Immediate-early MEK-1-dependent Stabilization of Rat Smooth Muscle Cell Cyclooxygenase-2 mRNA by Gαq-coupled Receptor Signaling*

Activation of Gαq-coupled P2Y nucleotide receptors strongly (>100-fold) induces the rat vascular smooth muscle cell cyclooxygenase-2 (COX-2) mRNA, yet transcription is induced only ∼3-fold over 1 h. Intact cell decay analysis of tetracycline-suppressible luciferase chimera mRNAs shows that regulated stabilization of the intrinsically unstable mRNA contributes to this response. Deletion mapping of the 2468-base COX-2 mRNA 3′-untranslated region (UTR) shows that a distal, 130-base AU-rich region functions as a cis-acting regulated stabilization response element, which under basal conditions serves as the dominant instability determinant for the 3′-UTR. Regulation of this response is through the p42/44 MAP kinases, whereas the p38 MAP kinases are not involved. The stabilization response element binds avidly and specifically to a prominent nuclear-enriched ∼90-kDa factor and several less abundantly labeled mRNA binding proteins that are unaffected by P2Y receptor signaling. Although other instability determinants are located throughout the rat COX-2 mRNA 3′-UTR, mitogen signaling only interferes with rapid decay mediated by its most distal 130 bases. A complex of nuclear factors that bind this mRNA region specifically may include candidate targets for regulatory modulation. These observations support the general notion that the rapid induction of immediate-early gene expression through mitogenic receptors involves simultaneous activation of transcriptional and post-transcriptional mechanisms.

Depending upon the initial stimulus and cell phenotype, most major signaling pathways have been purported to participate in regulating COX-2 gene expression. These include pathways controlled by cytokines, various second messengers, reactive oxygen species, and each of the final common mitogenactivated protein (MAP) kinase modules (p42/44, p38, and c-Jun NH 2 -terminal kinase/stress-activated protein kinase) (8 -14). Although among the more robustly induced mRNAs known, the effector mechanisms responsible ultimately for pleiotropic regulation of COX-2 mRNA expression remain unsettled. Transcriptional induction at the COX-2 locus is usually quite modest as assessed directly by run-on transcription and/or indirectly through promoter-reporter activity studies (2,5,11,(15)(16)(17)(18). Several studies, most focusing on the effects of cytokines and also employing actinomycin D mRNA chase analysis of the native transcript, have raised the possibility that the COX-2 mRNA is post-transcriptionally regulated (9, 19 -23). Some information points to a role of the 3Ј-UTR in this effect, but the putative regulatory elements in this large region of mammalian COX-2 mRNAs are not clearly defined (19 -21).
Cell surface receptor signaling is widely known to regulate scores of factors that participate in controlling rates of gene transcription (24). The broad ranges of stability exhibited by various mRNAs also play an important factor in the steadystate equation. These depend on many processes, including translational status, cellular compartmentalization, and the presence of various cis-acting mRNA stability determinants (25,26). The possibility that cell surface receptor signaling directly modulates mRNA turnover rates, similar to its influence on rates of transcription, receives growing attention. Yet, as in the specific case of purported COX-2 mRNA stability regulation, very few studies support the concept unambiguously. Uncertainties invariably arise due to the collateral effects caused by general transcription inhibitors that are used in mRNA stability studies, which interfere with both signaling and mRNA decay processes (27)(28)(29)(30)(31)(32).
More selective and less toxic recombinant approaches are now available to isolate better post-transcriptional from transcriptional regulation (33). For example, the tetracycline-regulated system allows for transcriptional suppression using a class of drugs that neither activate signaling pathways nor appear to interfere with known mRNA decay processes (34). Results using these newer approaches contribute to a very limited body of data for mRNAs that are modulated by receptor signaling pathways through post-transcriptional control (35)(36)(37)(38)(39)(40).
In this study, a retroviral vector-based, anhydrotetracyclinesuppressible system is used to examine whether post-transcriptional mechanisms in smooth muscle cells contribute to COX-2 mRNA modulation by mitogenic receptor signaling. We find that p42/44 MAP kinase-regulated mRNA stabilization contributes to immediate-early COX-2 gene expression following activation of a mitogenic G␣ q -coupled receptor. This response is mediated by a distal 3Ј-UTR stability determinant.

