Pak2 Links TCR Signaling Strength to the Development of Regulatory T Cells and Maintains Peripheral Tolerance

Defined by their expression of the transcription factor, Foxp3, Foxp3⁺ regulatory T cells (Tregs) are a class of suppressive T cells that limit the proliferation and effector functions of conventional T cells in the periphery (1, 2). The importance of Tregs in preventing autoimmunity has been highlighted in studies that reported a severe and lethal lymphoproliferative disease in humans or mice harboring an inactivating mutation in the Foxp3 gene, a mutation that severely abrogated Treg development (3–5). Retroviral transduction of the Foxp3 gene caused non-Tregs to adopt a suppressive phenotype, whereas the small number of CD4⁺CD25⁺ Tregs in Foxp3-deficient mice exhibited limited suppressive activity (6–8), suggesting Foxp3 as a master regulator of Treg development and function.

Although the role of Foxp3⁺ Tregs in preventing autoimmunity is well understood, the molecular mechanisms defining Foxp3 induction and diversion into the Treg lineage remain enigmatic. Tregs are generated in the thymus (termed thymus-derived Tregs, tTregs) around day 3 after birth (9, 10) and seem to originate from CD4 single-positive (CD4SP) thymocytes, although a minor population of Foxp3⁺ tTregs can be found at the double-positive (DP) stage (11). How tTregs develop is still not clearly defined, but one prevailing idea is that TCR affinity plays an instructive role in this process (12–14). According to this TCR instructive model, upon migration into the medulla, tTregs are preferentially selected by high-affinity TCRs directed toward self-antigens, whereas most autoreactive conventional T cells are eliminated by negative selection (13, 15). Recent developments within this model have proposed a two-step process for tTreg differentiation in the thymus, stating that tTreg precursor cells first receive strong TCR signals that condition cells to become responsive to cytokines through upregulating expression of the IL-2 receptor chain, CD25 (16, 17). Upon exposure to IL-2, as well as other common γ-chain cytokines such as IL-15 and IL-7, tTreg precursors fully commit to the tTreg lineage through STAT5-mediated expression of Foxp3 that occurs through permissive chromatin remodeling at the Foxp3 locus (18). The importance of IL-2 in tTreg development has been demonstrated in studies in which CD25 and STAT5, the downstream transcription factor activated by IL-2, were deleted. These mice developed very few Foxp3⁺ tTregs in the thymus and manifested with severe widespread peripheral autoimmunity (19–21).

How developing thymocytes transmit the information of increased TCR affinity to the nucleus to facilitate their commitment to the Treg lineage has not been clearly elucidated, but recent findings highlighted specific requirements of distinct signaling pathways induced by TCR engagement. Development of tTregs was severely impaired in skg mice that carry a W163C mutation in the ZAP70 allele (22) or in knock-in mice that carry tyrosine to phenylalanine mutations of Y315 and Y319 in ZAP70 (23). Furthermore, knock-in mice that carry a mutation at the phospholipase C (PLC)γ1 binding site (Y136) of LAT significantly reduced the frequency and number of tTregs as well as their function (24–26). Interestingly, these Y136 LAT mutant mice showed only a partial defect in positive selection, suggesting that the signaling requirement for the development of tTregs might be different from those for positive selection of conventional T cells. Mice deficient in PLCγ1 also exhibited defects in tTreg development (27), underscoring the importance of a signaling pathway leading to activation of PLCγ1 in the development of tTregs. NFAT, which is activated by calcineurin following Ca²⁺-mediated activation of PLCγ1, enters the nucleus and binds to numerous DNA elements, including the Il2, Cd25, and Ctla4 loci as well as the Foxp3 promoter in cooperation with Foxp3 (28). Although there is functional redundancy between the four known NFAT isoforms, double deficiency of NFATc2 and NFATc3 manifested with severe lymphadenopathy and additional signs of inflammatory disease, suggesting that these two NFAT isoforms may be essential for tTreg development and/or function (29).

Although the aforementioned studies demonstrated the importance of TCR signaling for tTreg development, the specific signaling pathway that governs tTreg development remains to be fully investigated. Recently, our laboratory reported the necessity of p21-activated kinase 2 (Pak2) in regulating TCR-driven cytoskeletal signaling (30). In the absence of Pak2, thymocyte development or maturation was impaired depending on the stage at which Pak2 deletion occurred. Deletion of Pak2 at the DP stage using the Cd4-Cre transgene did not affect positive selection of CD4SP thymocytes or the numbers of DP and SP thymocytes, but rather it impaired the maturation of CD4SP thymocytes from the semimature to mature state (30). Interestingly, we found that Pak2 was required for actin cytoskeletal remodeling of CD4SP thymocytes during TCR-triggered cell spreading. Defect in actin cytoskeletal remodeling following TCR stimulation was accompanied by reduced PLCγ1 and Erk1/2 activation in Pak2-deficient CD4SP thymocytes. Given the importance of actin cytoskeletal rearrangement in generating optimal TCR-mediated signaling events, including activation of PLCγ1 (31), we hypothesized that Pak2 was required for the development of tTregs by providing high-affinity TCR-induced signals that are a prerequisite for the development of tTregs.

