Functionally Independent AU-rich Sequence Motifs Regulate KC (CXCL1) mRNA*

Certain pro-inflammatory chemokine mRNAs containing adenine/uridine-rich sequence elements (AREs) in their 3′ untranslated regions (3′-UTRs) are known to exhibit constitutive instability and sensitivity to proinflammatory stimuli resulting in the stabilization of the message. Using tetR-regulated transcription we now show that the 3′-UTR of the mouse CXCL1 (KC) mRNA contains at least two ARE motifs that are structurally and functionally distinct. A fragment of 77 nucleotides containing 4 clustered AUUUA pentamers located at the 5′-end of the KC 3′-UTR is only modestly unstable yet promotes markedly enhanced, post-transcriptional protein production in response to either interleukin-1α (IL-1α) or lipopolysaccharide (LPS), suggesting translational regulation. In contrast, a fragment containing 3 isolated AUUUA pentamers corresponding to the residual 3′ 400 nucleotides of the KC 3′-UTR confers both instability and is stabilized in response to IL-1α. Although the clustered AUUUA pentamers in the upstream region are required for stimulus sensitivity, mutation of all three pentamers in the downstream region has little or no effect on either instability or stimulus sensitivity. The upstream region is comparably stabilized in response to either IL-1α or LPS, whereas the AUUUA-independent downstream determinant is differentially more sensitive to IL-1α. Finally, using UV-induced RNA cross-linking, these functionally independent sequences exhibit different patterns of interaction with RNA-binding proteins. Collectively, these findings document the presence of multiple independent determinants of KC mRNA function and demonstrate that these operate via distinct mechanisms.

Tissue inflammation is orchestrated via the production of multiple cytokines and chemokines that control the trafficking of leukocytes to sites of tissue injury and infection (1)(2)(3)(4). The production of these secreted mediators is stringently regulated at multiple mechanistic levels, including gene transcription, mRNA translation, and ultimately selective degradation of specific mRNAs (5)(6)(7)(8). Hence the pattern of gene expression depends critically upon appropriate engagement of each of these regulatory steps and deficiencies at any specific stage have been demonstrated to profoundly impact normal function (9,10). mRNA degradation is now widely recognized as an important regulatory step in controlling gene expression, and this is particularly true for short lived mRNAs such as those encoding cytokine and chemokine proteins (6,7,11,12). The instability of such mRNAs is determined by sequence motifs frequently located in the 3Ј untranslated region (3Ј-UTR) 1 of the message. The best studied of these sequences are known as adenine/ uridine rich elements (AREs) and have been demonstrated to confer marked instability and, in some cases, potent sensitivity for stabilization in response to extracellular stimuli (6,8,11,13). Moreover, ARE sequences have been reported to regulate translational efficiency that also exhibits potent stimulus sensitivity (14 -16). More than 1000 ARE-containing mRNAs have been defined within the human genome (17). Not surprisingly, these exhibit marked structural heterogeneity, and at least three broad classes have been identified (18). Class I AREs contain multiple independent repeats of the AUUUA pentamer. Class II AREs are defined on the basis of their content of multiple overlapping or closely juxtaposed AUUUA motifs. Class III AREs contain no pentamer motifs but do contain stretches of AU or U rich sequence. A number of recent reports suggest that there is substantial functional heterogeneity among ARE-containing mRNAs (19 -23). Certainly such mRNAs exhibit a broad range of decay rates and likewise show considerable variability with respect to their sensitivity for modulation of decay in response to extracellular stimulation.
The functional activity of ARE sequences reflects their interaction with proteins exhibiting appropriate sequence recognition specificity. A number of ARE-binding proteins have been identified over the last 15 years, and these have been correlated with either altered decay or translation of target mRNAs (24 -28). Despite much interest, however, clear cause and effect relationships between individual ARE-binding proteins and the behavior of specific target mRNAs have been established in only a few cases (29 -33). In light of the number of AREcontaining mRNAs, the structural heterogeneity they exhibit, and the number of RNA-binding proteins showing preference for AU-rich sequences, it is likely that ARE sequences confer a broad array of behaviors and provide significant diversity to the regulation of gene expression.
