IDENTIFICATION OF DUAL-SPECIFICITY PHOSPHATASE 5 (DUSP5) AS A REGULATOR OF MITOGEN- ACTIVATED PROTEIN KINASE ACTIVITY IN INTERLEUKIN-2 SIGNALING*

Interleukin (IL)-2, IL-4, IL-7, IL-9, IL-15, and IL-21 form a family of cytokines based on their sharing the common cytokine receptor chain, c, which is mutated in X-linked severe combined immunodeficiency (SCID). As a step toward further elucidating the mechanism of action of these cytokines in T-cell biology, we compared the gene expression profiles of IL-2, IL-4, IL-7, and IL-15 in T cells using cDNA microarrays. IL-2, IL-7, and IL-15 each induced a highly similar set of genes, whereas IL-4 induced distinct genes correlating with differential STAT protein activation by this cytokine. One gene induced by IL-2, IL-7, and IL-15 but not IL-4 was dualspecificity phosphatase 5 (DUSP5). In IL-2-dependent CTLL-2 cells, we show that IL-2-induced ERK-1/2 activity was inhibited by wild type DUSP5 but markedly increased by an inactive form of DUSP5, suggesting a negative feedback role for DUSP5 in IL-2 signaling. Our findings provide insights into the shared versus distinctive actions by different members of the c family of cytokines. Moreover, we have identified a DUSP5-dependent negative regulatory pathway for MAPK activity in T cells.

The common cytokine receptor ␥ c chain (␥ c ) 1 is essential for normal immune development and function. Mutations in ␥ c result in X-linked severe combined immunodeficiency (XSCID) (1), a disease in which affected individuals are highly susceptible to infections resulting from the defective development of T and NK cells and nonfunctional B cells. ␥ c is a component of the receptors for multiple cytokines, including IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 (2). These cytokines have both redun-dant and distinctive actions on lymphocytes. IL-2 is a growth and survival factor for activated T lymphocytes and is also essential for activation-induced cell death (AICD) and the prevention of T cell anergy, processes involved in the maintenance of peripheral tolerance (3). IL-4 is also a T-cell growth factor and is necessary for the development of Th2 cells (4), which regulate humoral immune responses, whereas IL-7 is required for the development and growth of T lymphocytes but can also support the growth of peripheral T lymphocytes (5)(6)(7). IL-15 has overlapping functions with IL-2, but, unlike IL-2, IL-15 is essential for NK-cell differentiation and is more important than IL-2 in the generation of CD8 ϩ memory T cells (8). Although transgenic expression of IL-9 causes thymomas, suggesting that IL-9 can regulate lymphocyte growth (9), IL-9 knock-out mice do not manifest abnormalities in T-cell development or function but exhibit defective mast cell proliferation and mucus production (10). Based on in vitro studies, IL-21 has potential roles for T-cell, B-cell, and NK-cell biology (11).
Only limited information is available regarding the genes that are activated by ␥ c -dependent cytokines. To investigate the basis for overlapping and distinctive actions of ␥ c -dependent cytokines, we used the "Lymphochip" microarray, a specialized cDNA array enriched for genes expressed in lymphocytes (12), to identify genes regulated by IL-2, IL-4, IL-7, and IL-15 in peripheral blood T lymphocytes. IL-2, IL-7, and IL-15 regulated most genes in a similar manner, whereas the pattern seen with IL-4 was more distinctive. Although most ␥ c -dependent genes are redundantly regulated by various stimuli, certain cytokine-specific patterns of gene expression were noted. One gene induced by IL-2, IL-7, and IL-15 but not IL-4 was dualspecificity phosphatase 5 (DUSP5) (13,14). IL-2 is known to activate several signaling pathways, including Ras-MAP kinase, Jak-STAT, and phosphoinositol 3-kinase/Akt/p70 S6 kinase pathways (15), and we now provide evidence for a role for DUSP5 as part of a negative feedback loop that controls IL-2induced MAPK activity in T cells. This is the first demonstration of a role for DUSP5 in T-cell biology and identifies a mechanism by which IL-2-mediated MAPK activation can be controlled. studies. Cells were then not stimulated or stimulated with IL-2 (100 units/ml), IL-4 (100 units/ml), IL-7 (200 units/ml), or IL-15 (200 units/ ml) for 4 h. We also separately performed more detailed time course experiments (0.5, 2, 4, 6, or 8 h) with IL-2 and IL-4. In some instances, PBMCs were stimulated for 1, 3, 6, or 24 h with phorbol 2-myristate 3-acetate (PMA) (50 ng/ml) and ionomycin (1.5 M) (PI).
