The 3 * -Untranslated Region of Murine Cyclooxygenase-2 Contains Multiple Regulatory Elements That Alter Message Stability and Translational Efficiency*

Renal mesangial cells regulate their expression of the pro-inflammatory gene cyclooxygenase-2 (COX-2) through mechanisms involving gene transcription and post-tran-scriptionalevents.Post-transcriptionalregulationofCOX-2isdependent,inpart,onsequenceswithinthe3 * -untrans-lated region (3 * -UTR) of the COX-2 mRNA. Insertion of the entire 3 * -UTR of COX-2 into the 3 * -UTR of luciferase resulted in a 70% decrease in luciferase enzymatic activity. Measurement of steady-state reporter gene mRNA levels suggested that the loss of activity was due to decreased translational efficiency. Deletion analysis identified the first 60 nucleotides of the 3 * -UTR of COX-2 as a major translational control element. This region of the 3 * -UTR of COX-2 is highly conserved across species; is AU-rich; and contains multiple repeats of the regulatory sequence AUUUA, reported to confer post-transcriptional control. In addition, we identified regions of the 3 * -UTR of COX-2 outside of the first 60 nucleotides that altered message stability. Some of these regions contained AUUUA consensus sequences, whereas others did not, and represent novel control elements. These results suggest that expression of COX-2 in mesangial cells depends on the complex integration of multiple signals derived from the 3 * -UTR of the message. Results were equivalent to by normalization to protein content, and data normalized protein levels reported. Quantitative Reverse Transcriptase-PCR— Total RNA was isolated from cells by a modified single-step acid/guanidine thiocyanate/phenol/ chloroform method using RNA STAT-60 reagent (Tel-Test, Inc., Friend- swood, TX). Total RNA was treated with DNase I (10 units, 37 °C, 30 min), followed by re-isolation of RNA with RNA STAT-60 reagent. DNase I treatment was repeated twice to eliminate amplification of reporter plasmid DNA and genomic DNA. RNA (0.5 m g) was reverse- transcribed with avian myeloblastosis virus reverse transcriptase (Pro-mega) using random hexamer primers. After first-strand synthesis, the cDNA was quantified by TaqMan real-time PCR using gene-specific primers and the double-stranded DNA-binding dye SYBR green I. Fluorescence was detected with an ABI Prism 7700 sequence detection system (PE Biosystems, Foster City, CA). Luciferase amplification primers were GCCTGAAGTCTCTGAT-TAAGT for the forward primer and ACACCTGCGTCGAAGATGT for the reverse primer. Amplification primers for glyceraldehyde-3-phos-phate (GAPDH) were TGGCAAAGTGGAGATTGTTGCC for the for- ward primer and AAGATGGTGATGGGCTTCCCG for the reverse primer. The amplicon was designed to be , 150 base pairs and to have a melting temperature of 78–84 °C. Primer pairs were tested to ensure a robust amplification signal of the expected size with no additional bands. Melting curves were generated to determine the temperature that maximized fluorescence from SYBR green I binding to the ampli- con and that minimized fluorescence due to primer dimers. The amount of luciferase message in each RNA sample was quantified and normal- ized to GAPDH content. Relative amounts of luciferase cDNA were calculated by the comparative C T method (49) and are expressed as a percentage of luciferase cDNA measured in

Cyclooxygenase-2 (COX-2) 1 catalyzes the conversion of arachidonate to prostaglandin H, the rate-limiting step in prostaglandin biosynthesis. COX-2 was identified as an immediateearly response gene whose synthesis is rapidly increased in response to various cytokines and mitogenic factors (1)(2)(3). Transcription of COX-2 can be activated by a variety of extracellular ligands (4 -12). Signal transduction occurs through many different signaling pathways, most of which require activation of protein kinase cascades. Studies using pharmacological inhibitors of MAPKs or their upstream protein kinase activators and dominant-negative mutant forms of protein kinases demonstrated the role of ERK, JNK, and p38 MAPK in transcriptional activation of COX-2 (5,9,(13)(14)(15)(16)(17)(18).
Recently, it has been reported that MAPK signaling pathways are involved in regulating gene expression at the post-transcriptional level (19 -24). In the majority of these investigations, activation of one or more MAPKs resulted in stabilization of the target mRNA, which was dependent on regulatory elements contained within the 3Ј-UTR of the message.
