Cytoplasmic Accumulation of the RNA-binding Protein HuR Stabilizes the Ornithine Decarboxylase Transcript in a Murine Nonmelanoma Skin Cancer Model*

Ornithine decarboxylase (ODC) is the first and usually rate-limiting enzyme in the polyamine biosynthetic pathway. Under normal physiological conditions, polyamine content and ODC enzyme activity are highly regulated. However, the induction of ODC activity is an early step in neoplastic transformation. The studies described here use normal mouse keratinocytes (C5N cells), and spindle carcinoma cells (A5 cells) to explore the regulation of ODC in nonmelanoma skin cancer development. Previous results have shown that induction of ODC activity is both necessary and sufficient for the promotion of skin tumors. We see a marked increase in ODC enzyme activity in A5 cells compared with C5N keratinocytes, which correlates with a 4-fold stabilization of ODC mRNA. These data suggest that ODC is post-transcriptionally regulated in skin tumor development. Thus, we sought to investigate whether the ODC transcript interacts with the RNA-binding protein HuR, which is known to bind to and stabilize its target mRNAs. We show that HuR is able to bind to the ODC 3′-UTR in A5 cells but not in C5N cells. Immunofluorescence results reveal that HuR is present in both the nucleus and cytoplasm of A5 cells, whereas C5N cells exhibit strictly nuclear localization of HuR. Knockdown experiments in A5 cells showed that when HuR is depleted, ODC RNA becomes less stable, and ODC enzyme activity decreases. Together, these data support the hypothesis that HuR plays a causative role in ODC up-regulation during nonmelanoma skin cancer development by binding to and stabilizing the ODC transcript.

Ornithine decarboxylase (ODC) 2 catalyzes the conversion of the amino acid ornithine to the diamine putrescine, which is subsequently converted to the higher polyamines spermidine and spermine (1,2). Polyamines are small, ubiquitous polycations and are necessary for normal cell growth and development. Thus, the ablation of ODC is lethal in utero (3). Because of their charge, polyamines are able to bind to RNA, DNA, and proteins and thereby influence gene expression (4). Both polyamine content and ODC enzyme activity are tightly regulated in cells, and ODC is induced in response to a variety of proliferative stimuli by alterations in its transcription, translation, and protein degradation (5)(6)(7)(8)(9)(10). A link between neoplastic transformation and increased ODC enzyme activity, as well as increases in intracellular putrescine and spermidine, have been well documented in animal models of skin carcinogenesis and other epithelial tumors (11)(12)(13)(14)(15)(16)(17). Moreover, a link between human nonmelanoma skin cancer (NMSC), the most prevalent cancer in the United States, and elevated ODC enzyme activity levels has been described (18). The induction of ODC activity during tumor development has been attributed to both increased Odc gene transcription and translation of the ODC mRNA (9,19).
Interestingly, changes in intracellular polyamine pools have also been shown to affect the localization of RNA-binding proteins (RBPs), which influence the stability of their target mRNA transcripts (20 -22). Control of mRNA decay is a rapid means of regulating the expression of mRNAs (23). RBPs bind to adenosine-and uracil-rich elements (AREs), sequences classically located within the 3Ј-UTR of labile mRNAs (24). These sequences behave as cis-acting elements and are located in numerous proto-oncogene, cytokine, and transcription factor mRNAs (25). Three categories of RBPs have been defined: those that stabilize mRNA, those that destabilize mRNA, and those that can both stabilize and destabilize mRNA (26). Additionally, it has been shown that some RBPs can alter the translational efficiency of their bound RNA transcript (27,28).
HuR (human antigen R) belongs to the family of Hu proteins that are highly homologous to the embryonic lethal abnormal vision family of proteins. Unlike its family members HuB, HuC, and HuD, which are expressed exclusively in neurons, HuR is ubiquitously expressed (28). The HuR protein contains three RNA recognition motifs (RRMs) (29). RRM1 and RRM2 have been shown to recognize AREs located within the mRNA, whereas it has been suggested that RRM3 recognizes the poly(A) tail of the mRNA transcript (29). The hinge region of HuR, located between RRM2 and RRM3, contains a nucleocytoplasmic shuttling sequence. Both a nuclear localization signal and nuclear export signal are contained in this region (30). HuR is mainly nuclear but has been shown to shuttle between the nucleus and cytoplasm in response to various stress stimuli such as UVB irradiation and H 2 O 2 (31,32). In the cytoplasm, HuR can stabilize and/or increase the translational efficiency of its bound mRNA transcript (29). Interestingly, a correlation between HuR localization and neoplastic transformation has been described. HuR has been shown to bind to and stabilize numerous transcripts whose overexpression contribute to tumorigenesis, such as TSP1, COX2, and ProT␣ (31,33,34). The studies described here support the hypothesis that HuR plays a major role in the regulation of ODC mRNA decay in skin tumor development.
