p38α Mitogen-Activated Protein Kinase Is Activated by CD28-Mediated Signaling and Is Required for IL-4 Production by Human CD4⁺CD45RO⁺ T Cells and Th2 Effector Cells¹

Full activation of T cells requires signaling through both the TCR/CD3 complex and the CD28 costimulatory receptor (reviewed in Refs. 1, 2). The mechanism of CD28 signaling has only been partially defined, but is thought to involve the guanine nucleotide exchange factor complex Grb-2/SOS, phosphatidylinositol 3-kinase, and inducible T cell kinase (1). Further downstream, members of the mitogen-activated protein kinase (MAPK)³ family, including the extracellular signal-regulated protein kinases (ERKs) and the stress-activated protein kinases/c-Jun NH₂-terminal kinases (SAPKs/JNKs), have been implicated in CD28-mediated signaling. CD3 signaling alone has been shown to activate ERK, and the combination of CD3 and CD28 signaling can synergistically activate JNK in T cell lines and clones (3, 4). In these studies, CD28 signaling alone was shown to minimally activate ERK and JNK (<2-fold), while others have reported much more significant ERK activation (5). Kinases known to be upstream of JNK, including p21-activated kinase and MAPK/ERK1 kinase, have been shown to be activated by CD28 signaling in Jurkat cells (6). The activities of ERK, JNK, and a third type of MAPK, p38 MAPK, are drastically reduced in anergic T cell clones (4, 7), suggesting roles for all three MAPK families in T cell activation.

p38 MAPK (p38/p38α/CSBP2/RK) was originally identified as a serine/threonine kinase activated by stimulation of monocytes with LPS and was later shown to regulate production of the proinflammatory cytokines IL-1β and TNF-α (8, 9). The p38 family also includes CSBP1, p38β, p38β2/p38-2, p38γ/ERK6/SAPK3, p38δ/SAPK4, and Mxi2 (9, 10, 11, 12, 13, 14, 15, 16, 17). In human peripheral leukocytes and lymphoid organs, p38α mRNA expression is relatively high, p38δ expression is intermediate, and expression levels of p38β and p38γ are low (17). p38β2 is 97% identical to p38β, lacking the 8-aa insert found in p38β that is unique among the MAPKs (11, 16). Mxi2 is an alternatively spliced form of p38α and is expressed at much lower levels than p38α in all tissues tested (15).

Activation of p38 occurs during many cellular responses, including those of lymphocytes, and often mirrors the activation of JNK. For example, both p38 and JNK are activated by various forms of environmental stress, IL-1β and TNF-α (18, 19, 20, 21, 22, 23), by the Ag receptor in human B cells, and by coligation of the Ag receptor and CD19 in murine splenic B cells (24, 25). Activation of p38 is also implicated in signaling during mouse thymocyte development (26), suggesting a role for p38 in T cell survival, growth, and differentiation. p38 activation has been observed in p38-transfected Jurkat cells treated with anti-CD3 mAb alone (27), in Jurkat cells and in mouse T cells treated with anti-CD3 and anti-CD28 mAbs (28, 29, 30), and in mouse T cells stimulated with Con A or with PMA and ionomycin (31). However, the role of p38 during activation of normal human T cells had not been assessed until recently, when it was reported that p38 was activated by CD28 signaling alone in human total peripheral blood T cells (32).

Many cellular functions of p38 have been defined through the use of specific p38 kinase inhibitors, the pyridinyl imidazoles. One such compound, SB 203580, blocks the activity of p38 by specifically binding to the p38 ATP-binding site (33). Three amino acid residues in the p38 ATP-binding pocket found to be important for SB 203580 binding are Thr¹⁰⁶, His¹⁰⁷, and Leu¹⁰⁸. These residues are shared by p38α, p38β, and p38β2 but not by p38γ, p38δ, or the more distantly related ERKs and JNKs (34, 35). Correspondingly, SB 203580 inhibits p38α and p38β2, partially inhibits p38β, and does not inhibit p38γ or p38δ (16) or the ERKs and JNKs (36, 37). Pyridinyl imidazoles have been shown to inhibit LPS-stimulated IL-1 and TNF-α biosynthesis in human monocytes (9, 38, 39), by a mechanism involving inhibition of mRNA translation (40, 41, 42), and to have therapeutic activity in inflammatory disease models (43). However, SB 203580 was shown to poorly block TNF-α production by a human T cell clone in vitro, suggesting an alternative mechanism for regulating TNF-α production in T cells as compared with monocytes (32). SB 203580 has also been shown to partially block IFN-γ production by mouse Th1 cells at the transcriptional level (31). Thus, p38 may be involved in the production of various cytokines and at multiple levels of biosynthesis.

