Tristetraprolin (TTP) coordinately regulates primary and secondary cellular responses to proinflammatory stimuli

TTP-promoted decay of transcripts is a critical post-transcriptional regulatory mechanism in the response of fibroblasts to TNF released from primary immune cells.

Abstract

TTP is an anti-inflammatory protein that acts by binding to AREs in its target mRNAs, such as Tnf mRNA, and promoting their deadenylation and decay. TNF released from inflammatory cells can then stimulate gene expression in tissue cells, such as fibroblasts. To determine whether TTP could affect the decay of TNF-induced transcripts in fibroblasts, we exposed primary embryonic fibroblasts and stable fibroblast cell lines, derived from WT and TTP KO mice, to TNF. The decay rates of transcripts encoded by several early-response genes, including Cxcl1, Cxcl2, Ier3, Ptgs2, and Lif, were significantly slowed in TTP-deficient fibroblasts after TNF stimulation. These changes were associated with TTP-dependent increases in CXCL1, CXCL2, and IER3 protein levels. The TTP-susceptible transcripts contained multiple, conserved, closely spaced, potential TTP binding sites in their 3′-UTRs. WT TTP, but not a nonbinding TTP zinc finger mutant, bound to RNA probes that were based on the mRNA sequences of Cxcl1, Cxcl2, Ptgs2, and Lif. TTP-promoted decay of transcripts encoding chemokines and other proinflammatory mediators is thus a critical post-transcriptional regulatory mechanism in the response of secondary cells, such as fibroblasts, to TNF released from primary immune cells.

Introduction

TTP is the best-characterized member of a small family of CCCH tandem zinc finger proteins. TTP and its widely distributed family members, ZFP36L1 and ZFP36L2, as well as the rodent- and placenta-specific protein ZFP36L3, can bind to AREs at a consensus nonamer site, UUAUUUAUU, in 3′-UTRs of certain mRNAs, leading to their accelerated deadenylation and degradation [1–4]. Tnf mRNA is the first-discovered and best-studied example of a TTP target transcript [5–7]. Its mRNA has a highly conserved ARE that contains multiple overlapping copies of the consensus TTP binding motif in its 3′-UTR [6, 7]. As a result of abnormal TNF transcript stabilization and the resulting protein hyperproduction, TTP-deficient mice exhibit many of the hallmarks of systemic TNF excess, including severe arthritis, cachexia, dermatitis, autoimmunity, and myeloid hyperplasia [5]. Virtually the entire syndrome can be prevented by repeated injections of anti-TNF antibodies into newborn TTP KO mice [5] or by interbreeding the KO mice with mice lacking both types of TNFR [8].

TNF is best known as the toxic mediator of septic shock and as a more subacute cause of chronic diseases, such as rheumatoid arthritis, Crohn’s disease, ankylosing spondylitis, psoriasis, and others. In many of these conditions, direct neutralization of TNF with use of recombinant antibodies and binding proteins has been demonstrated to be therapeutically effective. Less well appreciated is the role of TNF in stimulating the expression of genes in secondary cells, such as fibroblasts, at local sites of inflammation; these genes often encode other inflammatory mediators and chemokines, some of which can promote the recruitment of leukocytes into the inflammation site. These leukocytes can, in turn, release additional proinflammatory mediators, including TNF itself, as part of the concerted response to local inflammation. We have found that mice with the TTP deficiency syndrome exhibit many such sites of increased local leukocyte infiltration, exemplified in our recent finding of leukocyte infiltration into the left-sided heart valves of TTP-deficient mice, a phenomenon apparently mediated by TNF [9]. We proposed in that paper that heightened, local TNF production by resident macrophages and related cells could stimulate local leukocyte infiltration by activating nonimmune stromal cells, such as fibroblasts, followed by increased production of chemoattractants and chemokines.

Fibroblasts are resident cells that are well known for their activities in extracellular matrix production and structure support. Recent evidence has highlighted important roles for fibroblasts as secondary response cells in the innate immune response, particularly as sources of chemoattractant cytokine secretion [10]. Recently, Hao and Baltimore [11] categorized the induction characteristics of TNF-responsive genes in cultured mouse fibroblasts. Many of the most rapidly and transiently induced genes encoded cytokines and chemokines, and many of these rapidly induced mRNAs contained AREs in their 3′-UTRs. These authors found that mRNA stability, apparently controlled, at least in part, by the ARE contents of 3′-UTRs, was a key determinant of the kinetics of gene expression induced by TNF.

In this study, we investigated the possibility that TTP could facilitate the decay of early and intermediate response transcripts following TNF stimulation in mouse fibroblasts. Our major goal was to explore a potential role for TTP in the secondary cellular response to TNF released from macrophages and related “first responder” cells in response to primary stimuli, such as bacterial endotoxin. We were also interested in identifying novel physiologic target transcripts of TTP. To accomplish these goals, we analyzed turnover rates of TNF-induced transcripts in paired TTP-deficient (KO) fibroblasts and WT control cells, examined the sequences of these for potential TTP binding sites, and confirmed TTP binding to several transcripts that satisfied these criteria. Several of the target transcripts identified by this means encode chemokines or in other ways, participate in the recruitment, activation, and maturation of leukocytes in the sites of local inflammation. We conclude from these data that TTP not only regulates the primary cellular response to innate immune stimuli, exemplified by TNF elaborated from macrophages, but also regulates the expression of transcripts involved in the secondary response of mesenchymal cells to the TNF stimulation.

MATERIALS AND METHODS

Fibroblast cell culture and stimulation

Primary cultures of MEFs were prepared from embryos at E15.5 of gestation, where E0.5 was the date of detection of the vaginal plug. The embryos were the products of matings of Zfp36 ^+/− mice [5]; these mice have been maintained on a C57BL/6 background for >20 generations. The genotypes for individual fetuses from each litter were determined by evaluation of tail DNA from each embryo [5]. MEFs, at passages 2–4, were used in these experiments. Stable mouse fibroblast cell lines were derived from MEF cultures from E14.5 TTP KO and littermate WT embryos, as described previously [1, 12]. These stable cell lines have been cultured for>200 passages and are well matched in terms of growth rates, morphology, and responses of rapidly inducible genes, such as Fos, to serum stimulation [1]. Both primary MEFs and stable fibroblast cell lines were maintained at 37°C (5% CO₂) in DMEM (Invitrogen, Carlsbad, CA, USA) containing 10% FBS (Invitrogen), 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 mM L-glutamine. For serum deprivation experiments, cells at 70–80% confluence were washed once in serum-free DMEM and then incubated for at least 16 h in DMEM containing 0.5% (vol/vol) FBS [1, 12]. The cells were then stimulated with 10 ng/ml mouse TNF (R&D Systems, Minneapolis, MN, USA) for various times. As shown previously, these mouse fibroblasts do not express Tnf, Csf2, or Il2 mRNAs at levels readily detectable by routine Northern blotting techniques [1]. All of the animal breeding and procedures were approved by the Institutional Animal Care and Use Committee of the National Institute of Environmental Health Sciences.

