CD11c-Expressing Cells Affect Regulatory T Cell Behavior in the Meninges during Central Nervous System Infection Free

Regulatory T cells (Tregs) have potent suppressive capacity that is capable of limiting effector T cell responses and immune-mediated pathology in the context of immune homeostasis, as well as in infectious and noninfectious inflammatory processes. Although multiple suppressive mechanisms have been attributed to Tregs (1, 2), only a limited number of reports have examined Treg behavior in vivo, where Tregs have been imaged in the bone marrow, spleen, lymph nodes in diabetes and graft-versus-host models, and in tumors (3–7). In many CNS inflammatory conditions, Tregs are recruited to the brain, where it has been proposed that their presence represents one mechanism to limit the catastrophic consequences of inflammation in this site (8, 9). For example, in mice infected intracranially with murine hepatitis virus, the depletion of Tregs leads to an increase in self-reactive T cell responses and more severe pathology in the brain (10). Although the importance of Tregs in many experimental models that involve the CNS has been demonstrated (10–15), the behavior of these cells within the brain remains unexplored.

Toxoplasma gondii is a protozoan parasite that establishes a chronic infection within the CNS. In mice, cytotoxic T cells and T cell production of IFN-γ are required to control parasite replication within the brain (16–18). Several studies have established that Tregs contribute to the regulation of effector T cells during acute toxoplasmosis (19–21) and that, during many intracellular infections, there is the emergence of a population of Th1-like Tregs that express T-bet, IFN-γ, IL-10, and CXCR3 (20–22); however, there are open questions about the specificity of these populations (23, 24). During acute toxoplasmosis, expansion of the Treg population is associated with an increase in parasite burden within the brain (21, 25). These latter observations suggest that Tregs can suppress the protective T cell response required to control T. gondii, but it is unclear whether this is a general regulatory effect or is mediated locally within the brain. The studies presented in this article reveal that, unlike parasite-specific effector T cells, during toxoplasmic encephalitis (TE), Tregs were localized predominantly to the meninges and perivascular cuffs where they maintained interactions with CD11c-expressing cells, which influence the migratory behavior of Tregs. These observations suggest that Treg–dendritic cell (DC) interactions are an important component of Treg function during TE and this may be broadly relevant to Treg functions in other inflammatory settings that affect the CNS.

C57BL/6, CD11c-YFP, actin-CFP, and IL-10eGFP “Tiger” mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Foxp3-GFP mice were originally obtained from Vijay Kuchroo (Harvard University) and crossed with the CD11c-YFP strain. IL-10eGFP reporter VertX mice were obtained from Christopher Karp (Cincinnati Children’s Hospital). All procedures were performed in accordance with the guidelines of the University of Pennsylvania and University of Virginia Institutional Animal Care and Use Committees. OVA-expressing Prugnauid strain parasites expressing Tomato fluorescent protein were generated as previously described (26) and maintained by serial passage in human fibroblast cell monolayers in DMEM containing 10% FBS. Prior to infections, parasites were purified from human fibroblast culture by filtration through a 5.0-μm filter (Nucleopore, Clifton, NJ). Mice were infected i.p. with 10³ tachyzoites in 200 μl of PBS. Anti–LFA-1 Abs (Bio X Cell) and normal rat IgG (Sigma) in PBS were administered i.p. 4 h prior to imaging experiments.