CD4⁺ Th cell responses can be divided into distinct effector classes, Th1 or Th2, defined by the selective production of either IL-2 and IFN-γ, which primarily promote cell-mediated immunity (Th1), or IL-4 and IL-5, which promote IgE production and eosinophilia (Th2) (reviewed in Refs. 44, 45). Of these two classes, the development of the Th2 response is especially dependent on CD28 costimulation. In several studies, human CD4⁺ T cells stimulated in vitro without CD28 costimulation developed a Th1 phenotype, producing only IL-2 and IFN-γ, whereas the addition of an anti-CD28 Ab to the culture converted the population to a Th2 phenotype, inducing production of IL-4 and IL-5. (46, 47, 48). Injection of mice with the fusion protein CTLA-4Ig, which blocks signaling through both CD28 and CTLA-4 molecules by neutralizing their ligands B7-1 (CD80) and B7-2 (CD86), resulted in diminished IL-4 production, B cell activation, and IgE production in response to the nematode parasite Heligmosomoides polygyrus and to immunogenic anti-IgD Abs (49) and suppressed the Th2 response in mouse strains susceptible to leishmaniasis, while having no effect on the maturation of Th1 cells (50). T cells from CD28-deficient mice were able to produce normal amounts of IFN-γ, but severely reduced amounts of IL-4 and IL-5 (51). In another study, CD28 knockout mice exhibited greatly reduced IL-4, IgE, and IgG1 production in response to anti-IgD Abs, although responses to H. polygyrus were normal (52), indicating that particular Th2 responses may vary in their dependence on CD28.

Here we report the activation of p38α in response to CD28 signaling alone in highly purified human peripheral blood CD4⁺ T cells. In addition, two different p38-specific inhibitors completely blocked CD28-induced cell proliferation as well as CD3/CD28-induced IL-4 production, largely at the mRNA level. The IC₅₀ values for inhibition of these cellular responses were identical to those for inhibition of CD28-activated p38α kinase in immune complex kinase assays (20–100 nM for SB 203580), strongly suggesting that the blockade was due to inhibition of p38α and not due to an undiscovered cross-reactivity with another enzyme. Both of these p38-dependent responses were attributed to CD4⁺CD45RO⁺ Tcells. Furthermore, IL-4 production by Th2 effector cells was preferentially blocked by SB 203580. These findings identify a specific signaling pathway involved in CD28 stimulation that can lead to IL-4 production and may in part explain the observed link between CD28 signaling and Th2-type responses.

T cells were purified from human peripheral blood leukocytes (Sera-Tec Biologicals, North Brunswick, NJ) by negative selection as previously described (32). Leukocytes were shipped at 4°C by overnight courier and used within 24 h of bleeding. CD8⁺ T cells were magnetically immunodepleted using anti-CD8 Dynabeads M-450 (Dynal, Lake Success, NY) according to the manufacturer’s instructions. Unbound cells were washed and determined to be >98% CD3⁺CD4⁺CD28⁺ using a FACSort flow cytometer (Becton Dickinson, Mountain View, CA). Flow cytometry using anti-CD19, anti-HLA-DR, and anti-CD14 mAbs (Becton Dickinson) verified the absence of B cells and monocytes. Purified T cells were also determined to be free of APC by their inability to proliferate or produce IL-2 in response to CD3 cross-linking alone (Table I) and by the inability of LPS to activate p38 or elicit TNF-α production from these cells (data not shown). CD45RA⁺ and CD45RO⁺ T cell subsets were purified from the CD4⁺ T cell pool by magnetic immunodepletion using anti-CD45RO or anti-CD45RA mAbs (PharMingen, San Diego, CA), respectively, prebound to sheep anti-mouse IgG Dynabeads M-450 (Dynal). Flow cytometry indicated that the CD45RA⁺ subset contained <9% CD45RO⁺ cells, and the CD45RO⁺ cells contained <1% CD45RA⁺ cells.

Purified T cells (5–7.5 × 10⁶ per sample) were cultured overnight in 1 ml AIM-V medium (Life Technologies, Rockville, MD), pelleted, and resuspended in RPMI 1640 (Sigma, St. Louis, MO) with or without the p38 inhibitor SB 203580 (CalBiochem, La Jolla, CA) at 1 μM. For SB 203580 titration, all samples contained a final DMSO concentration of 0.1%. Cells were incubated for 45 min at 37°C/5% CO₂, then for 15 min on ice. Mouse IgG1κ and IgG2aκ isotype controls (PharMingen), anti-CD3ε mAb OKT3 (isotype IgG2aκ, Ortho Pharmaceutical, Raritan, NJ), anti-CD28 mAbs CD28.2 (isotype IgG1κ; PharMingen) or ANC28.1 (isotype IgG1κ; Ancell, Bayport, MN), and anti-ICAM-1 mAb HA58 (isotype IgG1κ; PharMingen) were added to cells at a final concentration of 10 μg/ml for each mAb and incubated on ice for 15 min. Cells were pelleted and resuspended in warm RPMI 1640 containing 30 μg/ml F(ab′)₂ goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA) and incubated in a 37°C water bath for 10 min unless otherwise indicated. Hydrogen peroxide (Sigma) was used at a final concentration of 500 μM in PBS for 30 min at 37°C. Cell lysis, immune complex kinase assays, and Western blotting were performed as previously described (32). Briefly, cells were lysed in 20 mM HEPES, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 1 mM Na₃VO₄ 1× EDTA-free complete protease inhibitor mixture (Boehringer Mannheim, Indianapolis, IN). p38α was immunoprecipitated with anti-p38α polyclonal rabbit Ab C-20 (Santa Cruz Biotechnology, Santa Cruz, CA), and its activity was measured using kinase-inactive GST-MAPK activated protein kinase-2 (MAPKAPK-2; Upstate Biotechnology, Lake Placid, NY) as substrate. MAPKAPK-2 was immunoprecipitated using anti-MAPKAPK-2 polyclonal sheep Ab (Upstate), and its kinase activity measured using recombinant human heat shock protein 27 (Hsp27; StressGen, Victoria, British Columbia, Canada) as a substrate. Western blotting of p38α was performed using the anti-p38α polyclonal rabbit Ab C-20.