In the present study, we have evaluated the functional heterogeneity of sequences within the mRNA encoding the mouse chemokine CXCL1 or KC. The KC gene (scyb1) encodes a potent neutrophil chemoattractant and serves as a functional homologue of human IL-8 in the mouse (1)(2)(3)(4). This mRNA is highly unstable in leukocytes, endothelial cells, fibroblasts, and epithelial cells and can be stabilized very effectively in response to several pro-inflammatory stimuli, including lipopo-lysaccharide (LPS) and IL-1␣ (34 -36). Although we have previously reported that a clustered set of four overlapping AUUUA pentamers is a critical determinant of this behavior, the 3Ј-UTR of KC contains three isolated pentamer motifs as well as additional AU-rich sequence (37). Functional dissection of the full 3Ј-UTR of KC mRNA reveals two independent determinants that confer mechanistically different post-transcriptional regulation of KC gene expression. Plasmids-Radiolabeled cDNA probes for use in Northern hybridization analysis were prepared from plasmids containing fragments of GAPDH and KC in the Bluescript vector. Plasmids used to drive expression of different versions of KC were prepared in pTRE2 (Clontech Inc.). The parent clone was created by insertion of the full KC 5Ј-UTR and coding region (residues 1 through 359) into the BamH1/NotI sites of pTRE2, and the 3Ј-UTR was provided from the rabbit ␤-globin gene. Additional constructs were created by excising the rabbit ␤-globin region with XbaI and Sap1, and different versions of the KC 3Ј-UTR sequence were inserted in the remaining EcoRV site. The full-length KC 3Ј-UTR (designated FL) contained residues 360 -952. The cluster only (CLU) clone contained residues 406 -483 and 868 -950, and the ⌬1 clone contained residues 467-950. The sequential deletions of the ⌬1 fragment were made at residues 538 (⌬2), 634 (⌬3), 807 (⌬4), and 868 (⌬5) and extended to residue 950. Mutant versions of CLU and FL were prepared by PCR or oligonucleotide site-directed mutagenesis as described previously (34,38). The FL(CLUmt) was prepared by substitution of the sequence TATTC-GATCGAGATATCTCC for residues 445-470 in the wild-type sequence eliminating the clustered AUUUA motifs. The CLUmt contained residues 406 -483 with the same mutation in the cluster of AUUUA pentamers described above. The ⌬1 mutant substituted CG for UU residues in the three remaining AUUUA pentamers (see Fig. 1). The 5ϫ B luciferase construct was described previously (39).

Reagents
Cell Culture and Transfection-HEK293 C6 cells stably expressing human IL-1R1 were prepared as described previously (40) and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin, and streptomycin in humidified 5% CO 2 . HEK293 C6 cells stably expressing the tetR-VP-16 fusion protein (293tetoff) were prepared as previously described and maintained in G418 and puromycin (39). HEK293 C6 cells stably expressing TLR4 and MD2 were prepared as described previously (39) and served as the parent line for development of the 293-TLR4/MD2-tet-off cells by transfection with the tetR-VP-16 fusion protein and selecting with puromycin. 293tet-off cells were also used to prepare lines stably expressing versions of KC mRNA containing the CLU or ⌬1 fragments by selection with hygromycin. Transfections were done using SuperFect transfection reagent.
Measurements of RNA Stability-Pools of 293tet-off cells were transiently transfected using SuperFect transfection reagent according to the manufacturer's protocol. Three hours after transfection, the cultures were subdivided into 60-mm dishes and rested for 24 h prior to individual treatments. KC mRNA transcription was terminated by the addition of Dox, and total RNA was prepared at the indicated times using Tri reagent following the manufacturer's instructions. Total RNA preparations were digested with RNase-free DNase to eliminate residual plasmid DNA prior to analysis of specific mRNA content by Northern blot hybridization as described previously (19,38).
Post-transcriptional KC Protein Secretion-mRNA decay was estimated by measuring the secretion of KC protein before and after the termination of reporter gene transcription with doxycycline as described previously (39). Pools of 293tet-off cells were transiently transfected as described above. Twenty-four hours after transfection, the supernatants were removed and replaced by fresh medium. After 3-h incubation, the supernatants were harvested and saved for later determination of KC protein secretion by enzyme-linked immunosorbent assay prior to the termination of transcription. The plates were washed, and fresh medium containing Dox with or without IL-1␣ or LPS was added for 3 h to allow mRNA decay to occur in the absence of transcription. The supernatants were discarded, and cultures were washed extensively prior to addition of fresh medium containing Dox with or without stimuli for a final 3 h, and supernatants were harvested again for determination of KC protein. The ratio of protein secretion before and after the addition of Dox provides a quantitative estimate of residual RNA.
