DUSP Meet Immunology: Dual Specificity MAPK Phosphatases in Control of the Inflammatory Response

Down-regulation of MAPK signaling: dual specificity phosphatases (DUSP) and other MAPK phosphatases (MKP)

Inactivation of MAPK occurs primarily through dephosphorylation of the TxY motif. Because MAPK have to be phosphorylated on both threonine and tyrosine residues for kinase activity, inactivation can be brought about by members of different phosphatase families. MKP include serine-threonine phosphatases (PP2A and PP2C), the tyrosine phosphatases PTPN5, PTPN7, and PTPRR (14, 15), and members of the DUSP family that is the focus of this study. (A confusing number of synonyma exists for some of the DUSP genes or proteins. We advocate the use of the official gene symbol that is also the abbreviation of the protein. In Table I⇓ and on first appearance in the text, the most common synonyms are also given.) All DUSP genes encode a catalytic domain harboring a signature HCxxxxxR motif. Of the 30 protein-coding DUSP genes found in the human genome (16), 11 are bona fide MKP because in addition to the DUSP domain they contain a MAPK binding domain (MKB). The 19 atypical DUSP are much smaller proteins that lack such a MKB, but may still function as MKP as already demonstrated for DUSP14/MKP6 (17), DUSP3/VHR (18), and DUSP22/VHX (19). Although more atypical DUSP may turn out to deactivate MAPK, most seem to have unrelated substrates including RNA in the case of DUSP11/PIR (16).

The biochemistry and structural biology of DUSP proteins have been recently reviewed (15, 20) and are therefore touched on in this study only briefly. The presence or absence of MKB, DUSP, and additional N- or C-terminal domains provides a reasonable basis for grouping DUSP-MKP into four different types (Table I⇑) (15). The subcellular localization and preference for one or the other MAPK as a substrate have also been used to subdivide the DUSP family (14, 20); however, these are not always well-defined and may depend on cell-type, activation state, and the presence of interacting proteins.

The 3D structure of the catalytic DUSP domain has been solved for DUSP6/MKP-3, DUSP2/PAC-1, and DUSP3/VHR, revealing that for the atypical DUSP3 it is in a catalytically active conformation (reviewed in Ref. 15). In contrast, in the absence of substrate, the catalytic domain of DUSP2 and DUSP6 is in a disordered state. As shown for DUSP6, binding to the substrate ERK2 MAPK then strongly increases the catalytic activity (21). The type III MKP DUSP10/MKP-5 contains a unique N-terminal domain of unknown function that is suspected to contain regulatory sites (22). In contrast to DUSP6, the catalytic activity of DUSP10 is not increased by binding to p38 or JNK (22). The larger type IV MKP DUSP8/VH5 and DUSP16/MKP-7 have a C-terminal region with a number of interesting features including proline/glutamine/serine/threonine rich, nuclear localization, and leucine-rich nuclear export sequences that play a role in stability of the protein, interaction with scaffold proteins, and control nuclear-cytoplasmic shuttling of the protein (23, 24, 25, 26, 27).

Patterns of DUSP expression: changes during development, between cell types, and in response to external stimuli

In contrast to their substrate MAPK that are expressed ubiquitously, many DUSP genes show regulated expression during development, in a cell type-specific manner or in response to cellular activation. A good example for the latter is the prototypic MKP, DUSP1/MKP-1, that was cloned as immediate-early gene 3CH134 from serum stimulated human fibroblasts (28, 29). Similarly regulated by growth factors or in response to cell stress are DUSP2/PAC1, DUSP4/MKP-2, and DUSP5/VH3. DUSP9/MKP-4 is highly expressed in the developing liver and placenta, but at low levels in other organs, illustrating developmentally regulated expression (30, 31).

The availability of microarray technologies was the prerequisite for the identification of DUSP as differentially regulated genes under many different experimental conditions through large scale gene expression profiling. Individual DUSP were found up-regulated in various cancers, in response to hypoxia, heat shock, and other perturbations of cells and tissues. Of importance, DUSP have surfaced in several studies investigating innate and adaptive immune responses. Kovanen et al. (32, 33) analyzed T cell transcriptional responses to cytokine signaling through the common γ-chain and identified DUSP5/VH3 and DUSP6/MKP-3 as up-regulated by IL-2, IL-7, and IL-15. Tanzola and Kersh (34) performed a real-time PCR profile of DUSP expression during thymic development that showed expression and regulation of 7 of 10 DUSP genes at various stages of T cell development and following stimulation of the TCR complex. In a comprehensive microarray experiment of human leukocyte subsets, Jeffrey et al. (35) analyzed the regulation of DUSP genes following activation of B cells, T cells, mast cells, eosinophils, macrophages, and DC, and found high levels of induction for the nuclear MKP DUSP1, DUSP2, DUSP4, and DUSP5 in activated immune cells. Mining of transcriptome datasets from primary mouse macrophages stimulated with LPS or the RAW 264.7 macrophage cell line after activation with a set of TLR ligands, our laboratory identified DUSP1, DUSP2, and DUSP16 as the most strongly induced MKP; thus, there is substantial overlap between these studies in the human and mouse system (36, 37). Finally, genome-wide analyses of the host pathogen interaction have revealed induction of several DUSP family members in macrophages infected with mycobacteria (38), Gram-negative and -positive bacteria (39), and virus-infected HeLa cells (40).