T cells were plated in duplicate in flat-bottom 96-well tissue culture plates in complete medium (RPMI 1640 containing 10% FCS, 50 U/ml penicillin G, 50 μg/ml streptomycin, and 2 mM glutamine) at 2 × 10⁵ cells/well. SB 203580 (CalBiochem) or RWJ 67657 (R. W. Johnson Pharmaceutical Research Institute) was serially diluted in complete medium, at a constant final DMSO concentration, and added to cells for a 1-h pretreatment at 37°C 5% CO₂ for a final volume of 0.2 ml/well. OKT3 (1 μg/ml), IgG2aκ (1 μg/ml), CD28.2 (10 μg/ml), or IgG1κ (10 μg/ml) were added, maintaining a total IgG concentration in all samples at 11 μg/ml, followed by F(ab′)₂ goat anti-mouse IgG (30 μg/ml). Cells were cultured at 37°C 5% CO₂ for 3 days unless otherwise stated, and supernatants were harvested (0.1 ml/well). Cytokine levels were measured using ELISAs for IL-2 and TNF-α (Genzyme, Cambridge, MA), IL-4 and IL-5 (PharMingen), and IFN-γ (Endogen, Woburn, MA). ELISA plates were read on a VMax kinetic microplate reader (Molecular Dynamics, Sunnyvale, CA). Cytokine levels were reported as the mean ± SD from duplicate samples, as calculated by SoftMax microplate reader software (Molecular Dynamics). After supernatants were harvested, proliferation was measured by culturing cells with [³H]thymidine (1 μCi/well) (Amersham) for 18 h from day 3 to day 4 unless otherwise stated. [³H]-labeled cells were harvested onto filter mats and counted in a 1205 Betaplate liquid scintillation counter (Wallac, Gaithersburg, MD). Cell proliferation levels were reported as the mean ± SD cpm values, as calculated by the 1205 Betaplate reader software (Wallac).

Purified CD4⁺CD45RO⁺ T cells (1 × 10⁷ per sample) were pretreated in complete medium with either 0.01% DMSO, 1 μM SB 203580, or 1 μM RWJ 67657 for 1 h, then stimulated with either isotype control mAbs IgG2aκ (1 μg/ml) and IgG1κ (10 μg/ml) or OKT3 (1 μg/ml) and CD28.2 (10 μg/ml) followed by the addition of F(ab′)₂ goat anti-mouse IgG (30 μg/ml). Three days later, supernatants were harvested for IL-4 ELISA, and RNA was prepared using the Qiashredder spin column homogenizer and the RNeasy total RNA purification kit (Qiagen, Chatsworth, CA). Reverse transcription was performed on total RNA from each sample using the cDNA cycle kit (Invitrogen, Carlsbad, CA). PCR was performed on total cDNA from each sample using Taq DNA polymerase (PCR reagent system; Life Technologies) and the human IL-4 amplimer and ribosomal protein S9 control primer pairs, using conditions recommended by the manufacturer (Clontech, Palo Alto, CA). Thermal cycles were performed on a GeneAmp PCR System 9600 (Perkin-Elmer, Norwalk, CT). IL-4 PCR product was quantitated by competitive PCR using the human IL-4 mimic template (Clontech), which is a DNA fragment containing the same 5′ and 3′ PCR primer annealing sequences as the IL-4 target cDNA. The mimic was titrated in 2-fold serial dilutions, from 0.35 to 0.01 amol/μg cDNA, against a constant amount of cDNA (500 ng) for each reaction, and the resulting PCR products were separated on a 2% agarose gel. Ratios of mimic PCR product to IL-4 PCR product were quantitated by scanning laser densitometry of photographs of the ethidium bromide-stained gels on a Storm 840 PhosphorImager System. Data were plotted as the PCR product ratio vs mimic concentration, and linear regression was performed to calculate the mimic concentration at which the PCR product ratio is equal to one. At this point, the initial mimic concentration was equivalent to the initial IL-4 cDNA concentration. Errors were estimated using 95% confidence limits on the linear regression.