In renal mesangial cells, the induction of COX-2 is important for modulation of glomerular inflammation. COX-2 is rapidly induced in response to IL-1␤ and phorbol 12-myristate 13acetate (16 -18, 34). The cellular mechanism of IL-1␤ signaling in renal mesangial cells, although not yet fully defined, includes activation of both JNK and p38 MAPK signaling pathways (18). Blocking either of these pathways attenuates the IL-1␤-induced expression of COX-2 protein and reduces COX-2 mRNA levels. It is believed that increased COX-2 expression is due to regulation of both transcriptional and post-transcriptional events. We have previously shown that IL-1␤ increases the half-life of COX-2 mRNA and is associated with the induction of RNA-binding proteins that interact with sequences in the 3Ј-UTR of COX-2 (34). These binding proteins interact with the first 150 nucleotides of the 3Ј-UTR, which contains highly conserved adenosine-and uridine-rich elements (AREs). The results support a critical role for the AREs of COX-2 in IL-1␤dependent gene expression in mesangial cells. Thus, it appears that a common mechanism for control of gene expression by MAPK signaling pathways is through post-transcriptional gene regulation, which requires the 3Ј-UTR of the target gene.
Other investigators have shown the importance of AREs in COX-2 gene expression (35)(36)(37)(38)(39). The majority of the AREs of COX-2 reside within the first 100 nucleotides of the 3Ј-UTR. This region of the 3Ј-UTR was shown to regulate message stability and translational efficiency of hybrid reporter genes. The entire 3Ј-UTR of COX-2 encompasses Ͼ2000 bases, and it seems likely that regions of the 3Ј-UTR outside of the AREs may also play a role in regulating COX-2 expression. To determine whether additional regions of the 3Ј-UTR of COX-2 regulate gene expression in mesangial cells, we constructed a series of reporter gene expression vectors containing various regions of the 3Ј-UTR of COX-2. Based on the results of reporter gene expression, we determined that the 3Ј-UTR of COX-2 contains multiple control elements that regulate message stability and message translation, many of which represent novel control elements that lie outside of the first 100 nucleotides of the 3Ј-UTR. Thus, the level of expression for COX-2 in renal mesangial cells is determined in part by integration of multiple signals regulating post-transcriptional events that are dependent on sequences that reside in the 3Ј-UTR of the message.

EXPERIMENTAL PROCEDURES
Reagents-Unless indicated, all reagents used for biochemical methods were purchased from Sigma, VWR, or Fisher. Restriction enzymes were obtained from New England Biolabs Inc. (Beverly, MA) and Promega (Madison, WI). The plasmid pGL3-control, which encodes firefly luciferase, was purchased from Promega. Cell culture medium and fetal bovine serum were from Life Technologies, Inc. Human recombinant IL-1␤ and DNase I were purchased from Roche Molecular Biochemicals. The wheat germ in vitro translation kit was from Ambion Inc. (Austin, TX).
Reporter Gene Construction-Bluescript SK Ϫ containing DNA that encodes the 3Ј-UTR of murine COX-2 was generated as previously described (34). Various regions of the DNA were amplified by PCR using primers terminating in an XbaI recognition sequence. PCR products were ligated into the TOPO TA cloning vector (Invitrogen, Carlsbad, CA) and subsequently excised with XbaI. DNA fragments were purified by agarose gel electrophoresis and extracted using a Geneclean III kit (BIO 101, Inc., Vista, CA). DNA inserts were ligated into the unique XbaI site of the pGL3-control vector, located in the 3Ј-UTR of the firefly luciferase gene.
Transient Transfections-Mesangial cells were transiently transfected using SuperFect transfection reagent (QIAGEN Inc., Valencia, CA). Cells were plated in six-well cluster plates at a density of 2 ϫ 10 5 cells/well and incubated overnight. Mixtures of 2.5 g of reporter gene plasmid DNA in 75 l of serum-free medium and 15 l of SuperFect reagent were incubated for 5-10 min at room temperature, followed by dilution to 0.5 ml with complete medium. The DNA⅐SuperFect complex was layered onto mesangial cells (2.5 g of DNA/well); after a 2-3-h incubation, the medium was changed, and cells were incubated overnight for gene expression.
