Requirement for CD28 in Effector Treg Differentiation, CCR6 induction, and Skin Homing

Abstract

The skin, like most non-lymphoid tissues, contains substantial numbers of T cells. Among these, memory T cells serve a sentinel role to protect against pathogens, and regulatory T cells terminate immune responses as a check against unrestrained inflammation. Previously, we created conditional knockout mice with Treg-specific deletion of CD28. Although these mice have normal numbers of Tregs, these cells have lower levels of CTLA-4, PD-1 and CCR6, and the animals develop systemic autoimmunity characterized by prominent skin inflammation. Here, we have performed a detailed analysis of the skin disease in these mice. Our data show that Treg-expressed CD28 is required for optimal maturation of CD44^loCD62L^hi central Tregs (cTreg) into CD44^hiCD62L^lo effector Tregs (eTregs), and induction of CCR6 among the cells that do become eTregs. While CD28-deficient Tregs are able to regulate inflammation normally when injected directly into the skin, they fail to home properly to inflamed skin. Collectively, these results suggest a key role for CD28 costimulation in promoting a cTreg to eTreg transition with appropriate upregulation of appropriate chemokine receptors such as CCR6 that are required for tissue homing.

Introduction

The skin is the major contact point between “self” and the environment, and continually encounters foreign antigens and inflammatory substances. While the stratum corneum and other layers of the epidermis serve a primary barrier function, inevitably there are numerous breaches. Consistent with this, the skin is an immune cell-rich organ, with large numbers of T lymphocytes that serve multiple functions. Among other cell populations, the skin contains large numbers of memory T cells (Tmem), which like other tissue resident Tmem, are able to respond rapidly in situ to antigen challenge.

Prior work has shown that the skin also contains significant numbers of CD4⁺Foxp3⁺ regulatory T cells (Tregs)(1). As with naïve T cells and Tmem, they gain access to the skin via complex regulation of chemokine receptors, including upregulation of CCR4 and CCR6, and downregulation of CCR7. Importantly, almost all cutaneous Tregs bear a memory phenotype, and are thought to exist to counterbalance and terminate locally induced inflammation(2, 3).

Like non-regulatory CD4 T cells, Tregs constitutively express the CD28 costimulatory receptor. Previous studies have shown that CD28 is required for thymic Treg development(4, 5). Recently, our own group created a mouse with conditional targeting of CD28. Using Foxp3-Cre to delete CD28 in Tregs, we uncovered an obligate cell intrinsic role for CD28 in Treg function and Treg survival(6). Specifically, mice lacking CD28 in Foxp3⁺ cells (termed CD28-ArialΔTreg mice) developed lymphadenopathy and splenomegaly characterized by the accumulation of activated effector T cells (Teff) and Tmem. In addition, tissue-specific inflammation was seen in the lung, and more prominently, the skin. Consistent with this, the most notable phenotypic alterations in CD28-ΔTregs were decreased expression of PD-1, CTLA-4, and CCR6 (of particular interest, given the role of CCR6 in Treg migration to the skin).

Recently, based on the expression of CD62L and CD44, Treg cells were divided into two subsets with distinct characteristics: CD44^loCD62L^hi central Tregs (cTreg), mainly recirculating in secondary lymphoid tissues, and CD44^hiCD62L^lo effector Tregs in nonlymphoid tissues (7). Quiescent cTregs (which express CCR7 to enable lymphoid homing and retention) can differentiate into eTregs under inflammatory signals and migrate to nonlymphoid tissues, but the mechanism of how eTregs are activated and function in lymphoid and nonlymphoid organs remains unknown.

Here we have performed a detailed characterization of the skin disease seen in CD28-ΔTreg mice. We find that loss of CD28 in Tregs leads to failure to induce appropriate maturation of cTregs into eTregs as well as defective induction of CCR6 among the cells that do become eTregs. While CD28-deficient Tregs are able to regulate inflammation normally when injected directly into the skin, they fail to home properly to inflamed skin. Together, these results suggest a key role for CD28 costimulation in promoting a cTreg to eTreg transition with appropriate upregulation of appropriate chemokine receptors required for tissue homing.