Cell-free RNA Decay-Cell-free mRNA decay assays were adapted from Mukherjee et al. (41) as described previously (38). The substrates were generated using either the ⌬1 or the CLU fragments in pBS by in vitro transcription in the presence of cap analogue (7meGpppG) following linearization with XbaI to produce a 5Ј-capped RNA substrate internally labeled with 32 P. The RNA was purified from a 6% polyacrylamide urea gel, and a poly(A) tail was added with the poly(A) taillabeling kit (Ambion) according to the manufacturer's protocol. The specific activity of the substrate was ϳ0.2 ϫ 10 6 cpm/ng. For the in vitro decay assay, 32 P-labeled RNA (1 ϫ 10 4 cpm) was incubated with S100 extract (10 g of protein) in a 25-l reaction volume containing 10 mM HEPES (pH 7.9), 50 mM KCl, 1 mM MgCl 2 , 1 mM dithiothreitol, 0.6 mM ATP, 1.5 g of poly(A), and 15 mM creatine phosphate for the indicated times. The decay of RNA substrates in this cell-free system was quantified by determining residual trichloroacetic acid-precipitable radioactivity as described previously (38).
UV Cross-linking-S100 extracts were prepared according to Ford and Wilusz (42). Protein concentration was measured by the method of Bradford (43). The extracts were stored in aliquots at Ϫ80°C. Plasmids containing either the CLU or ⌬1 fragments (see Fig. 1) (or their corresponding mutant versions) were used for in vitro transcription of RNA probes. pBS containing the ⌬1 fragment was cut with BamH1 and transcribed using T3 polymerase, whereas pBS containing the CLU fragment was linearized with XhoI and transcribed with the T7 polymerase in the presence of cap analogue as described above and isolated following 6% urea-PAGE. The specific activity of the RNA probe was 0.5-1.0 ϫ 10 6 cpm/ng. UV cross-linking was carried out as described before (44). Each reaction (20 l) contained 20 g of S100 protein, 1 ϫ 10 4 cpm RNA probe in binding buffer (10 mM Hepes, 100 mM KCl, 3 mM MgCl 2 , 2 mM dithiothreitol, 5% glycerol) for 20 min at 4°C. Heparin (30 g/ml) and yeast tRNA (50 g/ml) were added, and the mixture was incubated further for 10 min at 4°C. Each reaction was subjected to UV irradiation with a hand-held UV lamp (254 nm) for 20 min on ice, digested with a mixture of RNase A/T1 at room temperature for 30 min, and separated on 10% SDS-PAGE. The gels were stained with Coomassie Blue, destained, dried, and analyzed by autoradiography.

RESULTS
The 3Ј-UTR of KC contains 7 AUUUA pentamers, including a cluster composed of two juxtaposed overlapping AUUUA pairs and 3 isolated AUUUAs (see Fig. 1) (34,35,37). The full 3Ј-UTR sequence is responsible for both constitutive instability and sensitivity to stabilization in response to pro-inflammatory agents such as IL-1. This is clearly demonstrated in Fig. 2 where a KC cDNA transgene containing the full 3Ј-UTR was placed under control of a tetracycline regulated promoter allowing Dox-mediated suppression of transcription. Using 293C6 cells stably expressing both the type I IL-1R and a tetR-VP-16 fusion protein (293tet-off), transient transfection of the KC cDNA resulted in high level expression of specific mRNA. Following the addition of Dox to terminate transcription, KC mRNA was observed to decay rapidly. In cells that were stimulated with IL-1␣ at the time of Dox addition, the mRNA was stabilized ( Fig. 2A). A separate construct was pre-pared by deleting all 7 AUUUA pentamers from KC cDNA (construct ⌬5 in Fig. 1); when examined under the same conditions, this mRNA, containing only a truncated version of the 3Ј-UTR, was stable and insensitive to stimulation. These results were confirmed by measuring the amount of KC protein produced before and after the addition of Dox (Fig. 2B).
The role of clustered AUUUA pentameric sequence motifs in determining both mRNA instability as well as sensitivity for stabilization in response to extracellular stimulation has been demonstrated in multiple prior studies (13,(21)(22)(23). We have previously reported that the pair of overlapping AUUUA-containing clusters located in the 3Ј-UTR of the mouse KC mRNA are important for both instability and stimulus-induced stabilization or destabilization in mouse macrophages (34,35). In the present study we address the role of the residual three isolated pentamer elements located downstream of the cluster. Two different experimental strategies were employed. In the first, we prepared a KC cDNA plasmid construct containing the full 3Ј-UTR with mutations destroying the 4 clustered AUUUAs (FL(CLUmt)). Surprisingly, KC mRNA derived from this construct remained highly unstable and exhibited comparable or greater sensitivity to IL-1␣ (Fig. 3A). In the second strategy, a construct was prepared in which the region of the KC 3Ј-UTR containing the cluster was deleted (this construct is termed "⌬1," see Fig. 1). Consistent with the behavior of the FL(CLUmt), this deletion mutant also retained both instability and IL-1 sensitivity. Decay rates from these experiments are shown graphically in Fig. 3B. The quantitative estimate of mRNA stability provided by measurement of the residual secretion of KC protein after Dox-mediated inhibition of transcription (Fig. 3C) confirms the RNA behavior. In these assays both the FL(CLUmt) and ⌬1 fragments of KC appear to exhibit comparable instability and even enhanced sensitivity to IL-1␣ for stabilization as compared with the full-length wild type control. These finding strongly suggest that the KC 3Ј-UTR contains two independent sequence determinants controlling post-transcriptional mRNA behavior: the clustered AUUUA pentamers located in the 5Ј-end and the remaining 3 isolated pentamers (including AU-rich regions) in the more 3Ј region.