The mouse keratinocyte cell model used in these studies is comprised of two cell lines: C5N cells, which have a wild-type keratinocyte morphology and contain no detectable mutations in the H-ras gene, which is mutated at a high frequency in NMSC; and A5 cells, which have a spindle carcinoma phenotype and were isolated from the tumor of a mouse that had been subjected to the classical two-stage protocol of initiation and promotion (35). Here we show that ODC enzyme activity is induced, and ODC mRNA is markedly stabilized in A5 cells when compared with C5N cells. Using both in vitro and intracellular assays, we show that HuR associates with the 3Ј-UTR of ODC in transformed cells, but not in normal keratinocytes, and causes the ODC transcript to be stabilized. We further demonstrate that the binding status of HuR to the ODC mRNA transcript correlates with HuR localization, because HuR is exclusively nuclear in C5N cells but is both nuclear and cytoplasmic in A5 cells. Thus, these studies show for the first time that ODC is post-transcriptionally regulated by the stabilizing RBP HuR in a mouse nonmelanoma skin cancer model.

EXPERIMENTAL PROCEDURES
Cell Culture-The C5N and A5 mouse keratinocytes (a generous gift from Dr. Allan Balmain, University of California-San Francisco, San Francisco, CA) were cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA), 1% penicillin/streptomycin, and 1% glutamine. These cells have been described previously (35). Passages 5-20 were used in the experiments. Stock flasks were incubated at 37°C in a humidified atmosphere of 95% air, 5% CO 2 .
Assay for ODC Enzyme Activity-ODC enzyme activity was determined by radiometric assay measuring the amount of 14 CO 2 released from each reaction in which L-[1-14 C]ornithine is enzymatically converted to 14 CO 2 by ornithine decarboxylase. Collection and analysis were carried out as described (36).
RNA Extraction, Northern Blot Analysis, and Actinomycin D Experiments-RNA was extracted using the TRIzol reagent according to the manufacturer's instructions (Invitrogen). For Northern blot analysis 20 g of RNA was loaded onto a 1.2% (v/v) agarose-formaldehyde gel and subsequently transferred to a Hybond Nϩ membrane (GE Healthcare). After transfer, the membrane was UV cross-linked using a Stratalinker (Stratagene, La Jolla, CA). Northern analysis was conducted using 32 P-labeled cDNA probes synthesized for ODC and cyclophilin A. The primers are listed in supplemental Table S1. The bands were quantitated using Syngene Software (Syngene, Frederick, MD), and the ODC mRNA data were normalized to that of cyclophilin A. The stability of the ODC mRNA was assessed by the addition of actinomycin D (5-10 g), and RNA was extracted at 0, 4, and 8 h. ODC mRNA half-life was determined using curve fit analysis.
Real Time PCR Analysis-Total RNA was isolated as described above and used in reverse transcription and PCR amplification. qRT-PCRs were performed by the Step One Plus real time PCR system (Applied Biosystems, Foster City, CA), using the same primers for ODC and cyclophilin A described above. Analysis was performed by using the Step One Plus real time PCR software (Applied Biosystems). The data were normalized to cyclophilin A.
Western Blotting-Western blotting was conducted as described previously (37). To prepare cytoplasmic and nuclear extracts, the cells were fractionated using the Pierce fractionation kit as per the manufacturer's instructions (Pierce). To check the extracts for nuclear contamination, Western blot analysis was conducted on 20 g of extract, and ␣/␤-tubulin and histone H3 antibodies were used. The antibodies recognizing HuR (Santa Cruz Biotechnology, Santa Cruz, CA), GAPDH (BioDesign International, Saco, ME), histone H3, and ␣/␤-tubulin (Cell Signaling, Danvers, MA) were all used at a 1:1000 dilution.
