Negative Feed-forward Control of Tumor Necrosis Factor (TNF) by Tristetraprolin (ZFP36) Is Limited by the Mitogen-activated Protein Kinase Phosphatase, Dual-specificity Phosphatase 1 (DUSP1)

Background: ZFP36 negatively regulates AU-rich element-containing mRNAs. Results: DUSP1 silencing increases ZFP36 expression and enhances negative feed-forward control of TNF. Conclusion: MAPK inhibition initially attenuates TNF mRNA, but reduced feed-forward control subsequently produces an increase. Significance: Understanding such networks is essential to develop novel anti-inflammatory therapies or understand TNF repression by glucocorticoids, which may not require DUSP1 or ZFP36 expression. TNF is central to inflammation and may play a role in the pathogenesis of asthma. The 3′-untranslated region of the TNF transcript contains AU-rich elements (AREs) that are targeted by the RNA-binding protein, tristetraprolin (also known as zinc finger protein 36 (ZFP36)), which is itself up-regulated by inflammatory stimuli, to promote mRNA degradation. Using primary human bronchial epithelial and pulmonary epithelial A549 cells, we confirm that interleukin-1β (IL1B) induces expression of dual-specificity phosphatase 1 (DUSP1), ZFP36, and TNF. Whereas IL1B-induced DUSP1 is involved in feedback control of MAPK pathways, ZFP36 exerts negative (incoherent) feed-forward control of TNF mRNA and protein expression. DUSP1 silencing increased IL1B-induced ZFP36 expression at 2 h and profoundly repressed TNF mRNA at 6 h. This was partly due to increased TNF mRNA degradation, an effect that was reduced by ZFP36 silencing. This confirms a regulatory network, whereby DUSP1-dependent negative feedback control reduces feed-forward control by ZFP36. Conversely, whereas DUSP1 overexpression and inhibition of MAPKs prevented IL1B-induced expression of ZFP36, this was associated with increased TNF mRNA expression at 6 h, an effect that was predominantly due to elevated transcription. This points to MAPK-dependent feed-forward control of TNF involving ZFP36-dependent and -independent mechanisms. In terms of repression by dexamethasone, neither silencing of DUSP1, silencing of ZFP36, nor silencing of both together prevented the repression of IL1B-induced TNF expression, thereby demonstrating the need for further repressive mechanisms by anti-inflammatory glucocorticoids. In summary, these data illustrate why understanding the competing effects of feedback and feed-forward control is relevant to the development of novel anti-inflammatory therapies.

ZFP36 expression is rapidly induced by multiple pro-inflammatory stimuli, including IL1B or lipopolysaccharide (LPS), in various cells, including macrophages, fibroblasts, and A549 pulmonary epithelial cells (14 -17). Given the ability to reduce the expression of ARE-containing mRNAs, this means that ZFP36 is a negative (incoherent) feed-forward regulator of inflammatory gene expression (Fig. 1). Whereas increasing ZFP36 reduces the expression of inflammatory genes, ZFP36 protein expression is itself highly dependent on mitogen-activated protein kinase (MAPK) activation (16,18). Following pro-inflammatory stimulation, ZFP36 protein appears initially as a ϳ40-kDa protein, which becomes phosphorylated and migrates at ϳ45 kDa on SDS-PAGE (16,19). Phosphorylation is suggested to enhance ZFP36 stability and to promote targeting of ARE-containing transcripts (19). However, such a MAPKdependent negative feed-forward regulatory loop suggests that MAPK activation may act to reduce the expression of AREcontaining genes via increased ZFP36 activity (Fig. 1). Conversely, reducing MAPK activity may produce opposing effects and could, by reducing negative feed-forward control, promote expression of ARE-containing mRNAs. This scheme is further complicated by the fact that MAPKs are subject to feedback inhibition via a number of processes, including up-regulation of the dual-specificity MAPK phosphatase, DUSP1, which is itself dependent on MAPK activation (20 -23). Thus, pro-inflammatory stimuli, including IL1B and LPS, increase DUSP1 expression to dephosphorylate and inactivate MAPKs (Fig. 1). In A549 cells, knockdown of IL1B-induced DUSP1 expression transiently increased the appearance of phosphorylated MAPKs, and this increased the expression of inflammatory mRNAs at 1 h post-IL1B (24). However, 6 h post-IL1B, this loss of DUSP1 decreased the expression, relative to control, of multiple inflammatory mRNAs. This observation is consistent with the concept that MAPKs may increase ZFP36 expression to subsequently down-regulate ARE-containing mRNAs and is tested in the current study (Fig. 1).
In the context of glucocorticoids, reduced expression of multiple inflammatory genes is central to anti-inflammatory activity (25). However, increased DUSP1 expression is often considered as a key anti-inflammatory effector mechanism (23,26), yet this appears inconsistent with the above idea that MAPKs increase ZFP36 to increase feed-forward control of inflammatory genes. Certainly, glucocorticoids do induce DUSP1 expression, including in A549 cells (27,28). This does reduce MAPK activity, and this does play a role in the repression of at least some inflammatory genes in vitro and in vivo (24,29). However, in addition to DUSP1, glucocorticoids induce expression of multiple effector genes, and this may lead to redundant actions (30). Indeed, ZFP36 is modestly up-regulated by glucocorticoids in the human airway epithelial cells as well as in pulmonary A549 and bronchial BEAS-2B epithelial cells and in the airways following glucocorticoid inhalation (16,31,32). 3 Furthermore, a role for ZFP36 in the repression of inflammatory gene expression is indicated (31,33). Given interest in therapeutically targeting MAPK pathways in inflammatory disease and the fact that glucocorticoids induce DUSP1 to reduce MAPK activity, we have used TNF as a model ZFP36 target gene to explore the relationship(s) between the regulation of MAPK activation by DUSP1 with the expression of ZFP36 and effects on downstream gene expression.

Experimental Procedures
Gene Nomenclature-Official Human Genome Organization (HUGO) gene nomenclature committee gene symbols have been used for all genes and gene products.
Cell Culture and Drugs-A549 cells were grown in complete DMEM (Gibco) containing 10% fetal calf serum and 2.0 mM L-glutamine at 37°C in 5% CO 2 , 95% air. Human bronchial epithelial (HBE) cells, isolated from non-transplanted normal human lungs obtained using the tissue retrieval service at the International Institute for the Advancement of Medicine (Edison, NJ), were cultured in bronchial epithelial cell growth medium (Lonza, Allendale, NJ) as described previously (34). Prior to experiments, cell growth was arrested by incubating the cells in serum-free medium overnight. At this point, cell counts from four independent experiments indicate that there were 1.90 Ϯ 0.05 ϫ 10 6 cells/well (6-well plates of A549 cells) or 6.1 Ϯ 0.53 ϫ 10 5 cells/well (6-well plates of HBE cells), and the medium was switched to fresh serum-free medium containing IL1B (1 ng/ml) and/or experimental drugs. IL1B (R&D Systems) was dissolved in phosphate-buffered saline (PBS) plus 0.1% bovine serum albumin (BSA) (both Sigma). Dexamethasone (Sigma) was dissolved in Hanks' balanced salt solution (Sigma), and SB203580 (Calbiochem), Jun N-terminal kinase 3   as well as the negative feedback regulator, DUSP1, and the feed-forward regulator, ZFP36. By binding and promoting mRNA decay and/or translational arrest of ARE-containing mRNAs, ZPF36 is a negative, incoherent, regulator of inflammatory gene expression. Following MAPK activation, the increased expression of DUSP1 is one mechanism by which MAPK activity is restored to basal level. Expression of ZFP36 depends on MAPK activation, and this limits the expression of ARE-containing inflammatory mRNAs, such as TNF.