To further document the separate function of the CLU and ⌬1 fragments, the CLU fragment was placed independently in a construct terminated by the polyadenylation signal (⌬5 fragment) and compared with that of the ⌬1 fragment following stable transfection into 293tet-off cells. As expected, each construct exhibited constitutive instability and sensitivity for enhanced stability in response to IL-1␣ (Fig. 4, A and B). Surprisingly, however, we noted that there was a significant quantitative difference in the half-lives of mRNAs derived from the two constructs; whereas the ⌬1 fragment was highly unstable, as seen in Fig. 3, the CLU fragment was more stable. Moreover, whereas the ⌬1 fragment exhibited strong stabilization in response to IL-1␣ stimulation, the stabilization of the CLU-derived message was more modest, in keeping with its more limited rate of decay. Of interest, however, was the finding that IL-1 stimulated a comparable increase in post-Dox KC protein secretion from both constructs. This finding suggests that IL-1␣ may regulate the translation of CLU-containing mRNA, a possibility consistent with prior reports demonstrating that some ARE motifs possess such activity (14 -16).
Because both the CLU and the ⌬1 fragments contain multiple repeats of the AUUUA motif, we assessed the importance of the pentamers in each fragment for their contribution to instability and/or stimulus sensitivity. Wild type and mutated versions of the CLU or the ⌬1 fragments were prepared, and their function A, the sequence of the 3Ј-UTR of the mouse KC mRNA is outlined starting immediately after the translational termination codon. The sequence numbering corresponds to the full-length KC mRNA as defined previously (34). Pentamer structures are presented in bold type and labeled as cluster and pentamer number (P1 to P3). AU-rich regions of the sequence are underlined. B, cDNA constructs utilized in this study. All constructs contain the full 5Ј-UTR and coding region of KC. The nucleotide residues that define each construct are provided under "Experimental Procedures." The positions of the pentamer cluster and individual pentamers are indicated above the FL construct. evaluated following transient transfection in 293tet-off cells (Fig.  4, C and D). As expected from prior studies, mutation of the four pentamers clustered in the CLU fragment resulted in a loss of sensitivity to IL-1␣ (34,35). This was most evident at the protein production level (Fig. 4, C and D). Although limited, the CLUmt construct retained RNA instability. Of greater interest was the observation that mutation of all 3 pentamers present in the ⌬1 fragment had only modest impact on the instability of mRNA or sensitivity to IL-1␣ for stabilization. Because the determinants of instability, stabilization, and/or translational control depend upon structurally distinct nucleotide sequences, these findings further support the possibility that the two fragments operate through distinct mechanisms.
The finding that mutation of all three pentamers in the ⌬1 fragment caused little or no change in instability and did not impact stimulus-induced stabilization suggests that there are other regulatory sequence elements located within this fragment. As illustrated in Fig. 1, each of the pentamers in the ⌬1 fragment is associated with an AU-rich region either upstream (P1 and P2) or downstream (P3). In addition, there are two separate AU-rich regions (one located between P2 and P3 and one located downstream of P3). We therefore surveyed the mutant form of the ⌬1 fragment for functional regions by preparing a series of deletion constructs in which the mutated pentamer or AU-rich regions were sequentially removed. The results show an incremental reduction in instability as each segment is removed and complete loss of instability following the deletion of the last region of AU-rich character (excluding the polyadenylation signal) (Fig. 5). mRNA containing only the last 82 nucleotides (termed "⌬5") was also shown in Fig. 1 to be stable. Interestingly, although each successive deletion results in an incrementally more stable message, all constructs that were unstable could be stabilized by IL-1␣. These findings indicate that there are multiple instability determinants within the ⌬1 fragment. Stimulus sensitivity, however, appears to reside in the last AU-rich segment.
