Polar opposites: Erk direction of CD4 T cell subsets^{^{<a href="#FN2" rid="FN2" class=" fn">1</a>}}

Introduction

The differentiation of T cells in response to infectious agents constitutes an essential step in effective immune responses. Minimally, antigen stimulated T cells can take on one of four programs of cytokine production and effector function: IFNγ-secreting Th1 cells effective against viruses and intracellular microbes; IL-4-secreting Th2 cells most effective against helminths; IL-17-secreting Th17 cells effective against extracellular microbes, and induced T regulatory (iTreg) cells that limit inflammation and autoimmunity (1). These differentiation programs are engaged in response to cytokines and the conditions of antigen presentation, but the T cell intrinsic signaling pathways that are operative are not entirely understood.

The signaling pathway from membrane-bound receptors to the activation of extracellular signal-regulated kinase (Erk) is found in every cell type throughout metazoan species. Erk interacts with many substrates and regulators to modulate signal transduction in the cytoplasm, and to program gene expression in the nucleus (2, 3). It constitutes a highly connected node in the matrix of cell signaling. The effects include cell-cycle progression, lineage specification and cellular differentiation, survival, chemotaxis, learning and memory, and the elaboration of cell-type specific functions.

The two Erk isoforms appear to act additively in cell cycle control (4), although a loss of Erk1 (Mpk3) function reveals few phenotypic changes, whereas even a hemizygous loss of Erk2 (Mpk1) causes embryonic lethality in some strains of mice (5–8). The origin of the Erk2 loss-of-function lethality was due in part to the failure of the polar trophectoderm cells to proliferate (6). In several other physiological models there are differential requirements for Erk1 and Erk2 that do not correlate with dramatic differences in the relative amounts of Erk1 and Erk2 expressed (9).

The strength of signal through the TCR has been found to influence the outcome of differentiation such that a strong and prolonged Erk signal gave rise to Th1 cells, whereas weak antigen stimulation or an attenuation of Erk gave rise to IL-2-dependent Stat5 phosphorylation, Gata3 expression, and IL-4 production (10, 11). Alternatively, an inhibitor of the Erk kinases, Mek1,2, was found to enhance Foxp3 expression and Treg cell differentiation (12, 13). Here we used genetic ablation of Erk1 or Erk2 to study the T cell intrinsic role of this pathway in proliferation, expansion, and differentiation of four types of CD4 effector T cells. We find that Erk2 is central to the contingencies governing the differentiation of Th1 and iTreg cells, and this is reflected in its effects on the subset-characteristic programs of gene expression. In particular, we show that a loss of Erk2 caused a large-scale increase in gene expression specifically under conditions of TGFβ signaling.

Materials and Methods

Mice and Viral Infection

CreER^, Erk1^, Erk2 ^{allele mice were previously described (}8, 14, 15). Deletion of loxP-flanked alleles in Erk2CreER ^{mice was induced by i.p. injection of 2 mg tamoxifen} q.d. (Sigma) for 6 consecutive days, and mice were analyzed 2 days later. Where indicated, mice were infected i.p. with 2×10 ^{PFU lymphocytic choriomeningitis virus (LCMV-Armstrong). Bone marrow chimeras were made by reconstitution of lethally irradiated (10 Gy)} Rag1 ^{mice with T cell-depleted bone marrow cells from wildtype (CD45.1) or} Erk2CreER ^{(CD45.2) mice separately or mixed at a 1:1 ratio. Mice were analyzed 8 weeks after reconstitution. For the Treg cell induction,} Erk2CreER ^{mice were crossed with Smarta transgenic mice (TCR transgenes specific for H2A bound with LCMV glycopeptide 61–80) (}16). Animal work was performed according to UCSD guidelines.

Cell culture

CD4 T cells (purity >90%) from lymph nodes and spleens were MACS purified (Miltenyi Biotec) by either positive selection for CD4 or by depletion using biotin-labeled B220, CD8, CD25, CD69, DX5, and MHCII with streptavidin microbeads. T cell proliferation was stimulated with plate-bound anti-CD3 (5 ng/ml) plus or minus anti-CD28 added to culture (1 μg/ml) as described (17). Where indicated, IL-2 was added at 10 U/ml, PMA at 20 ng/ml, and ionomycin at 100 ng/ml. For T cell differentiation, cells were stimulated with anti-CD3 and anti-CD28 plus: for Th1, IL-12 (10 ng/ml), IL-2 (50 U/ml) and anti-IL-4; for Th2, IL-4 (10 ng/ml) IL-2 and anti-IFNγ; and for Th17, IL-6, TGFβ, anti-IFNγ and anti-IL-4. To generate iTregs, CD4 cells were stimulated with anti-CD3, anti-CD28, TGFβ, IL-2 (10 U/ml) in the presence or absence of anti-IL-4 and anti-IFNγ. Cytokines were measured in the supernatant by ELISA (eBioscience).

Flow cytometry

Lymphocytes were explanted and stained with antibodies specific for CD62L, CD44, CD4 and CD8. For the intracellular staining of Foxp3, cells were fixed and permeabilized with Foxp3 buffer set (eBioscience). For Erk2 or phos-Smad2,3 staining, cells were fixed with 2% p-formaldehyde and permeabilized with Perm III buffer (BD Biosciences). Cells were stained with Erk2 antibody (Santa Cruz) or phos-Smad2,3 (Cell Signaling Technology) followed goat anti-rabbit IgG-PE (SouthernBiotech). For intracellular cytokines, cells were fixed and permeabilized with the BD Cytofix/CytopermSolution Kit (BD Biosciences) and stained with for IFNγ, IL-4, or IL-17 (eBiosciences). MHC Class II, H2-A ^{tetramers loaded LCMV gp66–77 or human CLIP peptide 103–117 were provided by the NIH tetramer facility. Samples were collected on a FACS Calibur or LSRII (BD Biosciences) and analyzed by FlowJo software (Tree Star). For CFSE analysis, cultures were resuspended in equal volumes and analyzed for a constant period of time. The numbers on the ordinate of histograms indicate the number of T cells per interval of intensity, where the area under the curve equals the total number of T cells collected.}

Regulatory T cell assays

In vitro regulatory T cells were assayed as previously described (18). To generate Foxp3 ^{T cells} in vivo, 2 × 10 ^{naïve Smarta CD4 T cells (CD45.2) were transferred into B6 mice (CD45.1). The next day, recipient mice were immunized with 10 μg glycoprotein peptide (gp61–80) or sterile PBS. Five days later, spleens were harvested and analyzed for Foxp3 expression by flow cytometry.}

Inflammatory bowel disease (IBD)

Naïve CD4 T cells (CD25 ^{CD45RB) and regulatory T cells (CD25CD45RB) were sorted on a FACS Aria (BD Biosciences). Naïve T cells (3 × 10) were transferred into} Rag1 ^{mice alone or together with 2 × 10 wildtype (WT) or} Erk2 ^{regulatory T cells (}19). Mice were weighed weekly and euthanized if their body weight decreased by more than 20%.

SDS-PAGE and Western blotting

SDS-PAGE and Western blotting was carried out as described (18). The primary antibodies are used as follows: anti-Erk1, 2 (Santa Cruz), anti-phos-Smad2, anti-phos-Rsk3, anti-phos-Mnk1 (Cell Signaling Technology), anti-ZAP70 (BD Biosciences), anti-Bim (Sigma-Aldrich).

RNA isolation and quantitative real-time PCR

PCR and qPCR was carried out as described (18). Samples were normalized to Gapdh or Ppia (cyclophilin A). Primers used are available upon request.

Microarray analysis

Total RNA from wildtype and Erk2 deficient CD4 T cells under indicated conditions were purified with an RNeasy kit (Qiagen). Samples were processed by the BIOGEM core, UC San Diego using MouseRef-8 beadchip kit (Illumina). Interchip data were normalized by UC San Diego microarray core and further analyzed with Excel, Genepattern software suite (20), and TM4 microarray software suite (21). The data are available through GEO (http://www.ncbi.nlm.nih.gov/geo/), accession number: {"type":"entrez-geo","attrs":{"text":"GSE37554","term_id":"37554"}}GSE37554.