Enzyme-linked Immunosorbent Assay (ELISA)-ELISA for TNF was performed on 100 l of supernatant using DuoSet ELISA kits (R&D Systems). For HBE cells, TNF release was determined following concentration of 900 l of cell supernatant to less than 100 l using Corning Spin-X UF 500 Concentrator 5000 (catalog no. 431477) columns. Concentrated supernatants were adjusted to a final volume of 110 l using 1% (w/v) BSA/PBS, and ELISA for TNF was performed using 100 l of concentrated supernatant. Cell-associated TNF in A549 cells was detected by lysing the cells in 100 l of 1ϫ firefly luciferase assay buffer (Biotium) containing 1ϫ Complete TM protease inhibitor mixture and phosphatase inhibitors (50 mM NaF, 2 mM Na 3 VO 4 , and 20 mM Na 4 O 7 P 2 ) followed by one freeze-thaw cycle. ELISA for TNF was carried out using 50 l of cell lysates diluted with 50 l of 1% (w/v) BSA/PBS. Standard curves were generated using the same firefly luciferase assay buffer BSA/ PBS mix.
RNA Isolation, cDNA Synthesis, and SYBR Green Real-time PCR-Total RNA was extracted using the RNeasy minikit (Qiagen), and 0.5 g was used to produce cDNA as described (36). Resultant cDNA was diluted 1:4 with RNase-free water, and PCR was carried out on 2.5 l of cDNA using SYBR GreenER Mastermix (Invitrogen) with a StepOnePlus TM PCR system (Applied Biosynthesis). Relative cDNA concentrations were derived from standard curves generated by serial dilution of an IL1B-treated sample. Amplification conditions were as follows: 50°C for 2 min, 95°C for 10 min, and then 40 cycles of 95°C for 15 s and 60°C for 1 min. Primer specificity was assessed by dissociation (melt) curve analysis: 95°C for 15 s and 60°C for 20 s followed by ramping to 95°C over 20 min. Adenoviral Infection-As described previously (35), A549 cells at ϳ70% confluence were incubated with the indicated multiplicity of infection of DUSP1-expressing adenoviral vector (Ad5-DUSP1) (Seven Hills Bioreagents) or a green fluorescent protein (GFP)-expressing vector (Ad5-GFP) (Qbiogene) for 24 h in a serum-containing medium. Before further treatments, cells were incubated overnight in serum-free medium.
Analysis of Unspliced Nuclear TNF RNA-Unspliced nuclear RNA, or nascent transcript, accumulates transiently in the nucleus following transcriptional activation and may be measured as a surrogate of transcription rate (28,35). Unspliced nuclear RNA was analyzed using SYBR Green primers that crossed the exon 3/intron 4 junction for TNF. Expression was normalized to the abundant small nuclear RNA, U6. Because these primer sets detect both unspliced RNA and genomic DNA, the signal due to the contaminating genomic DNA was assessed in each sample. Each RNA sample was subject to reverse transcription, both in the presence and the absence of reverse transcriptase. The presence of an amplification product in the reverse transcription-negative samples was attributed to genomic DNA contamination, and samples with greater than 10% genomic contamination for U6 were excluded from further analysis. RNA extraction, cDNA synthesis, and SYBR Green real-time PCR were carried out as described above. Primer sequences were as follows: unspliced nuclear TNF (forward, TCT CGA ACC CCG AGT GAC A; reverse, CAT CAG CCG GGC TTC AAT) and U6 (forward, AAT TGG AAC GAT ACA GAG AAG ATT AGC; reverse, GGA ACG CTT CAC GAA TTT GC).
siRNA-mediated Gene Silencing-A549 cells were grown in 12-well plates to ϳ60 -70% confluence and transiently transfected with 1 ml of serum-free medium containing DUSP1 or ZFP36 or lamin A/C (LMNA) siRNA (control siRNA) at a final concentration of 25 nM. Each siRNA was mixed with Lipofectamine TM RNAiMAX (1 l of 1 g/l) (Invitrogen) in 100 l of serum-free DMEM and incubated at room temperature for 30 min prior to dilution to 1 ml and addition to cells. After 24 h, the medium was changed to fresh serum-free medium prior to cytokine and drug treatments. Sequences for siRNA targeting were as follows: DUSP1 siRNA 1 (SI00374801; 5Ј-TAG CGT CAA GAC ATT TGC TGA-3Ј) and DUSP1 siRNA 2 (SI00374808; 5Ј-CTG TAC TAT CCT GTA AAT ATA-3Ј) (both Qiagen); LMNA siRNA (control siRNA) (5Ј-AAC TGG ACT TCC AGA AGA ACA-3Ј) (Qiagen); and ZFP36 siRNA 1 (5Ј-ACC GAC GAT ATA ATT ATT ATA-3Ј) and ZFP36 siRNA 2 (5Ј-ACG ACT TTA TTT ATT CTA ATA-3Ј) (both Qiagen). Because the expression of ZFP36 and TNF induced by IL1B or IL1B plus dexamethasone was unaltered by LMNA siRNA (data not shown), treatment with IL1B or IL1B plus dexamethasone in the absence of LMNA siRNA was excluded from further analyses.
Data Presentation and Statistical Analysis-GraphPad Prism version 5 software (GraphPad Software Inc., La Jolla, CA) was used for all statistical analyses. All data are plotted as means Ϯ S.E. One-way ANOVA with a Bonferroni post hoc test was used for comparing five or fewer comparisons. Because the Bonferroni post hoc test gives high and increasingly inappropriate false negative rates (i.e. type II or ␤ error) for greater than five comparisons, ANOVA with a Newman-Keuls multiple comparison test was used for greater than five comparisons, as is recommended for greater power in hypothesis testing (Prism version 5, GraphPad Software). ANOVA with a Dunnett post hoc test was used for comparisons against a single control column. Two-tailed, paired Student's t test was used for comparing two treatment groups.

Characterization of TNF Expression in the Presence of IL1B
and Dexamethasone in A549 Cells-As a prelude to analyzing the regulation of TNF expression by DUSP1 and ZFP36, we characterized the expression of TNF following IL1B treatment in the absence and presence of the synthetic glucocorticoid, dexamethasone. IL1B rapidly increased TNF mRNA ( Fig. 2A), which reached a peak at around 2 h post-stimulation and then declined steeply toward basal levels over the following 4 h. This was accompanied by a transient elevation of unspliced nuclear TNF RNA and suggests that rapid enhancement of TNF transcription contributes to the increase in TNF mRNA (Fig. 2B). Analysis of TNF mRNA stability, by actinomycin D chase methodology, showed TNF mRNA induced by IL1B to be apparently stable at 30 min post-stimulation (Fig. 2C). However, by 90 min post-IL1B, and times thereafter, TNF mRNA was reduced to ϳ50% of initial levels within 30 -40 min of the actinomycin D addition (Fig. 2C). In terms of these actinomycin D chase experiments, it is notable that spliced TNF mRNA levels appeared to continue rising following the addition of actinomycin D, an effect that was most apparent after 30 min of IL1B. We interpret this to mean that although actinomycin D prevents RNA polymerase II-dependent transcription, the rapid build-up of unspliced/unprocessed TNF RNA (see Fig. 2B) allows the ongoing production of mature TNF mRNA to continue briefly following the addition of actinomycin D. This interpretation is consistent with the fact that splicing and processing to produce mature mRNAs can take several minutes or often longer (37). Thus, the actinomycin D chase experiments should more accurately be considered as reflecting post-transcription RNA processing and maturation as well as mRNA degradation. Under basal conditions, TNF release into the supernatant was undetectable, and at the peak of TNF mRNA expression, only low levels of TNF protein were detected (Fig. 2D, left). By 4 h post-IL1B, TNF release was maximal and remained at this level for up to 18 h. TNF protein inside or associated with the cells was analyzed following the removal of supernatants, washing, and then lysis in a soft lysis buffer that was compatible with the ELISA. TNF protein in the cell lysates was first detected at 1 h (Fig. 2D, right). This was further increased at 2 h before reaching a maximum at 4 h and then declining toward 6 h. Because soluble TNF is released into the supernatant after processing of the membrane-tethered, uncleaved, ϳ25-kDa form of TNF by the metalloproteinase, TNF-converting enzyme (38), we also examined the expression of uncleaved TNF, associated with the cells, by Western blotting (Fig. 2E, left). Similar to the cell-associated TNF, detected by ELISA, uncleaved TNF was first detected at 1 h post-IL1B treatment (Fig. 2E, left). By 2 h, The effect of IL1B ϩ dexamethasone for TNF mRNA (A, right) (n ϭ 12) and unspliced nuclear TNF RNA (B, right) (n ϭ 4) is plotted as a percentage of IL1B for the indicated times. Significance, relative to time-matched IL1B-treated samples, was tested using a paired t test. *, p Ͻ 0.05. C, A549 cells were treated with IL1B (1 ng/ml) for 30, 90, 120, and 180 min. Actinomycin D (Act D; 10 g/ml) was then added (t ϭ 0), and the cells were harvested as indicated. RNA was extracted for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 3) were normalized to GAPDH and are plotted as a percentage of t ϭ 0 for each treatment as means Ϯ S.E. (left panels). A549 cells were also treated with IL1B (1 ng/ml) or IL1B plus dexamethasone (1 M) for 90 min. After 90 min, actinomycin D (10 g/ml) was added (t ϭ 0), and the cells were harvested at the indicated times for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 4) were normalized to GAPDH and are plotted as a percentage of t ϭ 0 for each treatment as means Ϯ S.E. (middle panel). The effect of IL1B plus dexamethasone at 45 min post-actinomycin D treatment is plotted as a percentage of t ϭ 0 for 45 min. Significance, relative to IL1B-treated samples, was tested using a paired t test. *, p Ͻ 0.05 (right). D, supernatants (1 ml) from the cells in A were harvested for TNF release measurement (left). Alternatively, cells were harvested after 1, 2, 4, or 6 h in soft lysis buffer, and total protein was prepared for the detection of cell-associated TNF via ELISA (right). Data (n ϭ 4), expressed in pg/ml (for supernatant) and pg/well (for cell-associated TNF), are plotted as means Ϯ S.E. Significance, relative to time-matched IL1B-treatment, was tested using a paired t test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. E, A549 cells were either not treated or stimulated with IL1B (1 ng/ml) for 1, 2, or 6 h (left), or A549 cells were either not treated or stimulated with IL1B (1 ng/ml) or a combination of IL1B and dexamethasone (1 M) for 2 h (right), and total protein was prepared for Western blot analysis of uncleaved (ϳ25 kDa) TNF and GAPDH. Blots representative of four such experiments are shown. Following densitometric analysis, data (n ϭ 4) were normalized to GAPDH and plotted as means Ϯ S.E. Significance, relative to non-treated (left) or IL1B-treated (right) samples, tested using ANOVA with a Dunnett post hoc test (left panel) or paired t test (right panel), is indicated. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. expression of uncleaved ϳ25-kDa TNF was maximal, and by 6 h, this had largely returned to basal levels.
