Post-transcriptional Regulation of Meprin α by the RNA-binding Proteins Hu Antigen R (HuR) and Tristetraprolin (TTP)*

Background: Meprin α is a metalloproteinase implicated in the pathogenesis of inflammatory bowel disease (IBD). Results: RNA-binding proteins post-transcriptionally regulate meprin α, and RNA stability decreases after induction of an inflammatory response. Conclusion: Inflammatory stimuli downregulate meprin α by inducing TTP expression and binding to the transcript. Significance: Determining how inflammation alters meprin α regulation is crucial to understanding its role in IBD. Meprins are multimeric proteases that are implicated in inflammatory bowel disease by both genetic association studies and functional studies in knock-out mice. Patients with inflammatory bowel disease show decreased colonic expression of meprin α, although regulation of expression, particularly under inflammatory stimuli, has not been studied. The studies herein demonstrate that the human meprin α transcript is bound and stabilized by Hu antigen R at baseline, and that treatment with the inflammatory stimulus phorbol 12-myristate 13-acetate downregulates meprin α expression by inducing tristetraprolin. The enhanced binding of tristetraprolin to the MEP1A 3′-UTR results in destabilization of the transcript and occurs at a discrete site from Hu antigen R. This is the first report to describe a mechanism for post-transcriptional regulation of meprin α and will help clarify the role of meprins in the inflammatory response and disease.

Meprins are zinc metalloproteases that are members of the astacin family of proteinases (1,2). The two subunits, encoded by the human genes MEP1A (meprin ␣; chromosome 6p) and MEP1B (meprin ␤; chromosome 18q), are evolutionarily related and highly conserved at the amino acid level (2). Both subunits have similar domain structure and organization (1), with the exception that meprin ␣ has an inserted domain with a proteolytic cleavage site that results in secretion from the cell, whereas meprin ␤ remains bound to the plasma membrane (3,4). The ␤ subunit can form disulfide-linked dimers at the cell surface that are referred to as meprin B. The ␣ subunit also forms disulfide-linked homodimers (homomeric meprin A), and can further oligomerize to form large species. In addition, the ␣ and ␤ subunits can associate to form heterotetramers that are linked to the cell surface (heteromeric meprin A).
Meprins are highly expressed in the brush border membrane of the kidney and intestinal epithelium (5,6), and have also been identified in other tissue and cell types, such as liver, epidermis, and leukocytes (7)(8)(9). Their predominant location at sites where there is a barrier between the host and the outside and expression in leukocytes are suggestive of a function in host defense and the immune response (8,10). Meprins are capable of cleaving a wide variety of substrates, including gastrointestinal peptides, extracellular matrix proteins, cell junction molecules, cytokines, and chemokines (1,11,12). Thus, meprins have the potential to modulate the immune response by generation of either active or inactive protein fragments (1,13). Dysregulation and abnormal meprin expression have been implicated in urinary tract infection, kidney disease, IBD, 2 and cancer (8,10,14,15).
IBD is a group of illnesses characterized by aberrant inflammation within the gastrointestinal tract, and several lines of evidence suggest a role for meprin ␣ in the disease. Studies with meprin knock-out (KO) mice indicate that meprin ␣ expression in the intestine may be protective, as both ␣KO and ␣␤KO mice developed a more severe dextran sulfate sodium-induced colitis compared with wild-type (WT) or ␤KO mice (15,16). In addition, there is a strong correlation between decreased MEP1A mRNA expression in inflamed mucosa of patients with IBD relative to unaffected tissue from all groups (16). There is strong evidence for a genetic component in IBD (17), and genomewide associations have consistently identified chromosome 6p as an IBD susceptibility locus (18 -20). Two recent reports have established that the MEP1A gene is associated with IBD (15,16). One of these studies found a significant association between a single nucleotide polymorphism (SNP) in the 3Ј-UTR and IBD (16), raising the question of whether posttranscriptional regulation of meprin ␣ might play a role in IBD.
