Complex contribution of the 3'-untranslated region to the expressional regulation of the human inducible nitric-oxide synthase gene. Involvement of the RNA-binding protein HuR.

Cytokine stimulation of human DLD-1 cells resulted in a marked expression of nitric-oxide synthase (NOS) II mRNA and protein accompanied by only a moderate increase in transcriptional activity. Also, there was a basal transcription of the NOS II gene, which did not result in measurable NOS II expression. The 3'-untranslated region (3'-UTR) of the NOS II mRNA contains four AUUUA motifs and one AUUUUA motif, known to destabilize the mRNAs of proto-oncogenes, nuclear transcription factors, and cytokines. Luciferase reporter gene constructs containing the NOS II 3'-UTR showed a significantly reduced luciferase activity. The embryonic lethal abnormal vision (ELAV)-like protein HuR was found to bind with high affinity to the adenylate/uridylate-rich elements of the NOS II 3'-UTR. Inhibition of HuR with antisense constructs reduced the cytokine-induced NOS II mRNA, whereas overexpression of HuR potentiated the cytokine-induced NOS II expression. This provides evidence that NOS II expression is regulated at the transcriptional and post-transcriptional level. Binding of HuR to the 3'-UTR of the NOS II mRNA seems to play an essential role in the stabilization of this mRNA.

constitutively, but their expression level can be adjusted to meet local demand (2). In marked contrast, the high output NOS II enzyme is not normally present in resting cells. Instead, the cell must be activated to express the enzyme, whose activity is then independent of elevated Ca 2ϩ concentrations. NOS II expression can be induced in different types of cells and tissues following exposure to immunologic and inflammatory stimuli such as cytokines or lipopolysaccharide. Once synthesized, the enzyme is active for hours to days and generates large amounts of nitric oxide that can have beneficial effects, such as antimicrobial, antiatherogenic, or antiapoptotic actions. However, inappropriate NOS II induction can have detrimental consequences, such as cellular injury in arthritis or colitis, and in septic shock (3)(4)(5)(6).
Traditionally, it has been believed that the expression of the NOS II is regulated mainly at the transcriptional level. The human NOS II promoter contains potential binding sites for a number of transcription factors, such as signal transfer and activator of transcription-1, activator protein-1, (jun/fos transcription factor), and nuclear factor-B, which is known to participate in the induction of NOS II expression by cytokines. Interaction of these transcription factors with their cis-elements activates (or represses) the transcriptional machinery leading to mRNA synthesis (7)(8)(9)(10)(11)(12).
However, the steady-state levels of a particular mRNA depend not only on its synthesis but also on its rate of degradation. Several lines of evidence suggest that regulation of mRNA stability may contribute to NOS II expression. Using RNase protection assays or Northern blot experiments, different laboratories including our own were unable to detect NOS II mRNA in noninduced cells. However, the gene is not totally silent, and significant basal transcription has been detected. After cytokine induction, the transcription rate is only increased 2-5-fold compared with the estimated 20 -100-fold increase in the level of NOS II mRNA. These findings point to an instability of NOS II transcripts in resting cells (8,9,12).
Adenylate/uridylate-rich elements (AU-rich elements) are found in the 3Ј-untranslated region (3Ј-UTR) of many mRNAs encoding proto-oncogenes, nuclear transcription factors, and cytokines (13)(14)(15). Using RNA binding assays, several groups have identified proteins that interact with AU-rich elements (ELAV-like family, AUF1, heat shock proteins, among others) (16 -18). Many of these proteins have been implicated in the regulation of mRNA stability. The 3Ј-UTR of the human NOS II mRNA contains five AUUUA and one AUUUUA repeat, classical motifs found in AU-rich elements of different mRNAs, which have a short and fine regulated half-life. So far, little is known about the role of these motifs in the regulation of NOS II expression. Recently, Nunokawa et al. (19) designed a reporter gene construct containing not only the 5Ј-flanking region (Ϫ1 kb) of the human NOS II gene but also the 3Ј-UTR and ϳ1 kb of the 3Ј-flanking sequence. This construct showed a reduced basal reporter gene activity and a higher apparent rate of induction in response to cytokines. Although these authors proposed the existence of a cooperative interaction between the 5Ј-promoter and the 3Ј-region for the induction of the gene, the mechanism by which the 3Ј-UTR may regulate the expression has not yet been investigated (19).
