Diversity in sequence-dependent control of GRO chemokine mRNA half-life

Regulation of GRO chemokine half-life is determined by multiple sequence regions that confer IL-1α -dependent and -independent instability via distinct RNA binding proteins.

Abstract

Neutrophil trafficking to sites of injury or infection is regulated, in part, by the closely related GRO family of chemokines (CXCL1, -2, and -3). Expression of the GRO chemokine genes is known to be determined by transcriptional bursts in response to proinflammatory stimulation, but post-transcriptional mechanisms that regulate mRNA half-life are now recognized as important determinants. mRNA half-life is regulated via distinct sequence motifs and sequence-specific, RNA-binding proteins, whose function is subject to regulation by extracellular proinflammatory stimuli. Moreover, such mechanisms exhibit cell-type and stimulus dependency. We now present evidence that in nonmyeloid cells, GRO2 and GRO3 isoforms exhibit at least two patterns of mRNA instability that are distinguished by differential sensitivity to specific mRNA-destabilizing proteins and stimulus-mediated prolongation of mRNA half-life, respectively. Although the 3′ UTR regions of GRO2 and GRO3 mRNAs contain multiple AREs, GRO2 has eight AUUUA pentamers, whereas GRO3 has seven. These confer quantitative differences in half-life and show sensitivity for TTP and KSRP but not SF2/ASF. Moreover, these AUUUA determinants do not confer instability that can be modulated in response to IL-1α. In contrast, IL-1α-sensitive instability for GRO2 and GRO3 is conferred by sequences located proximal to the 3′ end of the 3′UTR that are independent of the AUUUA sequence motif. These regions are insensitive to TTP and KSRP but show reduced half-life mediated by SF2/ASF. These sequence-linked, post-transcriptional activities provide substantial mechanistic diversity in the control of GRO family chemokine gene expression.

Introduction

The inflammatory response involves the recruitment of leukocytes to sites of injury or infection, followed by instruction as to their behavior upon arrival, and these activities are dependent on a complex network of cytokines and chemokines [1, 2]. Chemokines that recruit the earliest inflammatory leukocytes (predominantly neutrophils) are typically induced in a wide selection of cell types resident within tissues (e.g., stromal, epithelial, endothelial, myeloid) in response to proinflammatory stimuli that include ligands for members of the TLR family, as well as endogenous, proinflammatory cytokines, such as IL-1α and/or TNF-α [3, 4]. Individual chemokines orchestrate inflammation by targeting well-defined leukocyte subsets, and hence, the pattern of specific chemokine expression can determine the spectrum of distinct leukocytes present within the tissue at different stages of the response [2, 5]. The infiltration and degranulation of proinflammatory neutrophils may cause unnecessary tissue damage, and hence, chemokine expression must be controlled stringently [6]. This regulation is focused at different mechanistic levels that include gene transcription and mRNA turnover [7, 8]. Whereas the induction of early neutrophil-directed chemokine expression requires a rapid and potent burst of transcription, it has become clear that the pattern of expression can be modified dramatically by alterations in mRNA half-life [7–9].

The turnover of specific cytoplasmic mRNAs is regulated by sequence elements, commonly found in the 3′UTR that promote or inhibit degradation [10, 11]. The most-thoroughly studied sequence elements responsible for rapid mRNA decay are the AU-rich sequences (also known as AREs), which typically contain the core pentamer AUUUA [10, 12–14]. Such motifs are present in 5–8% of human genes and are likely to regulate a wide range of cellular processes, including growth, inflammation, and stress responses [10, 12, 15]. AREs have been grouped into three classes, largely on the basis of sequence. Class I AREs contain several scattered copies of the AUUUA pentamer, often flanked by additional A and U residues; Class II AREs possess at least two overlapping UUAUUUA(U/A) (U/A) nonamers, and Class III AREs contain AU-rich regions but no AUUUA motif [10, 12, 13].

The function of ARE sequences is mediated through specific RNA-binding proteins, which by virtue of interaction with other effector molecules, modulate the rates of mRNA degradation [8, 16]. Multiple proteins that recognize, bind, and promote degradation or stabilization of specific messages have been identified. Those that have been observed in multiple studies to regulate mRNA turnover include TTP (destabilizing), human antigen R (stabilizing), ARE/poly(U)-binding/degradation factor-1/heterogeneous nuclear RNP D (both), and the KSRP (destabilizing) [17–20]. Prior reports have demonstrated that instability and stimulus-induced stabilization of chemokine mRNAs can vary between cell populations and exhibit substantial mechanistic diversity [7, 8]. For example, we have determined recently that the mCXCL1 mRNA has multiple distinct sequence regions that control instability and sensitivity to proinflammatory stimuli that operate distinctly in myeloid cell types and in nonmyeloid cells, including fibroblasts and epithelial cells [21, 22]. The regulation of instability in nonmyeloid cells was shown to involve the action of the splicing regulatory protein SF2/ASF [23].