In this study, we report that Pak2 plays a central role in the development and maintenance of Foxp3⁺ Tregs by a T cell–intrinsic mechanism. In the absence of Pak2 in T cells, both the numbers and percentages of Foxp3⁺ tTregs were markedly reduced and differentiation of in vitro–induced Tregs (iTregs) was greatly inhibited. This was accompanied by the development of spontaneous colitis in T cell–specific Pak2-deficient mice, suggestive of a break in peripheral immune tolerance. Mechanistically, we used TCR signaling reporter mice to show that loss of Pak2 greatly impaired high-affinity TCR signals that are required for the generation of CD25⁺Foxp3⁻ tTreg precursors and further inhibited their development into CD25⁺Foxp3⁺ tTregs in the thymus. Interestingly, Pak2 deficiency in tTreg-committed CD25⁺ CD4SP thymocytes also inhibited optimal STAT5 activation following IL-2 stimulation, suggesting that, in addition to generating strong TCR signals, Pak2 regulates tTreg development by controlling IL-2–mediated signals. Moreover, inducible deletion of Pak2 inhibited maintenance of Foxp3⁺ Tregs in the periphery as well as the differentiation of iTregs. Collectively, we propose that Pak2 links TCR signal strength to the proper development, maintenance, and differentiation of Tregs and contributes to peripheral tolerance. These findings demonstrate the novel role of an actin cytoskeletal signaling pathway, involving Pak2, in the commitment of developing thymocytes to the tTreg lineage and the maintenance of peripheral tolerance.

Mice were bred and used under the animal study protocol approved by the Northwestern University Animal Care Use Committee. All mice were housed in the specific pathogen-free facility at Northwestern University according to the university and National Institutes of Health guidelines. The generation of T cell–specific Pak2-deficient mice (Pak2^F/F;Cd4-Cre) has been previously described (30). C57BL/6 mice (congenic marker CD45.2⁺) were crossed with BoyJ mice (congenic marker CD45.1⁺) to generate CD45.1⁺CD45.2⁺ heterozygous mice that were used in bone marrow chimera experiments. Pak2^F/F;Cd4-Cre mice in the C57BL/6 background were crossed with Nur77-GFP reporter mice, which were a gift from the Julie Zikherman’s laboratory at the University of California, San Francisco. Nur77-GFP BAC transgenic (Tg) mice were described previously (32). Mice bearing UBC-Cre-ER^T2 and Rosa26-YFP have been described previously (33). Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice were generated by introducing the floxed Pak2 allele by mating. Pak2^F/F;Cd4-Cre mice were crossed with Bcl2-Tg mice, obtained from Astar Winoto’s laboratory (34), to generate Pak2^F/F;Cd4-Cre;Bcl2-Tg mice. C57BL/6, Cd4-Cre, and BoyJ (B6 mice with the CD45.1⁺ congenic marker B6.SJL-Ptprc^a Pepc^b/BoyJ) that were used for breeding were originally obtained from The Jackson Laboratory.

The genotyping protocol for the floxed Pak2 gene has been previously described (30). The Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice were genotyped by genomic PCR isolated from tail clips. Amplification for the Cre gene and Rosa26-YFP was carried out by standard PCR protocol. Primers used to screen the Cre gene are: forward primer, 5′-TGGGCGGCATGGTGCAAGTT-3′, reverse primer, 5′-CGGTGCTAACCAGCGTTTTC-3′. Primers used to screen the Rosa26-YFP transgene are: Rosa26-YFP transgene-specific forward primer, 5′- GCGAAGAGTTTGTCCTCAACC-3′, wild-type (WT) forward primer, 5′- GGAGCGGGAGAAATGGATATG-3′, reverse primer for the transgene and wt, 5′-AAAGTCGCTCTGAGTTGTTAT-3′. The Lck-Bcl2 transgene in Pak2^F/F;Cd4-Cre;Bcl2-Tg transgenic mice was amplified by using forward primer 5′- CATGTGTGTGGAGAGCGTCAAC-3′ and reverse primer 5′- TAGCCATTGCAGCTAGGTGAGC-3′.

To obtain thymocytes or T cells, thymi or peripheral lymphoid organs (spleen, peripheral/mesenteric lymph nodes) were dissociated through a 40-μm nylon mesh into RPMI 1640 (supplemented with 10% FBS). Cells were washed with FACS buffer (2% FBS, 2% NaN₃, 2 mM EDTA) prior to Ab staining. For analysis of T cells from the lamina propria (LP), the large intestine (extending from the rectum to the cecum) was filleted, the feces were removed, and the colon was washed in PBS by vigorous shaking for 2 min. The tissue was incubated in 1 mM DTT/PBS with gentle agitation for 10 min at room temperature, followed by two incubations in 30 mM HEPES/10 mM EDTA PBS solution at 37°C for 10 min each, shaking at 200 rpm. The tissue was then transferred to complete medium (RPMI 1640 plus 10% FBS) supplemented with collagenase type VIII (200 collagenase digestion units/ml; Sigma-Aldrich) and DNase I (150 μg/ml; Sigma-Aldrich) and incubated for 90 min at 37°C. The tissue was further dissociated by vigorously shaking and filtering through 100-μm nylon mesh. To isolate T cells from the homogenate, the digested tissue was resuspended in 40% Percoll solution, overlaid onto 80% Percoll solution, and centrifuged at 600 × g for 20 min at room temperature. Single-cell suspensions that included T cells were isolated by collecting cells at the interface between the two Percoll solutions.