Here we describe that post-transcriptional mechanisms account for an important part of the cytokine-induced expression of the human NOS II in intestinal epithelial DLD-1 cells, and the 3Ј-UTR of the gene seems to be essential for this regulation. In search for potential intracellular proteins able to interact with the NOS II 3Ј-UTR, we found that the ELAV-like RNAbinding protein HuR binds with high affinity and specificity to AU-rich elements present in the NOS II 3Ј-UTR. We also present evidence that the HuR protein is involved in the posttranscriptional regulation of the expression of human NOS II.
Cell Culture, Cytokine Treatment, and RNA Isolation-Human DLD-1 cells (ATCC, no. CCL-221; stable colon adenocarcinoma cells) were grown in Dulbecco's modified Eagle's medium with 10% inactivated fetal bovine serum, 2 mM L-glutamine, penicillin, and streptomycin. Eighteen hours before cytokine induction, the cells were washed with phosphate-buffered saline solution and incubated with Dulbecco's modified Eagle's medium containing 2 mM L-glutamine in the absence of serum and phenol red. DLD-1 cells were induced with the cytokines interferon-␥ (100 units/ml), interleukin 1-␤ (50 units/ml), and TNF-␣ (10 ng/ml) for the corresponding time periods depending on the experiment. Afterward, a supernatant of the cells (300 l) was used to measure N O 2 . by the Griess reaction and cells were processed for RNA isolation by guanidinium thiocyanate/phenol/chloroform extraction as described (12,20) or for protein extraction as described below.
RNase Protection Analyses-To generate a luciferase cDNA plasmid for in vitro transcription, a 230-bp HinfI fragment of pGL2-Basic (positions 306 -535) and pGL3-Basic (positions 318 -547) were cloned after blunting into the EcoRV site of pCR-Script, generating pCR-Luc-pGL2 and pCR-Luc-pGL3, respectively. A plasmid containing a human HuR cDNA fragment was generated from pGEX-HuR (21) by cloning a 207-bp BamHI-EcoRI fragment (positions 892-1096 of the HuR mRNA) into pCR-Script, generating pCR-HuR. DNA sequences of the clones were determined using the dideoxy chain termination method with a sequencing kit from Amersham Pharmacia Biotech.
Western Blot Experiments-To study the expression of NOS II and HuR proteins, total cell protein was fractionated in nuclear and cytoplasmic extracts as described (23). For NOS II or HuR Western blots, 10 -50 g of cytoplasmic or nuclear protein were separated on 7.5 and 12% polyacrylamide gels, respectively, and transferred to nitrocellulose membranes by semidry electroblotting. All further steps were performed as described (24). For the detection of human NOS II an anti-NOS II antibody (N-20) at 1:200, dilution was used. HuR protein was detected using a monoclonal anti-HuR-antibody at a concentration of 1 g/ml. For detection of the immunocomplexes a 1:10,000 dilution of phosphatase alkaline-coupled goat anti-rabbit or anti-mouse IgG was used and developed with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium substrate.
Nuclear Run-on Assay and Hybridization of de novo Radiolabeled RNA-Nuclear run-on assays were performed as described (23). The radiolabeled RNA was hybridized at 65°C for 48 h to the linearized pCR-Script (as negative control) or plasmids containing either the whole human NOS II cDNA (kindly provided by Dr. David Geller (25) or the whole human ␤-actin cDNA, immobilized on nitrocellulose filters as described (26). The reaction was carried out in 6ϫ SSC (0.9 M NaCl; 0.09 M sodium citrate, pH 7.0), 5ϫ Denhardt's reagent (0.1 g of Ficoll 400; 0.1 g of polyvinylpyrrolidone; 0.1 g of bovine serum albumin, Fraction V, in 100 ml of H 2 O), and 0.1% SDS. After hybridization, the filters were washed twice with 2ϫ SSC and 0.1% SDS at room temperature for 30 min, followed by two washes with 0.5ϫ SSC and 0.1% SDS at 65°C for 1 h. Filters were air-dried and exposed to x-ray films. Densitometric analyses were performed using a PhosphoImager (Bio-Rad).