Single-cell suspensions were generated from spleen and lymph node by macerating the tissues through 40-μm nylon mesh filters (BD Falcon, Bedford, MA). Spleen samples were subjected to hypotonic RBC lysis. Brain mononuclear cells (BMNCs) were isolated as previously described (27). Briefly, perfused brains were cut into small pieces, passed through an 18-gauge needle, and digested with collagenase/dispase and DNase (Roche) for 90 min. Following digestion, the cells were washed and strained through a 70-μm filter. Subsequently, cells were resuspended in 60% Percoll, overlaid with 30% Percoll, and centrifuged at room temperature for 25 min at 2000 rpm. BMNCs were collected from the interface, washed, and enumerated. For flow cytometry, 1–2 × 10⁶ cells were washed with FACS buffer (1× PBS, 0.2% BSA, and 2 mM EDTA) and incubated in Fc block (0.1 μg/ml 24G2 Ab) for 15 min prior to surface staining with CD4-FITC, ICAM-PE, CD11a-PE, CD25-PE, CD8-PerCpCy5.5, CD8–eFluor 780, CD45-allophycocyanin, CD11c-PECy7, CD11b–Alexa Fluor 780, CD3-FITC, CD19-FITC, NK1.1-FITC, and MHC class II–eFluor 450 (eBioscience). T. gondii–specific cells were identified with a PE-conjugated I-A^b–AVEIHRPVPGTAPPS tetramer reagent (National Institutes of Health Tetramer Facility, Atlanta, GA). For intracellular cytokine staining, cells were cultured for 4 h in the presence of PMA, ionomycin, and brefeldin A. Following surface staining, cells were fixed in eBioscience fixation and permeabilization buffer. Cytokines were detected with IFN-γ–PE–Cy7 and IL-10–PE. For GFP staining, cells were stained with rabbit anti-GFP (eBioscience), followed by a goat anti-rabbit Alexa Fluor 488 (Life Technologies) Ab. For transcription factor staining, cells were fixed with fixation and permeabilization buffer and stained with Foxp3–Pacific Blue, Foxp3–Alexa Fluor 488, or T-bet–PE (eBioscience). Flow cytometry was performed on a BD LSR II Fortessa or FACSCanto using FACSDiva 6.0 software (BD Biosciences, San Jose, CA). Statistical analysis was performed using FlowJo software (TreeStar, Ashland, OR).

Immune cells were isolated from the meninges of T. gondii–infected mice, as previously described (28). Cells were stained with Abs against CD3, CD4, and Foxp3 and sorted on a Becton Dickinson Influx cell sorter. DNA from Foxp3⁻ and Foxp3⁺CD4⁺ T cells was purified using a QIAGEN DNA Micro Kit. The TCR-β CDR3 regions were sequenced with an Immunoseq Assay (Adaptive Biotechnologies, Seattle, WA).

The TCR-β CDR3 sequences obtained were analyzed in the following ways. First, the presence/absence of amino acid sequences in the TCR sequencing was determined, giving us a binary matrix of 0s and 1s, with a 1 indicating that the sequence was measured at least once in that sample. The Jaccard index was used to quantify the similarity of the detected sequences in one sample versus another and was visualized with the R package corrplot (29). The binary matrix was then used to create the binary heat map showing the presence/absence of amino acid sequences. Before creating the heat map, amino acid sequences with low counts across all samples were removed; specifically, a sequence was removed if it did not make up 0.3% of the total measured sequences in at least one of the six samples. The heat map plot itself was produced with the R package pheatmap (30). The UpSet plot (31) used to visualize the overlap of amino acid sequences between the samples (set comparisons) was created with the R package UpSetR (32). As opposed to the binary heat map, all of the measured sequences were used to create the UpSet plot.

For immunohistochemistry, organs were embedded in OCT and flash frozen. Six-micrometer sections were cut using a Leica 3050 M cryostat (Leica Microsystems). Sections were fixed with a solution of 75% acetone and 25% ethanol. Sections were then stained with anti-laminin (CEDARLANE), anti-Foxp3 (eBioscience), and anti-CD4 (eBioscience). Anti-rabbit Alexa Fluor 488 (Invitrogen), anti-rat Cy3, or biotinylated anti-rat (Jackson ImmunoResearch) was used as a secondary Ab for fluorescence staining. A Cy3 tyramide signal amplification kit (Perkin Elmer) was used to amplify Foxp3 staining. DAPI (Invitrogen) was used to visualize nuclei. Images were captured using standard fluorescence microscopy using a Nikon Eclipse E600 microscope (Nikon, Melville, NY) equipped with a Photometrics CoolSNAP EZ CCD camera (Photometrics, Tucson, AZ) or an inverted Leica DMI4000 B microscope equipped with a Hamamatsu camera. MetaMorph and Nikon NIS Elements software were used to captures images, and Imaris (Bitplane) or ImageJ image analysis software was used to overlay images.

