DDX3, a DEAD Box RNA Helicase with Tumor Growth–Suppressive Property and Transcriptional Regulation Activity of the p21 Promoter, Is a Candidate Tumor Suppressor

DDX3 is a DEAD box RNA helicase with diverse biological functions. Using colony formation assay, our results revealed that DDX3 inhibited the colony formation ability of various tumor cells, and this inhibition might be due to a reduced growth rate caused by DDX3. Additionally, we identified p21, a cyclin-dependent kinase inhibitor, as a target gene of DDX3, and the up-regulation of p21 expression accounted for the colony-suppressing activity of DDX3. Moreover, DDX3 exerted its transactivation function on p21 promoter through an ATPase-dependent but helicase-independent mechanism, and the four Sp1 sites located within the 123 to 63 region, relative to the transcription start site of p21 promoter, were essential for the response to DDX3. Furthermore, DDX3 interacted and cooperated with Sp1 to up-regulate the promoter activity of p21 . To determine the relevance of DDX3 in clinical cancers, the expression profile of DDX3 in various tumors was also examined. A declined expression of DDX3 mRNA and protein was found in f58% to 73% of hepatoma specimens, which led to the reduction of p21 expression in a manner independent of p53 status. Additionally, an alteration of subcellular localization from nuclei to cytoplasm was also observed in >70% of cutaneous squamous cell carcinoma samples. Because DDX3 exhibits tumor suppressor functions, such as a growth-suppressive property and transcriptional activation of the p21 promoter, and is inactivated through down-regulation of gene expression or alteration of subcellular localization in tumor cells, all these features together suggest that DDX3 might be a candidate tumor suppressor. (Cancer Res 2006; 66(13): 6579-88)


Introduction
DDX3, also named as CAP-Rf (1) or DBX (2), is a hepatitis C virus core protein-associated cellular factor that belongs to the DEAD box RNA helicase family (1,3,4).DDX3 harbors nucleotide triphosphatase-deoxynucleotide triphosphatase activity (1), ATPasedependent RNA helicase activity, and a nuclear-cytoplasmic shuttling localization function (5).However, the precise biological functions of DDX3 are still not fully understood.It has been reported that the yeast homologue of DDX3, Ded1, is essential for translation initiation (6), and DDX3 is able to rescue Ded1-deficient yeast mutants (4).Additionally, DDX3 has also been suggested to be involved in pre-mRNA splicing because of its association with functional spliceosome (7) and its possession of an RS-like domain (1).Recently, DDX3 has been reported to play a role in mRNA migration (8) and is required for HIV-1 Rev-RRE export function (5).Moreover, DDX3 is identified as a subunit of mammalian Mediator complex (9) and has a transcriptional activation function (1).These findings indicate a regulatory role for DDX3 in gene expression.
The cell cycle progression of eukaryotic cells is well controlled by cyclin, cyclin-dependent kinases (cdk), and cdk inhibitors.p21 waf1/cip1 is one of the critical cdk inhibitors.Through interaction with the cdk/cyclin complexes, p21 waf1/cip1 modulates their kinase activity, resulting in cell growth arrest (10).Additionally, p21 waf1/cip1 also prevents DNA synthesis and regulates DNA replication through interaction with proliferating cell nuclear antigen (11).Therefore, p21 waf1/cip1 may serve as a growth inhibitor and cell cycle checkpoint.In addition to the modulation of cell proliferation, p21 waf1/cip1 is also involved in the regulation of apoptosis, differentiation, and stress response (11).Furthermore, although mutation of p21 waf1/cip1 is uncommon in cancers (12), p21 waf1/cip1deficient mice show susceptibility to spontaneously or chemically induced tumor formation (13)(14)(15).Hence, p21 waf1/cip1 acts as a tumor suppressor.
Several factors have been reported to induce the expression of p21 waf1/cip1 and the most important one is the tumor suppressor p53 (16).Additionally, p53-independent stimulation including certain extracellular signals, chemical drugs (17), and viral products (18,19) also induce p21 waf1/cip1 expression.The mechanism of p21 waf1/cip1 induction caused by these factors often involves the action of corresponding transcription factors on the p21 waf1/cip1 promoter (17).These transcription factors include signal transducers and activators of transcription, nuclear hormone receptors, Smad3, and the ubiquitous transcription factor Sp1. Sp1 binds to the GC-rich sequence on the proximal region of p21 waf1/cip1 promoter and acts as a mediator for the induction of p21 waf1/cip1 gene caused by transforming growth factor h (TGF-h; ref. 20) and several chemical drugs including mitogen-activated protein kinase inhibitor and histone deacetylase inhibitor (21,22).Furthermore, Sp1 also interacts with other regulatory transcription factors of p21 waf1/cip1 promoter physically and functionally (23)(24)(25).Therefore, Sp1 plays a critical role in the regulation of p21 waf1/cip1 promoter activity.