Statistics

Prism 4.0c software (Graphpad software; San Diego, CA) was used for Student’s T-test analyses. p values are indicated in the figure legends.

Results

Erk2 deficient CD4 T cells from Erk2 ^{CreER mice}

We previously analyzed the effects of Erk2 deletion on the function of CD8 ^{T cells, using a distal Lck promoter-Cre transgene (dLck-Cre); however, this transgene caused deletion in only 80% of CD4 T cells (}22). To analyze the role of Erk2 signaling in peripheral CD4 T cells, we crossed Erk2 ^{mice to} CreER ^{mice, and the deletion of the} loxP-flanked exons of Erk2 was induced by tamoxifen (Fig. 1A) (15). By PCR, western blot, and flow cytometry the deletion of Erk2 was uniform within the population and virtually complete. Tamoxifen treated Erk2CreER ^{mice are referred to as} Erk2^. Erk2 deletion led to reduced thymic cellularity, preferentially affecting the CD4^{CD8 population (}8); however, Erk2 ^{mice displayed similar proportions and numbers of CD4 and CD8 T cells within the secondary lymphoid tissues (}Supplemental Fig. 1A). Similarly, naïve and effector/memory cell populations based on CD44 and CD62L expression were also unchanged (Supplemental Fig. 1B).

FIGURE 1

Erk2 ^{T cells require co-stimulation for cell cycle progression. (}A) Purified CD4 T cells from lymph nodes and spleen were analyzed for the presence of Erk2 by PCR, Western blotting and flow cytometry. (B) Purified WT or Erk2 ^{CD4 T cells were stimulated} in vitro. CFSE profiles were analyzed for CD4 ^{T cells. (}C) T cells (WT, Erk2^{) were analyzed for CD25 and CD44 expression on day 3 following the stimulation as indicated. Similar results were found on day 4. The data are representative of three or more independent experiments.}

Erk2 is required for Th1, but not for Th2 or Th17 development in vitro

Since T cells can proliferate in the absence of Erk2, we were able to examine the role Erk1 or Erk2 in Th1, Th2 or Th17 differentiation. There were no differences observed comparing WT and Erk1 ^{cells when activated under the different T helper (Th1, Th2, Th17) conditions (}Supplemental Fig. 2B, 2C, 2D) (23). In contrast, Erk2 ^{CD4 T cells displayed normal proliferation but a survival defect when activated under Th1 and Th2 polarization conditions (}Fig. 2A). This survival defect may be the result of increased levels of the pro-apoptotic protein Bim that were observed in Erk2 ^{cells relative to WT (}Fig. 2B).

FIGURE 2

Erk2 ^{CD4 T cells display impaired survival and polarization to the Th1 subset. (}A) CD4 T cells were cultured under different conditions for 3 days. CFSE profiles were analyzed as in Fig. 1 (representative of 3 experiments). (B) Purified CD4 T cells from WT and Erk2 ^{mice were cultured for Th1 polarization. Cell lysates were analyzed by western blotting, and the amounts of BimEL and BimL/S were normalized to ZAP70 (representative of 3 separate experiments). (}C) Cells were cultured in polarizing conditions, and CD4 ^{T cells shown (representative of 3 experiments). (}D) The percentage of IFNγ ^{T cells and the amount of IFNγ from cultures of re-stimulated Th1 cells shown (mean ± SEM from 3 experiments; *, p<0.05; **, p<0.001). (}E, F) Percentage of IL-4 and IL-17-producing cells (mean ± SEM for 3 experiments). (G) T cells were analyzed for T-bet, Gata-3, phospho-Stat1 (pStat1) and phospho-Stat5 (pStat5) at day 3. Isotype (mouse IgG1) control was used for T-bet and Gata-3 staining; goat anti-rabbit PE was used for the phospho-Stat staining control. Representative of three experiments.

We further examined whether the loss of Erk2 affected the differentiation of Th1, Th2 and Th17 cells as measured by the intracellular production of IFNγ, IL-4 or IL-17 (Fig. 2C). The proportion of IFNγ-producing Erk2-deficient CD4 T cells and amount of IFNγ per cell was reduced when compared to WT and there was a 4-fold decrease in IFNγ levels measured in the culture supernatant (Fig. 2D). Despite this defect, T-bet and phospho-Stat1 were induced in Erk2 ^{T cells to an extent equivalent to WT T cells (}Fig. 2G). In addition, under Th1 conditions, Erk2 ^{CD4 T cells exhibited higher Gata-3 expression, consistent with increased IL-4 independent Th2 differentiation found under conditions of Erk attenuation (}10, 11). In contrast, the amount of phospho-Stat5 was reduced (Fig. 2G). Under optimal, polarizing Th2 conditions, there was a trend toward increased Th2 differentiation, but the difference did not reach significance given the number of trials (Fig. 2C, 2E). There was no difference in Th17 differentiation in the absence of Erk2 (Fig. 2C, 2F).

Impaired Th1-driven viral response in Erk2-deficient mice

To examine differentiation of Th1 cells in vivo, we infected WT, Erk1 ^or Erk2 ^{mice with LCMV Armstrong and assessed the number of IFNγ producing T cells at day 8 post-infection.} Erk1 ^{mice displayed similar CD4 responses as WT mice, whereas there was a substantial reduction in the proportion and the absolute cell number of IFNγ producing cells in} Erk2 ^{mice (data not shown). To determine whether this effect was T cell intrinsic,} Rag1 ^{mice were reconstituted with bone marrow cells from WT, or} Erk2CreER ^{mice, or an equal mixture of both. Eight weeks post-reconstitution, the mice were treated with tamoxifen and challenged with LCMV. The number of antigen-specific CD4 T cells, as measured by MHC Class II gp66–77 tetramers or anti-viral IFNγ producing Th1 cells, was dramatically reduced in the mice that were reconstituted with} Erk2CreER ^{cells (}Fig. 3A, B). Even within the mixed bone marrow chimeras, the population of antigen-specific Erk2 ^{T cells was reduced compared to WT T cells in the same animals. We note that the number of tetramer positive cells was approximately equivalent to the number of IFNγ T cells, and thus, the absence of Erk2 did not redirect differentiation, but it either diminished Th1 differentiation or survival.}

FIGURE 3

Impaired anti-viral Th1 responses in absence of Erk2. (A, B) Single or mixed bone marrow chimeras were treated with tamoxifen for six days and then rested 2 days followed by infection with LCMV. (A) CD4 ^{T cells were analyzed by staining for CD44 and H2A tetramer bound with gp66–77 or CLIP peptide (control). (}B) CD4 ^{T cells were stained for CD44 and intracellular IFNγ following re-stimulation with the LCMV peptides gp61–80 and NP309–327. Data were plotted as mean ± SEM of antigen-specific CD4 T cells (n=3 per group) present in spleen for a single experiment (*, p<0.05; **, p<0.001). Data shown are representative of three independent experiments carried out on three different sets of mice.}

Increased iTreg differentiation in absence of Erk2

The role of Erk in the differentiation of induced regulatory T cells (iTreg) in culture was examined. Erk2 ^{but not} Erk1 ^{T cells generated increased proportions and numbers of CD25Foxp3 iTreg cells (}Fig. 4A, B, and data not shown). iTreg cells were also induced in vivo by transferring naïve Smarta CD4 T cells (depleted of CD4^{CD25 cells) and immunizing mice with LCMV gp61–80. Under these conditions there was an increased proportion of} Erk2 ^{Smarta T cells that converted to Foxp3 cells (}Fig. 4C). In addition, there was an increase in the proportion of Foxp3 ^{T cells (nTregs) present in lymphoid organs of} Erk2Cd4Cre mice (data not shown).

