The work was completed at the Kennedy Institute of Rheumatology when it was a division of Imperial College London; the Institute became a division of the University of Oxford on August 1, 2011.
Bone formation and destruction are usually tightly linked; however, in disorders such as rheumatoid arthritis, periodontal disease, and osteoporosis, elevated osteoclast activity leads to bone destruction. Osteoclast formation and activation are controlled by many signaling pathways, including p38 MAPK. Dual-specificity phosphatase 1 (DUSP-1) is a factor involved in the negative regulation of p38 MAPK. The purpose of this study was to examine the effect of Dusp1 deficiency on bone destruction.
Penetrance, onset, and severity of collagen-induced arthritis were recorded in DUSP-1+/+ and DUSP-1−/− mice. Bone destruction was assessed by histologic and micro–computed tomographic examination of the joints. The in vitro formation and activation of osteoclasts from DUSP-1+/+ and DUSP-1−/− precursors were assessed in the absence or presence of tumor necrosis factor (TNF).
The formation and activation of osteoclasts in vitro in the presence of TNF were enhanced by Dusp1 gene disruption. DUSP-1−/− mice exhibited higher penetrance, earlier onset, and increased severity of experimental arthritis, accompanied by greater numbers of osteoclasts in inflamed joints and more extensive loss of bone. A DUSP-1−/− mouse colony of mixed genetic background also demonstrated striking spontaneous osteolytic destruction of distal phalanges.
DUSP-1 is a critical regulator of osteoclast activity and limits bone destruction in an experimental model of rheumatoid arthritis. Defects in the expression or activity of DUSP1 in humans may correlate with a propensity to develop osteolytic lesions in arthritis.
The delicate balance between osteoblast-mediated bone synthesis and osteoclast-mediated bone destruction is disturbed in many chronic inflammatory conditions, such as rheumatoid arthritis (RA), psoriatic arthritis, periodontal disease, and cachexia (1, 2). Osteoclast activity becomes excessive, causing a net loss of bone, either at localized sites of inflammation, at sites throughout the body, or both. Skeletal homeostasis normally depends upon a continuous process of remodeling, in which ∼3% of cortical bone and 25% of trabecular bone are replaced per year of adult life (3). This process is dependent upon the tightly coupled and mutually regulated activities of osteoblasts and osteoclasts.
In healthy individuals, several mechanisms of communication between osteoblasts and osteoclasts ensure the tight coupling of their respective activities and the homeostatic maintenance of bone mass and integrity. The most important mechanism of communication involves RANK, a member of the tumor necrosis factor receptor (TNFR) superfamily that is expressed on the cell surface of osteoclast precursors. RANKL, the ligand for RANK, is expressed on the cell surface of osteoblasts, and it provides a mechanism for contact-dependent communication between osteoblasts and osteoclasts (4, 5). RANKL can also be expressed by other mesenchymal cells, such as fibroblasts, and by T lymphocytes (6–8). In RA and in experimental models such as collagen-induced arthritis (CIA), proinflammatory cytokines, including TNFα, interleukin-1 (IL-1), and IL-17, stimulate RANKL expression in synovial fibroblasts and, hence, promote osteoclast activation. The same proinflammatory cytokines may also directly activate osteoclasts, particularly in cooperation with RANKL (9, 10). These indirect or direct mechanisms are thought to contribute to the periarticular bone erosions that are often observed in RA.
When engaged by its ligand, RANK recruits TNFR-associated factor 6 and initiates the activation of the NF-κB and MAPK signaling pathways. The p38 MAPK pathway mediates the up-regulation of the transcription factor NF-ATc1, a factor that is crucial for osteoclast differentiation (11). In addition, p38 MAPK phosphorylates microphthalmia-associated transcription factor (MITF), assisting the recruitment of NF-ATc1 to target sites in chromatin (12, 13). Both NF-ATc1 and MITF regulate expression of genes that are necessary for osteoclast function, cooperating with other transcription factors, such as PU.1 and upstream stimulatory factor, which is itself a target of the p38 MAPK pathway (13, 14). Hence, p38 MAPK plays a crucial role in osteoclast activation.