Defining essential functions for DUSP in immunity: the examples of DUSP1, DUSP10, and DUSP2

During the first decade of research into DUSP-MKP, the biochemistry of the interaction with MAPK was elucidated for some members in great detail. Much less is known about the physiological roles of the individual DUSP, which is only beginning to be revealed through the generation and phenotypic analysis of gene-targeted mice. Given the large number of DUSP-MKP, often coexpressed in one cell type, redundancy and functional compensation could be expected when just a single DUSP gene is inactivated. As reviewed in the following sections, for the DUSP-KO mice studied in immunological models to date, strong phenotypes were observed that suggest a surprising degree of nonredundancy in the function of individual DUSP genes (Table II⇓).

DUSP1/MKP-1: from no phenotype to molecular brake in inflammation

The first DUSP MKP was discovered as immediate early gene in the fibroblast serum response and initially thought to dephosphorylate ERK and only later shown to deactivate MAPK in the order p38>JNK≫ERK (41). This initial misconception may also underlie the failure to investigate the effect of DUSP1-deficiency on the different MAPK more closely in the first description of gene targeted mice. From the apparent health of the mice and the unaltered kinetics of ERK activation, it was concluded that DUSP1 does not have nonredundant functions (42). Indeed, a recent analysis of MEFs from DUSP1^−/− mice showed enhanced phosphorylation of p38 and JNK, but not ERK, after osmotic stress and anisomycin, that caused increased cell death (43).

DUSP1/MKP-1 is expressed in many hemopoietic and epithelial cell types, and up-regulated in response to hypoxia (44), heat shock (45), several growth factors, and in various malignancies. In macrophages, the first detection of DUSP1 mRNA was reported in Listeria-infected cells (46). Celada and colleagues (47, 48) then showed that DUSP1 is induced by signaling through the M-CSF receptor as well as in response to LPS, and additionally demonstrated the involvement of PKC in this process. TLR stimuli increase DUSP1 expression through MyD88- or, in the case of poly(I:C), TRIF-dependent signaling (49). Whether DUSP1 is a NF-κB-dependent target of TLR signaling has not been addressed. MAPK signaling controls DUSP1 expression at various levels. In fibroblasts, activation of the JNK pathway, but not of ERK, resulted in increased DUSP1 mRNA (50). Similarly, inhibition of MEK-1 by PD98059 did not interfere with LPS-induced expression of DUSP1 in macrophages (47, 48). However, the role of ERK signaling in DUSP1 mRNA expression is not entirely clear, because the MEK-1 inhibitor U0126 was reported to block LPS-induced increases of DUSP1 at the mRNA and protein level (51). Although TLR signaling activates the DUSP1 promoter, mRNA levels may also be controlled by prolonging the very short half-life of the transcript (36). ERK-dependent signals also increase the stability of DUSP1 protein through phosphorylation (52, 53). Consistent with a negative regulatory function for DUSP1 in macrophage activation, anti-inflammatory glucocorticoids increase DUSP1 expression in macrophages, and mast cells at the mRNA and protein level (36, 54, 55, 56). Recently, we found (36) that the immunosuppressive cytokine IL-10 enhanced and prolonged DUSP1 expression in TLR-stimulated macrophages, acting in synergy with dexamethasone. Finally, the anti-inflammatory effects of the endocannabinoid anandamide on microglia have recently been correlated to up-regulation of DUSP1 protein levels, suggesting that signaling through the CB2 receptor is another pathway that leads to dampened MAPK activation through increased DUSP1 levels (57).