The presence of dexamethasone co-incubated with IL1B resulted in a partial loss of TNF mRNA at all times ( Fig. 2A). This effect was also observed for unspliced nuclear TNF RNA, suggesting that transcriptional repression accounts for much of the repressive effect of dexamethasone on TNF mRNA (Fig.  2B). Similarly, although the decay of TNF mRNA, following 90 min of IL1B, revealed a t1 ⁄ 2 of 30 -40 min, this was further reduced by dexamethasone, which significantly increased the loss of TNF mRNA observed at 30, 45, and 60 min after actinomycin D addition (Fig. 2C, right). At the level of TNF protein, dexamethasone was highly effective, with nearly complete repression of cell-associated, uncleaved, and secreted TNF, observed at all times (Fig. 2, D and E). These data suggest translational or possibly post-translational repression by dexamethasone occurring in addition to transcriptional and post-transcriptional mechanisms of repression.
Analysis of DUSP1 Expression in the Presence of IL1B and Dexamethasone in A549 Cells-DUSP1 protein was strongly induced by IL1B and dexamethasone in A549 cells. Post-IL1B treatment, DUSP1 expression reached a peak at 1 h, before declining by 2 h and reaching basal levels by 6 h (Fig. 3A). Dexamethasone alone produced a small increase in DUSP1 at 1 h, and by 2 h, this was significantly increased. In the case of IL1B plus dexamethasone treatment, DUSP1 protein was strongly induced at 1 h and declined by 2 h post-treatment. However at 6 h, whereas IL1B-induced DUSP1 protein was undetectable, dexamethasone alone produced a significant increase in DUSP1 that was maintained in the presence of IL1B. These results are consistent with previous observations (24,28).

Analysis of ZFP36 Expression in the Presence of IL1B and
Dexamethasone in A549 Cells-Western blot analysis showed ZFP36 expression to be essentially undetectable in untreated cells, but expression, comprising a protein doublet, was dramatically induced within 1 h of IL1B treatment (Fig. 3B). Expression remained high at 2 h but was mostly shifted to the upper, presumably phosphorylated (19,39), form of ZFP36. By 6 h, ZFP36 expression was reduced. With dexamethasone alone, there was no apparent expression of ZFP36 protein at either 1 or 2 h (Fig. 3B). In the presence of IL1B, dexamethasone significantly reduced ZFP36 expression at 1 and 2 h (Fig. 3B). By 6 h, dexamethasone alone produced a small increase in ZFP36 expression, and the IL1B-induced expression of ZFP36, although modest, was no longer repressed.
Effect of MAPK Inhibitors on DUSP1 and ZFP36 Expression-Because dexamethasone reduces MAPK activation, the effect of selective inhibitors of each of the three major MAPK pathways was tested on both IL1B-induced DUSP1 and ZFP36 expression. A549 cells were treated with maximally effective concentrations of SB203580 (10 M), U0126 (10 M), or JNK inhibitor 8 (JNK-IN-8) (10 M), which inhibit the p38, extracellular signal-regulated kinase (ERK), and JNK MAPK pathways, respectively. SB203580 and JNK-IN-8 were without obvious effect on DUSP1 expression induced by IL1B at 1 h (Fig. 3C). However, UO126 consistently resulted in a partial inhibition of DUSP1, and in the presence of all three inhibitors, IL1B-induced DUSP1 protein expression was totally prevented. These data confirm the requirement for MAPK activity in the induction of DUSP1 expression by IL1B.
The presence of SB203580 completely prevented ZFP36 expression at all times, whereas the effects of U0126 or JNK-IN-8 were more modest, without significant inhibition of ZFP36 expression being achieved at any time (Fig. 3D). With all three inhibitors together, there was a complete loss of IL1Binduced ZFP36 expression at all times. These data are therefore consistent with the concept that reduced MAPK activity promotes loss of ZFP36 and could lead to enhanced mRNA stability and expression of ARE-containing mRNAs (Fig. 1).
Analysis of DUSP1 Expression in the Presence of IL1B and Dexamethasone in Primary HBE Cells-To validate results obtained in the A549 cells, DUSP1 mRNA and protein expression was examined in primary HBE cells. DUSP1 mRNA was strongly induced by both IL1B and dexamethasone at 1 h, following which the mRNA expression declined rapidly by 2 h post-treatment (Fig. 4A, left). At 18 h, although IL1B-induced DUSP1 mRNA had returned to basal levels, dexamethasone strongly induced DUSP1 mRNA. In the presence of IL1B, dexamethasone enhanced DUSP1 mRNA at multiple time points. However, at 18 h, DUSP1 mRNA expression was primarily dexamethasone-dependent, with no effect of IL1B. Likewise, DUSP1 protein was strongly induced by IL1B and IL1B plus dexamethasone (Fig. 4B, top panels). Dexamethasone alone induced DUSP1 expression at 6 and 18 h post-treatment, and there was little or no further effect of IL1B.
Analysis of ZFP36 Expression in the Presence of IL1B and Dexamethasone in Primary HBE Cells-In HBE cells, ZFP36 mRNA was strongly induced by IL1B (Fig. 4A, right). This was maximal at 1 h post-stimulation and then declined rapidly to basal levels. Dexamethasone alone produced no changes in ZFP36 mRNA expression until 18 h and also did not affect ZFP36 expression induced by IL1B. In the case of dexamethasone co-treatment with IL1B, ZFP36 mRNA was strongly induced at 1 h and declined steeply over the following 18 h post-treatment. Whereas the effect of IL1B was predominant at early time points, the late phase expression of ZFP36 mRNA at 18 h was largely dexamethasone-dependent. Similar to A549 cells, ZFP36 protein was also rapidly induced by IL1B at 1 and 2 h and declined thereafter (Fig. 4B, bottom). At 18 h, ZFP36 expression was variable. Dexamethasone alone did not produce any apparent increase in ZFP36 protein expression, and there was no obvious effect on IL1B-induced ZFP36.