Inhibitors of p38 MAPK block the formation of mature osteoclasts from precursor cells in vitro (15–17) and decrease the expression of cathepsin K, tartrate-resistant acid phosphatase (TRAP), and osteoclast-associated receptor (12–14, 18, 19). Not surprisingly, p38 MAPK inhibition or genetic ablation of the pathway reduces bone loss in experimental models of arthritis and other inflammatory diseases (20–24). Inhibitors of p38 MAPK have long been considered potential treatments for inflammatory osteolysis (25). Regrettably, due to toxicity or unsustained efficacy, none have yet reached the market (26).
Endogenous mechanisms of constraint of p38 MAPK function may point the way toward new treatments for inflammatory bone loss. Activation of p38 MAPK is mediated by upstream MKKs, which phosphorylate both a threonine and a tyrosine residue in the activation loop. DUSP-1, a member of a family of dual-specificity phosphatases, catalyzes the dephosphorylation of both threonine and tyrosine residues, causing the inactivation of p38 MAPK (27). In macrophages, the expression of DUSP-1 was found to be essential for limiting the strength and duration of p38 MAPK signaling and the expression of proinflammatory gene products (28–33). Both the incidence and severity of CIA were increased in mice in which the Dusp1 gene was disrupted (32). Recently, these mice were reported to display increased bone loss in response to ovariectomy (34) or lipopolysaccharide injection (35). Adenovirus-mediated overexpression was shown to protect against bone loss in the latter experimental model of periodontal disease (36). We describe herein an unusual spontaneous osteolytic condition in a DUSP-1−/− mouse line and provide evidence that DUSP-1 limits inflammatory cytokine–induced osteoclastogenesis and the development of osteolysis in an experimental model of RA.
MATERIALS AND METHODS
DUSP-1+/+ and DUSP-1−/− mice were maintained at a temperature of 21°C (±2°C) on a 12-hour light/12-hour dark cycle and were given food and water ad libitum. All experimental procedures were approved by the local Ethical Review Process Committee and by the UK Home Office.
In vitro osteoclast TRAP assay.
Bone marrow was isolated from 10–13-week-old mice by flushing femora and tibiae with α-minimum essential medium (α-MEM). Cells were counted and plated at a density of 5 × 105 cells/well in 96-well plates (3 wells per treatment) and incubated in α-MEM supplemented with 10% fetal bovine serum, 1% L-glutamine, and 1% penicillin/streptomycin and containing 25 ng/ml of soluble murine macrophage colony-stimulating factor (M-CSF) either alone (as a control) or with 5 ng/ml of RANKL and with or without 2 pg/ml of TNFα. Cells were cultured for 7 days with partial changes of the medium every 3 days. At the end of the culture period, cells were fixed with 4% formaldehyde, permeabilized with a 1:1 (volume/volume) mixture of acetone:ethanol, and osteoclast formation was measured by TRAP staining (0.1 mg/ml naphthol-AS-MX phosphate, 0.4 mg/ml of fast red violet LB salt, 1% dimethylformamide in TRAP buffer) for 15 minutes at 37°C. The numbers of TRAP-positive multinucleated cells containing ≥3 nuclei were counted at 20× magnification.
In vitro osteoclast resorption assay.
The resorptive activity of osteoclasts was determined by a lacunar resorption assay on dentin slices. Cells were cultured for 14 days on dentin slices in the presence of 25 ng/ml of M-CSF either alone (control) or with 5 ng/ml of RANKL and with or without 2 pg/ml of TNFα. Osteoclasts were then lysed, and the dentin slices were rinsed in distilled H2O. Resorption lacunae were marked with a marker pen, and the resorption area was analyzed using ImageJ software version 1.45d.
Induction of CIA.
Male mice ages 10–12 weeks were immunized by subcutaneous injection of 100 μl of chicken type II collagen (2 mg/ml) emulsified in Freund's complete adjuvant; 2 sites at the base of the tail were injected as previously described (37). Beginning at the onset of arthritis, animals were scored for clinical signs of inflammation. Each limb was scored on a scale of 0–3, where 0 = normal, 1 = slight swelling, 2 = pronounced edematous swelling, and 3 = pronounced swelling and joint rigidity.
