Two type II strains of Toxoplasma were used to allow the quantification of parasite-specific T cells and to maximize the ability to see cysts in the brain. First, a strain engineered to secrete ovalbumin (Pru-OVA) (29) was maintained in vitro in human foreskin fibroblasts (HFF) grown in complete DMEM (90% DMEM, 10% fetal bovine serum, 1% penicillin/streptomycin). After infecting HFFs, parasites were grown in D10 media (70% DMEM, 20% M199, 10% fetal bovine serum, 5% penicillin/streptomycin, 5% gentamycin) with chloramphenicol. Parasites were purified for infection by passage through a 22.5-gauge needle followed by passage through a 5.0-mm nylon filter. After centrifugation at ~2,000 × g for 10 min at 4°C, parasites were counted and resuspended in an appropriate volume of 1× PBS. Ten thousand tachyzoites were intraperitoneally (i.p.) injected in 200 µL PBS. Second, the Me49 strain of parasites was maintained in CBA mice. For infections, brains were harvested from chronically infected mice and homogenized in 3 mL of PBS by needle passage. After counting, cysts were resuspended in an appropriate volume of PBS to infect i.p. at 20 cysts per mouse in a 200 µL volume. C57Bl/6 and CBA mice were obtained from the Jackson Laboratory (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). Mice were maintained in a pathogen-free environment under IACUC established protocols at the University of California, Riverside.

Before harvest, mice were intracardially perfused with 20 mL ice-cold PBS, with perfusion confirmed by the white appearance of the brain and lack of red blood cells during tissue processing. Mononuclear cells were isolated from the brain by mincing and subsequent homogenization via passage through an 18-gauge needle in complete RPMI (86% RPMI, 10% FBS, 1% penicillin/streptomycin, 1% l-glutamine, 1% NEAA, 1% sodium pyruvate <0.01% β-mercaptoethanol). The resulting suspension was incubated at 37°C with 3 mg DNAse and 100 µg collagenase for 1 h and 45 min. After incubation, the suspension was passed through a 70-µm strainer, and mononuclear cells were isolated using a density gradient spun at 2,000 rpm for 25 min with no brakes. The density gradient consisted of a 60% Percoll solution in cRPMI overlayed with a 30% Percoll solution in PBS. Brain mononuclear cells (BMNCs) were isolated from the interphase, counted and washed in FACS buffer, blocked for 10 min in F_c block (BD Biosciences), and then incubated with a panel of antibodies against CD3, CD8, CD4, CD69, and CD103 (eBioscience) for 30 min protected from light. Samples were then washed and fixed in 4% PFA. Dextramer staining was performed as follows: Prior to surface staining, samples were incubated with ovalbumin-specific (SIINKFEKL) MHC I dextramer (Immudex) at room temperature for 45 min, protected from light. The finished samples were resuspended in FACS buffer and analyzed using a FACS Canto from BD Biosciences. Flow cytometry analyses were performed using FlowJo 10.1, and statistical analyses were performed using Prism 6.

For sorting of cell populations for RNA-Seq, BMNCs were isolated as described above. Splenocytes were isolated as follows: spleens were homogenized in a 40-µm strainer using the blunt end of a 3-mL syringe. The homogenate was washed with cRPMI, and red blood cells were lysed using ACK buffer from Lonza. Cells were then washed and counted. CD8⁺ T cells were isolated from spleen and BMNCs using negative selection columns from R&D systems. The purified CD8⁺ T cells were then incubated with anti-CD103 antibodies according to the protocol described above. Cells were suspended at a concentration of 3 × 10⁶ cells/mL and then sorted using a FACS Aria from BD Biosciences. Cell sorting was performed at the Institute for Integrative Genome Biology at UC Riverside.