Polarization of human T cells to the Th1 or Th2 phenotype was performed according to a modification of the procedure described by Sallusto et al. (53). Adult human PBMC were stimulated at 3 × 10⁶/ml in complete medium with 1 μg/ml PHA-L (Sigma), 100 U/ml recombinant human IL-2 (Endogen, Woburn, MA), and either 10 ng/ml recombinant human IL-12 (R&D Systems, Minneapolis, MN) and 200 ng/ml anti-human IL-4 mAb (PharMingen) for Th1 conditions or 10 ng/ml recombinant human IL-4 (Endogen) and 2 μg/ml anti-human IL-12 mAb (R&D Systems) for Th2 conditions. This stimulation was repeated on days 7 and 14. On days 3, 10, and 17, fresh complete medium was added to double the culture volume. Cells were harvested on day 18, and CD4⁺ T cells were purified as described above.

Purified Th1 or Th2 cells (1 × 10⁶/sample) were stained with 1 μg anti-CD3-FITC and either anti-CD28-PE or mouse IgG2a-PE isotype control mAb (Becton Dickinson) in PBS/5% FCS on ice for 30 min, washed, and stained with 1 μg/ml propidium iodide (Sigma). CD28 expression was analyzed using a FACSsort (Becton Dickinson), gating on CD3-positive, propidium iodide-negative events.

To address the involvement of p38α in primary human CD4⁺ T cell activation, we isolated CD4⁺ T cells from human peripheral blood by magnetic immunodepletion, stimulated them with anti-CD3 and anti-CD28 mAbs alone or in combination, and measured p38α activity using an immune complex kinase assay. While anti-CD3ε mAb (OKT3) alone had a small effect on p38α activity, anti-CD28 mAb (CD28.2) alone caused a large increase in p38α activity (Fig. 1,A). Another anti-CD28 mAb (ANC28.1) also caused activation of p38α, indicating that the ability to stimulate p38α through CD28 was not unique to one mAb. The combination of OKT3 and CD28.2 activated p38α, but to a lesser degree than CD28.2 mAb alone (Fig. 1,A, lane 5). This reduction in p38α activity in the presence of anti-CD3 Ab was observed in all experiments. This reduction did not occur in the presence of an isotype control mAb (IgG2aκ) (Fig. 1,A, lane 6), suggesting that the reduction in p38α activity was specifically due to CD3 cross-linking. A control mAb of the same isotype as CD28.2 and ANC28.1, but specific for ICAM-1, did not activate p38α (Fig. 1,A, lane 7). SB 203580 (1 μM), a specific p38 inhibitor, dramatically reduced the p38α activity in cells stimulated via CD3 and CD28 (Fig. 1,A, lane 8). The level of CD28-mediated p38α activation was comparable to that stimulated by hydrogen peroxide (Fig. 1 A, lane 9), a known inducer of p38α activity (17). CD28.2 mAb was determined to be free of LPS contamination by testing in PBMC culture for stimulation of TNF-α production. Additionally, polymyxin B sulfate (which binds to and inactivates LPS) did not block anti-CD28-induced p38α activation at concentrations that completely blocked optimal LPS-induced TNF-α production in PBMC cultures (data not shown), indicating that the observed CD28-mediated p38α activation was not due to contamination by endotoxin.

MAPKAPK-2 is an in vivo substrate of p38 that phosphorylates Hsp27 (22, 54) and is a known substrate for p38α in vitro (17). Therefore, activation of the p38α pathway was further assessed by an immune complex kinase assay using MAPKAPK-2 precipitated from T cell lysates. The same conditions that activated p38α also activated MAPKAPK-2 in the cells (Fig. 1 B), providing further evidence for p38α activation in T cells following CD28 signaling.

CD28-mediated p38α activation peaked at 10 min, with a return to baseline by 20 min (Fig. 2,A), similar to the kinetics of p38 and JNK activation in other systems (10, 55). The p38 inhibitor SB 203580 was a potent inhibitor of CD28-induced p38α activity, with an IC₅₀ of 20–100 nM (Fig. 2 B), identical to its IC₅₀ for inhibition of LPS-induced TNF-α production from human PBMC (56).