FIGURE 4

Enhanced differentiation and normal function of induced Treg in absence of Erk2. (A, B) Purified CD4 T cells were stimulated under iTreg conditions for 5 days and analyzed for CD4, CD25 and Foxp3 expression. (A) Representative profiles. (B) Accumulated results from all four experiments (mean ± SEM; **, p<0.001). (C) Induction of Smarta transgenic iTreg cells in vivo. Representative flow cytometry profiles (gated on CD4 ^{and CD45.2 cells) with the mean percentages of Foxp3 within the donor Smarta T cells graphed. Data were accumulated from two independent experiments (PBS n=8, gp peptide n=10; **, p<0.001). (}D) WT Teff cells (CD4^{CD25) were co-cultured with WT or} Erk2 ^{Treg cells (CD4CD25), and proliferation measured by the incorporation of H-TdR. The data are representative of three independent experiments. (}E) Weight loss associated with IBD. The data are accumulated from three independent experiments (total mice: WT eff n=14, WT eff +WT Treg n=11, WT eff + Erk2 ^{Treg n=9).}

To determine whether Erk2 ^{Treg cells were functional, we employed cell culture and} in vivo measures of Treg activity. As shown, CD4^{CD25 cells from} Erk2 ^{mice were at least as effective as WT Treg cells in their ability to inhibit T cell proliferation (}Fig. 4D). Erk2 ^{Treg cells also suppressed weight loss in an inflammatory bowel disease model to the same extent as WT Treg cells (}Fig. 4E) (24). We can conclude that Erk2 ^{T cells more readily differentiate into iTreg cells, and} Erk2 ^{nTreg cells exhibit at least equivalent function when compared to those from WT mice.}

Erk2 dependence of Dnmt expression and Smad signaling

The regulation of Foxp3 depends importantly on methylation such that, in the absence of Dnmt1, Foxp3 is efficiently expressed in activated CD4 and CD8 T cells (25–27). We thus considered the possibility that enhanced iTreg induction in Erk2 ^{T cells was the result of reduced} Dnmt1 expression. Analysis by qPCR indicated that Dnmt1 expression is progressively induced with time, and the induction is partially Erk2 dependent (Fig 5A) at 2 d in culture. As shown below, Foxp3 is induced by 1d, and its expression is enhanced in Erk2 ^{cells. Thus, simple regulation of} Dnmt1 mRNA does not appear to explain the TGFβ-induced Foxp3 induction.

FIGURE 5

Erk2 regulates DNMT and phosphorylation of Smad2. (A) WT or Erk2 ^{cells from individual mice were stimulated under iTreg conditions for 0–5 days. The mRNA samples were analyzed by real-time quantitative PCR for} Dnmt1 in duplicate and normalized to the Ppia gene (±SEM, WT n=3, Erk2 ^{n=3). The data are representative of two separate experiments (*, p<0.05; **, p<0.01). (}B) WT or Erk2 ^{naïve CD4 T cells were stimulated under iTreg conditions and analyzed for pSmad2,3 expression by flow cytometry (representative of 3 experiments). (}C) Purified CD4 T cells from WT and Erk2 ^{mice were stimulated under iTreg conditions for 30 minutes. Cell lysates were analyzed by western blotting at the indicated time points. ZAP70 was used as loading control. Plots show the phos-Smad2 density normalized to Smad2 or ZAP70. Relative expression level was normalized by WT day 0 expression, as 1. The data plotted in the bar graphs were accumulated from three independent experiments (*, p<0.05).}

Another potential mode of regulation is direct inhibition of TGFβ signaling through the inhibitory phosphorylation of Smad2 and Smad3 at their linker regions as shown by the analysis of Mekk2,3 deficient T cells (28). As Mekk2,3 are upstream of all the Map Kinases, the possibility exists that Erk constitutes the primary mediator of this effect. Smads are activated by TGFβRI-mediated phosphorylation at the carboxyterminal SXS motif, and this in turn causes nuclear localization and downstream gene activation. As a means of detecting activated Smad2 and Smad3, we probed the amount of SXS phosphorylation of Smad2/3 both by intracellular fluorescence staining and immunoblotting. The results showed an increased induction of pSmad2 in Erk2 deficient cells compared to WT by 30 minutes, and this difference was enhanced at 2–3 days after stimulation (Fig. 5B, 5C). We conclude that inactivation or removal of Erk2 promotes greater Smad signaling, and this in turn, enhances Treg cell development.

iTreg cell gene expression is suppressed by Erk2

To further understand the role of Erk2 in T cell differentiation, we performed microarray analyses of WT and Erk2 ^{T cells in 8 different conditions. Basic analyses revealed that the data set reflects Erk2-dependent, T cell specific mRNA expression under multiple stimulatory and polarizing conditions (}Supplemental Fig. 3A). To identify a set of genes dependent on Erk2 for expression, we averaged the ratios (WT:Erk2^{) across all 8 conditions. There were 49 Erk2-dependent probes (mean ratio greater 1.5), and the highest 14 are shown as a correlated heat map (}Supplemental Fig. 3B). These include: 3/3 Mapk1 (Erk2); 2/2 Dusp6—an Erk dual-specific phosphatase (29); Egr2—a direct Erk target; Ccl1—encoding a ligand of Ccr8; Gpr68-a G-coupled proton sensor, and Tnf. We note the alternating pattern of Erk2 dependence. In contrast, there were only 10 Erk2-suppressed probes (mean ratio of less than 0.67) (Supplemental Fig. 3C). These include: Ccr8; Ebi2—an oxysterol-specific nuclear receptor controlling T and B cell migration (30); Gata3; and Klf2—a Foxo1 target controlling T cell homing (31). We conclude from these analyses that the data set reflects Erk2-dependent T cell specific mRNA expression under multiple stimulatory conditions.

Scatter plots comparing Erk2 ^{to WT values under 1d Th0 or Th1 conditions identified a small number of probes for which} Erk2 ^{was greater than 1.5 or less than 0.67 compared with WT: Th0, 71 probes (45 Erk2-suppressed, 26 Erk2-dependent) and Th1, 77 probes (42 Erk2-suppressed, 35 Erk2-dependent). By contrast, the same comparison under 1d iTreg conditions identified 643 probes, more than an eight-fold increase over the number identified under Th0 or Th1 conditions (544 Erk2-suppressed, 99 Erk2-dependent) (}Fig. 6A). Using Venn diagrams, we graphed the number of overlapping or unique probes that were Erk2 suppressed or Erk2 dependent under each of the polarizing conditions. The Erk2-suppressed changes in gene expression specific to iTreg cells, 503 probes, constituted by far the largest category (Fig. 6B).

FIGURE 6

Microarray analysis of gene expression in WT and Erk2 ^{CD4 T cells. (}A–D) RNA samples were isolated from WT or Erk2 ^{(KO) T cells cultured as follows: naïve (0d); Th0 (CD3, CD28, IL-2) 3h; iTreg (Th0, TGFβ) 3h; Th0 1d; Th1 (Th0+IL-12+anti-IL-4) 1d; iTreg 1d; Th0 1d + TGFβ 1h; Th0 1d + TGFβ 3h. (}A) The values Erk2 ^{vs. WT for each probe are presented as scatter plots for Th0, Th1 and iTreg conditions at 1d. Numbers indicate genes with a difference in expression of 1.5-fold or more (Erk2 suppressed, top left, red dots) or 0.67 fold or less (Erk2-dependent, bottom right, blue dots). (}B) The numbers of Erk2-dependent or Erk2-suppressed genes from the scatter analysis (A) are shown in Venn diagrams to indicate the overlap between each category. (C) The values of probes for Dnmt1 and Dnmt3b were normalized to the median for each probe and charted for all 8 conditions indicated on the figure alternating between WT and KO. (D) The Erk2-suppressed iTreg genes were identified and the expression values were normalized to the maximum for each probe, ordered by hierarchical clustering (Pearson Correlation) and depicted as a correlated heat map. The criteria were: iTreg-inducible, iTreg 1d Erk2 ^{> (1.5) 0d} Erk2^{; TGFβ-dependent, iTreg 1d} Erk2 ^{> (1.5) Th0 1d} Erk2^{; not Th1, iTreg 1d} Erk2 ^{> (1.5) Th1 1d} Erk2^{; and Erk2-suppressed, iTreg 1d} Erk2 ^{> (1.3) iTreg 1d WT. The criteria were chosen for biological relevance and to identify a manageable set of genes. Asterisks indicate the number of probes identified for each gene if greater than 1.}

This large-scale increase in gene expression at 1d in Erk2 ^{T cells could be consistent with a hypomethylated genome as a result of reduced} Dnmt1; however, the qPCR data showed no Erk2-dependent difference in Dnmt1 expression at this time-point (Fig. 5A). The array data showed that, of the Dnmt isoforms queried, only Dnmt1 and Dnmt3b were expressed over background, and neither was affected by the addition of TGFβ (Fig. 6C)—equivalent induction was seen at 1 d under Th0, Th1, and iTreg conditions of stimulation. Moreover, the effects of Erk2 deletion on Dnmt1 and Dnmt3b expression were small compared with the magnitude of induction. We conclude that Erk2 signaling constitutes an important suppressor of gene expression in the presence of TGFβ, although the extent to which this results from Dnmt1 regulation is unknown (discussed below).

