Migration of CD8⁺ T Cells into the Central Nervous System Gives Rise to Highly Potent Anti-HIV CD4^dimCD8^bright T Cells in a Wnt Signaling–Dependent Manner Free

The brain is not an immune-privileged site. Lymphocytes survey the brain, and in the context of inflammation and/or infection, they home to the brain in greater numbers (1). In HIV infection, CD8⁺ T cells home to the brain, as demonstrated in human postmortem HIV-infected brains and in the brains of SIV-infected macaques (2–12). The consequence of CD8⁺ T cell–mediated anti-HIV responses in the brain is unclear. On the one hand, CD8⁺ T cells control HIV infection in the brain (4, 6, 11), whereas on the other hand, heightened anti-HIV responses are likely to lead to neuronal injury.

Prior to entering the brain, CD8⁺ T cells differentiate from naive cells to effector and eventually effector memory, central memory, or terminal effector memory cells (13). Considerable data from our laboratory (14–18) and others (19–26) described a unique subset of highly activated mature CD8⁺ T cells, which dimly express CD4 on their surface. This population, known as CD4^dimCD8^bright T cells, constitutes ∼3–5% of total CD8⁺ T cells and 1–3% of all PBLs (24, 25). These CD4^dimCD8^bright T cells are not prematurely released thymocytes, as they are negative for the thymocyte marker CD1a (18). They have an αβ TCR and an αβ CD8 molecule (17). They are negative for CD16, CD56, and 6B11 and therefore are not NKT cells (14). Further, we have shown that CD4^dimCD8^bright T cells are highly enriched in antiviral responses to both HIV and CMV (15). CD4^dimCD8^bright T cells constitute ∼60% of the anti-HIV tetramer responses (15) and are polyfunctional, as determined by coexpression of IL-2, IFN-γ, TNF-α, or surface expression of CD107ab in peripheral blood in response to pooled HIV peptides (15).

β-Catenin, the central mediator of the Wnt/β-catenin pathway, mediates CD4 expression on the surface of mature CD8⁺ T cells (18). The brain and especially astrocytes, which constitute 40–60% of the brain resident cells, are a rich source for Wnt ligands (27). Wnt ligands are small secreted evolutionarily conserved glycoproteins, some of which mediate a signal transduction cascade that culminates in stabilization of β-catenin and consequently its association with members of the T cell factor/lymphoid enhancer–binding factor transcription factors and binding to target genes to regulate gene expression (18). Further, Wnts promote communication between cells, leading to many cellular processes, including development, proliferation, survival, regeneration, wound healing, and stress (28–30).

In recognition of robust secretion of Wnts in the brain, we investigated their role in the generation of the CD4^dimCD8^bright T cell phenotype in the CNS and compared the role of CD8 single positive (CD4^negativeCD8^bright) versus CD4^dimCD8^bright T cells in HIV control in the brain. We used NOD/SCID/IL-2rcγ^−/− mice reconstituted with human PBMCs (NSG-huPBMC) to conduct these studies. We show that all three populations are present in the brain and that both CD4 single positive T cells (e.g., CD4^brightCD8^negative) and CD4^dimCD8^bright T cells are HIV infected in the CNS. The Wnt-rich environment in the CNS induces CD8 single positive T cells to become CD4^dimCD8^bright T cells both in vitro and in vivo. Most importantly, we show that it is the CD4^dimCD8^bright T cells, not the CD8 single positive T cells, that control HIV in the brain. Brain CD4^dimCD8^bright T cells of HIV-infected NSG-huPMBC mice are cytotoxic and are terminally differentiated but not senescent in comparison with uninfected mice. These findings point to a significant role of CD4^dimCD8^bright T cells in HIV control in the CNS and demonstrate that the brain microenvironment is capable of altering CD8⁺ T cell phenotype to generate a potent antiviral CD8⁺ T cell subset.

Research involving human subjects was conducted in accordance with institutional (IRB-L06080703) and U.S. government guidelines on human research.

NOD/SCID/IL-2rcγ^−/− and NOD.Cg-Prkdc^scidIL2rγ^tm1Wjl/SzJ (NSG) mice were purchased from The Jackson Laboratory (Bar Harbor, ME; stock number 005557) and housed under pathogen-free conditions in accordance with the Institutional Animal Care and Use Committee (IACUC #14-014) at Rush University Medical Center and the ethical guidelines for care of laboratory animals at the National Institutes of Health.

PBMCs were isolated from 12 healthy seronegative donors by Ficoll–Hypaque density gradient centrifugation. CD8⁺ T cells were isolated using the untouched CD8⁺ T Cell Isolation Kit II according to the manufacturer’s instructions (Miltenyi Biotec, Cambridge, MA). CD8⁺ T cells were suspended in complete RPMI 1640 medium (cRPMI; Lonza BioWhittaker, Walkersville, MD), which includes 10% heat-inactivated FBS (Gemini Bio Products, Calabasas, CA), and 1% penicillin/streptomycin (Sigma-Aldrich, St. Louis, MO). Cultures were stimulated with 20 U/ml IL-2 (AIDS Reagent Program, Germantown, MD) and, when indicated, 1 μg/ml soluble α-CD3/α-CD28 Abs (BD Biosciences, San Jose, CA) and maintained for 24 h–6 d, as indicated.