CD4⁺CD25⁻ cells from mice expressing cyan fluorescent protein (CFP) under the actin promoter were isolated using enrichment columns (Miltenyi Biotec, San Diego, CA). The cells were activated with plate-bound anti-CD3 (1 μg/ml) and anti-CD28 (1 μg/ml) and cultured with anti–IL-4 (1 μg/ml) and 20 U/ml recombinant human IL-2 (Proleukin) for 4 d. One million activated CFP⁺ cells were transferred to mice chronically infected with T. gondii and imaged 7 d later.

Mice were sacrificed by CO₂ asphyxiation, and the brains were removed immediately, with minimal mechanical disruption, and placed in heated chamber where they were constantly perfused with warmed (37°C) oxygenated media (phenol-red free RPMI 1640 supplemented with 10% FBS; Life Technologies). The temperature in the imaging chamber was maintained at 37°C using heating elements and monitored using a temperature control probe. Intravital imaging experiments were performed using the thinned skull technique, as previously described (33). Briefly, mice were anesthetized, and a region of skull bone (0.5–1 mm in diameter) was thinned with a dental drill and surgical blade until ∼30 μm of skull remained. All imaging was performed using a Leica SP5 2-photon microscope system (Leica Microsystems, Mannheim, Germany) equipped with a picosecond laser (Coherent, Santa Clara, CA) and external nondescan detectors that allow simultaneous detection of emissions of different wavelengths and second harmonic signals (secondary harmonic generation ∼ 460 nm). Enhanced GFP, YFP, and Qdots were excited using a 920-nm laser. Images were obtained using a 20× water-dipping lens. Four-dimensional imaging data were collected by obtaining images from the x-, y-, and z-planes, with a z- thickness of 68 μm and step size of 4 μm to allow for the capture of a complete z-series every 20 s for a period of 15 min. The resulting images were analyzed with Volocity (PerkinElmer, Waltham, MA) or Imaris software. Videos of T cell migration, mean migratory velocity, and cell contact duration were calculated using the software.

Statistical analyses were performed using Prism software. The p values <0.05 were considered significant. The tests used in each experiment are denoted in the figure legends.

To characterize Treg responses during chronic toxoplasmosis, mononuclear cell preparations were isolated from the spleen, cervical lymph nodes (CLNs), and brains of mice infected with T. gondii. The numbers and phenotype of these populations were assessed using flow cytometry for intracellular Foxp3 (Fig. 1). Based on the use of MHC class I and II tetramers that contain defined parasite epitopes, activated Ag-specific CD4⁺ and CD8⁺ effector cells were readily detected in the spleen and CLNs, and the Tregs represented 12–18% of the total CD4⁺ population (data not shown). In the brains of uninfected mice, T cells are confined to the meningeal spaces (28, 34), and BMNC preparations contained very few conventional effector/memory T cells, Tregs, or DCs (Fig. 1A, data not shown). During the early stages of infection, a sizable population of effector CD4⁺ T cells was recruited to the brain by day 14 (Fig. 1A). Tregs made up ∼1.5% of the CD4⁺ T cell compartment at days 14 and 21 postinfection. During chronic infection (day 28 postinfection), Tregs made up ∼8–10% of the total CD4⁺ T cell population, and these cells express high levels of CD25 (Fig. 1B). To further characterize this population, IL-10–eGFP reporter mice (VertX or Tiger mice) were infected, and IL-10 and IFN-γ production was assessed and compared with basal IL-10 production in the naive spleen. We observed negligible IL-10 production from conventional CD4⁺ T cells in the spleens of naive and infected mice. In contrast, Tregs in the spleen did produce some IL-10 at baseline (∼5%) that increased to nearly 20% during infection (Fig. 1C). In the infected CNS, ∼4% of Foxp3⁻ CD4⁺ cells were eGFP⁺, whereas 45% of Tregs in the CNS expressed IL-10–eGFP (Fig. 1C). Although the numbers of Foxp3⁺ and Foxp3⁻ IL-10–producing cells are similar, the mean fluorescence intensity of IL-10–eGFP was higher in the Tregs isolated from the infected brain (Fig. 1D). Analysis of IFN-γ production revealed that, although 57% of effector CD4⁺ T cells in the infected brain produced IFN-γ, only 12% of the Tregs at this site produced IFN-γ (Fig. 1C). This analysis also highlighted the presence of a small subpopulation of Tregs that produce both cytokines in the brain, which was not observed in peripheral tissues at this time point (Fig. 1C). Consistent with the highly polarized Th1 response that is generated against T. gondii, during TE, Tregs in the brain express the canonical Th1 transcription factor T-bet (Fig. 1E) and the T-bet–dependent chemokine receptor CXCR3 (Fig. 1F, 1G). Moreover, Tregs isolated from the infected CNS express higher levels of CXCR3 in comparison with effector CD4⁺ and CD8⁺ T cells (Fig. 1F), and CXCR3 expression on CNS Tregs was higher in comparison with Tregs isolated from the spleen and CLNs of infected mice (Fig. 1G). Together, these data demonstrate that a Th1-polarized Treg is recruited to the CNS during chronic T. gondii infection.