Although full activation of T cells in general requires both TCR/CD3 and CD28 signals, certain peripheral blood T cells can respond to CD28 cross-linking alone. For example, CD28 signaling alone can induce proliferation of CD4⁺CD45RO⁺ T cells (57), memory cytotoxic T cell activity (58), and increased expression of the HIV-1 tat gene in T cells from HIV-1 infected individuals (59). To examine the role of p38 in the biological responses of CD4⁺ peripheral blood T cells to anti-CD3 and anti-CD28 mAbs individually or together, proliferation and cytokine production were monitored. Cross-linking CD3 alone induced no cell proliferation and stimulated very little cytokine production (Table I), with the exception of IFN-γ. An equivalent amount of IFN-γ was also produced by cells treated with isotype control mAbs alone, but not by untreated cells, indicating that the CD4⁺ T cells were not constitutively producing IFN-γ, but rather responded nonspecifically to Ab (data not shown). Cross-linking CD28 alone produced relatively low levels of cytokine. However, cross-linking CD28 did cause significant cell proliferation, ∼30% the level seen with anti-CD3 and CD28 mAbs together (Table I), in agreement with previous reports (57). SB 203580 blocked this CD28-induced cell proliferation in a dose-dependent fashion (IC₅₀ = 10–80 nM), yet did not affect proliferation stimulated by CD3 plus CD28 (Fig. 3,A). The dose-response curve for inhibition of p38α kinase activity (Fig. 2 B) paralleled that observed for inhibition of CD28-induced proliferation, suggesting that the inhibition of proliferation was due to inhibition of p38α activity and not due to inhibition of an unknown enzyme.

Because TCR/CD3-independent proliferation has previously been attributed to the CD4⁺CD45RO⁺ (memory) T cell population (57), the CD4⁺ T cell pool was divided into naive CD45RA⁺ and memory CD45RO⁺ T cells by magnetic immunodepletion. The subsets were pretreated with titrated amounts of SB 203580, then stimulated by CD28 cross-linking alone. As expected, the CD45RO⁺ proliferative response was 5- to 8-fold higher than that of the CD45RA⁺ cells (Fig. 3,B). However, CD28 stimulation activated p38α to a similar extent in the CD45RA⁺ and CD45RO⁺ cells, and SB 203580 inhibited p38 to a similar degree in both populations (Fig. 3,C), indicating that p38α activation alone was insufficient to cause cell proliferation. To substantiate that SB 203580 blocked CD28-induced proliferation due to inhibition of p38, CD4⁺CD45RO⁺ T cells were pretreated with SB 203580 or a different p38 inhibitor, RWJ 67657 (S. A. Beers, E. A. Malloy, W. Wu, M. P. Wachter, D. Cavender, P. Lalan, S. Wadsworth, and J. Siekierka, manuscript in preparation). Like SB 203580, RWJ 67657 inhibits p38α and p38β, but not p38γ, p38δ, other MAPKs, lck, or Itk, and is 10-fold more potent than SB 203580 for inhibition of TNF-α production by PBMC in response to LPS. CD28-induced proliferation was again inhibited in a dose-dependent fashion (IC₅₀ = 0.5–4 nM), with the degree of inhibition reflecting the difference in potency between the two inhibitors (Fig. 3 D), again suggesting that the block in proliferation was due to inhibition of p38α.

To examine the effects of p38 inhibition on cytokine production, CD4⁺ T cells were pretreated with SB 203580 before stimulation with anti-CD3 and anti-CD28 mAbs. IL-4 production was completely blocked by SB 203580 (IC₅₀ = 20–100 nM), whereas IL-2 production was unaffected (Fig. 4,A). Similar results were obtained using purified T cells stimulated with staphylococcal enterotoxin B and anti-CD28 mAb (data not shown), indicating that the results were not dependent on the use of anti-CD3 mAb as a TCR/CD3 signaling agent. To determine whether the production of other cytokines was suppressed, IFN-γ, IL-5, and TNF-α levels were analyzed. IFN-γ, IL-5, and TNF-α were all inhibited by SB 203580, though to a lesser degree than IL-4 (Fig. 4,B). Inhibition of IFN-γ, IL-5, and TNF-α was incomplete even at the highest dose of compound, and the data did not fit the classical sigmoidal shape of a simple single-site inhibition curve. Thus, of the cytokines examined, IL-4 production was the most dependent on p38 activity. The dose-response for inhibition of IL-4 production was comparable to that for inhibition of p38α kinase activity (Fig. 2 B), suggesting that the effect of SB 203580 was specific for p38α and not due to inhibition of other signaling mechanisms.

Memory CD45RO⁺ T cells have previously been shown to produce larger amounts of IL-4 than naive CD45RA⁺ T cells (60, 61). Similarly, in our system, the CD45RO⁺ cell subset produced 10-fold more IL-4 than the CD45RA⁺ subset (Fig. 4,C). Production of IL-4 by CD45RO⁺ T cells was inhibited by either SB 203580 or the 10-fold more potent p38 inhibitor RWJ 67657 (Fig. 4 D), confirming the p38-dependence of this response.