The iTreg cell dependence on Erk2 thus provided a tool to more finely characterize a program of iTreg-specific genes. We identified those genes that were induced in 1d iTreg conditions, but not under Th0 or Th1 conditions, and Erk2 suppressed (Fig. 6D). This algorithm identified 45 unique probes, including all three probes specific for Foxp3, and in addition, both probes specific for the Treg-specific neuropilin-1 (Nrp1). Gpr83, Ecm1, Cmtm7, Nkg7, Socs2 and Glrx were previously found to be iTreg cell specific (32, 32) and of these, we also identified Gpr83 and Nkg7. Socs2 was strongly (>10-fold) induced in Th0, Th1, and iTreg conditions (see below), and thus not iTreg specific (Fig. 7D). Ecm1 was induced at 3h but not d1. Cmtm7 and Glrx were relatively unchanged across all conditions (data not shown).

FIGURE 7

T cell subset array analyses. (A) The array values for Tbx21, Gata3, Rorc, and Foxp3 were charted. See legend to Fig. 6. (B) The array values for signature cytokines, Ifng, Il4, Il10, Tgfb1, and Il17a were charted. (C–E) Correlated heat maps of subset-genes were identified and as described legend to Fig. 6. (C) Th0, Erk2-dependent probes: inducible, Th0 1d WT > (2) 0d WT; Erk2-dependent, Th0 1d WT > (1.5) Th0 Erk2 ^{1d. (}D) Th0, Erk2-suppressed probes: inducible, Th0 1d WT > (2) 0d WT; Erk2-suppressed, Th0 1d Erk2 ^{> (1.3) Th0 1d WT. (}E) Th2-like probes: Erk2-suppressed, Th0 1d Erk2 ^{> (1.5) Th0 1d WT; Not Th1, Th0 1d} Erk2 ^{> (3) Th1 1d WT. (}F) Th1 selected probes: inducible, Th0 1d WT > (3) 0d WT and Th1 1d WT > (3) 0d WT; Th1 not iTreg, Th1 1d WT > (2) iTreg 1d Erk2^{; Early-onset Erk2-dependent, Th0 3h WT > (1.3) Th0 3h} Erk2^.

Other genes identified in this subset that may be involved in immune function included: Dscr1l2 that inhibits calcineurin-dependent transcriptional responses; Ctsw (cathepsin W) which functions in T cell cytolytic activity; Faim3, a Caspase-8 inhibitor; Rcan3, which binds and inhibits calcineurin A; Tnfrsf1b (TnfRII); and Tnfrsf9 (4-1BB) (Fig. 6D). We also analyzed gene expression at 3 h post activation, with or without the addition of TGFβ. The rationale was that gene expression directly affected by TGFβ signaling would be induced early, yet none of the probes specific to iTreg induction at 1d were already induced at 3h in the presence of TGFβ. All of the probes induced at 3h with TGFβ were also induced in its absence under Th0 conditions (Fig. 6D). Likewise, we examined whether stimulation with TGFβ for 1h or 3 h after 1d Th0 activation would induce a subset of the signature genes, and we identified 5 of the genes listed in Fig. 6D. These included Ccr8, Ctsw, Dscr1l2, Emp1, and Edod1. Foxp3 and Nrp1 were not induced under these conditions. We conclude that the iTreg cell program of gene expression becomes elaborated in naïve T cells between 3h and 1d post TGFβ stimulation.

A separate analysis of the array data based on an unbiased hierarchical clustering of the entire data set (Pearson Corrlelation, Average linkage) was carried out and a heat-map of a cluster containing Foxp3 was selected (Supplemental Fig. 4A). Although there is a large overlap in the genes identified using the algorithm described above (e. g., Nrp1), there were also unique genes identified. Notably, this list includes Ccr6, encoding a chemokine receptor that plays a role in inflammatory bowel disease (33), and three transcription factors not previously known to play a role in T reg cell differentiation: Cux1, Stat5, and Zfhx3 discussed below.

Erk2 dependent cell survival

Since there was a marked loss of T cell survival correlated with increased Bim expression in Erk2 deficient T cells under Th1 conditions (Fig. 2B), we plotted the Bcl-2 family members, and found five that were highly regulated under the conditions tested (Fig. 8). Bcl2 and BclX_L (Bcl2l1) were mirror images with respect to T cell activation and Erk dependence, with Bcl2 being weakly Erk2 suppressed and BclX_L being weakly Erk2 dependent. Proapoptotic Bim was moderately Erk2 suppressed, consistent with the data presented in Fig. 2B and loss of survival under Th1 conditions. Interestingly, we found the prosurvival genes, Bcl2a1c and Bcl2a1d, to have expression patterns almost identical to that of Bim. The regulation of mitochondrial death in T cells in different phases of activation and differentiation is clearly complex.

FIGURE 8

The expression of Bcl-2 family members

The median normalized array values for Bcl2, Bcl2l1 (BclxL), Bcl2a1c, Bcl2a1d, and Bim were charted.

Erk2 deficient CD4 T cells from Erk2 ^{CreER mice}

We previously analyzed the effects of Erk2 deletion on the function of CD8 ^{T cells, using a distal Lck promoter-Cre transgene (dLck-Cre); however, this transgene caused deletion in only 80% of CD4 T cells (}22). To analyze the role of Erk2 signaling in peripheral CD4 T cells, we crossed Erk2 ^{mice to} CreER ^{mice, and the deletion of the} loxP-flanked exons of Erk2 was induced by tamoxifen (Fig. 1A) (15). By PCR, western blot, and flow cytometry the deletion of Erk2 was uniform within the population and virtually complete. Tamoxifen treated Erk2CreER ^{mice are referred to as} Erk2^. Erk2 deletion led to reduced thymic cellularity, preferentially affecting the CD4^{CD8 population (}8); however, Erk2 ^{mice displayed similar proportions and numbers of CD4 and CD8 T cells within the secondary lymphoid tissues (}Supplemental Fig. 1A). Similarly, naïve and effector/memory cell populations based on CD44 and CD62L expression were also unchanged (Supplemental Fig. 1B).

FIGURE 1

Erk2 ^{T cells require co-stimulation for cell cycle progression. (}A) Purified CD4 T cells from lymph nodes and spleen were analyzed for the presence of Erk2 by PCR, Western blotting and flow cytometry. (B) Purified WT or Erk2 ^{CD4 T cells were stimulated} in vitro. CFSE profiles were analyzed for CD4 ^{T cells. (}C) T cells (WT, Erk2^{) were analyzed for CD25 and CD44 expression on day 3 following the stimulation as indicated. Similar results were found on day 4. The data are representative of three or more independent experiments.}

Co-stimulation can replace the requirement for Erk2 in T cell proliferation

Consistent with our analysis of CD8 T cells (22), Erk2 ^{CD4 T cells, stimulated with anti-CD3 in the absence of added co-stimulation, accumulated at the undivided stage (}Fig. 1B) and displayed incomplete induction of CD25 and CD44 (Fig. 1C). Higher concentrations of anti-CD3 did not induce proliferation (data not shown). However, the defect in Erk2 ^{CD4 T cell proliferation was rescued by the addition of a CD28-mediated co-stimulatory signal (}Fig. 1B, 1C). In fact, Erk2 ^{T cells appeared to undergo more rounds of division when compared with WT T cells, and one explanation is suggested by the Erk2-dependence of the cyclin-dependent kinase inhibitor, p21 as shown below. Contrary to our expectations, similar results were obtained using T cells deleted for both} Erk1 and Erk2 (data not shown). These results imply that a signaling pathway downstream of the co-stimulatory receptor CD28 can replace a requirement for Erk activation in TCR-mediated cell cycle progression.