Post-transcriptional regulation (PTR) is recognized as an important means by which cells can rapidly adjust to changing conditions, as well as an additional level of control for genes whose expression must be transient (21,22). This form of regulation is known to be effected through RNA-binding proteins (RBPs) and regulatory sequences in the 3Ј-UTR of target mRNA broadly categorized as AU-rich elements (AREs) (23,24). The consequences of PTR may be a change in RNA shuttling, stability, and/or translation efficiency depending on the protein(s) bound (23,25). Two ubiquitously expressed RBPs are TTP, which is known to destabilize transcripts directly by promoting their deadenylation (26), and ELAVL1/HuR (27,28), which often functions to counteract the actions of destabilizing proteins by promoting adenylation, competing with other RBPs for binding sites, or enhancing translation (29 -31). Both HuR and TTP have been demonstrated to regulate a number of cytokines and chemokines, and post-transcriptional regulation is recognized to be important in the inflammatory response (32,33). Mice with a deletion of the tumor necrosis factor (TNF) ARE develop both chronic inflammatory arthritis and IBD (34), and TTP knockout mice develop severe inflammatory arthritis and autoimmunity (35). Thus, dysregulation of ARE-binding proteins has implications for multiple disease processes, including IBD and cancer (21,36,37). The studies herein were conducted to determine whether expression of meprin ␣ is subject to PTR and, using a tissue culture model of IBD, whether this regulation is altered in response to inflammatory stimuli.
Cloning of Luciferase Reporters and TTP Expression Construct-The luciferase reporter vector pmirGLO was purchased from Promega for expression of both firefly and Renilla luciferase from the same plasmid under the control of separate promoters. This reporter was used as both a reference control in transfection experiments (described below) and as the backbone vector for the hMEP1A 3Ј-UTR construct. The vector pSGG-MEP1A-3UTR (SwitchGear Genomics) was used as template for amplification of the target sequence by Phusion DNA polymerase (Finnzymes Oy) and subcloning into the pmirGLO vector. The following primers were purchased from Integrated DNA Technologies and used to introduce restriction endonuclease sites for DraI and XbaI (bases changed/ added in lowercase; recognition sites underlined): 5Ј-aaatttA-AAGGCCAAGGAAGTGACCTG-3Ј and 5Ј-aatctagAAGG-TGCTAACTCAAT-3Ј. Plasmid DNA was isolated from multiple colonies using the Zyppy Plasmid Miniprep kit (Zymo Research) for screening via restriction analysis, and all positive clones were confirmed by sequencing at the Penn State University Nucleic Acid Facility. The final construct, pmG-hMEP1A-3UTR-WT, consisted of the pmirGLO vector with the complete human MEP1A 3Ј-UTR sequence (GenBank accession number NM_005588.2) cloned downstream of the firefly luciferase coding region.
To generate the TTP expression construct, Caco-2 cDNA was used as template for amplification of the TTP coding sequence by Phusion DNA polymerase (Finnzymes Oy) and subcloning into the mammalian expression vector pcDNA3.1/V5-His (Invitrogen). The following primers were purchased from Integrated DNA Technologies to first amplify TTP cDNA: 5Ј-acttcagcgctcccactctc-3Ј and 5Ј-cttgattactgtcccccaagtc-3Ј. A second set of primers was then used to introduce restriction endonuclease sites for KpnI and SacII (bases changed/added in lowercase; recognition sites underlined): 5Ј-aaaaATGGggtACCGTTACACC-3Ј and 5Ј-aaaaccgCGGGC-AGTCACTTTG-3Ј. Plasmid DNA was isolated from multiple colonies using the Zyppy Plasmid Miniprep kit (Zymo Research) for screening via restriction analysis, and all positive clones were confirmed by sequencing at the Penn State University Nucleic Acid Facility. The final construct, pcDNA-hTTP, includes the ORF from GenBank accession number NM_003407.2 with an intact stop codon such that the V5-His tag is not translated.
Transient Transfections-Silencing experiments were conducted with TTP, HuR, or negative control shRNA constructs obtained from Genecopoeia. For each target, three shRNA constructs were tested. Cells were grown to 70 -80% confluence in 6-well plates, and 2 g of each plasmid was transfected using Lipofectamine 2000 (Invitrogen) as described by the manufacturer. Medium was changed after 16 h, and cells were harvested at 48 h post-transfection for isolation of both RNA and protein.
Efficacy of silencing was assessed by Western blot for either HuR or TTP. The shRNA construct that gave the greatest silencing was used in subsequent experiments.
For reporter mRNA stability assays, A549 cells were grown to 70% confluence in 6-well plates and cotransfected by Attractene reagent with 50 ng of either pmirGLO or pmG-hMEP1A-3UTR-WT and 200 ng of either negative control shRNA, HuR shRNA2, pUC19, or TTP expression plasmid where appropriate. Actinomycin D assays were performed 24 h post-transfection, and half-lives were calculated from the decay curves generated.
