Gain-of-Function Mutation of Tristetraprolin Impairs Negative Feedback Control of Macrophages In Vitro yet Has Overwhelmingly Anti-Inflammatory Consequences In Vivo

ABSTRACT The mRNA-destabilizing factor tristetraprolin (TTP) binds in a sequence-specific manner to the 3′ untranslated regions of many proinflammatory mRNAs and recruits complexes of nucleases to promote rapid mRNA turnover. Mice lacking TTP develop a severe, spontaneous inflammatory syndrome characterized by the overexpression of tumor necrosis factor and other inflammatory mediators. However, TTP also employs the same mechanism to inhibit the expression of the potent anti-inflammatory cytokine interleukin 10 (IL-10). Perturbation of TTP function may therefore have mixed effects on inflammatory responses, either increasing or decreasing the expression of proinflammatory factors via direct or indirect mechanisms. We recently described a knock-in mouse strain in which the substitution of 2 amino acids of the endogenous TTP protein renders it constitutively active as an mRNA-destabilizing factor. Here we investigate the impact on the IL-10-mediated anti-inflammatory response. It is shown that the gain-of-function mutation of TTP impairs IL-10-mediated negative feedback control of macrophage function in vitro. However, the in vivo effects of TTP mutation are uniformly anti-inflammatory despite the decreased expression of IL-10.

One important feedback mechanism involves the secretion of the anti-inflammatory cytokine interleukin 10 (IL-10). Particularly in tissues like those of the gut, which are continuously exposed to PAMPs, IL-10 is essential for restraining inflammatory responses to commensal microorganisms (8,9). One critical function of IL-10 is the suppression of the Il12b gene, which codes for the shared p40 subunit of the cytokines IL-12 and IL-23. The constraint of IL-12 and IL-23 expression is required to prevent excessive Th1-and Th17-mediated intestinal inflammation (10)(11)(12)(13). Macrophagespecific knockout of the IL-10 receptor (IL-10R) results in severe spontaneous gut inflammation (14,15). Similarly, polymorphisms in the human IL-10 -IL-10R axis are associated with susceptibility to inflammatory bowel disease (16). The anti-inflammatory actions of IL-10 are mediated by the transcription factor STAT3 (signal transducer and activator of transcription 3), which is phosphorylated and activated by the tyrosine kinases JAK1 and TYK2 upon the engagement of the IL-10 receptor. A number of STAT3-regulated target genes have been implicated as mediators of anti-inflammatory functions of IL-10 (6,(17)(18)(19). The other side of the coin is that increased expression of IL-10 may contribute to pathogenesis, for example, by impairing antitumor immunity (20). Several viruses express IL-10 orthologues, apparently as a means of evasion of the innate immune system (21). The expression of endogenous IL-10 is tightly regulated by different mechanisms in several cell types that can express this anti-inflammatory cytokine (22,23).
Much is known about how transcription factors regulate the complex programs of gene expression in activated macrophages (3,(24)(25)(26)(27). It is often overlooked that the dynamic regulation of mRNA stability also plays an important role (28,29). Many inflammatory mediator mRNAs have adenosine/uridine-rich elements (AREs) in their 3= untranslated regions, which act as binding sites for various RNA-destabilizing factors, the best characterized of which is tristetraprolin (TTP) (30,31). TTP promotes the rapid turnover of its substrate transcripts by recruiting a number of nuclease complexes, notably the CCR4/NOT complex, which mediates poly(A) tail shortening, the ratelimiting step in the degradation of most mRNA species. In resting macrophages, TTP is expressed at very low levels. Its upregulation in response to proinflammatory stimuli constitutes a negative feedback mechanism to limit the duration and strength of inflammatory gene expression (30)(31)(32). Myeloid cell-specific disruption of the TTP gene (formally known as Zfp36) causes extreme sensitivity to lethal endotoxic shock, although there is no evident phenotype in the absence of inflammatory challenge (33,34). In contrast, global disruption of the Zfp36 gene causes a spontaneous and severe inflammatory syndrome with some features of rheumatoid arthritis (35). Tnf was the first mRNA shown to be posttranscriptionally regulated by TTP, and dysregulation of tumor necrosis factor (TNF) expression plays an important role in both myeloid cellspecific and global Zfp36 knockout phenotypes (33)(34)(35)(36). Nevertheless, it is clear that very many additional targets exist and are dysregulated in cells lacking TTP (30,37,38). Puzzlingly, IL-10 was identified as a target of posttranscriptional regulation by TTP, which was overexpressed by Zfp36 Ϫ/Ϫ macrophages (38)(39)(40)(41). Enhanced IL-10-mediated negative feedback in Zfp36 Ϫ/Ϫ cells resulted in the paradoxical underexpression of IL-6, IL-12, and IL-23 (39,41).