Culture and differentiation of BMDM

BM cells were isolated from the femurs of 8- to 12-week-old M-TTP KO mice and their littermate WT controls. Culture and differentiation of BM cells were conducted in RPMI-1640 medium supplemented with 30% (vol/vol) L929 conditioned medium, as described previously [13, 14]. Differentiated BMDMs were harvested by gently scraping the cells from the dishes by use of a rubber “policeman” and seeded at 4 × 10⁴/well into 24-well culture plates for experiments. After cells reached 70–80% confluence, cells were subjected to serum starvation with RPMI-1640 medium containing 1% FBS for at least 20 h before stimulation with TLR agonists. Endotoxin-free TLR agonists Pam₃CSK₄ (TLR1/2), LTA-SA (TLR2), HKLM (TLR2), PIC low molecular weight (TLR3), and imiquimod (R387; TLR7) were purchased from InvivoGen (San Diego, CA, USA). LPS (serotype 055:B5; TLR4) was obtained from Sigma (St. Louis, MO, USA). Culture supernatants were collected after 24 h treatment and stored at −20°C for cytokine measurements. In the experiments for transcript analysis, cells were subjected to similar treatments and harvested at hourly intervals for a total of 4 h [13, 14]. Protein lysates were prepared from BMDM plated into 100 mm dishes, as described previously [14].

RNA extraction, Northern blotting, and real-time RT-PCR

Total cellular RNA was isolated by use of the illustra RNAspin Mini RNA Isolation Kit (GE Healthcare, Piscataway, NJ, USA), according to the manufacturer’s instructions, and residual genomic DNA was removed by on-column digestion with RNase-free DNase I (GE Healthcare). First-strand cDNAs were synthesized by use of oligo(dT)_12–18 primers and SuperScript II RT (Invitrogen). Northern blotting was performed as described previously [1]. Real-time PCR was performed by use of SYBR Green and the ABI Prism 7900 Sequence Detection System (Applied Biosystems, Grand Island, NY, USA). Results were normalized with the Actb transcript as an internal control and were then used to calculate expression levels according to the ΔΔ comparative threshold method [15]. All data were expressed in terms of fold change relative to the data from the unstimulated WT samples, which were set as 1 unless specified otherwise. The primers used were validated for their amplification efficiency and specificity before being used in the study; primer sequences are listed in Supplemental Table 1.

Measurement of mRNA decay

To determine the decay rates of selected transcripts, actinomycin D (10 μg/ml; Sigma) was added directly to the cell cultures that had been serum deprived overnight with 0.5% (vol/vol) FBS, as described above, and then stimulated with mouse rTNF for 30 min; the TNF was not removed at the time of actinomycin D addition. Cells were harvested at 10, 20, 30, 45, 60, 90, and 120 min after the addition of actinomycin D; transcript levels were measured by real-time RT-PCR as described above, and results were normalized to those of Actb mRNA. In all experiments, each sample represented 3 combined, 60-mm dishes of cells.

RNA EMSAs

Biotinylated RNA ARE probes were designed from the 3′-ARE sequences of respective transcripts and synthesized with 5′ biotin end-labeling (Invitrogen). Transfection of HEK 293 cells with the hTTP expression plasmid CMV.hTTP.HA and its zinc finger mutant C124R (a single amino acid mutation at position 124) and preparation of cell extracts were performed as described previously [7, 16]. Approximately 1 µg cytosolic protein, prepared from HEK 293 cells transfected with vector alone or with expression constructs driven by the CMV promoter, was incubated with 0.6 ng 5′ biotin-labeled mouse ARE probes at room temperature for 30 min in a 20 µl total reaction volume consisting of 10 mM HEPES (pH 7.6), 40 mM KCl, 2.5% (v/v) glycerol, 3 mM MgCl₂, 2.5 µg/µl heparin, and 50 ng/µl yeast tRNA. The reaction mixture was then loaded onto a 6–8% nondenaturing acrylamide gel and subjected to electrophoresis at 160 V for 90 min in 0.4× TBE buffer. Gels were transferred to Biodyne B nylon membranes (0.45 µm; Thermo Scientific, Rockford, IL, USA) in 0.4× TBE buffer at 80 V for 1 h. Unbound probe and RNA protein binding complexes were detected with the Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific), per the manufacturer's instructions, and were exposed to BioMax MR film from Kodak. Supershift experiments involving HA-tagged hTTP were performed by preincubation of the cell lysates with anti-HA antibody (clone Y-11; 1:5 dilution; Santa Cruz Biotechnology, Santa Cruz, CA, USA) on ice for 10 min, followed by the addition of biotinylated RNA ARE probes.

Western blotting

Whole-cell extracts were prepared in a hypertonic lysis buffer consisting of 100 mM Tris-HCl (pH 8.0), 0.5 M NaCl, 5 mM EDTA, 1.25% NP-40, and 1 protease inhibitor cocktail tablet/10 ml buffer (Roche, Indianapolis, IN, USA). In some experiments, cytosolic extracts were prepared in radioimmunoprecipitation assay buffer [50 mM Tris (pH 8.0), 150 mM NaCl, 1% NP-40, 5 mM EDTA, and 5% glycerol] containing protease inhibitors (Roche). Protein concentrations were determined by the Bradford assay (Bio-Rad Laboratories, Hercules, CA, USA). For the detection of the endogenous mouse TTP protein, 180 µg protein in SDS-denatured cell lysates was separated on 10–20% Criterion Tris-HCl precast gels (Bio-Rad Laboratories) and transferred onto nitrocellulose membranes. Blots were incubated at 4°C overnight with a rabbit antiserum raised against a mouse rTTP-maltose binding protein fusion protein [1, 17]. Goat polyclonal anti-immediate early-response gene X-1, mouse monoclonal anti-HuR, rabbit polyclonal anti-PTGS2, and mouse monoclonal anti-ACTB were from Santa Cruz Biotechnology, Cayman Chemical (Ann Arbor, MI, USA), and Abcam (Cambridge, MA, USA), respectively. The protein levels were visualized with SuperSignal West Pico chemiluminescent substrate (Thermo Scientific).

Biotinylated RNA-protein pull-down assays

Protein pull-downs with the use of biotinylated RNA ARE probes and streptavidin-coated magnetic beads were performed according to the manufacturer’s instructions for the Pierce Magnet RNA-Protein Pull-Down Kit (Thermo Scientific) with some modifications. In brief, 100 pmol biotinylated RNA ARE probes were incubated with prewashed, streptavidin-coated magnetic beads (40 or 50 μl; Thermo Scientific) at room temperature for 30 min. After washes, 500 or 900 μg cytosolic lysate protein, prepared from serum or TNF-treated, primary-cultured MEFs in polysome lysis buffer, consisting of 100 mM KCl, 5 mM MgCl₂, 10 mM HEPES (pH 7.0), 0.5% NP-40, 20% (v/v) glycerol, 10 mM DTT, 100 U/ml RNase inhibitor (Applied Biosystems), and 400 μM vanadyl ribonucleoside complexes (Sigma), supplemented with protease inhibitors (Roche), was added into the RNA-streptavidin bead complexes in the gel-shift binding buffer and incubated at 4°C for 3 h under constant rotation. The RNA protein binding complexes immobilized on the streptavidin beads were then washed 4 times in wash buffer, supplemented with 10 mM (Ptgs2 ARE probe), 250 mM (Lif ARE probe), 300 mM (Cxcl1 ARE probe), or 500 mM NaCl (Cxcl2 ARE probe). The pull-down proteins were then eluted in 30 μl 2× Laemmli sample buffer (Bio-Rad Laboratories), supplemented with β-ME and analyzed by Western blotting for the presence of TTP and HuR. HuR protein, a highly conserved ARE binding protein, served as a loading control. Serum-treated MEF lysates were used as positive controls for the pull-downs, and the A → C mutant probes were used as negative controls for the validation of binding specificity by TTP.

Immunohistochemistry

Tissues from TTP KO mice and triple KO mice deficient in TTP and both types of TNFRs [8] were fixed in 10% phosphate-buffered formalin and processed by use of standard procedures. Paraffin sections of tissues (5 μm) were immunostained with anti-CD45 antibody, as described previously [14]. Immunolabeled complexes were visualized by use of diaminobenzidine. The sections were lightly counterstained with hematoxylin before analysis.