The chemokine family that specifically recruits neutrophils is known as the CXC ELR+ chemokine subgroup and consists of seven members, including three closely related genes encoding the GRO1–3 (also known as CXCL1, -2, and -3), as well as CXCL5 (epithelial-derived neutrophil-activating factor 78), CXCL6 (granulocyte chemotactic protein 2), CXCL7 (neutrophil-activating polypeptide 2), and CXCL8 (IL-8) [2, 5]. All of these chemokines possess proinflammatory properties and function mainly through interaction with the CXCR2 receptor (IL-8 is the exception that can also use CXCR1) [2, 5]. The nucleotide sequences for GRO1–3 are highly conserved; GRO2 and GRO3 exhibit only 13 and 14 aa differences, respectively, compared with GRO1 (out of 107 residues) [24]. GRO chemokines are expressed by a diverse selection of normal cells, including epithelial, endothelial, stromal, and myeloid cells [4, 25–28], and have also been shown to be produced constitutively by multiple tumor cells [29, 30]. In this context, they are believed to be important features of the contribution of inflammation to the promotion of cancer [31, 32].

Although the coding sequences of the three GRO chemokine genes share substantial homology, the noncoding 3′UTR sequences diverge considerably. Furthermore, whereas GRO1 and GRO2 3′UTRs share ∼80% identity, they both differ substantially from GRO3. Despite these differences, all three GRO chemokine mRNAs have been shown to have rapid turnover rates and to be stabilized in response to stimulation with IL-1β [33]. In the present report, we have used nonmyeloid cells to compare the 3′UTR sequences from GRO2 and GRO3 with respect to their ability to confer instability and/or sensitivity to stabilization in response to extracellular stimulation by IL-1α. GRO2 and GRO3 3′UTR sequences contain at least two functionally distinct determinants of instability that differ in terms of their sensitivity to destabilizing proteins and to IL-1α-driven prolongation of half-life. The ARE motifs in both GRO mRNAs are recognized by TTP and KSRP, and these sequence regions confer stimulus-insensitive instability when examined in HeLa cells. Furthermore, sequence motifs in GRO2 and GRO3 mRNAs that confer IL-1α-sensitive instability, are independent of the AUUUA motifs, are insensitive to the action of KSRP or TTP, but are selectively destabilized by the action of SF2/ASF. Hence, these findings demonstrate significant functional and mechanistic heterogeneity in the control of chemokine mRNA half-life that can operate through multiple regulatory sequence motifs within a single mRNA.

MATERIALS AND METHODS

Reagents

DMEM, Dulbecco's PBS, and antibiotics were obtained from Central Cell Services of the Lerner Research Institute (Cleveland, OH, USA). Neomycin sulfate (G418), formamide, dextran sulfate, MOPS, diethyl-pyrocarbonate, and Act D were obtained from Sigma Chemical Co. (St. Louis, MO, USA). FBS was purchased from BioWhittaker (Walkersville, MD, USA). dox and the vector pTRE2 were obtained from Clontech Laboratories (Mountain View, CA, USA). Random priming kits were purchased from Stratagene (La Jolla, CA, USA). Nylon transfer membrane was purchased from Micron Separation (Westborough, MA, USA). PolyFect transfection reagent was obtained from Qiagen (Valencia, CA, USA), and TRI Reagent was purchased from Molecular Research Center (Cincinnati, OH, USA). Salmon sperm DNA was obtained from Ambion (Austin, Tx, USA). Human rIL-1α was purchased from R&D Systems (Minneapolis, MN, USA). Protein G agarose beads were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA), and antibody for GAPDH was from Millipore (Billerica, MA, USA). Anti-SF2 and anti-TTP antibodies were obtained from Santa Cruz Biotechnology, whereas anti-KSRP antibody was obtained from Cell Signaling Technology (Beverly, MA, USA). DuPont-New England Nuclear (Boston, MA, USA) was the source of [α-³²P]deoxycytidine 5′-triphosphate. ProtoGel, SequaGel (acrylamide, N,N-methylene bis-acrylamide, urea), and related buffers were obtained from National Diagnostics (Atlanta, GA, USA). Protein assay reagents were purchased from Bio-Rad Laboratories (Hercules, CA, USA), and restriction enzymes were obtained from Roche Applied Science (Germany).

Plasmids

Plasmids, containing the coding region from mCXCL1 as a reporter linked with different versions of GRO2 and GRO3 3′UTRs, were prepared in pTRE2 (Clontech Laboratories). The parental clone (designated RBG) was created by insertion of the full mCXCL1 5′UTR and coding region (residues 1–359) into the BamHI/NotI sites of pTRE2, and the 3′UTR was provided from the rabbit β-globin gene. Additional constructs were created by inserting different versions of the GRO2 and GRO3 3′UTR sequence in the BglII site of the rabbit β-globin region. The sequences of the GRO2 and GRO3 3′UTRs are provided in Supplemental Fig. 1, with annotation to indicate the location of functional elements (pentamers; clustered pentamers), as well as positions for deletion constructs, as outlined below. The GRO2FL 3′UTR contained residues 613–1179, which excludes the last 23 nucleotides containing the polyadenylation signal. The sequential deletions of the GRO2FL fragment were made at residues 720 (GRO2Δ1), 837 (GRO2Δ2), 997 (GRO2Δ3), and 1035 (GRO2Δ4) and extended to residue 1179. The pentamer-free deletion mutant (GRO2Δ6) contained residues 997–1151. GRO2Δ5 contained residues 613–1005. The GRO3FL 3′UTR contained residues 487–1134, missing the last 17 nucleotides that contained the polyadenylation signal. The sequential deletions of the GRO3FL fragment were made at residues 664 (Δ1), 891 (Δ2), and 951 (Δ3) and extended to residue 1134. GRO3Δ4 contained residues 487–804. A mutant version of GRO3Δ2 was prepared by changing the AUUUA pentamers into AGCUA, using a site-directed mutagenesis method (QuickChange site-directed mutagenesis kit, Stratagene) and termed GRO3Δ2Pmt. The coding region of human SF2/ASF (SRSF1) was linked with a HA epitope tag in the expression plasmid pcDNA3.1.