In this study, we have identified a tumor growth-suppressive property of DDX3.Additionally, we also identified p21 waf1/cip1 as a downstream gene of DDX3.DDX3 enhances the expression of p21 waf1/cip1 gene and up-regulates the promoter activity of p21 waf1/cip1 through an ATPase-dependent but helicase-independent mechanism.The modulation of p21 waf1/cip1 gene expression accounts for the growth-suppressive effect of DDX3.Moreover, our data also revealed that DDX3 transactivates the p21 waf1/cip1 promoter through its multiple Sp1 sites, and DDX3 collaborates with Sp1 on p21 waf1/cip1 promoter activation.Finally, we also showed that DDX3 expression was decreased in hepatocellular carcinoma (HCC) specimens, and the reduced expression of DDX3 might lead to a p53-independent decrease of p21 waf1/cip1 expression.Additionally, an extensive loss of its nuclear localization was observed in cutaneous squamous cell carcinoma (SCC) compared with normal squamous cells.Together, our results show that DDX3 exerts tumor suppressor functions, including growth inhibition and transcriptional modulation of the promoter activity of p21 waf1/cip1 , and is inactivated through deregulated expression or nuclear exclusion in tumor cells.All these features strongly suggest that DDX3 might act as a tumor suppressor.
Antibody preparation.To generate anti-DDX3 antibody, glutathione S-transferase fusion protein containing the NH 2 -terminal 1-226 amino acid fragment of DDX3 [GST/DDX3(1-226)] was used as antigen for immunization.Immunization and collection of antisera were done by the Custom Antibody Services of Protech Technology Enterprise Co., Ltd.(Taipei, Taiwan).Anti-DDX3 sera were depleted with purified GST protein before application to the immunohistochemistry analysis.
Reporter assay.For the luciferase reporter assay, correct amounts of cells (HuH-7, HeLa, NIH 3T3, 293T, HCT116, and HepG2 cells) were transfected with an appropriate amount of reporter plasmid, p21-Luc, or its derivatives and either the empty parental vector or the DDX3 expression construct.The TK-Relina reporter (Promega) was also cotransfected for normalization of transfection efficiency.Transfected cells were harvested at 48 hours posttransfection and a dual luciferase assay was done according to the instruction of the manufacturer.The luciferase activity was measured with an AutoLumat LB953 luminometer (Berthold) and normalized with the cotransfected Relina activity.
Determination of growth rate in DDX3-producing stable clones.HuH-7 and HeLa cells were transfected with pcDNA-SRa/FLAG-DDX3 or parental vector pcDNA-SRa/FLAG and selected with G418 (400 Ag/mL) for 3 to 4 weeks.Independent clones with exogenous DDX3 expression were isolated and maintained in G418 (400 Ag/mL)-containing medium.To compare the growth rate between DDX3-producing cells and control cell lines, cells were seeded at a density of 1.0 Â 10 5 onto 6-cm plates.At each 24-hour interval, the cells were collected, stained with trypan blue, and counted with a hemocytometer.At least two independent experiments were done.
In vitro pull-down assay.GST, GST/Sp1 fusion proteins were purified as previously described (29).For the in vitro GST pull-down assay, 20 AL of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) with immobilized GST or GST/Sp1 fusion proteins (4 Ag each) were incubated with HuH-7 whole-cell extracts (500 Ag) at 4jC for 6 hours.The beads were extensively washed.Proteins bound to the beads were then analyzed by Western blotting using anti-DDX3 polyclonal antibody.
Cancer profiling array.The cancer-profiling array II (BD Biosciences Clontech, Palo Alto, CA) includes normalized cDNA from 154 tumor specimens and corresponding normal tissues from individual patient.To address the mRNA expression profile of DDX3 in these tissues, a 0.7-kb 32 Plabeled cDNA probe spanning the 3 ¶-coding region (400 bp) and 3 ¶untranslated region of DDX3 (300 bp) was generated by the Rediprime II probe labeling kit (Amersham Pharmacia Biotech) and used to probe the cancer profiling array according to the user instruction.