Characterization of TNF Expression in the Presence of IL1B and Dexamethasone in HBE Cells-Similar to A549 cells, treatment with IL1B rapidly increased TNF mRNA (Fig. 4C, left), which reached a peak at 2 h post-stimulation and then declined steeply over the following 4 h. In both the A549 and HBE cells, IL1B-induced TNF were detected by real-time PCR with C T values of 22-23 cycles, suggesting very similar levels of expression (data not shown). Under basal conditions, TNF release into the supernatant was undetectable, and low levels of TNF protein were detected at 18 h after IL1B treatment (Fig. 4D). These data are consistent with previous findings in HBE cells showing low release of TNF in virus-treated HBE cells (40). Like the A549 cells, we also examined the expression of uncleaved ϳ25-kDa TNF by Western blotting (Fig. 4E). IL1B treatment rapidly increased uncleaved TNF expression, which was clearly detectable at 1 h, reached a peak of expression at 2 h, and was reduced by 6 h. Interestingly, these blots were performed in parallel with those for the A549 cells presented in Fig. 2E. Because all of the blotting and detection conditions, including exposure time, were identical, we can conclude that uncleaved ϳ25-kDa TNF protein expression is similar in the A549 and HBE cells. Dexamethasone in the presence of IL1B produced a partial loss of TNF mRNA at all times (Fig. 4C, right). At the level of TNF protein, dexamethasone also produced partial repression of uncleaved and soluble TNF at 2 and 18 h, respectively (Fig. 4, D  and E, right).
A Role for ZFP36 in Regulating TNF mRNA Expression-To assess the role of ZFP36 in the regulation of TNF mRNA expression, a siRNA-based strategy was employed. Two independent siRNA molecules directed to ZFP36, but not an unrelated LMNA siRNA, produced robust knockdown of IL1B-in- . The effect of IL1B ϩ dexamethasone for TNF mRNA (n ϭ 4) is plotted as a percentage of IL1B for the indicated times (right). Significance, relative to time-matched IL1B-treated samples, was tested using a paired t test. *, p Ͻ 0.05. D, supernatants, from the cells in A, were harvested for TNF release measurement. E, HBE cells were either not treated or stimulated with IL1B (1 ng/ml) for 1, 2, or 6 h (left). Alternatively, cells were not treated or stimulated with IL1B (1 ng/ml) or a combination of IL1B and dexamethasone (1 M) for 2 h (right). Total protein was prepared for Western blot analysis of uncleaved (ϳ25 kDa) TNF and GAPDH. Blots representative of at least four such experiments are shown. Following densitometric analysis, data (n ϭ 4) were normalized to GAPDH and plotted as means Ϯ S.E. Significance, relative to non-treated (left panel) or IL1B-treated (right panel) samples was tested using a paired t test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.
duced ZFP36 expression at all times tested (Fig. 5A). With respect to TNF mRNA induced by IL1B, ZFP36 knockdown had no effect at 1 h but produced marked and significant increases in TNF mRNA expression at 2 h. Although this effect was lost by 6 h post-IL1B, the data are consistent with the appearance of phosphorylated and active ZFP36 at 2 h. This confirms ZFP36 as a negative regulator of TNF mRNA expression. Parallel analysis of unspliced nuclear TNF RNA induced by IL1B revealed no effect of ZFP36 knockdown, and this is consistent with a post-transcriptional role for ZFP36 (Fig. 5A, bottom panels). To explore possible post-transcriptional regulation of TNF mRNA by ZFP36, A549 cells were treated with LMNA or ZFP36-targeting siRNAs prior to treatment with IL1B for 90 min prior to actinomycin D chase (Fig. 5B). As in Fig. 2B, TNF mRNA decayed with a t1 ⁄ 2 of 30 -40 min, and this was not altered by the LMNA siRNA. However, both the ZFP36-targeting siRNAs significantly reduced the loss of TNF mRNA (Fig. 5B). This indicates a role for ZFP36 in the posttranscriptional regulation of IL1B-induced TNF mRNA in A549 cells.
The effect of ZFP36 knockdown was also tested on TNF mRNA and unspliced nuclear TNF RNA expression in the presence of dexamethasone and IL1B (Fig. 5A). At each time, dexamethasone produced a significant repression of TNF mRNA expression, but the level of TNF mRNA was unaltered by ZPF36 knockdown. In terms of the percentage inhibition by dexamethasone, the elevated levels of expression in the presence of the ZFP36-targeting siRNA resulted in an increased percentage repression by dexamethasone (Fig. 5A, right panels). There were no significant effects of ZFP36 knockdown observed at the level of unspliced nuclear TNF RNA (Fig. 5A, bottom panels).
Effect of DUSP1 Overexpression on ZFP36 and TNF mRNA Expression-Because dexamethasone induces DUSP1 expression to reduce IL1B-dependent activation of MAPKs in A549 cells (24), the effect of DUSP1 overexpression was assessed on ZFP36 protein expression. DUSP1 overexpression was confirmed following infection with DUSP1-overexpressing adenovirus but not control (Ad-GFP) virus (Fig. 6). Substantial losses of p38, JNK, and ERK MAPK phosphorylation, and therefore activity have been shown previously and were not reassessed here (24). Untreated cells showed no ZFP36 expression, and this was unaltered by either adenovirus (Fig. 6). The ability of IL1B to induce ZPF36 protein was unaffected by Ad-GFP, but Ad-DUSP1 produced a nearly complete loss of ZFP36 expression at all times (Fig. 6). Parallel analysis of TNF mRNA revealed no effect of either virus on basal TNF mRNA expression (Fig. 6, bottom panels). Likewise, Ad-GFP had no significant effect on TNF expression, but the DUSP1-overexpressing adenovirus significantly reduced TNF mRNA expression at 1 h (Fig. 6, bottom panels). This repression was reduced at 2 h, and by 6 h, TNF mRNA expression appeared unaltered by Ad-DUSP1. Although these data support a role for MAPKs in the early induction of TNF mRNA expression by IL1B, it appears that DUSP1 overexpression may overcome this effect on TNF mRNA expression at longer time points.
Effect of Small Molecule MAPK Inhibitors on TNF Expression-To further explore roles for MAPK pathways in the regulation of TNF mRNA expression by IL1B, A549 cells  ). B, A549 cells were incubated with LMNA or ZFP36-specific siRNAs for 24 h before being treated with IL1B (1 ng/ml) for 90 min. Actinomycin D (Act D; 10 g/ml) was then added (t ϭ 0), and cells were harvested at the indicated times. RNA was extracted for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 4), normalized to GAPDH, are plotted as a percentage of t ϭ 0 for each treatment as means Ϯ S.E. Significance, relative to time-matched IL1B-treated samples was tested using a paired t test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 (left panel). The effect of IL1B plus LMNA siRNA and IL1B plus ZFP36 siRNAs at 45 min post-actinomycin D treatment is plotted as a percentage of t ϭ 0 for 45 min. Significance, relative to IL1B-treated samples, was tested using ANOVA with a Dunnett post hoc test. **, p Ͻ 0.01 (right).
were treated with maximally effective concentrations of SB203580 (10 M), U0126 (10 M), or JNK-IN-8 (10 M). TNF mRNA expression was profoundly induced by IL1B at all times, yet this was essentially unaffected by any of the MAPK inhibitors at either 1 or 2 h post-IL1B (Fig. 7A). However, by 6 h, SB203580 significantly elevated expression of TNF mRNA relative to IL1B. Because both dexamethasone and DUSP1 overexpression inhibit all three MAPK pathways, the combined effect of the three MAPK pathway inhibitors was assessed on TNF mRNA expression. One hour after IL1B treatment, this resulted in a substantial and significant 81.2 Ϯ 2.4% loss of TNF mRNA (Fig. 7A, right). However, by 2 h after IL1B treatment, TNF expression was 54.7 Ϯ 6.2% of IL1B control, and at 6 h, there was an enhancement to 183.2 Ϯ 22.2%. Thus, the combined inhibition of MAPKs reduced TNF mRNA expression at early time points, but at later times, this effect was both reversed and overcome.