Flow cytometric analyses were done at the Robert H. Lurie Flow Cytometry Core Facility using the LSRFortessa (BD Biosciences) flow cytometry system. For surface staining, cells were incubated with fluorochrome-conjugated Abs for 30 min at 4°C using 1:200 dilutions of each Ab (unless otherwise specified). Cells were washed twice in FACS buffer prior to being analyzed or stained intracellularly using the Foxp3/transcription factor staining buffer set (eBioscience). Dead cells were excluded either using DAPI (Life Technologies) or Live/Dead fixable blue dead cell stain kit (Life Technologies). Forward scatter and side scatter were used to identify live lymphocytes. Data analysis was performed using FlowJo (version 9.6.2) software (Tree Star). Abs against mouse CD4 (GK1.5), anti-CD8α (53-6.7), anti-TNFR2 (CD120b; TR75-89), anti-TER119 (TER119), anti-CD11c (HL3), anti-CD62L (MEL-14), and anti-CD44 (IM7) were from BD Biosciences. Abs against mouse CD25 (PC61), anti-CD3ε (145-2C11), and anti-B220 (RA3-6B2) were from BioLegend. Abs against mouse GITR (DTA-1), anti-CTLA4 (UC10-4B9), anti-CD122 (TM-b1), anti-CD127 (eBioSB/199), anti-Foxp3 (FJK-165), anti-OX40 (OX-86), anti-GR1 (RB6-8C5), anti-CD11b (M1/70), anti–IFN-γ (XMG1.2), anti–IL-17A (eBio17B7), anti-IL4 (11B11), and anti-CD69 (H1.2F3) were from eBioscience. An Ab against phospho-STAT5 (Y694; C71E5) was from Cell Signaling Technology. Annexin V apoptosis detection kit was from BD Biosciences. For optimal detection of GFP reporter signal, cells were surface stained and immediately fixed in 2% paraformaldehyde for 15 min at room temperature prior to intracellular staining.

For mixed bone marrow chimera experiments, bone marrow was harvested from the tibias and femurs of WT Pak2^+/+ (CD45.1⁺CD45.2⁺) and Pak2^F/F;Cd4-Cre (CD45.2⁺) KO mice and mixed at a 1:1 ratio prior to transferring cells into lethally irradiated BoyJ recipient mice (B6.SJL-Ptprc^a Pepc^b/BoyJ, CD45.1⁺). Mixed bone marrow cells were resuspended at 4 × 10⁷ cells/ml in sterile PBS, and 100 μl (4 × 10⁶ mixed bone marrow cells) was injected i.v. via the tail vein into recipients that had received two doses of lethal gamma radiation (600 rad per dose), 3 h apart. Following 8 wk engraftment, cells from the thymus and peripheral lymphoid organs were harvested and analyzed by flow cytometry as previously described.

To analyze peripheral tissue inflammation in Pak2^F/F;Cd4-Cre mice, the Mouse Phenotyping and Histology Laboratory at Northwestern University performed a comprehensive organ necropsy, including H&E staining of the thymus, spleen, lymph nodes, liver, heart, pancreas, salivary glands, kidneys, esophagus, stomach, and intestines. For acquisition of H&E staining images, an Olympus BX41 stereo compound microscope with UPLFLN ×4, ×10, and ×40 objectives with air at 25°C was used. NIS-Elements Documentation imaging software was used as acquisition software.

To induce UBC-Cre-ER^T2–mediated deletion of the Pak2^F/F alleles and expression of YFP, mice were administered with tamoxifen (100 mg/kg; Sigma-Aldrich), dissolved in corn oil, by oral gavage for 5 consecutive days. Animals were sacrificed for analysis after a 5-d rest period. Pak2^F/F;Rosa26-YFP mice that did not contain the UBC-Cre-ER^T2 transgene were treated with tamoxifen and used as a WT control.

Naive CD4⁺CD25⁻ T cells were isolated by magnetic depletion of cells labeled with a biotin-conjugated Ab mixture (CD8α, CD11b, CD11c, CD19, CD45R, CD49b, CD105, MHC class II, Ter-119, and TCRγδ; Miltenyi Biotec), anti-CD25 (eBioscience), and anti-biotin microbeads (Miltenyi Biotec). For iTreg differentiation, 1 × 10⁶ cells/well were cultured in 24-well plates precoated with 2 μg/ml anti-CD3 Ab (2C11; BD Biosciences) and supplemented with 2 μg/ml soluble anti-CD28 (37.51; BD Biosciences) and 10 U/ml IL-2 (National Institutes of Health) in the presence or absence of 10 ng/ml recombinant human TGF-β1 (R&D Systems) for 3 d.

Statistical analysis and graphs were generated using Prism 6 (GraphPad Software).

We have previously reported that T cell–specific deletion of Pak2 using the Cd4-Cre transgene resulted in severe T cell deficiency. When we compared mice ∼4–6 wk after birth, we did not find significant changes in total cell numbers from the thymus, spleen, and axial, brachial, and inguinal lymph nodes between WT (Pak2^F/F) and T cell–specific Pak2 KO (Pak2^F/F;Cd4-Cre) mice, except that some Pak2 KO mice showed an increase in cell numbers within mesenteric lymph nodes. Although total cell numbers were not changed, numbers of CD4 and CD8 T cells were significantly decreased in all secondary lymphoid organs (30). Starting at 2 mo of age, ∼50% of T cell–specific Pak2 KO mice (Pak2^F/F;Cd4-Cre) spontaneously developed symptoms of severe inflammatory disease, including rectal prolapse, colon distension, splenomegaly, lymphadenopathy, and thymic atrophy (Fig. 1A, 1B). The overall penetrance of colitis onset was ∼50%, with male mice showing a greater tendency for developing colitis (75%) compared with female mice (33%). Upon disease onset, total cell numbers were markedly higher in the spleen and lymph nodes, a consequence of increased numbers of B cells, erythrocytes, and polymorphonuclear leukocytes, whereas T cell numbers remained reduced (Fig. 1C–F). When mice develop colitis associated with splenomegaly and lymphadenopathy, their myeloid populations were expanded as shown by the increase in the GR1⁺CD11b⁺ population in the colon (Fig. 2B) as well as in secondary lymphoid organs (Fig. 1D–F, Supplemental Fig. 1C). When we examined the architecture of the spleen and lymph nodes, Pak2 KO mice that had developed colitis showed marked disarray of splenic architecture with disruption of lymphoid architecture as well as foci of lymphocyte apoptosis (Supplemental Fig. 1A, 1B).