The mitogen-activated protein kinase (MAPK) p38 signaling pathway controls the expression of inflammatory mediators via the phosphorylation and inactivation of TTP (42,43). Hence, MAPK p38 inhibitors decrease the expression levels of several proinflammatory mediators and destabilize the corresponding mRNAs in Zfp36 ϩ/ϩ but not Zfp36 Ϫ/Ϫ macrophages (41,44). The critical phosphorylations, at serine 52 and serine 178 of murine TTP, are carried out by MAPK-activated protein kinase 2 (MK2), which is downstream of the MAPK p38 itself (43,(45)(46)(47). They result in the recruitment of 14-3-3 proteins, the disruption of the interaction between TTP and the CCR4/NOT complex, and the consequent protection of the target transcript's poly(A) tail (46,(48)(49)(50). The coupling between MAPK p38 signaling and TTP activity contributes to the precise timing of gene expression in activated macrophages (33,38). Prolonged MAPK p38 activation in macrophages lacking the MAPK phosphatase dual-specificity phosphatase 1 (DUSP1) promotes the overexpression of several inflammatory mediators in vitro and in vivo, causing exaggerated responses to experimental challenges such as LPS injection (5,51). To a large extent, this dysregulated inflammatory response is caused by the increased phosphorylation and inactivation of TTP, the increased stability of proinflammatory mRNAs, and a delay to the off phase of gene expression (52).
We recently used homologous recombination to generate a knock-in mouse strain in which Ser52 and Ser178 of the endogenous TTP protein are replaced with nonphosphorylatable alanine residues (53). The mutated Zfp36 locus is known as Zfp36 aa , and its protein product is known as TTP aa . Since it cannot be inactivated by the MAPK p38 pathway, TTP aa is a constitutive, dominant mRNA-destabilizing protein. Its expression protects mice from harmful consequences of systemic LPS injection (53), experimental inflammatory arthritis (54), and inflammatory lung pathologies (43). Differential gene expression was assessed by microarray analysis of Zfp36 aa/aa macrophages, which express the alanine-substituted mutant form of TTP from the endogenous Zfp36 locus (52,53). Several known TTP target mRNAs, such as F3, Tnf, Cxcl1, Cxcl2, Ier3, and Il10, were underexpressed by Zfp36 aa/aa macrophages, and some putative novel TTP targets were identified.
We and others have discussed the possibility of therapeutic targeting of TTP in order to exert anti-inflammatory effects (31,43). A proof of principle was generated by the treatment of experimental arthritis using two compounds that promote PP2A-mediated dephosphorylation and activation of TTP (54). A significant concern is that altering TTP function could give rise to unwanted proinflammatory effects by disrupting endogenous negative feedback mechanisms, for example, those involving IL-10 (37,(39)(40)(41).
Here we investigated in vitro and in vivo consequences of the impaired expression of IL-10 caused by a gain-of-function mutation of TTP.

RESULTS
Anomalous expression of certain TTP targets by Zfp36 aa/aa macrophages. Microarray analysis was performed to identify transcripts differentially expressed by Zfp36 aa/aa macrophage colony-stimulating factor (M-CSF)-differentiated bone marrowderived macrophages (M-BMMs) after stimulation with LPS. The microarray data were deposited at the Gene Expression Omnibus, and some aspects of their analysis were previously reported (52,53). Figure 1 illustrates the analysis of LPS-induced transcripts (Ͼ2-fold increase of expression in response to LPS; P Ͻ 0.05). The data are presented in the form of volcano plots, in which transcripts with a Ͼ1.5-fold difference of expression between Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs and a corrected P value of Ͻ0.05 are shaded black. Differential expression data are summarized in Table S1 in the supplemental material. As expected for a dominant negative regulator of gene expression, many transcripts were underexpressed by Zfp36 aa/aa M-BMMs. Several of the changes of expression at 1 h were previously reported, including the underexpression of Tnf, Il10, and Cxcl1 mRNAs (53). By 4 h after the addition of LPS, the levels of both Tnf and Cxcl1 mRNAs had almost returned to basal levels, and differences between Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs were no longer statistically significant (also see Fig.  4A). Il10 mRNA was significantly underexpressed by Zfp36 aa/aa M-BMMs at both the 1-h and 4-h time points. Contrary to expectation, Il6, which was previously characterized as a TTP target (38,41,55,56), was not significantly underexpressed by Zfp36 aa/aa M-BMMs (Fig. 1). Several other transcripts were significantly overexpressed by Zfp36 aa/aa M-BMMs. One example, Il12b, is highlighted in Fig. 1.