A negative effect of DUSP1 on macrophage activation was suggested by overexpression of DUSP1, which was shown in a number of studies to inhibit phosphorylation of MAPK and the production of the cytokines TNF-α and IL-6 in response to diverse TLR stimuli (51, 55, 58). A requirement for DUSP1 as a physiological regulator of MAPK activation and cytokine production in macrophages was then demonstrated by a series of studies that used DUSP1^−/− mice generated in the Bravo laboratory (49, 59, 60), or an independently created DUSP1^−/− mouse line (61). Together, these studies show that p38 MAPK is the primary target of DUSP1 in activated macrophages, followed by JNK with no or very little effect on ERK activation. DUSP1-deficient macrophages secrete increased amounts of TNF-α, IL-6, and IL-10 after stimulation with TLR-ligands. In the high-dose LPS shock model in vivo, the absence of DUSP1 renders mice highly susceptible to the lethal effects of endotoxin, with a corresponding, substantially elevated production of certain cytokines and mediators. Consistent with the lethal outcome, DUSP1^−/− mice displayed the hallmarks of septic shock, such as depressed circulation, kidney failure, and inflammatory infiltrates in the lung and other tissues, much more pronounced than the surviving wild-type mice (60). Salojin et al. (61) further demonstrated in a model of autoimmune disease that collagen-induced arthritis is markedly more severe in DUSP1-deficient mice, with increased joint swelling and production of IL-6 and TNF-α. DUSP1 is therefore an essential endogenous negative regulator of the systemic and local innate inflammatory response, phenotypically similar in effect to the cytokine IL-10 (62, 63) or the phosphatase SHIP (64). Of note, however, not all cytokine and chemokine responses to the challenge with LPS are dysregulated in the absence of DUSP1, because serum levels of IFN-γ, IL-12p40, and IP-10 after LPS challenge were comparable or even reduced. Moreover, microarray analysis of spleen RNA revealed at the genome-wide level that DUSP1 controls a subset of LPS-induced genes (59, 60). An interesting aspect of these data is the counterintuitive, overwhelming release of IL-10 in the absence of DUSP1, which might be expected to bring about down-regulation of the inflammatory reaction but obviously fails to do so. Although elevated levels of known IL-10 induced genes (37) such as SOCS3, NFIL3, or Bcl-3 suggest intact IL-10 signaling in the absence of DUSP1, an essential fraction of IL-10-induced macrophage deactivation may therefore depend on functional DUSP1. The mutual regulation of expression between IL-10 and DUSP1 is connected in a module with p38 MAPK (Fig. 1⇓), whose contribution to IL-10 expression was also recently substantiated by work from the Ivashkiv and colleagues (6) laboratory.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Schematic overview: regulation of immune cell activation by DUSP1 (A), DUSP2 (B), and DUSP10 (C). Selectivity for the different MAPK and regulation of transcriptional responses are shown as described in the text. A, DUSP1 expression in macrophages is synergistically induced by TLR and IL-10 or glucocorticoid receptor signaling. DUSP1 acts primarily on p38 MAPK, thereby limiting the activity of CREB and AP-1 transcription factors and controlling the expression of a subset of LPS target genes. B, Control of JNK-ERK cross-talk by DUSP2 in mast cells and macrophages. DUSP2 is induced in macrophages by TLR ligands and in mast cells by FcεRI ligation. DUSP2 protein associates MAPKs and blocks JNK-mediated inhibition of ERK activation. In effect, DUSP2 increases expression of inflammatory genes in both cell types. C, Impact of DUSP10 on innate and adaptive immune cells. In macrophages (right), TLR ligands induce DUSP10 expression that inhibits JNK activity and thereby constrains the production of cytokines and may reduce costimulation of T cells. In T cells (left), DUSP10 is constitutively expressed and inhibits early JNK activation after TCR ligation. Increased JNK activity in the absence of DUSP10 results in reduced proliferation and increased AP-1-dependent production of T cell cytokines.