In human monocytes, p38 inhibitors block IL-1 and TNF-α biosynthesis (38), without affecting levels of IL-1 and TNF-α mRNA (40). p38 also controls VCAM-1 expression in endothelial cells in a posttranscriptional manner, without affecting mRNA accumulation (62). However, in other systems p38 controls gene expression at the mRNA level. For example, a p38 inhibitor prevents production of inducible NO synthase by mouse astrocytes by reducing the accumulation of inducible NO synthase mRNA (63). To determine whether p38 regulates IL-4 mRNA levels, CD4⁺CD45RO⁺ T cells were stimulated with anti-CD3 and anti-CD28 mAbs in the presence or absence of SB 203580 or RWJ 67657. After 3 days, total cellular RNA was purified, reverse transcribed into cDNA, and used in an IL-4-specific PCR amplification. IL-4 message was not detectable in cDNA from unstimulated cells, but was detected in cDNA from cells stimulated with anti-CD3 and anti-CD28 mAbs (Fig. 5,A, lane 2). Treatment of the cells with SB 203580 or RWJ 67657 reduced the level of IL-4 cDNA detected, but had no effect on the level of ribosomal protein S9 cDNA (Fig. 5,A, lanes 3–4). When a fragment of mimic DNA (containing the same primer-annealing sites as IL-4 cDNA but including extra nucleotides to yield a PCR product larger than that derived from IL-4 cDNA) was added to the IL-4 amplification as an internal standard for the PCR, a difference in relative IL-4 cDNA levels between SB 203580- and RWJ 67657-treated cells became apparent (Fig. 5,A, lanes 5–8). Titration of the mimic DNA into the IL-4 PCR allowed quantitative analysis of the IL-4 cDNA levels from stimulated cells treated with DMSO, SB 203580, or RWJ 67657 (Fig. 5, B and C), as described in Materials and Methods. The attomoles of IL-4 mimic DNA required to yield a band equivalent to the IL-4 cDNA band was used to calculate the amount of IL-4 message present in each cell sample (Fig. 5,D). Treatment of CD4⁺CD45RO⁺ T cells with 1 μM SB 203580 resulted in a 66% reduction of IL-4 mRNA levels, while treatment with 1 μM RWJ 67657 resulted in an 86% reduction (Fig. 5,D). This suggests that p38 regulates either IL-4 transcription or mRNA stability. A small but consistent difference in the degree of reduction in IL-4 protein levels, as compared with IL-4 mRNA levels, was observed (Fig. 5 D), indicating a possible role for p38 inhibitors in the blocking of IL-4 translation. Furthermore, the p38 inhibitors reduced IL-4 mRNA levels in a manner consistent with their rank order potency for inhibition of IL-4 protein production, again suggesting that the observed effect was due to inhibition of p38α rather than an unknown cross-reactivity with another enzyme.

To determine whether the observed block in IL-4 production by the treatment of freshly isolated peripheral human CD4⁺ T cells with p38 inhibitor was due to inhibition of IL-4 production by T cells already primed for IL-4 expression, rather than inhibition of differentiation toward the Th2 phenotype in vitro, T cells were polarized toward the Th1 or Th2 phenotype in vitro in the absence of SB 203580. Adult human PBMC were cultured for 18 days with PHA and IL-2, plus exogenous IL-12 and neutralizing anti-IL-4 mAb for Th1 conditions or exogenous IL-4 and neutralizing anti-IL-12 mAb for Th2 conditions. CD4⁺ T cells were then purified and stimulated with anti-CD3 and anti-CD28 mAbs. Following Th1 polarization, T cells produced high levels of IFN-γ and IL-2, very low amounts of IL-4, and no detectable IL-5 (Table II). Conversely, T cells grown in Th2-promoting conditions produced large amounts of IL-4 and IL-5, a low amount of IFN-γ, and no detectable IL-2. This cytokine expression profile is comparable to those previously reported for human Th1 and Th2 cell lines (53, 64). Anti-CD28 mAb was essential for stimulating high levels of cytokine production (Table II), indicating that under these conditions, full activation of Th1 and Th2 cells was dependent upon CD28 costimulation.

In both Th1 and Th2 cells, cross-linking CD28, but not CD3, resulted in p38α activation (Fig. 6,A). The degree of p38α activation was greater in Th1 cells than in Th2 cells (6.5-fold vs 2.6-fold). However, flow cytometry revealed a higher CD28 expression level in Th1 cells than in Th2 cells, with median fluorescence values of 1382 and 813 for Th1 and Th2 cells, respectively (Fig. 6,B). To examine the p38 dependence of Th1 and Th2 cytokine production, Th1 and Th2 effector cells were treated with SB 203580 during stimulation with anti-CD3 and anti-CD28 mAbs. In Th2 cells, SB 203580 strongly inhibited IL-4 production (81% inhibition at 1 μM), while IL-5 production and cell proliferation were not affected (Fig. 6,C). In Th1 cells, SB 203580 partially inhibited production of IL-2 (39% inhibition at 1 μM) and IFN-γ (48% inhibition at 1 μM), while cell proliferation was not affected (Fig. 6 D). Therefore, among the cytokines analyzed, IL-4 produced by Th2 cells was the most p38 dependent.