Luciferase Assay-Luciferase activity was determined using a luciferase assay system (Promega) following the manufacturer's protocol. Briefly, cell monolayers in six-well clusters were removed by scraping into 100 l of reporter lysis buffer. Cells were lysed by freeze-thawing, and cellular debris was removed by centrifugation for 30 s at 12,000 ϫ g. Luciferase activity was measured using a Lumat LB 9507 luminometer (EG&G Wallac, Gaithersburg, MD). Assays were performed by injecting 100 l of luciferase assay reagent into 20 l of supernatant diluted 1:10. Light output was measured over a 10-s time period. Activity is expressed as relative light units and was normalized to cell protein. In some instances, cells were cotransfected with a Renilla luciferase reporter gene (pRL-TK, Promega), and firefly luciferase activity was normalized to Renilla luciferase activity (data not shown). Results were equivalent to those obtained by normalization to protein content, and only data normalized to protein levels are reported.
Quantitative Reverse Transcriptase-PCR-Total RNA was isolated from cells by a modified single-step acid/guanidine thiocyanate/phenol/ chloroform method using RNA STAT-60 reagent (Tel-Test, Inc., Friendswood, TX). Total RNA was treated with DNase I (10 units, 37°C, 30 min), followed by re-isolation of RNA with RNA STAT-60 reagent. DNase I treatment was repeated twice to eliminate amplification of reporter plasmid DNA and genomic DNA. RNA (0.5 g) was reversetranscribed with avian myeloblastosis virus reverse transcriptase (Promega) using random hexamer primers.
After first-strand synthesis, the cDNA was quantified by TaqMan real-time PCR using gene-specific primers and the double-stranded DNA-binding dye SYBR green I. Fluorescence was detected with an ABI Prism 7700 sequence detection system (PE Biosystems, Foster City, CA). Luciferase amplification primers were GCCTGAAGTCTCTGAT-TAAGT for the forward primer and ACACCTGCGTCGAAGATGT for the reverse primer. Amplification primers for glyceraldehyde-3-phosphate (GAPDH) were TGGCAAAGTGGAGATTGTTGCC for the forward primer and AAGATGGTGATGGGCTTCCCG for the reverse primer. The amplicon was designed to be Ͻ150 base pairs and to have a melting temperature of 78 -84°C. Primer pairs were tested to ensure a robust amplification signal of the expected size with no additional bands. Melting curves were generated to determine the temperature that maximized fluorescence from SYBR green I binding to the amplicon and that minimized fluorescence due to primer dimers. The amount of luciferase message in each RNA sample was quantified and normalized to GAPDH content. Relative amounts of luciferase cDNA were calculated by the comparative C T method (49) and are expressed as a percentage of luciferase cDNA measured in cells transfected with pGL3-control.
Reporter Gene mRNA Decay-Mesangial cells were transiently transfected and incubated for 24 h for luciferase gene expression. Transcription was inhibited at this point by adding actinomycin D (10 g/ml) to the cell culture medium. Total RNA was isolated at various times after actinomycin D addition, and luciferase mRNA content was determined by quantitative reverse transcriptase-PCR as described above. Luciferase mRNA levels were normalized to GAPDH mRNA content and are expressed as a percentage of the mRNA level at the 0-h time point.
In Vitro Translation-Total RNA was translated in vitro using the wheat germ translation kit according to the manufacturer's protocol. Translation reactions contained 24 l of wheat germ extract, 1.2 l each of Master MixϪLeu and Master MixϪMet (a mixture of all the amino acids except the one indicated), 100 mM potassium acetate, and 4 -8 g of total RNA in total volume of 50 l. Reactions were incubated for 60 min at 30°C and stopped by placing tubes on ice. Duplicate samples containing 20 l of the translation reaction were assayed for luciferase activity using the luciferase assay described above. The amount of translated protein is expressed as luciferase activity normalized to total RNA content.
Statistical Analysis-All experiments were performed at least three times, each in duplicate. Data are expressed as the means Ϯ S.E. Comparison of means was performed using Student's t test.