To assess the contribution of transcriptional and post-transcriptional events, the accumulation of unspliced nuclear TNF RNA and TNF mRNA half-life was assessed. In the presence of each MAPK pathway inhibitor alone, there was little or no effect on the accumulation of unspliced nuclear TNF RNA at 1 or 2 h (Fig. 7B). Although there was little effect of JNK-IN-8 at 4 or 6 h post-IL1B, SB203580 and U0126 produced quite substantial, but not significant, enhancements relative to IL1B at 4 or 6 h. In the context of the three inhibitors combined, there was significant inhibition of unspliced nuclear TNF RNA induced by IL1B at 1 h and little or no effect at 2 h, followed by enhanced unspliced nuclear TNF RNA accumulation at 4 and 6 h (Fig. 7B). These data suggest that MAPK-dependent tran-scription of TNF occurs at early times but that MAPK inhibition leads to enhanced TNF transcription at later times.
Analysis of TNF mRNA stability (Fig. 2C) showed that following short, 30-min, IL1B treatment times, TNF mRNA appeared stable. In Fig. 7C, this is confirmed with essentially no overall loss of TNF mRNA over 60 min following actinomycin D addition. However, in the presence of SB203580, there was a reduction in the post-actinomycin D accumulation combined with an overall ϳ50% reduction in TNF mRNA over this period. Although this result may represent a mixed effect on post-transcription nuclear processing and/or mRNA stability (see above), the net effect, following transcriptional arrest, is that p38 inhibition considerably enhances loss of TNF mRNA. Pilot actinomycin D chase experiments suggest that SB203580 resulted in no effect on the loss of TNF mRNA at either 2 or 4 h after IL1B treatment (data not shown). Therefore, the effect of SB203580 was assessed on the loss of TNF mRNA after actinomycin D addition following IL1B and IL1B plus SB203580 treatments of 90, 120, 180, and 240 min. As observed at 30 min of IL1B, the presence of SB203580 significantly reduced TNF mRNA loss following 90-min IL1B stimulation but revealed no effect following 120-, 180-, or 240-min treatments (Fig. 7C,  right). These data suggest that p38 inhibition can destabilize and/or reduce nuclear RNA processing at 30 and 90 min of IL1B but that there was no apparent stabilization of TNF mRNA at longer treatment times. Thus, MAPK inhibition results in enhanced transcription of TNF at longer time points but may not, despite a dramatic loss of ZFP36 expression, produce an obvious stabilization of TNF mRNA. One explanation for this could be that in the absence of prior p38-dependent mRNA stabilization (occurring at 30 -90 min post-IL1B), there is no possibility of any subsequent p38-dependent mRNA destabilization. These data therefore support the existence of MAPKdependent feed-forward mechanisms of inhibition of TNF mRNA but do not specifically support mRNA stabilization as a mechanism. However, because dexamethasone co-treatment with IL1B has little initial repressive effect on MAPKs yet produces profound MAPK inhibition at longer times, it is likely that marked differences in MAPK-dependent feed-forward and feedback regulatory loops may occur. Likewise, DUSP1 knockdown in A549 cells only transiently enhanced IL1B-induced MAPK activity, and this could also result in temporally distinct effects on ZFP36 and ZFP36 target genes (24).
Effect of DUSP1 Knockdown on ZFP36 Protein and TNF mRNA Expression-To directly address the effects of reducing DUSP1 expression on TNF mRNA expression, cells were transfected with LMNA control or DUSP1-targeting siRNAs. Following IL1B treatment, DUSP1 protein was robustly induced at 1 h, and this was largely blocked by the two DUSP1-targeting siRNAs (Fig. 8A). Analysis of TNF mRNA revealed no effects of DUSP1 knockdown at either 1 or 2 h post-IL1B. However, by 6 h, the DUSP1-targeting siRNAs substantially and significantly reduced TNF mRNA expression relative to LMNA control siRNA (Fig. 8B). Analysis of unspliced nuclear TNF RNA also showed no significant effects of DUSP1 knockdown at 1 or 2 h, although a trend toward reduced unspliced nuclear TNF RNA was observed at 6 h. ZFP36 protein expression revealed little or no effect of DUSP1 knockdown at 1 h post-IL1B treatment (Fig.   FIGURE 6. Effect of DUSP1 overexpression on ZFP36 and TNF mRNA expression. A549 cells were either not infected or infected with Ad5-DUSP1 or Ad5-GFP (control) at a multiplicity of infection of 10 for 24 h before IL1B treatment (1 ng/ml). Cells were harvested after 1, 2, or 6 h, and total protein was prepared for Western blot analysis of DUSP1, ZFP36, and GAPDH. Blots representative of at least four such experiments are shown. NS, nonspecific band. Following densitometric analysis for ZFP36, data (n ϭ 4) were normalized to GAPDH and plotted as means Ϯ S.E. (error bars) (middle panels). Significance, using ANOVA with a Dunnett post hoc test, is indicated. *, p Ͻ 0.05. Cells were also harvested for real-time PCR analysis of TNF and GAPDH (bottom panels). Data (n ϭ 4), normalized to GAPDH, are plotted as means Ϯ S.E. Significance relative to time-matched IL1B and Ad5-GFP-treated samples, was tested using ANOVA with a Bonferroni post hoc test. **, p Ͻ 0.01; ***, p Ͻ 0.001. 8A). However, by 2 h post-IL1B treatment, the two DUSP1targeting siRNAs both markedly increased expression of the upper ZFP36 band observed on Western blots (Fig. 8A). These data suggest that a ZFP36-dependent post-transcriptional effect could contribute to the loss of TNF mRNA following DUSP1 knockdown. This is addressed below. The effect of IL1B ϩ MAPK inhibitors for TNF mRNA is plotted as a percentage of IL1B at the indicated times. B, RNA from the experiments in A were subjected to real-time PCR analysis of unspliced nuclear (un) TNF RNA and U6. Data (n ϭ 4) were normalized to U6 and are plotted as means Ϯ S.E. The effect of IL1B ϩ MAPK inhibitors for unspliced nuclear TNF mRNA is plotted as a percentage of IL1B at the indicated times. For both A and B, significance, relative to either non-treated or IL1B-treated samples, was tested using ANOVA with a Bonferroni post hoc test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. C, A549 cells were treated with IL1B (1 ng/ml) or pretreated for 30 min with SB203580 at 10 M prior to IL1B stimulation for 30 min. Following this (t ϭ 0), actinomycin D (Act D; 10 g/ml) was added, and cells were harvested at the indicated times. RNA was extracted for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 4) normalized to GAPDH, are plotted as a percentage of t ϭ 0 for each treatment as means Ϯ S.E. (left). Data following 45 min of actinomycin D treatment are shown (right). A549 cells were also treated with IL1B or IL1B ϩ SB203580 at 10 M for 90, 120, 180, and 240 min, prior to the addition of Act D (t ϭ 0) for 45 min and analyzed as above (right panels). Significance, relative to time-matched IL1B-treated samples, was tested using a paired t test. *, p Ͻ 0.05. Data (n ϭ 4 -8), normalized to GAPDH (for TNF mRNA) or U6 (for unspliced nuclear (un) TNF RNA), were expressed as a percentage of LMNA siRNA plus IL1B-stimulated for each time and are plotted as means Ϯ S.E. (error bars). Significance was tested using ANOVA with a Newman-Keuls multiple comparison test. Significance for specific comparisons is indicated. **, p Ͻ 0.01; ***, p Ͻ 0.001. The percentage of IL1B plus dexamethasone/IL1B for the LMNA siRNA was also plotted and compared with that for each of the DUSP1-specific siRNAs using ANOVA with a Dunnett post hoc test. **, p Ͻ 0.01 (right).
The effect of DUSP1 knockdown induced in the presence of IL1B plus dexamethasone (Fig. 8A) was also examined on ZFP36 expression and the repression of TNF mRNA by dexamethasone. Thus, dexamethasone reduced the level of IL1Binduced ZFP36, and in the presence of the DUSP1-targeting siRNA, at 2 h, ZFP36 expression was enhanced (Fig. 8A). However, although significant repression of IL1B-induced TNF mRNA by dexamethasone was observed at all times, there was no effect of the DUSP1-targeting siRNAs. Nevertheless, at 6 h, the large loss of IL1B-induced TNF mRNA in the presence of each DUSP1-targeting siRNA meant that there was a greatly reduced percentage repression in the presence of dexamethasone (Fig. 8B, right panels). No significant effects were apparent with respect to unspliced nuclear TNF RNA, suggesting that TNF transcription rate was not affected.