Immediately following excision, sagitally bisected brain tissue was flash-frozen in a bath of isopentane cooled with dry ice. Frozen organs were then put into a standard Tissue-Tek cryomold, filled with optimal cutting temperature solution (also from Tissue-Tek), put on dry ice, and subsequently stored at −80°C. Serial sections of 12 µm were prepared on a standard Cryostat machine (LEICA/CM1850). For in situ immunofluorescent staining, frozen tissue sections were fixed in 75% acetone/25% ethanol and then blocked for 10 min in 10% donkey serum prior to staining. For the primary antibody incubation, conjugated anti-mouse CD8 and CD103 were added to three panels containing anti-mouse biotinylated E-cadherin, anti-mouse laminin, or anti-T. gondii. Antibodies against CD8, CD103, and E-cadherin were from eBioscience, antibodies against laminin from Cedarlane Laboratories, and anti-Toxoplasma from Abcam. The sections were incubated for 3 h at room temperature. After washing, the sections were incubated for 1 h at room temperature with appropriate secondary antibodies (streptavidin for anti E-cadherin). Samples were mounted in ProLong Gold with DAPI (Invitrogen) for nuclear counterstaining. Images were collected on either a Leica DMI 6000B epifluorescent or a Leica SP5 scanning confocal microscope (Leica Optics), and data were analyzed and quantified using Volocity 6.1 (Perkin-Elmer).

To obtain sufficient RNA for analysis, CD8 T cells were isolated from pooled leukocytes of brain and spleen of at least n = 5 mice. This was repeated to obtain four trials to serve as biological replicates in analysis. The isolated CD8 T cells were sorted according to expression of CD103, and RNA was extracted from sorted groups using an RNeasy Mini Kit from Qiagen. Multiplexed cDNA libraries were generated, which included brain CD103⁺, brain CD103⁻, spleen CD103⁺, and spleen CD103⁻ CD8 T cells. Of the four trials, libraries of sufficient quality for sequencing could not be generated for two brain CD103⁺ samples and one brain CD103⁻ sample. RNA concentration and quality were analyzed by running an RNA Nano chip on an Agilent 2100 Bioanalyzer. mRNA was enriched using a RiboGone Mammalian kit from Clontech. cDNA libraries were generated using a SMARTer stranded RNA-Seq kit, also from Clontech. Concentration and quality of libraries were measured with a High Sensitivity DNA chip (Agilent). The resulting libraries were multiplexed, and single-end 50 bp sequencing was performed at the UC Riverside Genomics Core facility using the Illumina HiSeq 2500. After demultiplexing and QC of the resulting FASTQ files, alignment and differential gene expression analysis were performed with the systemPipeR workflow¹ (30). Quality control and trimming were conducted using FastQC² and Trim Galore,³ respectively. The reads were then aligned to the mouse genome using Tophat2 (31), and transcripts per gene were counted using GenomicRanges package in R (32). Differential gene expression analysis was performed following TMM normalization using edgeR (30, 33). MetaCore software (Thomson Reuters) was used for functional enrichment analysis. To conduct functional enrichment analysis, only DEGs with a fold change (FC) >2 and FDR <5% were used. We further excluded genes with very low counts by only selecting gene with a mean reads per kilobase of transcript per million mapped reads (RPKM) >1. RPKM served as a normalized value for read count and was obtained via GenomicRanges.

Microarray data from the studies by Wakim et al. and MacKay et al. were obtained from the NCBI Gene Expression Omnibus database under accession number GSE39152 (25) and GSE47045 (26). Differential gene expression analysis was conducted using GEO2R. Genes with a corrected p < 0.05 were used for subsequent analyses, including functional enrichment analysis. For the Wakim et al.’s data, differential gene expression analysis was conducted between brain CD8⁺ CD103⁺ (n = 5) and brain CD8⁺ CD103⁻ (n = 3) samples. For MacKay et al. data, differential gene expression analysis was conducted between skin T_RM (n = 3) and spleen T_CM plus spleen T_EM (n = 6) samples.

Brain mononuclear cells were collected from mice by 5 weeks postinfection and incubated for 6 h in complete T cell media (86% DMEM, 10% FBS, 1% penicillin/streptomycin, 1% l-glutamine, 1% NEAA, 1% sodium pyruvate, <0.01% β-mercaptoethanol) with 10 µg/mL BFA. Cells were treated with 10 µg/mL αCD3/αCD28 (BD Biosciences) antibody or media alone. After restimulation, cells were collected, and surface staining was performed as described above. Intracellular cytokine staining was performed with the FoxP3/transcription factor staining buffer kit from eBioscience. Antibodies against IFN-γ and TNF-α were obtained from eBioscience. Flow cytometry analyses were performed using FlowJo 10.1, and statistical analyses were performed using Prism 6.