MATERIALS AND METHODS
Cell Culture-A primary line of rat thoracic aorta smooth muscle cells was used between passages 8 and 23 (24 -69 population doublings since inception) in a growth medium containing 10% calf serum as described in detail (41). The cells were grown on 6 ϫ 35 mm plates for RNA samples or on 150-mm diameter plates for harvesting nuclear and cytoplasmic extracts essentially as described (42). Phoenix amphotropic retroviral producer cells (43) obtained from the ATCC were grown in the same media except using 10% fetal bovine rather then calf serum.
mRNA Reporter Plasmids-A full-length 4.4-kilobase pair rat COX-2 cDNA in pBluescript (5), which we term pRDCOX-2, was kindly provided by Ray DuBois (Vanderbilt University, Nashville, TN). Its distal 3Ј-UTR sequence was determined and is deposited as an update with previously reported data. COX-2 mRNA 3Ј-UTR cDNA fragments were cloned into a SalI-ApaI-HpaI-NsiI multicloning site inserted in the retroviral vector pXF40-Luc just after the luciferase open reading frame and immediately upstream of polyadenylation signals derived from the SV40 genome (41). The vector pKX51a has a 2.49-kilobase pair ScaI-XhoI fragment from the rat COX-2 cDNA essentially blunt-cloned into the SalI-HpaI sites of pXF40. This XhoI site is derived from pBluescript and lies immediately 3Ј adjacent to a 33-base poly(A) ϩ tail remnant in the cDNA. pKX54 has a 1.47-kilobase pair ScaI-ApaI fragment 5Ј bluntcloned into SalI and then ApaI. pKX55 has a 0.76-kilobase pair ApaI-HpaI COX-2 3Ј-UTR fragment, and pKX56 contains the most distal 0.26-kilobase pair sequence as an HpaI-XhoI fragment cloned blunt into HpaI and NsiI sites. The COX-2 3Ј-UTR inserts for pKX67 and pKX68 were generated by polymerase chain reaction with the oligonucleotides pairs, respectively, H38 (5ЈACGAGATGACGCCGGTGAACTT) and H37 (5ЈCAAATGCATCGGCAGCAGTCACATACTTA) or H58 (5ЈTTGTTA-ACTGCTGCCGTGCCTCAGAG) and H21 (5ЈGCGAATATGCATTTA-AAGACACAATCAAACTT) before cloning into pXF40-Luc using an HpaI-NsiI strategy. All constructs were verified by sequencing.
Retrovirus Production and Smooth Muscle Cell Infection-By using a transient retrovirus production technique adopted from others (44) as described previously (45), Phoenix producer cells were transfected with retroviral plasmids by a CaPO 4 precipitation method. Smooth muscle cells expressing the tetracycline transactivator protein (40) were infected with conditioned supernatants from the retroviral producer cells. Cells expressing chimera mRNA were studied over 3-4 repeated passages following an initial 7-day selection period using 200 mg/ml G418 (Life Technologies, Inc.), exploiting a neomycin resistance gene driven by the viral long terminal repeat promoter in the pXF40-derived vectors.
RNase Protection Assays-Media were aspirated after cell treatments at 37°C in a 5% CO 2 atmosphere before adding 1.0 ml of the Trizol Reagent (Life Technologies, Inc.). Total RNA (5-15 g) was hybridized with a mixture of sequencing gel-purified [ 32 P]UTP-labeled riboprobes antisense to luciferase and cyclophilin mRNA before RNase digestion, using recommended procedures and reagents supplied with the RPAII kit (Ambion, Inc.). After sample electrophoresis on sequencing gels, wet gels were exposed to storage phosphor screens (Molecular Dynamics, Inc.) for ϳ5 h before collecting phosphorimages. Hybridization signals were quantified by volume integration using ImageQuaNT software, using the cyclophilin mRNA signals in each sample to normalize the luciferase mRNA signals. The mRNA data are presented as a percentage of these ratios in untreated control samples (taken as 100%) within each experiment. Curve fitting routines using the Graph-Pad Prism version 3.0 software package were used to determine decay rate constants.