Our results indicate that signaling through the CD28 molecule can activate p38α, in the absence of signaling through the TCR/CD3 complex, in human peripheral blood CD4⁺ T cells. CD28-induced p38α activation was previously reported for total human peripheral blood T cells (32). We now show using two different highly specific p38 inhibitors that p38 activity appears to be required for both the proliferation of CD4⁺CD45RO⁺ T cells induced by CD28 stimulation alone, as well as IL-4 production by CD4⁺CD45RO⁺ cells or Th2 effector cells induced by CD3 plus CD28 signaling. The identical IC₅₀ values for inhibition of these cellular responses and the inhibition of p38α kinase activity in immune complex kinase assays strongly suggests that the cellular effects were due to inhibition of p38α and were not due to an unknown cross-reactivity with another enzyme. Although SB 203580 is capable of inhibiting p38β and p38β2 activity, it is more potent against p38α than against these other isoforms (16). Expression levels of p38β and p38β2 are very low in peripheral blood T cells, as measured by immunoprecipitation and Western blot (data not shown), consistent with the low level of p38β mRNA expression in peripheral leukocytes observed by northern blot (17). Thus, although a role for a p38 isoform other than p38α in the cellular responses described here cannot be entirely ruled out, it is most likely that the observed effects of p38 inhibitors were due to blockade of p38α activity.

The inability of SB 203580 to inhibit IL-2 production or proliferation of cells stimulated via CD3 plus CD28 signaling indicates that p38 is not required for all cellular responses, but is selectively involved in pathways leading to the production of particular cytokines, especially IL-4. The partial blockade of IFN-γ production in freshly isolated CD4⁺ T cells, as well as in differentiated Th1 cells, suggests that p38 is also required for maximal IFN-γ expression. The degree of IFN-γ inhibition observed here is in agreement with a previous study, which found that SB 203580 partially inhibited IFN-γ production by mouse Th1 cells (60% inhibition at 1 μM) (31). However, in contrast to our findings, the previous study showed that IL-4 production by mouse Th2 cells (stimulated with Con A) was not inhibited by SB 203580. This difference is most likely due to the fact that the human Th2 effectors required CD28 costimulation for IL-4 production (Table II), but the mouse Th2 effector cells did not (31). We have observed that IL-4 production by the mouse Th2 clone D10.G4.1 is independent of CD28 costimulation, and is insensitive to p38 inhibitors (data not shown). Therefore, p38 appears to be important for CD28-dependent, but not CD28-independent, IL-4 production. We have also found that although mouse and human p38α kinase activity were equivalently sensitive to inhibition by SB 203580, IL-4 production by mouse splenic T cells was ∼20-fold less sensitive than human T cell IL-4 production to inhibition by this compound (data not shown). These data strongly suggest that mouse IL-4 production is much less dependent on the p38α signaling pathway than human IL-4 production.

The substantial activation of p38α by CD28 stimulation alone and the lack of activation by CD3 stimulation observed in peripheral blood CD4⁺ T cells differs from the requirements for p38 activation found in other systems. For example, in p38-transfected Jurkat cells, p38 activation was observed with anti-CD3 mAb alone (27). In mouse Th1 clones, p38 was activated in response to CD3 stimulation alone and activation was not increased by CD28 costimulation (29). Another study, using mouse splenic and lymph node T cells previously stimulated with anti-CD3 plus IL-2 in vitro, demonstrated that p38 could be activated by CD3 stimulation and augmented by CD28 costimulation (28). Interestingly, we consistently observed that concomitant stimulation of CD3 and CD28 resulted in lower p38α activation than CD28 signaling alone (Fig. 1 A). This reduction was not observed when staphylococcal enterotoxin B was used instead of anti-CD3 (32), indicating that the decrease may have been due to reduced CD28 cross-linking in the presence of CD3 Abs. Alternatively, this decrease in p38α activity may have been due to ERK-induced MAPK phosphatase activation (65) or to different kinetics of p38α activation following CD3/CD28 costimulation as compared with CD28 stimulation alone. Ultimately, the difference between our results and previous reports may be due to the different nature of signaling in primary peripheral blood T cells as compared with T cell lines or clones. For example, we have previously demonstrated that the human T cell clone HA-1.70, Jurkat cells, and human peripheral blood T cells have differing CD3 and CD28 signaling requirements for p38α activation (32). We found that clone HA-1.70 expressed high levels of both B7-1 and B7-2, Jurkat cells expressed high levels of B7-1, and peripheral blood T cells expressed neither. The presence of CD28 ligands on T cell lines and clones might elevate basal p38 activity, thereby resulting in a lower apparent p38 response to additional CD28 signaling. High constitutive p38 activity has been reported for freshly isolated mouse thymocytes, as well as mouse splenic and lymph node T cells, and it has been suggested that this is due either to signals delivered in tissue microenvironments in vivo or due to stress signals generated during the death of the mouse or during the mechanical dissociation of the organs into cell suspensions (26, 28). The use of human peripheral blood T cells in our study, coupled with our overnight culture to allow recovery from the stress of the isolation procedure, avoids these complications and may be preferable for studies of stress-activated kinases in T cells.