Construction of Reporter Plasmids, Transient and Stable Transfection of DLD-1 Cells-The plasmid pGL3-3.6-NOS-II-luc-neo was generated by cloning a 3.6-kb KpnI fragment from pNOS2(16)Luc (8) into pGL3-neo (27) digested with KpnI. A NOS II cDNA fragment containing only the 3Ј-UTR sequence of the human NOS II gene was generated by PCR from reverse transcriptase products using the primers CCAAGCT-TGAGGGCCTACAGGAGGGG (5P1, sense) and CCAAGCTTGATTA-AAGTAAAATGCAATTCATG (3P1, antisense), corresponding to the positions 3667-3684 and 4122-4145 of the human NOS II mRNA (25) (positions 1-16 and 454 -477 of the NOS II 3Ј-UTR, see Fig. 3) flanked by HindIII restriction sites. The PCR product was digested with Hin-dIII and cloned into the HindIII site of pCR-Script to give pCR-NOS II-3Ј-UTR. The insert of this plasmid was further cut with HindIII, Klenow-filled, and then recloned in both orientations into the XbaI site of the plasmid pGL3-3.6-NOS-II-luc-neo to give pGL3-3.6-NOS-II-lucneo-3Ј-UTR A and B. Also, DNA fragments corresponding to the NOS II coding sequence (345 bp, positions 3127-3471) and the pCR-Script sequence (459-bp HinfI fragment, positions 177-636) were cloned into the XbaI site of pGL3-3.6-NOS-II-luc-neo.
DLD-1 cells were transfected by lipofection with DOTAP according to the manufacturer's recommendations. For transient experiments, 4.5 g of the plasmid were combined with 0.5 g of the renilla reporter gene plasmid pRL-SV40 for normalization of the transfection efficiency. After overnight transfection incubation, cells processed for luciferase activity were lysed in 1ϫ passive lysis buffer provided by the dual luciferase reporter assay system, and firefly and renilla luciferase activities were determined in 40-and 20-l extracts, respectively. The light units of the firefly luciferase were normalized by those of the renilla luciferase after subtraction of extract background. Cells used for quantification of luciferase mRNA levels by RNase protection assays were lysed with guanidium thiocyanate buffer as described (12).
To generate DLD-1 cells stably transfected with a construct containing a 16-kb fragment of the human NOS II promoter cloned in front of a luciferase reporter gene (pGL2 system), cells were transfected as described above with 4.5 g of pNOS2(16)Luc (8) and 0.5 g of pRc-CMV (containing a neomycin resistance gene). The transfected cells were selected by G 418 treatment (1 mg/ml). Different cell clones were analyzed for luciferase activity and checked for integration of the transfected DNA by PCR.
Purification of GST-HuR Protein-Purified GST-HuR fusion protein was prepared using the plasmid pGEX-HuR as described (21). The yield of the purification procedure was determined by comparison to a bovine serum albumin standard on Coomassie Blue-stained SDS-polyacrylamide gel electrophoresis. The electrophoresis revealed a 62-kDa band corresponding to the fusion protein. The same procedure was used to purify GST protein from Escherichia coli cultures transformed with the plasmid pGEX-2T.
To generate radiolabeled NOS II and c-fos 3Ј-UTR sense probes for RNA binding experiments, 0.5-1 g of DNA (linearized plasmids or PCR fragments) were in vitro transcribed as described above. Radiolabeled transcripts were analyzed by urea-denaturing electrophoresis to estimate the yield and the specific activity. Incorporated radioactivity in transcripts was usually higher than 80%, and the specific activity ranged from 0.2 to 0.5 Ci/pmol. The AUFL c-fos transcript for competition experiments was synthesized following the same procedure but using cold UTP with trace amounts of radioactivity to accurately determine molarity.