Histological examination of the distal colon revealed submucosal edema with dilated blood vessels and superficial erosion, accompanied by severe depletion of goblet cells, increased abscesses, and architecturally compromised mucosal crypts with remarkable neutrophil infiltration and lymphocyte apoptosis, suggesting that Pak2-deficient mice develop colitis (Fig. 2A). Inflammatory lymphocyte and neutrophil infiltrates were found in the large intestine and not distributed to the lung, pancreas, and heart (data not shown). Further analysis of cellular components of the LP from Pak2-deficient mice that had developed colitis confirmed increased neutrophil and T cell infiltration in the colon (Fig. 2B). Expression of the proinflammatory cytokine IL-17 was elevated in T cells derived from the inflamed LP of Pak2-deficient mice that developed colitis, but expression of IFN-γ was not (Fig. 2C). T cells from the spleen (data not shown) and lymph node of Pak2-deficient colitis-developed mice produced increased amounts of both IL-17 and IFN-γ (Fig. 2D). Additionally, T cells from the LP, spleen, and lymph node from Pak2-deficient mice that had not developed colitis also showed increased production of IL-17 and IFN-γ, suggesting that there was ongoing inflammation in these mice even though apparent symptoms had not manifested (Fig. 2E, 2F). These data suggested that tolerance mechanisms that regulated proinflammatory cytokine production in peripheral T cells were broken, although it is not clear whether dysregulated cytokine production is the cause or consequence of the colitis seen in T cell–specific Pak2 KO mice. Of note, we did not find any sign of elevated serum autoantibody titers or elevated IL-4, even though the numbers of B cells were increased in the spleen and lymph nodes of Pak2^F/F;Cd4-Cre mice with colitis (data not shown, Fig. 1D–F), excluding the possibility of B cell–mediated autoantibody production as a mechanism for the colitis.

Under physiological conditions, active inflammation is controlled by several peripheral tolerance mechanisms, including immunosuppression by Tregs (35). To determine whether Pak2 contributed to the development of tTregs, we used Foxp3 and CD25 as markers to identify Foxp3⁺ tTregs. Pak2-deficient mice possessed significantly fewer CD25⁺Foxp3⁺ cells within CD4SP thymocytes (Fig. 3A). The percentage of CD25⁺GITR⁺ or CD25⁺CTLA4⁺ (Fig. 3A) or Foxp3⁺GITR⁺ or Foxp3⁺CTLA4⁺ cells within CD4SP thymocytes was also greatly reduced (Fig. 3B). Cell numbers and frequency of Foxp3⁺ tTregs within CD4SP thymocytes (either defined by CD25⁺GITR⁺Foxp3⁺ or CD25⁺Foxp3⁺) were markedly reduced (Fig. 3C, 3D, and data not shown). This reduction was not a result of fewer CD4SP thymocytes, because no difference in conventional Foxp3⁻ CD4SP thymocyte numbers was observed between WT and Pak2-deficient mice (Fig. 3E). Numbers of Foxp3⁺ DP thymocytes remained unchanged (Fig. 3F).

Pak2 deficiency significantly reduced the numbers and frequencies of CD25⁺Foxp3⁺ tTregs, but not CD25⁺Foxp3⁻ tTreg precursors within CD4SP thymocytes, although there was a trend for tTreg precursors from Pak2^F/F;Cd4-Cre mice to be reduced compared with WT mice (Fig. 3C, 3D, Supplemental Fig. 2). Furthermore, expression of CTLA4 within CD25⁺Foxp3⁺ tTregs was markedly reduced whereas GITR and OX40 expression remained comparable to WT mice (Fig. 3G). Although the mean fluorescence intensity of CD25 (IL-2Rα) was comparable between WT and Pak2-deficient Tregs, the percentage of cells expressing the highest levels of CD25 were reproducibly reduced (Fig. 3G, Supplemental Fig. 3). Expression of CD122 (IL-2Rβ) and CD132 (common γ) were similar (Fig. 3G, data not shown). Moreover, when gating on the CD25⁺GITR⁺ population within CD4SP thymocytes, expression of Foxp3 was reduced in the absence of Pak2 (Fig. 3G).