During infection with T. gondii, parasite-specific CD4⁺ and CD8⁺ T cells, as well as Tregs, accumulate in the CNS; however, little is known about the specificity of the local Treg populations. To address whether the Tregs may be specific for parasite Ag, immune cells isolated from the CNS were stained with an MHC class II tetramer reagent specific for the AVEIHRPVPGTAPPS peptide. Parasite-specific CD4⁺ effector cells were detected, but very few CD4⁺ T cells were tetramer⁺ and Foxp3⁺ (0.018 ± 0.0059%) (Fig. 2A). Although this analysis was performed using a single tetramer, these results suggest that Tregs in the CNS may not recognize the same Ags as do the Toxoplasma-specific effector CD4⁺ T cells. To further explore the clonality and potential specificity of the Tregs in the CNS, effector and regulatory CD4⁺ T cells were sorted from meninges of infected mice, and the CDR3 regions of the TCR-β gene were sequenced. The TCR sequences of effector T cells and Tregs from three individual experiments were compared at the amino acid level. First, the similarity of the TCR sequences was compared between samples, where the presence of a sequence, but not the frequency of the sequence within the population, was considered. Although many unique sequences were detected in each population, the greatest similarity was found in the effector CD4⁺ T cell populations (Fig. 2B, 2C, Supplemental Table I). When comparing TCR sequences from effector T cells and Tregs, we found very little overlap (∼1%) between the populations (Fig. 2D), which is in agreement with results from MHC class II tetramer staining. Moreover, if a sequence was detected in both effector and Treg populations, the number of reads for that sequence was overrepresented primarily in the Treg population (Fig. 2E). These results suggest that Tregs found in the CNS during chronic infection are rarely from the same clonal lineage as the effector population, and some Tregs may lose Foxp3 expression and resemble effector T cells. Unexpectedly, we found that the CDR3 regions of Treg TCRs had very little sequence overlap (0.2–0.65%) among three separate T. gondii infections (Fig. 2B, 2C, 2F), suggesting that the Treg repertoire recruited to the CNS during infection varies greatly among experimental infections. Although the TCR sequences do not identify the Ag specificity of the Tregs, these experiments reveal that, during each individual infection in C57BL/6 mice, a unique repertoire of Tregs that is largely distinct from effector cells is recruited to the CNS.

To understand the spatial organization of Tregs during TE, immunohistochemistry was performed for CD4, parasite Ag, Foxp3, and laminin to identify basement membranes and, thus, demarcate the meninges and blood vessels and identify CD4⁺ T cell location within the brain. Staining for CD4 revealed the presence of T cells in the meninges and perivascular cuffs; however, the largest numbers of effector T cells were present in the parenchyma (∼75.7%), where parasite cysts and replicating tachyzoites are found (Fig. 3A, 3E). In contrast, Foxp3⁺ cells were largely (89%) localized to the inflamed meninges and perivascular cuffs and, thus, were more rare in the brain parenchyma in comparison with effector CD4⁺ T cells (Fig. 3B–E). Together with the data presented in Fig. 1, these data indicate that the Tregs present in the CNS during TE are Th1 polarized like the effector population, but their distribution within the brain is distinct from effector CD4⁺ Th1 cells.