Hind paws from arthritic mice and mice with toe swelling were assessed histologically by TRAP staining for osteoclasts and by hematoxylin and eosin (H&E) staining for inflammatory infiltration and joint damage. Hind paws were fixed in 10% neutral buffered formalin and decalcified in 10% EDTA for 21 days. Paws were then embedded in paraffin and cut into 3-μm sections. Sections were deparaffinized in xylene, rehydrated, and stained for TRAP or with H&E (38). Histologic changes on H&E-stained sections were scored on a scale of 0–3, where 0 = normal, 1 = mild inflammation with no joint damage, 2 = moderate inflammation with some joint erosions, and 3 = severe inflammation and loss of joint architecture. Sections of hind paws affected by toe swelling were also stained with Masson's trichrome or neutrophil elastase, and nuclei were stained with hematoxylin.
Micro–computed tomography (micro-CT).
Hind paws from mice with psoriatic onychopachydermoperiostitis (POPP) and from mice with arthritis were harvested 10 days after the onset of disease and imaged using a SkyScan 1072 scanner at a resolution of 16μ. Micro-CT scanning was performed by Kevin MacKenzie (University of Aberdeen, Aberdeen, UK). Images were analyzed using SkyScan CT Analyzer software version 188.8.131.52. For measurement of bone volume, a 1.5-mm region of the paws that included all 5 metacarpal joints was selected. A threshold of 115 density units was selected to distinguish mineralized from nonmineralized tissue.
Data are reported as the mean ± SEM. GraphPad Prism software was used for all statistical analyses. Disease onset, histologic scores, and TRAP-positive cell counts in histologic sections were analyzed by Student's unpaired 2-tailed t-test. The in vitro TRAP and resorption assay results, the clinical scores, and the bone volume measurements were analyzed by one-way analysis of variance followed by Bonferroni post hoc test. Disease incidence was analyzed by chi-square test. P values less than 0.05 were considered significant.
Spontaneous osteolytic disease in a DUSP-1−/− mouse colony.
DUSP-1−/− mice were provided by Bristol-Myers Squibb (39), and the strain was rederived by implantation of embryos into pseudopregnant C57BL/6 females. At this point, the genetic background of the strain was found to be ∼75% C57BL/6 and 25% 129Sv, typical of knockout strains that have not been extensively backcrossed to create a pure genetic background. A breeding colony from these mixed background mice was established from 6 animals obtained in the rederivation. To investigate the function of the Dusp1 gene, we backcrossed the DUSP-1−/− mice against the C57BL/6 background and then interbred heterozygotes and performed genotyping to identify DUSP-1+/+ and DUSP-1−/− littermates for use in experiments (29). Subsequently, 8 further generations of backcrossing against C57BL/6 were also used to generate DUSP-1+/+ and DUSP-1−/− strains with a >99.9% C57BL/6 genetic background, and these strains have recently been used in the study of DUSP-1 function (40).
In the original DUSP-1−/− breeding colony of mixed genetic background (C57BL/6 and 129Sv), we noted a spontaneous phenotype characterized by severe swelling of the toes, giving a “drumstick” appearance (Figure 1A). The number of affected digits ranged from 1 to 20, and the toenails were typically dystrophic or even absent from the affected digits. The swelling did not resolve, although erythema noted during the early phase tended to disappear, over time. In a cohort of 73 female mice kept for 20 months, the median age at onset was 6 months, and the cumulative incidence was close to 60% (Figure 1B). In 6 years since the phenotype was first observed, only 2 males from the same breeding colony have developed similar symptoms. This phenotype is absent from the DUSP-1−/− line that has been extensively backcrossed against C57BL/6. Other researchers who have maintained DUSP-1−/− mouse strains on different genetic backgrounds have not observed any similar condition (Cato A, Bennett A: personal communications).
Patterns of behavior of affected and unaffected females were studied using a Laboratory Animal Behavior Observation Registration and Analysis System (LABORAS) instrument, which measures time spent immobile, grooming, eating, drinking, running, and climbing. The only difference detected was an absence of climbing in mice with swollen toes, probably a simple reflection of physical incapacity (data not shown). There was no evidence that the condition caused significant discomfort, and affected females were also of similar weight as unaffected females (data not shown). Micro-CT was used to assess bone structure in affected mice. This technique dramatically illustrated the catastrophic breakdown of the architecture of the terminal phalanges, with extensive osteolysis as well as uncoordinated bone formation leading to the stellate appearance of the bone (Figure 1C).