Role of ZFP36 in the Loss of TNF mRNA following DUSP1 Knockdown-To explore a possible role for ZPF36 in mediating the enhanced loss of TNF mRNA following DUSP1 knockdown, cells were transfected with DUSP1-and ZFP36-targeting siRNAs either alone or in combination. In each case, robust knockdown of DUSP1, ZFP36, or both together was achieved (Fig. 9A). As in Fig. 5A, loss of IL1B-induced ZFP36 had little effect on TNF mRNA expression at 6 h (Fig. 9A, bottom).
Equally, targeting of DUSP1 alone resulted in a profound and significant reduction in TNF mRNA at 6 h (Fig. 9A, bottom). This is consistent with the data in Fig. 8B. However, this repressive effect of the DUSP1 siRNAs was significantly attenuated in the additional presence of each of the ZFP36-targeting siRNAs (Fig. 9A, bottom). This confirms a role for ZFP36 in the enhanced loss of TNF mRNA following knockdown of DUSP1.
Given that ZFP36 is an mRNA-destabilizing protein, the effect of DUSP1 knockdown was assessed on TNF mRNA stability. A549 cells were therefore treated with IL1B for 2 h prior to the addition of actinomycin D (t ϭ 0), and cells were harvested after 45 min (Fig. 9B). As shown in Figs. 2 and 5, TNF mRNA was reduced to around 50% during this 45-min time period. However, in the presence of the DUSP1-targeting siRNAs, there was significantly enhanced loss of TNF mRNA, indicating that in addition to enhancing ZFP36 expression at 2 h (Fig. 8A), the DUSP1 siRNAs also reduce TNF mRNA stability. To examine the role of ZFP36 in this loss, the effects of ZFP36-targeting siRNAs were tested alone and in the presence of the DUSP1 siRNA. As previously shown (Fig. 5B), the ZFP36 siRNAs significantly reduced the loss of TNF mRNA, and this confirms a role for ZFP36 in the negative regulation of TNF mRNA stability (Fig. 9B). However, in the presence of DUSP1 FIGURE 9. Role of ZFP36 in the enhanced loss of TNF mRNA following DUSP1 knockdown. A, A549 cells were incubated with LMNA (control) or DUSP1and/or ZFP36-specific siRNAs for 24 h before being treated with IL1B (1 ng/ml). Although not indicated, LMNA siRNA was added to all of the treatment groups except for the combination siRNA treatments to maintain a constant concentration of siRNAs across all of the treatments. Cells were harvested after 1 h, and total proteins were prepared for Western blot analysis of DUSP1, ZFP36, and GAPDH. Blots representative of at least four such experiments are shown (top). NS, nonspecific band. Cells were also harvested after 6 h for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 4) normalized to GAPDH were expressed as a percentage of LMNA siRNA plus IL1B and plotted as means Ϯ S.E. (error bars). Significance was tested using ANOVA with a Newman-Keuls multiple comparison test (bottom). B, A549 cells were treated as in A before being treated with IL1B (1 ng/ml) for 2 h. Following this (t ϭ 0), actinomycin D (Act D; 10 g/ml) was added, and cells were harvested after 45 min. RNA was extracted for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 6) normalized to GAPDH are plotted as a percentage of t ϭ 0 for each treatment as means Ϯ S.E. (error bars). For both A and B, significance between LMNA control siRNA plus IL1B and each of the DUSP1and/or ZFP36-targeting siRNAs plus IL1B is shown (#). Other comparisons are specifically indicated. * or #, p Ͻ 0.05; ##, p Ͻ 0.01; *** or ###, p Ͻ 0.001. C, A549 cells were incubated with LMNA (control) or DUSP1 and ZFP36-specific siRNAs for 24 h before being treated with IL1B (1 ng/ml) or IL1B plus dexamethasone (Dex; 1 M) as indicated. Cells were harvested after 1 h, and total proteins were prepared for Western blot analysis of DUSP1, ZFP36, and GAPDH. Blots representative of at least four such experiments are shown. D, A549 cells were treated as in C and harvested after 1, 2, or 6 h for real-time PCR analysis of TNF and GAPDH. Data (n ϭ 4) normalized to GAPDH were expressed as a percentage of LMNA siRNA plus IL1B-stimulated for each time and are plotted as means Ϯ S.E. Significance was tested using ANOVA with a Newman-Keuls multiple comparison test. Significance for specific comparisons is indicated. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 (left). The percentage of IL1B plus dexamethasone/IL1B for the LMNA siRNA was also plotted and compared with that for each of the DUSP1and ZFP36-specific siRNAs using ANOVA with a Dunnett post hoc test. *, p Ͻ 0.05 (right). plus ZFP36 siRNAs, the repressive effects of the DUSP1 siRNAs were blocked, and TNF mRNA decayed to the same extent as for ZFP36 knockdown alone (Fig. 9B). These data therefore support a role for ZFP36 in the enhanced decay of TNF mRNA following DUSP1 knockdown.
The effect of combined knockdown of both DUSP1 and ZFP36 was also assessed in the presence of IL1B and IL1B plus dexamethasone at 1, 2, and 6 h. In each case, clear knockdown of DUSP1 and ZFP36 was observed (Fig. 9C). With respect to the TNF mRNA induced by IL1B, the combined knockdown produced a modest but significant repression at 1 h, but no effect at 2 h, with, as shown in Fig. 9A, significant repression observed at 6 h (Fig. 9D, left). In the presence of dexamethasone, TNF mRNA was significantly repressed at all times in a manner that is consistent with Fig. 2A. At 1 and 2 h, there was no effect of the combined DUSP1 plus ZFP36 siRNAs on TNF mRNA in the presence of IL1B plus dexamethasone (Fig. 9D,  left). However, as a percentage of IL1B ϩ LMNA at 6 h, there was significantly reduced expression of TNF mRNA induced with IL1B plus dexamethasone (16.7 Ϯ 1.0%) compared with 9.7 Ϯ 0.6 and 8.3 Ϯ 1.2% in the presence of the DUSP1/ZFP36 combined knockdown (Fig. 9D, left). However, accounting for the changes in IL1B-induced TNF mRNA with each siRNA, the loss of IL1B-induced TNF expression at 1 h produced a reduction in the percentage repression by dexamethasone, whereas there were no apparent changes in the percentage repression by dexamethasone at either 2 or 6 h (Fig. 9D, right panels).
Roles for MAPKs, DUSP1, and ZFP36 in the Regulation of TNF Protein Expression-The above analyses address the regulation of TNF mRNA expression by DUSP1 and ZFP36 in the context of IL1B and dexamethasone. Here, the effects on the regulation of TNF protein release are assessed. TNF protein accumulation in the supernatants was totally prevented by SB203580 and significantly inhibited by 74.0 Ϯ 8.8 and 69.0 Ϯ 20.2% by U0126 and JNK-IN-8, respectively (Fig. 10A). Likewise, the combination of SB203580, U0126, and JNK-IN-8 resulted in a nearly complete loss of TNF release. Similarly, adenoviral expression of DUSP1 also significantly inhibited TNF release, whereas the control, Ad-GFP virus, was without effect (Fig 10B). These data support a major role for MAPKs in the IL1B-induced release of TNF protein.