To further evaluate the functional distinction between the two separate determinants in the KC 3Ј-UTR, we compared their sensitivity to either IL-1␣ or LPS, another inflammatory stimulus known to promote enhanced stability of short-lived ARE-containing mRNAs (38,46). This was accomplished using a 293tet-off cell line constructed to express TLR4 and MD2, cell surface proteins that together confer sensitivity to LPS (39). Cells transfected with the CLU construct exhibited comparable sensitivity to either IL-1␣ or LPS in terms of KC protein secretion after Dox (Fig. 6A). In contrast, cells transfected with the ⌬1 construct showed greater sensitivity to IL-1␣ than to LPS. This differential behavior did not reflect differential cell sensitivity to LPS, because treatment with LPS or IL-1␣ promoted comparable activation of a co-transfected B-dependent luciferase reporter (Fig. 6B). Similar differential sensitivity to the two stimuli was also observed when examined by Northern hybridization (Fig. 6C). Although the CLU-containing mRNA was more stable than that containing the ⌬1 fragment, the CLU fragment exhibited comparable (though modest) stabilization in response to either IL-1␣ or LPS. The mRNA containing the ⌬1 fragment was highly unstable, well stabilized by IL-1␣ but relatively insensitive to stimulation with LPS. The differences in half-life as compared with protein secretion confirm the possibility that the behavior of the CLU fragment involves translational control.
The differential stimulus sensitivity of the two sequence determinants was independently documented in cell-free mRNA degradation assays (41,42). Using post-polysomal fractions of untreated, IL-1␣-treated, or LPS-treated cells, we evaluated the decay of 5Ј-capped, polyadenylated in vitro transcripts from the CLU or ⌬1 fragments. The CLU fragment was readily degraded using extracts obtained from untreated cells and showed comparably enhanced stability in extracts from cells treated with either IL-1␣ or LPS (Fig. 6D). In contrast, the ⌬1 fragment was more unstable in extracts from untreated cells and showed significantly greater stability when assayed with extracts from IL-1␣-treated cells as compared with extracts from LPS-treated cells.
The regulatory function of mRNA sequences controlling their rates of decay must operate through recognition by and interaction with factors that mediate this behavior. Indeed, a number of proteins that recognize ARE sequences have been identified, cloned, and evaluated (24 -28). Because of the different structural and functional characteristics of the two fragments composing the KC mRNA 3Ј-UTR, we sought to examine the spectrum of proteins capable of interacting with each sequence. Radiolabeled in vitro transcripts corresponding to the CLU and ⌬1 fragments were prepared, and their interaction with proteins present in post-polysomal S100 extracts from 293tet-off cells was assessed by UV radiation-induced cross-linking as described under "Experimental Procedures." The RNA-tagged proteins were separated by SDS-PAGE and evaluated by autoradiography (Fig. 7). Although a number of cross-linked proteins interact comparably with both fragments, two show sig-

FIG. 2. The KC 3-UTR containing 7 AUUUA pentamers confers instability and stimulus sensitivity.
A, 293tet-off cells were transfected in separate pools with plasmid constructs FL or ⌬5 (see Fig. 1) and after 3 h were separated into 3 individual Petri dishes and cultured overnight. Dox (1 g/ml) with or without IL-1␣ (10 ng/ml) was added and total RNA was prepared following further incubation for the indicated times. KC and GAPDH mRNA levels were determined in each sample by northern hybridization as described under "Experimental Procedures." B, groups of 293tet-off cells transfected as in A were separated into four individual 35-mm Petri dishes. After overnight incubation, the cultures were washed and KC protein secretion before and after the addition of Dox Ϯ IL-1␣ were determined as a quantitative estimate of the rate of KC mRNA degradation as described under "Experimental Procedures." Similar results were obtained in three separate experiments. nificant and reproducible differences in multiple experiments (see arrows in figure). A protein of ϳ90 kDa is the most abundant protein binding to both the CLU and ⌬1 fragments but exhibits quantitative preference for the CLU fragment. The specificity of this interaction is indicated by a loss of binding interaction with mutant transcripts of the CLU fragment (Fig.  7B). Interestingly, the 90-kDa protein binds to both the wild type and the mutant version of ⌬1 fragment. A second protein of ϳ60 kDa binds selectively to the ⌬1 fragment but not the CLU fragment. This protein interaction was retained in the version of ⌬1 containing mutated AUUUA pentamers. When the ⌬4 fragment was used as a probe for protein binding, the 60-kDa protein interaction was lost (Fig. 7C). The behavior of these two proteins is consistent with their potential roles in mediating either stimulus sensitivity or instability. The binding of the 90-kDa protein is lost upon mutation of the CLU, and this mutation also destroys stimulus sensitivity for this fragment in intact cells and instability when assayed in the cellfree mRNA decay system. The 90-kDa protein binding is not lost in the mutant ⌬1 fragment where both instability and stimulus sensitivity are retained. Likewise, the 60-kDa protein shows specificity for the ⌬1 fragment that exhibits instability/ stimulus sensitivity in both the wt and mutant versions. The loss of binding to the ⌬4 fragment suggests that this interaction does not confer stimulus sensitivity. The pattern of protein binding to either fragment was not altered when cross-linking studies were performed using extracts from IL-1␣-or LPSstimulated 293TLR4/MD2tet-off cells (Fig. 7D). DISCUSSION The importance of ARE sequences in enhancing mRNA decay kinetics is well established, and such sequences are also known FIG. 3. The CLU region is not required for KC mRNA instability or IL-1␣-stimulated stabilization. A, 293tet-off cells were transfected with FL, FL(CLUmt), or ⌬1 constructs as described in the legend to Fig. 2. After overnight incubation, the cells were treated with Dox Ϯ IL-1␣ as in Fig. 2 and incubation continued for the indicated times prior to harvest and determination of KC and GAPDH mRNA levels as described in Fig.  2. B, the autoradiographs shown in A were quantified using the National Institutes of Health Image software and the levels of KC mRNA normalized to those of GAPDH in each sample. The ratio is presented as the percentage remaining mRNA. C, 293tet-off cells were transfected as in A and used to measure KC protein secretion before and after the cessation of KC transcription by the addition of Dox either with or without IL-1␣ as described in the legend to Fig. 2. Similar results were obtained in two experiments.