Cell culture

HeLa tet-off cells were obtained from Clontech Laboratories and were maintained in complete DMEM.

Measurements of RNA stability

Cultures of HeLa tet-off cells were transfected with the indicated plasmids using PolyFect and 4 h later, were subdivided into 60-mm dishes and rested 24 h prior to individual treatments. Act D (5 μg/ml) or dox (1 μg/ml) was added to terminate transcription of endogenous and reporter genes, respectively, and total RNA was isolated at indicated times, prepared, and analyzed by Northern blot, as described previously [34, 35]. RNA half-lives were determined by scanning autoradiographs, followed by analysis with NIH ImageJ software.

RNAi

HeLa cells were cotransfected with the KSRP-specific siRNA (KSRP target sequence: GATCAACCGGAGAGCAAGA), TTP-specific siRNA (TTP target sequence: CGCUGCCACUUCAUCCACA), or a sequence control siRNA (Dharmacon, Lafayette, CO, USA) and GRO reporter constructs. Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) was used as a transfection agent, and transfection was performed, according to the manufacturer's instructions. The cells were incubated with transfection complexes overnight and then split to 60-mm dishes. The cells rested 48 h prior to individual treatment.

RNA-binding assays

The ability of TTP, KSRP and SF2 to bind to RNA in vivo was evaluated as described previously [23]. Briefly, cells were transiently transfected with GRO2FL and GRO3FL reporter constructs. Twenty hours after transfection, 2 × 10⁶ cells were trypsynized, washed twice, and resuspended in 10 ml ice-cold PBS. Cells were fixed by addition of formaldehyde (0.1%) for 15 min at room temperature, and the cross-linking reaction was stopped with glycine (pH 7.0, 0.25 M). The cells were then washed twice with ice-cold PBS, resuspended in 2 ml RIPA buffer (50 mM Tris-HCl, pH 7.5, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.05% SDS, 1 mM EDTA, 150 mM NaCl, and proteinase inhibitors), and sonicated. The lysate was centrifuged (15 min, 4°C, 16,000 g), and 1 ml each supernatant was immunoprecipitated overnight at 4°C, using protein G-agarose beads preincubated with anti-TTP, anti-KSRP, or anti-SF2 antibody. The beads were washed five times with 1 ml RIPA buffer and resuspended in 150 μl elution buffer (50 mM Tris-Cl, pH 7.0, 5 mM EDTA, 10 mM DTT, 1% SDS). Cross-linking was reversed by incubation at 70°C for 45 min, and RNA was purified from immunoprecipitates with TRI Reagent and treated with RNase-free DNase, and 10% of the total RNA sample was reverse-transcribed with MMLV RT. The RT product (2 μl; 10%) was subjected to quantitative real-time PCR. The primers for the CXCL1 coding region were 5′-CTGGCCACAGGGGCGCCTATC-3′ (forward) and 5′-GGACACCTTTTAGCATCTTT-3′ (reverse) and for GAPDH were 5′-TCACCATCTTCCAGGAGCGAGAT-3′ (forward) and 5′-GTTGGTGGTGCAGGAGGCATTGCT-3′ (reverse).

RESULTS

All three GRO mRNAs are unstable and can be stabilized in response to stimulation with IL-1α

The half-lives for GRO mRNAs and their sensitivity to IL-1α-mediated prolongation were examined by evaluating the kinetics of turnover in the presence or absence of IL-1α. We chose to determine expression patterns in an epithelial cell population, as such cells are frequently the source for GRO expression in normal tissues and in malignancy. Expression of endogenous GRO mRNA was induced in HeLa cells by treatment with TNF-α; we have reported previously that treatment with TNF-α drives transcription but does not stabilize mRNA in several human-cultured cell populations [36]. Further contribution of transcription to the GRO mRNA pool was stopped by addition of fresh medium containing the transcriptional inhibitor Act D in the presence or absence of IL-1α. Total RNA was isolated at various times after Act D treatment, and the levels of GRO mRNA remaining were determined by Northern hybridization (in this experiment, the radiolabeled hybridization probe detects all three GRO mRNAs as a result of high nucleotide sequence homology within the coding regions). GRO mRNA decayed rapidly in cells treated with TNF-α, followed by Act D alone, but in cultures where IL-1α was added along with Act D, the half-life was prolonged (Fig. 1A).