Quantitative real-time PCR.cDNAs prepared from 45 adjacent normaltumor paired hepatoma samples were provided by different sources (30,31) and used in quantitative RT-PCR assay.mRNA expression of DDX3 and p21 waf1/cip1 in adjacent normal liver tissues and HCC samples was examined using quantitative real-time PCR (LightCycler FastStart DNA Master SYBR Green I, Roche Applied Science, Mannheim, Germany) according to the instruction manual using the primer set DDX3-4987F, TTCTCAGATGT-TTGTTGTGTGGATT; DDX3-5142R, AAACTTGCTCAAATGCTATTGCTG; p21-F, AATCCCAGCTACTTGGAAGGC; and p21-R, GCTGACTGCAACCTC-TGCC. Ribosomal protein S24 (RPS24) gene served as an internal control for quantitation using the primer set TGGCTTTGGCATGATTTATGAT and CTTTTTGCCAGCACCAACATT.Real-time detection of the emission intensity of SYBR Green was done with a LightCycler 2.0 instrument (Roche Applied Science).Each experiment was repeated in triplicate, including a no-template negative control.
Identification of p53 mutations in HCC specimens.To determine the p53 status in individual HCC samples, cDNA templates (1 Ag) from 45 HCC samples were mixed with the primer set p53-1-F, ATGGAGGAGCCGCAGT-CAG, and p53-1181-R, TCAGTCTGAGTCAGGCCCTTC, and high-fidelity Easy-A Hi-Fi PCR cloning enzyme (Stratagene) for PCR reactions.Fulllength p53 cDNA fragments were subsequently cloned with pGEM-T easy vector system (Promega) to generate p53 constructs.At least 10 constructs of each individual HCC sample were sequenced and mutations appearing in more than five constructs were considered as somatic mutations of the p53 gene, not PCR-induced errors.
Immunohistochemistry. Tissue sections were placed in 10 mmol/L citric acid buffer (pH 6.0) and subjected to antigen retrieval by microwaving for 20 minutes after deparaffinization and rehydration.After incubation with 3% hydrogen peroxide for 10 minutes and 0.2% horse serum for 20 minutes, a 1:200 diluted rabbit anti-DDX3 polyclonal antibody was applied for 1 hour at room temperature and followed by incubation with biotinylated linker and streptavidin-horseradish peroxidase (LSAB2 system-HRP, DAKO Cytomation, Carpinteria, CA) for 30 and 10 minutes, respectively.The signals were visualized using AEC + (DAKO Cytomation) as the substrate-chromagen at room temperature for 13 minutes.Sections were counterstained with hematoxylin and mounted.A semiquantitative evaluation was used to score the immunoreactivity.For analysis of the normal-tumor paired HCC samples, the staining intensity was compared directly between the tumor part and the normal part.Assessment of the DDX3 staining in the normal skin and cutaneous SCC specimens was done in two ways.Specifically, the percentage of nuclear immunoreactive cells was graded into five categories: 0% to 10% cells stained, 11% to 25% cells stained, 26% to 50% cells stained, 51% to 75% cells stained, and 76% to 100% cells stained.The intensity of nuclear staining was divided into four levels: negative, slight, moderate, and strong.In addition, a m 2 test was used to analyze the loss of nuclear localization in SCC specimens.Differences with P < 0.05 were accepted as statistically significant.

Results
DDX3 exerts an inhibitory effect on cell growth.To explore the biological function of DDX3 on cell growth control, a colony formation assay was done in several tumor cell lines (HuH-7, HCT116, and HeLa) and normal murine fibroblast cells (NIH 3T3) under the selection of G418.As shown in Fig. 1A, forced expression of DDX3 inhibited colony formation activity of HuH-7 and NIH 3T3 cells in a dose-dependent manner and a strong suppression of colony formation was also observed in HeLa and HCT116 cells.This inhibition of colony formation was due to the growth-suppressive ability of DDX3 on tumor cells but was not a result of the inhibitory effect of DDX3 on the expression of G418 resistance gene because DDX3 up-regulated, but not down-regulated, the SV40 early promoter that directs G418 resistance gene expression (data not shown).Moreover, we also established DDX3-overproducing clones in HuH-7 and HeLa cells (Fig. 1B, top) and determined their growth rates by cell counting.As shown in Fig. 1B (bottom), the growth rates of DDX3-overproducing clones, including HuH-7/DDX3#6, HuH-7/ DDX3#7, HuH-7/DDX3#8, HeLa/DDX3#9, and HeLa/DDX3#14, were much slower than the control cell line harboring parental vector without DDX3.This may account for the inhibition of colony formation by DDX3.Taken together, these results indicate that DDX3 is a negative regulator of cell growth control.