to confer sensitivity for mRNA stabilization in response to extracellular stimuli (6,7,11,12). AREs are, however, structurally diverse and are recognized with variable specificity by a relatively broad spectrum of RNA-binding proteins (24 -28). Although it is likely that this structural diversity translates into substantial functional diversity, the relationship between FIG. 4. CLU and ⌬1 fragments are independent determinants of KC mRNA instability and stimulus-induced stabilization. A, KC protein secretion from 293tet-off cells stably expressing either the CLU or ⌬1 KC constructs was determined before or after the addition of Dox either with or without the addition of IL-1␣ as described in the legend to Fig. 2. B, 293tet-off cells stably expressing either CLU or ⌬1 derivatives of KC cDNA as in A were plated in 100-mm Petri dishes and stimulated with nothing (NT) or IL-1␣ (10 ng/ml) in the presence of Dox for the indicated times prior to preparation of total RNA and analysis of KC and GAPDH (not shown) mRNAs by northern hybridization. C, 293tet-off cells were transiently transfected with either wild type or mutant versions of the CLU or ⌬1 constructs and, following 3 h, each transfection pool was divided into separate Petri dishes for measurement of KC protein secretion before and after the addition of Dox in the presence or absence of IL-1␣ as described in the legend to Fig. 2. D, 293tet-off cells were transiently transfected with wt and mt versions of CLU or ⌬1 constructs as in C and divided into separate Petri dishes for determination of KC mRNA levels after the addition of Dox with or without IL-1␣. After the indicated times, total RNA was prepared and used to determine the levels of KC, and GAPDH mRNAs by Northern hybridization. Autoradiographs were quantified with the National Institutes of Health Image software, and levels of KC mRNA normalized to that of GAPDH in the same samples. Similar results were obtained in two separate experiments.
FIG. 5. Deletion mutagenesis of the ⌬1 fragment. 293tet-off cells were transiently transfected with the indicated plasmids containing the KC coding region and incrementally smaller regions of the ⌬1 fragment prepared as described under "Experimental Procedures" and illustrated in Fig. 1B. KC protein secretion before and after treatment with Dox in the presence or absence of IL-1␣ was determined as an estimate of KC mRNA decay as described in the legend to Fig. 2. Similar results were obtained in two separate experiments. specific sequences, sequence-specific-binding proteins, and the pattern of decay for individual mRNAs is still poorly understood. In the present study we began to explore these issues by examining the relative importance of the 7 AUUUA pentamers found in the mouse KC chemokine mRNA in governing the post-transcriptional regulation of its function. We now report that KC mRNA contains at least two functionally independent determinants, each of which exhibits sensitivity to extracellular stimuli. These two determinants are functionally distinguished by several criteria indicating that they operate through separate mechanisms. These conclusions are supported by the following findings. 1) Deletion or mutation of the 4 clustered AUUUA pentamers does not affect the capacity of the remaining sequence to confer instability nor reduce its sensitivity to stimulation with IL-1␣. The cluster alone, however, confers only modest instability but marked sensitivity to IL-1␣ or LPS that appears to involve regulation of translation. 2) Although the behavior of the CLU fragment depends upon the clustered AUUUA pentamer content, the function of the second element (⌬1) is only modestly altered by mutation of the AUUUA structures.
3) The two sites exhibit differential sensitivity to two stimuli; LPS and IL-1␣ comparably modulate function of the CLU fragment, whereas IL-1␣ is more effective than LPS in promoting enhanced stability of the ⌬1 fragment. 4) The two sequence fragments exhibit RNA-binding protein specificities that correlate with their functional differences. Collectively these findings establish the functional heterogeneity of the ARE sequence motifs.