Figure 1. GRO chemokine mRNAs are unstable and stabilized by IL-1α in HeLa cells.

(A) HeLa tet-off cells were treated with TNF-α (10 ng/ml) for 2 h prior to addition of Act D (5 μg/ml), alone [no treatment (NT)] or with IL-1α (10 ng/ml) for the indicated times. Total RNA was prepared and used to determine levels of total GRO mRNA and GAPDH mRNA by Northern hybridization using a probe for the coding region of GRO1 (this detects all three GRO chemokines, as a result of high-sequence identity). Autoradiographs were quantified using Image J software, and half-life for total GRO under each condition was determined by linear regression. Results in the graph are the mean (listed above each bar) from three independent experiments ± sem. (B) HeLa tet-off cells were transfected with plasmids encoding the mCXCL1 reporter linked with FL versions of the 3′UTRs from each of GRO1, GRO2, or GRO3 mRNAs, as described in Materials and Methods. After overnight rest, the cultures were treated with dox (1 μg/ml), alone or in the presence of IL-1α for the indicated times prior to preparation of total RNA and analysis of reporter mRNA levels as described in A.

Because of the high degree of homology among the three GRO mRNAs and the differential expression of each version in specific cell types, we chose to examine the individual behavior of each sequence using a transgenic expression system. For this purpose, we used the tetracycline-regulated transcription system known as tet-off [35, 37]. Plasmid constructs were prepared in the pTRE2 vector that allows expression of a transgene under control of a tet-responsive promoter via the action of a constitutively expressed tet repressor VP16 fusion protein. When dox is added, the tetR-VP16-driven transcription is terminated, allowing assessment of the disappearance of specific mRNA in the absence of transcriptional poison, such as Act D. Constructs contained the reporter sequence (mCXCL1 5′UTR and coding region) linked with FL 3′UTRs from each of the three GRO mRNAs. In HeLa tet-off cells stably producing the tet-transactivator, these reporter RNAs were strongly expressed but decayed rapidly upon the addition of dox (Fig. 1B). A control construct, which contains the reporter sequence linked to the rabbit β-globin 3′UTR, is quite stable, demonstrating that the instability is encoded within the 3′UTRs of the GRO mRNAs. In cultures treated with dox along with IL-1α, however, the half-lives of mRNAs derived from all three constructs were prolonged substantially (roughly twofold). These results suggest that instability and stimulus sensitivity are encoded within the 3′UTRs of each GRO mRNA.

GRO2 and GRO3 3′UTRs exhibit comparable instability/IL-1α sensitivity features

As GRO1 and GRO2 3′UTRs are closely related and distinct from the GRO3 3′UTR, we focused further studies on comparison of GRO2 and GRO3. We prepared a series of tet-regulated constructs, in which distinct regions of GRO2 or GRO3 3′UTR were deleted and used these to identify regions contributing to message instability in the presence or absence of IL-1α. The sequences of the GRO2 and GRO3 3′UTRs with regions of interest annotated are provided in Supplemental Fig. 1. Sequential deletion of segments from the 5′ end of GRO2 3′UTR resulted in modest, incremental loss of instability, while retaining sensitivity for stabilization in response to IL-1α stimulation (Fig. 2A). The removal of a 38-nt segment from the Δ3 construct generating the Δ4 version, however, resulted in stable mRNA. Interestingly, deletions of the clustered and isolated AUUUA pentamers resulted in only modest reduction in instability, whereas sensitivity to IL-1α-mediated stabilization was maintained, suggesting that for GRO2, the major instability and stimulus sensitivity region is located in the very 3′ end of the message. This was confirmed with the GRO2Δ5 construct, in which a 174-nt segment was removed from the 3′ end of GRO2FL 3′UTR, leading to substantial loss of instability and complete abrogation of stimulus sensitivity. The analysis of the GRO3 3′UTR produced outcomes similar to those obtained with the GRO2 3′UTR constructs. Incremental deletion of regions containing the AUUUA pentamers from GRO3FL also reduced instability while retaining IL-1α sensitivity (Fig. 2B). Also similar to GRO2, deletion of a 60-nt segment from the Δ2 construct to create the Δ3 construct eliminated instability and IL-1α sensitivity, locating the stimulus-sensitive instability determinant in the distal 3′UTR segment. Interestingly, whereas GRO2Δ5 and GRO3Δ4 were insensitive to stabilization by IL-1α, they exhibited substantially different half-lives (180 vs. 60 min, respectively) that apparently result from differential function of the ARE motifs.

Figure 2. Localization of sequence regions contributing to GRO2 and GRO3 mRNA instability and IL-1α-induced stabilization.