ELISA assays

Sandwich ELISA assays for CXCL1/KC and CXCL2/MIP2 production into the cell culture medium were performed by use of the mouse DuoSet ELISA Development kits (R&D Systems), according to the manufacturer’s instructions. These assays are claimed to have sensitivities of 7.8 pg/ml, with no significant cross-reactivity with other known mouse chemokines. Measurement of LIF protein levels in the cell culture supernatant was performed by use of the Quantikine mouse LIF immunoassay kit (R&D Systems), with a lowest detection limit of 21.9 pg/ml. Concentrations of TNF in the culture supernatants released from primary macrophages were determined by sandwich ELISA, as described previously [14].

Statistical analysis

Data were expressed as mean ± sem. Differences in the average normalized transcript levels between the WT and TTP KO fibroblast cells were evaluated by paired Student's t-test, and P < 0.05 was considered statistically significant. In the mRNA decay experiments, the transcript turnover rates were calculated based on the nonlinear fit 1-phase exponential decay curves by use of GraphPad Prism software version 6.0 (GraphPad Software, San Diego, CA, USA). The decay rates of transcripts, expressed as times to 50% mRNA decay, were calculated for each experiment, and the differences between WT and KO groups were compared by 2-tailed paired Student’s t-test from 5 individual experiments.

RESULTS

Overproduction of TNF in TTP-deficient macrophages upon exposure to different classes of TLR agonists

We first determined the role of TTP in modulating the expression of TNF from macrophages stimulated with different types of TLR agonists. The agonists used were derived from bacterial or viral membrane or cell-wall components or their nucleic acids (Fig. 1). As expected, there was little detectable TNF in the culture supernatants of BMDM before stimulation. After 24 h of treatment, TNF accumulation was stimulated most dramatically by the 2 TLR2 ligands: LTA and HKLM. It was also readily stimulated by LPS in a dose-dependent manner. Less TNF production was observed in the Pam₃CSK₄ treatment group, representing a TLR1/2 ligand, and minimal increases in TTP-dependent TNF production were noted in the imiquimod (R387) and PIC groups. Remarkably, TTP deficiency in these cells resulted in dramatic, statistically significant increases in the levels of TNF in the culture supernatants from the WT cells for almost all of the TLR challenge groups, with fold changes ranging from 3.3- to 17.1-fold increases when compared with control cells (Fig. 1A). We concluded that not only LPS, as an activator of TLR4 pathways, but also activators of TLR1 and TLR2 pathways could stimulate TNF secretion from these macrophages in a manner that was greatly exaggerated by the TTP deficiency. The induction of TTP mRNA and protein was confirmed in the BMDM with some selected TLR agonists, including LTA, imiquimod, and PIC (Fig. 1B and C). When similar experiments were performed with LPS-stimulated macrophages also deleted in both types of TNFRs, the cells lacking TTP and both TNFRs secreted approximately as much TNF as the single TTP KO cells, suggesting that the effects of secreted TNF on further TNF production were minimal under these conditions (Supplemental Fig. 1).

Figure 1. TTP and TNF induction in WT and TTP-deficient macrophages after stimulation with TLR agonists.

(A) M-TTP KO macrophages and their control WT cells were treated with Pam₃CSK₄, affecting the TLR1/2 receptors, LTA-SA (TLR2), HKLM (TLR2), PIC (TLR3), imiquimod (R387; TLR7), or LPS (TLR4) at the indicated doses in serum-free RPMI-1640 media for 24 h, and the culture supernatants were collected for cytokine measurement by ELISA. Results shown are representative of mean ± sem of 3–5 independent experiments; data for each experiment were obtained from cells in triplicate wells. TNF protein levels in the media of unstimulated WT and M-TTP KO macrophages were not detectable (data not shown). *P < 0.05, and **P < 0.01 when comparing values from WT versus M-TTP KO cells. (B) TTP mRNA induction was measured in WT and M-TTP KO macrophages after stimulation with 5 μg/ml LTA, 100 ng/ml LPS, 100 μg/ml PIC, and 5 μg/ml R387 at the indicated time-points. The results show the average fold inductions of TTP mRNA from 3–4 independent experiments and are expressed as fold increases relative to the WT Time 0 samples, which were set at 1 after normalization to the internal control Actb mRNA. (C) TTP protein production was examined in WT and M-TTP KO macrophages after 6 h treatment with the TLR agonists at the indicated concentrations by use of Western blotting. ACTB served as a loading control.

Leukocyte infiltration in nonlymphoid organs of TTP KO mice

TTP deficiency is associated with local infiltration of leukocytes in many tissues, exemplified by the left-sided cardiac valvulitis observed in these mice, a phenomenon that appears to be dependent on intact TNFRs [9]. As another example of this phenomenon, we examined the extent of leukocyte infiltration into myocardium of TTP KO mice and “triple KO” mice that are deficient in TTP and 2 two types of TNFRs by use of the anti-CD45 antibody as a general leukocyte marker (Fig. 2). Minimal leukocyte staining was observed in the myocardium of WT mice, whereas areas of dramatic leukocyte infiltration were seen scattered throughout the myocardium of the TTP KO mice. Foci of leukocyte infiltration were similarly observed in several other nonlymphoid tissues, such as liver, lung, and kidney (Fig. 2). These infiltrates were not seen in the mice deficient in TTP and both TNFRs (triple KO), suggesting that TNF activity is a major contributor to the local leukocyte infiltration seen in the absence of TTP.

Figure 2. Leukocyte infiltration in nonlymphoid tissues of TTP KO mice.

Representative photomicrographs are shown for the immunostaining of CD45-positive leukocytes (brown) in the ventricular myocardium; liver, lung, and kidney from WT female mice at 5 months of age (heart), or male mice at 3 months of age (liver, lung, and kidney); their TTP KO littermates; and mice of the same age, sex, and genetic background in which TTP and both types of TNFR genes were deleted (Triple KO). Sections were counterstained with hematoxylin. Original magnification, ×20 (heart, liver) or ×40 (lung, kidney).

TNF-stimulated induction of TTP in mouse fibroblasts

We first determined the time course of TTP mRNA expression after stimulation of primary MEFs and stable fibroblast cell lines with mouse rTNF. TTP induction after TNF stimulation was reported earlier in macrophages [6], and it was a member of the most rapidly induced set of genes responding to TNF in 3T3 MEF cells [11]. Under our experimental conditions, TNF caused a rapid increase of TTP mRNA in the WT stable fibroblast cell line (Fig. 3A and C) and in the WT primary MEFs (Fig. 3B), with peak mRNA levels observed at ∼30 min, decreasing rapidly to basal levels within 60 min (Fig. 3A–C). In both experiments, primer pairs and probes were used that would detect the Zfp36-neo fusion KO transcript described previously in these mice [5]; this transcript was detected at low levels in both types of cells derived from KO mice (Fig. 3A–C). Parallel induction patterns were observed in the fibroblast cell lines when they were stimulated with TNF at concentrations of 0.1, 1, and 10 ng/ml (Fig. 3A, and data not shown); the greatest increase in mRNA levels was observed with 10 ng/ml, a concentration that was therefore used for the rest of the experiments described here. In contrast to the rapid induction of TTP mRNA in fibroblasts, there were minimal changes seen with Zfp36l1 and Zfp36l2 mRNAs in the same time-course studies following TNF treatment (Fig. 3C).

Figure 3. Induction of TTP in mouse fibroblasts in response to TNF.