Similar to CD8 T cells, proliferation and survival were only partially rescued by the addition of IL-2 (Supplemental Fig. 1C) (22). However, Erk2 ^{deficient CD4 T cells proliferated to the same extent as WT T cells upon PMA and ionomycin stimulation (}Supplemental Fig. 1D). Finally, in marked contrast to the results with Erk2 ^{T cells, there was no effect of} Erk1 deletion on CD4 ^{T cell proliferation in response to TCR-mediated stimulation, with or without co-stimulation (}Supplemental Fig. 2A).

Erk2 is required for Th1, but not for Th2 or Th17 development in vitro

Since T cells can proliferate in the absence of Erk2, we were able to examine the role Erk1 or Erk2 in Th1, Th2 or Th17 differentiation. There were no differences observed comparing WT and Erk1 ^{cells when activated under the different T helper (Th1, Th2, Th17) conditions (}Supplemental Fig. 2B, 2C, 2D) (23). In contrast, Erk2 ^{CD4 T cells displayed normal proliferation but a survival defect when activated under Th1 and Th2 polarization conditions (}Fig. 2A). This survival defect may be the result of increased levels of the pro-apoptotic protein Bim that were observed in Erk2 ^{cells relative to WT (}Fig. 2B).

FIGURE 2

Erk2 ^{CD4 T cells display impaired survival and polarization to the Th1 subset. (}A) CD4 T cells were cultured under different conditions for 3 days. CFSE profiles were analyzed as in Fig. 1 (representative of 3 experiments). (B) Purified CD4 T cells from WT and Erk2 ^{mice were cultured for Th1 polarization. Cell lysates were analyzed by western blotting, and the amounts of BimEL and BimL/S were normalized to ZAP70 (representative of 3 separate experiments). (}C) Cells were cultured in polarizing conditions, and CD4 ^{T cells shown (representative of 3 experiments). (}D) The percentage of IFNγ ^{T cells and the amount of IFNγ from cultures of re-stimulated Th1 cells shown (mean ± SEM from 3 experiments; *, p<0.05; **, p<0.001). (}E, F) Percentage of IL-4 and IL-17-producing cells (mean ± SEM for 3 experiments). (G) T cells were analyzed for T-bet, Gata-3, phospho-Stat1 (pStat1) and phospho-Stat5 (pStat5) at day 3. Isotype (mouse IgG1) control was used for T-bet and Gata-3 staining; goat anti-rabbit PE was used for the phospho-Stat staining control. Representative of three experiments.

We further examined whether the loss of Erk2 affected the differentiation of Th1, Th2 and Th17 cells as measured by the intracellular production of IFNγ, IL-4 or IL-17 (Fig. 2C). The proportion of IFNγ-producing Erk2-deficient CD4 T cells and amount of IFNγ per cell was reduced when compared to WT and there was a 4-fold decrease in IFNγ levels measured in the culture supernatant (Fig. 2D). Despite this defect, T-bet and phospho-Stat1 were induced in Erk2 ^{T cells to an extent equivalent to WT T cells (}Fig. 2G). In addition, under Th1 conditions, Erk2 ^{CD4 T cells exhibited higher Gata-3 expression, consistent with increased IL-4 independent Th2 differentiation found under conditions of Erk attenuation (}10, 11). In contrast, the amount of phospho-Stat5 was reduced (Fig. 2G). Under optimal, polarizing Th2 conditions, there was a trend toward increased Th2 differentiation, but the difference did not reach significance given the number of trials (Fig. 2C, 2E). There was no difference in Th17 differentiation in the absence of Erk2 (Fig. 2C, 2F).

Impaired Th1-driven viral response in Erk2-deficient mice

To examine differentiation of Th1 cells in vivo, we infected WT, Erk1 ^or Erk2 ^{mice with LCMV Armstrong and assessed the number of IFNγ producing T cells at day 8 post-infection.} Erk1 ^{mice displayed similar CD4 responses as WT mice, whereas there was a substantial reduction in the proportion and the absolute cell number of IFNγ producing cells in} Erk2 ^{mice (data not shown). To determine whether this effect was T cell intrinsic,} Rag1 ^{mice were reconstituted with bone marrow cells from WT, or} Erk2CreER ^{mice, or an equal mixture of both. Eight weeks post-reconstitution, the mice were treated with tamoxifen and challenged with LCMV. The number of antigen-specific CD4 T cells, as measured by MHC Class II gp66–77 tetramers or anti-viral IFNγ producing Th1 cells, was dramatically reduced in the mice that were reconstituted with} Erk2CreER ^{cells (}Fig. 3A, B). Even within the mixed bone marrow chimeras, the population of antigen-specific Erk2 ^{T cells was reduced compared to WT T cells in the same animals. We note that the number of tetramer positive cells was approximately equivalent to the number of IFNγ T cells, and thus, the absence of Erk2 did not redirect differentiation, but it either diminished Th1 differentiation or survival.}

FIGURE 3

Impaired anti-viral Th1 responses in absence of Erk2. (A, B) Single or mixed bone marrow chimeras were treated with tamoxifen for six days and then rested 2 days followed by infection with LCMV. (A) CD4 ^{T cells were analyzed by staining for CD44 and H2A tetramer bound with gp66–77 or CLIP peptide (control). (}B) CD4 ^{T cells were stained for CD44 and intracellular IFNγ following re-stimulation with the LCMV peptides gp61–80 and NP309–327. Data were plotted as mean ± SEM of antigen-specific CD4 T cells (n=3 per group) present in spleen for a single experiment (*, p<0.05; **, p<0.001). Data shown are representative of three independent experiments carried out on three different sets of mice.}

Increased iTreg differentiation in absence of Erk2

The role of Erk in the differentiation of induced regulatory T cells (iTreg) in culture was examined. Erk2 ^{but not} Erk1 ^{T cells generated increased proportions and numbers of CD25Foxp3 iTreg cells (}Fig. 4A, B, and data not shown). iTreg cells were also induced in vivo by transferring naïve Smarta CD4 T cells (depleted of CD4^{CD25 cells) and immunizing mice with LCMV gp61–80. Under these conditions there was an increased proportion of} Erk2 ^{Smarta T cells that converted to Foxp3 cells (}Fig. 4C). In addition, there was an increase in the proportion of Foxp3 ^{T cells (nTregs) present in lymphoid organs of} Erk2Cd4Cre mice (data not shown).

FIGURE 4

Enhanced differentiation and normal function of induced Treg in absence of Erk2. (A, B) Purified CD4 T cells were stimulated under iTreg conditions for 5 days and analyzed for CD4, CD25 and Foxp3 expression. (A) Representative profiles. (B) Accumulated results from all four experiments (mean ± SEM; **, p<0.001). (C) Induction of Smarta transgenic iTreg cells in vivo. Representative flow cytometry profiles (gated on CD4 ^{and CD45.2 cells) with the mean percentages of Foxp3 within the donor Smarta T cells graphed. Data were accumulated from two independent experiments (PBS n=8, gp peptide n=10; **, p<0.001). (}D) WT Teff cells (CD4^{CD25) were co-cultured with WT or} Erk2 ^{Treg cells (CD4CD25), and proliferation measured by the incorporation of H-TdR. The data are representative of three independent experiments. (}E) Weight loss associated with IBD. The data are accumulated from three independent experiments (total mice: WT eff n=14, WT eff +WT Treg n=11, WT eff + Erk2 ^{Treg n=9).}

To determine whether Erk2 ^{Treg cells were functional, we employed cell culture and} in vivo measures of Treg activity. As shown, CD4^{CD25 cells from} Erk2 ^{mice were at least as effective as WT Treg cells in their ability to inhibit T cell proliferation (}Fig. 4D). Erk2 ^{Treg cells also suppressed weight loss in an inflammatory bowel disease model to the same extent as WT Treg cells (}Fig. 4E) (24). We can conclude that Erk2 ^{T cells more readily differentiate into iTreg cells, and} Erk2 ^{nTreg cells exhibit at least equivalent function when compared to those from WT mice.}

Erk2 dependence of Dnmt expression and Smad signaling

The regulation of Foxp3 depends importantly on methylation such that, in the absence of Dnmt1, Foxp3 is efficiently expressed in activated CD4 and CD8 T cells (25–27). We thus considered the possibility that enhanced iTreg induction in Erk2 ^{T cells was the result of reduced} Dnmt1 expression. Analysis by qPCR indicated that Dnmt1 expression is progressively induced with time, and the induction is partially Erk2 dependent (Fig 5A) at 2 d in culture. As shown below, Foxp3 is induced by 1d, and its expression is enhanced in Erk2 ^{cells. Thus, simple regulation of} Dnmt1 mRNA does not appear to explain the TGFβ-induced Foxp3 induction.