Materials and methods

Mice

Mice with conditional targeting of CD28 in Foxp3⁺ cells (CD28^fl/fl x Foxp3^YFP-Cre, termed CD28-ΔTreg mice) have been previously described (6). CD28-ΔTreg mice, and CD45.1 or CD45.1/2 congenic B6 mice were bred in our facility and maintained under specific pathogen free conditions. All experiments described were approved by the Institutional Animal Care and Use Committee at the Massachusetts General Hospital.

Media, reagents, antibodies, and flow cytometry

Cell culture media was RPMI 1640 (Mediatech Inc.) supplemented with 10% heat-inactivated FBS, 100 U/ml penicillin, 100 mg/ml streptomycin, 2 mM L-glutamine, and 50 mM 2-mercaptoethanol (Sigma-Aldrich). Fluorescent anti-CD4, anti-CD8α, anti-CD8β, anti-CD25, anti-CD44, anti-CD45, anti-CD45.1, anti-CD45.2, anti-CD62L, anti-CD103, anti-CCR4, anti-CCR6, anti-CCR7, anti-CCR-9, anti-TCRβ and anti-TCRγδ antibodies were purchased from Biolegend. Anti-CTLA-4, anti-PD-1, anti-GITR, anti-CD127, anti-IFNγ, anti-CXCR3 and anti-CXCR4 were purchased from BD Biosciences-Pharmingen. Anti-MHCII, anti-CD207, anti-Gr-1, anti-CD11b, and anti-Foxp3 staining kit were purchased from eBioscience. Murine recombinant IL-2, IL-6 and E-selection-FC were purchased from R&D Systems. BD Cytofix/Perm buffer was used for intracellular staining. Cells were run on a BD LSR II flow cytometer or a Beckman Coulter Gallios flow cytometer and analyzed by Flowjo (Flowjo LLC).

T cell isolation

Spleen and lymph nodes were harvested from 4 to 8 week old CD28-ΔTreg mice and age-matched Foxp3-YFP-Cre littermates. Total cells were enriched for CD4⁺ cells by negative selection using a mouse CD4⁺ enrichment kit (Stemcell Technologies, Canada). For in vitro Treg activation assays, cells were stained with CD44-Percpcy5.5, CD62L-PE and CD4-A700. CD4⁺YFP⁺ CD44^hiCD62L^lo (eTregs) and CD4⁺YFP⁺CD44^loCD62L^hi (cTregs) were sorted by flow cytometry on a FACSAria Cell Sorter (BD Biosciences). Cell purity was routinely greater than 90%.

Tissue digestion and cell isolation

As previously reported(8), ears or skin grafts were minced into small pieces and put into a gentle MACS C tube (Miltenyi) with 3 to 5 ml DMEM digestion solution containing 2% FCS, 10mM Hepes, 0.1mg/ml Liberase TM (Sigma), 0.1 mg/ml DNAse I (Sigma) and 0.5 mg/ml hyaluronidase (Sigma). After 1.5 hour shaking at 37°C, ear tissues were further broken down in a gentle MACS dissociator (Miltenyi) and debris were filtered. Cells were washed in FACS buffer and stained for flow cytometry. To separate the epidermis and dermis, ears were split into dorsal and ventral halves and incubated with 20 mM EDTA in PBS at 37°C for 2 hr. Then the separated epidermis and dermis were digested as above.

Adoptive transfer of T cells and in vivo migration

Enriched CD4 T cells were adoptively transferred to congenic B6 mice by intraperitoneal injection.

Skin Transplantation

Skin transplantation was performed as described previously(9). In brief, recipient mice were anesthetized with isoflurane, and a 1 × 1 cm area of skin was removed from the lateral trunk. A full-thickness donor skin graft was sutured to the exposed s.c. tissue bed using 4.0 chromic absorbable suture, and animals were bandaged after application of antibiotic ointment to the graft.

DTH responses

Mice were sensitized on days 0 and 1 by painting 50 μl 0.5% 1-fluoro-2, 4-dinitrobenzene (DNFB, Sigma) in acetone on a shaved abdomen. On day 4, ears were painted with 20 μl of 0.2% DNFB in acetone and ear thickness was measured every 24 hours using a micrometer (Mitutoyo, Japan) for up to 7 days.