Preparation of Synthetic RNA Transcripts-cDNA from C5N and A5 cells was used as a template for PCR amplification of the 5Ј-UTR, coding region, and 3Ј-UTR of ODC. The sense primers for each reaction contained the T7 RNA polymerase promoter sequence on the 5Ј end and are listed in supplemental Table S1. The PCR templates were used to transcribe biotinylated RNAs by using T7 polymerase and biotin-labeled CTP (Invitrogen) as reported previously (22). To elucidate the binding site of HuR on the 3Ј-UTR of ODC, various fragment and deletion constructs were created using sense primers containing the T7 promoter sequence and antisense primers that created varying length fragments of the ODC 3Ј-UTR. GAPDH and c-Myc were utilized as negative and positive controls, respectively. All of the primers used are listed in supplemental Table S1.
Biotin-labeled RNA Protein Binding Assay-The biotin-labeled RNA transcripts (15 l) were incubated with 120 g of cytoplasmic lysate for 30 min at room temperature. The RNAprotein complexes were then harvested by using paramagnetic streptavidin-conjugated Dynabeads (Invitrogen) and were analyzed by Western blotting using the HuR antibody described above.
RNA Protein Binding Assays-For immunoprecipitation (IP) of endogenous RNA-protein binding complexes, 2000 g of cytoplasmic lysate was incubated for 2 h at room temperature with 50% (v/v) suspension of protein A-Sepharose beads (Sigma-Aldrich) that had been precoated with 30 g of either mouse IgG (Invitrogen) or HuR antibody as described (38). The RNA IP material was analyzed by washing the beads with NT-2 buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , and 0.05% Nonidet P-40) and further incubated with 0.5 mg/ml proteinase K, NT-2 buffer, and 0.1% SDS for 15 min at 55°C. RNA was precipitated using phenol-chloroform extraction in the presence of glycoblue (Qiagen). RNA from the IP was then reverse transcribed to detect the presence of ODC mRNA. The primers used for the ODC coding region are listed in supplemental Table S1.
Immunofluorescence Staining-A5 and C5N cells were seeded onto glass coverslips that had been precoated with poly-L-lysine (Sigma-Aldrich). The cells were allowed to grow to 70% confluency and were then fixed in 50/50 acetone/methanol solution at Ϫ20°C for 5 min. The samples were incubated overnight at 4°C in 5% goat serum in PBS. HuR antibody was placed on the samples (diluted 1:500 in PBS containing 0.1% BSA) for 2 h. The samples were then incubated in Cy-2 secondary antibody at a 1:200 dilution in PBS containing 0.1% BSA for 1 h (Jackson Immunoresearch Laboratories, West Grove, PA). The coverslips were incubated with DAPI diluted to 1 g/ml for 10 min at room temperature. The coverslips were washed and then mounted using Aqua Poly/Mount (Polysciences Inc., Warrington, PA). The slides were viewed through the Nikon Eclipse E800 microscope, and the pictures were taken using the Nikon ACT-1 Software (Nikon Instruments Inc., Melville, NY). The images were processed using Windows Photo Gallery (Microsoft, Redmond, WA).
RNA Interference-A validated siRNA duplex corresponding to the coding region of HuR and a negative control duplex were purchased from Dharmacon Research (Thermoscientific, Lafayette, CO). siHuR contained the sequence 5Ј-GGUU-GAAUCUGCAAAGCUU-3Ј. For RNAi experiments, C5N and A5 cells were allowed to grow to 70% confluency and transfected with the siRNA (final concentration, 100 nM) specific for the HuR coding region or a negative control according to the manufacturer's instructions. Mock transfected cells were treated with the transfecting agent DharmaFECT 1 (Thermoscientific) only. Knockdown efficiency was assessed 72 h posttransfection through qRT-PCR, Northern blot, or Western blot analysis. For Northern blot analysis and qRT-PCR, RNA was harvested using the methods described above. For Western blot analysis, protein was harvested in 1ϫ radioimmune precipitation assay buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM Na 2 EDTA, 1 mM EGTA, 1% Nonidet P-40, 1% sodium deoxycholate, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 1 g/ml leupeptin, 1 mM PMSF, 1ϫ protease inhibitor mixture) (Cell Signaling Technology). ODC enzyme activity was tested in mock, negative control, or siHuRtreated cells as described above. ODC mRNA stability in mock, negative control, or siHuR-treated cells was also assessed by treating each sample with 10 g/ml of actinomycin D. RNA was harvested at 0, 4, and 8 h, and ODC half-life was determined as described above.