RNA binding experiments were carried out as described (21) using 5-8 fmol of labeled RNA (about 60,000 cpm) and protein as indicated. RNA-protein complexes were analyzed on 1% agarose gels in buffer (40 mM Tris acetate, 1 mM EDTA). After electrophoresis the gels were dried and exposed to x-ray films. In competition experiments, GST-HuR was preincubated with the RNA competitor for 15 min at room temperature.
Northwestern experiments were carried out as described (29). Briefly, 25 g of recombinant HuR or GST were fractionated by 12% polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes by semidry blotting. The membranes were then incubated at 4°C for 1 h with renaturation buffer (10 mM Tris, pH 7.5, 1 mM EDTA, 50 mM NaCl, 0.1% Triton X-100, 0.25 mg/ml bovine serum albumin, and 0.25 mg/ml tRNA). Subsequently, the blotted proteins were probed overnight at 4°C with 3Ј-UTR, non-AU, AU, subfragment A, B, or C (150,000 -200,000 cpm/ml) in the same buffer with occasional shaking. Unbound RNA was removed by three cycles of washing in TBS buffer (5 min/wash), and the membrane was then exposed to x-ray film and quantified by densitometry using a PhosphoImager.
In Northwestern experiments with cell extracts, 75-100 g of protein (nuclear and cytoplasmic extracts) diluted in 200 l final volume with immunoprecipitation buffer (10 mM Tris, pH 7, 4, 0.1% Triton X-100, 0.5% Nonidet P40, 150 mM NaCl) were precleared by adding 12 l of goat anti-mouse IgG bound to agarose (1:18, 1-h incubation, 4°C with shaking), and then centrifuged at 200 ϫ g for 2 min. Supernatants were incubated for 1 h with anti-HuR antibody (1:200) at 4°C and then with 12 l of anti-mouse IgG-agarose (1:18, 1-h incubation, 4°C with shaking). Agarose beads were washed five times with 1 ml of immunoprecipitation buffer and then resuspended in 10 l of loading buffer. The immunoprecipitates were analyzed for RNA binding as described above.
Establishment of Cell Lines Overexpressing Antisense or Sense HuR mRNA-A 1.6-kb ApaI fragment of HuR9 (21, 30) (containing the entire coding sequence of HuR) was cloned in either the sense or antisense orientation in pZeoSV2(Ϫ) and introduced by lipofection into DLD-1 cells. Stable transfectants were selected with zeocin (200 g/ml). As a control, DLD-1 cells stably transformed with the pZeoSV2(Ϫ) vector were also generated (Zeo). Pooled population of cells and single clones were characterized by HuR protein expression and DNA analyses.

NOS II Induction by Cytokines in
Human DLD-1 Cells-As a human model for studying the regulation of the NOS II gene, we chose the intestinal epithelial cell line DLD-1 (9,12,(31)(32)(33). As shown in Fig. 1, no NOS II mRNA or protein was detected in unstimulated DLD-1 cells, and virtually no nitrite production was seen with the Griess reaction (B). A triple cytokine mixture (CM) consisting of interferon-␥ (100 units/ml), interleukin 1-␤ (50 units/ml), and TNF-␣ (10 ng/ml) produced a marked NOS II expression. Using quantitative RNase protection assays with a NOS II 3Ј-UTR probe (see "Experimental Procedures"), significant levels of NOS II mRNA were already detected after 2 h of incubation with CM. Maximal induction was reached between 6 and 12 h (Fig. 1, A and B, open circles). Western blot studies (Fig. 1C) showed NOS II protein after 4 h of incubation with CM peaking at 12 h. This time course was also reflected in the accumulation of nitrite as detected by the Griess reaction (Fig.  1B, closed circles).