Radiographic imaging showed a rather diffuse, “fluffy” signal in the region of the affected digits (Figure 2A). As expected from the micro-CT imaging results, histologic inspection showed that the distal phalanges were extensively eroded and fragmented (Figure 2C). Figure 2B shows sagittal sections of unaffected and affected toes stained with Masson's trichrome, showing keratin as red and showing bone and collagen as blue or green. There was general enlargement of the affected toe (Figure 2B, bottom), with hypertrophy of the nailbed, thickening of the dermis, fragmentation of the distal phalanx, and with preservation of the distal interphalangeal joint, all of which are consistent features of the condition.
We did not detect inflammation of the interphalangeal joints, pannus formation, periarticular bone erosion, or indeed osteolysis of any bone other than the distal phalanx. Vestigial bone fragments were embedded in large masses of connective tissue that stained pale green (Figure 2C, top). Scanning electron microscopy confirmed the presence of abundant collagen fibers within this tissue (results not shown). Affected toes were markedly infiltrated by neutrophils, as demonstrated by neutrophil elastase staining (Figures 2A and C). Neutrophils were particularly abundant around the nailbed, but were also closely associated with bone fragments (Figure 2C, middle). TRAP staining demonstrated the presence of multinucleated osteoclasts decorating the surfaces of bone fragments (Figure 2C). In some cases, erosion of the distal phalanges could be detected by radiographic imaging or histologic examination in the absence of macroscopically obvious swelling. In all cases where swollen toes were inspected, extensive bone destruction was found, even if the mice were killed within 24 hours of the appearance of swelling and histologic assessment performed. It is therefore likely that osteolysis is a very early event, followed later by fibrosis and swelling.
Negative regulation of in vitro osteoclast activation by DUSP-1.
The low penetrance of the condition and its variable age at onset made it difficult to capture the early stages of pathology. The absence of similar symptoms in other DUSP-1−/− lines also made it difficult to be certain about the contribution of Dusp1 gene disruption. Because hyperactivation of osteoclasts appeared to be an early event in the pathologic changes of the disease, we investigated the function of osteoclasts derived from DUSP-1−/− and DUSP-1+/+ colony mice on a pure C57BL/6 background (>99.9% C57BL/6). Osteoclasts were differentiated from bone marrow cells in the presence of M-CSF and soluble RANKL, with or without TNFα. The proinflammatory cytokine TNFα was included in these experiments because it had previously been shown to stimulate the formation of human osteoclasts in the presence of suboptimal concentrations of RANKL (41). Differentiation of osteoclasts was assessed by counting multinucleated cells staining positive for TRAP, a marker of osteoclast activation. Function was assessed by plating the cells on dentin slices and then measuring the area of resorption after 14 days.
As expected, the formation of osteoclasts was not promoted by M-CSF alone, but required the addition of RANKL (Figures 3A and B). As previously described (34), RANKL induced both the activation of p38 MAPK and the up-regulation of Dusp1 gene expression in cultures of cells from wild-type mice (data not shown). Again, consistent with previous findings (34) and with the known function of the DUSP-1 protein, RANKL-induced p38 MAPK activation appeared to be prolonged in cultures of cells from DUSP-1−/− mice (data not shown). In the presence of both M-CSF and RANKL, the numbers of multinucleated, TRAP-positive osteoclasts and the area of dentin resorption were not significantly different in cultures of cells from DUSP-1−/− mice as compared to wild-type mice (Figures 3B and D). However, the addition of TNFα (2 pg/ml) showed a striking cooperation with M-CSF and RANKL to promote the formation and activation of osteoclasts in cultures of cells from DUSP-1−/− mice, whereas a similar synergy was not observed in cultures of cells from DUSP-1+/+ mice (Figure 3). Thus, in the presence of all 3 agonists, M-CSF, RANKL, and TNF, the numbers of TRAP-positive osteoclasts and the area of dentin resorption were both significantly greater in cells from DUSP-1−/− mice than in those from DUSP-1+/+ mice. Therefore, DUSP-1 normally limits the activation of osteoclasts in vitro, a function that is likely to be particularly important under proinflammatory conditions.
Enhanced osteolysis in mice with Dusp1 gene disruption and CIA.