Nuclear Run-on Assays-Nuclei were harvested from cells treated with 100 M UTP for the indicated times and labeled with [␣-32 P]UTP and unlabeled ribonucleotides before extracting radioactive RNA as described (46). This was hybridized overnight to Hybond N slot blots of GAPDH, pBluescript (as an empty vector control), and pRDCOX-2 denatured plasmids at 56°C using solutions supplied in an Ambion RNA hybridization kit. The blots were washed at 66°C before exposure on phosphorimage cassettes for 5 h. Relative transcription rates were determined by calculating ratios of COX-2 to GAPDH transcription in each sample by volume integration of the hybridization signals, which were normalized to the ratio determined within non-stimulated cells.
In Vitro Transcription-Transcription reactions used T7 or T3 RNA polymerase, 50 Ci of 800 Ci/mmol of [␣-32 P]UTP (final total UTP concentration of 0.1 mM for radioactive probes and 1.0 mM for unlabeled transcripts), and reagents and directions supplied in the Maxiscript kit (Ambion, Inc.). The template for the antisense-luciferase riboprobe is a 540-base pair XbaI-EcoRI fragment from poLuc (47) cloned into pBluescript (vector pKX37), which was linearized with XbaI. The pTri-Cyp vector (Ambion, Inc.) served as template to make the antisense-cyclophilin riboprobe. The template for sense and antisense 263-base distal 3Ј-UTR probes (from bases 4142-4404) used in mRNA protein binding assays were constructed by digestion of pRDCOX-2 with SmaI and HpaI before recircularization. Digestion of pRDCOX-2 with HindIII and EcoRI before recircularization created a template to make a slightly smaller 186-base transcript (from bases 4218 -4404) for the distal 3Ј-UTR. A template for the proximal 3Ј-UTR AU-rich region control probe (bases 1922-2173) was created by polymerase chain reaction with the primers 5ЈGCGAATTCCGTTCAACTGAGCTGTAAGAG and 5ЈATCTC-GAGTCATTTCCCTTCTCACTGGC and cloned into the EcoRI and XhoI sites of pBluescript(SK).
mRNA/Protein Binding Assays-These reactions are adapted from published protocols (48) as described in detail (49). Sequencing gelpurified transcripts (ϳ10 5 to 10 6 cpm.; ϳ0.1-1.0 M final concentration) were incubated in 96-well plates for 15 min at 22°C with the indicated concentrations of extracted protein in a final volume of 20 l. After adding heparin, the plates were placed in an ice bath and exposed to UV light using a Stratalinker (Stratagene). After 30 min incubation at 37°C with 2.5 g of RNase A and 2.5 units of RNase T1, UV crosslinked samples were heated in a standard loading buffer and resolved by SDS-polyacrylamide gel electrophoresis before autoradiography.
p42/44 and p38 MAP Kinase Assays-Confluent cells in serum-free media as above were grown in 35-mm diameter dishes and treated as indicated. After lysis using 150 l of a buffer recommended in the PhosphoPlus p42/44 MAP kinase assay kit (New England Biolabs, Inc. Natick, MA), 20-l aliquots were separated by electrophoresis on SDS-12% polyacrylamide minigels, before transfer to Immobilon P membranes and performing Western analysis using antibodies and directions from the kit. After treating the cells similarly, the samples were lysed in a buffer supplied with the New England Biolabs p38 Kinase Assay Kit to measure p38 MAP kinase activity. Using the manufacturer's directions, phosphorylated p38 was immunoprecipitated and used to phosphorylate exogenous glutathione S-transferase-ATF-2 substrate, which was detected after Western blotting using an anti-phospho ATF-2 antibody. Signals were developed using the Amersham Pharmacia Biotech ECL reagents and film.

RESULTS
Gene expression responses following stimulation of our smooth muscle cell preparation with the nucleotide UTP are mediated by mitogenic G␣ q -coupled P2Y receptors (42). A bolus stimulation with UTP (100 M) elicits a sharp increase in COX-2 mRNA expression (137 Ϯ 33 (mean Ϯ S.E.; n ϭ 3)-fold over basal at 60 min). This robust effect is transient, reducing to 20 Ϯ 6-fold over basal at later time points (Fig. 1, A and B). Nuclear run-on assays performed after 20, 40, and 60 min of UTP stimulation, representing the upstroke mRNA response phase, show only a modest ϳ3-fold transcriptional induction in response to this mitogen agonist ( Fig. 1, C and D). These latter observations suggest that transcriptional regulation alone would unlikely explain the robust nature of P2Y receptormediated COX-2 gene induction in this preparation. Here, we test the hypothesis that immediate-early COX-2 mRNA induction also involves regulation of a post-transcriptional mechanism by receptor signaling.