T cell–specific deletion of Pak2 using the Cd4-Cre transgene resulted in peripheral T cell lymphopenia due to defects in egress and maturation of CD4SP thymocytes, resulting in only 10–20% of conventional T cells in the periphery compared with WT mice (30). Similar to the reduction in conventional CD4 T cells, the numbers of CD25⁺Foxp3⁺ Tregs were markedly reduced (Fig. 4A, 4B), even though the percentage of CD25⁺Foxp3⁺ Tregs within CD4 T cells was comparable between spleens and mesenteric lymph nodes from WT and Pak2^F/F;Cd4-Cre mice and increased in the peripheral lymph nodes of Pak2^F/F;Cd4-Cre mice (Fig. 4C, 4D). Whereas absolute Treg numbers were reduced, CD25⁺Foxp3⁺ Tregs of the lymph node or spleen of Pak2-deficient mice displayed increased expression of CD44, GITR, or CTLA4 (Fig. 4E). These changes in expression might indicate that there is an increase in autoreactivity among mature T cells due to a reduction in the number of peripheral Tregs. Alternatively, this increase might be a consequence of lymphopenia-induced proliferation. At this point, we cannot distinguish between these two possibilities. Taken together, these findings suggest that Pak2 is required for the proper development of tTregs in the thymus and its absence results in a significantly reduced Foxp3⁺ Treg pool in the periphery.

To determine whether the defect in tTreg development was T cell intrinsic, we performed competitive bone marrow repopulation experiments in which irradiated hosts were reconstituted with hematopoietic stem cells from WT and Pak2-deficient mice at a 1:1 ratio. Although Pak2 mRNA expression was reduced by 90% at the DP stage in Pak2^F/F;Cd4-Cre mice and >95% at semimature CD4SP stage (data not shown), Pak2-deficient donor cells were able to repopulate the DP or semimature CD4SP stage efficiently (Fig. 5C). In contrast, Pak2-deficient donor cells showed a selective disadvantage in generating Foxp3⁺CD4SP thymocytes (Fig. 5A, 5B). Within CD4SP thymocytes, generation of CD25⁺Foxp3⁻ tTreg precursor cells as well as CD25⁺Foxp3⁺ tTregs was impaired (Fig. 5B, 5C). Expression of CD25 was markedly reduced in Foxp3⁺ CD4SP thymocytes developed from Pak2-deficient donor cells, suggesting that Pak2 was required for signals that increase CD25 (Fig. 5D). Likewise, expression of Foxp3 within CD25⁺GITR⁺CD4SP thymocytes was decreased (Fig. 5D). These results suggested that Pak2 might be required for two stages of tTreg development: first, to receive signals that are required to increase CD25, and second to induce Foxp3 expression. Because we observed a significant difference in generating tTreg precursors only under competing conditions in these 1:1 chimeras (Fig. 5C) but not under noncompeting conditions in Pak2^F/F;Cd4-Cre mice (Fig. 3C, 3D), it is likely that the presence of WT cells outcompeted Pak2-deficent cells from receiving signals to become tTreg precursors. Furthermore, because Pak2-deficient thymocytes developed under conditions in which WT thymocytes were present at an equal ratio, these results indicated that impaired tTreg development in Pak2-deficient mice was T cell intrinsic and not a consequence of absent exogenous factors, such as IL-2.

T cell lymphopenia in the periphery of Pak2^F/F;Cd4-Cre mice prevented addressing whether reduced numbers of Foxp3⁺ Tregs were due to reduced numbers of CD4 T cells or rather a specific defect in Tregs. To determine whether Pak2 plays a role in the homeostasis of Foxp3⁺ Tregs in the periphery, we generated a tamoxifen-inducible Pak2 KO mouse model (Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP). The Cre recombinase–human estrogen receptor (Cre-ER^T2) chimeric molecule under the UBC promoter allowed deletion of Pak2 following tamoxifen administration (Fig. 6A). The Rosa26-YFP reporter was also introduced, with the gene encoding YFP preceded by a 5′ Cre excisable “stop” cassette inserted into the Rosa26 locus. When tamoxifen is administered, Cre becomes active and deletes both Pak2 genes and the transcriptional stop signal preceding the YFP reporter, marking cells that are Pak2 deficient with YFP. Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice did not express YFP in the absence of tamoxifen administration (data not shown). Approximately 40–70% of CD4⁺ T cells from the lymph nodes, spleen, and thymus of Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice (Cre⁺) exhibited YFP fluorescence following tamoxifen administration (Fig. 6B, left panels). Importantly, purified YFP⁺ peripheral T cells from tamoxifen-administered Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice showed efficient deletion of Pak2 (Fig. 6B, right panel, Supplemental Fig. 4). Because purified YFP⁻ peripheral T cells from tamoxifen-administered Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP mice also decreased Pak2 expression (Supplemental Fig. 4), we did not use these cells as WT controls. Instead, we used Pak2^F/F; Rosa26-YFP mice that did not contain the UBC-Cre-ER^T2 transgene as WT controls (Cre⁻, Fig. 6B, left panels). As expected, T cells from Pak2^F/F; Rosa26-YFP mice expressed sufficient Pak2 and did not express YFP following tamoxifen administration (Fig. 6B, Supplemental Fig. 4). Approximately 10% of CD4 T cells were CD25⁺Foxp3⁺ Tregs in the periphery of Pak2 WT (Pak2^F/F; Rosa26-YFP) mice (Fig. 6C, left panel). Percentages of CD25⁺Foxp3⁺ Tregs in the periphery were reduced by >50% when Pak2 was deleted following tamoxifen administration in Pak2 inducible KO (Pak2^F/F;UBC-Cre-ER^T2;Rosa26-YFP) mice (Fig. 6C, left panel). The effect of inducible Pak2 deletion was smaller in total CD4 T cells (Fig. 6C, right panel). Moreover, CTLA4 expression was reduced in Foxp3⁺ Tregs when Pak2 was inducibly deleted, whereas CD25, GITR, and OX40 expression remained similar (Fig. 6D, data not shown), supporting that increased GITR and CTLA4 expression in Tregs in Pak2^F/F;Cd4-Cre mice (Fig. 4E) was due to lymphopenic-induced proliferation. These results imply that continuing signals by Pak2 are required for homeostasis of Foxp3⁺ Tregs and continued expression of CTLA4 in the periphery.