Transcriptional and Post-transcriptional Components of the NOS II Induction, Role of the 3Ј-UTR Region-To investigate whether the induction of the NOS II gene is mainly because of transcriptional activation or, on the other hand, also mechanisms at the post-transcriptional level take place, we examined the transcription rate of the gene in a nuclear run-on assay. A significant basal NOS II mRNA synthesis was observed in nuclei of unstimulated cells. Incubation with CM for 6 h resulted only in a modest 1.5-fold up-regulation of NOS II promoter activity (CM, 147.2 Ϯ 14.6; basal, 100.8 Ϯ 4.7, percent-age of basal as mean Ϯ S.D., n ϭ 6, p Ͻ 0.01), which is in agreement with previously published data (9). We have also established transfectants of DLD-1 cells expressing a luciferase reporter under the control of a 16-kb fragment of the human NOS II promoter (8). As shown in Fig. 2, these cells produced significant levels of luciferase mRNA (open circles) under nonstimulated conditions, whereas no NOS II mRNA (closed circles) was detected. After incubation with cytokines, both luciferase mRNA and endogenous NOS II mRNA were upregulated. Interestingly, different time courses were observed for the two mRNAs. Luciferase mRNA levels increased quickly after exposure of the cells to CM, peaking at 4 h. In contrast, NOS II mRNA levels increased more slowly reaching their maximum level at 8 h. The delay in the induction of the NOS II mRNA relative to the luciferase mRNA (both driven by the same promoter) cannot be explained by a longer processing time of the NOS II transcript (luciferase cDNA versus NOS II genomic sequence). We found the same time course with a NOS II 5Ј-UTR probe (data not shown). Taken together, these findings suggest the involvement of post-transcriptional mechanisms in the regulation of NOS II expression in human cells.
Because many post-transcriptional mechanisms (such as regulation of mRNA stability) are determined by cis-acting elements in the 3Ј-UTR, we focused on the 3Ј-UTR of the human NOS II gene. Fig. 3 depicts the 3Ј-UTR of the human NOS II mRNA. As potential regulatory elements, this sequence contains one AUUUUA and four AUUUA motifs, which are well conserved between human, dog, rat, and mouse. Similar AU-rich regions are present in the 3Ј-UTRs of cytokine and proto-oncogene mRNAs and have been shown to mediate a rapid mRNA turnover. We performed transient transfection studies with a luciferase reporter gene whose expression was driven by a 3.6-kb fragment of the human NOS II promoter (Fig. 4A). This plasmid construct is particularly interesting because it displays a significant basal activity but is not induced by cytokines (8,9,11,33). With this approach, the contribution of the 3Ј-UTR region can be analyzed without the influence of an inducible promoter. When the whole NOS II 3Ј-UTR (orientation A) was cloned downstream of the luciferase reporter gene, basal luciferase activity was markedly reduced (Fig. 4B). This reduction was not observed when the same nucleotide sequence was cloned in the opposite orientation (orientation B) or when fragments of similar size from the NOS II coding region (CDS) or from a vector sequence (vec) were used. Subsequently the 3Ј-UTR was dissected into two subfragments, one without AU repeats (non-AU) and the other containing these regulatory elements (AU, Fig. 3). Unlike the whole 3Ј-UTR, neither of these fragments reduced luciferase activity (Fig. 4B). After CM induction, none of above constructs showed an enhanced luciferase activity compared with basal conditions. As shown in Fig. 4C, DLD-1 cells transiently transfected with the 3Ј-UTR construct (in orientation A) also expressed significantly lower levels of luciferase mRNA when compared with cells transfected with control plasmids or plas-mids containing the 3Ј-UTR in orientation B. Therefore, mRNA destabilization by the 3Ј-UTR fragment (orientation A) accounts for the reduced luciferase activity.