B lymphocytes (Ͼ98% pure by flow cytometry) were isolated from PBMCs by negative magnetic selection using StemCell Technologies human B-cell enrichment mixture and cultured for 0, 1, 3, 6, or 24 h at 5 ϫ 10 6 cells/ml in complete RPMI medium containing 50 g/ml anti-IgM (Jackson Laboratories).
Cell Lines and Chronic Lymphocytic Leukemia (CLL) Samples-Cell lines (Jurkat, Ramos, SUDHL10, L428, MCF-7, and PC-3) were cultured in complete RPMI medium. OCI-Ly8 cells were grown in Iscove's modified Dulbecco's modified Eagle's medium in the presence of 20% human plasma and 1% penicillin/streptomycin. CLL cells were obtained from untreated patients, and CD19 ϩ cells were purified by magnetic selection.
RNA Isolation-For cDNA microarray analyses, mRNA was prepared using the Fast-Track kit (Invitrogen). Total RNA was isolated using the Trizol reagent (Invitrogen).
cDNA Microarray Analyses-Microarray analysis was performed as described in the "Microarray Procedures" section under "Methods" in Ref. 16, using Lymphochips that contained either 7,296 or 12,672 array elements. RNA samples from cytokine-stimulated T cell cultures were reverse transcribed, labeled with Cy5, and these probes were hybridized to microarrays together with Cy3-labeled probes generated from nonstimulated control cultures. Cy5-labeled probes were also generated from T cells stimulated with IL-2 or IL-4 for various time periods, PBMCs activated with PI, B-cells stimulated with anti-IgM beads, and cell lines and CLL cells. These were hybridized together with Cy3labeled cDNA generated from a common reference mRNA pool (16). This allowed us to compare the relative expression level of a given gene across all of our experiments. Microarrays were analyzed on a GenePix scanner (Axon Instruments), and data files were entered into a custom data base maintained at the National Institutes of Health (nciarray. nci.nih.gov). We used a standard global normalization approach for our expression data analogous to that previously used (16,17). We extracted data for clustering analysis (programs "Cluster" and "Tree-View") (19) that fulfilled the following requirements: spot size of at least 25 m, minimum intensities of 100 relative fluorescent units (RFU) in the Cy3 and Cy5 channels or minimum intensity of at least 1000 RFU in one of the two channels. The 100 RFU criteria are as previously used for the Lymphochip and Axon 4000A scanner (17). Unless stated otherwise, the expression of a gene was considered induced or repressed if the median induction or repression was at least 2-fold with any one cytokine in at least two of three experiments (five experiments were done with IL-2, three with IL-4, four with IL-7, and 3 three with IL-15).