Because DC populations are critical regulators of Treg homeostasis and are known targets of Treg suppression (2, 5, 35), Foxp3-GFP mice were crossed with CD11c-YFP reporter mice, and mice expressing both reporters were used for intravital imaging studies (Fig. 4A, Supplemental Video 1). First, these studies showed that there was close association of Foxp3⁺ cells with blood vessels in the CNS of mice with TE and that these cells were frequently colocalized with CD11c⁺ cells (Fig. 4A), confirming the results obtained with immunohistochemistry (data not shown). Next, imaging of Foxp3⁺ cells in the CLNs of infected mice revealed that they were highly motile and had many short-lived interactions with CD11c⁺ cells (Fig. 4B, 4C, Supplemental Video 2), similar to the behavior of Tregs in the lymph nodes of uninfected mice (3). However, Tregs localized within the meninges of infected mice had a slower velocity and formed long-lived associations with CD11c⁺ populations in this microenvironment in comparison with the CLN (Fig. 4D, Supplemental Video 3). In contrast with naive T cell interactions with DCs during priming (36), Foxp3⁺ cells were not stationary while interacting with CD11c⁺ cells in the meninges; they were frequently observed to move from one CD11c⁺ cell to the next while maintaining contact with these cells (Supplemental Video 3). It is relevant to note that, when CD4⁺ T cells from CFP-expressing mice were activated and expanded in vitro and transferred to infected Foxp3-GFP reporter mice, this CD4⁺CD44^hiCD62L^loFoxp3⁻ population was present in the same area as Tregs but remained highly migratory (Fig. 4E, Supplemental Video 4) and did not form sustained interactions with CD11c-YFP cells (data not shown). These data highlight that Tregs in the CNS during TE display a pattern of behavior that is distinct from Treg populations in the CLNs or effector CD4⁺ T cells present in the meninges (Fig. 4F).

The reduced migratory phenotype of Tregs observed in the CNS during TE may be explained by prolonged interactions with CD11c⁺ cells. Previous studies have demonstrated that treatment with anti–LFA-1 Abs leads to a loss of DCs in the CNS (37). Indeed, treatment with anti–LFA-1 Abs for 4 h leads to a significant decrease in DCs (CD3⁻CD19⁻NK1.1⁻CD11c⁺MHCII^hi) in the meninges (Fig. 5A). To determine whether Treg interactions with DCs in vivo affects their migratory behavior, mice were treated with LFA-1–blocking Abs 4 h prior to imaging (Supplemental Video 5). The loss of DCs resulted in a reduced duration of contact between Tregs and the remaining CD11c⁺ cells, with fewer Tregs maintaining contact for the duration of the imaging period and more cells making contacts of short duration (Fig. 5B). The average contact time was reduced significantly from 11.5 to 8.5 min. Moreover, Ab blockade resulted in a significant increase in the speed of Foxp3⁺ cell migration from to 2.2 to 3.3 μm/min (Fig. 5C). Together, these results suggest that interactions between Tregs and CD11c⁺ cells limit the migratory speed of Tregs in the CNS.

Multiple studies have associated the presence of Tregs in the CNS with the ability to limit inflammation in the context of infection (West Nile virus, murine hepatitis virus, and coronaviruses) and autoimmunity (experimental autoimmune encephalomyelitis [EAE] and multiple sclerosis) (10–14, 38, 39), but little is known about the localization and behavior of Tregs in these different disease settings. In addition, in many cases it has been difficult to discern local effects within the CNS from a role for this regulatory population in peripheral events (40). The studies presented in this article reveal that during TE, Foxp3⁺ T cells were restricted to the perivascular spaces and the meninges, unlike effector CD4⁺ T cells. Precedent exists for this observation, as Campbell and colleagues (41) reported a similar localization of Tregs in the CNS in a model of EAE and suggested that signaling through CXCR3 is critical to prevent Treg entry into the inflamed brain parenchyma. Consistent with this idea, Tregs present in the CNS during TE express high levels of CXCR3 and we have observed that, in CXCR3-knockout mice infected with T. gondii, the localization of Tregs is altered and that these populations are now present in the brain parenchyma (T.H. Harris, unpublished observations). Thus, although recent studies have proposed that CXCR3 expression on Tregs during Th1 inflammation allows these populations to access sites of Th1 inflammation (42), these observations suggest that CXCR3 is not only involved in entry to the tissue but can also regulate Treg location within the CNS.