Intradermal injection of LPS

To induce local skin inflammation, a single dose of 10 μg LPS (sigma) in 10μl PBS was injected into the ventral side of the ear using a 31 gauge syringe.

Quantitative PCR

FACS sorted CD4⁺YFP+CD44^lo CD62L^hi cTregs and CD4⁺YFP+CD44^hi CD62L^lo eTregs were lysed in TRIzol (Invitrogen), and total RNA was extracted using the PureLink RNA Mini kit (Ambion). For CCR6 induction, FACS sorted CD4⁺YFP⁺ Tregs were stimulated with plate-bound anti-CD3 (2 μg/mL) and soluble anti-CD28 (5 μg/mL) for 24 h before RNA extraction. Genomic DNA was removed using DNase I (Invitrogen), and cDNA was generated using the iScript cDNA Synthesis Kit (Bio-Rad). Quantitative PCR was performed on a Stratagene Mx3005P instrument, using the QuantiTect SYBR Green PCR Kit (Qiagen) and the following amplification protocol: 10 minutes at 95°C, 40 cycles (30 seconds at 95°C, 1 minute at 55°C, 1 minute at 72°C), followed by confirmation of amplicon melting temperature. Reactions were performed in triplicate, and genomic DNA contamination was excluded by observing no amplification from non-reverse-transcribed RNA. Primer sequences used were as follows: 18S rRNA Fwd: 5′-GTAACCCGTTGAACCCCATT-3′ and Rev: 5′-CCATCCAATCGGTAGTAGCG-3′; β-Actin Fwd: 5′-GCGAGCACAGCTTCTTTGC-3′ and Rev: 5′-TCGTCATCCATGGCGAACT-3′; and CCR6 Fwd: 5′-CCTCACATTCTTAGGACTGGAGC-3′ and Rev: 5′-GGCAATCAGAGCTCTCGGA-3′. Signals from the β-Actin reactions were used to normalize signals in the CCR6 PCR reactions of the same cDNA sample to determine relative CCR6 mRNA levels for each sample.

Statistical Analysis

Comparison of means between groups used the two-tailed Student t-test. Differences were considered statistically significant at p<0.05.

Results

Abundant TCRαβ⁺ T lymphocytes in inflamed ears of CD28-ΔTreg mice

We have previously shown that loss of CD28 in Tregs caused a marked autoimmune skin disease in CD28-ΔTreg mice as they age (6). One of the earliest sites in which this immune/inflammatory process manifests itself is the ears, with grossly visible lesions occurring in mice aged 8 to 12 weeks. To characterize skin inflammation, ears from wild type (WT) and CD28-ΔTreg mice were digested in Liberase TM solution, and lymphocyte populations were analyzed by flow cytometry. We observed a large increase in the absolute number and percentage of CD45⁺ hematopoietic-derived cells in the ears of CD28-ΔTreg mice (Fig. 1 and data not shown). Among CD45⁺ cells, Thy1.2⁺TCRβ⁺ T lymphocytes were the major population that accumulated in the inflamed ears of CD28-ΔTreg mice, with substantial numbers of both CD4⁺ and CD8⁺ cells (Fig. 1A and 1D), suggesting that both populations contributed to skin inflammation (Figs. 1B and 1C). Importantly, while the percentage of total CD4⁺ T cells was dramatically enhanced in the ears of CD28-ΔTreg mice, the percentage of Tregs was significantly lower than that observed in littermate controls (Fig. 1D and 1E). Finally, both Tregs and effector T cells demonstrated a similar relative distribution between the epidermis and dermis of CD28-ΔTreg mice compared with controls (Fig. S1).

Figure 1.

Characterization of T lymphocytes in the skin of CD28-ΔTreg mice. A. Representative sequential gating strategy to identify T lymphocytes in digested ear tissues. B and C. Percentage (B) or total cell number (C) of CD45⁺, TCRβ⁺ and/or CD4⁺ cells in ears as gated in A (n=5 per genotype). D and E. Percentage (D) and total cell number (E) of Tregs in ears. WT – wild type; cKO – CD28-ΔTreg. 5 mice of each phenotype were analyzed.