IL-2, IL-7, and IL-15 Induce a Highly Similar Set of Genes, whereas IL-4 Induces Both Overlapping and Distinct Genes-
Using cDNA microarrays, we identified 137 genes that were induced at least 2-fold by IL-2, IL-4, IL-7, or IL-15 in preactivated T lymphocytes (genes induced or repressed at least 2-fold are in Figs. 1-3 and will be made available at llmpp.nih.gov/ cytokines). A one-sample t test was performed on the log ratios to evaluate the significance of the observed fold changes. We observed that 99% of IL-2-regulated (five experiments), 100% of IL-7-regulated (four experiments), 74% of IL-4-regulated (three experiments), and 54% of IL-15-regulated (three experiments) genes were differentially expressed to a degree that achieved statistical significance (two-sided p Ͻ 0.05). We hypothesize that the lower significance for the IL-4 and IL-15 groups results in large part from the lower power associated with having only three experiments for each of these cytokines. The induced genes (Figs. 1 and 2) included many genes previously known to be regulated by IL-2, such as those encoding the IL-2R␣ chain, cyclin D2 (15), SOCS1 (19), CIS1 (20), and Pim-1 (21), indicating the validity of our analysis. IL-4R␣ was also induced by IL-4, as reported previously (22). We also identified 34 repressed genes (Fig. 3). Although the use of 2-fold induction criteria is common, smaller changes in gene expression levels can also be biologically significant, and certain known IL-2induced genes were not identified by the 2-fold criteria. We therefore searched the microarray data for genes whose expression was induced or repressed by more than 40% (based on microarray analysis) in at least any 8 of the 15 experiments. 95 additional genes fulfilled these less stringent criteria (data not shown but will be made available at llmpp.nih.gov/cytokines). This group included c-myc (15) and Bcl-X L (23), genes whose expression levels are known to be regulated by IL-2. Thus, it is important to evaluate genes whose mRNA levels are changed less than 2-fold as well.
The gene expression data were analyzed using average linkage hierarchical clustering (the Pearson correlation coefficient was used as the distance metric) (18), yielding dendrograms that group experiments and genes based on the degree of similarity of their expression patterns (see Fig. 1). We separately analyzed genes that were induced by IL-2, IL-4, IL-7, or IL-15 ( Fig. 1A), and the expression profiles seen with IL-2, IL-7, and IL-15 were very similar (see for example Fig. 1, B-D). Some genes were induced in a similar fashion by all four cytokines (Fig. 1B). A number of genes were strongly induced by IL-2, IL-7, and IL-15 but not induced or only weakly induced by IL-4 ( Fig. 1C). A third set of genes was preferentially induced by IL-4 ( Fig. 1D).
To determine whether the differences between the gene expression profiles for IL-2 and IL-4 represented differences in the kinetics of gene expression between the cytokines rather than induction of different genes, we studied gene expression 0.5, 2, 4, 6, and 8 h after cytokine treatment (Fig. 1, E and F).
In the four panels on the left, multiple experiments are shown at the 4-h time point for IL-2, IL-7, IL-15, and IL-4, whereas in the three panels on the right data are shown for two time courses for IL-2 and one for IL-4. Most genes showed increased expression in a cytokine-restricted manner between 0.5 and 2 h, and this increase was sustained for at least 8 h. For example, leukemia inhibitory factor (LIF) and biliary glycoprotein were induced by IL-2 at 30 min but were not induced by IL-4 at any time point (second and third rows from the top in Fig. 1E).
␥ c -Dependent Cytokines Induce a Group of Genes That Are Highly Expressed in Proliferating Cell Lines but Are Expressed at a Low Level in CLL Cells-Most of the genes we identified have not been previously linked to cytokine responses and many are functionally uncharacterized. One strategy for finding clues to the functions mediated by these genes is to define expression "signatures" characterizing cellular processes. One such expression signature is defined by a set of genes whose expression correlates with cell proliferation in that they are highly expressed in proliferating cell lines ( Fig. 2A, third panel from the left, lanes A-G) but are expressed at a low level in CLL cells (lanes H-L) which are relatively quiescent in their growth properties (16). ϳ20% of the genes induced by IL-2, IL-4, IL-7, and/or IL-15 fulfilled these criteria.
Most Genes Induced by IL-2, IL-4, IL-7, or IL-15 Are Induced by Multiple Stimuli-We sought to identify a set of genes that could distinguish a ␥ c -dependent cytokine response from other activation events, and we compared genes induced or repressed by IL-2, IL-4, IL-7, and IL-15 with those regulated in PBMCs (Ͼ70% T cells) by PI and in B cells after antigen receptor cross-linking (Fig. 2). We found that 73% of the genes induced by the cytokines were also induced in PI-stimulated PBMCs (Fig. 2, A-C, first column of panels from the left). The induction typically occurred within 1 h, minimizing the possibility that PI-dependent cytokine production was responsible for the PI effect. BCR stimulation of B cells induced less overlapping (41%) gene expression profiles (Fig. 2, A and B versus C and D, compare first and second panels from the left), consistent with the use of nonshared signaling pathways or lineage-specific differences in expression. 23% of the ␥ c -dependent genes showed a more restricted expression pattern, as they were not up-regulated by either PI treatment of PBMCs or BCR crosslinking of B cells.