T_RM are defined as a persisting population of T cells that remain localized to the tissue, that is, non-recirculating. To determine whether such a population exists in the brain during chronic T. gondii infection, we investigated whether a stable T_RM phenotype could be observed across the course of infection and whether such a phenotype would localize to the parenchyma. Therefore, we enumerated the distribution and kinetics of CD8 T cells in the brain expressing the residency markers CD103 and CD69. There are no T cells in a naïve uninfected brain (1, 34); however, by 2 weeks postinfection, representing the late acute stage, there is a small proportion (2 ± 0.4%) of CD8 T cells exhibiting a CD103⁺ CD69⁺ phenotype. This population significantly increases (p < 0.0001) over the course of infection, and at the late chronic stage (12 weeks postinfection), 38 ± 1% of CD8⁺ T cells were observed to be positive for CD103 and CD69 (Figures 1A,B). Thus, by the late stages of chronic infection, this phenotype forms a substantial percentage of the CD8 subset in the brain relative to all other phenotypes contained in the CD103⁻ group. In the periphery, CD103⁺ cells are observed to accumulate in the spleen and lymph node over time; however, such cells did not express CD69 in significant proportions (Figure S1 in Supplementary Material) and are more transcriptionally consistent with a quiescent/naïve population (Figure S2 in Supplementary Material) (35). In contrast, the brain has no population of CD103⁺ CD8 T cells that do not express CD69 (Figure 1A), and therefore, the use of this marker for subsequent analysis was redundant.

Previous work has suggested that the majority of T cells in the brain are specific for the parasite, and there is little evidence of a bystander population (36). However, to rule out the possibility that CD103-expressing cells are not directly responding to parasite infection, we used parasites engineered to express ovalbumin along with ovalbumin-specific (SIINFEKL) MHC I dextramer (29, 36) to test the antigen specificity of this population. At 12 weeks postinfection, approximately, 3% of CD8⁺ cells in the brain are bound by the SIINFEKL dextramer, a significant proportion of which express CD103 (Figure 1C). These data do not definitively establish that all CD103⁺ CD8 T cells in the brain are parasite specific nor does it indicate the full repertoire of parasite antigens to which CD103⁺ CD8 T cells could respond. Nevertheless, it does support the view that brain CD103⁺ CD8 T cells are not simply a bystander population and is consistent with the previous work demonstrating that retention of T cell populations in the brain requires cognate antigen at this site (Figure S7 in Supplementary Material) (36, 37).

To determine whether CD103⁺ CD8 T cells localize to distinct environments within the brain, we used in situ immunofluorescent staining. Expression of CD103 by CD8 T cells is observed at entry points to the brain, including the choroid plexus, where it associates with the CD103 ligand E-cadherin (Figure 2A). In addition, this population can be found within the perivascular space (Figure 2B) and within the parenchyma, despite little expression of E-cadherin at these sites (38). Finally, this population can be observed in proximity to cysts and individual parasites (Figure 2C). Quantification of the distance of CD103⁺ cells from cysts was highly variable (80–160 µm), and there was no correlation between cyst and CD103⁺ cell number (Figure S3 in Supplementary Material).

Together these data indicate that a population of CD103⁺ CD8 T cells can be observed throughout the chronic stage and form a substantial proportion of the CD8 subset in the infected CNS. In situ immunofluorescent staining revealed that CD103 expression is not confined to a certain area of the brain or specifically localized to the regions of infection. Our observation that CD103⁺ CD8 T cells are present within the perivascular space suggests that this phenotype may be acquired before entry into the parenchyma. Together, these data raised the possibility that CD103 did not define a homogenous population of T cells within the brain but rather a functionally diverse set of cells having in common expression of CD103.