Arthritis was induced by immunization of age-matched male C57BL/6 DUSP-1+/+ and DUSP-1−/− mice with chicken type II collagen. This protocol (37) is a variation on the classic CIA model, in which arthritis is induced using bovine type II collagen in susceptible mice of the DBA/1 genetic background. CIA is driven by proinflammatory cytokines (particularly TNFα) that are expressed by activated macrophages and infiltrating lymphocytes in the joint. It is characterized by joint inflammation and progressive destruction of cartilage and bone.
Consistent with a previous report (32), we found that disease penetrance was significantly higher in DUSP-1−/− mice than in DUSP-1+/+ mice (Figure 4A). In a total of 4 independent experiments, 75% of the immunized DUSP-1−/− mice (42 of 56) developed disease, whereas only 35% of the DUSP-1+/+ mice (24 of 69) did so. The latter is fairly typical of disease incidence in wild-type C57BL/6 mice, suggesting that disruption of the Dusp1 gene increases susceptibility to experimental arthritis. The onset of disease was significantly earlier in DUSP-1−/− mice (Figure 4B) and the initial progression significantly more rapid (Figure 4C). The maximum clinical score was reached within 3 days of disease onset in DUSP-1−/− mice, whereas DUSP-1+/+ mice reached this level of disease severity only after 6 or 7 days. At the macroscopic level, swelling of the affected limbs appeared to be more severe in DUSP-1−/− mice 6 days after disease onset, suggesting that increases in disease severity may not be reflected most effectively by the standard clinical scoring method (Figure 5A).
Ten days after disease onset, the mice were killed, and their paws were harvested and processed for histologic (Figure 5) or micro-CT (Figure 6) analysis. According to the standard histologic scoring system evaluating H&E-stained sections of the proximal interphalangeal joints, the disease was more severe in DUSP-1−/− mice than DUSP-1+/+ mice (Figures 5B and D). Inflammatory infiltration of the joint, hyperplasia of the synovium, and erosion of the cartilage (Figure 5B, higher-magnification view) and bone (Figure 5C, higher-magnification view) were all enhanced in DUSP-1−/− mice. In DUSP-1+/+ mice, the majority of osteoclasts detected by TRAP staining were on the interior surface of cortical bone and were therefore possibly engaged in normal bone turnover rather than periarticular erosion (Figure 5C). In DUSP-1−/− mice, the numbers of TRAP-positive osteoclasts in the joint were 3 times higher than in DUSP-1+/+ mice (Figure 5D), with many of these osteoclasts being clearly associated with the interface between the pannus and the periosteal surface of the bone (Figure 5C, higher-magnification view), consistent with the principal sites of bone erosion in RA and its experimental models.
Further analysis by micro-CT illustrated enhanced bone erosion in DUSP-1−/− mice as compared to DUSP-1+/+ mice (Figure 6A). The erosions were particularly evident around the metacarpophalangeal (MCP) joint, which was therefore used to calculate the net loss of bone volume by comparison to nonimmunized mice of the same age (Figure 6B). We found no significant difference in the appearance of paws from the nonimmunized DUSP-1−/− and DUSP-1+/+ mice. The induction of experimental arthritis was associated with a significant decrease in bone volume at the MCP joint in DUSP-1−/− mice but not in DUSP-1+/+ mice. These images highlight the significantly enhanced bone disease in the DUSP-1−/− mice and strengthen the case for a role of DUSP-1 in the control of inflammatory osteolysis.
Although the DUSP-1–knockout mouse was initially described as having no discernible phenotype (39), several recent publications indicate an important role for this phosphatase in the limitation of inflammatory responses (28, 42), thus emphasizing the importance of challenging gene-deficient animals. For example, DUSP-1−/− mice displayed exaggerated responses in experimental models of sepsis (43–45), asthma (46), and pulmonary hypertension (47). Together with previously published data (32, 34, 35), our findings indicate that DUSP-1 is also an important negative regulator of inflammatory bone loss.