Whether the exclusion of Tregs from the brain parenchyma is biologically relevant remains unclear, but this may be one mechanism that allows effector T cells present in the brain parenchyma to operate independently of the suppressive effects of Tregs and therefore to control parasite replication more effectively. In previous reports, the use of IL-2 complexes to expand Treg populations in infected mice led to an increased parasite burden in the CNS (21, 25), further suggesting an important role for Tregs in controlling effector T cell responses and resulting parasite burden. There are several possible ways in which Tregs in perivascular sites might influence parasite-specific effector responses in the parenchyma. In models of EAE and viral encephalitis, the ability of effector T cells to interact with APCs within these perivascular compartments allows effector T cells to proliferate and be retained in the CNS (43, 44). The ability of Tregs to limit DC activity at this site of T cell entry into the parenchyma of the brain, perhaps through the production of IL-10, may allow Tregs to serve as “gatekeepers” to the CNS.

Recent studies have shown that TCR signaling is required for Treg suppressor capacity in vivo (45), but it remains unclear whether Treg populations in the brain during TE are specific for parasite-derived Ags or whether these are self-reactive populations. Although reagents to detect parasite-specific CD4⁺ T cells are limited to a single tetramer reagent, our results do not indicate that Tregs are specific for this parasite Ag. Moreover, the results from TCR sequencing performed in this study suggest that, if Tregs are indeed parasite specific, they rarely share a clonal lineage with effector T cells. We also found that, when a TCR sequence is shared between effector T cells and Tregs, more reads were typically detected in the Treg population, which may suggest the loss of Foxp3 expression and acquisition of an effector T cell phenotype. In addition, unique Treg clones were identified in the CNS in each independent experimental infection, suggesting that variable(s) other than infection shape the repertoire of Tregs in the CNS. Our results are in agreement with several studies that also did not detect overlap in TCR sequences between regulatory and effector CD4⁺ T cell populations in diverse settings of tissue inflammation (46–49).

Several observations suggest that the unique behavior of Tregs in the CNS (compared with activated effector CD4⁺ T cells in the same areas or Tregs in secondary lymphoid organs) may be a function of the sustained interactions with CD11c⁺ populations. Indeed, several in vitro studies have also examined the influence of Tregs on DCs and found that Tregs decrease levels of MHC class II and expression of costimulatory molecules (18, 35, 50). This observation is consistent with early studies that demonstrated that the ability of Tregs to interact with CD11c⁺ cells is central to Treg suppression of effector T cell responses (3, 5, 51). In our studies, the loss of DCs from the meninges led to increased Treg velocity, suggesting that interactions between these two cell types influence Treg behavior at this site. The Treg interaction with DCs in this space may be important for suppressing DC function by downregulating the costimulatory capacity of these APCs or their production of cytokine, thereby regulating local effector T cell responses. Alternatively, recent studies have also demonstrated that DCs in peripheral tissues provide survival signals for Tregs that differ from those in the lymph nodes (52) and that these cells may promote the survival of Tregs in the brain. If these interactions are disrupted long-term, it is possible that Treg survival could be affected. Regardless, the ability to visualize how Tregs behave in inflamed tissues and determine the cell types with which they interact provides insight into how these cells operate to limit inflammatory processes and a better understanding of how local effector responses are regulated at sites of inflammation.

We thank Gordon Ruthel and the Penn Vet Imaging Core for assistance with microscopy experiments, Vijay Kuchroo and Christopher Karp for kindly providing mouse strains, and the members of our laboratories for feedback during the development of this manuscript.

The authors have no financial conflicts of interest.