3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium Bromide Viability Assays-The cells were treated as described above for RNA interference experiments. At 72 h post-transfection, the medium was aspirated, and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide solution dissolved in PBS was added to each well (final concentration, 0.5 mg/ml) (Sigma). The plate was then incubated at 37°C and 5% CO 2 for 4 h. The solution was aspirated, and the cells were solubilized in 2 ml of Me 2 SO. Upon homogeneity, 100 l of each sample was assayed in triplicate in a 96-well plate. The absorbance was read at a 570-nm wavelength using a SPECTRAmax PLUS 384 plate reader (Molecular Devices Corporation, Sunnyvale, CA). Cell viability was calculated as a percentage of the optical density value of untreated cells.
Statistics-The results are expressed as the means Ϯ S.E. from three to nine samples. Autoradiography and immunoblotting were repeated up to three times. Statistical analysis was performed using Student's t test. p values of Ͻ0.05 were considered significant.

ODC Enzyme Activity and mRNA Stability Are Increased in
Transformed Compared with Nontransformed Keratinocytederived Cell Lines-ODC enzyme activity assays revealed a 4-fold change in ODC enzyme activity between the two cell lines (Fig. 1A), with C5N keratinocytes displaying low ODC enzyme activity, whereas the A5 cells exhibited a significant induction of ODC enzyme activity. The up-regulation of ODC enzyme activity was not due to an increase in transcription because qRT-PCR revealed no significant difference in steadystate mRNA levels between the two cell lines (Fig. 1B). Using both Northern blot analysis with subsequent band quantitation and qRT-PCR, we determined that ODC mRNA half-life was 4 h in the C5N keratinocytes compared with 15 h in A5 cells (Fig. 1C). These data show that ODC mRNA is markedly less stable in the normal keratinocytes than in carcinoma cells and that the observed change in ODC stability alone could account for the difference in the ODC enzyme activity between the two cell lines.
HuR Is Able to Bind to the ODC 3Ј-UTR in A5 Spindle Carcinoma Cells but Not C5N Keratinocytes-Previous results have shown that the intracellular polyamine pool can affect the localization of the stabilizing RBP HuR in IEC-6 cells (21,22). HuR has been shown to preferentially bind to ARE sequences that are located in the 3Ј-UTR of a target mRNA transcript (28,39). Upon examination of the mouse 3Ј-UTR for ODC, one classical AUUUA ARE was located ( Fig. 2A). A biotinylated pulldown assay was conducted to determine whether HuR could bind to synthetic transcripts corresponding to the full-length ODC 3Ј-UTR or the ODC coding region. The data show that cytoplasmic extracts of C5N cells demonstrate no measurable binding of HuR to either the ODC 3Ј-UTR or coding region. Conversely, cytoplasmic A5 extracts exhibit strong HuR binding to the ODC 3Ј-UTR biotinylated probe (Fig. 2, B and C). The difference in binding between C5N and A5 cells was not due to a lack of HuR in the C5N cells, because whole cell lysates revealed that the relative level of HuR protein was similar between the two cell lines (Fig. 2D). Together, these data show that HuR is present in both cell lines but that only HuR present in the A5 cytoplasm is able to bind to the 3Ј-UTR of ODC.
To determine the location of the HuR-binding site in the 3Ј-UTR of ODC, we created overlapping biotinylated probes of ϳ150 -200 bases in length that scanned the entire ODC 3Ј-UTR (Fig. 3A). A5 cytoplasmic lysates were used to perform these mapping experiments because we did not observe HuR binding from C5N lysates (Fig. 2C). The association of HuR protein with a synthetic ODC mRNA transcript was detected by Western blot analysis of the pulldown material (Fig. 3B). The ODC full-length 3Ј-UTR transcript associated with HuR. Interestingly, fragments 1 and 2 exhibited even stronger HuR binding than the full-length probe. Fragments 3 and 4 also displayed lower levels of HuR binding, whereas the most distal fragment, fragment 5, did not appear to associate with the HuR protein. The c-Myc full-length 3Ј-UTR positive control exhibited HuR binding, and the GAPDH negative control demonstrated no binding (40). These data verify that HuR is able to bind to the more proximal region of the ODC 3Ј-UTR and suggest the presence of multiple binding sites.