The 3Ј-UTR of COX-2 Decreases Expression of the Luciferase
Reporter Gene-To determine the effect of the 3Ј-UTR of COX-2 on gene expression, reporter gene constructs were created by inserting DNA encoding various regions of the 3Ј-UTR of murine COX-2 message into the 3Ј-UTR of the luciferase gene ( Fig. 1). Each construct contained the luciferase coding sequence under the control of the SV40 promoter and enhancer elements, followed by the 3Ј-UTR of luciferase and the SV40 late poly(A) signal. The reporter constructs differed only in the regions of the 3Ј-UTR of COX-2 that were inserted into the luciferase 3Ј-UTR. Regions of the 3Ј-UTR of COX-2 included the full-length 3Ј-UTR (nucleotides 1-2232), serial deletions from the distal and proximal ends of the 3Ј-UTR, and two internal regions of the 3Ј-UTR. Since all reporter constructs contained identical promoter elements, differences in luciferase activity reflect differences in regulation as a result of post-transcriptional events.
We used a mouse immortalized mesangial cell line as a model system to study regulation of COX-2 in renal mesangial cells. Reporter gene constructs were transiently transfected into this cell line, incubated overnight, and assayed for luciferase activity. Luciferase activity measured in cells transfected with the pGL3-control vector lacking 3Ј-UTR sequences of COX-2 was designated as 100%. Inserting the entire 3Ј-UTR (nucleotides 1-2232) of COX-2 into the 3Ј-UTR of luciferase resulted in a 70% decrease in luciferase activity (Fig. 2), suggesting the presence of negative regulatory element(s) within the 3Ј-UTR of the COX-2 message. Compared with the fulllength 3Ј-UTR, truncation of the distal region of the 3Ј-UTR of COX-2 to nucleotide 1558 or 1384 had no additional effect on luciferase activity ( Fig. 2A). Removal of nucleotides 792 to 1384 caused a significant increase in luciferase activity (p Ͻ 0.05) to 66% of the control. This increase suggests the presence of a negative element between nucleotides 792 and 1384. In support of this conclusion, insertion of only nucleotides 792-1384 alone lowered luciferase activity to 24% of the control (Fig. 2B). Further truncation from nucleotide 792 to 373 resulted in a significant decrease in luciferase activity (p Ͻ 0.001) to a level of activity that was 12% of the control ( Fig. 2A). This result suggested the presence of a positive element between nucleotides 792 and 373. Surprisingly, inclusion of nucleotides 373-792 not only failed to increase luciferase activity, but rather decreased luciferase activity (Fig. 2B). Thus, this region appears to modulate regulation of gene expression in a positive manner through interaction with the first 373 nucleotides of the 3Ј-UTR, but cannot increase gene expression by itself. Inclusion of only the first 60 nucleotides of the 3Ј-UTR of COX-2 decreased reporter gene activity to 10% of the control ( Fig. 2A). This region of the 3Ј-UTR contains 7 of the 12 AUUUA consensus sequences, suggesting that the AREs in the 3Ј-UTR of COX-2 negatively regulate gene expression.
Truncation from the proximal region of the 3Ј-UTR of COX-2 revealed the presence of an additional negative element. Removal of the first 60 nucleotides resulted in an increase in luciferase activity to 70% of the control. This region contains seven AUUUA consensus sequences, suggesting that they account for the decreased luciferase activity measured with the construct containing the full-length 3ЈUTR. However, truncation to nucleotide 373 or 792 reduced luciferase activity to a level equivalent to that of the full-length 3Ј-UTR, and inclusion of only the terminal 674 nucleotides of the 3Ј-UTR of COX-2 (nucleotides 1558 -2232) caused a significant decrease in luciferase activity to nearly 5% of the control (Fig. 2B). These results indicate that additional negative elements are present within the 3Ј-UTR that are affected by the "context" in which they are presented. The terminal region contains three AUUUA consensus sequences, further supporting a negative role for AREs in the 3Ј-UTR of COX-2.
Decreased Expression of the Reporter Gene Occurs through Changes in Both mRNA Stability and Message Translation-Measurement of luciferase activity was used to quantitate the amount of reporter enzyme synthesized by transfected cells. Changes in luciferase activity could be due to either alterations in message stability or rates of mRNA translation. To distinguish between these two possibilities, we measured the steadystate levels of luciferase mRNA using reverse transcriptase, followed by real-time PCR analysis. In this technique, the PCR product was measured as it accumulated, allowing for accurate quantitation of mRNA levels without the ambiguities associ-ated with traditional reverse transcriptase-PCR. Luciferase mRNA levels in cells transfected with reporter gene constructs were normalized to GAPDH mRNA levels and are expressed as a percentage of luciferase mRNA measured in cells transfected with the pGL3-control vector (Fig. 3). If loss of luciferase activity were due to decreased message stability, then we would observe comparable changes in message levels. If decreased luciferase were due to inhibition of translation, then we would expect to measure no change or a disproportionate change in luciferase mRNA levels.