To begin, we focused on the role of the COX-2 mRNA 3Ј-UTR which comprises 2.49 kilobases of the 4.4-kilobase rat COX-2 mRNA. The 3Ј-UTR region is highly enriched in A and U content (A ϩ U ϭ 66%), including 19 AU-rich motifs defined as AU n A (where n ϭ 3-6). No AU-rich motifs exist in either the 5Ј-UTR or coding region. Chimera mRNAs were created by fusing various COX-2 mRNA 3Ј-UTR fragments downstream of a luciferase coding region ( Fig. 2A). Transcribing these in smooth muscle cells from an anhydrotetracycline (Antet)-suppressible minimal cytomegalovirus promoter (34) provides a more selective approach to suppress transcription than using general transcriptional inhibitors and thus measures intrinsic mRNA stability and how it might be affected by receptor signaling.
By using retroviral vectors, cells first prepared to express a tetracycline transactivator protein constitutively (40) were infected subsequently with vectors containing the Antet-suppressible chimera mRNA transgenes. Antet prevents enhancer binding of the tetracycline transactivator and has been shown directly to inhibit transcription in this preparation, which has a "leak" of transgene expression that is typically between 3% and no more than 10% of maximal transcriptional activity (40,41,50). As shown by the results of Antet mRNA chase experiments in Fig. 2B, incorporation of COX-2 mRNA sequences intrinsically destabilizes the luciferase mRNA (XF40). Most notably, those mRNAs containing the most distal COX-2 mRNA 3Ј-UTR (KX51a, KX56, and KX68) decay more rapidly than other constructs tested. As revealed below, these more unstable mRNAs are also the only that are stabilized in response to P2Y receptor signaling.
Steady-state levels of the control luciferase mRNA (XF40) are unaffected by UTP stimulation alone (Fig. 3). After adding Antet, this control mRNA decays with a half-life of 65 min, a rate that is not significantly different from that in cells cotreated with Antet and UTP ( Fig. 3; see also Table I), establishing that UTP stimulation has no effect on luciferase mRNA stability. In contrast, steady-state levels of the KX51a chimera mRNA, which incorporate the full-length 2.49-kilobase COX-2 mRNA 3Ј-UTR, are induced maximally 2-fold within 40 min after adding UTP (Fig. 4, A and B). This chimera mRNA also decays three times faster (t1 ⁄2 ϭ 22 min; Table I) than the control luciferase mRNA after treatment with Antet. As shown in Fig. 4, Antet-induced decay of the full-length 3Ј-UTR chimera is significantly delayed during co-treatment with UTP (t1 ⁄2 ϭ 41 min; see also Table I), indicating P2Y receptor activation stabilizes the chimera mRNA.
UTP treatment alone has no effect on steady-state levels of the KX54 and KX55 mRNAs, which respectively possess the FIG. 1. UTP-stimulated COX-2 mRNA induction in rat smooth muscle cells. Cells were stimulated for the indicated times with 100 M UTP before measuring COX-2 and cyclophilin (Cyp) mRNAs by RPA, or COX-2 and GAPDH gene transcription by nuclear run-on followed by hybridization to cDNA slot blots. A, representative RPA phosphorimage of steady-state mRNA. B, quantified data from three independent RPA experiments (mean Ϯ S.E.), expressed as fold over unstimulated levels. COX-2 mRNA signals are normalized by the cyclophilin signal in each sample and expressed as fold induction over basal. C, representative nuclear run-on phosphorimage. D, quantified data from two independent run-on experiments. [ 32 P]UTP-labeled COX-2 nuclear transcripts were normalized by that for GAPDH. Vector, empty pBluescript slots.

FIG. 2. Chimera construction and deletion analysis strategy.
