Tim-3 Is Required for Regulatory T Cell–Mediated Promotion of T Cell Exhaustion and Viral Persistence during Chronic Lymphocytic Choriomeningitis Virus Infection

Regulatory T cells (Tregs), a subset of CD4⁺ T cells, are critical mediators of immune tolerance under homeostatic conditions. Tregs can also modulate the balance between immunity and tolerance during infection. The importance of Tregs in T cell–mediated viral clearance is highlighted by a study showing that transient in vivo ablation of Tregs during chronic LCMV infection enhances virus-specific T cell responses (1). This study served as a proof of concept that the suppressive activity of Tregs might contribute to chronic viral infection. Consistent with such a model, in people with untreated chronic hepatitis B virus, hepatitis C virus, or HIV there is an increase in the peripheral Treg frequency that also correlates with higher viral load, as well as a decrease in virus-specific T cell responses (2–6). Ex vivo depletion of human Tregs improves viral-specific T cell responses to hepatitis C virus (2, 4), hepatitis B virus (3), and HIV (5, 7, 8). Altogether, these studies suggest that the presence of Tregs limits effective antiviral immunity in certain chronic viral infections.

T cell Ig and mucin domain-containing protein 3 (Tim-3) is a transmembrane protein that can be either constitutively or inducibly expressed on multiple different immune cell types, including Tregs. In murine and human chronic infections, the expression of Tim-3 by CD8⁺ effector T cells is associated with poor control of viral burden and impaired antiviral immune responses (9–11). Nevertheless, the biological and clinical relevance of Tim-3 on Tregs during chronic infection remains unknown. Data from our group and others show that Tim-3⁺ Tregs have increased expression of CD39, PD-1, CTLA-4, ICOS, Ki67, and CD44 in comparison with Tim-3⁻ Tregs (12–17). More recently, our group and others showed that in people with HIV on antiretroviral therapy (PWH-ART) there is an increase in the frequency of peripheral Tim-3⁺ Tregs that is not normalized by antiretroviral therapy (ART) (18, 19). In contrast, a small subset of people with HIV known as long-term nonprogressors, who have low viral load, high CD4⁺ T cell counts, and are ART-naive, do not exhibit increased peripheral Tim-3⁺ Treg frequency (19). We recently reported that Tim-3⁺ Tregs have a robust suppressive phenotype compared with Tim-3⁻ Tregs from the same individual (18). In addition, we showed that anti–Tim-3 mAb reversed the in vitro suppression capacity of Tregs in PWH-ART. Thus, Tim-3⁺ Tregs may represent a target for the treatment of chronic inflammation and viral persistence.

Based on these findings, we hypothesized that Tim-3 expression on Tregs promotes viral persistence and limits antiviral T cell responses during chronic infection. Employing conditional Treg-specific Tim-3 knockout (KO) mice, we demonstrate, to our knowledge, for the first time, that deletion of a single surface protein in Tregs is sufficient to limit lymphocytic choriomeningitis virus (LCMV) viral persistence. Deficiency of Tim-3 in Tregs also led to an increase in the activation phenotype and function of virus-specific T cells during chronic LCMV infection. Lack of Tim-3⁺ Tregs reduced morbidity, boosted T cell function, and limited the development of T cell exhaustion. Therefore, our data revealed that Tim-3 expression in Tregs contributes to LCMV viral persistence, the development of T cell exhaustion, and poor virus-specific T cell function.

To evaluate the specific contribution of Tim-3 deletion in Tregs during chronic LCMV infection we employed a conditional Treg-specific Tim-3 loss-of-function mouse model, as previously reported (12). Havcr2^flox mice were generated by the University of Pittsburgh Department of Immunology Transgenic and Gene Targeting core. Briefly, CRISPR/Cas9 technology was applied for inserting LoxP sites on either side of the exon 4 of the Havcr2 gene in C57BL/6J zygotes. Deletion of exon 4 by Cre-mediated recombination results in splicing of exon 3 into exon 5, which results in a premature stop codon 20 bp into exon 5, before the transmembrane domain. A premature stop codon before the last exon leads to nonsense-mediated mRNA decay, resulting in the absence of transcript and lack of expression of a potential soluble truncated protein. Finally, incorporation of the LoxP sites was confirmed by PCR followed by sequencing of the targeted region. Foxp3^{eGFP-Cre-ERT2} mice were then crossed with Havcr2^flox/flox mice to generate mice homozygous for both genetic modifications. Foxp3^{eGFP-Cre-ERT2} mice had been previously backcrossed to C57BL/6 mice for at least 10 generations.