The heterogeneity of infiltrating T cells during chronic infection (8, 36, 39, 40) required we exclude the possibility of CD103 expression masking a group of functionally and phenotypically diverse cells that are not T_RM. Furthermore, expression of CD103 is not the limiting phenotype for T_RM, that is, a memory T cell can be resident (non-recirculating) without expressing CD103 (13). To address whether CD103 defines a transcriptionally distinct population within the brain that is uniquely characteristic of T_RM, RNA-Seq analysis was performed on sorted cells from infected mice at 4 weeks postinfection.

Three comparisons between CD8⁺ subsets were carried out (Figure 3A). The first comparison, brain CD103⁺ to brain CD103⁻, specified whether CD103 defined a transcriptionally distinct population within the brain. The second comparison, brain CD103⁺ to spleen CD103⁺, specified whether CD103 defined a transcriptionally distinct CD8 T cell population when compared to a putatively naïve peripheral population of the same phenotype. Finally, the brain CD103⁺ to spleen CD103⁻ comparison specified whether CD103 defined a transcriptionally distinct population of CD8 T cells when compared to a putatively activated peripheral population. The comparison of brain CD103⁺ to spleen CD103⁻ yielded the greatest number of DEGs (2,350) (Figure 3A). This provided preliminary evidence that the brain CD103⁺ CD8 T cell group is transcriptionally dissimilar to both local and peripheral CD8 T cells. To confirm these data, principal component analysis was performed to determine the variance in RPKM across all groups (Figure 3B). The resulting plots demonstrate that the brain CD103⁺, brain CD103⁻, spleen CD103⁺, and spleen CD103⁻ populations are transcriptionally distinct from each other. PC1 appears to define the organ, spleen versus brain, whereas PC2 appears to define CD103 expression. PC1 had a loading of 12.90, and PC2 had a loading of 14.14. The most variable gene in PC1 was transthyretin, whereas the most variable gene in PC2 was S1pr5 (Figure S4 in Supplementary Material). These data lend support to CD103 expression defining a distinct and homogenous population of CD8 T cells within the chronically infected brain.

Next, DEGs were analyzed between CD103⁺ CD8 T cells and CD103⁻ CD8 T cells within the brain. Functional enrichment analysis revealed that the DEGs enriched to GO terms such as cell migration, proliferation, and differentiation (Figures 4A,B). These processes are critical aspects of T cell response to infection, T cell signaling, and activation. In addition, these DEGs had significant overrepresentation of notable disease terms—“RNA virus infection” (Figures 4A,C) and “arteriosclerosis” (Figures 4A,D). In support of the view that this is a memory population, the terminal differentiation marker Klrg1, expression of which defines terminally differentiated effector cells, is downregulated in the brain CD103⁺ population (FC = −3.5), whereas Il-7r, a receptor that defines memory cells is upregulated (FC = 2.1) (Figures 4C,E). Furthermore, consistent with tissue residency is the downregulation of Klf2 (FC = −2.5) and S1pr1 (FC = −4.4) in the brain CD103⁺ population (Figures 4B–E), inhibiting recirculation of cells back to lymphoid organs (Schenkel and Masopust). To validate differential expression, flow cytometry of BMNCs was conducted. This confirms downregulation of both S1PR1 and KLRG1 at the protein level in CD103⁺ CD8⁺ T cells (Figure 4F). The percentage of CD103⁻ S1PR1⁺ CD8 T cells is approximately threefold higher that of the CD103⁺ S1PR1⁺ subset (0.88 versus 3.32%) (Figure 4F). A total of 37.50% of CD103⁻ CD8 T cells were KLRG1⁺, whereas 1.05% of CD103⁺ CD8 T cells were KLRG1⁺ (Figure 4F).

Additional changes of interest in the brain CD103⁺ population included upregulation of the co-inhibitory receptor Ctla4 (FC = 2.14) and the transcription factor Irf4 (FC = 2.1) and downregulation of the transcription factor Atf3 (FC = −2.6) (Figures 4C–E). Ctla4 is among the first co-inhibitory receptors to be expressed and could provide evidence of increased antigen engagement in this subset (41). Although several co-inhibitory receptors were upregulated in the brain CD103⁺ group compared to the spleen CD103⁺ group, including Pdcd1 (FC = 77.7), Lag3 (FC = 9.2), and Tigit (FC = 27.7) (not shown), only Ctla4 was uniquely upregulated relative to other CD8 T cells in the brain.