␥ c -Dependent Cytokines Repress Certain Genes, Some of Which Are Highly Expressed in CLL Cells-We also identified 34 genes that were consistently repressed by IL-2, IL-4, IL-7, and/or IL-15 (Fig. 3), most of which were also repressed in PBMCs treated with PI or B cells treated with anti-IgM. Analogous to many of the activated genes being poorly expressed in CLL cells, many of the repressed genes were more highly expressed in CLL cells than in highly proliferating cell lines (Fig.  3, third panel from the left, lanes H-L versus A-G), suggesting a correlation between the expression of these genes and establishing or maintaining a more quiescent state.
Confirmation of Microarray Results by Northern Blotting-We confirmed the induction or repression of select genes (those encoding IL-2R␣, TRAIL, MAPKAPK3, DUSP5, Mal, IL-4R␣ and TSC-22R; blue circles in Figs. 1-3) by Northern blot analysis (Fig. 4). IL-2R␣, TRAIL, and DUSP5 were more potently induced by IL-2, IL-7, and IL-15 than by IL-4. Mal and IL-4R␣ were most strongly induced by IL-4. MAPKAPK3 was induced by all four cytokines, and TSC-22R was repressed by all of the cytokines.
DUSP5 Regulates IL-2-Dependent Phosphorylation and Catalytic Activity of ERK-1/2-One of the genes induced by IL-2, IL-7, and IL-15 but not by IL-4 was that encoding DUSP5, a dual-specificity phosphatase originally cloned from mammary epithelial (13) and liver (14) cell lines. DUSP5 is also known as hVH-3 (14) and is a dual-specificity phosphatase induced by serum stimulation and heat shock (13). DUSP5 can hydrolyze proteins at both phosphotyrosine and phosphoserine/threonine residues, and recombinant DUSP5 can decrease the catalytic activity of purified ERK-1 protein in vitro (13,14); but its physiological role has not been evaluated, and it has not previously been shown to be expressed in T cells.
We next investigated whether DUSP5 could regulate IL-2induced ERK-1/2 phosphorylation using an IL-2 receptor reconstitution system (24) in 293T cells. In this setting, ERK-1/2 phosphorylation was induced by IL-2 (Fig. 5D, lane 2 versus  lane 1), and this phosphorylation was inhibited when the cells stimulation of B cells but showed variable expression in the cell lines and CLL samples. C, cytokine-inducible genes that are induced Ͼ2-fold by PI treatment of PBMCs but not by BCR cross-linking of B cells. D, genes induced Ͼ2-fold with BCR cross-linking of B cells but not induced by PI. E, genes specifically induced by IL-2, IL-4, IL-7, or IL-15. Cytokine-inducible genes were assigned to this cluster based on lack of induction by PI stimulation of PBMCs or BCR cross-linking of B cells and variable expression cell lines and CLL samples. In panels A and B the blue circles correspond to genes whose expression was also studied by Northern blotting (Fig. 4). The time-course for PI stimulation and BCR cross-linking was 0, 1, 3, 6, and 24 h. The letters below the cell lines and CLL cells correspond to following: were also transfected with wild type DUSP5 (Fig. 5D, lanes 3  and 4). Transfection of an inactive (C263S) mutant of DUSP5 did not affect ERK-1/2 phosphorylation (Fig. 5D, compare lane 6 to lane 2), which is consistent with the fact that 293T cells express very low levels of endogenous DUSP5 (data not shown). As expected, 293T cells transfected with constitutively active MEK1 showed increased ERK-1/2 activity that was independ-ent of IL-2 (Fig. 5D, lanes 7 and 8).
