in CL-1 cells reduced their growth rates
We showed the effect of overexpression of NRDG3, we then examined the effect of knocking down NDRG3 expression in prostate cancer cells. As we showed previously, the endogenous level of NDRG3 protein is the highest in CL-1 cells compared to the 3 other prostate cancer cell lines (Fig. 1), and so, we decided to use CL-1 cells for the NDRG3 knockdown experiment. CL-1 cells seeded in 6-well plates were transfected with NDRG3-siRNA and GFP-siRNA (as a negative control). Three days after transfection, the cells were harvested for western blotting analysis. As shown in Figure 3a, NDRG3-siRNA knocked down most of the endogenous NDRG3 protein when compared to the negative control. However, NDRG3-siRNA did not affect the level of β-actin.
We then examined the effect of knocking down NDRG3 on cell-growth rate. CL-1 cells were seeded at 10,000 cells per well in 96-well plates. After 24 hr, they were transfected with NDRG3-siRNA and were continuously incubated until cell counting at Day 3 and Day 5. The CL-1 cells treated with transfecting reagent alone (mock control) grew slower compared to the untreated cells (data not shown). The numbers of CL-1 cells transfected with GFP-siRNA (vector control) increased by 89% and 3 days after the transfection, whereas the numbers of CL-1 cells transfected with NDRG3-siRNA did not change significantly (Fig. 3b). After incubation for additional 2 days, the cell numbers of the 2 groups of cells increased 332% and 42%, respectively. As a result of different growth rates, the total numbers of CL-1 cells transfected with NDRG3-siRNA is approximately 2-fold and 3-fold less than that from cells transfected with GFP-siRNA at 3 and 5 days after transfection, respectively, indicating that knockdown of NDRG3 expression in CL1 cells reduced their growth rates.
Immunohistochemistry of NDRG3 in prostate cancer tissues
The data we obtained so far showed that overexpression of NDRG3 promoted cell growth, increased cell migration in vitro and enhanced tumor formation of prostate cancer cells in in-vivo mouse models. We next attempted to determine the role of NDRG3 in human prostate cancer. To investigate this, we first performed a comparative analysis of prostate cancer and BPH tissues by immunohistochemistry. NDRG3 expression was detected in 13.2% (5/38) of the BPH specimens, but in 58.6% (41/70) of the prostate cancer specimens. NDRG3 expression was detected in the epithelial cells in BPH tissue, and no significant staining was observed in the stromal cells in BPH tissue (Fig. 4a). Among the 41 NDRG3-positive prostate cancer specimens, expression of NDRG3 was detected only in epithelium in 18 (43.9%) specimens, only in stromal cells in 9 (22.0%) specimens, and both in epithelium and stromal cells in 14 (34.1%) specimens (Fig. 4c). The results suggest that NDRG3 is overexpressed both in epithelium and strome of prostate caner, compared to that in BPH. However, we did not observe enhanced intensity of staining with increasing tumor grade.
We further analyzed the cellular distribution of NDRG3 in these specimens. The staining was observed in the cytoplasm of the luminal glandular cells in BPH (Fig. 4b). In prostate cancer cells, NDRG3 expression was heterogeneous with majority of the staining being detected in epithelium (Figs. 4c–4h). The staining seems to be stronger in tumor cells than that in BPH cells. In addition, we detected that NDRG3 is localized both in nucleus and cytoplasm of epithelial (Figs. 4f and 4g) and stromal (Fig. 4d) cells in prostate cancer specimens, whereas the nuclear staining is much stronger than cytoplasmic staining. In 18.8% (6/32) NDRG3-positive prostate epithelial cells, NDRG3 expression was detected mostly in epithelial cell nucleus and rarely in the cytoplasm (Fig. 4h). These data may imply that nuclear localization seems to be the predominant location of NDRG3 expression in prostate cancer specimens. Finally, we showed that the antibody is specific to NDRG3 by demonstrating a loss of signal when the antibody was premixed with peptides (Fig. 4i).
Differentially expressed genes regulated by NDRG3 overexpression
We compared gene-expression profiles between PNK and PN10 cells to study gene expression induced by NDRG3 overexpression. We identified a list of 109 genes that were up-regulated more than 3-fold by overexpression of NDRG3 (Supp. Info. Table I). We identified 667 genes that were down-regulated by NDRG3 overexpression (Supp. Info. Table II). To further validate the NDRG3-responsive gene expression detected by microarrays, we selected 5 genes (FGFBP1, CXCL1, CXCL3, CXCL5 and ADAM9) and determined their expression by RT-PCR in PN10 cells and the 2 negative controls (PNK and the parental PC-3 cells). We confirmed over expression of FGFBP1, CXCL1, CXCL3 and CXCL5 in PN10 comparing to that in the controls (Fig. 5).