(A) HeLa tet-off cells were transfected with plasmids encoding the indicated GRO2 3′UTR regions linked to the mCXCL1 coding region [for detailed information about the full GRO2 and GRO3 3′UTR sequences and the positions of the indicated pentamers (P) and deletions, see Supplemental Fig. 1]. After overnight rest, the cells were treated with dox (1 μg/ml), alone or along with IL-1α (10 ng/ml) for the indicated times prior to analysis of CXCL1 mRNA by Northern hybridization. Autoradiographs were quantified, as described in the legend to Fig. 1. (B) HeLa tet-off cells were transfected with plasmids encoding the indicated GRO3 3′UTR regions linked to the mCXCL1 coding region. After overnight rest, the cells were treated with dox (1 μg/ml), alone or along with IL-1α (10 ng/ml) for the indicated times prior to analysis of CXCL1 mRNA by Northern hybridization. Autoradiographs were quantified as described in the legend to Fig. 1.

The AUUUA pentamers are not involved in IL-1α-mediated stabilization of GRO2 and GRO3 mRNAs

As GRO2Δ3 and GRO3Δ2 mutants represent the minimal 3′UTR segments that confer instability to the reporter mRNA, while retaining sensitivity for IL-1α-driven stabilization, these constructs were chosen to further examine the role of AUUUA motifs as determinants of sensitivity to IL-1α. The importance of AUUUA pentamers in the GRO2Δ3 construct was tested by deleting a 28-nt segment from the 3′ end to generate GRO2Δ6, in which the one remaining AUUUA segment (P4) was removed. The mRNA derived from this plasmid retained instability, and its half-life could be prolonged in cells treated with IL-1α (Fig. 3). In the GRO3Δ2 construct, the sixth and seventh AUUUA pentamers (P6 and P7) were mutated to AGCUA, and as with the GRO2 pentamer deletion, GRO3Δ2Pmt remained unstable and sensitive to IL-1α stimulation (Fig. 3). Together, these results suggest that the Class I and II AREs in GRO2 and GRO3 3′UTRs are dispensable for sensitivity to IL-1α-driven stabilization.

Figure 3. AUUUA pentamer sequence motifs do not contribute to IL-1α-sensitive instability of GRO2 and GRO3 mRNAs.

HeLa tet-off cells were transfected with plasmids encoding mCXCL1 reporter genes containing the indicated WT and mutant regions from the GRO2 and GRO3 3′UTRs. After overnight rest, the cells were treated with dox (1 μg/ml), alone or along with IL-1α (10 ng/ml). After the indicated times, total RNA was prepared and used to determine CXCL1 and GAPDH mRNA levels. The autoradiographs were quantified as described in the legend to Fig. 1.

The pentamer-rich domains of GRO2 and GRO3 3′UTR confer sensitivity to TTP-mediated instability

TTP is a RNA-binding protein that exhibits binding specificity for AUUUA-containing sequences and has been shown to promote more rapid degradation of a number of ARE-containing mRNAs [38–40]. To determine whether TTP contributes to the instability of stimulus-sensitive regions of the GRO2 and GRO3 3′UTRs, HeLa tet-off cells were separately cotransfected with siRNA specific for TTP mRNA or control siRNA and one of several GRO2 or GRO3 constructs. Half-lives of pentamer-rich GRO2FL, GRO3FL, GRO2Δ5, and GRO3Δ4 mRNAs were markedly prolonged in cells where the TTP protein levels were depleted by specific siRNA compared with the nonspecific control siRNA (Fig. 4A and B). In contrast, the half-lives of mRNAs containing the pentamer-free GRO2Δ6 and GRO3Δ2Pmt fragments remained unchanged (Fig. 4A and B). Interestingly, treatment with IL-1αα markedly increased levels of TTP protein (Fig. 4C), but this change does not appear to be associated with any alteration in mRNA decay (see Fig. 2). The TTP-specific siRNA substantially diminished the amount of TTP protein compared with nonspecific control siRNA (Fig. 4D). TTP protein was able to bind to mRNAs containing the FL GRO2 and GRO3 3′UTR sequences, as determined by coimmunoprecipitation of TTP protein and reporter mRNA sequence (Fig. 4E). This binding was reduced very modestly in cells treated with IL-1α for 2 h (Fig. 4E). Taken together, these observations indicate that TTP contributes only to the instability of AUUUA-rich regions and is therefore not likely to be the target for IL-1α-mediated stabilization.

Figure 4. Distinct instability motifs within GRO2 and GRO3 3′UTRs are differentially sensitive to the action of TTP.

(A) HeLa tet-off cells were transiently cotransfected with plasmids encoding GRO2 3′UTR regions (GRO2FL, GRO2Δ5, GRO2Δ6) and control or TTP-specific siRNA. After 48 h, the cells were treated with dox (1 μg/ml), and total RNA was prepared after the indicated times and used to determine levels of mCXCL1 or GAPDH mRNAs by Northern hybridization. Autoradiographs were quantified as described in the legend to Fig. 1. (B) HeLa tet-off cells were transiently cotransfected with plasmids encoding GRO3 3′UTR regions (GRO3FL, GRO3Δ4, GRO3Δ2Pmt) and control or TTP-specific siRNA. After 48 h, the cells were treated with dox (1 μg/ml), and total RNA was prepared after the indicated times and used to determine levels of mCXCL1 or GAPDH mRNAs by Northern hybridization. Autoradiographs were quantified as described in the legend to Fig. 1. (C) HeLa tet-off cells were treated with dox (1 μg/ml), alone or with IL-1α (10 ng/ml) for the indicated times prior to preparation of whole-cell lysates and analysis of TTP protein levels by Western blot. Rx, treatment. (D) HeLa tet-off cells were cotransfected with control (CTRL) or TTP-specific siRNA and GRO2FL reporter. After 48 h, whole-cell lysates were prepared and used to determine levels of TTP protein by Western blot. (E) HeLa tet-off cells were transfected with the GRO2FL or GRO3FL reporter constructs. After overnight rest, the cells were left untreated or treated for 60 min with IL-1α (10 ng/ml). Whole-cell lysates were prepared following fixation and subjected to analysis for RNA-TTP binding using immunoprecipitation with nonspecific IgG or anti-TTP antibodies, followed by real-time RT-PCR analyses of GRO2FL and GRO3FL containing mRNAs. Results are presented as mean fold enrichment of reporter mRNA in the immunoprecipitations (IPs) using specific versus nonspecific IgG ± one-half range of duplicate determinations.