To examine possible roles for ZFP36 and DUSP1, TNF release was measured from the experiments shown in Figs. 5 and 8, respectively. Knockdown of ZFP36 significantly enhanced the release of TNF following IL1B treatment (Fig.  10C) and is consistent with the enhanced expression of TNF mRNA observed at 2 h in Fig. 5. Conversely, targeting of DUSP1 was without obvious effect on TNF release (Fig. 10D). The effects of targeting ZFP36 and DUSP1 were also reconfirmed in experiments where the effect of combined knockdown of ZFP36 and DUPSP1 was also examined (Fig. 10E). Thus, as shown in Fig. 10C, ZFP36 siRNA enhanced IL1B-induced TNF release, and DUSP1 siRNA was without any marked effect (Fig.  10E). In the presence of ZFP36 plus DUSP1 siRNA, TNF release Significance, using ANOVA with a Bonferroni multiple comparison test, is indicated. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. C-E, A549 cells were incubated with LMNA (control) or DUSP1-and/or ZFP36-specific siRNAs for 24 h before being treated with IL1B (1 ng/ml) or IL1B plus dexamethasone (Dex; 1 M) as indicated. Cell supernatants were harvested after 6 h for TNF release measurement. Data (n ϭ 4) expressed in pg/ml are plotted as means Ϯ S.E. Significance, using ANOVA with a Newman-Keuls multiple comparison test, is indicated. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 (left panels). In each case, the percentage of IL1B plus dexamethasone/IL1B for the LMNA siRNA was also plotted and compared with that for each of the DUSP1-and ZFP36-specific siRNAs using ANOVA with a Dunnett post hoc test (right panels). For E (left), LMNA siRNA was added in all of the treatment groups except for the combination siRNA treatments in order to maintain a constant siRNA concentration across all of the treatments.
was more than following DUSP1 knockdown alone, but in each case, this was not significantly different from the effect of targeting ZFP36 alone (Fig. 10E, left).
In the presence of IL1B plus dexamethasone, siRNA targeting of ZFP36 and DUSP1 either individually (Fig. 10, C and D) or in combination (Fig. 10E, middle) showed no significant effect on TNF release. Moreover, the percentage repression by dexamethasone was also unaltered by either treatment (Fig. 10, C-E, right panels).

Discussion
Appropriate regulation of inflammatory gene expression is central to inflammation and its resolution. Equally, understanding these processes will aid in the identification of therapeutic agents that target inflammation. However, whereas the regulation of signal transduction and gene expression involves a myriad of signal-ing molecules leading to transcriptional, post-transcriptional, translational, and often post-translational control of gene expression, it is routine to depict these assemblies as simple linear pathways. Using TNF biosynthesis as an example, regulation involves the integrated effect of feedback and feed-forward control loops (2,41). The current study documents interplay between the phosphatase, DUSP1, and the mRNA-destabilizing protein, ZFP36, in the regulation of TNF mRNA as a model ARE-containing transcript. This network behaves in a non-linear fashion, such that modulation of individual factors may produce opposing outputs (Fig.  11A). Because the expression of DUSP1, ZFP36, and TNF are similarly regulated by IL1B and dexamethasone in both A549 cells and primary HBE cells, our findings are likely to be physiologically and therapeutically relevant.
In a prior study, siRNA depletion of DUSP1 expression resulted, as expected for a feedback regulator of MAPKs, in FIGURE 11. Regulation of TNF gene expression following DUSP1 silencing and glucocorticoid treatment. Schematics representing possible regulatory networks are shown along with summarized data from the present and previous work (24). A, with reduced DUSP1 expression, there is reduced negative feedback control of MAPKs, leading to enhanced MAPK activity (at 1 h) (boldface type) at the times where DUSP1 expression would have been elevated. This could promote enhanced TNF expression. However, enhanced MAPK activity enhances expression of ZFP36 (at 2 h) (boldface type), which may decrease TNF expression (at 6 h). Actual expression of TNF is dependent on the temporal interplay of these competing regulatory processes. B, in the presence of glucocorticoid co-treatment, DUSP1 expression is enhanced (boldface type) and promotes inactivation of MAPKs (gray). Loss of MAPK activity reduces the expression of feed-forward negative control genes, such as ZFP36. However, ZFP36 is also modestly induced by glucocorticoid alone (in brackets), and this may help maintain feed-forward control. Given the net loss of IL1B-induced ZFP36 expression in the presence of glucocorticoids (at 1 and 2 h), additional glucocorticoid effectors must exist to maintain repression of TNF (at later times). Repression of MAPKs is also maintained by non-DUSP1-dependent mechanisms. Positive signaling/ expression (blue) is represented by arrows. Negative effects are indicated (red) by lines ending in a T-bar. Time course expression of phospho-p38, DUSP1, and ZFP36 protein and TNF mRNA in the presence of IL1B and IL1B plus DUSP1 siRNA (A) or IL1B plus dexamethasone (B) is shown. The protein expression data for DUSP1 and ZFP36 were generated following densitometric analysis of Western blots showed in Fig. 3. For phospho-p38, Western blots were taken from Shah et al. (24). For phospho-p38, DUSP1, ZFP36, and TNF, at each time, the effect of IL1B plus Dex expressed as a percentage of IL1B is plotted as a mean (B, right panels). enhanced activity of MAPK pathways following IL1B treatment (24). However, this effect was transient and did not necessarily produce robust increases in the mRNA expression of inflammatory genes whose expression was MAPK-dependent. More surprisingly, loss of DUSP1 reduced, relative to control, the mRNA expression for a number of inflammatory genes, including CCL2, CXCL3, and PTGS2, at longer (6 h) times (24). Given that 1) such mRNAs contain AREs, 2) ZFP36 can destabilize ARE-containing mRNAs, and 3) ZFP36 is induced in A549 cells in a p38 MAPK-dependent manner, ZFP36 represents a candidate to explain this enhanced loss of inflammatory mRNAs following DUSP1 knockdown (16,(41)(42)(43). We confirm, in A549 cells, that ZFP36 is induced by IL1B to reduce TNF mRNA stability and exerts feed-forward inhibition of TNF mRNA and protein expression. Furthermore, inhibition of MAPK pathways, in particular using SB203580 to inhibit p38 MAPK, prevents ZFP36 expression and increased TNF mRNA expression 6 h following IL1B treatment. Despite this, we could not document any increases in TNF mRNA stability following p38 MAPK inhibition. Rather, combined MAPK inhibition reduced TNF mRNA expression at short times (1 or 2 h), and the increased TNF mRNA expression at 6 h appeared to be primarily due to enhanced transcription. These data therefore suggest the existence of as yet uncharacterized MAPK-dependent mechanisms that negatively regulate TNF transcription.
Despite the above, inhibition of MAPKs, in particular p38 MAPK, using kinase inhibitors or by DUSP1 overexpression, prevents MAPK activity at all times. However, this may produce different functional outcomes compared with a transient modulation of MAPK activity (41). Thus, following 30 min of IL1B stimulation, TNF mRNA did not decline below the 100% level within 45 min of actinomycin D chase. This contrasts with longer IL1B treatment times, where ZFP36 is expressed and TNF mRNA decayed to 50% within 30 -40 min of actinomycin D treatment. In the presence of the p38 MAPK inhibitor, SB203580, this apparent mRNA stabilization did not occur, and at longer times, there was no destabilization of what was presumably an already unstable transcript. This is consistent with findings in macrophage, where a transient stabilization of TNF mRNA occurred 30 -60 min post-LPS treatment but was lost by 2 h (44). Likewise, many reports indicate that p38 MAPK plays a role in the ARE-dependent stabilization of mRNAs, including IL8, PTGS2, and TNF (12,42,(45)(46)(47). By comparison, following silencing of DUSP1, the initial MAPK activation by IL1B occurs normally, but due to reduced feedback control, MAPK activity is not adequately down-regulated (i.e. at 1 h after IL1B treatment in A549 cells) (24) (Fig. 11A). In this situation, we observe increased expression of ZFP36 at 2 h, followed by a subsequently enhanced loss of ARE-containing mRNA expression at 6 h, including, as is now shown, for TNF (Fig. 11A). This scheme is consistent with a role for p38 MAPK in inducing ZFP36 expression and activity to destabilize ARE-containing mRNAs, such as TNF (18,48). Thus, the dramatically enhanced loss of TNF mRNA following DUSP1 knockdown was not primarily associated with reduced TNF transcription, although some loss of transcription was evident and is consistent with the effect of total inhibition by the MAPK inhibitors. Rather, the enhanced loss of TNF mRNA following DUSP1 knockdown was attenuated by the further knockdown of ZFP36. This was associated with reduced TNF mRNA stability, an effect that was also attenuated by ZFP36 silencing. These data directly confirm that the loss of DUSP1 enhances ZFP36 expression to increase negative (incoherent) feed-forward control of TNF by reducing TNF mRNA stability (Fig. 11A). However, despite these data, loss of ZFP36 did not completely ablate all of the effects of DUSP1 knockdown, suggesting that other non-ZFP36-dependent effects may also impact TNF mRNA expression.