DDX3 up-regulates the p21 waf1/cip1 gene expression and inhibits the colony formation activity of tumor cells in a p21 waf1/cip1 -dependent manner.Recently, several RNA helicases, such as RNA helicase A and CHAMP, have been shown to modulate the expression of cell cycle regulator p16 INK4a or p21 waf1/cip1 (32,33).To address the mechanism of growth inhibition by DDX3, we examined whether the promoter activity and the expression level of the cell cycle regulator p21 waf1/cip1 are regulated by DDX3.As shown in Fig. 2A, exogenous expression of DDX3 in various cell lines led to a 2-to 4-fold up-regulation of p21 waf1/cip1 promoterdriven luciferase activity.Additionally, Western blot analysis indicated that the expression level of p21 waf1/cip1 gene was also enhanced by the ectopic expression of DDX3 in HuH-7, HeLa, and 293T cells (Fig. 2B).Because DDX3 up-regulates the promoter activity and the expression level of p21 waf1/cip1 gene without cell type specificity, this suggests that DDX3 acts as a general regulator of p21 waf1/cip1 gene expression.
To address the essential role of p21 waf1/cip1 in DDX3-mediated colony-suppressing activity, a colony formation assay was also carried out in HCT116 and its derived cell line HCT116 p21 À/À that is deficient in p21 waf1/cip1 expression.As shown in Fig. 2C, forced expression of DDX3 in HCT116 cells led to an 85% reduction of colony formation, whereas in HCT116 p21 À/À cells this DDX3mediated suppression of colony formation was dramatically decreased to 20%.This result strongly suggests that DDX3 exerts its suppressing activity on colony formation in a p21 waf1/cip1dependent manner.
The ATPase activity, but not the helicase activity, of DDX3 contributes to DDX3-mediated transcriptional modulation on the p21 waf1/cip1 promoter.Because DDX3 harbors ATPase and RNA unwinding activities, we are interested to know whether these two enzymatic activities are required for the transcriptional regulatory effect of DDX3 on p21 waf1/cip1 promoter.To address this issue, we generated mutations within the DEAD box or SAT motif.
The conversion of DEAD motif to DQAD sequence impairs both ATPase activity and ATPase-dependent helicase activity (34).On the other hand, a point mutation of the SAT motif to AAA sequence leads to a loss of RNA unwinding activity but retains the ATPase activity of the helicase (34).Using these two mutants, we are able to distinguish which activity is required for the transcriptional activation of DDX3 on p21 waf1/cip1 promoter.To this end, increasing amounts of expression constructs of DDX3/DQAD, DDX3/AAA and wild-type DDX3 were cotransfected with p21 waf1/cip1 promoterdriven luciferase reporter.Interestingly, our data revealed that at a higher dosage (2 Ag/well), the wild-type DDX3 and mutant DDX3/ AAA up-regulated the activity of p21 waf1/cip1 promoter f12-fold whereas the DDX3/DQAD mutant only activated the p21 waf1/cip1 promoter activity f5-fold (Supplementary Fig. S1).Because a similar expression level of these three DDX3 expression constructs was detected (data not shown), their differential induction of p21 waf1/cip1 promoter activity suggests that the ATPase activity, but not the RNA helicase activity, contributes to the transcriptional activation of DDX3 on the p21 waf1/cip1 promoter.
As noted, several regulatory elements including six Sp1 sites, two E2F sites, and one AP-2 site reside within the DDX3 responsive region of the p21 waf1/cip1 promoter (17).To delineate which element is essential for the transactivation of DDX3 on p21 waf1/cip1 promoter, p21 waf1/cip1 promoter-directed reporter plasmids harboring single mutation within Sp1-1, Sp1-2, Sp1-3, Sp1-4, Sp1-5/6, E2F(b)-1, E2F(b)-2, and AP-2 sites (see Supplementary Fig. S3A) were assayed for luciferase activity in the presence or absence of DDX3.Interestingly, our data indicated that the induction of reporters destroying AP-2, E2F, and each individual Sp1 site was comparable with that of wild-type reporter (Supplementary Fig. S3B and C).However, for those reporters harboring more than two Sp1 sites mutations, the DDX3-mediated induction of reporter activity was decreased compared with the wild-type reporter, and this decreased level was proportional to the numbers of Sp1 sites destroyed (Fig. 3D).All together, our results indicate that DDX3 transactivates p21 waf1/cip1 promoter through multiple Sp1 sites.