(A and B) Time courses of TTP mRNA levels were examined in serum-deprived fibroblasts of the WT and KO stable cell lines (A) and MEFs (B) after exposure to TNF treatment at 1 and 10 ng/ml (A) or 10 ng/ml only (B). RNA was prepared at various time-points after addition of TNF, and TTP mRNA was measured with real-time RT-PCR. Results shown are means ± sem of values from 4–5 independent experiments, and data are expressed as fold increases relative to the WT Time 0 samples, which were set at 1 after normalization to the internal control Actb mRNA. (C) Time courses of Zfp36, Zfp36l1, and Zfp36l2 mRNA expression (arrowheads) after TNF (10 ng/ml) stimulation were measured in the stable fibroblast cell lines by Northern blotting. The TTP-neo fusion mRNA seen in the KO cell line (top left panel) is indicated by the arrow. (D) Shown in the upper panel are TTP protein levels in a similar experiment, as determined by Western blotting of total cellular lysates from fibroblast cell lines that were stimulated with TNF (10 ng/ml) for the indicated times. The positions of protein size standards are indicated to the left of the blot. The multiple TTP species, thought to represent differentially phosphorylated forms of the protein, are indicated by the arrows. Immunoreactive ACTB (lower panel) served as a loading control.

At the protein level, TTP protein was undetectable at Time 0 in the WT fibroblast cell line but became evident in multiple forms of different relative molecular mass within 30 min of TNF stimulation and then increased to an apparent maximum between 60 and 120 min (Fig. 3D). The multiple species have been attributed to progressively increased phosphorylation of the protein [18]. As in other systems, e.g., LPS-stimulated macrophages [6, 17] or serum-stimulated fibroblasts [1], the TTP protein levels remained relatively constant and elevated after 90 min, whereas the transcript levels returned to near baseline by 1 h (Fig. 3A–C). As seen previously in cells from these mice, there was no detectable immunoreactive TTP protein in the KO cells before or after TNF stimulation (Fig. 3D).

Time courses of gene expression following TNF challenge

To evaluate the effect of TTP on the decay of TNF-induced transcripts in this system, we selected 12 from the total of 25 early-response genes from the Hao and Baltimore study [11]. These were selected on the basis of their rapid induction in response to TNF and also on the presence of at least 1 copy of the core heptamer of the ideal TTP binding site, UAUUUAU, in their 3′-UTRs (Table 1). Eleven of the studied genes were from the Class I genes (Hao and Baltimore [11]), with kinetic features of early induction and rapid decrease after TNF stimulation. We included the transcript encoding LIF to the study list, as this had been identified previously as a likely TTP target transcript in this cell type [1]. Finally, we added 2 genes from their Class II or intermediate group, encoding ICAM1 and CSF1.

TABLE 1.

TTP binding motifs in TNF-responsive genes

mRNA	Alias or description	GenBank Accession #	Class	3′-UTR annotation
				7-mer	9-mer
Tnf	Tumor necrosis factor	NM_013693	I	8	5
Cxcl2	Chemokine (C-X-C motif) ligand 2	NM_009140	I	7	4
Cxcl1	Chemokine (C-X-C motif) ligand 1	NM_008176	I	6	2
Ptgs2	Cox-2	NM_011198	I	6	1
Ier3	Immediate early response 3	NM_133662	I	4	0
Lif	Leukemia inhibitory factor	NM_008501	I	5	1
Ereg	Epiregulin	NM_007950	I	3	1
Zfp36	Tristetraprolin	NM_011756	I	3	1
Slc25a25	Solute carrier family 25	NM_146118	I	3	0
Fos	FBJ osteosarcoma oncogene	NM_010234	I	2	1
Tnfaip3	A20	NM_009397	I	2	1
Icam1	Intercellular adhesion molecule 1	NM_010493	II	2	0
Csf1	Colony-stimulating factor 1	NM_007778	II	1	0

The 3′-UTRs of Class I and Class II transcripts (from ref. [11]) were analyzed for the presence of TTP binding 7-mers (UAUUUAU) and 9-mers (UUAUUUAUU). The Class I transcripts exhibited rapid induction kinetics with peak levels at 0.5 h after TNF stimulation; Class II transcripts peaked later at ∼2 h. Transcripts highlighted in bold type were documented to be stabilized in TTP-deficient fibroblasts in the present study.

Consistent with the earlier findings [11], we found that TNF produced rapid elevations in transcript levels from the 11 early-response genes in WT fibroblast cell lines (data not shown) and in WT MEFs (Figs. 4A and 5A and Supplemental Fig. 2), with all transcript levels reaching their peaks 30 or 60 min after TNF, generally decreasing thereafter. These studies suggested that 30 and 60 min after TNF stimulation would be reasonable time-points for later studies involving actinomycin D inhibition of transcription. The rapid return to basal levels at 2 h for most of these transcripts indicates that these ARE-rich transcripts are unstable, suggesting the involvement of additional trans-acting instability factors other than TTP. Levels of the Class II Icam1 and Csf1 mRNAs were increased gradually until 1–2 h and sustained thereafter (Supplemental Fig. 2). In general, the induction kinetics of these genes were comparable between early passages of primary MEFs and the stable fibroblast cell lines, suggesting that the 67^+/+ fibroblast cell line is a reasonable model for primary MEFs [1].

Figure 4. Time courses of early-response transcripts (Fos, Cxcl1, and Cxcl2) following TNF stimulation.

(A) Induction of Fos, Cxcl1, and Cxcl2 mRNAs at various time-points after TNF stimulation was analyzed by use of real-time RT-PCR in primary MEFs. Results are expressed as means ± sem of values from 5 independent experiments. The asterisks and P values represent significant differences between the mean values from WT and KO primary MEFs at 1 h. (B) Concentrations of CXCL1 and CXCL2 proteins in the supernatants from cultured, primary MEFs were measured by ELISA after TNF stimulation (10 ng/ml). ND, Not detectable; *P < 0.05, and **P < 0.01, WT versus KO.

Figure 5. Time courses of early-response transcripts (Ier3, Ptgs2, and Lif) following TNF stimulation.

(A) Induction of Ier3, Ptgs2, and Lif mRNAs at various time-points after TNF stimulation was analyzed by use of real-time RT-PCR in primary MEFs. Results are expressed as means ± sem of values from 5 independent experiments. The asterisks and P values represent significant differences between the means of values from WT and KO in primary MEFs at 1 h. (B and C) Cytosolic extracts and hypertonic cell lysates prepared from fibroblast cell lines were used for Western blotting of IER3 and PTGS2 protein in TNF-stimulated fibroblast cell lines, respectively. FBS stimulation was used as a positive control. Immunoreactive ACTB [bottom (B)/lower (C)] served as a loading control. (D) Concentrations of LIF protein in the supernatants from cultured, primary MEFs were measured by ELISA at various time-points after TNF stimulation (10 ng/ml). *P < 0.05, WT vs. KO.

In the MEFs, there were consistent and significant increases of Cxcl1, Cxcl2, and Ier3 transcript levels in TTP KO primary MEFs compared with control cells harvested 1 h after TNF stimulation (Figs. 4A and 5A), suggesting a possible role for TTP deficiency in the greater increase and/or delayed return to normal of those potential target transcript levels. A trend toward up-regulation of Ptgs2 and Lif mRNAs was also noted 1 h after TNF, which did not reach statistical significance (Fig. 5A). In contrast, TTP KO primary MEFs did not exhibit abnormal induction kinetics or altered response magnitude of certain other early-response genes that contain fewer or no TTP binding motifs, including those encoding Fos, Ereg, Tnfaip3, and Slc25a25 mRNAs (Fig. 4A and Supplemental Fig. 2).