To confirm these results and to further narrow the region of HuR binding, a second mapping experiment was performed using additional deletion constructs of either the proximal or distal end of the ODC 3Ј-UTR. These deletion constructs ranged from 50 to 500 bases in length (Fig. 4A). The full-length ODC 3Ј-UTR, deletion 1, deletion 2, and deletion 4 constructs all exhibited HuR binding, whereas deletion constructs 3 and 5 demonstrated little to no HuR binding (Fig. 4B). Both sets of binding site data suggest that HuR binds to multiple sites on the 3Ј-UTR of ODC between bases 1851 and 2151 (Fig. 4C).
The endogenous intracellular association of ODC mRNA and HuR was assessed through immunoprecipitation of HuR in an assay that allows the association between HuR and its target  C5N and A5 cells were fractionated as described under "Experimental Procedures." To check the extracts for nuclear contamination, Western blots for ␣/␤-tubulin (cytoplasmic marker) and histone H3 (nuclear marker) were used. C, to assess HuR binding, cytoplasmic lysates (120 g) prepared from C5N and A5 cell lines were incubated with 15 l of synthetic biotin-labeled ODC 3Ј-UTR or ODC coding region (CR) transcript for 30 min at room temperature. The resulting ribonucleoprotein complexes were pulled down using streptavidin beads, and the presence of HuR was assessed via Western blot analysis. This experiment was done in duplicate with similar results. D, whole cell lysates were prepared in 1ϫ radioimmune precipitation assay buffer. HuR was detected via Western blot analysis. The blot was then stripped and reprobed for GAPDH to confirm equal loading of protein. This experiment was conducted three times with similar results. mRNAs in cytoplasmic ribonucleoprotein complexes to be maintained. In agreement with the biotinylated pulldown data, little or no ODC mRNA was immunoprecipitated from the C5N keratinocyte cell line cytoplasmic lysate, but a strong ODC band was detected in the immunoprecipitated material from A5 cells, as measured by both conventional PCR and qRT-PCR (Fig. 5). In fact, the qRT-PCR revealed a 9-fold increase in ODC mRNA in the A5 pulldown material when compared with the C5N. GAPDH was present at very low levels in both lysates. This is in agreement with previously reported results, because GAPDH can act as a partial contaminant in IP materials (41). The positive control c-Myc mRNA associated strongly with HuR protein in A5 lysates (Fig. 5A) (40).

HuR Cellular Localization Differs between Normal Mouse Keratinocytes and Skin
Tumor-derived Cells-The results above show strong binding of HuR to ODC mRNA in transformed A5 cells, but little or no association in C5N keratinocytes, despite having similar levels of HuR protein as seen in the whole cell lysates of each cell line (Fig. 2D). To address the issue of preferential binding of HuR to the A5 ODC 3Ј-UTR, the ODC 3Ј-UTR was sequenced in both cell lines. An alignment of the two sequences revealed a 100% sequence identity (data not shown). Given that the ODC 3Ј-UTR was not mutated in the A5 cells, another possible explanation is that the two cell lines display different cellular localizations of the HuR protein. HuR has been shown to shuttle between the nucleus and the cytoplasm in response to various stimuli or intrinsic cellular conditions. Additionally, cytoplasmic HuR in human tumors correlates with an increased tumor grade and a decreased survival rate (42,43). Based on this, we decided to investigate the cellular localization of the HuR protein in C5N and A5 cells.
The immunofluorescence results show that HuR is almost exclusively nuclear in C5N cells, whereas in A5 cells HuR is both nuclear and cytoplasmic (Fig. 6). These data affirm our previous findings in Figs. 2 and 5, because both the biotinylated pulldown assay and ribonucleoprotein complex assay use cytoplasmic extract. This difference in localization likely accounts for the inability of HuR to bind to ODC mRNA in C5N cytoplasmic extracts. In fact, when C5N nuclear extracts were used to investigate HuR binding to the ODC mRNA transcript, we show that HuR is able to bind to the ODC 3Ј-UTR (Fig. 7A). The C5N nuclear extract also displayed lower level binding to the 5Ј-UTR and coding region of the ODC mRNA transcript (Fig. 7B).