Functional cooperation and physical interaction between DDX3 and Sp1.It is well known that Sp1 and Sp3 can transactivate the p21 waf1/cip1 promoter through Sp1 sites (35).To examine whether DDX3 transactivates the p21 waf1/cip1 promoter activity through Sp1 site-interacting proteins, the Sp1 or Sp3 expression plasmid together with the DDX3 expression construct and the p21 waf1/cip1 promoterdriven luciferase reporter were introduced into HuH-7 cells.As shown in Fig. 4A, transient expression of DDX3 alone resulted in a 4-fold activation of p21 waf1/cip1 promoter activity whereas transiently expressed Sp1 or Sp3 of increasing amount led to a 1.6-to 1.8-fold activation of the p21 waf1/cip1 promoter.Interestingly, the induction effect of DDX3 on the p21 waf1/cip1 promoter was increased to 10-fold when coexpressing with Sp1; however, this coactivation effect was not found when coexpressing with Sp3 (Fig. 4A).This result strongly suggests that Sp1, but not Sp3, is involved in the transcriptional regulation of DDX3 at the p21 waf1/cip1 promoter.
The functional cooperation between DDX3 and Sp1 proteins strongly implies the physical interactions of these two factors.To test this possibility, an in vitro GST fusion protein pull-down assay was done.As shown in Fig. 4B, endogenous DDX3 protein was pull downed by GST-Sp1 but not by GST control beads, suggesting a specific interaction between DDX3 and Sp1 in vitro.The in vivo Figure 2. p21 waf1/cip1 is a target gene of DDX3.A, DDX3 up-regulates the p21 waf1/cip1 promoter activity.p21 waf1/cip1 promoter-driven luciferase reporter (p21-Luc; 0.05-0.5 Ag) and increasing amount (0.1-2 Ag) of FLAG-DDX3 expression construct were cotransfected into various cell lines as indicated.The total amount of transfected DNA was kept constant by adding control vector (pcDNA3-SRa/FLAG).Luciferase activity was measured at 48 hours posttransfection.In all cases, the relative luciferase activity shown is presented as fold activation relative to the control transfection and is derived from at least three independent experiments done in triplicate.B, the expression of endogenous p21 waf1/cip1 is enhanced by forced expression of DDX3.HuH-7, HeLa, and 293T cells were transfected with increasing amount (10 or 20 Ag) of FLAG-DDX3 expression plasmid.After 48 hours of transfection, cell extracts were prepared and the expression level of FLAG-DDX3, p21 waf1/cip1 , and a-actin was detected by immunoblotting with anti-FLAG, anti-p21 waf1/cip1 , and anti-a-actin (Santa Cruz Biotechnology, Santa Cruz, CA) antibodies.The expression level of a-actin was served as a loading control.C, DDX3 exerts the colony suppressive effect through a p21 waf1/cip1 -dependent mechanism.HCT116 and HCT116 p21 À/À cells were cotransfected with pSilencer 3.1-H1 puro and pcDNA-SRa/FLAG-DDX3 (FLAG-DDX3) or its parental vector at indicated amount by Effectene and selected with puromycin for 14 days.Colonies were fixed, stained with crystal violet, and the colony number was counted.Relative colony number in the absence of FLAG-DDX3 transfection was arbitrarily assigned as one.Results were derived from at lease two independent experiments done in triplicate.
interaction between DDX3 and Sp1 was further examined by a coimmunoprecipitation assay.As shown in Fig. 4C, transiently expressed FLAG-Sp1 and endogenous DDX3 were coimmunoprecipitated by anti-FLAG antibody, suggesting that DDX3 could interact with Sp1 in vivo.Thus, the in vitro and in vivo binding analyses clearly show that there is an interaction between DDX3 and the transcription factor Sp1.