Of note, there was no significant change in the percentage of MHCII⁺CD207⁺ Langerhans cells in the ears of CD28-ΔTreg mice (Fig. S2A). In contrast, the percentage of TCRγδ cells among T cells was reduced in ears of CD28-ΔTreg mice (Fig. S2B), which can be attributed to the increased number of TCRβ cells in ears, as the overall numbers of TCRγδ cells in the ears of CD28-ΔTreg mice was comparable to that of control mice (data not shown). To determine localization of cell types within the skin we separated the dermis and epidermis prior to cell isolation and analysis by flow cytometry (Fig. S2C–E). This revealed significantly increased numbers of TCRαβ cells, but not TCRγδ cells, in both the dermis and epidermis of CD28-ΔTreg mice (panel D). Because of the unchanged absolute numbers of TCRγδ cells (panel C) in the face of increased TCRαβ cells, the percentage of TCRγδ cells was decreased (panel E).

One explanation for the paucity of Tregs in the skin of CD28-ΔTreg mice could have been a failure of the cells to expand. We have previously shown that CD28-ΔTreg cells in lymph node had a similar turnover rate as WT Tregs (6), and consistent with this, Ki67 and vital dye staining of skin T cells demonstrated that both CD28-ΔTregs and effector T cells in CD28-ΔTreg mice had slightly higher levels of proliferation and cell death respecitvely (although neither was statistically significant) than the same populations in wild-type mice (Fig. S3), most likely due to localized inflammation in the CD28-ΔTreg animals. Thus, we conclude that CD28-ΔTregs are able to undergo normal proliferation in skin.

Detailed phenotypic characterization of CD4⁺ T cells in the skin of CD28-ΔTreg mice

It has been reported that cutaneous resident T cells, including Tregs, express surface markers characteristic of effector/memory cells (2, 10, 11). Therefore, we next examined chemokine receptors and activation markers of Tregs in the skin of CD28-ΔTreg mice. Previously, we found that CD28-ΔTregs in lymph nodes exhibited reduced expression of PD-1, CTLA-4 and CCR6 (6). In contrast, while CD28-ΔTregs in the skin continued to display reduced CCR6 levels, they demonstrated comparable expression of PD-1 and CTLA-4 as wild-type Tregs (Fig. 2A). Of note, while it was previously reported that lymph node CD28-ΔTregs expressed normal levels of CD25 and CD44, Tregs in inflamed skin showed reduced expression of each of these molecules (Fig. 2A). As expected, differences in cell phenotype were restricted to Tregs, as relevant surface markers of Teff populations in the LN and skin were largely similar in wild-type and autoimmune 8 week old CD28-ΔTreg mice (Fig. 2B).

Figure 2.

Cutaneous T cell phenotypes. A/B. Activation markers and chemokine receptors in cutaneous Tregs (defined as by sequential gating of CD45⁺Thy⁺TCRβ⁺CD4⁺CD25⁺YFP⁺ cells as per figure 1) from wild type and CD28-DTreg mice. Representative of 5 mice each phenotype. C/D. Chemokine receptors and adhesion molecules in CD4⁺ non-Tregs from lymph nodes and ears. Representative of 5 mice each phenotype. E/F. Selected markers in cutaneous Tregs from ears of CD28^fl/fl X Foxp3^YFP-Cre/+ female mice. CD28-sufficient and CD28-deficient Tregs were discriminated using a YFP gate in the Foxp3⁺ population. 3 mice were analyzed. G/H. Female (CD28^+/+ X Foxp3^YFP) mice (CD45.2) received skin transplants from CD45.1 congenic mice. Animals were sacrificed 7 days later and Tregs in ears and grafts were analyzed for CCR6 expression. 4 mice were analyzed. N.S.= not significant, p > 0.05.