Comparison of the results from luciferase activity measurements (Fig. 2) and quantitation of luciferase mRNA levels (Fig.  3) indicated that decreased luciferase expression occurred through multiple mechanisms. Insertion of the entire 3Ј-UTR of COX-2 into the 3Ј-UTR of luciferase had no effect on steadystate mRNA levels (Fig. 3A, 1-2232 versus pGL3c). Likewise, luciferase message levels using the reporter construct containing nucleotides 1-792 were not significantly different compared with the control. Thus, the decreased luciferase activity measured using these reporter constructs was presumed to reflect a decreased rate of message translation. In contrast, inclusion of nucleotides 1-60, 373-792, 792-1384, or 1558 -2232 caused a significant and dramatic drop in luciferase mRNA levels compared with the luciferase gene alone (pGL3-control) and compared with the construct containing the entire 3Ј-UTR of COX-2 (nucleotides 1-2232). In most cases, the magnitude of the decrease in luciferase mRNA levels was nearly equal to the corresponding decrease in luciferase activity, suggesting that reporter gene expression with these constructs is strongly dependent on message stability. Truncation of the 3Ј-UTR to nucleotide 60 lowered luciferase mRNA levels to 30% of the control (Fig. 3A), whereas luciferase activity decreased further to 10% of the control ( Fig. 2A), suggesting both altered message stability and translational regulation.
To confirm that changes in steady-state luciferase mRNA levels reflect altered message stability, we directly measured message degradation in cells transfected with three different constructs (Fig. 4). Mesangial cells were transiently transfected, incubated for 24 h, and treated with actinomycin D to stop transcription. Luciferase mRNA was measured at various times after inhibition of transcription. Luciferase message without any COX-2 sequences was very stable and exhibited little or no decay over the 10-h treatment period. Adding the full-length 3Ј-UTR of COX-2 (nucleotides 1-2232) to the luciferase message had no significant effect on its stability. In contrast, insertion of the proximal 60 nucleotides of the 3Ј-UTR of COX-2 into the 3Ј-UTR of luciferase mRNA caused a dramatic decrease in message stability. These findings are in complete agreement with steady-state mRNA measurements.
The results presented above suggest that various regions of the 3Ј-UTR of COX-2 regulate gene expression by altering message stability and/or translational efficiency. Table I shows the ratio of luciferase mRNA levels and luciferase activity levels in cells expressing the various reporter constructs. This quotient reflects the relative contribution of translation and message stability to the regulation of reporter gene activity. A larger number indicates that a greater contribution of message translation occurred, and the closer the ratio is to 1, the greater the dependence of reporter gene expression was on message stability. Cells expressing reporter constructs containing re-gions 1-2232 and 1-60 had the highest ratio of mRNA levels to luciferase activity, indicating a strong effect on message translation. Reporter constructs containing regions 1-373, 60 -2232, 792-2232, 1558 -2232, 792-1384, and 373-792 of the 3Ј-UTR of COX-2 expressed luciferase mRNA at levels that corresponded to an equivalent change in luciferase activity. Accordingly, these constructs had ratios nearly equal to 1, suggesting that the loss of reporter expression is due to decreased message stability. Results from the other constructs generated ratios that fell in between these extremes, suggesting a mixed effect on both message stability and message translation.
Translational Regulation Occurs in Vitro-Comparison of luciferase activity and mRNA measurements suggested that insertion of the entire 3Ј-UTR of COX-2 into the 3Ј-UTR of luciferase altered reporter gene expression by decreasing translational efficiency. To directly measure changes in trans-

FIG. 2. Inclusion of sequences from the 3-UTR of COX-2 in the chimeric reporter message causes a decrease in luciferase activity.