PBMCs were isolated by density gradient centrifugation from human donor venous blood. NSG mice, 6–8 wk old, were injected i.p. with 2 × 10⁷ human PBMCs. At 1–2 wk after reconstitution, mice were bled by retro-orbital perfusion, and the extent of reconstitution was determined by flow cytometry. Reconstituted mice were infected with 10⁴ Tissue Culture Infectivity Dose 50 per mouse of HIV_BaL i.p.

At 2 wk postinfection, mice were anesthetized by isoflurane inhalation and bled retro-orbitally using EDTA-coated glass pipettes. Then 1 wk later, at 3 wk postinfection, mice were again anesthetized by isoflurane inhalation, and 300 μl of blood was removed from mice by cardiac puncture. The mice were then perfused with 30–50 ml ice-cold PBS for dissection of other organs. Blood was collected into microcentrifuge tubes with 15 μl EDTA. Blood was spun at 12,000 RPM for 15 min, and serum was collected. Viral copy number in the blood of HIV-infected animals was determined using TaqMan (Life Technologies, Carlsbad, CA) based real-time PCR. For RNA extraction from blood serum, 100 μl serum was processed in a QIAAmp viral RNA Mini Kit (QIAGEN, Valencia, CA) according to the manufacturer’s instructions. The sample was then reverse transcribed using qScript RT Master Mix (Quanta Biosciences, Gaithersburg, MD) and used to amplify viral RNA via real-time PCR. To generate a standard curve, a 500-bp gBlock gene fragment encompassing the GAG region and representing unspliced HIV RNA was synthesized by IDTDNA (Integrated DNA Technologies, Coralville, IA). The fragment was suspended in sterile TE buffer (Life Technologies) at dilutions ranging from 50 to 1 × 10⁶ copies per microliter. Copy number was calculated using the DNA copy number Web tool at Thermo Scientific (Thermo Fisher Scientific, Waltham, MA) and converted to RNA copy number by multiplying by 2. The dilutions were aliquoted and stored at −80°C. The primers and probe were designed using the PrimerQuest tool at IDTDNA (Integrated DNA Technologies). The sequence of the primers was as follows: F, 5′-CCCAGAAGTGATACCCATGTT-3′, and R, 5′-GCTTCCTCATTGATGGTCTCT-3′; and the probe was 5′/56-FAM/ATTTGCATG/ZEN/GCTGCTTGATGTCCC/3IABkFQ/-3′. Real-time PCR reactions were performed in a 20-μl solution containing 10 μl TaqMan Gene Expression Master Mix (Life Technologies), 900-nm primers and 500-nm probe, and 7.5 μl cDNA sample or standard. Cycling proceeded in an Applied Biosystems 7900HT sequence detection system (Thermo Fisher Scientific) using SDS2.3 software, under the following parameters: 50°C for 2 min, 95°C for 10 min, followed by 55 cycles of 95°C for 15 s and 60°C for 1 min. Samples were run in duplicates, and standards were included with each run. The lower limit of signal detection for standard was 40 copies per reaction and for sample was 2000 copies per milliliter.

Mice were anesthetized by inhalation of isoflurane and then perfused with 30–50 ml ice-cold PBS through the left ventricle. Brains were collected and minced with scissors before treatment with Liberase TL (Roche, Indianapolis, IN) and DNase I (Invitrogen) at 37°C on a tube rotator (Miltenyi Biotec) for 30 min. Samples were pushed through 100-μM Nytex Cell Strainers (BD Biosciences) to yield a single-cell suspension. Mononuclear cells were isolated using a 30%/70% Percoll gradient (Amersham, Piscataway, NJ) centrifuged at 2500 × g for 20 min at room temperature without break. Cells were counted using trypan blue exclusion (Life Technologies Invitrogen, Carlsbad, CA). Spleens were collected from mice at the time of perfusion. Spleens were mashed and RBCs were lysed using NH₄Cl to obtain a single-cell suspension of spleen-infiltrating lymphocytes.