A, depiction of a chimera mRNA expression cassette, with its Antetsuppressible minimal cytomegalovirus promoter (tetOp), luciferase open reading frame, COX-2 mRNA 3Ј-UTR region (filled line) and SV40-derived polyadenylation signals (SV40pAϩ). Approximate locations of the 3Ј-UTR AU-rich elements and restriction sites are indicated. Sequence numbering corresponds to that in GenBank TM accession number AF233596. B, comparison of intrinsic stability for all chimera mRNAs in unstimulated cells, measured after treating with Antet (1 g/ml) for the indicated times. Each point represents the mean Ϯ S.E. from 2 or 3 independent experiments. proximal 1.47-kilobase and middle-late 0.76-kilobase regions of the COX-2 mRNA 3Ј-UTR. These two mRNAs decay only at a slightly faster rate than does the control luciferase mRNA in response to Antet treatment (Table I) but are unaffected by co-treatment with UTP. These observations indicate that the proximal 2.23 kilobases of 3Ј-UTR sequence represented in both KX54 and KX55 mRNAs possess functional instability elements but lack sequences mediating mRNA stabilization in response to P2Y receptor signaling.
The KX56 mRNA incorporates the most distal 263 bases of the COX-2 3Ј-UTR, which includes a 33-base remnant of the poly(A) tail derived from the original cDNA clone. This mRNA also decays more rapidly (t1 ⁄2 ϭ 20 min) than the control luciferase mRNA (Table I) and as rapidly as the full-length 3Ј-UTR chimera mRNA KX51a. This suggests the distal element possesses the dominant instability determinant for the entire 3Ј-UTR. P2Y receptor stimulation induces KX56 mRNA steadystate levels 2.5-fold and also delays the rate of decay evoked by Antet treatment (t1 ⁄2 ϭ 50 min) (Fig. 5, A and B). The data for the KX67 mRNA show that the proximal 100 bases of this distal 263 base distal region cannot account for this particular response, because it lacks both the greater instability and regulated stabilization capacities associated with the KX56 mRNA ( Table I). The 130-base COX-2 3Ј-UTR sequence cloned into KX68 lacks both the 33-base remnant poly(A) ϩ cDNA sequence and also the proximal 100 bases represented in the KX67 mRNA. Steady-state levels of the KX68 mRNA are induced 3.5-fold over basal in response to P2Y receptor activation alone (Fig. 5, C and D). As shown in Table I, the Antet-mediated rate of KX68 mRNA decay (t1 ⁄2 ϭ 21 min) is very rapid and is also delayed by co-treatment with UTP (t1 ⁄2 ϭ 56 min), indicating this response is due to mRNA stabilization.
Attempts to dissect a smaller response element within this 130-base region further by deletion analysis are in progress, but the region appears complex at this time. A series of three overlapping 50-base fragments individually display neither rapid intrinsic decay nor significant P2Y receptor-induced stabilization. Thus, its intrinsic and regulated stability properties may require coordination among multiple epitopes throughout its 130 bases, but further experiments are necessary to clarify these issues.
To gain insight into the signaling pathway responsible for regulating 3Ј-UTR-mediated mRNA stabilization, we tested whether two selective MAP kinase inhibitors attenuate the response using the KX68 mRNA as a reporter of the phenomenon. As shown in Fig. 6A, P2Y receptor stimulation transiently activates the p42/44 MAP kinase for up to 15 min, as assessed using phospho-specific antibodies. In this series of experiments, treatment with UTP for 60 min induced the KX68 mRNA 1.94 Ϯ 0.16 (mean Ϯ S.E.; n ϭ 4)-fold over basal, a response that is inhibited dose-dependently by the MEK1 inhibitor PD09508 (Fig. 6E). The dose effect relationships for inhibition of p42/44 activation and inhibition of mRNA stabilization are similar (Fig. 6B). Although UTP treatment also transiently activates the p38 MAP kinase (Fig. 6C), the selective p38 MAP kinase inhibitor SB202190 has no effect on P2Y receptor-regulated stabilization of the KX68 chimera mRNA (Fig. 6E) at concentrations that inhibit phosphorylation of the p38 MAP kinase substrate ATF-2 (Fig. 6D).
trans-Acting factors may be associated with this phenomenon. To gain insight, UV cross-linking based mRNA/protein binding assays were performed using a [ 32 P]UTP-labeled sense RNA probe representing the most distal 263 base segment of the 3Ј-UTR (bases 4142-4404). Initial exploratory experiments demonstrated relatively weak binding activity in cytosolic extracts compared with strong binding activities present in nuclear extracts, although the protein species in each fraction appeared similar in size (see below). Fig. 7A shows a high resolution separation by SDS-polyacrylamide gel electrophoresis of mRNA binding activity as a function of increasing nuclear extract protein. The most prominent of these mRNA binding proteins migrates at approximately 90 kDa, but several labeled proteins ranging in size between ϳ40 and ϳ75 kDa were also detected.