RNA Isolation, cDNA Synthesis, and Quantitative Real Time PCR-Total RNA was isolated using TRIzol reagent (Invitrogen) as described by the manufacturer and quantitated by UV absorbance at 260 nm. Within an experiment, equivalent amounts of RNA were reverse transcribed using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems) according to the manufacturer's instructions. Quantitative real time PCR was performed using template cDNA, 2ϫ SensiMix SYBR Master Mix with fluorescein (Bioline), and the gene-specific primers in Table 1. All reactions were run in 96-well plates on the MyiQ2 Two-Color Real-Time Detection System (Bio-Rad) using the following program: 2-min hot start at 95°C, 40 cycles of 15 s at 95°C and 60 s at 60°C, and 2 min at 10°C. Data were analyzed by the ⌬⌬Ct method with normalization to the endogenous control, GAPDH.
Biotin Pulldown Experiments-To generate the biotinylated RNA oligos for biotin pulldown experiments, cDNA templates were first amplified using the primers shown in Table 2 to add a T7 site for RNA polymerase. Templates were then in vitro transcribed to RNA using the MAXIscript T7 kit and a ratio of 4:1 CTP:biotin-CTP, followed by digestion with DNase I and reaction cleanup using NucAway spin columns as described by the manufacturer (both from Ambion). RNA integrity was confirmed via agarose gel electrophoresis, and concentrations were determined via spectrophotometry. Biotin affinity pulldown experiments were performed as described previously (38,39).
Immunoblotting-Cell pellets were lysed in 1ϫ PLB (100 mM KCl, 5 mM MgCl 2 , 10 mM HEPES, pH 7.0, 0.5% Nonidet P-40) for 10 min on ice in the presence of EDTA-free protease inhibitors (Roche Applied Science), then centrifuged for 5 min at maximum speed (4°C) to clear cellular debris. Samples were prepared for Western blotting by boiling for 5 min in SDS-PAGE sample buffer and subjected to denaturing polyacrylamide gel electrophoresis using precast 4 -20% Ready Gels (Bio-Rad) for ϳ45 min at 200 V or 10% SDS-polyacrylamide gels for 45 min at 45 mA. The following antibodies were used at a 1:1000 dilution: HuR (sc-5261; Santa Cruz Biotechnology), TTP (T5327; Sigma), ␤-actin (A5441; Sigma), donkey anti-rabbit IgG (NA934; GE Healthcare), and donkey anti-mouse IgG (sc-2314; Santa Cruz Biotechnology). For detection of secreted meprin A, 1 ml of medium was collected from Caco-2 cells stimulated with PMA (50 ng/ml) or control for 24 h. Samples were dialyzed into a 0.01% SDS solution using a 100-kDa-molecular mass-cutoff cellulose ester membrane (Spectrum Laboratories) and concentrated 100-fold by vacuum drying (Speed-Vac concentrator, Savant). The SDS solution was necessary to prevent protein from precipitating during concentration. After addition of SDS-PAGE loading buffer containing 100 mM DTT, the sample was separated by SDS-PAGE using an 8% acrylamide gel and subjected to Western blot as described previously (40). Blots were scanned and analyzed by ImageJ for densitometric analysis of relative protein expression as indicated.
Ribonucleoprotein Immunoprecipitation-For ribonucleoprotein immunoprecipitation (RNP-IP), cell pellets were lysed in 1ϫ PLB as above with the addition of both protease inhibitors (Roche Applied Science) and RNaseOUT (Invitrogen). Antibody-bead complexes were formed by combining 75 l of DynaBeads Protein G (Invitrogen) and 10 g of either normal mouse IgG (sc-2025; Santa Cruz Biotechnology), HuR antibody (sc-5261; Santa Cruz Biotechnology), normal rabbit IgG (sc-2027; Santa Cruz Biotechnology), or TTP antibody (T5327; Sigma) in PBST (PBS with 0.05% Tween 20) for 30 min at room temperature with gentle mixing. Beads were washed once with PBST and combined with ϳ400 g of cell lysate with RNase-OUT in a total volume of 500 l per IP. Lysates were incubated with the antibody-bead complexes for 30 min at room temperature with mixing, then beads were washed three times with PBST. An aliquot of the beads from each IP was saved from the third wash step for Western blot analysis to confirm that the target protein was successfully immunoprecipitated. The remaining beads from each IP were subjected to digestion with DNase I and proteinase K, followed by phenol-chloroform extraction and RNA precipitation in the presence of Glycoblue co-precipitant (Ambion). The entire volume of RNA was then used for cDNA synthesis and quantitative real time PCR (qRT-PCR) as described above. In parallel, cells not lysed for the RNP-IP were used for total RNA isolation, reverse transcription, and qRT-PCR to verify that the RNA transcripts of interest were present at a detectable level in the starting sample. Fold enrichment was calculated by taking 2 Ϫ⌬Ct where ⌬Ct is the difference in average Ct values between the HuR-IP or TTP-IP and isotype control-IP for a given transcript.