Proteins Interacting with the 3Ј-UTR of Human NOS II mRNA, the ELAV-like Protein HuR-The stability of many lymphokine, cytokine, and proto-oncogene mRNAs is modulated through specific binding of proteins to the AU-rich regulatory elements present in their 3Ј-UTRs (13)(14)(15). Recently, HuR, a 36-kDa RNA-binding protein, has been cloned, which binds to AU-rich elements with high affinity and selectivity. Through this interaction, HuR modulates mRNA turnover (21,30). To investigate whether HuR can bind to putative regulatory elements in the NOS II 3Ј-UTR, purified recombinant GST-HuR fusion protein was incubated with labeled transcripts comprising different nucleotide regions (see Fig. 3) and HuR-RNA complex formation was assayed by RNA gel shift. In preliminary experiments, we detected complex formation between recombinant GST-HuR protein and the whole 3Ј-UTR transcript, and this binding activity was not observed with the same concentration of GST (n ϭ 5, data not shown). To localize the HuR binding site within the NOS II 3Ј-UTR, the region was first dissected into two subfragments, one without AU repeats (non-AU) and the other containing these putative regulatory elements (AU). Only the subfragment containing the AU repeats interacted with the HuR protein (Fig. 5B). Subsequently, the AU fragment was dissected into three subfragments: subfragment A (232-329, with no AU repeats, but relatively well conserved among species), subfragment B (327-428, with three AUUUA motifs), and subfragment C (387-477, with one AUUUA and the AUUUUA element). As shown in Fig. 6, HuR binding was only found for subfragment C. Because subfrag- ment B did not show any significant binding activity and subfragments B and C share a short nucleotide sequence (387-428), the HuR binding site is likely to be located between positions 439 -477, with one or both of the AU repeats being involved in the complex formation. As shown in Fig. 7, binding activity of a transcript comprising positions 443-477 (Fig. 7B, wild type) was not affected when the AU-rich element at positions 449 -453 was mutated to a GC-rich element (Fig. 7C, mut  1). In contrast, mutation of the second AU element between positions 464 and 469 (Fig. 7D, mut 2) reduced complex formation considerably. The double mutant (Fig. 7E, mut 1/2) showed virtually no binding activity.
The specific nature of HuR binding was further confirmed by RNA gel shift experiments performed with the well characterized HuR binding site of the AUFL element of the c-fos 3Ј-UTR (21,34). Fig. 8A shows that our preparation of HuR recombinant protein was able to bind radiolabeled AUFL transcript in the same concentration range determined for the interaction with the NOS II 3Ј-UTR mRNA. Furthermore, unlabeled c-fos AUFL transcript in 5-or 10-molar excess reduced significantly the complex formation between HuR and labeled NOS II 3Ј-UTR or subfragments AU or C, indicating that both RNAs compete for the same binding site on HuR protein (Fig. 8B). Because it has been described that the affinity of HuR for the AUFL element of the c-fos 3Ј-UTR is in the nanomolar range, we can assume that the NOS II 3Ј-UTR interaction affinity falls also in the same concentration range (21).
HuR Expression in DLD-1 Cells-Next we studied the ex-pression of HuR in DLD-1 cells. As shown in Fig. 9A, DLD-1 cells produced high levels of HuR mRNA under basal conditions, which were down-regulated by cytokine treatment. Fig.  9B shows a Western blot performed with a HuR-specific antibody (19F12). DLD-1 cells were found to express constitutively HuR protein, which was localized almost exclusively in the nuclear fraction with a minor fraction present in the cytoplasm (Ͻ5%). Cytokine treatment induced a down-regulation of HuR protein after 12-24 h of incubation in both nuclear and cytoplasmic fractions.
Northwestern experiments were performed to analyze potential interactions of recombinant and endogenous HuR and the NOS II 3Ј-UTR. As shown in Fig. 10A, the NOS II 3Ј-UTR probe bound a 62-kDa protein band corresponding to the HuR-GST fusion protein but did not bind GST alone. The same results were obtained with radiolabeled probes of subfragment AU and subfragment C (data not shown). Fig. 10B demonstrates that endogenous, immunoprecipitated HuR also bound the NOS II 3Ј-UTR. A band corresponding to the molecular mass of HuR (about 36 kDa) was detected by the NOS II 3Ј-UTR probe. The same results were obtained with radiolabeled probes of fragment AU and fragment C (data not shown). Northwestern experiments performed with nuclear and cytoplasmic fractions from cells incubated under basal or CM conditions showed no measurable difference in the binding capacity of endogenous HuR (data not shown).