TABLE I Kinetic parameters of luciferase/COX-2 3Ј-UTR chimera mRNA decay
Decay rate constants (k) were derived by combining data points from each replicate experiment (n) and performing a single nonlinear regression analysis using a one-site exponential decay function. Half-lives (t 1/2 ) were calculated from the relationship t 1/2 ϭ 0.693/k. A two-tailed t test analyzed the goodness of fit and was used to evaluate whether decay rate constants for a given mRNA differed between the two treatments. Whether UTP treatment alone stabilized the chimera is indicated as is the fold induction. The experiment shown in Fig. 7B indicates this binding activity is specific, because it is inhibited by an unlabeled competitor composed of the sense mRNA strand from bases 4218 -4404 but not by an antisense mRNA prepared from the same region. Fig. 7C shows results from a binding assay that incorporates several controls and also establishes the substantial differences between binding activities in nuclear and cytoplasmic extracts. Binding to the 4142-4404-base probe, as alluded to above, is substantially greater in nuclear (NUC) than in cytosolic (S100) extracts. Furthermore, treatment of the cells with UTP for 30 min has no discernible affect on the binding in either fraction (compare lanes 4 and 5 and lanes 6  and 7). Similarly, 10-and 60-min treatments were without effect on binding (data not shown). We also compared the binding to the mRNA stabilizer region (4142-4404-base sequence of the COX-2 3Ј-UTR) against that of a similarly sized proximal AU-rich COX-2 mRNA 3Ј-UTR segment composed of bases 1922-2173. This is from a region that does not appear to confer the regulated stabilization function (see data on mRNA KX54) and includes the 6 times AU 3 A sequences depicted in Fig. 2, which lies just downstream of the COX-2 open reading frame. The binding to this proximal probe (Fig. 7C, lanes 8 -14) is substantially weaker compared with that of the stabilizer region (Fig. 7C, lanes 1-7). Other controls showing no significant binding include the use of labeled antisense probes (lanes 3 and 10) and complete competition with a 100-fold excess of unlabeled in vitro transcribed sense cognate transcripts (lanes 2 and 9). Shown in Fig. 7D are data indicating the 186-base distal region (bases 4218 -4404) has a binding capacity similar to that of the larger 256-base probe, including preferential interactions with nuclear proteins and no affect by UTP stimulation.

DISCUSSION
Robust but transient COX-2 mRNA induction by G␣ q -coupled receptor signaling provides a model to study the dynamic integration of coordinated and opposing gene expression control mechanisms regulated by this important and large family of cell surface proteins. Rapid loss of the signal after the peak response at 60 min suggests that the P2Y receptor-stimulated inductive mechanisms are transient and also that the COX-2 mRNA is highly unstable (see Fig. 1). However, we observe relatively weak (ϳ3-fold) induction of COX-2 gene transcription as measured both by nuclear run-on and promoter/reporter 2 approaches in this preparation, similar to findings in other cells (2,5,11,(15)(16)(17)(18). What accounts for maximal 100-fold mRNA induction when a highly unstable mRNA is comparatively induced weakly through transcription? Identification of a regulatory element in the COX-2 mRNA that mediates transcript stabilization in response to P2Y receptor signaling represents a significant step toward resolving this paradox. This post-transcriptional response is detectable as early as 20 min after receptor stimulation, and as such represents a component of the immediate-early response. When tetracycline transactivator-mediated expression is permitted, our data show that this element mediates a rapid readjustment to a 2-3-fold higher steady state in response to UTP, which is sustained for as long as 2 h after receptor activation. Even so, this stabilization process cannot overwhelm the inherent strength of intrinsic mRNA decay processes, because stabilized chimera mRNAs still decay, albeit more slowly, when transcription is suppressed simultaneously with UTP signaling.