The above findings highlight a number of key points with respect to the regulation of TNF mRNA expression by glucocorticoids. Dexamethasone attenuated the accumulation of unspliced nuclear TNF RNA induced by IL1B treatment. This effect was largely reflected in TNF mRNA levels, which showed significant repression by dexamethasone at all times. Although transcriptional repression is clearly important, actinomycin D chase experiments also showed dexamethasone to increase the loss of TNF mRNA. Thus, as observed for other inflammatory genes, transcriptional and post-transcriptional processes mediate the repressive effects of glucocorticoids on TNF mRNA (30,49). Furthermore, because MAPK activation is reduced by glucocorticoids in many cell types, including in A549 cells, the implication of MAPK pathways in these processes points to possible repressive mechanisms (27,29,50). However, with cotreatment, as used here, there is little repression of MAPK phosphorylation by dexamethasone at 30 min post-IL1B treatment (Fig. 11B) (24,28,35). Therefore, although inhibition of MAPKs reduces TNF transcription at this time (see Fig. 7B, right), this is not the mechanism for the observed repression of TNF mRNA at 30 min by dexamethasone. Indeed, one plausible explanation could be classical transrepression because the loss of IL1B-induced TNF mRNA occurs in the presence of the translational blocker, cycloheximide, and this is likely to rule out effects due to glucocorticoid-induced gene expression (51). Nevertheless, and irrespective of the mechanisms repressing TNF mRNA, dexamethasone induces DUSP1 expression, and this inhibits MAPK activation (Fig. 11B) (24). Given that ZFP36 expression is sensitive to MAPK inhibition, dexamethasone produces a considerable reduction in ZFP36 expression (Fig. 11B). This is clearly observed at 1 and 2 h following IL1B treatment and has been shown previously in A549 cells and in LPS-treated macrophages (15,16). Contrary to this, a number of groups report ZFP36 to be induced by glucocorticoids (16,31,32,52). Although this glucocorticoid-inducibility seems relatively minor in A549 cells, the response may contribute toward restoring or maintaining ZFP36 expression at longer times following IL1B plus dexamethasone treatment (Fig. 11B). Nevertheless, switching off MAPKs by the dexamethasone-dependent induction of DUSP1 and other mechanisms will reduce negative feed-forward control by ZFP36 (Fig. 11B) (and potentially other factors acting on TNF transcription and mRNA stability). Although this should promote TNF mRNA expression at longer times, this does not happen in the presence of glucocorticoid. Indeed, neither the silencing of DUSP1 or ZFP36 nor the silencing of both together affected the repressive effect of dexamethasone on TNF mRNA expression. Therefore, a requirement for additional, non-DUSP1/non-ZFP36-depen-dent effector mechanisms of glucocorticoid inhibition is indicated (26,30) (Fig. 11B).
Synthesis and release of TNF protein by A549 cells into the supernatants occurs rapidly following peak mRNA expression and lead to ϳ100 pg/ml (ϳ50 pg/10 6 cells) TNF. Such levels are within the range (30 -900 pg/10 6 cells) that are produced by mast cells, which are believed to be a physiologically relevant source of TNF (53)(54)(55)(56)(57). Interestingly, whereas in human airway epithelial cells, low nanomolar levels of TNF produce maximal responses, including cytokine release and activation of NF-B, it is clear that low picomolar levels (ϳ10 nM) are capable of eliciting responses, particularly in the context of co-stimulants, such as histamine (58). Certainly, TNF will be released in the context of multiple mediators in vivo. Furthermore, local concentrations of TNF secreted into a limited volume of fluid in the immediate vicinity of the epithelium may exceed those reached following release into a large volume of cell culture medium. Thus, it is plausible that even the low amounts of TNF produced by the primary HBE cells could be sufficient to elicit paracrine effects on nearby cells. However, because membranetethered TNF is biologically active, we also examined the expression of the uncleaved, ϳ25-kDa, TNF that is believed to exist in a membrane-tethered form (59 -61). Western blotting showed that expression of the 25-kDa TNF was rapidly induced by IL1B with a peak in expression at 2 h. This effect was common to both A549 and primary HBE cells, and in each case, peak expression was significantly and highly repressed by dexamethasone. Moreover, these levels of protein expression were similar in both cells and are consistent with the similar levels of mRNA expression observed by PCR. Furthermore, although expression of the 25-kDa TNF had largely disappeared by 6 h, in the A549 cells, we found little change in the level of TNF associated with the cell. Therefore, although an analysis of the functional relevance is clearly beyond the scope of the current study, this membrane-tethered and/or cell-associated TNF may be expected to play key roles in the rapid activation of resident or infiltrating cells.
In addition, while causing partial repression of TNF mRNA, dexamethasone produced a nearly complete loss of IL1B-induced TNF protein. This implies translational and/or posttranslational mechanisms that produce profound inhibition of TNF expression by glucocorticoids, as has been suggested previously (62). Indeed, similar effects are reported for a number of inflammatory genes (28,63). However, although TNF expression was strongly inhibited by MAPK inhibitors, and despite the considerable loss in mRNA expression occurring 6 h post-IL1B, silencing of DUSP1 had no effect on TNF release. This contrasts with ZFP36 silencing, which significantly increased TNF release and indicates a key role in the negative regulation of TNF translation. How this repression is achieved remains unclear, but it may involve ZFP36-dependent poly(A) tail loss, an effect that should decrease translation efficiency and mRNA stability (10,13), both effects that are implicated in the control of TNF biosynthesis (64). Furthermore, this may account for the lack of effect of the DUSP1 siRNAs. Although transiently increased activation of MAPKs resulting from DUSP1 silencing should promote MAPK-dependent TNF synthesis, the corresponding increases in ZFP36 expression could counterbalance this effect (Fig. 11A). Indeed, in the further presence of ZPF36 siRNA, there was increased TNF protein expression, suggesting that the scheme in Fig. 1 holds for TNF protein synthesis.
Turning to the repressive effect of dexamethasone, neither the ZFP36-nor the DUSP1-targeting siRNAs alone produced any change in the level of TNF released by IL1B plus dexamethasone. Thus, neither molecule appears to be an overriding determinant for glucocorticoid repression of TNF protein.
Indeed, in the presence of ZFP36-targeting siRNAs, the increased level of IL1B-induced TNF release produced by the ZFP36-targeting siRNAs meant that the percentage repression by dexamethasone actually increased (Fig. 10C). Similarly, combined knockdown of both ZFP36 and DUSP1 did not significantly alter the level of TNF release induced by IL1B plus dexamethasone or the percentage repression by dexamethasone (see Fig. 10E, right). This confirms that additional glucocorticoid-induced repressive mechanisms must account for the repression of TNF release (Fig. 11B).
Confirmation of these findings, with respect to the effects of IL1B and dexamethasone on DUSP1, ZFP36, and TNF expression in primary HBE cells provides considerable confidence that the regulatory mechanisms described for the A549 cells are physiologically relevant to primary human airway epithelial cells. In addition, recent data in other cell lines, primary human airway smooth muscle cells, and mouse models also collectively support this scheme ( Fig. 1) (65, 66). In particular, the interaction between feedback control by DUSP1 and feed-forward control by ZFP36 reveals why modulation of a single signaling component can lead to opposing effects on gene expression. Given an apparent loss of feed-forward control with respect to multiple inflammatory genes (24), these data illustrate the need to accurately model network behavior in order to predict biological outcomes. Finally, ZFP36 did not completely account for the effects of DUSP1 silencing, and MAPK inhibition resulted in enhanced transcriptional responses. Thus, there are likely to be a number of additional MAPK-dependent processes providing negative regulatory control of inflammatory gene expression. Such data add to the complexity of the system and confirm the need for detailed modeling to assess therapeutic strategies (41). With respect to the anti-inflammatory effects of glucocorticoids, these data suggest that neither DUSP1 nor ZFP36 have dominant repressive effects, at least on TNF expression. Importantly, the data clearly show that additional mechanisms of repression by glucocorticoids must also exist in order to maintain repression, particularly at longer times, where negative feed-forward control may be down-regulated by the early effects of glucocorticoid treatment.