To determine if the increases in the chemokine Cxcl1 and Cxcl2 mRNA levels in the KO MEFs were reflected in increased chemokine secretion into the culture medium, we measured their accumulation at various time-points after TNF stimulation. CXCL1 levels were elevated significantly in the supernatants from TTP-deficient MEFs at Time 0 and at most of the time-points examined until 8 h. The basal level of CXCL2 was too low to be detected at Time 0 but became detectable 1 h after TNF stimulation (Fig. 4B). After 2 h, there were significant increases in immunoreactive CXCL2 in the TTP KO cell supernatants (Fig. 4B). These data confirm that the increases in Cxcl1 and Cxcl2 mRNAs seen in the KO cells after TNF stimulation were reflected in increased secretion of the chemokines into the culture medium.

To determine whether the increased steady-state levels of the Ier3, Lif, and Ptsg2 transcripts in the TTP-deficient MEFs were associated with increases in the respective protein concentrations, we determined protein levels in cytosolic extracts or cell culture supernatants (Fig. 5B–D). Increased levels of IER3 were readily detectable in the KO cells after 30 min treatment with TNF compared with the minimal change observed in WT cells; however, the TNF-induced increase in IER3 protein was not as dramatic as that seen in KO and WT cells after 90 min of serum stimulation when TTP protein levels were already maximal (Fig. 5B). In contrast, the levels of PTGS2 protein in KO cell lysates and the accumulation of LIF production in the KO cell supernatants at various time-points after TNF stimulation did not differ significantly from those seen in the WT cells (Fig. 5C and D).

Increased stability of early-response transcripts after TNF stimulation of TTP-deficient fibroblasts

To evaluate the potential role of TTP in promoting the instability of these TNF-induced transcripts, both types of cultured fibroblasts were then treated with actinomycin D, 30 min after TNF stimulation. This time was chosen for the mRNA decay assays, as the levels of these transcripts had reached near maxima at this time after TNF stimulation; a caveat is that TTP protein levels had not yet reached maxima by 30 min in the WT cells (Fig. 3C), although immunoreactive TTP was readily detectable by this time. The Fos transcript has been used as a rapidly induced, negative control in previous studies of this cell type [1]. In the present study, this transcript decayed very rapidly after actinomycin D treatment, with decay rates virtually superimposable in WT and TTP KO cell lines and in WT and TTP KO MEFs (Fig. 6 and Table 2). These studies support previous conclusions that TTP has little or no effect on Fos mRNA decay in this cell type, at least after serum or TNF stimulation.

Figure 6. Stability of early-response transcripts after TNF stimulation in the presence and absence of TTP.

To inhibit transcription, actinomycin D was added after 30 min of TNF treatment of serum-starved fibroblasts (A, stable cell lines; B, MEFs). The cells were harvested for total RNA extraction at the indicated times, and mRNA levels were measured with real-time RT-PCR. Transcript concentrations were normalized to those of the Actb transcripts and were expressed as fractions of abundance in TNF-treated samples before the addition of actinomycin D. The results shown are means ± sem of 5 independent experiments. Statistical differences, determined by Student’s t-test, are indicated (*P < 0.05, and **P < 0.01, WT vs. KO).

TABLE 2.

Times to 50% decay of ARE-containing transcripts in mouse fibroblasts

mRNA	7-mers in 3′UTR	Time to 50% mRNA decay in cell lines, min			Time to 50% mRNA decay in primary MEFs, min
		67 WT	66 KO	P	WT	KO	P
Fos	2	11.8 ± 1.7	12.5 ± 2.6	0.493	11.9 ± 0.6	15.1 ± 0.8	0.074
Zfp36	3	9.3 ± 0.9	8.7 ± 0.7	0.285	14.4 ± 0.8	18.9 ± 1.6	0.104
Cxcl1	6	34.9 ± 4.6	84.6 ± 13.0	0.011	17.2 ± 1.4	33.6 ± 2.9	0.011
Cxcl2	7	25.5 ± 2.7	46.0 ± 8.6	0.030	25.0 ± 1.9	76.4 ± 14.2	0.017
Ier3	4	15.5 ± 1,8	27.5 ± 3.3	0.002	18.6 ± 1.5	40.6 ± 3.8	0.012
Lif	5	20.2 ± 3.0	31.9 ± 6.2	0.040	33.0 ± 2.1	52.5 ± 4.9	0.007
Ptgs2	6	43.1 ± 11.4	73.7 ± 15.2	0.033	86.3 ± 12.7a	Stable	NA
Tnfaip3	2	17.1 ± 1.1	20.2 ± 3.1	0.202	30.4 ± 1.6	31.1 ± 1.9	0.704
Slc25a25	3	29.8 ± 3.4	39.5 ± 5.5	0.106	51.4 ± 1.8	68.6 ± 5.3	0.039
Ereg	3	78.5 ± 14.3	80.0 ± 13.3	0.933b	Stable	Stable	NA
Icam1	2	Stable	Stable	NA	Stable	Stable	NA
Csf1	1	38.1 ± 8.1	56.2 ± 14.9	0.168	Stable	Stable	NA

Actinomycin D was added after 30 min of TNF stimulation of serum-starved, stable fibroblast cell lines and primary MEFs. The times to 50% mRNA decay of these TNF-responsive mRNAs shown are the averages of nonlinear fit 1-phase decay curves from 5 independent experiments unless otherwise indicated. “7-mer,” UAUUUAU. For the TTP mRNA, the WT cells expressed the endogenous mRNA, whereas the TTP-Neo fusion transcript was expressed in the KO cells. “Stable,” No measurable decay during these experiments. NA, not applicable.

^aCalculated half-life from 1 averaged decay curve from 5 experiments.

^bAverages from 3 computable 1-phase decay curves out of 5 independent experiments.

In contrast, although there were rapid decreases in Cxcl1, Cxcl2, Ier3, and Lif mRNA levels in WT cells from the stable cell line and primary MEFs after actinomycin D treatment, the decay rates of these transcripts in the TTP-deficient cells were decreased significantly (Fig. 6 and Table 2). For example, the times to 50% mRNA decay for the Cxcl1 transcripts were increased from ∼35 min in the WT cell line to 85 min in the TTP KO cell line and from 17 min in the WT MEFs to 34 min in the KO MEFs. For the Cxcl2 transcripts, the times to 50% transcript decay increased from 26 min in the WT cell line to 46 min in the KO cell line and from 25 min in the WT MEFs to 76 min in the TTP KO MEFs (Table 2).

We showed previously that that TTP can promote Ier3 mRNA turnover in fibroblasts after serum stimulation [1]. In the present study, the decay rates of Ier3 transcripts seen after TNF stimulation were decreased in the absence of TTP in primary MEFs and in the stable fibroblast cell lines (Fig. 6). The mean times to 50% decay were increased from 16 min in the WT cell line to 28 min in the TTP KO cell line and from 19 min in the WT MEFs to 41 min in the KO MEFs (Fig. 6 and Table 2).

Smaller but nonetheless significant slowing in mRNA turnover rates in the TTP KO cells were also noted for the Lif and Ptgs2 transcripts (Fig. 6 and Table 2). For both transcripts, significant changes were observed in both the stable cell lines and the primary MEFs.

The other early and intermediate response transcripts evaluated did not exhibit effects of TTP on their decay rates after TNF stimulation (Supplemental Fig. 3). Tnfaip3 and Slc25a25 mRNAs decreased rapidly in a parallel manner in WT and KO cells from stable cell lines and MEFs (Supplemental Fig. 3). The Ereg, Icam1, and Csf1 transcripts were relatively stable in the WT cells but also did not exhibit significant decreases in decay rate in the TTP-deficient cells (Supplemental Fig. 3).