KSRP promotes degradation of AUUUA-containing regions of GRO2 and GRO3 mRNAs

KSRP has been recently shown to bind AU-rich sequences and to promote the decay of several ARE-containing mRNAs, including the CXC ELR+ family member IL-8 [41]. To determine the role of KSRP in stimulus-sensitiveGRO mRNA turnover, the effect of RNAi-mediated depletion of KSRP on the half-life of the GRO construct-derived mRNAs was assessed using siRNA, similar to the strategy used for examining TTP. HeLa tet-off cells were cotransfected separately with siRNA specific for KSRP mRNA or control siRNA and each of the GRO constructs (GRO2FL, GRO2Δ5, GRO2Δ6, GRO3FL, GRO3Δ4, or GRO3Δ2Pmt). The KSRP-specific siRNA strongly diminished the amount of KSRP protein compared with nonspecific control siRNA (Fig. 5D). The half-lives of the AUUUA-containing mRNAs, derived from GRO3FL, GRO2Δ5, and GRO3Δ4 plasmids, were prolonged in cells depleted of KSRP compared with the control siRNA-treated cells. The half-life of GRO2FL-derived mRNA did not show significant difference, although this may reflect the very short half-life (<30 min). In contrast the half-lives of the GRO2Δ6 and GRO3Δ2Pmt derivatives remained unchanged (Fig. 5, A and B). Treatment with IL-1α did not alter the levels of KSRP protein (Fig. 5C). As was observed for TTP, KSRP was able to bind with GRO2FL- and GRO3FL-derived mRNAs, but this was not appreciably altered in IL-1α-treated cells (Fig. 5E).

Figure 5. Distinct instability motifs within GRO2 and GRO3 3′UTRs are differentially sensitive to the action of KSRP.

(A) HeLa tet-off cells were cotransfected with control or KSRP-specific siRNA and one of the GRO2 reporter plasmids (GRO2FL, GRO2Δ5, GRO2Δ6). After 48 h, the cells were treated with dox (1 μg/ml) for the indicated times prior to determination of mCXCL1 or GAPDH mRNA levels by Northern hybridization. The autoradiographs were quantified as described in the legend to Fig. 1. The results are presented as the mean ± 1 sd of triplicate determinations. (B) HeLa tet-off cells were cotransfected with control or KSRP-specific siRNA and one of the GRO3 reporter plasmids (GRO3FL, GRO3Δ4, GRO3Δ2Pmt). After 48 h, the cells were treated with dox (1 μg/ml) for the indicated times prior to determination of mCXCL1 or GAPDH mRNA levels by Northern hybridization. The autoradiographs were quantified as described in the legend to Fig. 1. The results are presented as the mean ± 1 sd of triplicate determinations. Levels of KSRP protein were determined by Western blot analysis. (C) HeLa tet-off cells were treated with dox (1 μg/ml), alone or with IL-1α (10 ng/ml) for the indicated times prior to preparation of whole-cell lysates and analysis of KSRP protein levels by Western blot. (D) HeLa tet-off cells were cotransfected with control or KSRP-specific siRNA and GRO2FL reporter. After 48 h, whole-cell lysates were prepared and used to determine levels of KSRP protein by Western blot. (E) HeLa tet-off cells, transfected with the GRO2FL or GRO3FL reporter constructs, were left untreated or treated for 60 min with IL-1α (10 ng/ml). Whole-cell lysates were prepared following fixation and subjected to analysis for RNA-KSRP binding using immunoprecipitation with nonspecific IgG or anti-KSRP antibodies, followed by real-time RT-PCR analyses of GRO2FL- and GRO3FL-derived mRNA. Results are presented as mean fold enrichment of reporter mRNA in the IPs using specific versus nonspecific IgG ± one-half range of duplicate determinations.