The expression or subcellular localization of DDX3 is altered in tumor cells.Because DDX3 exhibits tumor suppressor functions, it is interesting to know whether DDX3 is inactivated in tumor cells by alteration of gene expression.To address this issue, a DDX3-specific DNA probe was hybridized against the cancer-profiling array II (see Materials and Methods).Interestingly, the mRNA expression of DDX3 was declined in the liver tumor specimens (three of three samples) and SCC of the vulva (three of five samples; Supplementary Fig. S4).To further confirm the alteration of DDX3 expression in cancer specimens, the expression profiles of DDX3 were examined in normal-tumor paired HCC samples by immunohistochemical staining analysis using specific anti-DDX3 antibody.As shown in Fig. 5A, decreased expression of DDX3 at the protein level was found in 73% of HCC from 26 patients.Additionally, the mRNA expression of DDX3 was also examined by quantitative RT-PCR, and our data revealed that the DDX3 mRNA level was reduced in 57.8% of 45 HCC specimens (Fig. 5A and B).These results clearly indicate that the expression of DDX3 is down-regulated in liver tumor cells.Furthermore, we also delineated the relationship between DDX3 and p21 waf1/cip1 mRNA expression in these HCC cases by quantitative RT-PCR (Fig. 5B).Our result showed that, among 26 cases with decreased DDX3 expression status, 20 cases (77%) also showed simultaneous p21 waf1/cip1 down-regulation, and the correlation between the expression status of DDX3 and p21 waf1/cip1 was statistically Figure 3. Multiple Sp1 sites (Sp1-1, Sp1-2, Sp1-3, and Sp1-4) located at À123 to À63 related to transcription start site are essential for the transactivation of DDX3 on p21 waf1/cip1 promoter.A, schematic representation of serial p21 waf1/cip1 promoter-driven luciferase reporters used in transactivation assay: p21-Luc contains the 2.3-kb full-length of p21 waf1/cip1 promoter (from À2,326 to +10 nucleotide related transcription start site); its derivatives (À159/+8)p21-Luc, (À123/+8)p21-Luc, (À84/+8)p21-Luc, (À76/+8)p21-Luc, (À63/+8)p21-Luc, and (À56/+8)p21-Luc contain a series of deleted promoters of p21 waf1/cip1 .(À2,326/+8)D(À127/À63)p21-Luc is also a derivative of p21-Luc reporter containing the full-length p21 waf1/cip1 promoter lacking À127 to À63 region.B and C, HuH-7 cells were cotransfected with a p21 waf1/cip1 promoter-driven reporter (p21-Luc) or its derivatives represented in (A) (0.25 Ag of each) together with DDX3 expression construct (pcDNA3-SRa/ FLAG-DDX3) or control vector (pcDNA3-SRa/FLAG; 2 Ag).Luciferase assay was done at 48 hours posttransfection.The fold transactivation of the various p21 waf1/cip1 promoter-directed reporter constructs by DDX3 is shown as fold activation relative to the control transfection and is derived from at lease three independent experiments done in triplicate.D, analysis of the transactivation effect of DDX3 on p21 waf1/cip1 promoters mutated at multiple Sp1 elements.Top, schematic representation of generated mutations within DDX3 responsive region of the human p21 waf1/cip1 promoter.Nucleotide substitutions introduced into the regulatory elements of p21 waf1/cip1 promoter region in mutant constructs are shown in boldface and underlined.The DDX3-mediated transactivation of p21 waf1/cip1 promoter driven by mutant constructs defective in two Sp1 sites (MUT1/2 and MUT3/4), three Sp1 sites (MUT1/2/3 and MUT2/3/4), or four Sp1 sites (MUT1/2/3/4) was measured as described in (B and C ).
Next, the expression of DDX3 in SCC patients was examined.Several cutaneous SCC and their matched normal skin specimens were stained with DDX3 antibody.Our results revealed that among the DDX3-positive cells of normal epidermis, DDX3 was localized mainly in the nuclei and less in the cytoplasm (Fig. 6A).In contrast, in SCC cells, the nuclear distribution of DDX3 was significantly decreased or completely absent, and DDX3 was predominantly detected in the cytosolic compartments instead (Fig. 6A).To further confirm this observation, the intensity of nuclear staining and the percentage of nuclear positive cells obtained from immunohistochemistry analysis of 17 normal epidermis and 34 SCC specimens were analyzed statistically.As shown in Fig. 6B, both the nuclear staining intensity of DDX3 and the percentage of cells positive in DDX3 nuclear staining were much higher in normal epidermis as compared with neoplastic squamous cells (normal versus SCC, 94% versus 18% cases for strong and moderate grade intensity; 94% versus 12% cases with >50% DDX3 nuclear staining positive cells; P < 0.001).This observation indicates that the loss of DDX3 nuclear localization is a frequent event in cutaneous SCC.Together, our results suggest that the decreased expression or loss of nuclear localization of DDX3 may play a regulatory role in the formation of liver tumor and cutaneous SCC.