Statistical Analysis-Hypothesis testing was performed primarily using the Student's unpaired two-sample t test (twotailed). Analysis of qRT-PCR data from RNP-IP experiments was done using a paired, one-tailed t test only when comparing the fold enrichment of a transcript in the HuR-IP or TTP-IP versus control-IP, as these were conducted using cell lysate from the same sample.

RESULTS
PMA Regulates MEP1A-MEP1A mRNA is known to be expressed by normal human colonic epithelium (5,6), and expression has previously been demonstrated in the human intestinal epithelial cell line Caco-2 (14). To determine whether meprin ␣ is regulated by inflammatory stimuli, treatment with PMA was used to induce an inflammatory response in Caco-2 cells. PMA has been used as a model of IBD and has been shown  GAPDH 5Ј-GAG TCA ACG GAT TTG GTC GT-3Ј  238  5Ј-TTG ATT TTG GAG GGA TCT CG-3Ј  MEP1A  5Ј-CAG GTG GAC GTT CCC CAT TC-3Ј  167   to induce colitis in both rabbits and rats (41,42). Caco-2 cells were treated for 0, 4, and 16 h with 50 ng/ml PMA or vehicle control. Total RNA was isolated, reverse transcribed to cDNA, and quantified by qRT-PCR with normalization to the housekeeping gene GAPDH, which was unchanged with PMA treatment (data not shown). Levels of CCL2 mRNA were also evaluated as a positive control, as PMA has been demonstrated to upregulate expression of CCL2 (43). As expected, CCL2 mRNA was strongly increased after 4 h PMA treatment (data not shown). In contrast, MEP1A mRNA levels were decreased following both 4 and 16 h treatment with PMA relative to control (Fig. 1A). Protein levels of both meprin ␣ in whole cell lysates (Fig. 1B) and secreted meprin A protein (Fig. 1C) were similarly decreased at 24 h. These results indicate that MEP1A is regulated in response to PMA treatment. As our hypothesis is that the MEP1A 3Ј-UTR contains regulatory elements that impact RNA stability, an Actinomycin D assay was performed to test whether treatment with PMA had an effect on the half-life of the MEP1A transcript. Actinomycin D inhibits global transcription (44), thus allowing the half-life of existing transcripts to be determined. Based on the finding that MEP1A mRNA levels are decreased with 4 h PMA treatment, cells were pretreated with either 50 ng/ml PMA or vehicle for 4 h, then changed to medium containing 5 g/ml Actinomycin D and harvested for RNA isolation at 0, 1, 2, and 3 h posttreatment. Analysis was performed by qRT-PCR and normalized to GAPDH, which was unchanged through 3 h of Actinomycin D treatment (data not shown). The percent transcript remaining was calculated for the control (ϪPMA) and treated (ϩPMA) samples with respect to 0 h Actinomycin D treatment. Pretreatment with PMA resulted in a significant decrease in the percent transcript remaining at 3 h of Actinomycin D treatment (Fig. 1D) and a shorter estimated half-life relative to the untreated control at 3.5 Ϯ 1.3 versus 17.4 Ϯ 5.6 h, respectively (Fig. 1E). Taken together, these results provide evidence that MEP1A is regulated post-transcriptionally in response to PMA, and that PMA has a destabilizing effect on the transcript.