Naive conventional Foxp3⁻ CD4 T cells can be converted into Foxp3⁺ Tregs if the appropriate cytokine milieu is present (36, 37). We determined whether Pak2 was required for the generation of iTregs using CD4 T cells from tamoxifen-inducible Pak2-deficient (KO) mice. When cultured in iTreg polarizing conditions, Pak2 KO naive CD4 T cells showed markedly reduced Foxp3 induction relative to WT CD4 T cells (Fig. 6E). This was particularly evident at 24 and 48 h following induction (Fig. 6E, 6F). Induction of CD25 and CTLA4 were also impaired in Pak2 KO naive CD4 T cells under iTreg differentiation conditions (Fig. 6E, 6F). When Foxp3⁺ cells were gated to identify Foxp3⁺ iTregs from CD4 T cells, CTLA4 expression was substantially decreased in the absence of Pak2, whereas CD25, GITR, and OX40 were comparable between Foxp3⁺ WT and KO iTregs (data not shown). Reduced iTreg numbers were not a consequence of increased apoptosis, because annexin V staining of WT and KO cells cultured under iTreg conditions remained comparable (data not shown). These data suggest that the absence of Pak2 not only impaired the development of tTregs in the thymus, but also impaired the differentiation of naive peripheral CD4 T cells into iTregs. Of note, it is possible that the CD4⁺CD25⁻ T cells that we used for iTreg differentiation might contain a small number of promiscuous revertant tTregs that unstably express Foxp3. Sorting CD4⁺Foxp3⁻ Tregs from Foxp3-GFP reporter mice will help to ensure that these results are specifically due to defects in differentiation of iTregs in the absence of Pak2.

Thymocytes that receive strong TCR signals during their development become tTreg precursors and tTregs, as shown by increased TCR signaling strength in both precursor and tTreg populations from a TCR signaling reporter mouse model (38). Given our results that Pak2 was necessary for proper tTreg development and differentiation of iTregs, it is possible that Pak2 may facilitate signaling downstream of the TCR. Indeed, we previously reported that CD4SP thymocytes from Pak2^F/F;Cd4-Cre mice showed decreased Erk and PLCγ1 activation accompanied by defects in actin cytoskeleton-mediated spreading following plate-bound TCR stimulation (30). However, these reduced signaling events within CD4SP thymocytes are not a direct measurement of the signaling capacity of tTreg precursors and tTregs. Thus, we looked for a more physiologically relevant means to directly determine the signaling capacity of tTreg precursors as well as tTregs. To determine whether Pak2 provided strong TCR signaling in cells that are destined to become tTregs by interacting with endogenous ligands, we introduced the Nur77-GFP transgene into T cell–specific Pak2-deficient mice (Pak2^F/F;Cd4-Cre;Nur77-GFP). In these Nur77-GFP mice, TCR signaling strength can be measured by the expression of GFP under the control of the orphan nuclear receptor Nur77 promoter, an immediate early gene induced by TCR stimulation (38, 39), allowing us to measure signaling capacity of tTreg precursors and tTregs directly. Consistent with previous reports, GFP intensity was greatly increased in CD4SP thymocytes compared with DP thymocytes from WT (Pak2^F/F; Nur77-GFP) mice, indicating that positively selected CD4SP thymocytes activated the TCR signaling cascade via endogenous ligand interactions and expressed high levels of GFP as a consequence of Nur77 induction in response to TCR stimulation (data not shown).

Next, we determined TCR signaling strength of tTregs by measuring GFP intensity in the CD4⁺CD25⁺GITR⁺ populations of the thymus given that this population contains most Foxp3⁺ cells. GFP intensity was markedly increased in CD4⁺CD25⁺GITR⁺ putative tTregs compared with conventional T cells (GITR⁻CD25⁻ CD4SP) from WT reporter mice (Pak2^F/F;Nur77-GFP), but it was substantially reduced in CD4⁺CD25⁺GITR⁺ tTregs from Pak2-deficient reporter mice(Pak2^F/F;Cd4-Cre;Nur77-GFP) (Fig. 7A, 7C). To determine TCR signaling strength in CD25⁺Foxp3⁻ tTreg precursors as well as CD25⁺Foxp3⁺ tTregs, we measured GFP intensity in cells stained with Foxp3 and CD25 (Fig. 7B). Compared with conventional CD25-Foxp3–CD4SP thymocytes in quadrant (Q)4, WT CD25⁺Foxp3⁻ (tTreg precursors in Q1) cells contained cells that exhibited strong TCR signaling intensity as indicated by increased GFP intensity. Most WT CD25⁺Foxp3⁺ tTregs in Q2 gate showed high GFP intensity, supporting that strong TCR signaling was required for the generation of tTregs, as previously reported (38, 40) (Fig. 7B). In stark contrast, Pak2-deficient tTreg precursors in Q1 had very few cells that increased GFP, suggesting that Pak2 was required for inducing strong TCR signals in tTreg precursors (Fig. 7B, 7C). Furthermore, GFP intensity was reduced in Pak2-deficient CD25⁺Foxp3⁺ tTregs in Q2. According to recent reports, cells within Q3 (CD25⁻Foxp3⁺) include a population that have induced Foxp3 expression, but undergo cell death due to the inability to receive IL-2–mediated survival signals (41). WT CD25⁻Foxp3⁺ thymocytes within Q3 exhibited increased GFP expression (Fig. 7B), indicating that these cells had already received strong TCR-induced signals. In contrast, Pak2-deficient CD25⁻Foxp3⁺ thymocytes displayed substantially decreased GFP intensity. Interestingly, GFP expression was comparable between WT and Pak2-deificient non-Tregs (Q4) that exhibited medium intensity TCR signals (Fig. 7B), suggesting that Pak2 was not required for conditions where intermediate TCR signaling intensity was required.