Given the significant effects of TTP deficiency in the stable cell lines and primary MEFs, we conclude from these studies that TTP is likely to play a key role in determining the turnover rates of Cxcl1, Cxcl2, Ier3, Lif, and Ptgs2 transcripts in fibroblasts after TNF stimulation. As the mRNA turnover was analyzed only 30 min after TNF stimulation, when TTP protein is just becoming detectable, it seems probable that these transcripts are direct binding targets of TTP under these conditions. This is supported by target sequence analysis and direct binding assays, as described below.

Conserved TTP binding sites within the 3′-UTRs of potential TTP target transcripts

We next examined the 3′-UTR sequences of the likely target transcripts for the presence and conservation of TTP binding sites. The mRNAs for the early-response genes Cxcl1 and Cxcl2 contained multiple copies of the minimal TTP binding motif, UAUUUAU, that were conserved across many mammalian species (Supplemental Fig. 4A and B). Similar potential TTP binding sites were noted previously in the 3′-UTR of the Ier3 mRNA, where the presence of 4 canonical UAUUUAU heptamers has been reported (ref. [1]; also see Supplemental Fig. 4C and Table 1). Likewise, the 3′-UTR sequences of mouse Ptgs2 and Lif mRNAs contained at least 3 copies of potential TTP binding heptamers (Supplemental Fig. 4D and E and Table 1). In contrast, the 3′-UTR sequences for the Ereg, Slc25a25, and Tnfaip transcripts, whose turnover rates were not affected significantly by the absence of TTP, contained fewer, if any, potential TTP binding sites; those that were present were often scattered throughout the 3′-UTRs rather than overlapping or closely spaced. The intermediate response transcripts, such as Icam1 and Csf1, had the fewest potential TTP binding sites in their 3′-UTRs, consistent with their sustained accumulation after TNF stimulation and lack of a TTP effect on their decay.

Binding of TTP to its putative target transcripts

We next tested the ability of hTTP, overexpressed in 293 cells, to bind to its predicted binding sites within the 3′-UTRs of the identified potential TTP target transcripts by use of RNA gel-shift experiments. Each of the synthetic ARE probes synthesized was comprised of 2 closely spaced heptameric binding sites within the respective target transcripts; mutated versions of these probes were also synthesized, which were otherwise identical except for the replacement of the core A residues with Cs (Fig. 7A–C). This modification has been shown previously to eliminate the binding of TTP to similar probes [12, 16]. A nonbinding mutant form of TTP, C124R, with a single amino acid mutation of cysteine 124 to arginine in the first TTP zinc finger was also used as a further negative control [16]. A probe derived from the mouse Tnf ARE served as a positive control [16].

Figure 7. Binding of TTP to potential target transcripts.

RNA gel-shift assays were performed with biotinylated RNA probes containing 2 heptameric potential TTP binding sites (underlined) derived from the 3′-UTRs of Cxcl1, Cxcl2 (A), Lif (B), and Ptgs2 (C) transcripts. The locations of the probe sequences in relation to the GenBank RefSeq mRNA accession numbers are indicated next to the gene names. The A→C mutations of the RNA probes are highlighted in black. Probes were incubated with cytosolic extracts prepared from HEK 293 cells, transfected with vector alone (BS+), the WT hTTP plasmid, or the mutant TTP plasmid with the C124R zinc finger mutation (C124R). The positions of free probes, TTP-containing binding complexes, and antibody-supershifted complexes are indicated to the right of the blots. A Tnf ARE probe was used as a positive control. (D) Biotinylated RNA-protein pull-down assays were performed by use of the biotinylated RNA ARE probes listed in A–C and 900 or 500 μg protein lysates prepared from primary MEFs after 1 h of TNF or serum treatment. The proteins were then eluted and analyzed using Western blotting for the presence of TTP and HuR. HuR protein served as a control for pull-downs and Western blotting.

The typical pattern of Tnf ARE probe binding to the lysates overexpressing WT TTP, which was shown with ³²P-labeled Tnf ARE probes in our previous studies [1, 7], was recapitulated with the otherwise identical, biotinylated ARE probes (Fig. 7A–C, Lanes 1–6). Binding complexes were readily detectable when the Tnf ARE probe was allowed to bind to WT TTP (Fig. 7A–C, Lanes 3 and 5), but no complexes were formed with the mutant TTP (C124R; Fig. 7A–C, Lanes 4 and 6) or with lysates transfected with vector alone (Fig. 7A–C, Lane 2). Both of the Cxcl1 and Cxcl2 RNA ARE probes formed at least 2 binding complexes with slightly different migration rates when incubated with WT TTP (Fig. 7A, Lanes 9 and 19). These complexes could be supershifted by an anti-HA antibody (Fig. 7A, Lanes 11 and 21). In contrast, neither of these RNA probes bound to the mutant TTP (C124R; Fig. 7A, Lanes 10 and 20). As expected, WT TTP did not form binding complexes with the mutant probes (Fig. 7A, Lanes 15 and 25).

The binding of an Ier3 ARE probe to TTP has been described in a previous study, in which the binding specificity of the Ier3 ARE probe was confirmed with mutant TTP and mutant probes [1]. Here, we tested the ability of TTP to bind to at least some of the potential binding sites within the 3′-UTRs of the Lif and Ptgs2 mRNAs, in which the sites are somewhat scattered within the 3′-UTRs (Supplemental Fig. 4). In the case of the Lif transcript, binding of TTP to 2 different probes was readily detectable in RNA gel-shift assays (Fig. 7B, Lanes 9 and 19); in both cases, the complexes were supershifted with the addition of the anti-HA antibody (Fig. 7B, Lanes 11 and 21). None of the protein lysates from vector alone or the C124R mutant transfection formed binding complexes with these probes (Fig. 7B, Lanes 8, 10, 18, and 20). Apparently, weaker binding complexes were formed with the Ptgs2 ARE probe, requiring a longer exposure time to detect the binding complexes (Fig. 7C, Lane 9). Nonetheless, these complexes could still be supershifted with the anti-HA antibodies (Fig. 7C, Lane 11). The specificity of the Ptgs2 probes was confirmed by the absence of binding of the normal probes to the mutant form of TTP, as well as the absence of binding of the mutant probe to WT TTP (Fig. 7C, Lanes 10 and 15).

To validate further the binding of TNF-induced endogenous TTP to these RNA ARE probes, we examined the protein lysates from primary MEFs after 1 h of TNF or serum treatment in the biotinylated RNA-protein pull-down assay. Protein lysates from serum-treated cells were used as a positive pull-down control, as a result of the much higher TTP protein levels seen after serum stimulation in fibroblasts. As shown in Fig. 7D, TNF- and serum-induced endogenous TTP protein was able to bind to the RNA ARE probes derived from Cxcl1, Cxcl2, Lif, and Ptgs2 transcripts; however, TTP did not bind to the A → C mutant probes, except for the mutant probe of Cxcl2. There was trace binding of TTP to the mutant Cxcl2 ARE probe, possibly as a result of the presence of an AU-rich sequence in the linkage region between the 2 optimal binding sites. HuR protein, a control for pull-down and Western blotting, was shown to bind equally to the WT and mutant ARE probes, suggesting that the binding of HuR to these types of ARE elements is less dependent on these specific sequences than TTP. An exception was the experiment in which HuR was tested with the Cxcl2 ARE probe in the setting of a high salt buffer, in which the binding of HuR to the WT Cxcl2 ARE probe was abolished, whereas its binding to the mutant probe was preserved. These data support the high binding affinity of TTP to its binding sites and the possibility of antagonism between TTP and HuR.