HuR Silencing Significantly Decreases Both ODC Enzyme Activity and the Stability of ODC mRNA in A5 Cells-To determine whether cytoplasmic HuR plays a causal role in increasing ODC enzyme activity and ODC mRNA stability in A5 cells, siRNA against HuR was utilized to reduce intracellular HuR levels. A transient transfection of siHuR caused a 90% knockdown of HuR message as assessed by qRT-PCR (Fig. 8A). Moreover, the knockdown efficiency of the siHuR was assessed through Western blot analysis and revealed a 70% decrease in HuR protein when compared with mock transfected and negative control siRNA-treated cells (Fig. 8, B and C). To rule out toxicity, the cell viability of A5 cells treated with siHuR and the negative control siRNA was assessed. No change in cell viability was measured in siHuR-treated cells when compared with either untreated or negative control siRNA-treated cells (Fig.  8D). These data affirm that siHuR treatment in A5 cells is not toxic.
To measure the level of ODC protein after the addition of siHuR to the A5 cells, we used an ODC enzyme activity assay, because ODC activity correlates with protein levels. On average, a 60% decline in ODC enzyme activity occurred with the addition of siHuR. Conversely, the control siRNA had no effect on the ODC enzyme activity in the A5 cells (Fig. 8E). In addition, the ODC mRNA stability decreased by ϳ60% when A5 cells were treated with siHuR compared with cells that had been treated with the negative control siRNA (Fig. 8F). These findings strongly suggest that HuR enhances the stability of the ODC mRNA transcript in mouse keratinocyte cells through direct binding to the ODC 3Ј-UTR and that this increased stability contributes to the observed induction of ODC enzyme activity in these cells.
To ensure that the knockdown of cytoplasmic HuR played a causal role in the decrease of both ODC enzyme activity and ODC mRNA stability in A5 cells, we assessed the effects of knocking down HuR in C5N cells. Because C5N cells express HuR almost exclusively in the nucleus (Fig. 6), we would not expect HuR knockdown in these cells to affect ODC mRNA stability. Treatment of C5N cells with siHuR resulted in a decrease in whole cell HuR levels (Fig. 9A). In fact, HuR protein levels decreased by more than 50% in siHuR-treated cells when compared with cells that had been treated with the negative  supplemental Table S1. The PCR product was then biotin-labeled as described under "Experimental Procedures." B, to determine the association of HuR, A5 cytoplasmic lysate (120 g) was incubated with 15 l of each biotin-labeled synthetic probe, and Western blot analysis for HuR was performed. The full-length 3Ј-UTR (FL) was used as a positive control, as was a c-Myc 3Ј-UTR. A GAPDH (Gpdh) biotinylated probe was used as a negative control. Whole cell lysate (WC) is also shown. This experiment was done in duplicate with similar results. F1-F5, fragments 1-5. control siRNA (Fig. 9B). However, the knockdown of HuR had no effect on the ODC mRNA stability in C5N cells. Both siHuR-treated and negative control siRNA-treated cells had the same ODC mRNA half-life of ϳ9 h (Fig. 9C). Correspondingly, no change in ODC enzyme activity was observed in siHuR-treated C5N cells when compared with the negative control siRNA-treated cells (data not shown).
The ODC mRNA half-life in C5N cells that had been treated with either the siHuR or the negative control siRNA was ϳ2-fold longer than untreated C5N cells. The presence of the lipid transfection reagent or the siRNA duplexes may explain this result. In fact, we also saw a stabilization of ODC mRNA in A5 cells treated with the same negative control siRNA (compare Fig. 1C with Fig. 8F). In any case, we show that ODC mRNA stability is unchanged in C5N cells treated with siHuR and that A5 cells treated with siHuR demonstrate a reduction in both ODC mRNA stability and ODC enzyme activity. These data support our hypothesis that the cytoplasmic accumulation of HuR in NMSC progression is at least in part responsible for the observed increases in ODC mRNA stability and ODC enzyme activity.

DISCUSSION
The data described here show for the first time that the rate-limiting enzyme in the polyamine biosynthetic pathway, ODC, is regulated by binding to the stabilizing RNAbinding protein HuR. These data are novel, because ODC mRNA has never before been shown to be regulated by such a cis-and trans-acting mechanism. Our results demonstrate that binding of the transacting factor HuR increases ODC enzyme activity and mRNA stability via cis-acting elements contained within the 3Ј-UTR of the ODC mRNA transcript (Figs. 1-4). Data from deletion constructs indicate that HuR may bind to multiple sites within the ODC 3Ј-UTR. Although the dogma is that HuR binds to classical AREs, the exact binding motif for HuR remains unclear. López de Silanes et al. (44) have shown that HuR binds to U-rich elements. Furthermore, Yeap et al. (45) demonstrated HuR binding to UC-rich motifs in the 3Ј-UTR of androgen receptor mRNA. The general area of strongest HuR binding on the ODC 3Ј-UTR, between bases 1851 and 2151, is also U-rich (Fig. 4C).