FIGURE 5

Erk2 regulates DNMT and phosphorylation of Smad2. (A) WT or Erk2 ^{cells from individual mice were stimulated under iTreg conditions for 0–5 days. The mRNA samples were analyzed by real-time quantitative PCR for} Dnmt1 in duplicate and normalized to the Ppia gene (±SEM, WT n=3, Erk2 ^{n=3). The data are representative of two separate experiments (*, p<0.05; **, p<0.01). (}B) WT or Erk2 ^{naïve CD4 T cells were stimulated under iTreg conditions and analyzed for pSmad2,3 expression by flow cytometry (representative of 3 experiments). (}C) Purified CD4 T cells from WT and Erk2 ^{mice were stimulated under iTreg conditions for 30 minutes. Cell lysates were analyzed by western blotting at the indicated time points. ZAP70 was used as loading control. Plots show the phos-Smad2 density normalized to Smad2 or ZAP70. Relative expression level was normalized by WT day 0 expression, as 1. The data plotted in the bar graphs were accumulated from three independent experiments (*, p<0.05).}

Another potential mode of regulation is direct inhibition of TGFβ signaling through the inhibitory phosphorylation of Smad2 and Smad3 at their linker regions as shown by the analysis of Mekk2,3 deficient T cells (28). As Mekk2,3 are upstream of all the Map Kinases, the possibility exists that Erk constitutes the primary mediator of this effect. Smads are activated by TGFβRI-mediated phosphorylation at the carboxyterminal SXS motif, and this in turn causes nuclear localization and downstream gene activation. As a means of detecting activated Smad2 and Smad3, we probed the amount of SXS phosphorylation of Smad2/3 both by intracellular fluorescence staining and immunoblotting. The results showed an increased induction of pSmad2 in Erk2 deficient cells compared to WT by 30 minutes, and this difference was enhanced at 2–3 days after stimulation (Fig. 5B, 5C). We conclude that inactivation or removal of Erk2 promotes greater Smad signaling, and this in turn, enhances Treg cell development.

iTreg cell gene expression is suppressed by Erk2

To further understand the role of Erk2 in T cell differentiation, we performed microarray analyses of WT and Erk2 ^{T cells in 8 different conditions. Basic analyses revealed that the data set reflects Erk2-dependent, T cell specific mRNA expression under multiple stimulatory and polarizing conditions (}Supplemental Fig. 3A). To identify a set of genes dependent on Erk2 for expression, we averaged the ratios (WT:Erk2^{) across all 8 conditions. There were 49 Erk2-dependent probes (mean ratio greater 1.5), and the highest 14 are shown as a correlated heat map (}Supplemental Fig. 3B). These include: 3/3 Mapk1 (Erk2); 2/2 Dusp6—an Erk dual-specific phosphatase (29); Egr2—a direct Erk target; Ccl1—encoding a ligand of Ccr8; Gpr68-a G-coupled proton sensor, and Tnf. We note the alternating pattern of Erk2 dependence. In contrast, there were only 10 Erk2-suppressed probes (mean ratio of less than 0.67) (Supplemental Fig. 3C). These include: Ccr8; Ebi2—an oxysterol-specific nuclear receptor controlling T and B cell migration (30); Gata3; and Klf2—a Foxo1 target controlling T cell homing (31). We conclude from these analyses that the data set reflects Erk2-dependent T cell specific mRNA expression under multiple stimulatory conditions.

Scatter plots comparing Erk2 ^{to WT values under 1d Th0 or Th1 conditions identified a small number of probes for which} Erk2 ^{was greater than 1.5 or less than 0.67 compared with WT: Th0, 71 probes (45 Erk2-suppressed, 26 Erk2-dependent) and Th1, 77 probes (42 Erk2-suppressed, 35 Erk2-dependent). By contrast, the same comparison under 1d iTreg conditions identified 643 probes, more than an eight-fold increase over the number identified under Th0 or Th1 conditions (544 Erk2-suppressed, 99 Erk2-dependent) (}Fig. 6A). Using Venn diagrams, we graphed the number of overlapping or unique probes that were Erk2 suppressed or Erk2 dependent under each of the polarizing conditions. The Erk2-suppressed changes in gene expression specific to iTreg cells, 503 probes, constituted by far the largest category (Fig. 6B).

FIGURE 6

Microarray analysis of gene expression in WT and Erk2 ^{CD4 T cells. (}A–D) RNA samples were isolated from WT or Erk2 ^{(KO) T cells cultured as follows: naïve (0d); Th0 (CD3, CD28, IL-2) 3h; iTreg (Th0, TGFβ) 3h; Th0 1d; Th1 (Th0+IL-12+anti-IL-4) 1d; iTreg 1d; Th0 1d + TGFβ 1h; Th0 1d + TGFβ 3h. (}A) The values Erk2 ^{vs. WT for each probe are presented as scatter plots for Th0, Th1 and iTreg conditions at 1d. Numbers indicate genes with a difference in expression of 1.5-fold or more (Erk2 suppressed, top left, red dots) or 0.67 fold or less (Erk2-dependent, bottom right, blue dots). (}B) The numbers of Erk2-dependent or Erk2-suppressed genes from the scatter analysis (A) are shown in Venn diagrams to indicate the overlap between each category. (C) The values of probes for Dnmt1 and Dnmt3b were normalized to the median for each probe and charted for all 8 conditions indicated on the figure alternating between WT and KO. (D) The Erk2-suppressed iTreg genes were identified and the expression values were normalized to the maximum for each probe, ordered by hierarchical clustering (Pearson Correlation) and depicted as a correlated heat map. The criteria were: iTreg-inducible, iTreg 1d Erk2 ^{> (1.5) 0d} Erk2^{; TGFβ-dependent, iTreg 1d} Erk2 ^{> (1.5) Th0 1d} Erk2^{; not Th1, iTreg 1d} Erk2 ^{> (1.5) Th1 1d} Erk2^{; and Erk2-suppressed, iTreg 1d} Erk2 ^{> (1.3) iTreg 1d WT. The criteria were chosen for biological relevance and to identify a manageable set of genes. Asterisks indicate the number of probes identified for each gene if greater than 1.}

This large-scale increase in gene expression at 1d in Erk2 ^{T cells could be consistent with a hypomethylated genome as a result of reduced} Dnmt1; however, the qPCR data showed no Erk2-dependent difference in Dnmt1 expression at this time-point (Fig. 5A). The array data showed that, of the Dnmt isoforms queried, only Dnmt1 and Dnmt3b were expressed over background, and neither was affected by the addition of TGFβ (Fig. 6C)—equivalent induction was seen at 1 d under Th0, Th1, and iTreg conditions of stimulation. Moreover, the effects of Erk2 deletion on Dnmt1 and Dnmt3b expression were small compared with the magnitude of induction. We conclude that Erk2 signaling constitutes an important suppressor of gene expression in the presence of TGFβ, although the extent to which this results from Dnmt1 regulation is unknown (discussed below).

The iTreg cell dependence on Erk2 thus provided a tool to more finely characterize a program of iTreg-specific genes. We identified those genes that were induced in 1d iTreg conditions, but not under Th0 or Th1 conditions, and Erk2 suppressed (Fig. 6D). This algorithm identified 45 unique probes, including all three probes specific for Foxp3, and in addition, both probes specific for the Treg-specific neuropilin-1 (Nrp1). Gpr83, Ecm1, Cmtm7, Nkg7, Socs2 and Glrx were previously found to be iTreg cell specific (32, 32) and of these, we also identified Gpr83 and Nkg7. Socs2 was strongly (>10-fold) induced in Th0, Th1, and iTreg conditions (see below), and thus not iTreg specific (Fig. 7D). Ecm1 was induced at 3h but not d1. Cmtm7 and Glrx were relatively unchanged across all conditions (data not shown).