Mouse mesangial cells were transiently transfected with reporter gene constructs, incubated for 24 h, and assayed for luciferase activity. A, luciferase activity results using constructs containing serial deletions from the distal end of the 3Ј-UTR; B, luciferase activity results using constructs containing serial deletions from the proximal end of the 3Ј-UTR and constructs containing internal regions of the 3Ј-UTR of COX-2. Luciferase activity was normalized to total cell protein and is expressed as a percentage of activity measured in cells transfected with the luciferase gene without COX-2 3Ј-UTR sequences (pGL3-control (pGL3c)). Results are the means Ϯ S.E. for 5-23 independent experiments, each measured in duplicate. a ϭ significantly different from pGL3-control, p Ͻ 0.001; b ϭ significantly different from pGL3-control, p Ͻ 0.05; c ϭ significantly different from construct 1-2232, p Ͻ 0.001; d ϭ significantly different from construct 1-2232, p Ͻ 0.01.

FIG. 3. Various sequences from the 3-UTR of COX-2 cause a decrease in chimeric reporter gene message levels.
Mouse mesangial cells were transiently transfected with reporter gene constructs and incubated for 24 h, and total RNA was extracted. RNA was reversetranscribed and assayed for luciferase and GAPDH cDNAs using Taq-Man real-time quantitative PCR. Luciferase message levels were normalized to GAPDH mRNA and are expressed as a percentage of the message level measured in cells transfected with luciferase gene without COX-2 sequences (pGL3-control (pGL3c)). A, luciferase mRNA levels using constructs containing serial deletions from the distal end of the 3Ј-UTR; B, luciferase message levels using constructs containing serial deletions from the proximal end of the 3Ј-UTR and constructs containing internal regions of the 3Ј-UTR of COX-2. Results are the means Ϯ S.E. for three to six independent experiments, each measured in duplicate. a ϭ significantly different from pGL3-control, p Ͻ 0.001; b ϭ significantly different from pGL3-control, p Ͻ 0.01; c ϭ significantly different from pGL3-control, p Ͻ 0.05; d ϭ significantly different from construct 1-2232, p Ͻ 0.005; e ϭ significantly different from construct 1-2232, p Ͻ 0.05. lational efficiency, we tested the ability of the 3Ј-UTR of COX-2 to alter message expression using an in vitro translation assay. Luciferase mRNA levels in cells transfected with the luciferase vector alone (pGL3-control) or with the luciferase vector containing the entire 3Ј-UTR of COX-2 (nucleotides 1-2232) were comparable (Fig. 3); and therefore, decreased translational efficiency would be directly correlated with decreased luciferase activity when using equal amounts of total RNA. Total RNA isolated from cells transfected with these two constructs was translated using a wheat germ lysate and assayed for luciferase activity. Luciferase activity measured following in vitro translation of RNA isolated from cells transfected with the reporter construct containing the full-length 3Ј-UTR was 27% of that measured with RNA from control cells (Fig. 5). This was similar to the luciferase activity measured in cell lysates (32%) and corroborated the indirect measurement of translational regulation in cultured cells. The fact that alterations in translation were observed in an in vitro system suggests that the ability to decrease message translational efficiency is inherent to the structure of the 3Ј-UTR of COX-2.
IL-1␤ Fails to Regulate Luciferase Expression-We have previously shown that rat primary mesangial cells express COX-2 at high levels in response to IL-1␤ (16 -18, 34). Cytoplasmic extracts derived from these treated cells exhibit an IL-1␤-dependent RNA gel shift using the 3Ј-UTR of murine COX-2 as   Fig. 3 (mRNA level) and Fig. 2 (luciferase level) are expressed as a ratio. This ratio was used as a gauge of the relative contributions of message destabilization and translational inhibition to overall reporter gene expression. Ratios equal to 1.0 indicate that protein levels were directly correlated to message levels. A high ratio value indicates a greater degree of translational silencing.  5. The full-length 3-UTR of COX-2 inhibits message translation in vitro. RNA isolated from mesangial cells transfected with reporter constructs containing luciferase alone (pGL3-control) or luciferase fused to the full-length 3Ј-UTR of COX-2 (nucleotides 1-2232) were translated in vitro using a wheat germ lysate. The translation product was determined by assaying luciferase activity. Results are expressed as a percentage of activity measured in cells transfected with pGL3-control and are shown in comparison with luciferase activity measured in cell lysates (in vivo translation). Results are the means Ϯ S.E. for three independent experiments, each measured in duplicate.
the target sequence (34). It has been suggested that IL-1␤ induces phosphorylation of cytosolic proteins that bind to the 3Ј-UTR of COX-2 and stabilize the message (34). We wanted to determine whether IL-1␤ regulates expression of the reporter gene constructs described above. If the 3Ј-UTR of COX-2 were sufficient for post-transcriptional regulation, then we would expect IL-1␤ to reverse the decrease in message stability and translational efficiency, resulting in an increase in measured luciferase activity.