Recent reports suggest that the expression of GITR corresponds with TCR signaling intensity (40). Thus, expression of GITR and Nur77-GFP was examined in four quadrants of cells stained with CD25 and Foxp3 (Fig. 7C). In WT tTreg precursors (Q1), the increase in GFP was proportional to GITR expression, with ∼60% of Treg precursors exhibiting high expression of both proteins. This suggested that, indeed, GITR^hi populations within Treg precursors were cells that received high-intensity TCR signals. In contrast, Pak2-deficient tTreg precursors were far inferior in their ability to induce high expression of GITR and GFP. Pak2-deficient CD25⁺Foxp3⁺ tTregs (Q2) were relatively competent at increasing GITR, but they failed to efficiently increase GFP expression. Pak2-deficient CD25⁻Foxp3⁺ CD4 T cells within Q3 reduced expression of both GFP and GITR compared with WT cells (Fig. 7C), similar to cells in Q1. These data showed that Pak2 was required for optimal TCR signaling in both Treg precursors and Treg-committed cells. Taken together, these results demonstrated that Pak2 was required for the generation of tTregs by providing strong TCR signals, which is a prerequisite for their development.

Once tTreg precursors increase CD25, the high-affinity receptor (IL-2Rα) for IL-2, IL-2 binds and activates STAT5. Activated STAT5 binds the Foxp3 locus and further facilitates the development of the tTreg lineage by inducing expression of Foxp3 as well as other survival factors, such as Bcl2 (16, 42–44). TCR signaling intensity was significantly decreased in tTreg precursors from Pak2^F/F;Cd4-Cre mice, but whether STAT5 activation differed in these cells was not known. Because WT tTregs express slightly higher levels of CD25, we sorted CD4SP thymocytes that expressed CD25 at a given range so that we only included CD25⁺ CD4SP thymocytes that exhibited comparable CD25 expression between WT and Pak2-deficient mice. We found that phosphorylation of STAT5 was greatly impaired in sorted cells from Pak2-deficient mice (Fig. 8A). Because activated STAT5 binds to regulatory regions of the Foxp3 locus and induces expression of Foxp3, these results indicated that Pak2 was also required for IL-2–mediated induction of Foxp3.

One of the functions of STAT5 is to promote the survival of developing tTregs by inducing expression of prosurvival factors, such as Bcl2 family members (43, 44). Indeed, IL-2 signals have been suggested to counteract Foxp3-mediated cell death by inducing expression of Bcl2 (41). Because we reported decreased STAT5 phosphorylation in CD25⁺ CD4SP thymocytes in the absence of Pak2, we asked whether IL-2–induced expression of survival factors, such as Bcl2, might be impaired. Interestingly, Bcl2 expression was similar in thymic tTregs and tTreg precursors, but it was greatly decreased in Pak2-deficient Foxp3⁺ Tregs in the periphery (Fig. 8B, data not shown). Because Bcl2 expression was markedly decreased in Pak2-deficient Foxp3⁺ Tregs in the periphery, we sought to determine whether providing exogenous Bcl2 into Pak2-deficient T cells could rescue the defects in Tregs from Pak2^F/F;Cd4-Cre mice. We introduced the Bcl2 transgene under the promoter of Lck into Pak2^F/F;Cd4-Cre mice by mating. Numbers and percentages of tTregs within CD4SP thymocytes were still markedly decreased, suggesting that providing prosurvival factor Bcl2 did not rescue defects in the generation of tTregs in the absence of Pak2 (Fig. 8C). Given that Pak2 is required for strong TCR signals that drive Treg lineage commitment, it is likely that IL-2–induced Bcl2-dependent prosurvival signals cannot compensate for the TCR signal that specifies the Treg fate in the thymus. Furthermore, introducing Bcl2 into Pak2-deficient mice did not fully restore reduced peripheral Foxp3⁺ Treg numbers to WT levels, but slightly increased numbers of Tregs compared with Pak2-deficient mice (Fig. 8D). Thus, providing prosurvival signals via Bcl2 was not sufficient to fully restore the function of Pak2 in maintaining Foxp3⁺ Tregs in the periphery, suggesting Bcl2-independent activity of Pak2.