To determine whether expression defects were cell-intrinsic due to loss of CD28, or alternatively due to a more inflammatory environment, cutaneous CD28-ΔTregs were also examined in female CD28^fl/fl X Foxp3^YFP-Cre/+ mice. In these animals, due to random X chromosome inactivation, only half of the Tregs delete CD28, and the complement of wild-type Tregs is sufficient to prevent disease (6). Interestingly, in this setting, all CD28-deficient skin Tregs expressed high levels of CCR6, equivalent to that of wild-type Tregs (Fig. 2C). This was surprising, as we had shown previously that most CD28-deficient lymph node Tregs in these mice had decreased CCR6 expression. We hypothesized that the few CD28-deficient Tregs that were CCR6 high were able to home to the skin, and that loss of CCR6 in CD28-deficient Tregs was a consequence of local inflammation. To more directly test this, we transplanted congenic skin onto mice with a CD28-sufficient Treg compartment, sacrificed the animals 7 days later, and examined Tregs in the grafts (which undergo sterile inflammation as a result of the transplantation procedure) and in their ears (a non-inflamed site). As shown in Figure 2D, CCR6 was downregulated in Tregs at the inflamed skin site but not in the quiescent one, indicating that maintenance of CCR6 on Tregs of healthy female CD28^fl/fl mice heterozygous for Foxp3-YFP-Cre likely is due to the complement of normal Tregs in these mice preventing systemic inflammation.

CD28 is required for the optimal expression of CCR6 in effector Tregs

Naive Tregs are activated in lymphoid organs and subsequently migrate to peripheral tissues via downregulation of CCR7 and upregulation of other chemokine receptors (12, 13). Recently, Smigiel et al. identified a population of CD44^loCD62L^hi Tregs as central Tregs (cTreg), found primarily in lymphoid organs, and CD44^hiCD62L^lo Tregs as effector Tregs (eTreg), residing primarily in tissues (7). Using this definition, we observed a somewhat lower percentage of eTregs in lymph nodes of CD28-ΔTreg mice compared with wild-type animals (Fig. 3A and 3B). Of note, in female CD28^fl/fl X Foxp3^YFP-Cre/+ mice, we observed higher percentages of CD28-sufficient vs. CD28-deficient cells among eTregs, suggesting that CD28 promotes the induction or maintenance of eTregs in a competitive environment (Fig. 3C).

Figure 3.

CD28-deficient eTregs are defective in CCR6 expression. A/B. Percentages of CD44^hiCD62L^lo eTregs and CD44^loCD62L^hi cTregs in lymph nodes of wild type and CD28-DTreg mice (<= 2 months old, 5 mice each genotype). C/D. Percentage of eTregs in CD28^fl/fl X Foxp3^YFP-Cre/+ female mice. Gating as per figure 2. 4 mice were analyzed. E/F/G. CCR6 protein expression (E/F) and mRNA expression (G) in cTregs and eTregs. H. Sorted WT Tregs were stimulated by plate-bound anti-CD3 antibody with/without soluble anti-CD28 antibody for 16 h. CCR6 mRNA expression is shown. 4 mice were analyzed. N.S.=Not significant.

As noted above, we have described lower CCR6 levels in total lymph node Tregs in CD28-ΔTreg mice (6). CCR6 has been identified as a marker for effector/memory Tregs and is involved in Treg trafficking to nonlymphoid tissues (14, 15). Consistent with this, in wild type mice we noted that while cTregs are almost all CCR6^lo/−, eTregs consist of two equal sized populations of CCR6^lo/− and CCR6⁺ cells. In contrast, CD28-deficient eTregs were a unimodal population of CCR6^lo/− cells (Fig. 3C) indicating that defective CCR6 expression in CD28-deficient Tregs occurs exclusively in the eTreg compartment.

To investigate one possible mechanism for alteration in CCR6 expression we analyzed CCR6 mRNA levels (Fig. 3G and 3H). As shown in Figure 3G, CCR6 mRNA is elevated in wild-type eTregs compared with wild-type cTregs, consistent with high expression of CCR6 protein in the former but not the latter. We observed that CCR6 mRNA is reduced in CD28-deficient cTregs and eTregs compared with their wild-type counterparts, however, because of variability in the amount of mRNA relative to housekeeping genes, only the differences seen in the cTregs were statistically significant. Figure 3H shows that CD28 provides a strong costimulatory signal for CCR6 mRNA induction compared with CD3 stimulation alone. Together these data suggest that at least part of the mechanism by which CD28 signals regulate CCR6 protein expression is control of mRNA levels.