NSG mice were reconstituted as described above with 2 × 10⁷ PBMCs from a single donor. At 2 wk later, PBMCs were isolated from the same donor and T cells were depleted by magnetic separation using CD3 microbeads according to the manufacturer’s instructions (Miltenyi Biotec). T cell–depleted PBMCs were then cultured overnight at 37°C without peptide or with 40 μg/ml CMVpp65 (HLA-A2: NLVPMVATV) or 2 μg/peptide/ml HIV pooled peptides for Gag, Nef, Pol, and Env (AIDS Regent Program) in complete DMEM with 8 μg/ml β₂-microglobulin (Sigma-Aldrich) to load target cells. Peptide loading on target cells was confirmed by upregulation of HLA-ABC, as determined by flow ctyometry. The following day, total lymphocytes isolated from HIV-infected mouse brain were cocultured with target cells without peptide or loaded with CMVpp65 irrelevant peptide or HIV pooled peptides in the presence of Abs for CD107a-FITC and CD107b-FITC (BD Biosciences). At 1 h of coculture, BD GolgiStop and BD GolgiPlug (BD Biosciences) were added to eliminate internalization/degradation of CD107ab. The culture was incubated for an additional 5 h at 37°C. Cells were then were stained with LIVE/DEAD Fixable Violet Dead Cell Stain (Life Technologies Invitrogen) according to the manufacturer’s instructions. After washing, the cells were stained for HLA-ABC-PE-Cy7, CD3 APC-H7, CD4-PE, and CD8-PERCP-Cy5.5 (BD Biosciences). CD3 was used to distinguish target cells from brain-isolated T cells. Data were acquired on an LSR II flow cytometer (BD Biosciences) with BD FACSDiva software (BD Biosciences) and analyzed using FlowJo Software (TreeStar, Ashland, OR).

PBMCs were isolated from the venous blood of healthy donors, as described above. CD8⁺ T cells were isolated from PBMCs prior to cell culture by an untouched CD8⁺ T cell Isolation Kit II according to the manufacturer’s instructions (Miltenyi Biotec). CD8⁺ T cells were then cultured overnight in cRPMI. Cells were next stained with BD Horizon Violet Proliferation Dye (VPD), CD8-Percp-Cy5.5, and CD4-PE (BD Biosciences) according to the manufacturer’s instructions. Cells were run on an LSR II flow cytometer to establish baseline proliferation stain, purity of CD8⁺ T cells in CD8 isolated cultures, and expression of CD4 on CD8⁺ T cells. Cells were resuspended to 3 × 10⁶ cells/10 μl in sterile PBS. The 10-μl Hamilton syringes with 32-gauge needles (Hamilton, Reno, NV) were autoclaved and cleaned with 70% ethanol and sterile PBS prior to surgery. Mice were anesthetized with 100 mg/kg ketamine and 20 mg/kg xylazine and placed in a stereotaxic frame. A scalp midline incision was made and a burr hole was drilled through the skull of the left hemisphere at 0.22 mm posterior to the bregma, 0.85 mm lateral to midline, and 2.35 mm ventral to the skull. With a motorized injection pump (Stoelting, Wood Dale, IL) CD8⁺ T cells or PBS was injected directly into the ventricle at a rate of 1 μl/min. The needle was left in place for 2 min after all volume was pushed into the brain before removal from the brain. At 6 d later, brains were processed as described above to isolate injected cells for flow cytometry.

Single-cell suspensions of cells isolated from mouse brain, spleen, or PBMCs from venous blood were stained with LIVE/DEAD Fixable Violet Dead Cell Stain (Life Technologies Invitrogen) according to the manufacturer’s instructions. Cells were washed with PBS and stained with CD3-APC-H7, CD4-PE or CD4-APC, and CD8-PERCP-Cy5.5, CCR7-PE, CD27-FITC, CD28-Alex Fluor 700, CD57-PE-Cy7, and/or CD45RA-PE-Cy7 (BD Biosciences) for extracellular stains. For intracellular detection of HIV-1 core Ag p24 cells were fixed and permeabilized using BD Cytofix/Cytoperm Solution Kit (BD Biosciences) according to the manufacturer’s instructions before staining with 1 μl per sample of HIV-1 core Ag–RD-1 (Beckman Coulter, Pasadena, CA). When cell number was limiting, a minimum of 5 × 10³ live gated events were collected in each sample and analyzed to maintain cell number between samples for comparison of different T cell populations. Data were collected on an LSR II flow cytometer with BD FACSDiva software (BD Biosciences) and analyzed using FlowJo Software (TreeStar).

Human progenitor-derived astrocytes (PDAs), provided by Dr. Eugene O. Major (National Institute of Neurological Disorders and Stroke, Bethesda, MD) were generated from neural progenitor cells, as previously described (31). Briefly, progenitors were seeded on poly-d-lysine–coated T 75 culture flasks at 2 × 10⁶ cells per flask and maintained in progenitor medium consisting of neurobasal media (Life Technologies Invitrogen), 25 ng/ml fibroblast growth factor, 20 ng/ml epidermal growth factor (R&D Systems, Minneapolis, MN), 50 μg/ml gentamicin (Lonza BioWhittaker), and 2 mM l-glutamine (Life Technologies Invitrogen). To induce differentiation, progenitor medium was replaced with PDA medium containing DMEM (Life Technologies Invitrogen) supplemented with 10% heat-inactivated FBS (Sigma-Aldrich), 2 mM l-glutamine, and 50 μg/ml gentamicin. Cultures were glial fibrillary acidic protein, glutamine synthetase, and EAAT2 (excitatory amino acid transporter-2) positive and nestin negative after 30 d of differentiation, making them comparable to commercially available primary fetal astrocytes (31). After differentiation, PDAs were maintained in PDA medium on poly-d-lysine–coated plates, as described above. Media were changed every 3 d, and cells were split when they reached 80–90% confluency. Astrocyte-conditioned media (ACM) were harvested from 3-d cultured PDAs. All ACM used in transfer experiments was taken from PDAs cultured in complete 10% heat-inactivated FBS containing DMEM.