FIGURE 7

T cell subset array analyses. (A) The array values for Tbx21, Gata3, Rorc, and Foxp3 were charted. See legend to Fig. 6. (B) The array values for signature cytokines, Ifng, Il4, Il10, Tgfb1, and Il17a were charted. (C–E) Correlated heat maps of subset-genes were identified and as described legend to Fig. 6. (C) Th0, Erk2-dependent probes: inducible, Th0 1d WT > (2) 0d WT; Erk2-dependent, Th0 1d WT > (1.5) Th0 Erk2 ^{1d. (}D) Th0, Erk2-suppressed probes: inducible, Th0 1d WT > (2) 0d WT; Erk2-suppressed, Th0 1d Erk2 ^{> (1.3) Th0 1d WT. (}E) Th2-like probes: Erk2-suppressed, Th0 1d Erk2 ^{> (1.5) Th0 1d WT; Not Th1, Th0 1d} Erk2 ^{> (3) Th1 1d WT. (}F) Th1 selected probes: inducible, Th0 1d WT > (3) 0d WT and Th1 1d WT > (3) 0d WT; Th1 not iTreg, Th1 1d WT > (2) iTreg 1d Erk2^{; Early-onset Erk2-dependent, Th0 3h WT > (1.3) Th0 3h} Erk2^.

Other genes identified in this subset that may be involved in immune function included: Dscr1l2 that inhibits calcineurin-dependent transcriptional responses; Ctsw (cathepsin W) which functions in T cell cytolytic activity; Faim3, a Caspase-8 inhibitor; Rcan3, which binds and inhibits calcineurin A; Tnfrsf1b (TnfRII); and Tnfrsf9 (4-1BB) (Fig. 6D). We also analyzed gene expression at 3 h post activation, with or without the addition of TGFβ. The rationale was that gene expression directly affected by TGFβ signaling would be induced early, yet none of the probes specific to iTreg induction at 1d were already induced at 3h in the presence of TGFβ. All of the probes induced at 3h with TGFβ were also induced in its absence under Th0 conditions (Fig. 6D). Likewise, we examined whether stimulation with TGFβ for 1h or 3 h after 1d Th0 activation would induce a subset of the signature genes, and we identified 5 of the genes listed in Fig. 6D. These included Ccr8, Ctsw, Dscr1l2, Emp1, and Edod1. Foxp3 and Nrp1 were not induced under these conditions. We conclude that the iTreg cell program of gene expression becomes elaborated in naïve T cells between 3h and 1d post TGFβ stimulation.

A separate analysis of the array data based on an unbiased hierarchical clustering of the entire data set (Pearson Corrlelation, Average linkage) was carried out and a heat-map of a cluster containing Foxp3 was selected (Supplemental Fig. 4A). Although there is a large overlap in the genes identified using the algorithm described above (e. g., Nrp1), there were also unique genes identified. Notably, this list includes Ccr6, encoding a chemokine receptor that plays a role in inflammatory bowel disease (33), and three transcription factors not previously known to play a role in T reg cell differentiation: Cux1, Stat5, and Zfhx3 discussed below.

Gene expression associated with Th0, Th1, and Th2 polarization

For each of the T cell subsets: Th1, Th2, Th17 and iTreg, there is a transcription factor that, if not a lineage specification factor, is characteristic and essential for function. As such, we charted the profiles of T-bet (Tbx21), Gata3, Rorγ (Rorc), and Foxp3 (Fig. 7A and Supplemental Fig. 4B) over the conditions tested. Remarkably, the profiles for Tbx21 (T-bet) and Gata3 were mirror images of one another—both for conditions of induction and Erk2 dependence. For all 8 conditions, Tbx21 was Erk2 dependent and Gata3 was Erk2 suppressed. This graphically illustrates the opposite polarity of these two T cell subsets, and reinforces the Erk2 dependence of Th1 differentiation contrasted with enhanced Gata-3 expression and Th2 differentiation found under conditions of Erk2 attenuation (Fig. 2C–G, Fig. 3A,B) (11). Equally striking was the profile of Foxp3. It was only induced under iTreg conditions for 1d, and consistent with increased induction of iTreg cells from Erk2 deficient precursors, its expression was further enhanced two-fold by the deletion of Erk2 (Fig. 7A, Supplemental Fig. 4A). In addition, under iTreg conditions, Erk2 deletion attenuated Tbx21 and enhanced Gata3, gene expression predicted to further promote iTreg differentiation (34, 35). Rorγ (Rorc) expression was not detected.

This analysis was extended by an examination of the signature cytokines for the different T cell subsets (Fig. 7B). IFNγ was only induced under Th1 conditions, but it was not appreciably Erk2 dependent. Since the MFI of intracellular IFNγ was higher in WT vs. Erk2-deficient T cells (Fig. 2C), there may be a post-transcriptional Erk2-dependent enhancement. IL-4 was induced at 3h under Th0 conditions, but it was close to baseline by 1d. TGFβ, sometimes characteristic of iTreg cells, was induced at 3h but not at 24h, and IL-10 was not induced under any conditions. These data reveal that there is not a strict correlation between the lineage-characteristic transcription factors and subset-specific cytokines, and this emphasizes the notion that T cell differentiation is not governed by the expression of a single, signature transcription factor (36, 37).

In addition, we analyzed those genes that are Erk2 dependent and induced under different conditions of T cell activation. There were 21 unique probes induced under Th0 conditions at 1d and Erk2-dependent (Fig. 7C). This set contained: Egr2; Ccl1; Cdkn1a (p21^{); and} Nfatc1. Further analysis showed that p21 ^{was the only cyclin-dependent kinase inhibitor that was Erk2 dependent under any of the conditions tested (data not shown). It is essential for the prevention of excess numbers of T cells and onset of a Lupus-like autoimmune syndrome (}38). The Th0 inducible probes that were Erk2 suppressed numbered 12 and included: Ccr8; Lif; and Socs2, a positive regulator of cytokine signaling (Fig. 7D) (39).

We did not analyze gene expression under Th2 conditions, but we looked for genes whose expression was similar to the pattern of Gata3 expression, i. e., Erk2 ^{Th0 was greater than WT Th0 and} Erk2 ^{Th0 was much greater than WT Th1. This identified only 5 probes including} Gata3 and Ccr8—a chemokine receptor found to be required for the recruitment of Th2 cells to allergen-inflamed skin (40) (Fig. 7E). A similar analysis was carried out for Th1 signature genes, looking for probes that were Erk2 dependent and preferentially induced in Th1 cells compared with iTregs (Fig. 7F). We identified 16 genes including: Ccl3, Ccl4, Cdkn1a (p21^), IL12rb1 (IL12 receptor) and Tbx21 (T-bet). We did not identify any further transcription factors that were specific to the Th1 subset.

Erk2 dependent cell survival

Since there was a marked loss of T cell survival correlated with increased Bim expression in Erk2 deficient T cells under Th1 conditions (Fig. 2B), we plotted the Bcl-2 family members, and found five that were highly regulated under the conditions tested (Fig. 8). Bcl2 and BclX_L (Bcl2l1) were mirror images with respect to T cell activation and Erk dependence, with Bcl2 being weakly Erk2 suppressed and BclX_L being weakly Erk2 dependent. Proapoptotic Bim was moderately Erk2 suppressed, consistent with the data presented in Fig. 2B and loss of survival under Th1 conditions. Interestingly, we found the prosurvival genes, Bcl2a1c and Bcl2a1d, to have expression patterns almost identical to that of Bim. The regulation of mitochondrial death in T cells in different phases of activation and differentiation is clearly complex.

FIGURE 8

The expression of Bcl-2 family members

The median normalized array values for Bcl2, Bcl2l1 (BclxL), Bcl2a1c, Bcl2a1d, and Bim were charted.