DUSP10/MKP-5: acting on innate and adaptive responses

DUSP10/MKP-5 is the only member of the type III group of DUSP-MKP, carrying an additional N-terminal domain of poorly characterized function that may control its cytoplasmic or nuclear location (22). JNK and p38 MAPK appear to be the preferred substrates of DUSP10 (22). Although DUSP10 is expressed at much lower levels in immune cells compared with many other DUSP, as judged from microarray data, gene targeting in mice revealed an interesting phenotype with distinct functions for DUSP10 in innate, and adaptive immunity (Fig. 1⇑) (65). DUSP10 is inducible by TLR stimulation in macrophages, which showed increased activation in its absence in terms of elevated cytokine production in vitro and in vivo, although the difference does not seem to be as big as in DUSP1^−/− mice. In CD4⁺ T cells, DUSP10 is constitutively expressed but down-regulated 24 h after TCR activation. In the absence of DUSP10, Th cells activate JNK more strongly early after activation. Although DUSP10-deficient T cells produced strikingly increased amounts of both Th1 and Th2 cytokines, their Ag-specific proliferation was impaired, consistent with an inhibitory role for JNK on T cell growth observed earlier (66). Impaired T cell expansion in DUSP10-deficient mice is also the likely cause of an amelioration of disease symptoms in the MOG-induced EAE model, where brain infiltration by Th cells was abrogated nearly completely. Finally, in viral lymphocytic choriomeningitis virus (LCMV) infection, strongly enhanced production of the T cell-derived cytokines IL-2, IFN-γ, IL-4, and TNF-α upon secondary infection resulted in the immune-mediated death of DUSP10-deficient mice. Contrary to what might be expected given the reduced proliferation of T cells in vitro, the frequency of LCMV-reactive CD8 and CD4 T cells was not reduced in the absence of DUSP10. This discrepancy may be explained by the increased APC function of DUSP10^−/− innate immune cells that compensated for T cell autonomous growth-inhibitory effects (Fig. 1⇑).

DUSP2/PAC1: from feedback inhibition to driving force of inflammation

Cloned in 1993 as an immediate early gene from TCR-activated T cells (67), DUSP2/PAC-1 localizes to the nucleus and primarily inactivates p38 MAPK and ERK in vitro (68). DUSP2 was found expressed in malignant tissues where high levels correlated with poor prognosis (69). It was also described as a transcriptional target of p53 that is required for the induction of apoptosis in a colon cancer cell line (70). In another study, DUSP2 expression was found to be controlled by ERK activation (71). In nontransformed cells, the expression of DUSP2 appears to be restricted predominantly to the hemopoietic lineages, where it is highly inducible in macrophages, mast cells, eosinophils, B, and T cells by activation of Ag or Fc receptors (35). This distinguishing feature of DUSP2 may already indicate a special function for this phosphatase in immune cells. Indeed, the recently reported (35) phenotype of the DUSP2^−/− mouse confirmed a significant role in the immune system, although the results contradicted expectations and introduced new twists into the tale of MAPK regulation by DUSP-MKP. Similarly to what was observed in the case of DUSP1 or DUSP10, the immune system of DUSP2^−/− mice develops normally. Contrasting to the effects of DUSP1 and DUSP10 deficiency, DUSP2^−/− macrophages and mast cells produced fewer inflammatory mediators after stimulation via TLR and FcεR, respectively. In addition, DUSP2^−/− mast cells failed to proliferate long term, which was due to an increased rate of apoptosis. In harmony with this impaired responsiveness in vitro, DUSP2^−/− mice were protected in the K/BxN arthritis model that particularly depends on the effector functions of mast cells and macrophages.

How can this surprising positive effect of DUSP2 be explained at the molecular level? From overexpression studies in vitro, DUSP2 appeared to dephosphorylate p38 and ERK (68). However, in the absence of DUSP2, a decreased or shortened activation of these MAPK was observed in macrophages and mast cells. A reduced activation of Elk- and AP-1 reporter genes showed the functional significance of reduced ERK activation in DUSP2^−/− cells. In contrast, JNK activation was increased and the use of JNK inhibitors rescued ERK phosphorylation and Elk reporter gene activation, suggesting that DUSP2 strengthens ERK activation by blocking JNK. Finally, the authors also showed that DUSP2 physically interacts with p38, JNK, and ERK, providing a basis for the cross-talk between MAPK that is suggested by the data (Fig. 1⇑). Such a negative cross-talk relationship between JNK and the ERK pathway had been previously described, seemingly dependent on Jun-induced gene expression (72). It is worth mentioning that a positive regulatory role for a DUSP in MAPK activation is not without precedent, as both human DUSP22/JSP-1 and its murine orthologue DUSP22/JKAP were found to specifically activate JNK upon overexpression (73, 74), and murine DUSP22 was indeed required for full activation of JNK in response to TNF-α (74). Together, these results are important in demonstrating that DUSP proteins not only inhibit biological responses by terminating MAPK activation, but also can decisively increase the activation of a specific MAPK.