eTregs are “produced” in the periphery from a population of thymic-generated cTregs (7). To test whether CD28 signals are involved in this process, cTregs were sorted and stimulated in the presence of APCs and anti-CD3 antibodies for 3 days. As shown in Figure 4, while CD28 is dispensable for the induction of eTregs (Fig. 4A – all comparisons between wild-type and CD28ΔTreg are non-significant), it is required for the induction of CCR6 on these cells (Fig. 4B). CD28 signaling cannot be replaced by IL-2 (Fig. 4B) and notably IL-2 actually suppresses induction of CCR6, which is of particular interest given the role of IL-2 in the maintenance of cTregs but not eTregs(7). The induction of expression of two other relevant cytokine receptors, CCR4 and CCR9, was minimally affected by the absence of CD28; both were somewhat lower in CD28-deficient Tregs compared with control Tregs, but the differences observed were not statistically significant (Fig. 4C). As well, for both CCR4 and CCR9, provision of IL-2 compensated for the absence of CD28 signaling (Fig. 4B). The ability to induce these chemokine receptors in the absence of CD28 may account for the fact that both CCR4 and CCR9 are expressed at comparable levels in Tregs directly isolated from CD28ΔTreg and littermate control mice (Fig. 2A, Ref. 6 and data not shown). Despite the lack of CCR6 induction, we found that when injected directly into the target site in vivo, sorted CD28-deficient CCR6^lo/− eTregs functioned as well as wild-type Tregs suppress ear swelling in a DTH model (Fig. 4C).

Figure 4.

CD28 signaling is required for CCR6 induction in eTregs. A. cTregs (CD44^loCD62L^hiYFP⁺) sorted from wild-type and CD28-DTreg mice were cultured in vitro in the presence of mitomycin-treated and T-depleted splenocytes in the indicated conditions. Cells were analyzed 3 days later. B/C/D. Differentiated eTregs from panel A (CD44^hiCD62L^lo gate) were analyzed for CCR6, CCR4 and CCR9 expression. N=6 experiments for CCR6 expression, 3 experiments for CCR6, plus CCR4 and CCR9. Data shown are from a representative experiment where all three were analyzed. E. Suppression of in vivo DTH response by eTregs. Mice were sensitized by abdominal painting with DNFR on days 0 and 1, and DTH was induced by DNFB painting of the ear on day 4. 5 × 10⁵ sorted CD4⁺YFP⁺CD44^hiCD62L^lo eTregs from wild type or CD28-ΔTreg mice were injected intradermally immediately after DNFB painting, and swelling was measured 24 hours later. Experiments were repeated three times. N.S.= Not significant.

Migration of CD28-ΔTreg cells

The data above showing intact function of CD28-deficient eTregs when delivered directly to the skin suggested a primary role of CD28 in enabling migration via upregulation of CCR6, and indicating that the defect seen in CD28-ΔTreg mice may be due to impaired Treg migration to the skin. To test the migratory capability of effector T cells and Tregs in a physiological environment, we adoptively transferred magnetic bead-enriched CD4⁺ T cells to congenic B6 mice and sacrificed the recipients 4 days later (Fig. 5A & 5B). Using lymph node cell populations pre-adoptive transfer as a baseline for comparison, we noted enhanced migration of CD28-ΔTregs to the skin (ear) and to a lesser extent in the lung, known sites of inflammation in CD28-ΔTreg mice. However, non-Tregs exhibited even more marked migration to the skin (likely due to the high proportion of activated Teffs in CD28-ΔTreg mice), and this could have biased Treg localization.

Figure 5.

CD28-deficient Tregs exhibit normal steady-state skin-homing in vivo. A - C. Magnetic bead-isolated CD4+ T cells from wild type and CD28-DTreg mice were adoptively co-transferred into congenic mice (A). Four days after adoptive transfer, animals were sacrificed after gating on Tregs or non-Tregs, and the donor cells were identified by CD45.1 or CD45.2 expression (B). The ratio of cKO-derived cells to WT cells was compared (C, n=3 experiments). D – F. As in panels A–C, except cells from wild type and CD28-ΔTreg donor mice were fractionated into Tregs and non-Tregs. Tregs or non-Tregs were then combined as depicted prior to adoptive transfer. N=3 experiments.