Recently, it has been reported that Treg homeostasis in the periphery requires continuous TCR signaling (45, 46). Although Pak2 contributes to generate strong TCR signals for the development of tTregs (Fig. 7A–C), it is not known whether Pak2 controls the homeostasis of Foxp3⁺ Tregs in the periphery by regulating continuous TCR signaling. To test this, we measured GFP intensity driven by the Nur77 promoter in CD25⁺GITR⁺ or Foxp3⁺CD25⁺ Tregs from the lymph nodes from Pak2^F/F;Cd4-Cre;Nur77-GFP mice (Fig. 9). WT CD25⁺GITR⁺ or Foxp3⁺CD25⁺ Tregs displayed increased GFP intensity compared with conventional T cells, suggesting that Foxp3⁺ Tregs continuously received stronger TCR signals compared with conventional T cells in the periphery. In contrast, Pak2 deficiency in CD25⁺GITR⁺ or Foxp3⁺CD25⁺ Tregs substantially decreased GFP intensity. Collectively, these data support that Foxp3⁺ Tregs in the periphery maintain continuous heightened TCR signals, perhaps toward endogenous self-antigens, and additionally suggest that Pak2 is required for this process.

In this study, we provide a platform for better understanding the mechanisms underlying how developing thymocytes transduce high-affinity TCR signals to become tTregs, which currently remains elusive. We identified Pak2, a kinase involved in cytoskeletal remodeling, as an essential regulator that connects TCR signal strength to the development of tTregs. Loss of Pak2 disrupted strong TCR signals, ultimately inhibiting the generation of CD25⁺Foxp3⁺ CD4SP thymocytes that were competent for activating STAT5. Upregulation of CD25 was reduced following TCR/CD28 stimulation in the absence of Pak2 under iTreg differentiation conditions, implying that Pak2-mediated signals controlled TCR/CD28-mediated induction of CD25. For thymic development of tTregs, Pak2 governs TCR-specific signals that are required for the specification of the tTreg lineage rather than providing a survival signal that may counteract Foxp3-mediated death signals in the thymus (41), because providing survival signals to developing thymocytes, by introducing exogenous Bcl2, did not rescue the development of tTregs in the absence of Pak2. In the periphery, Pak2 affects maintenance of Foxp3⁺ Tregs by providing at least two signals: first, prosurvival signals that can be replaced by exogenous Bcl2, and second, Bcl2-independent signals. As a result, expression of Bcl2 in peripheral T cells only partially rescued the defects of Pak2 deficiency. Because continuous heightened TCR signaling within Foxp3⁺ Tregs was inhibited by the loss of Pak2, it is possible that TCR triggering of Foxp3⁺ Tregs produces both Bcl2-dependent and -independent signals via Pak2 for Treg homeostasis. Additionally, temporal deletion of Pak2 decreased numbers of peripheral Foxp3⁺ Tregs as well as differentiation of naive CD4 T cells into iTregs. Collectively, these data posit Pak2 as a critical signaling component that plays a key role in the development, homeostasis, and differentiation of Foxp3⁺ Tregs.

Among the many signaling cascades that are activated by TCR ligation, only some signaling components proximal to TCR stimulation are required for the development of tTregs. For example, mice expressing a mutant form of LAT that disrupted the interaction between PLCγ1 and LAT showed defects in tTreg development (24), suggesting that activation of PLCγ1 was critical for the development of tTregs specifically. Consistent with this, PLCγ1-deficient T cells also showed defects in tTreg development (27). Because PLCγ1 is critical for the activation of NFAT, and given the importance of NFAT in tTreg development (28), it is likely that the disruption of PLCγ1-mediated NFAT activation within these signaling mutants might provide a common mechanism for the observed tTreg developmental defects. Consistent with these data, PLCγ1 activation was severely impaired in Pak2-deficient CD4SP thymocytes following plate-bound CD3/CD28 stimulation (30), suggesting that the function of Pak2 in regulating PLCγ1 activation might be the key mechanism in controlling Treg development.

Deletion of Pak2 at the DP stage using the Cd4-Cre gene did not impair positive selection of CD4SP thymocytes with polyclonal TCRs, but it markedly decreased generation of tTregs in the thymus, suggesting that Pak2 was specifically required for the generation of tTregs. This raises the question of why Pak2 would affect the development of the tTreg lineage specifically. Analysis of TCR strength using the Nur77-GFP reporter gave us an indication of how this specificity might have originated. We reported that the intermediate TCR signaling strength exhibited by positively selected CD4SP thymocytes that bear endogenous polyclonal TCRs was not affected by the loss of Pak2. However, TCR signaling strength of Foxp3⁺ Tregs, which require strong TCR affinities, was greatly affected by the loss of Pak2, suggesting that Pak2 specifically contributes to checkpoints that generate strong, high-affinity TCR signals.

The nature of the Pak2-mediated “strong” TCR signal that is required for tTreg development still needs to be defined biochemically. We propose that actin cytoskeletal regulation by Pak2 may be key to explaining this specificity. Pak2 was dispensable for activation of soluble TCR stimulation but was required for plate-bound TCR-mediated PLCγ1 stimulation (30). More importantly, Pak2 deficiency disturbed TCR-mediated spreading of CD4SP thymocytes (30). If tTreg precursors with high-affinity TCRs cannot interact strongly with self-antigen–bearing APCs, it is likely that tTreg generation will be impeded. Because AIRE-expressing medullary thymic epithelial cells have been suggested as APCs that drive tTreg development (47, 48), it would be interesting to investigate whether Pak2 possesses a role in promoting the interaction between tTreg precursors and AIRE-expressing medullary thymic epithelial cells to generate strong TCR signals.

This work was supported by the Northwestern University Mouse Histology and Phenotyping Laboratory.

The authors have no financial conflicts of interest.