HuR Binds to the MEP1A 3Ј-UTR and Increases Baseline Transcript Stability-As PMA treatment downregulated meprin post-transcriptionally, we sought to determine whether this effect was mediated by RBPs. A variety of RBPs, including HuR, can associate with 3Ј-UTRs of target genes to regulate mRNA stability. The 3Ј-UTR of MEP1A contains a class I ARE with two AUUUA motifs, sequences known to be sites of HuR binding (24,28,45), and is generally AU-rich. Therefore, the ability of HuR to bind the MEP1A 3Ј-UTR compared with CCL5, a negative control shown not to bind HuR (39), was evaluated by biotin pulldown experiments. The 3Ј-UTR of MEP1A and CCL5 were synthesized and biotinylated in vitro and incubated with untreated Caco-2 cell lysate. Following pulldown by streptavidin-coated beads and several washes, bound proteins were eluted and detected by Western blot. As seen in Fig. 2A, HuR protein was detected in both the cell lysate and the MEP1A 3Ј-UTR pulldown, but did not bind to the CCL5 3Ј-UTR. These results indicate that HuR is capable of binding to the MEP1A 3Ј-UTR in vitro. To determine whether HuR binds to the endogenous mRNA, a RNP-IP assay for HuR was performed. Ribonucleoprotein complexes were immunoprecipitated from cell lysates using either an antibody to HuR or normal mouse IgG as an isotype control. An aliquot of each IP prior to proteinase K digestion was also analyzed by Western blot to confirm that the target protein, HuR, was successfully bound (Fig. 2B). Following RNP-IP, RNA was isolated, reverse transcribed, and used for qRT-PCR analysis to determine the fold enrichment of GAPDH (negative control) and MEP1A mRNA in the HuR-IP versus control-IP (Fig. 2C). The MEP1A transcript was significantly enriched in the HuR-IP (p Ͻ 0.05), whereas GAPDH was FIGURE 2. HuR binds and stabilizes the MEP1A transcript. A, Caco-2 cell lysate was incubated with biotinylated RNA oligos of the 3Ј-UTR from either CCL5 (negative control) or MEP1A, followed by pulldown of RNA-protein complexes with streptavidin-coated magnetic beads. Washed beads were eluted in SDS-PAGE sample buffer and analyzed by Western blot to detect the presence of HuR in the cell lysate and RNA pulldowns. The data shown are a representative blot from three independent experiments. B and C, RNP-IP was performed using lysate from untreated Caco-2 cells and either HuR antibody or normal mouse IgG. B, representative Western blot of antibody-bound beads from control (IgG)-and HuR-IP showing specific binding of HuR. C, purified RNA samples were reverse transcribed and analyzed by qRT-PCR for the detection of MEP1A and GAPDH transcript levels. Fold enrichment was calculated as described under "Experimental Procedures." The MEP1A transcript was significantly enriched in the HuR-IP versus control-IP (n ϭ 4; *, p Ͻ 0.05). D, Caco-2 cells were transiently transfected with either negative control (Ϫ con) (nonspecific) or one of three HuR-specific shRNA vectors (sh1, sh2, and sh3) for 48 h. Cells were then harvested, lysed, and analyzed by Western blot for levels of HuR and ␤-actin. E, negative control and HuR shRNA vector sh2 were transiently transfected into Caco-2 cells for 48 h, followed by RNA isolation and qRT-PCR analysis. Data were normalized to an endogenous control (GAPDH), and mean relative expression of MEP1A mRNA was calculated with respect to mock transfection control (no vector). Silencing of HuR significantly decreased MEP1A mRNA levels compared with negative control (n ϭ 3; *, p Ͻ 0.05). F, schematic illustrating the firefly luciferase transcripts encoded by pmirGLO and pmG-hMEP1A-3UTR-WT reporter plasmids. G, A549 cells were transiently cotransfected with either the parent or MEP1A 3Ј-UTR reporter plasmids and negative (neg) control or HuR-specific shRNA for 24 h prior to performing Actinomycin D assays. RNA was isolated and used for qRT-PCR analysis of Fluc expression, and the percent transcript remaining was calculated with respect to the 0 h Actinomycin D treatment control (set to 100%; n ϭ 3). The addition of HuR sh2 significantly decreased the stability of the MEP1A 3Ј-UTR fusion transcript relative to negative control shRNA (*, p Ͻ 0.05), whereas the effect on the parent transcript was not significant. All error bars represent the S.E.

Post-transcriptional Regulation of Meprin ␣
FEBRUARY 15, 2013 • VOLUME 288 • NUMBER 7 unchanged. Taken together, these data demonstrate that HuR binds to the MEP1A transcript via the 3Ј-UTR.
To determine the functional consequence of HuR binding to MEP1A mRNA, silencing experiments were performed using Caco-2 cells transiently transfected with negative control (nonspecific) or HuR-specific shRNA plasmids and compared with a mock transfection without shRNA. As shown in Fig. 2D, HuR protein levels were strongly diminished compared with the mock transfection and negative control shRNA (normalized to the loading control ␤-actin), and HuR plasmid sh2 was used for subsequent experiments. Silencing of HuR significantly decreased the steady-state level of MEP1A mRNA (Fig. 2E), indicating that HuR increases baseline expression of MEP1A.