Table I. Functional Enrichment of NDRG3 Regulated Genes
Mitogen-activated protein kinase-activated protein kinase 2
Serine/threonine kinase 4
Microtubule associated serine/threonine kinase family member 4
Ribosomal protein s6 kinase, 90kda, polypeptide 2
snf1-like kinase 2
wnk Lysine deficient protein kinase 4
Aurora kinase a
Oxidative-stress responsive 1
Polo-like kinase 2 (drosophila)
src-Related kinase lacking c-terminal regulatory tyrosine and n-terminal myristylation sites
ptk6 Protein tyrosine kinase 6
rio kinase 3 (yeast)
Analysis of the enrichment of functional classification of these genes by the EASE program (david.abcc.ncifcrf.gov) revealed that 3 groups of genes have enrichment scores greater than 1. They are the core-histone group, the small chemokine/interleukin 8 protein group and the metallothionein superfamily protein group (Table I). There are 15 members of histone proteins that were up-regulated by NDRG3 expression, suggesting that NDRG3 is important for regulation of functions of histones. In the NDRG3 down-regulated genes, functional-enrichment analysis revealed that the protein-kinase group is enriched (Table I).
Further in-depth pathway analysis using the IPA software (www.ingenuity.com) revealed that the top IPA network affected by NDRG3 over-expression was the IPA's cancer pathway (Score 47) and the second top-ranking IPA network was the IPA's cellular movement pathway (Score 44). Figure 6 shows the IPA's cellular movement network with gene expression changes induced by NDRG3 over-expression overlaid in color (red, up-regulated; green, down-regulated). In the network, the key genes IL1B, ATF3, ADAMTS1, IKK, PTX1, CXCL1 and CXCL5 are all up-regulated by NDRG3 over expression.
We have shown that NDRG3 expression levels were highest in testis and second highest in the prostate in the 33 MPSS datasets that we have (Fig. 1a). Interestingly the MPSS tag, GATCACGAACCCACTCA, does not map to the 3′-most GATC site (the enzyme site used for generating 3′ MPSS tags) of the NDRG3 (NM_022477, 2928 bps) but maps to the nucleotide 1,023–1,039 of NM_022477, suggesting that the form we detected by MPSS is probably a short isoform of NDRG3. Zhao et al.16 previously showed by northern blot that even though a band of about 2.6 kb was the strongest in the human prostate, a 1.3 kb band was most abundant in testis. Our MPSS data probably identified the 1.3 kb isoform of NDRG3, and the expression pattern we observed is consistent with Zhao et al.'s findings. The different isoforms of NDRG3 that are reported in the Genbank differ in their 3′ UTR regions, but they all encode about 375 amino acids with a calculated molecular weight of 41,409 Da. By western blot using different polyclonal antibodies against different regions of NDRG3, we have only observed a specific band with molecular mass of about 41 kDa in the 5 aforementioned prostate cell lines, confirming the prediction that there is probably only a single protein isoform in human prostate cells. Selective and high expression of NDRG3 in the prostate tissue suggests potential involvement of NDRG3 in prostate development and carcinogenesis. Interestingly, both the expression of NDRG3 mRNA and protein are much higher in CL-1 cells than in LNCaP cells (Fig. 1b). This is also consistent with our MPSS analysis of LNCaP and CL1 data showing that NDRG3 expresses at 1 tpm (transcript per million) in LNCaP cells and 15 tpm in CL1 cells.20 CL-1 cells are androgen-independent derivatives of LNCaP cells and do not express androgen receptor.20, 23 This result suggests that NDRG3 expression also might be regulated via the mechanisms that are androgen-independent and NDRG3's role in androgen-dependent cells and androgen-independent cells may be different.
We examined the effects of NDRG3 on the growth and migration potential of prostate cancers by overexpression or knockdown of NDRG3 expression. We showed that overexpression of NDRG3 increased growth rates of PC-3 cells and promoted colony formation and migration properties of PC-3 cells in in-vitro cell-culture systems (Fig. 2). We also showed that over-expression of NDRG3 enhanced tumor growth in a nude-mouse model. Complementarily reduction of NDRG3 expression by siRNA suppressed the growth rate and the colony-forming abilities of CL-1 cells (Fig. 3). To study over expression of NDRG3 in mouse model, we had to select stably transfected NDRG3 expression clones first. It is possible that we selected a subpopulation of cells that are resistant to the presumed antiproliferative effect of NDRG3 based on inference from NDRG1's antiproliferative role. However, our siRNA know-down study, which is a transient study, is consistent with our finding that NDRG3 has a proliferative role. This suggests that NDRG3 indeed has a proliferative and tumor promoting role, which is not an artifact due to stable clone selections.