Discussion
DEAD box RNA helicases participate in disparate cellular functions (36).Here, we have shown by a colony formation assay that DDX3 RNA helicase harbors a growth-suppressive property.Forced expression of DDX3 leads to the suppression of colony formation activity of HCC, cervical carcinoma, colon cancer, and murine fibroblast cells (see Fig. 1A).Furthermore, the growth rate is delayed by ectopic expression of DDX3 (see Fig. 1B).These results support the notion that DDX3 plays a negative role in growth regulation.
In this study, we also identify p21 waf1/cip1 as a target gene of DDX3, and the up-regulation of p21 waf1/cip1 expression is linked to the growth-suppressive effect exerted by DDX3 (see Fig. 2).DDX3 enhances p21 waf1/cip1 gene expression through up-regulation of promoter activity of p21 waf1/cip1 .Because the activation of p21 waf1/cip1 promoter activity was shown in various cell lines with diverse p53 status (p53 wild-type in HepG2, HCT116, and NIH 3T3; p53 mutation in HuH-7; p53 degraded by human papillomavirus E7 in HeLa; p53 inactivated by SV40 large T antigen in 293T), this DDX3-mediated up-regulation on p21 waf1/cip1 promoter seems to be p53 independent.Supporting this notion is the fact that the DDX3induced activation level of serial 5 ¶-deleted p21 waf1/cip1 promoterdirected reporters that lack two p53 responsive elements (at the À2,301 and À1,394 nucleotide region upstream of the transcription start site; ref. 17) is comparable with the full-length p21 waf1/cip1 promoter-driven reporter, p21-Luc (see Supplementary Fig. S2B).Moreover, in this work, the DDX3-responsive region is delineated to 60 bp (between À123 and À63) of the p21 waf1/cip1 proximal promoter region (see Fig. 3B), and the integrity of multiple tandem Sp1 sites in this region is required for the activation induced by DDX3 (see Fig. 3C).This requirement of multiple Sp1 sites to confer DDX3-mediated transactivation of the p21 waf1/cip1 promoter is Because DDX3 harbors growth-suppressive ability and transcriptional modulation activity of the p21 waf1/cip1 promoter, in this work we further examined the deregulated expression of DDX3 in tumors by a cancer profiling array and immunohistochemical staining analysis.Our results clearly indicate that the expression of DDX3 is decreased in HCC and the subcellular localization of DDX3 is altered in cutaneous SCC (see Figs. 5 and 6).Cutaneous SCC is a clinically aggressive cancer, and the inactivation of p21 waf1/cip1 has been shown to play a critical role in the formation of skin tumor in knockout mice models (14,15).Because DDX3 harbors a dominant nuclear localization in normal squamous cells and exerts a transactivation function on p21 waf1/cip1 promoter, presumably the inactivation of transcriptional modulation function of DDX3 through cytoplasmic mislocalization is relevant to the deregulation of cell growth in cancerous squamous cells.
Recently, Huang et al. (44) have reported that the mRNA expression of DDX3 is up-regulated in HCC, and they considered DDX3 to be a promoter for hepatocarcinogenesis.However, according to our present study and recent report (45), both the mRNA and protein expressions of DDX3 are down-regulated in 58% to 73% of HCC specimens (see Fig. 5A).Furthermore, a tendency that DDX3 may regulate p21 waf1/cip1 expression independently from p53 is also noted in HCC specimens (see Fig. 5B-D).This observation is consistent with our in vitro studies (see Figs. 2 and 3), which indicate that DDX3 up-regulates p21 waf1/cip1 promoter activity in a p53-independent manner.Taking into account that it acts as a positive regulator of p21 waf1/cip1 expression and has growth-suppressive ability on liver cancer cells, presumably DDX3 is a candidate tumor suppressor gene in liver.Support for this notion comes from several studies that have indicated the involvement of the p21 waf1/cip1 inactivation in hepatocellular carcinogenesis (46,47).