DISCUSSION

The mRNA destabilizing, RNA-binding protein TTP has long been recognized as an important component of the ARE-mediated mRNA decay machinery (reviewed recently in refs. [4, 19, 20]). The first-identified and best-studied example of a TTP target transcript is the Tnf mRNA, initially studied in LPS-stimulated macrophages. In this experimental system, TTP is first induced in response to LPS and then binds to ARE within the 3′−UTR of the Tnf transcript, promoting its deadenylation and ultimate decay [6, 7]. Thus, TTP functions as an essential anti-inflammatory protein by limiting TNF production from macrophages upon exposure to endotoxin and other inflammatory stimuli. TNF is one of the most potent proinflammatory cytokines and can participate in systemic inflammation as well as in local responses to activators of innate immunity. Once it is released by primary response cells, primarily macrophages, TNF exerts a wide variety of actions on tissues and on diverse cell types, including immune cells and nonimmune cells.

Fibroblasts are a group of heterogeneous resident cells of mesenchymal origin that have various locations, diverse appearances, and distinctive activities. Because of their ubiquitous distribution as tissue cells, they are poised to respond to factors released by newly activated innate immune cells. This concept was studied recently in mouse 3T3 fibroblasts that were exposed to TNF [11]. The authors categorized TNF-responsive genes in these cells into at least 3 kinetically different groups and correlated the speed and duration of the induction responses with the ARE content of their 3′-UTRs. We took advantage of this dataset in the present study, in which we investigated transcripts that had been demonstrated to increase rapidly in response to TNF [11]. Our goal was to determine whether any of these rapidly responding transcripts in these potential “secondary response” cells were affected by the absence of TTP. We took advantage of the availability of 2 types of fibroblasts from matched TTP WT and KO mice: well-validated and matched stable cell lines [1], which are uncontaminated with other cell types, and early-passage MEFs, which may contain small concentrations of contaminating cells, such as macrophages.

Our data confirm the earlier data presented by Hao and Baltimore [11] in fibroblasts that TTP mRNA is rapidly induced in response to TNF. We confirmed this in the stable cell lines and MEFs and demonstrated commensurate increases in immunoreactive TTP protein. We showed previously that TTP responds in a similar manner to TNF stimulation of primary macrophages [6]. We then investigated the effects of TTP deficiency on the stability of the rapidly responding mRNAs from the Hao and Baltimore set [11], which contained expected TTP binding motifs. We found in both types of fibroblasts that the absence of TTP resulted in the stabilization of transcripts encoding CXCL1, CXCL2, IER3, LIF, and PTGS2 after TNF stimulation. In the cases of CXCL1 and CXCL2, we documented increases in the secretion of these chemokines into the culture medium in the absence of TTP; however, we were not able to demonstrate similar increases in LIF concentrations under these conditions. In cytosolic extracts, we demonstrated increases in the levels of IER3 protein in TTP-deficient fibroblasts relative to WT cells after TNF stimulation, but we were not able to document significant changes in PTGS2 protein in TTP-deficient fibroblasts compared with WT cells. We also demonstrated binding of TTP to ARE-containing RNA probes derived from the 3′-UTRs of several of these transcripts that use RNA gel-shift binding and biotinylated RNA-protein pull-down assays; direct binding to TTP of some of these transcripts had been described earlier [1]. Thus, the transcripts produced by Cxcl1, Cxcl2, Ptgs2, Ier3, and Lif have been identified and confirmed as highly likely physiologic target mRNAs for TTP in mouse fibroblasts.

These 5 early-response transcripts share some common features in their kinetic patterns in response to TNF, such as rapid induction and subsequent rapid decay. Consistent with their rapid turnover rates, all 5 transcripts contain at least 3 copies of closely spaced and often overlapping heptamer and/or nonamer potential TTP binding sites in their 3′-UTRs. In addition, scattered potential binding nonamers are often present elsewhere in the 3′-UTRs of these mRNAs. In contrast, the turnover rates of other early TNF response mRNAs that contained few such binding sites, including Tnfaip3, Fos, Ereg, and Slc25a25 mRNAs, were not affected by the presence or absence of TTP. The presence of multiple putative TTP binding motifs in the 3′-UTR of transcripts appears to increase the likelihood that these transcripts will be physiologic TTP targets.

Fibroblasts have been shown to be capable of modifying the quality, quantity, and duration of inflammatory infiltrates through their production of cytokines and chemokines [10, 21–23]. We propose here a role for TTP in regulating the stability of cytokines and chemokines produced by fibroblasts in response to TNF, implicating TTP in the regulation of inflammatory responses participated in by secondary response cells. CXCL1 (KC in mouse and GROα in human) is a member of the CXC class of chemokines; its mRNA stability has been shown previously to be regulated by TTP in murine macrophages by use of cells derived from the same TTP-deficient mice used in the current studies [24]. This mRNA was also identified as being among the top 23 “hits” in a global microarray analysis of stabilized transcripts in TTP-deficient fibroblasts after serum stimulation [1]. The present studies documented stabilization of the Cxcl1 mRNA in the absence of TTP in response to TNF in 2 types of mouse fibroblasts and demonstrated a commensurate increase in the secretion of the CXCL1 chemokine into the cell-culture medium in the absence of TTP. Originally identified as a potent neutrophil chemoattractant, CXCL1 has been demonstrated extensively to play central roles in the processes of angiogenesis, atherosclerosis, inflammation and immunity, wound healing, and tumorigenesis [25–27]. Local elevation of CXCL1 in synovial fluid and tissue has been associated with the pathogenesis of rheumatoid arthritis [27]. CXCL1 exerts its biologic activity through binding to a G-protein-coupled receptor, CXCR2, which appears to play major roles in neutrophil recruitment during arthritis; therefore, neutrophils are positioned to guide the development of immune responses through the production of various inflammatory mediators.

CXCL2 (MIP2-α) is another CXC chemokine family member; the mouse protein shares almost 70% identical nucleic and amino acid sequences with mouse CXCL1. Although CXCL2 is thought to be expressed mainly by macrophages and monocytes, we found that both types of fibroblasts secreted CXCL2 in response to TNF, and this secretion was increased in the absence of TTP. Likewise, the Cxcl2 transcript was greatly stabilized in both types of TTP-deficient fibroblasts. The Cxcl2 transcript has also been identified as one of the top 25 most stable mRNAs in TTP-deficient macrophages in a recent genome-wide analysis of mRNA decay in LPS-stimulated macrophages after treatment with actinomycin D and a p38 MAPK inhibitor [28]. CXCL2 is thought to act by a similar mechanism to CXCL1 in that both bind to CXCR2; CXCL2 also plays a significant role in mediating neutrophil recruitment in inflammation, with effects on leukocyte mobilization, homing, and trafficking [29]. It is also thought to be chemotactic for hematopoietic stem cells [30]. CXCL1 and CXCL2 have been implicated in the transformation of melanocytes, as well as in the proliferative and invasive properties of melanoma cells [31]. Both chemokines are thought to play crucial roles in cancer-related inflammation, as well as having effects on tumor development and progression.