We show here that NDRG3 seems to promote tumor growth, which is different from NDRG1, another member of the NDRG family, which has been shown to be a tumor/metastasis suppressor that is down-regulated in cancer.13, 24 However, a closer review of the literature suggest that there are great controversy in the NDRG protein family's expressions and roles in cancer. Whether NDRG1 is a tumor suppressor or promoter is controversial. A recent survey of NDRG1 expression in a large number (233) of prostate cancer specimens (223) failed to reveal up-regulation or down regulation of NDRG1 in these samples.25 In addition, NDRG1 expression was found to be higher in mouse skin carcinomas than in normal mouse skin.26 Two recent studies showed that NDRG1 protein was significantly increased in colorectal tumor compared to normal epithelium in Chinese, Japanese and US populations.27, 28 All these data suggest that NDRG1 may also have tumor promotion roles, which is similar to what we showed here for NDRG3's role in cancer. The NDRG family has 4 members NDRG1, 2, 3, 424 and each member may play different roles in carcinogenesis.
We detected NDRG3 expression in a prostate stromal cell line (Fig. 1b). We further showed that stromal staining of NDRG3 was detected in 56.1% of the 41 NDRG3-positive prostate cancer specimens. Stromal cells can exert an influence on nearby epithelial cells through paracrine growth-factor signals and the extracellular matrix.17, 22, 29–31 Crosstalk between carcinoma epithelial cells and adjacent stromal cells might alter tumor microenvironment to promote tumor progression.30, 32 It is possible that high levels of NDRG3 in the stromal cells promote epithelial prostate cancer. Whether artificially increasing expression of NDRG3 in stromal cells will actually increase epithelial prostate cancer growth remains to be investigated. Moreover, we detected nuclear localization of NDRG3 in epithelium and stroma of prostate cancer cells and found that the nuclear expression level is significantly higher than the cytoplasmic expression level (Fig. 4). But we did not observe this pattern in BPH. The results suggest that there is a significant accumulation of nuclear NDRG3 during prostate carcinogenesis.
The function of NDRG3 is not known. Microarray analysis revealed that NDRG3 could regulate the expression of a large number of genes. We found that up-regulated genes are enriched in core histones, small chemokines and metallothionein-superfamily proteins by EASE pathway analysis (Table I). We showed by microarray and by RT-PCR that NDRG3 over-expression up-regulates the expression of many members of the small-chemokine family including the C-X-C-chemokine-family proteins CXCL5, CXCL1 and CXCL3. The chemokines are a family of small molecules that mediate cell migration, cell activation, cell differentiation and angiogenesis. The CXC-chemokine family, a subfamily of chemokines, is unique in their ability to regulate angiogenesis. Angiogenesis is associated with tumorigenesis, but the effects differ depending on whether a specific CXC chemokine has either angiogenic or angiostatic properties. CXCL1, CXCL3 and CXCL5 are all angiogenic CXC chemokines.33 Up-regulation of these chemokines may enhance angiogenesis of tumors, and eventually promote tumor growth. Engl et al.34 showed that CXCL1, CXCL3, CXCL5 and CXCL6, chemokines with proangiogenic activity, were highly expressed in highly aggressive tumor cells DU-145 and PC-3 cells, but not in low aggressive LNCaP cells. They suggested that prostate tumor malignancy might be accompanied by enhanced synthesis of angiogenesis stimulating CXC chemokines. We showed that NDRG3 might be an important regulator of the expression of these angiogenesis chemokines, and therefore a potential target for modulating the expression of chemokines as a therapeutic approach for prostate cancer.
By the IPA, we found that the IPA's cellular movement was a key network involved in NDRG3 mediated function. In the network, 3 cytokine genes CXCL1, CXCL5, IL1B (interleukin 1, beta) were up-regulated by NDRG3. We have discussed the potential role of CXCL1 and CXCL5 in prostate cancer in the previous paragraph. For IL1B (interleukin 1, beta), Zabaleta recently showed that polymorphism in IL1B, together with that in IL10, has effect in conferring prostate cancer risk.35 In this network, we also showed that ADAMTS and PTX3 are 2 genes up-regulated by NDRG3. ADAMTS (ADAM metallopeptidase with thrombospondin Type 1 Motif, 1) and PTX3 (pentraxin-related gene, rapidly induced by IL-1 beta) are 2 genes expressed in the extracellular domains. These 2 genes may have roles in controlling cell movement and migration. In contrast, NDRG3 down regulated a few key genes PTGFR (prostaglandin F receptor), FGF18 (fibroblast growth factor 18) and GRB7 (growth factor receptor-bound protein 7). However, the function consequences of down-regulation of these proteins remain to be investigated.
The mechanism of NDRG1's action remains to be further investigated. We previously showed that NDRG1 directly binds to β-catenin and E-cadherin, and its interactome interacts with the androgen response program through common interactions with β-catenin and heat shock protein 90.21 Whether NDRG3 can bind to β-catenin and E-cadherin or interact with the androgen response program remains to be investigated.
The authors thank Dr. Xie Yi for providing them with the pGexT2-NDRG3 cDNA clone. R.L. acknowledges financial support from the National Basic Research Program of China, Shanghai Municipal Committee of Science and Technology and B.L. acknowledges financial support from the Ministry of Science and Technology, China.