Wnts were depleted from ACM, as described (32). Briefly, Pierce Protein A/G magnetic beads (Life Technologies Invitrogen) were washed two times for 1 h at 4°C with 1XGE Binding Washing Buffer (GE Life Sciences, Pittsburgh, PA). The beads were then coated with 4 μg anti-rabbit Wnt1, Wnt2b, Wnt3, Wnt5b, Wnt10b, or rabbit IgG1 isotype control overnight (Abcam, Cambridge, MA). Then 1 ml ACM was applied to the beads and incubated overnight at 4°C with continuous rotation at 30 rpm. The supernatant was next collected by separation of magnetic beads under a magnetic field, and depletion of Wnt(s) was confirmed by Western blot of supernatant, as previously shown (32).

Uninfected and HIV-infected NSG-huPBMC mice were anesthetized by isoflurane inhalation and perfused with 30–50 ml ice-cold PBS. Half of the brain was placed in TRIzol (Life Technologies Invitrogen) in Lysing Matrix D tubes (MP Biomedicals, Santa Ana, CA). Brain was homogenized using FASTPrep-24 tissue homogenizer (MP Biomedicals) for 5 min with 1 min rest at 2400 rpm, repeated three times. TRIzol/chloroform (Invitrogen Life Technologies) was then used to isolate mRNA from brain tissue. A_260/280 was used to measure RNA concentration, and 1 μg total RNA was used for downstream applications. DNA contamination was removed by DNase I (Sigma-Aldrich) treatment at room temperature for 15 min, followed by denaturation of DNase I at 70°C for 10 min. cDNA synthesis was performed using qScript cDNA supermix (QIAGEN) according to the manufacturer’s instructions. cDNA was diluted so that 1/20 of the original volume was used to perform real-time PCR using SsoFast EvaGreen Supermix with Low Rox kit (Bio-Rad, Hercules, CA) and ROX Passive Reference Dye (Bio-Rad) in an ABI 7500 Fast Real Time PCR System using 7500 software V2.01. The PCR conditions were as follows: stage 1: 95°C for 20 s; stage 2: 95°C for 5 s, 60°for 30 s; and one dissociation stage, which was at 95°for 15 s. This sequence was performed for 45 cycles. Melting curve analysis was conducted to ensure single and specific product amplification. The following primers were used: Gag: Forward 5′-AGAGAAGGCTTTCAGCCCAGAAGT-3′, Reverse 5′-TGCACTGGATGCACTCTATCCCAT-3′; GapDH: Forward 5′-TGACTTCAACAGCGACACCCACT-3′, Reverse 5′-ACCACCCTGTTGCTGTAGCCAAAT-3′. Fold change in mRNA expression was calculated using the comparative C_T method with GAPDH as endogenous control. A C_T off value of 35 was used to determine whether mRNA was expressed in tissue. For HIV fold change, all samples were first normalized to GAPDH and then further normalized to the sample with the lowest detectable HIV, thereby giving that sample an arbitrary value of 1, with the other higher 27 samples compared with that sample.

To measure β-catenin–dependent signaling activity, 5 × 10⁶ PBMCs cultured in cRPMI or cRPMI and ACM or cRPMI and Wnt-depleted ACM were transfected with 10 μg TOPflash reporter construct (Millipore, Billerica, MA) or FOPflash using the Amaxa nucleofection protocol (Amaxa, Gaithersburg, MD), as recommended by the manufacturer. At 3 d later, construct reporter activity was performed with a dual-luciferase reporter assay using 10–20 μl lysate (Promega, Madison, WI). The total protein concentration was measured using a Pierce BCA Protein Assay Kit (Thermo Scientific, Waltham, MA).

When the data were distributed normally, ANOVA and post hoc tests were used. When the data were not normally distributed, nonparametric analysis was performed. All tests assumed a two-sided significance level of 0.05 using GraphPad Instat 3 Software (San Diego, CA) for data analysis. Pearson correlation coefficients were determined using SAS analysis (SAS Institute, Cary, NC).