Gene ontology (GO) analysis was performed on transcripts differentially expressed by Zfp36 aa/aa M-BMMs (see Table S2 in the supplemental material). At both 1 h and 4 h, underexpressed transcripts were highly enriched in several GO terms relating to response to stimulus, defense response, or inflammatory response. The enrichment of these terms was a consequence of the differential expression levels of transcripts including Tnf, Cxcl1, Cxcl2, Clcf1, Ier3, S100a8, Zfp36, Mefv, Ltf, and Il10, most of which were previously described as TTP targets (30). These results are arguably trivial, given that LPS-induced transcripts were selected prior to the analysis of differential gene expression. GO analysis of the overexpressed transcripts was more informative. At 1 h, several GO terms related to cell migration were very significantly enriched (P Ͻ 10 Ϫ10 ). All of these enrichments arose from the increased expression levels of four transcripts, namely, Cited2, Cxcl10, Sdc4, and Trib1. At 4 h, the most significantly enriched GO terms were related to the phosphorylation of the transcription factor STAT5 (signal transducer and activator of transcription 5) or the regulation of the extrinsic apoptosis pathway (Table S2). The corresponding lists of differentially expressed transcripts were dominated by Il12b, Csf2, Cx3cl1, Kdr, and Rasip1. These findings suggest that the genes overexpressed by Zfp36 aa/aa M-BMMs are functionally distinct.
TTP can negatively regulate gene expression by inhibiting translation rather than destabilizing mRNA (37,57,58). The LPS-induced expression levels of the TNF, IL-6, IL-10, and IL-12p40 proteins between Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs were therefore compared. Multiple experiments were performed by using different numbers and densities of M-BMMs, depending on whether the primary aim was to harvest RNA for microarrays, RNA for quantitative PCR (qPCR), secreted protein for enzyme-linked immunosorbent assays (ELISAs) and multiplex assays, or intracellular protein for Western blotting. Regardless of the experimental conditions, the expression levels of both the TNF and IL-10 proteins were consistently diminished in Zfp36 aa/aa M-BMMs (by approximately 5-fold and 3-fold, respectively) ( Fig. 2A and B). In contrast, the pattern of expression of IL-6 was inconsistent, with unchanged expression by Zfp36 aa/aa M-BMMs in most cases but both overexpression and underexpression in some experiments. Overall, there was no statistically significant difference in the expression levels of IL-6 between Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs. (Fig. 2C). The expression level of the IL-12p40 protein was invariably higher in Zfp36 aa/aa than in matched Zfp36 ϩ/ϩ M-BMMs, although the extent of overexpression was quite variable (Fig. 2D).
Anomalous expression of certain TTP targets by Dusp1 ؊/؊ macrophages. In Dusp1 Ϫ/Ϫ macrophages, enhanced and prolonged MAPK p38 activation in response to LPS promotes the overexpression of several cytokines and chemokines via increased phosphorylation and inactivation of TTP (52). The defining characteristic of genes that are regulated by the DUSP1-TTP axis is that the overexpression caused by the disruption of the Dusp1 gene can be prevented by combining this with the Zfp36 aa/aa genotype. Dysregulated MAPK p38 signaling fails to promote the overexpression of target genes if TTP cannot be phosphorylated and inactivated. At the protein and mRNA levels, the expression levels of both TNF and the anti-inflammatory cytokine IL-10 conformed to the pattern described above, being elevated in Dusp1 Ϫ/Ϫ but not Dusp1 Ϫ/Ϫ -Zfp36 aa/aa M-BMMs (Fig. 3). Il6 mRNA levels were decreased in M-BMMs of all three genetically modified genotypes, but there were no significant differences in protein expression levels ( Fig. 3). At 4 h, the expression level of Il12b mRNA was actually decreased rather than increased in Dusp1 Ϫ/Ϫ M-BMMs and was increased rather than decreased in Zfp36 aa/aa M-BMMs, whereas in Dusp1 Ϫ/Ϫ -Zfp36 aa/aa M-BMMs, the expression level of Il12b mRNA was not significantly different from that in wild-type controls (Fig. 3). The pattern of expression of the IL-12p40 protein was similar.
Targeted mutation of TTP paradoxically increases transcription of the Il6 and Il12b genes. Tnf, Il6, Il10, and Il12b mRNAs can all be physically recognized by TTP (38) yet show very different responses to mutations that influence TTP phosphorylation and activity. To make sense of these disparate patterns, we monitored gene expression at the levels of steady-state mRNA, primary transcripts (as a surrogate for the transcription rate), mRNA stability, and protein secretion in LPS-treated Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs (Fig. 4).