The results presented here show that HuR binds to the ODC 3Ј-UTR of A5 spindle carcinoma cells, but no association was demonstrated in C5N keratinocytes (Figs. 2C and 5). The binding status of HuR to the ODC 3Ј-UTR corresponds to the localization of the HuR protein, because HuR was cytoplasmic in A5 cells but was strictly nuclear in C5N cells (Fig. 6). This phenomenon is in accordance with previous results showing that the cytoplasmic localization of HuR stabilizes p53 and nucleophosmin (22). Additionally, in studies investigating the role of HuR in human tumor samples, a strong correlation was found between tumor grade and HuR cytoplasmic accumulation in serous ovarian carcinoma, Merkel cell carcinoma, and human invasive ductal breast carcinoma. Nonneoplastic tissue controls exhibited little cytoplasmic HuR presence (42,46,47). C5N nuclear extracts, which have an abundance of HuR protein, demonstrate binding to a synthetic ODC 3Ј-UTR transcript (Fig. 7). Given that the sequences of the ODC 3Ј-UTR are identical in both cell lines, these data suggest that HuR localization dictates whether the HuR protein binds to and stabilizes the ODC transcript. It has been found previously, using a rat intestinal epithelial cell model, that polyamine depletion can cause a decrease in AMPK activity and thereby cause the retrograde movement of HuR from the nucleus into the cytoplasm via importin ␣1 (21). Furthermore, the depletion of the polyamine intracellular pool through the inhibition of ODC was shown to cause an increase in HuR cytoplasmic presence and an increase in ATF-2 mRNA stability (20). In contrast, we do not see HuR shuttle to the nucleus in response to polyamine depletion in our NMSC model. In fact, we see the opposite in our cells, because A5 carcinoma cells, where HuR is present in the cytoplasm, contain higher putrescine levels when compared with nontransformed C5N keratinocytes (data not shown), which contain measurable HuR only in the nucleus. Although the reason for this difference is not known, it is likely that other factors in addition to the polyamines play a role in controlling HuR shuttling. For example, HuR can shuttle from the nucleus to the cytoplasm during cell cycle progression or after UVB exposure (31,48,49).
We note in A5 spindle carcinoma cells that both ODC enzyme activity and ODC mRNA half-life are up-regulated (Fig. 1, A and C). The increase in ODC mRNA stability in A5 cells is due, at least in part, to the association of cytoplasmic HuR with the ODC transcript. Given that a mutated HuR protein has never been reported in cancer, our results suggest the possibility that HuR is post-translationally modified in A5 cells, thereby causing its cytoplasmic accumulation, where it can FIGURE 5. Association of HuR with endogenous ODC mRNA in A5 cells but not in C5N cells. A, cytoplasmic lysates from C5N and A5 cells were used for IP with anti-HuR or nonspecific mouse IgG. RNA in the IP material was extracted using phenol chloroform extraction, used in a PCR reaction, and visualized on a 2% agarose gel. The levels of ODC mRNA and GAPDH mRNA were assessed by PCR using primers that were specific for their respective coding regions. The level of c-Myc bound to HuR was assessed as a positive control. This experiment was conducted in triplicate with reproducible results. B, the above experiment was repeated and quantitated using qRT-PCR. ODC mRNA values were normalized to cyclophilin A. The values are the means Ϯ S.E. of data from two separate experiments (n ϭ 6). *, p Ͻ 0.005. FIGURE 6. HuR is exclusively nuclear in C5N normal keratinocytes but is both nuclear and cytoplasmic in A5 spindle carcinoma cells. The cells were grown on glass coverslips as described under "Experimental Procedures." HuR protein was detected using anti-HuR antibody (1:500). The Cy-2 secondary antibody was used (1:200) (green fluorescence). The nucleus was stained using DAPI (1:1000) (blue fluorescence). The images were taken at 40ϫ magnification. Representative pictures are shown. This experiment was done in triplicate with reproducible results. FIGURE 7. Nuclear extracts from C5N keratinocytes display HuR binding to the ODC mRNA transcript. A, a biotinylated pulldown assay using C5N nuclear extracts was conducted as described under "Experimental Procedures." 120 g of pulldown material as well as 20 g of C5N whole cell lysate were loaded onto a 10% SDS-PAGE gel, and HuR binding was assessed via Western blot analysis. A representative immunoblot is shown. This experiment was done in triplicate with similar results. B, the immunoblot described in A was quantitated using Syngene software. All of the bands were normalized to a GAPDH control. The 5Ј-UTR band quantitation was set to 1, and the coding region and 3Ј-UTR band quantitation is reported as a percentage of the 5Ј-UTR band. WC, whole cell lysate; CR, coding region.