We evaluated the brains of uninfected and HIV-infected NSG-huPBMC mice for CD8 single positive (CD4⁻CD8^bright), CD4 single positive (CD4^brightCD8⁻), and CD4^dimCD8^bright T cells. NSG mice were reconstituted with human PBMCs, and at 2 wk post reconstitution, the mice were left uninfected or infected with HIV_BaL by i.p. injection. Blood viremia was measured preinfection (negative control) and at 2 and 3 wk postinfection (Fig. 1A). HIV-infected mice had 2063 HIV RNA copies per milliliter at 2 wk postinfection and 67,400 HIV RNA copies per milliliter at 3 wk postinfection. These data are consistent with previously published work (33, 34). At 2 and 3 wk post HIV infection, the brains were harvested and T cells were isolated by Percoll gradient centrifugation. The total number of lymphocytes isolated from HIV-infected brains was 3- and 2-fold higher in comparison with uninfected brains, at 2 and 3 wk postinfection, respectively (Fig. 1B). CD4 single positive, CD8 single positive, and CD4^dimCD8^bright T cells were detected in both uninfected and HIV-infected brains (Fig. 1C–F). Flow cytometry gating strategy is shown in Supplemental Fig. 1. HIV predominantly infects CD4⁺ T cells and leads to their rapid depletion. We demonstrate that this also occurs in HIV-infected CD4⁺ T cells that infiltrate the brain. At 2 wk postinfection, the percentages of CD4 single positive, CD8 single positive, and CD4^dimCD8^bright T cells in the brain were comparable between HIV uninfected and infected mice (Fig. 1C–E). Specifically, for CD4 single positive it was 25% versus 17%, for CD8 single positive it was 55% versus 40%, and for CD4^dimCD8^bright T cells it was 8% versus 12% of CD3⁺ T cells between uninfected and HIV-infected mice, respectively. At 3 wk postinfection, however, a dramatic depletion of CD4 single positive T cells (35% versus 3.3%) (Fig. 1C) and an increase in the frequency of CD8 single positive T cells (37% versus 56%) (Fig. 1D) in the brain of HIV-infected mice were observed. CD4 single positive T cells were depleted by 90% and CD8 single positive T cells were increased by 68% in brains of HIV-infected mice in comparison with uninfected mice (Fig. 2B and data not shown). These findings are consistent with clinical data and data from the SIV-macaque model demonstrating expansion and activation of CD8⁺ T cells in the brain during acute infection (2–5, 10, 12). The frequency of CD4^dimCD8^bright T cells was also altered at 3 wk postinfection. CD4^dimCD8^bright T cells decreased from ∼13% of CD3⁺ T cells to ∼2% of CD3⁺ T cells in HIV-infected mouse brains (Fig. 1E), a loss of ∼75% (Fig. 2B). These data indicate that T cells (CD4 single positive, CD8 single positive, and CD4^dimCD8^bright) infiltrate the brain and that the HIV-susceptible target cells (e.g., CD4 single positive and CD4^dimCD8^bright T cells) are depleted postinfection. The current paradigm indicates that nonproductively infected monocytes and perivascular macrophages are the “Trojan Horses” of HIV, allowing for its entry into the CNS. Our findings in this article indicate that CD4⁺ T cells can also infiltrate the brain and undergo depletion. CD8⁺ T cells are expanded in the CNS, which can contribute to heightened inflammatory responses in the CNS and neuronal injury.

Expression of CD4 on CD8⁺ T cells renders them permissive to HIV infection (16, 20). Within the CNS, the extent of HIV infection between CD4 single positive (6%) and CD4^dimCD8^bright T cells (3%) at 2 wk postinfection is comparable, as determined by measuring intracellular HIVp24 expression (Fig. 2A). Although statistically insignificant, at 3 wk postinfection, there was a trend toward a higher percentage of infected CD4 single positive cells in comparison with CD4^dimCD8^bright T cells (1% of CD4 single positive T cells versus 4% of CD4^dimCD8^bright T cells expressed HIVp24; p ≤ 0.07) (Fig. 2A), which may be due to differential loss of CD4 single positive in comparison with CD4^dimCD8^bright T cells. We compared the percentage of CD4 single positive and CD4^dimCD8^bright T cells in the brain of uninfected NSG-huPBMC mice and HIV-infected NSG-huPBMC mice. At 2 wk postinfection, ∼60% of CD4 single positive T cells were lost, compared with only 20% of CD4^dimCD8^bright T cells. Further, at 3 wk postinfection, >90% of CD4 single positive T cells were lost, compared with 75% of CD4^dimCD8^bright T cells (Fig. 2B). These data indicate that although both CD4 single positive and CD4^dimCD8^bright T cells found in the CNS are HIV infected, CD4 single positive T cells are lost at an earlier time point than are CD4^dimCD8^bright T cells. Reduction of T cell numbers in the brain may be due to trafficking of cells out of the brain, HIV-dependent death, HIV-independent death, or loss of CD4 expression as HIV infection results in downregulation of CD4 on the cell surface. Further, this finding suggests that CD4^dimCD8^bright T cells may constitute a novel reservoir for HIV in the CNS, independent of resident brain cells (microglia, perivascular macrophages, and astrocytes).