Inclusion of DUSP in the equation: a more accurate description of MAPK driven immune activation

Recently, the field of signal transduction research has seen a growing appreciation of the specific and active roles of phosphatases in the regulation of physiological processes in general. This shift in attention from kinases to phosphatases was in part due to the realization that mammalian genomes encode a much larger number of phosphatases than previously thought. This is especially true in the case of the MAPK family, where the large family of DUSP MKP outnumbers the small group of MAPK. The inclusion of these regulators in the MAPK signaling module provides specification and flexibility in the downstream responses to MAPK-activating stimuli by introducing a number of additional variables. These include the following: 1) the set of constitutively expressed DUSP in a given cell that controls the initial response type; 2) the kinetics of inducible expression of DUSP as feedback inhibitors; 3) subcellular compartmentalization of negative regulation by the action of cytoplasmic and nuclear DUSP; and 4) DUSP-mediated cross-talk at the level of MAPK that can even result in positive effects on individual MAPK. DUSPs belong therefore to a larger group of molecules, including scaffold proteins and other inhibitors, that determine the subcellular localization, the intensity and the kinetics of differential MAPK activation in different cell types in response to various stimuli (75, 76).

The studies reviewed here demonstrate that DUSP fulfill an important function in innate and adaptive immune cell responses. Following the identification of different DUSP as genes differentially expressed and regulated in immune cells, the modulation and specification of immune receptor initiated MAPK signaling by this family of MKP is being revealed in ongoing biochemical and genetic studies in the mouse model. It is likely that phenomena of differential MAPK activation and biological responses to different stimuli or by different cell types may, at least in part, be attributable to the action of DUSP. For example, the findings that the balance between p38 and ERK activation in DC in response to TLR triggering determines the relative strength of immunosuppressive IL-10 or Th1-driving IL-12 production (9, 10) may be explained by regulated expression patterns of DUSP family members. DUSP1^−/− macrophages have increased p38 MAPK activation whereas ERK phosphorylation is unchanged, but contrary to what might be expected this leads to elevated production of IL-10 but not of IL-12 (59, 65). Therefore, the equations that translate expression levels of DUSP and activation states of MAPK into transcriptional responses may not be that simple. There is now a need to investigate the contribution of individual DUSP to the regulation of the MAPK module in inflammation and infection, and thereby generate the data necessary to establish these rules.

Impact of DUSP on other signaling pathways?

Another important question for future study is whether and how DUSP interfere with other signaling pathways, be it indirectly via changes in MAPK activity or directly by acting on additional substrates. At the moment, there is no evidence for direct binding and dephosphorylation of non-MAPK targets by the classical DUSP MKP that contain a MKB domain. This is different for the atypical DUSP that can dephosphorylate a wide array of target proteins and even RNA. It would therefore not be too surprising if one or the other of the classical DUSP MKP would also turn out to bind and inactivate novel targets. An indirect impact of the DUSP family is expected for pathways that are already known to cross-talk with the MAPK module. For example, p38-dependent histone modifications and chromatin remodeling in macrophages and DC after TLR stimulation facilitates access of NF-κB transcription factors to the promoters of certain cytokines and chemokines (77). We have observed that NF-κB binding sites are overrepresented in the promoters of DUSP1-contolled LPS-responsive genes, although there was no detectable difference in IκB degradation and nuclear translocation of RelA p65, implying that stronger p38 activation in the absence of DUSP1 may increase the recruitment of NF-κB transcription factors to the promoters of these genes (our unpublished data). Serine phosphorylation of STAT1 mediated by p38 MAPK represents another example of cross-talk that may be indirectly controlled by DUSP proteins and contribute to the shaping of the transcriptional responses (78, 79). Curiously, there are reports from two different groups suggesting that DUSP1 dephosphorylates STAT1 at tyrosine residues in hepatocytes and vascular smooth muscle cells, which would imply a more direct interaction with the JAK-STAT pathway (80, 81). However, these claims were contested by a yeast two hybrid-based investigation of DUSP1 binding determinants that concluded that direct interactions are limited to the MAPK family (82).

DUSP as targets for the manipulation of immune responses?

The phenotypic analysis of DUSP-deficient mice has already shown their importance in regulating immune responses in a few models of systemic and local inflammation. These mouse lines provide ideal tools to investigate the function of DUSP family members in additional models of autoimmune diseases. Likewise, the impact of DUSP on the host response to infection can now be investigated using a range of different pathogens. Because bacterial and viral pathogens induce DUSP1 expression (38, 40), a lack of this presumable immune evasion mechanism may be expected to shift the balance in these infections to a more robust protective response. It is an important question whether changes in the production of inflammatory cytokines and mediators depending on the functionality of DUSP confer a state of more efficient pathogen control to the innate immune system, or may in contrast lead to increased immunopathology. Because DUSP bring about the specification of MAPK signaling, targeting of this class of regulators may perhaps enable us to dissociate desirable protective responses from self-destructive inflammation.

Acknowledgments

We thank Reinhard Hoffmann for critical reading of the manuscript.