To determine whether HuR exerts its effects by stabilizing MEP1A mRNA, a reporter construct was used to assess the effects of HuR silencing. To test the hypothesis that the 3Ј-UTR of the MEP1A transcript contains regulatory elements, the entire 3Ј-UTR was cloned into the luciferase reporter vector pmirGLO. The resulting construct encodes a fusion transcript that places the MEP1A 3Ј-UTR downstream of the firefly luciferase coding region (Fig. 2F), allowing for testing of the effect of the 3Ј-UTR on luciferase expression. A549 cells were cotransfected with either pmirGLO (parent vector) or pmG-hMEP1A-3UTR-WT and negative control shRNA or HuR sh2 plasmid for 24 h followed by 0, 3, or 5 h treatment with 5 g/ml Actinomycin D. RNA was then isolated and used for qRT-PCR analysis of Fluc (luc2) expression. The percent transcript remaining was calculated for each combination with respect to 0 h Actinomycin D treatment. Neither reporter was found to be significantly affected by the addition of the negative control shRNA plasmid (data not shown). Both the parent Fluc and the MEP1A 3Ј-UTR fusion transcript were stable in the presence of the negative control shRNA (Fig. 2G, open symbols) with calculated halflives of 31.7 Ϯ 9.8 and 28.8 Ϯ 3.8 h, respectively. In contrast, silencing of HuR (Fig. 2G, closed symbols) led to a significant decrease in stability of the fusion transcript (5.4 Ϯ 1.6 h) but not the parent Fluc transcript (26.5 Ϯ 7.3 h), indicating that HuR stabilizes the MEP1A transcript via the 3Ј-UTR.
PMA Does Not Alter HuR Levels or Binding to MEP1A-To determine whether HuR is responsible for the PMA-induced destabilization of MEP1A mRNA, Caco-2 cells were treated with or without PMA for 4 h prior to isolation of RNA and protein. Although MEP1A mRNA levels decreased with treatment as before, PMA had no effect on the amount of either HuR mRNA (Fig. 3A) or protein (Fig. 3B).
As PMA was not found to change HuR abundance, the possibility that treatment alters the ability of HuR to bind to MEP1A mRNA was tested. Lysate from Caco-2 cells treated for 4 h with or without PMA was used for RNP-IP as described above, and association of MEP1A mRNA was assessed by qRT-PCR. The successful IP of HuR was confirmed by Western blot (Fig. 3C). Fold enrichment of each transcript in the HuR-IP versus isotype control-IP was calculated separately for the two treatment groups (Fig. 3D). Whereas the MEP1A transcript was again enriched in the HuR-IP, there was no significant difference between treatment groups (mean fold enrichment of 5.07 versus 3.61 for ϪPMA and ϩPMA, respectively; p ϭ 0.53). A biotin pulldown experiment was subsequently performed using lysates from cells treated with or without PMA for 4 h, and no difference in HuR binding was noted (Fig. 3E). Taken together, the results indicate that the effects of PMA on MEP1A are not due to a change in HuR expression or binding.
PMA Upregulates TTP and Destabilizes MEP1A-As the effects of PMA did not appear to be mediated by HuR, we investigated the possibility that a different RNA-binding protein, TTP, might be involved in this process. Following 4 h treatment of Caco-2 cells, PMA was found to increase expression of both TTP mRNA (Fig. 4A) and protein (Fig. 4B) relative to control. The ability of TTP to bind the endogenous MEP1A transcript was tested via RNP-IP assay using lysate from cells treated for 4 h with or without PMA and antibodies to either TTP or normal rabbit IgG control. Successful immunoprecipitation of TTP protein was confirmed by Western analysis (Fig. 4C), and qRT-PCR analysis showed that with PMA treatment the MEP1A transcript was significantly enriched (p ϭ 0.04) in the TTP-IP compared with no treatment and with the isotype control-IP (Fig. 4D). To further confirm this finding, biotin pulldown experiments were performed using the labeled MEP1A 3Ј-UTR and lysate from cells treated for 4 h with or without PMA. Consistent with the RNP-IP data, TTP binding to the 3Ј-UTR was strongly enhanced by PMA (Fig. 4E).