This latter point emphasizes that COX-2 gene transcription is necessary for manifestation of the post-transcriptional response. We therefore speculate that the peak immediate-early COX-2 mRNA accumulation represents synergistic integration of simultaneously activated mRNA stabilization and transcriptional induction mechanisms. The chimera strategy here shows ϳ3-fold mRNA stabilization, whereas transcriptional induction is maximally ϳ3-fold over basal levels. The more robust ϳ100fold induction of the COX-2 mRNA most likely reflects an exponential, rather than additive or multiplier, function of these two processes. However, unresolved in this study is whether the distal 3Ј-UTR mRNA regulatory element fully explains all aspects of post-transcriptional COX-2 mRNA regulation by G␣ q -coupled receptor signaling. Possibly, the distal 3Ј-UTR-mediated mechanism functions more efficiently in context with the remainder of the COX-2 mRNA sequence instead of the heterologous luciferase sequence. Accordingly, we have not ruled out the existence of additional regulated stability determinants within the COX-2 coding region or 5Ј-UTR. Given the large size of the 3Ј-UTR, we felt it was best to use luciferase chimeras as a first approach, which avoid potential complications associated with expressing recombinant COX-2 enzyme activity. This could produce higher levels of any of several autocrine factors and potentially complicate the otherwise selective control of the signaling stimulus achieved in this experimental design. Future studies can build upon the present observations to address potential roles for upstream sequences in post-transcriptional regulation of the COX-2 mRNA. At some point, incorporating COX-2 promoter regulatory sequences upstream of recombinant COX-2 mRNAs in which stability elements are carefully mapped may provide a direct experimental test for the notion that post-transcriptional and transcriptional regulation function synergistically.
Intrinsic destabilization of the luciferase mRNA by the AUrich COX-2 mRNA 3Ј-UTR is not unexpected given long standing observations about the relationship between AU-rich sequences and rapidly decaying mRNAs (51). The most unstable chimeric mRNAs studied have the most distal 130 bases of the COX-2 mRNA in common. Both the KX54 and KX55 mRNAs are slightly more unstable than is the luciferase mRNA, but they possess weaker instability determinants than the distal region. Unlike the latter, these are not influenced by P2Y receptor signaling, suggesting not all instability elements in the COX-2 mRNA are equivalent and that regulated stabilization only affects a subset of them. Our experiments do not establish whether both the weak and the strong instability determinants reflect AU-rich mediated decay. Although this seems quite possible, it is important to note that functional AU-independent mRNA destabilization motifs have been described even in otherwise AU-rich environments (52). This speculation, however, is consistent with the fact that AU-rich mediated mRNA decay itself does not reflect a simple monotypic process because various sequences can be functionally distinguished by, among other parameters, differential decay characteristics in vivo (53)(54)(55). The consensus strong instabil-ity element established in these reports is the motif UAUUUAU. Interestingly, the less unstable proximal region of the COX-2 mRNA 3Ј-UTR contains a 3-fold overlap of this UAUUUAU consensus, whereas the motif is absent in the more unstable distal regulatory region. Conceivably, G␣ q -coupled receptor signaling only modulates subsets of AU-rich instability element mediated decay. Recent reports support the plausibility of this notion insofar as cellular stresses can affect the distribution of two major groups of AU-rich element mRNA-binding proteins, the AUF and ELAV-related Hu factors (56 -58).
G␣ q -coupled receptor stabilization of the distal COX-2 regulatory element is mediated by the MEK1/p42/44 MAP kinase signaling pathway. Although it is formally possible that its sensitivity to PD098059 reflects inhibition of other kinase activities, the drug is purported to be fairly selective (59)  regulated mRNAs for which evidence of post-transcriptional modulation is reasonably direct, other regulatory kinases have also been implicated including c-Jun NH 2 -terminal kinase/ stress-activated protein kinase (60, 61), p38 MAP kinases (38), and the cAMP-dependent protein kinase A (36,40). The emerging picture from this limited data suggests effectors of multiple signaling enzymes exist within the mRNA decay pathways, which can connect changes in the extracellular environment to post-transcriptional modulation of gene expression control. The analogy to complexities associated with transcriptional regulation is increasingly striking.
Stabilization of the human COX-2 mRNA in response to cytokine signaling has been reported to involve the p38 MAP kinase (20,22,23). Current evidence suggests its 3Ј-UTR is somehow involved in this, although specific mRNA regulatory regions have yet to be clarified in detail (19,21,62). We see no evidence that p38 MAP kinase activity participates in the process uncovered here (Fig. 6). Methodological differences may account for this discrepancy, as all studies of the human mRNA rely on a fairly prolonged stimulus period before measuring mRNA decay by actinomycin D chase approaches, and some cellular compensation may be involved. In contrast, our focus is on a process activated within the first several minutes of G␣ qcoupled receptor stimulation. Species differences may also be important, since the distal 130 bases of rat and mouse COX-2 mRNA 3Ј-UTR are highly conserved but align poorly with the human mRNA sequence. A recent study of the rat COX-2 mRNA has shown that the proximal AU-rich 3Ј-UTR can also function as a regulated stability control element in cells in-duced to express a ras oncogene and stimulated with transforming growth factor-␤1 (63). Notably, this stabilization effect is manifested most prominently several hours after these treatments, perhaps due to longer term cellular compensation than occur in the immediate-early regulatory phase. Whether this represents a distinct regulatory process from that uncovered in the present study will require further study.