The block in production of IL-4, but not IL-2, by SB 203580 in freshly isolated total CD4⁺ T cells (Fig. 3,A), and the block in Th2 cell IL-4 production, but only partial inhibition of IL-2 production by polarized Th1 effector cells (Fig. 6,D), indicates that multiple MAPKs control distinct downstream T cell effector functions. IL-2 gene expression is controlled in part by the transcription factors NF-AT and AP-1, which are regulated by both ERK- and JNK-dependent mechanisms (3, 4, 66). The IL-4 gene promoter also contains NF-AT and AP-1 sites, but differs from the IL-2 promoter in that it can be activated by calcium mobilization alone (67, 68, 69). It has also been shown that IL-4 gene transcription is controlled by the transcription factor NF-ATc and the Th2-specific transcription factors c-Maf and GATA-3 (70, 71, 72, 73). Whether transcription factors such as these lie downstream of p38 and are responsible for the p38-dependent IL-4 transcription described here is under investigation. The SB 203580-mediated increase in IL-2 production by Th1 cells under certain conditions (Fig. 6 D) has also been observed in other related experiments (data not shown). Although we cannot explain this phenomenon at present, it is consistent with a recent report that p38 is involved in the nuclear export of NF-ATp, a transcription factor more closely associated with IL-2 production and Th1 development than with IL-4 production and Th2 development (74). The partial inhibition of IL-5, TNF-α, and IFN-γ production by SB 203580 may be due to joint regulation of these cytokines by p38 and other MAPKs such as JNK. Indeed, IFN-γ production was found to be impaired in both p38-dominant negative transgenic mice (31) and JNK2-deficient mice (75).

p38 may also be able to regulate IL-4 production at the level of mRNA translation, because p38 inhibitors caused a reduction in IL-4 mRNA levels that was not as great as that for IL-4 protein (Fig. 5). In monocytes, inhibition of TNF-α production by p38 inhibitors appears to occur at the translational level, mediated through an AUUUA repeat motif in the 3′-UTR of TNF-α mRNA (42). Because this AUUUA motif is also present in the 3′-UTR of IL-4 mRNA, a similar mechanism may control IL-4 biosynthesis. However, this sequence is also found in the mRNA of IL-2 and other cytokines, so translational control may be dependent on additional factors.

The inhibition of IL-4 production by p38 inhibitors in our study was not secondary to inhibition of IL-1β production or IL-1β signaling, because monocytes and macrophages, the producers of IL-1β, were shown to be absent from the cultures by the failure of anti-CD3 mAb to elicit IL-2 production in the absence of costimulation (Table I), by the lack of TNF-α production in response to LPS, and by flow cytometry (data not shown). IL-2 and IL-7 have also been shown to activate p38 in a cytokine-dependent mouse T cell line (37), and we therefore assessed their role in our system. p38 inhibitors did not block CD3/CD28-dependent cell proliferation (Fig. 3), IL-2 production (Fig. 4), or expression of either the low-affinity IL-2R CD25 or the high-affinity IL-2R CD122 (data not shown), thereby ruling out an essential function for p38 in IL-2R signaling. IL-7 was undetectable by both ELISA and RT-PCR, there are no reports of T cell production of IL-7 in the literature, and the addition of neutralizing anti-IL-7 Abs did not affect IL-4 production (data not shown). Collectively, these data indicate that the inhibition of IL-4 production by SB 203580 was not due to inhibition of IL-7 receptor signaling.

Our data show that two responses of CD4⁺CD45RO⁺ memory T cells, IL-4 production and CD28-induced cell proliferation, are highly p38 dependent. Importantly, this demonstrates that the observed p38 activation is functionally relevant. It has previously been reported that IL-4 is preferentially produced by CD4⁺CD45RO⁺ rather than CD4⁺CD45RA⁺ T cells in response to stimulation via CD3 plus CD28 (60), or to specific Ag (61). CD28-induced proliferation, which may play a role in Ag-independent T cell activation and expansion, occurs in response to the anti-CD28 Ab BW828 (57). These responses have been reproduced here and shown to be p38-dependent processes through the use of p38 inhibitors. Interestingly, although CD45RA⁺ cells did not respond functionally in the same manner as the CD45RO⁺ population, p38 was, in fact, activated by CD28 stimulation in these cells (Fig. 3 C). We have not yet identified a p38-dependent functional response in these cells.

The activation of p38 by CD28 signaling alone and its role in IL-4 production suggest an important role for p38 in the responses of CD4⁺CD45RO⁺ T cells and differentiated Th2 cells. Given the central role of IL-4 in promoting IgE synthesis by B cells and its association with atopic states (76), the observed inhibition of IL-4 production suggests that p38 inhibitors may be useful as therapeutic agents for allergy or other conditions where it is desirable to modulate this cytokine.

We thank M. Fahmy for excellent technical assistance, P. Peterson, A. August, D. Cavender, J. Davis, and C. Harris, for critically reviewing the manuscript, and M. Wachter, S. Beers, and W. Wu for synthesis of RWJ 67657.