Although interesting, the unusual osteolytic phenotype in our mixed genetic background DUSP-1−/− mouse colony is clearly not a straightforward consequence of disruption of the Dusp1 gene. To our knowledge, no similar phenotype has been observed in several laboratories that maintain either the original DUSP-1–deficient mice (39) on different genetic backgrounds or a second, independently derived knockout (32). It is formally possible that an entirely different genetic alteration occurred during our rederivation and reestablishment of the DUSP-1−/− colony from small numbers of animals and that this unknown mutation has become fixed in the population, directly causing the osteolytic phenotype. Although this cannot be ruled out without an extensive program of breeding and genetic analysis, it seems unlikely. DUSP-1−/− osteoclast precursors of almost pure (C57BL/6) genetic background are hyperresponsive to proinflammatory stimuli, DUSP-1−/− mice of the same genetic background display strong increases in osteoclast activity in an experimental model of RA, and the unusual, spontaneous osteolytic phenotype of our mixed-background DUSP-1−/− mouse colony is characterized by enhanced osteoclast activity in the context of neutrophil infiltration, a classic hallmark of inflammation. The most conservative hypothesis is that disruption of the Dusp1 gene and consequent dysregulation of MAPK signaling contribute to (but do not entirely explain) an increased susceptibility to spontaneous inflammatory osteolysis.
POPP, a very rare variant of psoriatic arthritis, is characterized by onychodystrophy, connective tissue thickening, periostitis of the distal phalanx, and complete sparing of the interphalangeal joints. The toes of affected individuals have a macroscopic drumstick appearance and a fuzzy appearance on radiographs, with evident erosion of the terminal phalanx but no damage to other bones or interphalangeal joints (48, 49). All of these features are common to the spontaneous osteolytic phenotype we observed in the mixed-background DUSP-1−/− strain. While POPP appears to be resistant to conventional psoriatic arthritis treatments, such as methotrexate, it has been successfully treated with neutralizing antibodies against TNFα (50). It would be premature to suggest that the mixed-background DUSP-1−/− strain is an experimental model of POPP, or even that a disease with fewer than 20 reported cases worldwide is seriously in need of an experimental model. However, it may be interesting to investigate whether there are any abnormalities of expression of the DUSP1 gene in humans with POPP.
Importantly, disruption of the Dusp1 gene may increase susceptibility to inflammatory osteolysis in two different ways. First, activated macrophages lacking DUSP-1 overexpress TNF and other mediators of inflammation (30–33, 40). Second, we recently reported that TNF enhances osteoclast formation in the presence of low concentrations of RANKL (41), and we show here that DUSP-1−/− osteoclast precursors are particularly sensitive to the combination of RANKL and TNF. We suggest that hyperactivation of osteoclasts in the context of an inflammatory response is likely to be a consequence of both overexpression of inflammatory cytokines by macrophages and enhanced sensitivity of osteoclasts to the effects of these factors. Osteoclast-specific deletion of Dusp1 is required for further validation of these findings in vivo. It is possible that an unknown mutation or genetic combination in our original mixed-background DUSP-1−/− colony causes an increased susceptibility to inflammation of the toes, for example by impairing the clearance of infections of the nailbed. According to this hypothesis, hypersensitivity of osteoclast precursors then leads to a striking, but highly localized, osteolytic reaction.
DUSP-1−/− mice with experimental arthritis were previously shown to express elevated levels of serum TNF (32). This, combined with the enhanced sensitivity of DUSP-1−/− osteoclast precursors to TNF, is likely to contribute to the increased numbers of activated periarticular osteoclasts and the enhanced erosion of periarticular bone seen in DUSP-1−/− mice with experimentally induced arthritis. An obvious and interesting question is whether the DUSP-1 protein in humans has the same function as its murine counterpart. If this were the case, differences in the expression or function of DUSP-1 could contribute to different rates of progression of bone erosion in RA and other chronic inflammatory diseases, and measurement of such differences could have some predictive utility: further studies in RA and osteoarthritis cohorts are required. Conversely, treatments aimed at increasing or sustaining the expression of DUSP-1 might effectively reduce the bone erosion that is such a debilitating feature of RA.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Horwood had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Vattakuzhi, Abraham, Clark, Horwood.
Acquisition of data. Vattakuzhi, Abraham, Freidin.
Analysis and interpretation of data. Vattakuzhi, Freidin, Clark, Horwood.
Special thanks to Ann Sandison and David Peston (Imperial College London, Histopathology Unit, Charing Cross Hospital) for their contributions to the histologic analysis of the mixed-background DUSP-1−/− mice. We are grateful to Bristol-Myers Squibb for permission to use the DUSP-1−/− mice and to Andy Cato (Karlsruhe Institute of Technology, Karlsruhe, Germany) for supplying the mice.