Biological Role of HuR, Overexpression of Sense or Antisense HuR mRNA-To determine if HuR plays a role in the regulation of NOS II gene expression, we generated stable transformants of DLD-1 cells that expressed constitutively sense HuR mRNA or antisense HuR mRNA under the control of a cyto- megalovirus promoter (pZeoSV2(Ϫ) expression plasmid). Pooled cell populations and single clones were analyzed for stable integration of the DNA and HuR expression. As a control, DLD-1 cells stably transfected with the pZeoSV2(Ϫ) vector were also generated (Zeo). Fig. 11 shows the HuR protein expression (A) and NOS II 3Ј-UTR binding activity (B) of such pooled cell populations as analyzed by Western and Northwestern blotting. HuR expression and RNA binding activity in sense HuR-cDNA-transfected cells increased to ϳ150% compared with control cells (transfected with the vector backbone; Zeo). Overexpression of antisense HuR-cDNA resulted in reduced HuR protein levels and in reduced RNA binding activity. NOS II induction correlated well with the levels of HuR expression. NOS II mRNA (Fig. 12A), protein (Fig. 12B), and activity (Fig. 12C) were 30 -40% reduced in cells transfected with HuR antisense cDNA. NOS II expression was increased by 30 -50% in cells transfected with HuR sense cDNA. Similar results were obtained with single clones (data not shown).
Cells stably transfected with HuR sense or antisense cDNA (see above) were transiently transfected with a 16-kb human NOS II promoter luciferase reporter gene construct. As shown in Fig. 12D, neither the basal reporter activity nor the inducibility of the promoter was modified by increased or decreased of HuR expression.

DISCUSSION
The data presented here demonstrate that post-transcriptional mechanisms account for an important portion of the cytokine-induced NOS II expression in human DLD-1 cells, and the 3Ј-UTR of the gene contributes in a complex manner to this post-transcriptional regulation. Cytokine incubation increases NOS II promoter activity only modestly (1.5-5-fold). In contrast, steady state NOS II mRNA levels increase 100 -200-fold. On the other hand, we and others have found a significant transcription of the NOS II gene under resting conditions, although this constitutive promoter activity does not result in an appreciable accumulation of mRNA (9,12). This seems to be a general property of the NOS II gene because similar results were obtained with other human cell types such as AKN liver cells (8) or A549 epithelial cells. 2 The above observations point to an efficient degradation of the transcript in resting cells, which is slowed down significantly upon cytokine stimulation. This mechanism, in conjunction with the moderate activation of the NOS II promoter results in the marked up-regulation of 2  the NOS II mRNA. This cooperative mechanism has also been observed with mRNAs encoding for cell cycle control factors, oncogenes, cytokines, etc. All of these have short half-lives that can be modulated by various stimuli (13)(14)(15)35).
Several agents have been shown to modulate NOS II mRNA stability in various cell types. For example, increased Ca 2ϩ concentrations reduce NOS II mRNA half-life in human articular chondrocytes, and 6(R)-5,6,7,8-tetrahydrobiopterin, an allosteric cofactor of the NOS II enzyme, stabilizes the transcript in rat vascular smooth muscle cells (36,37).
Oftentimes, mRNA decay signals are located in the 3Ј-UTR of the mRNAs, suggesting that the half-lives of most mRNAs are influenced by this region (13)(14)(15). This is also the case for the NOS II mRNA. Based on the current results, the luciferase mRNA and activity can be reduced dramatically when the NOS II 3Ј-UTR is placed downstream of the reporter gene. A common feature of different short lived mRNAs is the presence of AU-rich elements in their 3Ј-UTRs. AU-rich elements generally exhibit several copies of the AUUUA pentanucleotide or similar sequences and are considered as cis-regulatory determinants of mRNA turnover (38,39). Surprisingly, we found that the destabilizing effect of the NOS II 3Ј-UTR was not conveyed by the AU-rich elements alone but required the full 3Ј-UTR. Thus, the AU-rich elements are necessary but not sufficient for mRNA destabilization. Similar results have been reported for the 3Ј-UTR of plasminogen activator inhibitor type 2 mRNA. Mutation of a nonameric UUAUUUAUU element does not fully abrogate the destabilizing effect observed with the full 3Ј-UTR sequence (40). Also for human cyclooxygenase-2 mRNA, the destabilizing effect of the 3Ј-UTR cannot be explained by the AUUUA elements alone (41).