The observation that binding of HuR to the MEP1A 3Ј-UTR does not appear to be affected by treatment whereas TTP binding is primarily PMA-dependent suggests that HuR and TTP are not in competition for the same binding site. To test this hypothesis, biotinylated RNA oligonucleotides were generated for three smaller segments of the MEP1A 3Ј-UTR (mRNA positions 2286 -2515, 2485-2647, and 2620 -2927; Fig. 5A) and used for pulldown assays with lysate from cells treated for 4 h with PMA. As shown in Fig. 5B, TTP binding was mainly localized to segment A, whereas HuR binding was restricted to segment B. These results demonstrate that the MEP1A 3Ј-UTR can be bound by HuR and TTP concurrently rather than in a mutually exclusive fashion.   To determine whether binding of TTP to the MEP1A 3Ј-UTR had a functional effect on RNA stability, the luciferase reporter constructs were again utilized. As basal expression of TTP is low in cells, a TTP expression construct was first tested for transient overexpression of the protein in A549 cells (Fig.  6A). A549 cells were then transfected with either pmirGLO (parent vector) or pmG-hMEP1A-3UTR-WT in the presence or absence of the TTP expression plasmid for 24 h followed by 0, 3, or 5 h treatment with 5 g/ml Actinomycin D and qRT-PCR analysis of Fluc expression (Fig. 6B). Although the stability of the parent Fluc transcript did not change significantly with TTP overexpression (21.5 Ϯ 3.2 versus 42.6 Ϯ 21.9 h, p ϭ 0.17), a significant reduction in the stability of the MEP1A 3Ј-UTR fusion transcript was observed with TTP cotransfection (25.3 Ϯ 4.6 versus 5.8 Ϯ 2.3 h, p ϭ 0.019). These results suggest that TTP decreases mRNA stability by acting on the MEP1A 3Ј-UTR.
To determine whether TTP is required for downregulation of the endogenous MEP1A transcript, Caco-2 cells were tran-siently transfected with either negative control or TTP-specific shRNA constructs for 48 h and treated with or without PMA for 4 h prior to isolation of RNA and protein. Transfection of the TTP-specific sh1 construct significantly diminished TTP protein to 39 Ϯ 8% (p ϭ 0.017) of the level observed with treatment of the negative control shRNA, as determined by densitometric analysis of TTP Western blot (Fig. 6C). TTP silencing effectively prevented the PMA-induced decrease in MEP1A mRNA levels (Fig. 6D). Taken together, the results from these studies provide strong evidence that the PMA-induced destabilization of MEP1A is mediated through upregulation of TTP and subsequent binding to the 3Ј-UTR.

DISCUSSION
Due to the nature of the reactions catalyzed by meprin metalloproteases, their activity is highly regulated. Meprin ␣ is known to undergo several types of post-translational processing that modify localization and/or activity of the protein (46 -49), and MEP1A transcription is specifically induced during FIGURE 6. TTP destabilizes MEP1A mRNA. A, representative Western blot showing overexpression of TTP protein by pcDNA-hTTP in transiently transfected A549 cells compared with the loading control, ␤-actin. B, A549 cells were transiently cotransfected with either the parent or MEP1A 3Ј-UTR reporter plasmids with or without pcDNA-hTTP expression plasmid for 24 h prior to performing Actinomycin D assays. RNA was isolated and used for qRT-PCR analysis of Fluc expression, and the percent transcript remaining was calculated with respect to the 0 h Actinomycin D treatment control (set to 100%; n ϭ 3). Overexpression of TTP significantly decreased the stability of the MEP1A 3Ј-UTR fusion transcript (*, p Ͻ 0.05), whereas the effect on the parent transcript was not significant. C, Caco-2 cells were transiently transfected with either negative control (Ϫ con) (nonspecific) or one of two TTP-specific shRNA vectors (sh1 and sh2) for 48 h and then treated with or without PMA. Silencing of TTP was confirmed by Western blot using ␤-actin as a loading control. D, negative control and TTP shRNA vector sh1 were transiently transfected into Caco-2 cells for 48 h, then cells were treated for 4 h with PMA or DMSO (vehicle control) followed by RNA isolation and qRT-PCR analysis. Data were normalized to an endogenous control (GAPDH), and mean relative expression of MEP1A mRNA was calculated with respect to control (no treatment). Silencing of TTP significantly increased MEP1A mRNA levels with PMA compared with negative control and no vector control (*, p Ͻ 0.05; n ϭ 3). All error bars represent the S.E.
intestinal cell differentiation (50 -52). In contrast, there is little information about mRNA regulation, and this may be an important means by which protein levels are controlled. These studies demonstrate that the MEP1A transcript is post-transcriptionally regulated by the RNA-binding proteins HuR and TTP. In particular, we show that PMA destabilizes the MEP1A transcript by inducing TTP expression and binding to the 3Ј-UTR.