Peripheral CD4^dimCD8^bright T cells exhibit enhanced anti-HIV responses compared with CD8 single positive T cells. They constitute >60% of the CD8⁺ tetramer-positive response against HIV and are polyfunctional (15). To determine the association between CD4^dimCD8^bright T cells and HIV control in the CNS, we harvested brains from HIV-infected NSG-huPBMC mice at 3 wk postinfection. One hemisphere was used to measure HIV gag by quantitative RT-PCR and the other to isolate infiltrating lymphocytes. We show that although the percentage of CD8 single positive T cells was not associated with a change in HIV gag mRNA expression (R = −0.24; p < 0.27; Fig. 3A), the percentage of CD4^dimCD8^bright T cells was inversely correlated with a change in HIV gag mRNA expression (R = −0.62; p ≤ 0.001; Fig. 3B). Further, the frequencies of CD4^dimCD8^bright T cells and CD8 single positive T cells were positively associated (R = +0.99; p < 0.0001; Fig.3C). These data suggest that CD4^dimCD8^bright T cells play a significant role in anti-HIV responses in the brain. Moreover, CD4^dimCD8^bright T cells and CD8 single positive T cells may enter the brain at the same rate or, once in the brain, CD8 single positive T cells may give rise to the CD4^dimCD8^bright T cell phenotype.

Given the correlation between increased CD4^dimCD8^bright T cells and decreased HIV gag mRNA expression in the CNS, we assessed whether CD4^dimCD8^bright T cells exhibit anti-HIV–specific responses. T cells were isolated from the brain of HIV-infected NSG-huPBMC mice at 2 wk postinfection, cocultured with syngeneic CD3-depleted PBMCs preloaded with either an irrelevant peptide (CMVpp65) or HIV pooled peptides, or left unloaded. CD107ab expression was evaluated at 6 h post ex vivo culture. Surface expression of CD107a and b (also called lysosome-associated membrane protein) is a marker of cytotoxic potential, as CD107a and b translocate to the cell surface in response to activation-induced degranulation (35). Less than 2% of CD4^dimCD8^bright T cells cultured with targets without peptide or targets loaded with an irrelevant peptide expressed CD107ab, whereas 6–10% of CD4^dimCD8^bright T cells cultured with targets loaded with HIV pooled peptides expressed CD107ab (Fig. 4A and B). These data indicate that CD4^dimCD8^bright T cells in the brain of HIV-infected NSG mice exhibit cytotoxic anti-HIV–specific responses, which could contribute to the decreased level of HIV seen in the brain of animals with higher frequencies of CD4^dimCD8^bright T cells.

We next evaluated whether the brain microenvironment can induce CD8 single positive T cells to become CD4^dimCD8^bright T cells in vitro and in vivo. The brain, particularly astrocytes, is a rich source of Wnts (27, 32), which initiate a cascade of events to induce β-catenin expression. Given that CD4 expression on CD8⁺ T cells is mediated by Wnt/β-catenin signaling (18), we evaluated whether astrocytes mediate CD4^dimCD8^bright T cell expression through Wnts secretion. We cultured αCD3/αCD28-activated normal human PBMCs with or without ACM from primary human PDAs. ACM induced CD4^dimCD8^bright T cells by 2-fold in comparison with control (Fig. 5A). Transfecting PBMCs with a reporter for Wnt/β-catenin (TOPflash construct, which contains several T cell factor/lymphoid enhancer–binding factor binding sites linked to firefly luciferase) induced TOPflash activity in PBMCs cultured in ACM by ∼2.5-fold (Fig. 5B) compared with control, suggesting that Wnts in ACM may be driving CD4^dimCD8^bright T cell expression through increased β-catenin activity. To determine whether this induction is indeed mediated by Wnts in ACM, and because there are no neutralizing Abs for Wnts, we depleted ACM of Wnt 1, Wnt 2b, Wnt 3, Wnt 5b, or Wnt 10b by magnetic immunoprecipitation, as described (27). Depletion of IgG was used as a control throughout. Wnts depletion was confirmed by Western blot (data not shown) (27). Depletion of Wnts 1, 2b, 3, and 5b from ACM abrogated the ability of the ACM to induce the CD4^dimCD8^bright T cell phenotype (Fig. 5C). These data indicate that Wnts expressed by human astrocytes can skew CD8⁺ T cells toward the CD4^dimCD8^bright T cell phenotype.

We next determined whether human CD8⁺ T cells could become CD4^dimCD8^bright T cells in vivo. CD8 single positive T cells were sorted from healthy human PBMCs by magnetic separation and cultured overnight with IL-2. After 24 h in culture, the cells were labeled with VPD and then assessed by flow cytometry for the CD4^dimCD8^bright T cell population (Fig. 6A) and baseline VPD expression (Fig. 6D and E) directly before intracranial injection of 3 × 10⁶ cells in 10 μl of PBS in the lateral brain ventricle of unreconstituted NSG mice. Prior to injection, 0–2% of CD8⁺ T cells expressed CD4 (Fig. 6A and C). At 6 d later, cells were reisolated from the brain and analyzed by flow cytometry for percentage of CD4^dimCD8^bright T cells. We show that once injected into the brain, 13–20% of highly purified CD8 single positive cells become CD4^dimCD8^bright T cells. These CD4^dimCD8^bright T cells continue to proliferate in the brain at a rate comparable to that of CD8 single positive cells (Fig. 6D and E). These data indicate that the microenvironment of the brain is conducive to differentiating infiltrating CD8⁺ T cells to express CD4 on their surface, generating the CD4^dimCD8^bright T cell phenotype.