Numerous proteins have been implicated in mRNA turnover by virtue of their ability to bind regulatory elements in the 3Ј-UTR of specific mRNAs. Among them, particular interest was focused on factors interacting with AU-rich elements. Different proteins binding to these elements have been characterized using gel retardation, cross-link assays, and in vitro decay systems. Some of these proteins, such as AUF1, destabilize mRNAs. Others, for example the ELAV-like family of proteins, seem to reduce mRNA degradation (18,42,43). HuR is an ELAV-like RNA-binding protein containing three RNA recognition motifs. Unlike other members of the ELAV-like family (HuD, HuC and HuB, or Hel-N1), which are expressed exclusively in brain, HuR is expressed in a wide variety of human tissues. HuR exhibits a high binding affinity for AU-rich elements present e.g. in c-fos, interleukin-3, c-myc, TNF-␣, plasminogen activator inhibitor type 2, or vascular endothelial growth factor mRNAs (16,21,30,40,(43)(44)(45). Therefore, we sought to determine whether HuR may bind to the 3Ј-UTR of the human NOS II mRNA and through this binding modulate NOS II expression. Our in vitro results demonstrate that HuR binds with high affinity and specificity to an AU-rich element located at the 3Ј-end of the NOS II 3Ј-UTR. Interestingly, the NOS II 3Ј-UTR contains three other AUUUA repeats that did not seem to bind HuR protein. Thus, our results confirm previous observations suggesting that HuR protein does not recognize primary sequences but rather higher order structures (21). The biological significance of the HuR binding to the NOS II 3Ј-UTR was demonstrated using stably transfected DLD-1 cells overexpressing HuR sense-or antisense mRNA. Inhibition of HuR expression by antisense mRNA constructs reduced the cytokine-induced NOS II expression, whereas cells transfected with HuR sense mRNA constructs showed an increased NOS II expression. These effects were not because of a modification of NOS II promoter activity. These results suggest that HuR participates in the cytokine-induced up-regulation of the gene and that this mechanism takes place at a post-transcriptional level.
Interestingly, after longer incubations (12-24 h) of normal DLD-1 cells with cytokines, both HuR mRNA and protein expression were down-regulated. This may contribute to the termination of NOS II expression.
So far little is known about the mechanism by which HuR stabilizes AU-containing mRNAs. The protein contains a nucleocytoplasmic shuttling domain, which has been proposed to regulate the stimulus-induced nucleus to cytoplasm translocation, leading to mRNA stabilization in the cytoplasm (46,47).
In our NOS II model the HuR binding site was mapped to specific AU-rich elements present in the 3Ј-UTR. Nevertheless, the single presence of this sequence in a luciferase reporter construct did not confer inducibility to its expression, which suggest that HuR or associating proteins may require additional sequences that were not present in this construct. Similar data have been reported for the regulation of the TNF-␣ gene. Expression of this gene is also regulated in a very complex way by transcriptional and post-transcriptional mechanisms. TNF-␣ promoter responses are only weakly induced by lipopolysaccharide, and even in nonstimulated cells, a significant basal promoter activity was detected. An AU-rich element present in the 3Ј-UTR of the TNF-␣ mRNA confers translational repression and instability to the transcript, and this intrinsic instability of the 3Ј-UTR is not modified after stimulation of cells. On the other hand, the AU-rich sequence in the 3Ј-UTR of TNF-␣ mRNA is a well characterized binding site for ELAV proteins, such as HuR. The exact mechanism through which these regulatory elements participate in TNF-␣ expressional regulation is no yet clear, but it has been described that the presence of both promoter sequences and the 3Ј-UTR are required to confer inducibility to the reporter gene (48 -50). Whether this is also the case for the human NOS II gene and which sequences might be involved is subject to further investigation.
In conclusion, experiments presented here demonstrate that the NOS II expression is regulated at both transcriptional and post-transcriptional levels in human DLD-1 cells. The 3Ј-UTR of the NOS II mRNA seems to play an essential role in the post-transcriptional regulation. On the one hand it permanently destabilizes the NOS II mRNA, and on the other hand it provides a specific binding site for the protein HuR, which is likely to stabilize the transcript after stimulation.