Expression of the reporter gene in rat primary mesangial cells was similar to the results obtained with the murine cell line (Fig. 6). Insertion of the entire 3Ј-UTR of COX-2 into the 3Ј-UTR of luciferase resulted in a decrease in luciferase activity to 25% of the control. Luciferase activity decreased to 16 and 15% using constructs containing 3Ј-UTR regions bounded by nucleotides 1-373 and 1-60, respectively (Fig. 6). These results are in good agreement with the results obtained using murine mesangial cells. However, treatment with IL-1␤ for 24 h failed to reverse the negative regulation of reporter gene expression due to sequences within the 3Ј-UTR of COX-2 (Fig. 6), such that luciferase activity was indistinguishable from that measured in untreated cells. The inability to measure regulation of reporter gene expression by IL-1␤ under our conditions may be due to the fact that additional message sequences are required to fully restore IL-1␤ regulation.

DISCUSSION
Conserved ARE sequences have been identified within the 3Ј-UTR of many short-lived messages that provide signals that regulate message stability (41)(42)(43)(44)(45)(46)(47). The 3Ј-UTR of COX-2 is AU-rich and contains multiple copies of an AUUUA consensus sequence that are implicated in regulating post-transcriptional events. Several investigators have shown that insertion of the 3Ј-UTR of COX-2 within the 3Ј-UTR of a reporter gene alters its expression and that the ARE-containing region is crucial for this response (35)(36)(37)39). The majority of the AREs are located in the proximal portion of the 3Ј-UTR of COX-2. This region is highly conserved in mammalian species and contains a cluster of six or seven AUUUA sequences. The data presented here show that in addition to these conserved AREs, the 3Ј-UTR of COX-2 contains novel regulatory elements that are important for the expression of a chimeric reporter gene.
The ability of distinct regions of the 3Ј-UTR of COX-2 to alter reporter gene expression through both message stability and translational efficiency is summarized in Fig. 7. Regions of the 3Ј-UTR that caused a decrease in steady-state luciferase mRNA levels imply the presence of elements that alter message stability. Comparison of steady-state mRNA levels and rates of message decay for representative reporter constructs corroborated this implication. Regions that cause changes in luciferase activity, not accompanied by a similar decrease in message levels, reflect regulatory elements that alter translational efficiency. Although this constitutes an indirect measurement of translational rates, this interpretation was supported by the fact that the full-length 3Ј-UTR (nucleotides 1-2232) caused a comparable decrease in translational efficiency measured in vitro.
Other investigators have shown that the 3Ј-UTR of COX-2 can regulate gene expression. However, depending on the system and species of study, the sequences required for reporter gene regulation and the ability of extracellular signals to modify reporter gene expression were highly variable. In our hands, insertion of the entire 3Ј-UTR of COX-2 into the reporter gene 3Ј-UTR had no effect on either message degradation or steadystate mRNA levels. We found that the decay of luciferase reporter mRNA was very slow, similar to that reported by Balmer et al. (48). The inability of the full-length 3Ј-UTR of COX-2 to confer instability to a transiently transfected reporter gene was also reported by Dixon et al. (37). Interestingly, this same group was able to demonstrate destabilization of a luciferase reporter gene by the 3Ј-UTR of COX-2 when using stably transfected cell lines. Similarly, Xu et al. (38) were able to demonstrate that sequences from the 3Ј-UTR of COX-2 cause a change in the half-life of luciferase message expressed in stably transfected rat smooth muscle cells. These apparently contradictory results may be due to cell-specific differences in message degradation or may reflect the inability of the cell to degrade higher levels of reporter message in transiently transfected cells compared with stably transfected cells.