The SF2/ASF selectively destabilizes the ARE-independent GRO-instability determinants

We identified recently the SRSF SF2/ASF as a mediator of instability for the mCXCL1 mRNA [23]. As the stimulus-sensitive instability determinant in the mCXCL1 3′UTR is not dependent on AUUUA-containing AREs, we reasoned that SF2/ASF might also target the IL-1α-sensitive instability determinant in the GRO2 and GRO3 3′UTRs, particularly in light of the findings that these regions are not sensitive to TTP or KSRP (see Figs. 4 and 5). As SF2 is essential for cell viability, we were unable to use siRNA-mediated depletion effectively. Hence, to test this hypothesis, we cotransfected HeLa cells with an expression plasmid encoding human SF2/ASF and one of the GRO reporter mRNAs (GRO2FL, GRO3FL, GRO2Δ3, GRO3Δ2, GRO2Δ5, or GRO3Δ4). Whereas mRNAs from the pentamer-free constructs (GRO2Δ6 and GRO3Δ2Pmt) showed markedly shorter half-lives in cells overexpressing SF2/ASF, this effect was sequence-specific, as mRNAs derived from the pentamer-containing constructs (GRO2Δ5 and GRO3Δ4) were unaffected (Fig. 6A and B). Although mRNAs from GRO2FL and GRO3FL did not show any sensitivity to overexpression of SF2, the half-lives for these mRNAs are so short that further reduction could not be measured. IL-1α treatment did not affect endogenous SF2 protein levels (Fig. 6D). Importantly, SF2 protein could effectively bind the GRO2 and GRO3 3′UTR sequences, but in contrast to the findings with TTP and KSRP, treatment with IL-1α did produce a more substantial reduction in SF2 binding (Fig. 6E). Taken together, these observations demonstrate that SF2 has the capacity to promote AUUUA pentamer-independent instability. Moreover, the finding that SF2 interaction with GRO2 and GRO3 3′UTRs is reduced by treatment with IL-1α suggests that its dissociation from GRO mRNA results in stabilization.

Figure 6. Distinct instability motifs within GRO2 and GRO3 3′UTRs are differentially sensitive to the action of SF2/ASF.

(A) HeLa tet-off cells were transiently cotransfected with plasmids encoding GRO2 3′UTR regions (GRO2FL, GRO2Δ5, GRO2Δ6) and empty-vector (pcDNA3) or plasmid-encoding FL mouse SF2/ASF. After overnight rest, the cells were treated with dox (1 μg/ml), and RNA was prepared after the indicated times and used to determine levels of mCXCL1 or GAPDH mRNAs by Northern hybridization. Autoradiographs were quantified as described in the legend to Fig. 1, and the results presented are the mean ± one-half range. (B) HeLa tet-off cells were transiently cotransfected with plasmids encoding GRO3 3′UTR regions (GRO3FL, GRO3Δ4, GRO3Δ2Pmt) and empty vector (pcDNA3) or plasmid-encoding FL mouse SF2/ASF. After overnight rest, the cells were treated with dox (1 μg/ml), and RNA was prepared after the indicated times and used to determine levels of mCXCL1 or GAPDH mRNAs by Northern hybridization. Autoradiographs were quantified, as described in the legend to Fig. 1, and the results presented are the mean ± one-half range. (C) HeLa tet-off cells were treated with dox (1 μg/ml), alone or with IL-1α (10 ng/ml) for the indicated times prior to preparation of whole-cell lysates and analysis of SF2 protein levels by Western blot. (D) HeLa tet-off cells were cotransfected with pcDNA3 or SF2 and GRO2FL reporter. After overnight rest, whole-cell lysates were prepared and used to determine levels of SF2 protein by Western blot. (E) HeLa tet-off cells, transfected with the GRO2FL or GRO3FL reporter constructs, were left untreated or treated for 60 min with IL-1α (10 ng/ml). Whole-cell lysates were prepared following fixation and subjected to analysis for RNA-SF2 binding using immunoprecipitation with nonspecific IgG or anti-SF2 antibodies, followed by real-time RT-PCR analyses of GRO2FL- and GRO3FL-derived mRNA. Results are presented as mean fold enrichment of reporter mRNA in the IPs using specific versus nonspecific IgG ± one-half range of duplicate determinations.

DISCUSSION

Regulation of the expression of many inflammatory cytokines and chemokines is known to involve control, not only of the rate of mRNA synthesis (i.e., transcription) but also the rate of mRNA degradation or half-life [8, 10]. Moreover, the rapid degradation of mRNA can be abrogated in response to extracellular stimuli to enable sufficient accumulation of such messages. Whereas it is now well-established that specific sequences within the 3′UTR of such mRNAs confer these regulatory behaviors, the mechanistic diversity of such control is emerging in greater detail. In the present study, we evaluated determinants contributing to the control of GRO (CXCL1, -2, and -3) chemokine mRNA half-lives with particular emphasis on such behavior in nonmyeloid cells. The results support the following conclusions: (1) the 3′UTRs of all three GRO mRNAs confer instability and sensitivity to IL-1α stimulation for prolongation of half-life; (2) GRO2 and GRO3 mRNAs exhibit multiple AUUUA pentamer sequences in the 5′ proximal portion of their 3′ UTRs, and these regions confer sensitivity for enhanced decay via the action of TTP and KSRP, RNA-binding proteins with known specificity for ARE sequences; importantly, these instability determinants are not stabilized in response to IL-1α nor does IL-1α modulate the binding of TTP or KSRP to GRO2 or GRO3 3′UTRs; and (3) GRO2 and GRO3 3′UTRs also contain a 3′ proximal domain, contributing instability that can be stabilized in response to stimulation with IL-1α and not requiring the well-recognized ARE motif AUUUA. These 3′ proximal instability determinants show selective sensitivity for enhanced decay to the RNA SF2/ASF, recently shown to promote instability for mCXCL1 mRNA [23]. Moreover, the interaction of SF2/ASF with GRO mRNAs is reduced appreciably in cells treated with IL-1α.