A set of genes associated with extracellular adhesion and migration were upregulated in the brain CD103⁺ population, including Cdh1 (E-cadherin, FC = 2.9), Adam8 (FC = 2.3), Swap70 (FC = 2.8), Sema6d (FC = 3.7), and Lpar3 (FC = 13.9) (Figures 4B,E). Excluding Sema6d, these are novel upregulated genes in T cells (42–45). Both Sema6d and the E-cadherin gene were also upregulated relative to peripheral spleen CD8 T cells (both CD103⁻ and CD103⁺), supporting the possibility that these genes are unique determinants of residency for the brain CD103⁺ population in our model (Figure S5 in Supplementary Material). This was also the case for the downregulated genes S1pr1 and Klf2 (Figure S5 in Supplementary Material).

The original studies of T_RM in the brain by Wakim and colleagues included microarray analysis to generate a transcriptional profile of T_RM in a VSV model of acute and resolved brain infection (25). We compared DEGs from this model with the RNA-Seq data generated in our chronic model, focusing specifically on brain CD103⁺ CD8 T cells relative to brain CD103⁻ CD8 T cells. The results of our Venn analysis reveal a set of 27 genes (~5%) common to both transcriptional profiles (Figure 5A). S1pr1, Klf2, and Adam8 are contained within this set, suggesting similar mechanisms of residency between the two models (Figure 5B). Swap70 and Eomes are also represented in this group. A remaining 70 genes are unique to VSV, and a remaining 393 genes are unique to T. gondii. The top enrichment terms for these sets of genes were “immune system process” and “cell adhesion” (Figures 5C,D, respectively). When our data were compared to data from skin T_RM generated from HSV infection (26), we observed a similar set of 33 differentially expressed genes, comprising 4.1% of the total DEGs (Figure S6 in Supplementary Material). Although the majority of differentially expressed genes in each model were unique to the model, a core profile of genes associated with tissue residency was common to all models (Figure 5; Figure S6 in Supplementary Material).

Several cytokines and receptors including Tgfb3, Il1a, Il1b, Il22, and Csf1r are downregulated in brain CD103⁺ CD8 T cells relative to brain CD103⁻ CD8⁺ T cells (Figures 4B–D). The TGF-β receptor was upregulated (FC = 2.75) in brain CD103⁺ CD8 T cells, consistent with previous reports of the requirement for TGF-β in generation of this population (46, 47). Notably, the pro-inflammatory cytokine Tnf is significantly upregulated in CD103⁺ CD8 T cells (FC = 2.5 relative to brain CD103⁻) (Figures 4B–D; Figure S5 in Supplementary Material). T_RM are reported to produce this cytokine although their ability to do so relative to other CD8 T cell subsets has not been previously validated via flow cytometry (16, 25, 26).

To validate and confirm the increase in Tnf gene expression in the CD103-expressing cells, BMNCs from chronically infected mice were restimulated and production of IFN-γ and TNF-α was compared between the CD103⁺ and CD103⁻ subsets. The proportion of CD103⁺ T cells producing both IFN-γ and TNF-α was significantly greater than cells not expressing CD103 (19.6 ± 5.3 versus 9.3 ± 3.7%; p < 0.01) (Figures 6A,B). Furthermore, of CD103⁺ CD8 T cells expressed more total TNF-α and more total IFN-γ on a per cell basis than their CD103⁻ counterparts (for IFN-γ: 3,824 ± 787% versus 3,208 ± 746 MFI; p < 0.05) (for TNF-α: 8,778 ± 2,165% versus 5,966 ± 1,171 MFI; p < 0.05) (Figures 6C,D). Since both TNF-α and IFN-γ are critical for protection against toxoplasmic encephalitis (TE) (5, 48), these data suggest that the CD103⁺ T_RM population contributes to protection within the brain at the chronic stage.