Clearly, one region of the 3Ј-UTR that is crucial for regulating expression of COX-2 is the proximal 60 -150 nucleotides. Several groups of investigators have shown that this region can decrease the message stability of a normally stable reporter gene (6,35,37). Additionally, Dixon et al. (37) were able to demonstrate that a 116-nucleotide region bearing the conserved AREs from human COX-2 message can regulate message translation in transiently transfected cells. What we have shown is that the 3Ј-UTR region contained within nucleotides 1-60 caused a significant decrease in message stability, determined by measuring both message decay and steady-state mRNA levels. Furthermore, we have demonstrated that this region of the 3Ј-UTR also caused a decrease in translational efficiency. This represents the first report that this region can regulate both message stability and translation in a single cell system. Deletion of this region alone resulted in a dramatic increase in activity and no increase in message levels, indicating a loss of translational silencing. Therefore, it is likely that FIG. 7. Multiple control elements are found in the 3-UTR of COX-2. Regions of the 3Ј-UTR of COX-2 that decreased reporter gene mRNA levels to at least 30% of the control are indicated as stability control elements. The region bounded by nucleotides 1-60 caused a decrease in luciferase activity that was 3-fold greater than the decrease in luciferase mRNA and is designated as a translational control element. All other constructs exhibiting translational control of reporter gene expression contain this region, except for the construct containing nucleotides 373-2232. Truncation of this construct to nucleotides 792-2232 abolished the translational regulation, suggesting a second translational control element between nucleotides 373 and 792. Accordingly, this region is indicated to contain both a translational and stability control element. this region accounts for the majority of the translational effects measured with the other constructs, including the reporter construct containing the full-length 3Ј-UTR.
In addition to the highly conserved proximal region of the 3Ј-UTR of COX-2, we have identified other regions of the 3Ј-UTR that altered gene expression. Serial deletions from the proximal end of the 3Ј-UTR resulted in a steady decrease in luciferase mRNA levels. The strongest destabilizing effect was measured using the construct containing nucleotides 1558 -2232. Gou et al. (35) showed that insertion of the entire 3Ј-UTR of human COX-2 downstream of the luciferase coding region causes a strong reduction in luciferase activity when expressed in human endothelial cells. Truncation of ϳ600 bases from the distal end results in a chimeric message that is nearly as stable as the luciferase gene alone, indicating removal of a negative regulatory element. The area deleted completely overlaps with nucleotides 1559 -2232 of the murine 3Ј-UTR used in our study, supporting the idea that this region is important for regulating message stability across species. A recent report evaluating the ability of the 3Ј-UTR of rat COX-2 to regulate chimeric message stability identified a region in the very distal portion of the 3Ј-UTR that significantly decreases message half-life (38). Alignment of this rat sequence and the murine sequence used in our study indicates that this regulatory region is beyond the sequences contained in our full-length 3Ј-UTR; and therefore, it is not known whether similar regulation occurs in the mouse mesangial cell system.
Much emphasis has been placed on the role of the AREs in the 3Ј-UTR of COX-2 in regulating message stability. We identified four regions of the 3Ј-UTR of murine COX-2 that strongly decreased message stability (Fig. 7). The distal region (nucleotides 1558 -2232) caused the largest decrease in luciferase activity and an equal decrease in mRNA levels. This region also contains 3 of the 12 ARE consensus sequences that may be critical for message destabilization. The region within nucleotides 373-792 contains a single AUUUA motif, and the region containing the majority of the ARE sequences (nucleotides 1-60) caused a 70% reduction in message level, but also exhibited the strongest ability to regulate translation. The central region (nucleotides 792-1384) is unique in that it conferred a strong destabilizing effect, but contains no ARE sequence. Thus, the ability to cause message instability is not strictly associated with ARE consensus sequences, and the region between nucleotides 792 and 1384 contains a novel instability element.
The identification of specific regions of the 3Ј-UTR of COX-2 that alter the expression of a chimeric message increases our understanding of how mesangial cells regulate COX-2 expression. However, these sequences were not sufficient to confer regulation of luciferase expression in response to treatment with IL-1␤. This result suggests that other regions of the COX-2 message are required for IL-1␤-dependent post-transcriptional regulation. Similar results were found studying the expression of IL-2 in response to T-cell activation (19). In this case, regions within the 3Ј-UTR were able to cause a decrease in reporter gene half-life, but sequences within the 5Ј-UTR and the coding region were required to confer regulation of reporter gene expression following activation. Whether the 5Ј-UTR of COX-2 plays a role in regulating gene expression is not known and merits additional investigation.
The results presented here support the role of the 3Ј-UTR of COX-2 in regulating COX-2 production, both through altered message stability and translational efficiency. Separating various regions of the 3Ј-UTR and assigning unique functions to isolated sequences provide essential tools to further understand how individual regions of a single transcript contribute to its overall expression.