We see no overt evidence that P2Y receptor signaling stimulates or inhibits occupancy of the distal regulatory region by mRNA-binding proteins, but we cannot exclude the possibility that minor, less readily detected components bind differentially as a consequence of signaling pathway activation. Nevertheless, this binding activity to the distal 3Ј-UTR element appears to be fairly specific based on several criteria including its distinction from that to a similarly sized probe derived from an unregulated proximal region of the 3Ј-UTR, which includes 3-fold overlapping UAUUUAU motif. Furthermore, antisense probes from the distal regulatory region do not bind proteins in the extracts very well either. Finally, the binding activity correlates well with the functional stabilization activity, in which both require the distal 130 bases as the so far defined minimal elements.
The lower molecular weight cross-linked species are consistent with the sizes for members of the AUF and Hu protein families, which range in size between 32 and 45 kDa. To our knowledge the strongly labeled ϳ90-kDa factor is not consistent with any known AU-binding proteins, yet the prominent protein migrating around 75 kDa is similar to a 70-kDa AUbinding factor observed in human fibroblast extracts (64). Sub- FIG. 7. Specific binding of nuclear-enriched proteins to the distal 3-UTR stabilizer region. Each image is a representative 4-h exposure autoradiogram from ultraviolet cross-linking binding experiments using in vitro transcribed [ 32 P]UTP-labeled probes and is repeated independently 2-4 times. Except in A, the amount of extract protein used in each assay was 15 g. After transcription, all probes were purified on sequencing gels before use. A, increased cross-linked protein as a function of increasing nuclear extract resolved by SDS-12% polyacrylamide gel electrophoresis. B, binding to the radiolabeled distal 3Ј-UTR 263-base stabilizer region is specifically inhibited by an unlabeled sense but not antisense competitor probe prepared from the most distal 186 bases (4218 -4404) of the 3Ј-UTR. The molar excesses of competitors over labeled probe are 1-, 3-, 10-, and 30-fold. C, binding capacity of the distal 263-base stabilizer region (4142-4404, lanes 1-7) is greater than a probe prepared from a similarly sized, unregulated AU region  in the proximal 3Ј-UTR that lacks stabilizer function (lanes 8 -14). Negative controls include no protein (⌽; lanes 1 and 8), competition in nuclear extracts by 100-fold excess cognate unlabeled RNA (lanes 2 and 9), and use of antisense [ 32 P]UTP-labeled probes (lanes 3 and 10). Binding to the stabilizer region probe is greater in nuclear (lanes 6 and 7) than cytosolic (lanes 4 and 5) extracts, but the species appear similar. A 30-min stimulation prior to extraction of proteins had no effect on the level of binding (lanes 5 and 7) compared with untreated cells (lanes 4 and 6). D, a distal 186-base probe (4218 -4404) shows the same binding as the 263-base probe using nuclear extracts and binds the same species. Negative controls are similar to those in B. stantial work will be necessary to refine a structure-activity relationship between the various components of this mRNA-binding protein activity and functional receptor-regulated mRNA stabilization activity. It will also be of interest to understand why the binding activity is highly enriched in nuclei as opposed to cytosol and whether this bears any insight into the mechanism involved in regulated mRNA stabilization.
In summary, the findings here add the COX-2 mRNA to a relatively small group of mRNAs for which post-transcriptional regulation by cell surface receptor signaling is supported by rather unambiguous and direct evidence (36 -40). Although the mechanisms controlling COX-2 gene expression likely vary in stimulus and cell type-dependent contexts, our study provides substantial evidence supporting the notion that transcriptional induction alone is unlikely to account for the full repertoire of responses displayed by this ubiquitously inducible transcript. This may also prove to be the case for many of the related immediate-early response genes.