The demonstration of distinct and independent nucleotide sequence motifs regulating mRNA instability, translation, and stimulus-mediated stabilization underscores the diversity of mechanisms that are associated with ARE structure. Prior work has clearly demonstrated that specific mRNAs contain both ARE and non-ARE determinants of instability, and these may be differentially sensitive to stimuli such as IL-1␣ or LPS (21,(47)(48)(49). Furthermore, both the diversity of ARE structure and the number of ARE-containing mRNAs strongly suggest that ARE function and mechanism are likely to be diverse. For example, whereas many ARE-containing mRNAs are unstable, only a subset exhibit sensitivity for stabilization in response to stimulation (17,19,20). The present study extends the understanding and scope of mechanistic heterogeneity within ARE-dependent decay. The observation that the ⌬1 fragment is more unstable than is the CLU fragment suggests significant functional difference between the two sequences. This is more clearly demonstrated, however, by the finding that both fragments exhibit nearly com-FIG. 6. ⌬1 and CLU determinants exhibit differential sensitivity to stabilization by IL-1␣ and LPS. A, 293TLR4/MD2 tet-off cells were transiently transfected with either the ⌬1 or CLU constructs and used to determine the KC protein secretion before and after the addition of Dox with or without the addition of IL-1␣ (10 ng/ml) or LPS (1 g/ml) as described in the legend to Fig. 2. Similar results were obtained in three separate experiments. B, the 293tet-off cell lines described in A were co-transfected with a 5ϫ B luciferase reporter. After overnight incubation, the cells were stimulated with nothing (NT), IL-1␣ (10 ng/ml), or LPS (1 g/ml) for 6 h prior to cell lysis and measurement of luciferase activity. Similar results were obtained in two separate experiments. C, 293tet-off cells stably expressing TLR4/MD2 were transiently transfected with the ⌬1 or CLU constructs as described in A and used to determine the decay of mRNA following the addition of Dox with or without the addition of IL-1␣ or LPS. KC and GAPDH (not shown) mRNA levels were determined by Northern blot hybridization. Similar results were obtained in three separate experiments. D, 32 P-radiolabled substrate RNAs corresponding to either the CLU or ⌬1 regions of the KC 3Ј-UTR were prepared by in vitro transcription with a 5Ј-7-meG cap and a poly(A) tail as described under "Experimental Procedures." S100 protein extracts were prepared from 293tet-off cells stably expressing both TLR4/MD2 and IL-1R1 receptors either without treatment or following treatment for 2 h with IL-1␣ (10 ng/ml) or LPS (1 g/ml). Reactions were carried out in a total volume of 25 l containing 10 g of extract protein and 4 ϫ 10 4 cpm of RNA substrate for the indicated times, and residual substrate RNA was determined as described under "Experimental Procedures." Similar results were obtained in two separate experiments. parable increases in protein production following stimulation despite the marked differences in both mRNA stability and stabilization in response to stimulation with IL-1␣. This finding strongly suggests that stimulus sensitivity of the CLU fragment operates by modulation of translational efficiency. Stimulus-dependent translational control has been previously documented in the context of the tumor necrosis factor-␣ mRNA ARE motif (14,16,50). Interestingly, both the CLU and the tumor necrosis factor-␣ AREs possess two overlapping AUUUA pentamer sets (14,28). These findings strongly suggest that the functions of the two regulatory elements are carried out through distinct molecular mechanisms. This is further supported by the difference in AUUUA sequence requirements between the CLU and ⌬1 portions of the KC 3Ј-UTR. Moreover, the differential sensitivity of each element to stimulation with either IL-1␣ or LPS suggests that signaling pathways coupling with these post-transcriptional control mechanisms are distinguishable. Although the CLU element is comparably stabilized by LPS and IL-1␣, the ⌬1 fragment is more sensitive to IL-1␣ than to LPS. Although LPS and IL-1 both signal through Toll-Interleukin 1 family receptors that share many features in signal transduction, these pathways appear to couple with at least partially independent downstream mechanisms. The finding that separate stimuli differentially control the function of a single mRNA represents an extension of our earlier studies illustrating that a single stimulus can promote enhanced stability of two different sequences through separate pathways (19).