The role of AUUUA and related sequences in determining mRNA instability and the ability of extracellular stimuli to promote prolongation of cytokine and growth factor mRNA half-life is well established [8, 10]. The presence of multiple copies of the AUUUA pentamer sequence within the 3′UTRs of the GRO chemokines predicts that these sites would be important in determining instability and would also govern the sensitivity for stimulus-driven stabilization. Thus, it was surprising to find that much of the instability and all sensitivity for enhanced stability in response to IL-1α are provided by GRO2 and GRO3 3′UTR fragments that do not contain AUUUA sequences, as evidenced by deletion or mutation. Of particular interest is the similarity of the findings reported here with prior work examining the sequence dependence for instability and stimulus sensitivity (IL-17) of the mCXCL1 mRNA in nonmyeloid cell populations [21, 23]. As with GRO2 and GRO3, the sequence region conferring IL-17-sensitive instability in mCXCL1 does not contain or require AUUUA pentamer motifs and is localized in the terminus of the 3′ UTR of the message. Moreover, the mCXCL1 stimulus-sensitive instability motif is not sensitive to TTP or KSRP but does show selective sensitivity to the SF2/ASF. These similarities suggest that such features represent a common paradigm for the post-transcriptional control of the ELR-CXC chemokine family in nonmyeloid cell settings. However, comparison of these sequence regions does not reveal any common motifs. Surprisingly, perhaps, the IL-8 mRNA does not appear to exhibit these features and instead, the major instability and the sensitivity to TLR and IL-1R signaling are dependent on a motif in the more 5′ portion of the 3′ UTR that is enriched for AUUUA sequence elements [41–43].

An important finding in the present study is that the GRO mRNA 3′UTRs possess at least two distinct regions conferring potential for instability. The region conferring AUUUA-independent, IL-1α-sensitive instability, discussed above, is localized toward the distal end of the message. The second region of the GRO 3′UTRs, containing AUUUA instability determinants, is proximal to the coding region. As mentioned previously, we anticipated that deletion of this region would abrogate instability and also serve as the target for the cytokine-sensitive stabilization functions. The negative experimental findings, however, directed our attention toward the behavior of these AUUUA-rich portions of GRO 3′UTRs in the context of known, destabilizing RNA-binding proteins TTP and KSRP. GRO2 and GRO3 pentamer-rich segments conferred sensitivity to TTP, as their half-lives were increased markedly in HeLa cells, where the expression of TTP was depleted by transfection of specific siRNAs. Also, KSRP appears to contribute to the instability seen with reporter mRNAs containing the GRO2 and GRO3 ARE regions. Although the half-lives differ substantially for mRNAs containing the ARE regions of the two different GRO mRNAs, both are prolonged in cells transfected with siRNA specific for KSRP. Neither TTP nor KSRP depletion, however, impacts on the instability driven by the IL-1α-sensitive regions. This pattern was also observed in mRNA protein-binding experiments. Whereas all three of the tested proteins (TTP, KSRP, and SF2) exhibited strong binding to GRO mRNAs, only SF2 binding was reduced convincingly by IL-1α treatment, suggesting that stimulus-induced dissociation of SF2 from GRO mRNA is responsible for its stabilization.

The behavior of GRO mRNAs described in this study largely reflects patterns observed within nonmyeloid cells. These patterns can be distinguished from those that have been characterized previously within myeloid cells based on the requirements for AUUUA elements, the participation of the TRAF6/p38 MAPK pathway, and TTP. Although the specific mechanisms regulating the CXC chemokine mRNA half-life do not operate in a mutually exclusive fashion in myeloid and nonmyeloid cells (e.g., are not absolutely cell type-specific), it seems apparent that multiple pathways exist to control the degradation of cytokine and chemokine mRNAs that may be used differentially, in part, based on cell type. Thus for mCXCL1 and now the human GROs, we have identified distinct AUUUA-independent regions of the 3′UTR that provide stimulus-sensitive instability and differential roles for RNA-binding proteins that provide target recognition specificity. The use of these different mechanisms by nonmyeloid cells may reflect the relative expression of specific components (e.g., TTP), as well as the distinct role of epithelial and stromal cells in the initial stages of inflammatory response, where neutrophil-targeting chemokine production is an important and early event. Furthermore, encoding multiple regulatory mechanisms within a single gene provides opportunity for diversity in the control of gene expression under different physiologic and pathophysiologic conditions.

AUTHORSHIP

M.N. contributed to study performance. T. H. contributed to study conception, design, performance, and preparation of the manuscript. T.A.H. contributed to study conception, design, and preparation of the manuscript.