Metalloproteases play crucial roles in inflammatory responses. For instance, matrix metalloproteinases are usually proinflammatory in nature, and tend to be expressed at low levels under basal conditions and induced by inflammatory stimuli (53). Meprins, on the other hand, are normally well expressed in the gastrointestinal tract, but are downregulated with stress, injury, and inflammatory conditions. For example, meprin ␣ mRNA levels are significantly decreased in the inflamed mucosa of IBD patients (16) and in leukocytes from mesenteric lymph nodes in a mouse model of IBD (9). Our study findings are consistent with these results, and provide a potential mechanism by which expression is decreased with inflammation via post-transcriptional mechanisms.
These observations suggest that metalloproteases of the same superfamily (metzincins) have very different physiological roles and effects on pathology. In particular, homomeric meprin A is proposed to have an anti-inflammatory role in the gastrointestinal epithelium. It has been shown that ␣KO mice develop more severe colitis when challenged with dextran sulfate sodium than WT, ␤KO, or ␣␤KO littermates (15,16). These effects may be attributed to the ability of the enzyme to hydrolyze several cytokines and chemokines, such as CCL2 (MCP-1), IL-6, pro-IL-1␤, and CCL5 (RANTES) (16,55). An initial downregulation of meprin ␣ at the site of inflammation might be important for a robust immune response, and it is possible that chronic low levels of meprin ␣ might have adverse affects that favor the development or progression of IBD.
It is interesting to note that substrates of meprin A, such as CCL2 and IL-6, are also post-transcriptionally regulated (38,56). These transcripts are stabilized and upregulated by inflammatory stimuli, whereas MEP1A is downregulated. As such, it would be expected that after an inflammatory stimulus the expression of cytokines would be maximally enhanced by increasing gene expression and by preventing turnover on the protein level. This is consistent with the ability of PTR to act in a coordinate fashion to affect expression of inflammatory mediators as well as genes that regulate these mediators.
The interplay of functionally opposing RBPs such as TTP and HuR, and how their actions are altered by various stimuli are not well established. HuR is generally believed to be pro-inflammatory and TTP anti-inflammatory, though they are both capable of regulating many of the same transcripts. Although HuR is ubiquitously expressed, TTP expression can be induced by phorbol esters and other inflammatory stimuli (38,57), and this may be one method by which a stimulus can alter the posttranscriptional regulation of a target. For instance, TTP expression in Caco-2 cells is low, consistent with a previous report (58). However, both TTP mRNA and protein expression was strongly upregulated by PMA, resulting in the binding of TTP to MEP1A mRNA and subsequent repression.
Mapping experiments demonstrated that HuR and TTP bound distinct regions within the 3Ј-UTR. Somewhat surprisingly, neither protein associated with the segment containing ARE sequences. Furthermore, binding of TTP did not appear to displace HuR, suggesting that the two proteins do not simply compete for binding and that TTP can exert a destabilizing effect on mRNA even when HuR is present on the transcript. This finding contradicts current models of PTR mechanisms that suggest that TTP and HuR exert effects by competing for binding (54). Given the large number of transcripts that both RBPs regulate, it is possible that HuR and TTP act in conjunction as part of a multiprotein complex, that other RBPs play a role in modulating their function, or that another stimulus-dependent signal, such as phosphorylation, can alter the function of the proteins.
A working model of MEP1A PTR is presented in Fig. 7. We propose that HuR binds to the MEP1A transcript at baseline to promote meprin expression. Under inflammatory conditions, TTP expression and subsequent binding to the 3Ј-UTR occur, causing destabilization of the MEP1A transcript and decreased expression. In addition, other RBPs or microRNA may also bind to the MEP1A 3Ј-UTR. This model could prove to be very useful to study the interplay of TTP, HuR, and other elements of PTR, and will be the focus of future work. In summary, the results of this study demonstrate the first direct evidence that meprin ␣ is post-transcriptionally regulated and that regulation is altered by inflammatory stimuli. These data provide important insight as to how expression of meprin ␣ is controlled, and shed light on the interplay between the functions of the RBPs TTP and HuR. Greater understanding of meprin ␣ regulation and PTR in IBD will help clarify the disease pathogenesis, and could provide a target for the development of novel therapeutics.