Although ARE sequences have been known to confer mRNA instability for many years, the definition of specific sequences and their relationship to function remains poorly understood. There are a large number of mRNAs in the human genome that contain ARE sequences, and it is not surprising that the stability of such sequences will be highly diverse (17). Indeed, oligonucleotide and cDNA array analyses examining the behavior of these sequences demonstrate a broad spectrum of stability and sensitivity to extracellular stimulation (19,20). ARE-containing mRNA sequences fall into one of three classes (I, II, or III) based upon the content and organization of pentameric AUUUA sequences (18). Class II mRNAs, containing multiple overlapping clusters of AUUUA, are found in many pro-inflammatory cytokine genes, and such sequences have been most commonly linked with instability and sensitivity to stabilization in response to agents, including IL-1 and LPS. These structures have been characterized in detail in several mRNAs, including tumor necrosis factor-␣, granulocyte macrophage-colony stimulating factor, Cox2, and IL-8 (13,22,23,51,52). Although non-class II ARE-containing mRNAs are also known to exhibit stimulus-mediated stabilization, the linkage between any given ARE structure and stimulus sensitivity remains largely undefined. One objective of our study was to determine the relative importance of the pentameric AUUUA structures found in the mouse KC mRNA for its decay and stabilization behavior. Although our findings confirm the importance of AUUUA structure in the context of the isolated CLU fragment, the ⌬1 fragment confers both properties without requirement for any pentamer. Although this separate element contains 3 isolated pentamers, these can be disrupted by mutagenesis without significant effect on either instability or sensitivity to stabilization by IL-1. This demonstrates that FIG. 7. Analysis of protein binding to CLU, ⌬1, and ⌬4 fragments. A, 32 Pradiolabeled RNA fragments corresponding to wild-type versions of the CLU and ⌬1 fragments of the KC 3Ј-UTR were prepared by in vitro transcription, and each contained a 5Ј-7-meG cap. 5 ϫ 10 6 cpm of each fragment was incubated in a total volume of 25 l containing 20 g of S100 extract from 293tet-off cells and analyzed by UV-cross-linking and SDS-PAGE as described under "Experimental Procedures." B, 32 P-radiolabeled RNA fragments corresponding to wild-type or AUUUA mutant versions of the CLU and ⌬1 fragments of the KC 3Ј-UTR were prepared and used for UV cross-linking analysis as described in A. C, 32 P-radiolabeled 5Ј-capped RNA fragments corresponding to the ⌬1 and ⌬4 fragments of the KC 3Ј-UTR were incubated with S100 extracts from 293TLR4/MD2tet-off cells and analyzed for RNA-binding proteins as in A. D, 32 P-radiolabeled 5Ј-capped RNA fragments corresponding to wild-type versions of the CLU and ⌬1 fragments of the KC 3Ј-UTR were incubated with S100 protein extracts prepared from 293TLR4/ MD2tet-off cells either untreated or treated with IL-1␣ (10 ng/ml) or LPS (1 g/ml) for 2 h. Reactions were subjected to UV-cross-linking and analysis as described in A. Positions of molecular weight markers are indicated on the left of each figure. Arrows indicate the positions of bands that exhibit selective binding with CLU or ⌬1 fragments. Similar results were obtained in three separate experiments.
stimulus-specific responses can involve a broad range of ARE structures and do not depend specifically on the AUUUA motif.
It is fully expected that the sequence specific decay behavior of mRNAs will result from their recognition by and interaction with specific protein factors. Indeed, multiple ARE-binding proteins have been identified, and a subset has been studied in detail (16,24,25,27,28). In some experimental models, AREbinding proteins have been causally linked with specific message instability and/or stimulus-sensitive behavior (10,30,53). Indeed, a recent report suggests that the zinc finger protein Tristetraprolin may be a mediator of LPS-or IL-1-induced stabilization of ARE containing cytokine mRNAs (54). Because Tristetraprolin (and several related isoforms) are not present in the HEK293 cell line employed in the present study (45), there are likely to be additional factors mediating the regulation of KC mRNA decay. By comparing protein binding interactions using these two distinct sequences, we have identified two proteins whose binding specificity exhibit some correlation with function. A major protein with a molecular size of ϳ90 kDa binds to both CLU and ⌬1 sequences, and binding to the CLU is lost when the pentameric structure is mutated. This behavior correlates well with the loss of function. Furthermore, the 90-kDa protein binding interaction with the ⌬1 fragment is not lost when the three pentamers in this sequence are mutated, and this is consistent with the retention of both instability and stimulus sensitivity in the mutant fragment. The second protein exhibits a molecular size of ϳ60 kDa. This binding interaction appears to exhibit specificity for the ⌬1 fragment and could contribute to the functional differences in decay behavior seen between the ⌬1 and CLU fragments. Because the binding of the 60-kDa protein is lost in the ⌬4 fragment, which exhibits much of the stimulus sensitivity of the full ⌬1 fragment, it is unlikely that this protein participates in stimulus sensitivity. Rather, because some level of instability is lost by deletion of the region to which the 60-kDa protein binds, this factor may contribute to instability though the magnitude of the effect is modest. Further studies are underway to determine the identity and functional roles for these factors.