To further evaluate the possible role of DUSP5 in the regulation of IL-2-induced MAPK activity, we stably transfected IL-2-dependent CTLL-2 cells with wild type and inactive DUSP5. CTLL-2 is a murine T cell line that has been widely used to study IL-2 biology and signaling (see, for example, Refs. 25 and 26). We studied two clones that constitutively express FIG. 4. Northern blot analysis of selected genes. PBMCs were either not stimulated (Ϫ) or stimulated (ϩ) for 3 h with PI (lanes 1 and 2), and cultured T cells were either not stimulated (Ϫ) (lane 3) or stimulated with the indicated cytokines for 4 h (lanes 4 -7). Data are shown for one of three representative experiments.

DISCUSSION
In this study, we investigated genes regulated by IL-2, IL-4, IL-7, and IL-15. Although some data regarding genes regulated by these cytokines have been previously generated, comparative data on the effects of these cytokines on large numbers of genes have not been available. Genes induced by IL-2 have been most extensively studied, and IL-2 is known to induce a number of genes including, for example, those encoding c-Myc (15), c-Fos (15), c-Jun (15), IL-2R␣ chain (15), Pim-1 (21), Bcl-2 (15), and the SOCS family proteins SOCS1 (19) and CIS1 (20). Previous studies have sought to identify IL-2-regulated genes in a more systematic way (27)(28)(29) but have revealed largely nonoverlapping sets of IL-2-induced transcripts, suggesting that many other IL-2-regulated genes remained to be identified. In our microarray analysis137 genes appeared to be induced, and 34 genes appeared to repressed by IL-2, IL-4, IL-7, or IL-15. A significant number of these genes (20%) are related to cell proliferation based on their high level expression in proliferating cell lines but low expression in relatively quiescent CLL cells (16), in accord with the known mitogenic function of these cytokines on activated T cells.
Hierarchical clustering of these genes revealed that the induced genes fell into two major groups, i.e. those regulated preferentially by IL-2, IL-7, and IL-15, and those regulated by IL-4. It was noteworthy that IL-2, IL-7, and IL-15 induced almost identical gene expression patterns in T lymphocytes, at least at the doses we used. This supports the concept that multiple cytokines can similarly provide survival/growth signals and that the specificity of cytokine action is largely determined by cell type or developmental stage specific expression of cytokine receptor(s) and the availability of ligand(s) (30 -32). However, it is possible that cell type-dependent differences also exist. For example, it has been suggested that IL-15 uses different receptor or signaling pathways depending on cell type (33,34), which could also explain why IL-2 and IL-15 induce similar gene responses and proliferation in T cells, whereas only IL-15 is essential for NK cell differentiation.
Although IL-4 induced a set of genes overlapping those induced by IL-2, IL-7, and IL-15, there were also many differences. Most genes that were induced by IL-2, IL-7, or IL-15 were not induced by IL-4 or were induced at only a very low level. However, some genes (e.g. those endocing SOCS-1, CIS1, and Bcl-2) were induced in a similar fashion by all of the cytokines, whereas others (e.g. those encoding IL-4R␣ and Mal) were more strongly induced by IL-4. The basis for this more distinctive pattern for IL-4 may be explained by the fact that IL-4 activates primarily Stat6, whereas the other cytokines preferentially activate Stat3, Stat5a, and Stat5b (35). For example, the promoter for IL-4R␣, which is regulated by IL-4, contains Stat6 binding sites (36), whereas the IL-2R␣ promoter, which is regulated by IL-2, contains binding sites for Stat5a and Stat5b (37)(38)(39). It will be important to determine whether genes induced preferentially by IL-4, such as Mal, contribute to functions unique to IL-4 such as the induction of Th2 differentiation among T lymphocytes.