During the multistep process of carcinogenesis, decreased expression of tumor suppressors is often caused by epigenetic modification or abnormal degradation (48,49).Additionally, alterations of subcellular localization through aberrant splicing or genetic mutations also provide ways to inactivate tumor suppressor genes (50).In view of these, further work to delineate the deregulation of DDX3 expression or localization will shed light on diverse mechanisms for inactivation of DDX3 in different cancers.In conclusion, our study characterizes the biological function of DDX3 as a putative candidate tumor suppressor gene, at least in HCC and cutaneous SCC.Most importantly, due to the universal tumor growth-suppressive property of this gene in various kinds of neoplasm regardless of p53 status, DDX3 is a potential therapeutic target for gene therapy.

Figure 1 .
Figure 1.Ectopic expression of DDX3 inhibits cell growth.A, DDX3 inhibited the colony formation ability of several tumor and normal murine fibroblast cells.HuH-7, HeLa, HCT116, or NIH 3T3 cells were transfected with parental vector or FLAG-DDX3 expression plasmid at the indicated amount by the Effectene or calcium phosphate method (see Materials and Methods).The total amount of transfected DNA was kept constant by adding control vector.Transfected cells were selected with G418 for 2 to 4 weeks.Colonies were then fixed, stained with crystal violet, and counted.In all cases, the relative colony number in the absence of FLAG-DDX3 transfection was arbitrarily assigned as one.Results were derived from at lease three independent experiments done in triplicate.B, top, expression levels of DDX3 in various DDX3-overexpressing HuH-7 and HeLa cell lines and parental cells were examined by immunoblotting with anti-DDX3 specific antibody.Bottom, different clones of DDX3-overproducing cells were seeded and the numbers of viable cells were counted at various time intervals (see Materials and Methods).Results were derived from at least two independent experiments done in triplicate.

Figure 4 .
Figure 4. DDX3 interacts with Sp1 physically and functionally.A, functional interaction between DDX3 and Sp1 on p21 waf1/cip1 promoter.HuH-7 cells were cotransfected with FLAG-DDX3, FLAG-Sp1, FLAG-Sp3 expression construct, and p21-Luc reporter as indicated.The total amount of plasmid DNA was kept constant (2.25 Ag) by the addition of empty vector in each transfection.The luciferase assay was done at 48 hours posttransfection.The fold transactivation of the p21 waf1/cip1 promoter-driven reporter elicited by DDX3 alone or together with Sp1 or Sp3 is presented as fold activation relative to the control transfection and is derived from at lease three independent experiments done in triplicate.B, DDX3 interacts with Sp1 in vitro.GST and GST-Sp1 fusion proteins were purified and analyzed by SDS-PAGE and Coomassie blue staining (bottom ).Asterisk, purified GST-Sp1 fusion protein.Total cell lysates of HuH-7 cells were incubated with GST or GST-Sp1 fusion protein-prebound glutathione-Sepharose 4B beads and the bound fractions were subjected to immunoblotting with antibody against DDX3 (top ).Lane 1, input control (10% of cell extract without any treatment).C, Sp1 interacts with DDX3 in vivo .HuH-7 cells were transfected with 20 Ag of FLAG-Sp1 expression construct (lanes 2 and 4 ).HuH-7 cell extracts (lane 3) or FLAG-Sp1-containing cell extracts (lane 4; 1.5 mg each) were immunoprecipitated with anti-FLAG antibody-conjugated beads and the immunoprecipitates were analyzed by SDS-PAGE followed by immunoblotting with anti-FLAG antibody or anti-DDX3 antibody.Lane 1, input control for mock transfection; lane 2, input control for FLAG-Sp1 expression construct transfected cells.

Figure 6 .
Figure 6.The nuclear localization of DDX3 is altered in SCC.A, immunohistochemical staining of DDX3 in cutaneous SCC and paired normal skin.DDX3 was strongly immunolocalized in the nucleus and less intensively in the cytosol of squamous cells in normal skin, but the nuclear staining was lost or dramatically decreased in cutaneous SCC.N, normal skin; T, cutaneous SCC.B, DDX3 nuclear staining in normal and neoplastic squamous tissues of skin.The intensity of DDX3 nuclear staining and the percentage of DDX3 nuclear staining positive cells in normal skin and cutaneous SCC tissues are summarized.
À/À (kindly provided by Dr. B. Vogelstein, Howard Hughes Medical Institute, Chevy Chase and the Sydney Kimmel Comprehensive Cancer Center of the Johns Hopkins University School of Medicine, Baltimore, MD; ref. 26) were grown in DMEM supplemented with 10% fetal bovine serum or bovine calf serum (NIH 3T3 cells; Invitrogen-Life Technologies Corp., Carlsbad, CA).