The abundances of both Tnf and Il10 steady-state mRNAs were decreased in Zfp36 aa/aa M-BMMs (Fig. 4A). The expression levels of the corresponding proteins were diminished (Fig. 4D), as previously shown (53) (Fig. 2). There were no significant differences in Tnf or Il10 primary transcript levels (Fig. 4B), but the stabilities of Tnf and Il10 mRNAs were decreased in Zfp36 aa/aa M-BMMs (Fig. 4C), as previously reported (53). The behavior of both the Tnf and Il10 genes is therefore consistent with direct regulation by TTP at the level of mRNA stability. Since the Zfp36 gene mutation had a greater effect on the TNF and IL-10 proteins than on Tnf and Il10 mRNAs, the phosphorylation of TTP may also have some impact at the translational level, as was previously suggested (37, 57, 58). Only a minor difference in the steady-state Il6 mRNA abundance was detected at 4 h (Fig. 4A), and IL-6 secretion did not differ between Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs (Fig. 4D). There appeared to be an almost perfect balance between the increase in the rate of Il6 transcription (Fig. 4B) and the increase in the rate of Il6 mRNA degradation (Fig. 4C). Il12b mRNA was highly stable in both Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs (Fig.  4C). The increased expression levels of Il12b mRNA (Fig. 4A) and IL-12p40 protein (Fig.  4D) by Zfp36 aa/aa M-BMMs appeared to be driven by an enhanced transcriptional response to LPS (Fig. 4B).
Disruption of IL-10-mediated negative feedback contributes to the anomalous expression of IL-6 and IL-12p40 by Zfp36 aa/aa macrophages. IL-10 is a wellcharacterized target of TTP (38)(39)(40)(41), which is underexpressed by Zfp36 aa/aa M-BMMs (Fig. 5) and known to negatively regulate the expression of TNF, IL-12p40, and IL-6 (7,19). Noticeably, the expression of IL-12p40 in macrophages of differing genetic backgrounds was an almost perfect mirror image of IL-10 expression (Fig. 3). It was therefore speculated that a disruption of IL-10-mediated autocrine negative feedback mechanisms contributed to differences in IL-6 and IL-12p40 expression in vitro.
It is not known whether proinflammatory genes display the same dose response to inhibition by IL-10. Differential sensitivity to IL-10 could influence the outcomes of dysregulated IL-10 biosynthesis in macrophages. Therefore, Il10 Ϫ/Ϫ M-BMMs were used , and the levels of the corresponding proteins were measured by an ELISA or a multiplex bead assay (B). Graphs represent means Ϯ standard errors of the means of data from 3 independent M-BMM cultures. n.s., not statistically significant; *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.005 (by one-way ANOVA). The microarray experiment used to generate the data in panel A was previously described (52).
to investigate the relative sensitivity of TNF, IL-12p40, and IL-6 expression to exogenous IL-10 in the absence of interference from endogenous IL-10 (Fig. 5). The three proinflammatory cytokines displayed quite distinct dose responses. IL-12p40 was the most sensitive in terms of both the 50% effective concentration (EC 50 ) (estimated to be 0.15 ng/ml) and maximal inhibition (93%). TNF was inhibited by a maximum of 76% and with an EC 50 of approximately 0.29 ng/ml. IL-6 biosynthesis was insensitive to low  concentrations of IL-10. Maximal inhibition was 81%, and the estimated EC 50 was 1.52 ng/ml. Differentiation of BMMs in the presence of granulocyte-macrophage colonystimulating factor (GM-CSF) rather than M-CSF has been reported to generate a cell population (GM-BMMs) that expresses relatively little IL-10 but larger amounts of TNF, IL-6, and IL-12p40 in response to LPS (59,60). It was confirmed that wild-type GM-BMMs expressed 5-fold less IL-10 than did M-BMMs generated in parallel from the same donor mice (Fig. 6). The expression levels of the proinflammatory cytokines TNF, IL-6, and IL-12p40 were elevated 2.5-fold, 5.4-fold, and 56.3-fold, respectively, in GM-BMMs compared to M-BMMs generated in parallel. Zfp36 ϩ/ϩ and Zfp36 aa/aa GM-BMMs did not differ in their expression of IL-10 after stimulation with LPS. Therefore, in GM-BMMs, the targeted mutation of TTP cannot indirectly influence gene expression by perturbing IL-10-mediated autocrine/paracrine feedback mechanisms. Targeted mutation of the Zfp36 locus decreased the expression of TNF in both M-BMMs and GM-BMMs. The Zfp36 mutation did not affect IL-6 expression in M-BMMs, whereas in GM-BMMs, the mutation significantly decreased the expression of IL-6. The expression of IL-12p40 was elevated as a consequence of the Zfp36 mutation in M-BMMs. In contrast, in GM-BMMs, the Zfp36 mutation had no significant effect on IL-12p40 expression.