bind to a variety of target mRNAs, including ODC (50). We are currently investigating upstream targets that are activated in these cells and that may cause this localization change in HuR between our spindle carcinoma cells and normal keratinocytes. Several post-translational modifications to the HuR protein have been shown to influence the localization of HuR. For example, an S242A mutant of HuR accumulates in the cytoplasm and also increases its ability to bind to its targets cyclin A2 and B1 in HeLa cells (51).
It is well established that NMSC carries both a health and fiscal burden, costing approximately $1.5 billion to treat in the United States alone in 2004 (52). ODC has been shown to be both necessary and sufficient for the onset of NMSC in mice and is also elevated in human NMSC (12,14,(53)(54)(55). Thus, the link between ODC and NMSC is indisputable, validating ODC as both a chemoprevention and chemotherapeutic target. In fact, a recent phase III chemoprevention trial demonstrated that ␣-difluoromethylornithine, a suicide inactivator of ODC, markedly reduces the incidence of new basal cell carcinoma in subjects with a history of skin cancer (18). Earlier studies showed that topical ␣-difluoromethylornithine administration reduced the incidence of actinic keratosis, a precursor of squamous cell carcinoma, in subjects at high risk for NMSC (56). The studies provided here shed more light on the mechanism of ODC regulation in the skin and can provide an additional target for the development of chemopreventive and chemotherapeutic agents that alter ODC activity.
In summary, we describe a cisand trans-acting mechanism regulating ODC mRNA stability and subsequently ODC enzyme activity in the skin. We demonstrate, using a murine model of NMSC, that ODC activity is increased and mRNA is stabilized in carcinomas and that stabilization of the ODC transcript is attributed to the binding of HuR  (n ϭ 3). B, Western blot analysis to detect HuR protein knockdown. A5 cells were grown and transfected with siRNA duplexes specific for the HuR coding region or a negative control siRNA (denoted as siHuR and Ctrl siRNA, respectively). Mock transfected cells (Mock) and untreated A5 cells (WC) are also shown. Protein was harvested after 72 h in 1ϫ radioimmune precipitation assay buffer. The level of HuR was measured by Western blot analysis. Equal loading was monitored by GAPDH immunoblotting. C, quantitation of the Western blot analysis from B. All of the HuR bands were normalized to GAPDH. A representative blot and band quantitation are shown. D, cell viability does not change in A5 cells treated with siHuR. Cell viability was measured in untreated A5 cells, mock transfected cells (DharmaFECT1 only), siHuR-treated cells, and negative control siRNA-treated cells (Ctrl siRNA) using methods described under "Experimental Procedures." The optical density measurements for the untreated A5 cells were set to 1. The mock, siHuR, and negative control siRNA-treated cells are reported as mean percentages of the untreated cell optical density readings Ϯ S.E. (n ϭ 9). E, ODC enzyme activity decreases with HuR knockdown in A5 cells. At 72 h post-siHuR transfection, the cells were harvested in ODC harvest buffer, and ODC enzyme activity was assayed as described under "Experimental Procedures. to the ODC 3Ј-UTR. HuR is localized to the cytoplasm in spindle carcinoma cells, whereas in normal keratinocytes it is exclusively nuclear. This alteration in the localization of the HuR protein accounts for the discrepancy in HuR binding status to the ODC 3Ј-UTR between the two cell lines. Furthermore, we provide a functional link between HuR and ODC using RNA interference to knockdown HuR, resulting in the destabilization of the ODC message and a decrease in ODC activity in carcinoma cells but not in normal keratinocytes. These findings not only demonstrate a novel mechanism of ODC regulation but also provide another potential target for chemoprevention and chemotherapeutics for NMSC.