We evaluated the phenotype of brain CD4^dimCD8^bright T cells, particularly whether those cells are central memory, effector memory, or terminally differentiated effector memory (TEMRA) T cells, using conventional phenotypic markers to distinguish these subsets (36–39). We found that although most CD8 single positive T cells were effector memory cells (CCR7⁻CD27^loCD28^loCD45RA⁻) prior to injection (Fig. 7A), 6 d post injection both CD8 single positive (Fig. 7A) and CD4^dimCD8^bright (Fig. 7B) T cells largely re-expressed CD45RA, indicating that they represent terminal effector memory cells re-expressing CD45RA (TEMRA, CCR7⁻ CD27⁻CD28⁻CD45RA⁺). TEMRA cells are end-stage terminally differentiated cells with low proliferative capacity but high cytotoxicity owing to high expression of perforin and FasL and enhanced cytokine production. TEMRA cells have been shown to play a vital role in control of HIV infection in the blood (39). Accumulation of terminally differentiated CD8⁺ T cells with low expression of CD28 is often associated with an increase in senescent T cells (36). To assess whether CD4^dimCD8^bright T cells in the HIV-infected brain are senescent, we harvested lymphocytes from the brain of NSG-huPBMC mice that were uninfected or HIV infected and assessed expression of CD28^loCD57⁺, a common delineator of T cell senescence (37–41). At 3 wk postinfection, the percentage of CD8 single positive T cells that were CD28^loCD57⁺ was comparable in uninfected (47%) and HIV⁺ (40%) mice (Fig. 7C). This was also true in the CD4^dimCD8^bright subset, with ∼47% of brain-infiltrating CD4^dimCD8^bright cells in the uninfected mice and 51% in the HIV-infected mice (Fig. 7C). Collectively, these data indicate that the microenvironment of the brain can induce CD4 expression on CD8 T cells, which represent potent cytotoxic anti-HIV cells that are not senescent and are largely of the TEMRA phenotype in the brain.

CD8⁺ T cells play a critical role in antiviral responses, including in HIV infection (2–6, 10–12, 15). However, in chronic HIV infection, peripheral CD8⁺ T cells undergo senescence, and their lytic activity is diminished over time and does not normalize even after prolonged antiretroviral treatment (42). HIV is found in the brain of HIV-infected individuals within 7–12 d of acute infection (43). Immunologic determinants of HIV control in the brain are not clear. We used the NSG-huPBMC mouse model to assess the role of CD8⁺ T cell subsets in HIV control. We specifically focused on a unique subset of CD8⁺ T cells that coexpress CD4 on their surface (CD4^dimCD8^bright T cells) because peripheral CD4^dimCD8^bright T cells have potent anti-HIV responses (15). We show in this article that although both CD8 single positive and CD4^dimCD8^bright T cells are found in the brain of NSG-huPBMC HIV⁺ mice, it is the CD4^dimCD8^bright T cells, and not CD8 single positive T cells, that are associated with lower HIV content in the brain. Furthermore, brain CD4^dimCD8^bright T cells of HIV-infected mice exhibit anti-HIV specific responses ex vivo. These data indicate that in the CNS CD4^dimCD8^bright T cells control HIV infection. However, owing to expression of CD4 on their surface, CD4^dimCD8^bright T cells are susceptible to HIV infection. In fact, both CD4 single positive and CD4^dimCD8^bright T cells are susceptible to HIV infection and loss, albeit CD4 single positive cell loss occurs at an earlier time point than do CD4^dimCD8^bright T cells. Reduction of CD4^dimCD8^bright T cells in the brain may be independent of HIV infection, as p24-positive cells remain unchanged in this population. Although we did not directly assess whether CD4^dimCD8^bright T cells are more resistant to HIV infection, they have several properties that can confer resistance. CD4^dimCD8^bright T cells are enriched in β-catenin expression, which is a cotranscriptional factor linked to suppression of HIV in PBMCs (44). Further, CD4^dimCD8^bright T cells express higher levels of Bcl-X_L, which is an antiapoptotic factor under the control of β-catenin signaling that prevents activation-induced cell death (18, 45). As β-catenin represses HIV transcription (44), the high expression of β-catenin in CD4^dimCD8^bright T cells, coupled with the Wnt-rich microenvironment of the brain, may either reduce the population size of HIV-infected CD4^dimCD8^bright T cells or allow them to survive better postinfection. Moreover, other mechanisms contributing to reduction of brain CD4 single positive and CD4^dimCD8^bright T cells over time may be due to HIV-independent death (e.g., inflammation/activation), change in phenotypic markers (loss of CD4 on CD4^dimCD8^bright T cells due to HIV infection), and/or trafficking out of the brain.