Therefore, to exclude the potential impact of effector T cells on Treg migration, we sorted CD4⁺CD25⁻ non-Tregs and CD4⁺CD25⁺ Tregs from control mice and from CD28-ΔTreg mice for co-adoptive transfers based on cell type, i.e., Treg or non-Treg (Fig. 5C & 5D). In addition, to facilitate the detection of these small numbers of cells, we induced minor irritation in the ears of the recipient mice with LPS on the day prior to cell transfer. We observed a trend to preferential localization of CD28-ΔTregs to the ear and lung, suggesting that preferential homing is not due to the confounding effects of Teffs.

These results, failing to show a defect in cell migration, were surprising, and we considered they may have reflected a steady state, rather than a response to inflammation. To test more directly CD28-ΔTreg migration to inflamed sites, we performed syngeneic skin transplantation on CD28-ΔTreg mice (Fig. 6). As expected, on day 7 post-transplantation, hematopoietic cells in the graft were primarily derived from the recipient (data not shown). Consistent with increased skin inflammation in CD28-ΔTreg mice, we saw a large increase in the percentage of CD4⁺ cells infiltrating grafts of these animals compared with controls (Fig. 6B). Moreover, when gating on these cells, we observed a dramatic (~85%) reduction in Treg infiltration of the graft in CD28-ΔTreg mice compared with control mice (Fig. 6C).

Figure 6.

Defective migration of CD28-ΔTregs following transplantation. A. Tail skin from CD45.1 congenic donors was transplanted onto CD45.2⁺ wild type or CD28-ΔTreg hosts. B. Skin graft analysis day 7 post-transplantation. Graft mononuclear cells were gated as shown. C/D. Analysis of gated CD4⁺ T cells from grafts and draining lymph nodes. Tregs, identified as CD25⁺YFP⁺ cells, are circled in panel C. E. Diagram of tamoxifen treatment and skin transplantation in CD45.2⁺ wild type (CD28^+/+ X Foxp3^eGFP-ERT2 X Rosa26YFP) or CD28-iΔTreg (CD28^fl/fl X Foxp3e^GFP-ERT2 X Rosa26YFP) mice. F. Analysis of host CD4⁺ T cells in lymph nodes, ears and skin grafts at day 7 post-transplantation. G. Analysis of CD4⁺GFP⁺YFP⁺ cells as shown in panel F. All experiments were performed three times.

We were still concerned that the reduced percentage of CD28-deficient Tregs in transplanted skin might be an indirect effect due to increased numbers of Teffs in CD28-ΔTreg mice. Thus to study CD28-ΔTreg migration in a more controlled environment, we crossed Foxp3^eGFP-CreERT2 Rosa26^YFP mice (16) to CD28^fl/fl animals to create mice in which CD28 deletion in Tregs was not constitutive, but rather inducible by tamoxifen (Fig. 6D, termed CD28-iΔTreg mice). While GFP expression is constitutive in Tregs, YFP is not. However, upon tamoxifen treatment, activation of Cre under the Foxp3 promoter also leads to the deletion of a stop codon in the ubiquitously expressed Rosa locus and allows for expression of YFP (which is much expressed at higher levels than GFP(16) and data not shown). As with most tamoxifen-inducible conditional knockout mice, deletion does not occur in every targeted cell (data not shown), thus treated animals have both CD28-deleted Tregs (YFP⁺) and CD28-sufficient Tregs (YFP⁻).

Following a 5 day course of tamoxifen, we placed syngeneic skin grafts on control (CD28^+/+ X Foxp3^eGFP-CreERT2 Rosa26^YFP) mice or CD28-iΔTreg mice, which were then harvested 7 days later. As expected, tamoxifen treatment did not induce deletion of CD28 in Tregs of control mice (Fig. 6E and data not shown), and both YFP⁻ and YFP⁺ Tregs were detectable in lymph nodes (Fig. 6E). While in the case of control mice, the proportion of YFP⁺ cells infiltrating skin grafts mirrored that found in lymph nodes (~85%, p=0.48 Fig. 6F), among CD28-iΔTreg mice, the proportion of YFP⁺, and thus CD28-deficient) Tregs infiltrating skin was only ~30% of the amount found in lymph nodes (p=0.006). Together, this data demonstrates that deletion of CD28 in Tregs in a healthy host led to a marked loss of their ability to migrate to an inflamed site.