Most of the genes that we found to be induced by ␥ c -dependent cytokines were also induced by the combination of PMA plus ionomycin (74%), and many were induced after BCR stimulation of B lymphocytes (57%). One of these genes, the IL-2R␣ gene is regulated by at least five positive regulatory regions (PRRs) (15,(37)(38)(39)(40)(41). PRRI is presumably a T cell receptor response element as it contains an NF-B binding site that is required for IL-2R␣ promotor activity in response to PHA or PMA (15), whereas PRRIII and PRRIV (37)(38)(39)(40) are both required for IL-2-induced IL-2R␣ induction. A fifth element is a CD28 response element (41). Thus, in the IL-2R␣ gene, different enhancer-like elements differentially respond to different stimuli. The coexistence of antigen and cytokine response elements in other genes as well might account for the highly overlapping gene expression profiles between ␥ c -dependent cytokines and PI-stimulation. In this regard, IL-15 and T cell receptor stimulation were recently shown to induce many of the same genes in CD8 ϩ memory T cells (42).
Only a few of the genes whose expression was regulated by ␥ c -dependent cytokines were previously identified as functionally relevant target genes, such as those encoding IL-2R␣ (43)(44)(45) and Bcl-2 (46), and most of the genes we identified have not been characterized as part of a cytokine response. One such gene is DUSP5, which we now show is induced by IL-2, IL-7, and IL-15, but not by IL-4. DUSP5 was previously shown in vitro to be capable of dephosphorylating ERK-1, but the physiological context was not investigated (13,14). IL-2 signaling has been extensively studied, and along with the Jak-STAT and PI 3-kinase/Akt pathways, the MAP kinase pathway has been described as important (15). The activation of Stat5 proteins is mediated by phospho-tyrosine docking sites (Tyr-392 and Tyr-510) on the IL-2 receptor ␤ chain (47,48) and has been functionally linked to the regulation of proliferation and activation-induced cell death (48,49). Activation of the PI 3-K/Akt pathway regulates IL-2-dependent cell survival (50) and may contribute to cell proliferation (50). Regarding the MAPK pathway, Tyr-338 on IL-2R␤ directly binds the phospho-tyrosine binding (PTB) domain of Shc (48) and mediates the activation of ERK-1/2 by IL-2 (49,51,52). We provide evidence that DUSP5 induction by IL-2 may negatively regulate IL-2-dependent activation of ERK-1/2. Studies in the 32D myeloid cell line revealed that IL-2R␤ Tyr-338 is required for IL-2-dependent proliferation, suggesting that Shc-coupled MAPK activation is vital for proliferation (48). However, in primary T cells, simultaneous inactivation of Tyr-338 and Stat5 binding sites was required to reveal a decrease in proliferation, suggesting that in these cells either MAPK or Stat5-coupled pathways by themselves are sufficient for proliferation (49). This is consistent with our finding in CTLL-2 cells that DUSP5 alone did not decrease IL-2-mediated proliferation (data not shown). However, it is possible that MAPK and DUSP5 may have other effects in IL-2 biology.
It this study, we have shown that IL-2 potently activates ERK-1 and ERK-2 as well as DUSP5, which negatively regulates ERK-1 and ERK-2. In contrast, IL-4 does not potently activate either ERK-1 or ERK-2 nor does it induce DUSP5, thus providing an example of how differential gene regulation by ␥ c -dependent cytokines can modulate cytokine-specific actions. As noted above in our microarray analysis, DUSP5 was induced in T cells either following treatment with cytokines or PMA plus ionomycin. This suggests that DUSP5-dependent negative feedback regulation of MAPK is not restricted to cytokine signaling. ERK-1/2 activation has been indicated in a range of important functions in T-cells and NK cells (53). Remarkably, in ERK-1-deficient mice the only observed defect is in thymic T-cell development at the double positive stage resulting in ϳ50% reduction in the numbers of single positive lymphocytes (54). Accordingly, we recently generated mice expressing DUSP5 transgene, which in preliminary experiments show an even more complete block in T lymphocyte development with ϳ70% reduction in numbers of single positive lymphocytes (data not shown). This is consistent with our observation that DUSP5 regulates both ERK-1 and ERK-2 in T cells and suggest that DUSP5 regulation of MAPK is a general theme in T lymphocyte activation and signaling.