The Wnt/β-catenin pathway is important in thymocyte survival as thymocytes pass from the double negative to double positive stage (18, 45). We have previously shown that this process still occurs in the periphery in CD8⁺ T cells to drive CD4 re-expression on mature CD8⁺ T cells (18). Another population of specialized CD8⁺ T cells with high β-catenin activity, and high potency and durability, is the T stem cell memory (CD8⁺ Tscm) population. Tscm make up ∼2–4% of all circulating T cells and can rapidly differentiate into more mature central memory, effector memory, and effector T cells while maintaining their own pool (46). Like our study, previous work has shown that the frequency of circulating CD8 Tscm in untreated HIV⁺ individuals is inversely correlated with viral load, suggesting a contribution in protecting patients from HIV disease progression (47). Although CD4^dimCD8^bright T cells and Tscm have numerous similarities, it is unknown if they are the same population. Tscm are routinely identified by a CD8⁺CD45RA⁺CCR7⁺CD27⁺CD95⁺ phenotype in the blood of HIV-infected patients (47). CD4^dimCD8^bright T cells in the brain of NSG mice are largely CCR7⁻ and CD27⁻ but do express CD45RA⁺, which would indicate that at least in the brain these are not phenotypically the same population of cells.

The expression of CD4^dimCD8^bright T cells is positively associated with CD8 single positive T cells in the brain. This finding suggests that both populations may enter the CNS together or that once in the brain, CD8 T cells give rise to CD4^dimCD8^bright T cells or that both events could occur simultaneously. This is especially likely because the CNS and particularly astrocytes are a rich source for Wnts, which drive the CD4^dimCD8^bright T cell phenotype. Indeed, Wnts from ACM induced the CD4^dimCD8^bright T cell phenotype in vitro. Injecting a highly purified CD8 single positive T cell population into the brain also gave rise to CD4^dimCD8^bright T cells. Those cells are largely terminal effector memory cells re-expressing CD45RA. Re-expression of CD45RA in effector memory cells is a marker of terminal differentiation whereby effector memory cells display enhanced lytic capabilities but reduced proliferative potential (39, 48–51). Terminal differentiation of total CD8⁺ T cells and HIV-specific CD8⁺ T cells has been associated with slower progression of HIV (52, 53). Accumulation of terminally differentiated CD8⁺ T cells has also been linked to senescence, which is characterized by CD8⁺ T cells with shortened telomeres, loss of the costimulatory molecule CD28, and increased expression of CD57 (42). CD57 is a marker of proliferative history and often indicates a cell that has poor proliferative capacity (42). In this article, we show that brain-infiltrating CD4^dimCD8^bright T cells are not more senescent in the brain of HIV-infected mice. Furthermore, HIV infection actually inhibits senescence in T cell populations in the periphery (39). Hence, the brain microenvironment induces CD4^dimCD8^bright T cells with a potent terminal effector phenotype and the capability of lysing HIV-infected cells in the brain. In support of this observation, it has been previously reported that CD8⁺ T cells in the cerebrospinal fluid of HIV-infected patients have higher expression of CD38, HLA-DR, CXCR3, and adhesion markers, indicating they are more highly activated than CD8⁺ T cells in the periphery (11, 38). Although increased activation and increased expression of adhesion markers is necessary for activated CD8⁺ T cells to cross the blood–brain barrier into the HIV-infected brain, it is possible that the microenvironment of the brain, likely through Wnts, further skews the differentiation of CD8⁺ T cells to enhance their antiviral capacity. The consequence of heightened inflammatory CD8⁺ T cells in the brain is not clear. On the one hand, they can control HIV spread in the CNS, but on the other, secretion of inflammatory cytokines and lytic molecules may lead to neuronal injury. Recently, Schrier et al. (54) showed that CD8⁺IFN-γ⁺ T cells in the cerebrospinal fluid of HIV-infected patients positively correlated with neurocognitive impairment, whereas CD8⁺CD107ab⁺ T cells negatively correlated with neurocognitive impairment. Our data show that ∼10% of brain CD4^dimCD8^bright T cells ex vivo express CD107ab when cocultured with HIV pooled peptide-loaded targets. Further studies are warranted to determine whether the frequency of CD4^dimCD8^bright T cells is associated with a better neurocognitive performance score among HIV-infected individuals. Anti-HIV responses of CD4^dimCD8^bright T cells may be linked to lower brain viral burden. However, higher expression of inflammatory cytokines and cytolytic molecules such as CD107ab from CD4^dimCD8^bright T cells may be detrimental to CNS homeostasis and function (15, 54). It still remains to be determined whether CD4^dimCD8^bright T cells in cerebrospinal fluid will also be associated with better cerebrospinal fluid viral control and better neurocognitive outcome. Understanding the contribution of CD4^dimCD8^bright T cells to anti-HIV immunity and the biology of CD8⁺ T cells is critical to devising therapeutic strategies to heighten and/or preserve antiviral responses to a wide variety of viral infections.

The authors have no financial conflicts of interest.