Of the 3 other TTP target transcripts validated in this study, Ier3, Ptgs2, and Lif transcripts have been implicated as TTP target transcripts in other settings. For example, Ier3 and Lif mRNAs were identified as TTP targets in serum-stimulated fibroblast cell lines [1]. Likewise, Ptgs2, also known as Cox-2, has been described as a TTP target transcript in transient transfection studies of the human colorectal adenocarcinoma-7 [32]. Little is known about the biologic activities of IER3 in inflammation, although Ier3 KO mice were reported recently to exhibit aberrant immune regulation and enhanced inflammation [33]. LIF is a well-known secreted glycoprotein belonging to the IL-6 family of cytokines. It has been documented to possess proinflammatory and anti-inflammatory properties, and endogenous LIF is thought to be required for neuronal, glial, and skin responses to injury in terms of inflammatory cell infiltration, probably because of macrophage chemotactic activity [34]. Ptgs2 mRNA has long been known to be selectively inducible at sites of inflammation and is a well-known clinical therapeutic target for inflammatory inhibition. Although TTP has been shown to bind to and promote the decay of Ptgs2 transcripts in a cancer cell line [35], other mRNA binding proteins, such as HuR, T cell intracellular antigen-1 (TIA-1), and heterogeneous nuclear ribonucleoprotein D (ARE RNA binding protein 1, AUF1), are capable of binding to the same ARE region [35, 36], giving rise to questions about the specific roles of individual components in the modulation of Ptgs2 mRNA stability. This potential diversity in the post-transcriptional control of chemoattractant cytokine gene expression has been discussed recently by several groups [37, 38]. Indeed, we did not observe increases in PTGS2 protein levels in the TNF-stimulated TTP KO fibroblasts relative to WT cells; we have also failed to detect changes in the eicosanoid products of its activity (data not shown).

Given the close proximity of macrophages to stromal fibroblasts in tissue microenvironments, we suggest that TTP regulates certain aspects of the cross-talk between the 2 cell types that occurs with local activation of innate immunity (Fig. 8). In this scheme, macrophages act as the first-line responders to invading pathogens and their toxins as part of the innate immune response. These activated macrophages secrete TNF, which can then stimulate fibroblasts as secondary response cells. The role of TTP in regulating the secretion of TNF in response to bacterial pathogens is well established; we now show that TTP plays an important role in regulating the biosynthesis of important secondary response proteins in TNF-stimulated fibroblasts, such as CXCL1, CXCL2, LIF, PTGS2, and IER3. TTP has already been shown to play a similar role in the expression of another chemokine, GM-CSF, in another mesenchymal cell type, BM-derived stromal cells [39]. Taken together, these fibroblast-secreted chemokines and other molecules would then be expected to increase chemotaxis of leukocytes into the site of inflammation, perhaps also promoting differentiation of tissue monocytes into activatable macrophages in the process. These leukocytes could, in turn, release more TNF into the local site of inflammation, producing several levels of self-reinforcing cycles. TTP induction by the proinflammatory stimuli can influence the coordinated expression of inflammatory genes in both cell types and thus, the overall tissue response to inflammatory challenge. The critical role for mesenchymal cell responses in the pathogenesis of TNF-mediated inflammatory diseases has been highlighted in a recent study by use of selective Cre/loxP-mediated targeting of the type I TNFR expression in mesenchymal cells, in which fibroblasts and myofibroblasts of the joint and the intestine were recognized as the primary response cells sufficient for full pathogenic TNF/TNFRI signaling in arthritis, sacroiliitis, and Crohn's-like inflammatory bowel disease [40].

Figure 8. Coordinate regulation by TTP of primary and secondary cellular responses to proinflammatory stimuli.

A schematic representation is shown of the proposed cross-talk between macrophages and fibroblasts that is coordinated by TTP. In this scheme, macrophages serve as the primary cells responding to proinflammatory stimuli, such as LPS, and fibroblasts function as the secondary cells responding to the TNF released from the macrophages. TTP is responsible for regulating the primary release of TNF from the macrophages and for regulating the secondary release of chemokines and other chemotactic factors from stromal cells, such as fibroblasts. Thus, TTP induction by proinflammatory stimuli can coordinately regulate the expression of proinflammatory genes in both cell types.

Although we have focused on the post-transcriptional regulation of Cxcl1, Cxcl2, Ier3, Ptgs2, and Lif transcripts by TTP, we have consistently observed greater fold changes in protein levels than seen at the mRNA level, as seen with Cxcl1, Cxcl2, and Ier3 in this study. These findings are consistent with initial findings in the TTP KO mice, in which TTP-deficient BMDM secreted ∼5-fold more TNF, whereas the increase in Tnf mRNA levels was only ∼2-fold [5], suggesting a direct effect of TTP on translation. A well-recognized concept is that ARE elements within the 3′-UTR of transcripts are key elements in the regulation of transcriptional and translational processes [41]. We have proposed recently a post-transcriptional regulation model involving TTP, in which TTP is proposed to function as a signal-integrating switch, serving as the “link” between decay and translation of its mRNA ligands [4]. In this model, decreased p38 activity has been proposed to link with TTP-dependent transcript destabilization and translational repression [4, 42]. In the translational studies of Gm-csf and Tnf mRNAs, TTP was found to cooperate with DDX6 (DEAD [Asp-Glu-Ala-Asp] box polypeptide 6)/p54 to shut off protein production only after translational initiation in polysome profiles [43]. Further work is necessary to determine whether these instances of TTP-affected translation are universal among TTP target transcripts or are transcript specific, particularly under physiologic conditions.

In summary, a rapid increase in TTP biosynthesis is one of the immediate responses of fibroblasts to TNF stimulation. TTP can coordinately regulate the primary (macrophage) and secondary (fibroblast and stromal) cellular responses to proinflammatory stimuli and thus, participates in the resolution of tissue inflammation. Many questions still remain concerning potential antagonism, redundancy, and synergy between TTP and other RNA-binding proteins and factors in the post-transcriptional regulation of inflammatory processes. Examples of possible interactions between TTP and other RNA-binding proteins are the competitive binding of TTP and HuR to Tnf mRNA [42] and the binding of HuR and heterogeneous nuclear RNA A0 to Ptgs2, Cxcl2, and Tnf mRNAs, as shown in this study and others [44].

AUTHORSHIP

P.J.B. designed and supervised the study. L.-Q.Q. performed the experiments and analyzed the data. W.S.L. generated the stable fibroblast cell lines, prepared the overexpression lysates used in the gel-shift experiments, and performed the Northern blotting analysis of TTP family members. A.B. and D.C.Z. performed and analyzed the Western blot of experiments with PTGS2. L.-Q.Q. and P.J.B. wrote the manuscript.

Glossary

3′-UTR

3′-untranslated region

ACTB

β-actin

ARE

AU-rich element

BM

bone marrow

BMDM

bone marrow-derived macrophages

COX-2

cyclooxygenase 2

CXCL1

chemokine (C-X-C motif) ligand 1

CXCL2

chemokine (C-X-C motif) ligand 2

E

embryo day

Ereg

epiregulin

Fos

FBJ osteosarcoma oncogene

GRO

growth-related oncogene

HA

hemagglutinin

HEK

human embryonic kidney

HKLM

heat-killed Listeria monocytogenes

hTTP

human TTP

HuR

Hu antigen R

IER3

immediate early response 3

KC

keratinocyte-derived chemokine

KO

knockout

LTA-SA

lipoteichoic acid purified from Staphylococcus aureus

M-TTP KO

myeloid-specific, tristetraprolin-deficient

MEF

mouse embryonic fibroblast

NP-40

Nonidet P-40

Pam₃CSK₄

palmitoyl-3-cysteine-serine-lysine-4

PIC

polyinosinic:polycytidylic acid

PTGS2

PG-endoperoxide synthase 2

Slc25a25

solute carrier family 25

TBE

Tris-borate-EDTA

Tnfaip3

TNF-α-induced protein 3

TTP

tristetraprolin

WT

wild-type

ZFP36L1/2

zinc finger protein 36, C3H type-like 1/2

DISCLOSURES

The authors declare that they have no conflict of interest.