Discussion

Depending on cytokines and co-stimulation, strong activation can favor a Th1 response, whereas weaker activation favors either a Th2 response or iTreg cell differentiation (36, 41). Here we show that this strength of signaling contingency is mediated in part through Erk2. Although we found no phenotypic effect of Erk1 deletion, the effects of Erk2 deletion mimicked a weak signal, inhibiting Th1 differentiation while promoting iTreg differentiation. In addition, gene expression studies clearly illustrated the opposite polarity of Th1 vs. Th2 differentiation based on the presence or absence of Erk2. The origin of Erk1 and Erk2 differences is presently unknown, but their differential association with upstream kinase signaling complexes or localization within the cell may provide important information with respect to the signaling important for T cell differentiation.

Under conditions supporting Th1 differentiation both in culture and in response to LCMV infection, the loss of Erk2 caused a reduced accumulation and a decreased percentage of IFNγ ^{cells. This is consistent with a study showing that impaired H-ras and K-ras resulted in decreased IFNγ expression (}42). Reduced Th1 survival in the absence of Erk2 correlated with the increased expression of all three forms of Bim, including the highly apoptogenic Bim_s form (43). From the array data, this is at least partly due to differences in RNA expression as we showed for CD8 T cells (22). In addition, Bcl-2 expression was reduced upon T cell activation, whereas Bcl-X_L was increased (44), and these changes were inversely Erk2 dependent. The Bcl2a1 paralogs, like Bim, were strongly induced at 3 h, but their expression fell at 1d post activation. Bcl2a1d has a role in early thymocyte survival, and it is a direct target of NF-κB (45). Its expression pattern and promoter sequences suggest that it is also a direct target of Egr-2; however, the role of Bcl2a1d in mature T cell activation is unknown. The cell survival defect in Th1 cells was not universal since there was no Erk2-associated difference in survival or differentiation of IL-17 ^cells.

Under optimal conditions supporting Th2 differentiation, there was also reduced survival in the absence of Erk2, and a trend toward an increased percentage of IL-4 ^{cells. Previous work showed that Th2 differentiation in the absence of added IL-4 is enhanced by the attenuation of Erk (}10, 11, 46, 47). The dichotomy of Th1 vs. Th2 differentiation was clearly illustrated by the array expression patterns of Tbx21 and Gata3—almost perfect mirror images of one another with respect to Erk2 dependence and eight different conditions of activation.

The Foxp3 locus is regulated by methylation, and a model is that partial demethylation, accompanying TGFβ signaling and a decrease in Dnmt1 activity, is sufficient for iTreg differentiation (25–27, 48, 49). Furthermore, pharmacological inhibition of Erk promoted Foxp3 expression and suppressor function in naïve T cells, and this was also correlated with a diminished expression of Dnmt1—at least after 4 days of culture (12). Consistent with this, under iTreg conditions for 1d, the array value changes associated with a loss of Erk2 were much more numerous than those found for Th0 or Th1 conditions. Furthermore, the loss-of-Erk2 changes were overwhelmingly skewed toward increased expression. The implication is that there is suppression of gene expression, including Foxp3, by TGFβ that is Erk2 dependent. This would be consistent with diminished expression of the maintenance methyltransferase, Dnmt1, or possibly the de novo methyltransferase, Dnmt3b. However, neither Dnmt1 or Dnmt3b mRNA was suppressed by TGFβ at 1d, a time when Foxp3 has already been induced. In addition, array values and qPCR indicated that Erk2 deletion had a minimal effect on Dnmt1 mRNA expression at 1d compared to the magnitude of induction, although the small decrease in Dnmt1 in Erk2 ^{T cells was seen in all the array conditions that included TGFβ. At later time-points did we detect significant Erk2-dependent differences, but whether the small Erk2-dependent differences seen at 1d hold biological significance is unknown. We conclude that these results are not consistent with a simple model of iTreg differentiation based entirely on the amount of} Dnmt1 mRNA expressed; but we note that the regulation of Dnmt1 activity derives from a large allosterically controlled protein complex that assembles on chromatin in a cell-cycle dependent manner (50). Independent of mRNA amount, Dnmt1 activity may be regulated in an Erk and TGFβ-dependent manner in a manner not yet understood. Future studies will characterize the entire methylome of WT and Erk2 ^{T cells under alternate conditions of activation and differentiation.}

The set of genes that were specific to iTreg conditions and further induced in the absence of Erk2 included a number known to be part of the Treg cell signature, including Foxp3 and Nrp1. Other genes identified in this set may suggest mechanisms of T cell differentiation, but none encodes a DNA-binding transcription factor. A separate unbiased hierarchical cluster analysis was performed, and the cluster including Foxp3 identified an overlapping set of genes that also included three candidate transcription factors. One, Cux1, is TGFβ-regulated homeobox gene that gives rise to many different proteins with differing functions (51). One of the Cux1 knockout strains had multiple defects in T and B cell development, and we thus speculate that it controls part of the iTreg program (52). A second is Stat5, downstream of γ_C signaling and necessary for T reg cell function (53, 54), and a third is Zfhx3, Zinc finger homeobox 3, known to regulate p21 ⁽55). As Treg differentiation has been shown to depend on a higher order control that in turn induces Foxp3 (32, 56–58), a possibility is that these transcription factors as well as Foxo1,3 (18, 59, 60) initiate the program of Treg gene expression.

Another mechanism of Erk2 suppression involves TGFβ signaling—a central facet of iTreg cell differentiation. There was an increased amount of activated phospho-Smad2,3 in Erk2 deficient T cells, suggesting that Erk2 attenuates the TGFβR-mediated activation of Smads. A similar result was found for Mekk2 and Mekk3 deficient T cells that show impaired phosphorylation of Smad2,3 linker region, enhanced Smad transcription activity, and impaired IFNγ production (28). In addition, Mekk3 deficient T cells were impaired in their Erk activation and were defective in their Th1 responses (61). An implication from these studies and the present work is that Mekk2,3, upstream of all the Map Kinases, are specifically acting through Erk phosphorylation to affect TGFβ signaling. A further mechanism involves Gata3 shown to be essential for Treg cell function (34), and the absence of Erk2 there is a substantial increase in the expression of Gata3 under all conditions. Finally, Socs2 is a positive regulator of Stat3,5 signaling, and Erk2 suppresses its expression under Th0 and Th1 conditions. A possibility is that under conditions of limiting TGFβ, the signals such as those transmitted through the IL-2 receptor would be dampened by the activation of Erk2 (39). In sum, the strength of Erk2 signaling constitutes a contingency for T cell differentiation through its role as a highly connected node in signal transduction.

An unexpected result was that the most important role for Erk2 in T cells is survival and differentiation of effector T cells, separate from its role in cell proliferation. Erk is well established as an essential signal transducer in cell cycle entry and progression in all cells previously examined and under all stimulatory conditions thus far tested (62), and yet with the addition of co-stimulation through CD28, we found that T cells could still initiate cell cycle entry and complete division with a loss of Erk2 or even a complete loss of Erk1 and Erk2 (data not shown). This is not congruous with studies using pharmacological inhibitors of Mek1,2, suggesting that such inhibitors have off-target effects (63). Consistent with increased division in Erk2 deficient T cells, p21 ^{is strongly dependent on Erk2 for expression making it an important pathway for negative feedback control of T cell proliferation. This may partly explain the role p21 in tolerance and suppression of autoimmunity (}64). The mechanism underlying the bypass of Erk in T cell cycle progression is a topic of current study.

As finely tuned Erk signaling is essential for appropriate T cell differentiation, it is perhaps not surprising that defects in Erk expression are associated with autoimmunity. Studies have shown loss of Erk activation correlated with diminished Dnmt1 and DNA hypomethylation in patients with Systemic Lupus Erythematosus (65, 66). From the data presented here we would predict increased numbers of Treg cells in such patients, and in one study, newly diagnosed and untreated SLE patients had increased frequencies of CD4 ^{FOXP3 T cells that positively correlated with disease severity (}67). Why an increased proportion of Treg cells should positively correlate with autoimmune disease is unclear, but most likely this is correlative but not causative. Rather, changes in Erk activation underlie aberrant differentiation of T effector cells leading to a loss of tolerance. Overall, Erk signaling is clearly pivotal in T cell differentiation, and important for the regulation of T cell homeostasis.

Supplementary Material

1

Acknowledgments