The observations described above support the hypothesis that changes in the expression of IL-10 influence the phenotype caused by the substitution of two TTP phosphorylation sites. To test this hypothesis more rigorously, Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs were stimulated with LPS in the presence of an IL-10-neutralizing antibody or an isotype control (Fig. 7). In the presence of the neutralizing antibody, IL-6 was underexpressed by Zfp36 aa/aa M-BMMs. IL-10 neutralization increased the expression of IL-12p40 in Zfp36 ϩ/ϩ M-BMMs but had relatively little effect on Zfp36 aa/aa M-BMMs. As a consequence, the degree of overexpression of IL-12p40 by Zfp36 aa/aa M-BMMs was diminished. Together, these data indicate that the aberrant patterns of expression of IL-6 (unaltered in Zfp36 aa/aa M-BMMs) and IL-12p40 (elevated in Zfp36 aa/aa M-BMMs) are at least partly due to alterations in IL-10-mediated negative feedback. The expression of TNF appears minimally sensitive to this phenomenon.
Targeted mutation of TTP has overwhelmingly anti-inflammatory consequences in vivo. Next, Zfp36 ϩ/ϩ and Zfp36 aa/aa mice were injected intraperitoneally (i.p.) with LPS, and serum cytokine and chemokine levels were measured after 3 and 12 h (Fig. 8). Decreased expression levels of TNF and IL-10 in Zfp36 aa/aa mice were previously reported (53) (indicated by † in Fig. 8). CCL3, CCL4, CCL11, CSF2, gamma interferon (IFN-␥), IL-1␣, IL-5, and IL-13 were also significantly underexpressed by Zfp36 aa/aa mice. Most notably, the expression level of IL-6 was decreased by almost 4-fold at 3 h and by more than 200-fold at 12 h in Zfp36 aa/aa mice. The elevated expression of IL-12p40 in LPS-treated Zfp36 aa/aa M-BMMs was not recapitulated in vivo. Serum concentrations of IL-12p40 were not significantly different between LPS-treated Zfp36 ϩ/ϩ and Zfp36 aa/aa mice at 3 h but were significantly lower in Zfp36 aa/aa mice at 12 h.
Finally, to test whether cytokine expression in vivo was regulated by DUSP1 via alterations in TTP phosphorylation, wild-type, Dusp1 Ϫ/Ϫ , Zfp36 aa/aa , and Dusp1 Ϫ/Ϫ -Zfp36aa/aa mice were injected intraperitoneally with LPS and culled after 3 h, and the expression levels of Tnf, Il10, Il6, and IL12b mRNAs were measured in spleens (Fig. 9). Levels of Tnf and Il10 mRNAs were elevated in spleens of LPS-treated Dusp1 Ϫ/Ϫ mice and decreased in those of Zfp36 aa/aa mice. When Dusp1 Ϫ/Ϫ and Zfp36 aa/aa genetic modifications were combined, the Zfp36 aa/aa phenotype was dominant, and the expression levels of both mRNAs remained low. These patterns of gene expression are very similar to those previously described in vitro (Fig. 3A) and in serum of LPS-treated mice (52), consistent with DUSP1 regulating gene expression via modulation of the phosphorylation state of TTP in both contexts. Strikingly, the pattern of expression of Il6 mRNA was the same: it was increased in Dusp1 Ϫ/Ϫ spleens but decreased in both Zfp36 aa/aa and Dusp1 Ϫ/Ϫ -Zfp36 aa/aa spleens, indicating a similar mechanism of regulation. The level of the IL-6 protein was also very strongly elevated in sera of LPS-treated Dusp1 Ϫ/Ϫ mice but decreased in sera of both Zfp36 aa/aa and Dusp1 Ϫ/Ϫ -Zfp36 aa/aa mice (data not shown). Differences in the expression levels of Il12b mRNA among the four genotypes of mice were not significant. The dysregulation of Il12b gene expression in isolated Zfp36 aa/aa M-BMMs was therefore not recapitulated in vivo in Zfp36 aa/aa mice (Fig. 8).