Discussion

CD28 is required for thymic Treg development, and previously, by creating mice with conditional targeting of Cd28, we uncovered a cell-intrinsic role for CD28 in Treg maintenance and function(6). Those studies found that CD28-ΔTreg mice had normal numbers of Tregs, but nonetheless developed an autoimmune syndrome that manifested primarily in the skin (and to a lesser extent in the lung). Here, to gain further insight into the role of CD28 in Treg function in vivo, we have explored the characteristics and mechanism of skin autoimmunity in CD28-ΔTreg mice.

We find that the mononuclear infiltrate in inflamed skin of CD28-ΔTreg mice consists primarily of TCRαβ⁺ T cells. While absolute numbers of both conventional and regulatory cells are increased, the Treg percentage among infiltrating cells is greatly reduced. We examined multiple cell surface markers and found that the most striking phenotypic difference between cutaneous CD28-ΔTregs and control Tregs was in CCR6 expression, with markedly decreased expression of CCR6 on CD28-ΔTregs. This observation was consistent with our previously observed loss of CCR6 in CD28-deficient Tregs isolated from lymphoid organs (6). Additional studies suggest that when the small population of CCR6⁺ CD28-ΔTregs is able to localize to the skin, CCR6 is downregulated by local inflammation, as was suggested by prior investigations (17).

CCR6 is expressed in regulatory effector/memory-like T cells and mediates the migration of Tregs to inflammatory tissues (14, 15). While the lack of CD28 did not influence the percentage of eTregs in CD28-ΔTreg mice or their differentiation from CD62L^hi cTregs, it was required for CCR6 induction on eTregs. This suggested the scenario that the lack of CCR6 in CD28-deficient Tregs impaired their capacity to migrate to the skin and was the basis for skin inflammation in CD28-ΔTreg mice. Consistent with this, while CD28-ΔTregs were able to suppress a DTH response equivalently to wild-type cells when directly injected into the skin, a variety of adoptive transfer studies collectively showed impaired migration of CD28-ΔTregs to an inflamed site. Thus, CD28-deficiency within the lymph node may put Tregs at a disadvantage in the “race” with wild-type effector T cells to peripheral tissues, resulting in uncontrolled inflammation in the skin and lung.

The fact that CD28-deficient Tregs appear capable of mediating suppression, but fail to prevent disease in vivo, highlights the role of migration in Treg function, and also emphasizes the limits of in vitro suppression assays on predicting in vivo function. Using multiple different models of conditional gene targeting in Foxp3⁺ Tregs, we and others have found that intact suppression during an in vitro Treg-Tnaive co-culture assay does not predict normal in vivo immunoregulation by those same cells(6, 18, 19). Factors other than migration, including in vivo differentiation, survival and stability likely play a role as well in the frequent discordance between in vitro and in vivo findings.

A previous study reported that CCR6 expression was required on γδ effector T cells for migration within the skin to the epidermis in imiquimod-induced inflammation(20). Thus, microenvironments within peripheral tissues are key to disease, and directed T cell migration even after homing to tissues from the blood influences inflammatory outcomes. We did not observe deficient homing of CD28-ΔTregs to the epidermis in our model, and instead the epidermal:dermal ratio of Tregs was increased in CD28-ΔTreg mice, implying Treg distribution imbalance may play a role in skin inflammation. Our studies are not directly comparable however, as imiquimod-induced inflammation strongly involves the epidermis, while skin inflammation in our model is primarily located in the dermis. It appears that it is homing to the skin, not within the skin, that contributes to inflammation in our model.

CCR6 promotes the migration of T cells to multiple tissues including the gut (21), however the autoimmune inflammation we have observed in CD28-ΔTreg mice is limited to the skin and the lung(6). It is notable that CD28-deficiency in Tregs, which leads to failure to induce CCR6 in eTregs, does not also cause inflammation in other sites. While the reason for this observation is not yet clear, redundancy in other homing molecules (such as gut homing receptors CCR9 and integrin alpha4beta7(22)) which enable Treg trafficking to different sites and/or the presence of regulatory mechanisms other than Foxp3⁺ T cells (such as Tr1 cells or regulatory dendritic cells(23, 24)) may serve to maintain immune homeostasis in the gut and other tissues. Alternatively, differences in the microbiome between gut and skin may underlie aspects of tissue specific phenotypes.