DISCUSSION
The phosphorylation and dephosphorylation of TTP constitute a molecular switch for turning gene expression on and off during responses to inflammatory stimuli (43). In Dusp1 Ϫ/Ϫ cells, the dysregulation of MAPK p38 signaling pushes the switch in the FIG 8 In vivo responses to LPS are broadly impaired in Zfp36 aa/aa mice. Zfp36 ϩ/ϩ and Zfp36 aa/aa mice were injected intraperitoneally with 5 mg/kg LPS and humanely sacrificed after 3 or 12 h. Serum cytokine levels were measured by a multiplex bead assay or an ELISA. Graphs represent means Ϯ standard errors of the means for 5 untreated mice and 10 LPS-injected mice at each time point. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.005 (by a Mann-Whitney test). The IL-10 and TNF measurements (indicated by †) were previously reported (53).
"on" direction, promoting the phosphorylation and inactivation of TTP and enhancing and sustaining the expression of TTP target genes by increasing mRNA stability. In Zfp36 aa/aa cells, the switch is locked in the "off" position, and the expression of TTP target genes is diminished as a consequence of enhanced mRNA decay. This locked switch prevents the overexpression of TTP target genes in cells in which the two genetic modifications are combined (43,52). Here and in previous reports (52,53), the expression of both TNF and IL-10 conformed exactly to the pattern described above, whether investigated in vitro or in vivo. The behavior of these genes can be understood simply in terms of our working model of TTP function. However, it remains puzzling at first sight that the same molecular mechanism is used to regulate the expressions of a powerful proinflammatory cytokine and a powerful anti-inflammatory cytokine. It is likely that this shared regulation contributes to the resolution of inflammation. Any stimulus that provokes the strong expression of TNF via the activation of the MAPK p38 and the phosphorylation of TTP will also provoke the strong expression of IL-10, with a built-in delay owing to the slower transcriptional activation of the latter. IL-10 then acts as an autocrine or paracrine negative feedback regulator of the inflammatory response.
When gene expression was initially studied in Zfp36 aa/aa M-BMMs, we were puzzled that IL-6 did not behave as expected. IL-6 has been well characterized as a target of TTP (30,55,56), yet its expression was not consistently diminished in Zfp36 aa/aa M-BMMs and in many experiments was actually increased (Fig. 2). Microarray analysis also revealed the surprising overexpression of several transcripts by Zfp36 aa/aa M-BMMs (Fig.  1), among which Il12b was prominent. We hypothesized that an impairment of IL-10mediated negative feedback control contributed to these anomalies. Being a secondary response gene, the peak of expression of Il6 occurred at a time when the endogenous IL-10 protein had already accumulated in tissue culture supernatants (Fig. 4). IL-6 expression was sensitive to inhibition by IL-10, with a very steep dose response at FIG 9 Regulation of cytokine expression by the DUSP1-TTP axis in vivo. Dusp1 ϩ/ϩ -Zfp36 ϩ/ϩ , Dusp1 Ϫ/Ϫ -Zfp36 ϩ/ϩ , Dusp1 ϩ/ϩ -Zfp36 aa/aa , or Dusp1 Ϫ/Ϫ -Zfp36 aa/aa mice were injected intraperitoneally with 5 mg/kg LPS and humanely sacrificed after 3 h. Spleens were excised, RNA was prepared, and the levels of the indicated transcripts were quantified by qPCR, with normalization against B2m mRNA levels and then against the untreated wild-type (Dusp1 ϩ/ϩ -Zfp36 ϩ/ϩ ) control. Graphs represent means Ϯ standard errors of the means for 2 untreated and 4 LPS-treated mice of each genotype. n.s., not statistically significant; **, P Ͻ 0.01; ***, P Ͻ 0.005 (by ANOVA).
concentrations of between 1 and 3 ng/ml (Fig. 5), similar to the range of IL-10 concentrations produced by Zfp36 ϩ/ϩ and Zfp36 aa/aa M-BMMs (Fig. 4). As predicted, IL-6 was underexpressed by Zfp36 aa/aa GM-BMMs (Fig. 6) and by Zfp36 aa/aa peritoneal macrophages (data not shown), both of which produce very low levels of IL-10. Neutralization of IL-10 also restored the expected gene expression pattern, in which IL-6 was underexpressed by Zfp36 aa/aa M-BMMs (Fig. 7). It therefore appears that differences in the levels of IL-10 in tissue culture have a profound effect on the expression of the Il6 gene, resulting in variable under-or overexpression by Zfp36 aa/aa M-BMMs. Although Tnf gene expression is also negatively regulated by IL-10 ( Fig. 5) (19,61), its immediate early pattern of expression (61,62) (Fig. 4) presumably allows it to escape the influence of such fluctuations in the efficacy of negative feedback control. By the time that levels of endogenous IL-10 accumulate significantly, the transcriptional activity of the Tnf gene is already declining.
The production of the IL-12p40 protein was profoundly sensitive to IL-10-mediated inhibition (Fig. 5). The paradoxical overexpression of IL-12p40 by Zfp36 aa/aa M-BMMs was caused by an increase in transcription (Fig. 4). Like others (63,64), we found no evidence that TTP regulated Il12b mRNA stability (Fig. 4). This increase of expression was not observed in Zfp36 aa/aa GM-BMMs, which express little IL-10 ( Fig. 6). IL-10 has been reported to inhibit Il12b transcription via the induction of the transcriptional repressor NFIL3 (nuclear factor, IL-3 regulated) (13). Four hours after the addition of LPS, the expression level of Nfil3 mRNA was significantly lower in Zfp36 aa/aa than in Zfp36 ϩ/ϩ M-BMMs (2,135 Ϯ 35 versus 4,940 Ϯ 497 robust microarray analysis units [RMA]; P Ͻ 0.005). It is therefore possible that impaired negative feedback via NFIL3 contributes to the enhanced transcriptional induction of the Il12b gene in Zfp36 aa/aa M-BMMs (Fig. 4). However, neutralization of IL-10 only partially reversed the overexpression of IL-12p40 by Zfp36 aa/aa M-BMMs, suggesting that additional mechanisms may be involved. TTP has been reported to inhibit Il12b gene expression by modulating the function of the transcription factor NF-B (nuclear factor kappa light chain enhancer of activated B cells) (65). However, we were unable to detect any differences in NF-B function in Zfp36 aa/aa M-BMMs (53). Additionally, it is difficult to explain the increases in the expression levels of only a few genes by the increased activity of a transcription factor that regulates much of the macrophage response to LPS. Therefore, the aberrant expression of IL-12p40 by Zfp36 aa/aa M-BMMs has been only partially accounted for.
Previous reports identified IL-10 as a target of TTP (39,41), describing the increased expression of IL-10 and the consequent underexpression of both IL-6 and IL-12 by Zfp36 Ϫ/Ϫ macrophages. The phenomenon described here is a mirror image, in which a gain of function of TTP impairs IL-10-mediated negative feedback control in vitro and promotes the increased expressions of these two cytokines. As we and others are interested in the concept of therapeutic targeting of TTP (31,43,53), the impairment of negative feedback control was a source of concern, prompting us to test the in vivo effects of the gain-of-function mutation. Despite the fact that LPS-induced IL-10 expression was diminished in Zfp36 aa/aa mice, we detected no corresponding increase in the expression of either IL-6 or IL-12p40 ( Fig. 8 and 9). In fact, IL-6 was consistently expressed at very low levels by Zfp36 aa/aa mice following systemic LPS challenge (Fig.  8), the injection of zymosan into dorsal air pouches, or the induction of experimental arthritis (53). Increased circulating levels of either IL-12 or IL-23, which share the common IL-12p40 subunit, would be expected to promote Th1 and Th17 responses, characterized by elevated expression levels of IFN-␥ and IL-17. In fact, both of these effector cytokines were present at lower levels in sera of LPS-injected Zfp36 aa/aa mice (49) (Fig. 8).These findings suggest that agonists that activate TTP may exert antiinflammatory effects without incurring immune activation via the increased expression of IL-6, IL-12, or IL-23. It will be important to test this further in experimental settings such as inflammatory bowel disease, where impaired expression of IL-10 may have profound consequences (8,12).
A remaining question is why the behaviors of the Il6 and Il12b genes are so different when studied in vitro and in vivo. Although in vitro experiments are invaluable for gaining insight into molecular mechanisms of gene regulation, it is easy to forget how poorly they model inflammatory responses in vivo. In many senses, the in vitro assay system is artificial. The population of cells is homogeneous, their exposure to stimulus is synchronous, and they remain thereafter in a static milieu, in which local concentrations of cytokines and other effectors may become high and autocrine or paracrine effects may be amplified. In vivo, the population of cells responding to LPS is heterogeneous, their time of exposure to the agonist will vary somewhat, and the local accumulation of cytokines and other immune effectors will be limited by circulation. Whatever the explanation for the divergent behaviors of IL-6 and IL-12p40 in vitro and in vivo, our principal conclusion remains the same. A gain of function of TTP causes the dysregulation of IL-6 and IL-12p40, at least in part indirectly, via changes in IL-10 expression. However, aberrant overexpression of IL-6 and IL-12p40 does not occur in Zfp36 aa/aa mice.

MATERIALS AND METHODS
Materials. LPS (Escherichia coli serotype EH100) was purchased from Enzo Life Sciences. Other biochemicals were purchased from Sigma-Aldrich unless otherwise stated. All media and sera were routinely tested for endotoxin by using the Limulus amebocyte lysate test (Lonza) and were rejected if the endotoxin concentration exceeded 0.1 U/ml.
Generation of mouse strains. The generation of the Zfp36 aa/aa mouse strain in the C57BL/6 background was described previously (49). The Dusp1 Ϫ/Ϫ strain was a generous gift from Bristol-Myers Squibb and was backcrossed to the C57BL/6 strain for 10 generations prior to performing the experiments described here. The double-targeted Dusp1 Ϫ/Ϫ -Zfp36 aa/aa line was generated by three generations of crossing, with genotyping being performed by PCR of genomic DNA, and thereafter was maintained as a pure-breeding line. Bone marrow from Il10 Ϫ/Ϫ mice was generously provided by Fiona Powrie (Oxford).
In vivo experiments and cell isolation. All animal experiments were approved by local ethical committees and performed under UK Home Office project licenses. C57BL/6 mice were purchased from Harlan Laboratories. All mice used were between 6 and 12 weeks of age. To assess the systemic response to LPS, mice were injected i.p. with 5 mg/kg of body weight of purified LPS in 200 l sterile phosphatebuffered saline (PBS). Mice were humanely culled 3 h or 12 h after challenge, and peripheral blood was collected by cardiac puncture for serum isolation. Spleens were excised and snap-frozen in liquid nitrogen for later isolation of RNA.
Bone marrow was isolated from humanely culled mice, and BMMs (bone marrow-derived macrophages) were obtained by differentiation in vitro with either 100 ng/ml M-CSF (Peprotech), or 50 ng/ml GM-CSF in RPMI 1640 containing 10% heat-inactivated fetal calf serum (FCS) and penicillin-streptomycin for 7 days. BMMs were plated at a density of 1 ϫ 10 6 macrophages/ml in the appropriate cell culture plate at least 1 day prior to stimulation.
Neutralization of endogenous secreted IL-10 was achieved by in vitro incubation with an IL-10neutralizing antibody (catalogue number 501407; BioLegend) at 10 g/ml for the duration of the assays. Controls were treated with an equal concentration of isotype-matched antibody.
Measurements of mRNA levels. RNA was extracted from BMMs by using QIAshredder columns and an RNeasy minikit (Qiagen). cDNA was generated by using the iScript cDNA synthesis kit (Bio-Rad). Gene expression was quantified by reverse transcription-quantitative PCR (qRT-PCR) on a LightCycler 480 II instrument (Roche) using the Superscript III platinum RT-PCR kit and custom-synthesized oligonucleotide primers (Eurofins MWG) with SYBR Premix Ex Taq (Lonza). The relative gene expression level was calculated by using the ΔΔC T method with Gapdh or B2m mRNA for the normalization of RNA levels. Primary transcript PCR was performed by using primer pairs that crossed exon-intron boundaries, with an additional DNase I step to remove contaminating genomic DNA from RNA samples (Qiagen). Control PCRs were carried out in the absence of RT in order to monitor genomic DNA contamination, which in all cases was negligible. Sequences of oligonucleotides designed to detect primary or mature transcripts are available from the authors upon request.
The half-life of mRNA was estimated by using the transcription inhibitor actinomycin D (Sigma) at 5 g/ml, followed by measurement of mRNA levels by quantitative real-time PCR.
Microarray analyses were performed by using SurePrint G3 Mouse GE 8x60K slides (Agilent) and Partek Genomics Suite version 6.6, build 6.13.0315 (Partek), as previously described (49). For the generation of volcano plots, transcripts were first filtered for significant upregulation in response to LPS (Ͼ2-fold increase; adjusted P value of Ͻ0.05), and weakly expressed transcripts were removed by the application of an arbitrary filter of 200 RMA. Plots (log 2 fold difference of expression versus Ϫlog 10 analysis of variance [ANOVA] P value) were constructed by using ggplot2s in the R statistical package, with subset cutoffs at a P value of Ͻ0.05 and a fold difference of expression of Ͼ1.5.
Assessment of protein expression. Secreted factors in tissue culture supernatants and sera were quantified by an ELISA according to the manufacturer's instructions (eBioscience) or by using Bio-Plex bead capture assays and a Bio-Plex 200 analyzer (Bio-Rad).