Mice were bred in-house under specific pathogen-free conditions and used at 6–8 wk. All experiments were age and sex matched. All animal procedures were conducted in accordance with National Institutes of Health and University of Pittsburgh Institutional Animal Care and Use Committee guidelines. For inducing Cre-mediated Tim-3 deletion, tamoxifen was resuspended in sunflower oil and administered daily for 5 d i.p. (1 mg/d). After the last dose, mice were left to rest for 2 d, and were used for experiments on day 7 after tamoxifen was initiated. Validation of Tim-3 deletion is shown in Supplemental Fig. 2A and 2B.

LCMV Cl-13 was obtained from Rafi Ahmed (Emory University) and propagated as described previously (20). For chronic infections, mice were infected with 2 × 10⁶ PFU LCMV Cl-13 i.v. at day 0 and analyzed at the indicated time points postinfection. For acute infections, mice were infected with 2 × 10⁵ PFU i.p.

Spleen single-cell suspensions were stained with Ghost Dye Violet 510 (catalog no. 13-0870-T500; Tonbo Biosciences) and Fc block for 20 min. Samples were washed and stained for several surface markers using the following mAbs/clones: Tim-3 (RMT3-23, BioLegend), CD4 (GK1.5, BD Biosciences), CD8 (53-6.7, BD Biosciences), CD44 (IM7, BioLegend), ICOS (7E.17G9, Thermo Fisher Scientific), PD-1 (RMP1-14, Thermo Fisher Scientific), CD39 (Duha59, BioLegend), CD25 (PC61, BioLegend), KLRG-1 (2F1, Thermo Fisher Scientific), PD-L1 (10F.9G2, Cytek Biosciences), CD45.1 (A20, BD Biosciences), CD45.2 (104, BD Biosciences), CD11b (M1/70, BD Biosciences), Ly6G (1A8-Ly6g, Thermo Fisher Scientific), MHC class II (M5/114.15.2, BioLegend), CD24 (M1/69, Tonbo Biosciences), and F4/80 (BM8, Thermo Fisher Scientific). After performing extracellular staining, samples were fixed and permeabilized using a Foxp3 staining kit (eBioscience), and stained for intracellular transcription factors, cytokines, and phosphorylated proteins, including Foxp3 (FJK-16 s, Thermo Fisher Scientific), CTLA-4 (UC10-4B9, BioLegend), IL-10 (JES5-16E3, BioLegend), Bcl-2 (BCL/10C4, Thermo Fisher Scientific), Ki67 (B56, BD Biosciences), p-STAT3 (Y705) (4/P-STAT3, BD Biosciences), p-STAT5 (Y694) (47/Stat5 [p-Y694], BD Biosciences), TNF (MP6-XT22, BioLegend), IFN-γ (XMG1.2, Thermo Fisher Scientific), and granzyme B (GzmB; NGZB, Thermo Fisher Scientific). H-2D^b-gp33 tetramers were obtained from the NIH Tetramer Core Facility. Flow cytometry gating strategy was used for analyzing Tregs (Supplemental Fig. 4A), gp33-specific T cells (Supplemental Fig. 4B), P14 T cells (Supplemental Fig. 4C), and innate immune cells (Supplemental Fig. 4D). Flow cytometry data were obtained using a Cytek Aurora spectral flow cytometer, and results were analyzed using FlowJo (version 10.8.1).

Viral titers were measured from infected tissue after RNA extraction using TRIzol LS reagent, followed by RNA-to-cDNA conversion, and quantitative PCR (qPCR) for the GP protein. Viral copy number was calculated by comparing Ct values with a plasmid standard curve, as previously described (21). The VERO cell plaque assay was used to confirm the results obtained using the qPCR method in a selected group of samples, as previously reported (22).

For assessing T cell functional capacity, 5 × 10⁶ splenocytes from infected mice were stimulated with 100 ng/ml LCMV gp33 peptide (AnaSpec) for 5 h at 37°C in complete RPMI 1640 with bovine growth serum (HyClone). Cells were simultaneously treated with a protein transport inhibitor containing monensin (BD GolgiPlug, catalog no. 554724), following the manufacturer’s instructions, to improve cytokine detection by flow cytometry. Surface and intracellular staining was performed as described above.

Splenic live CD45.1 congenically marked P14 cells were stained, as described above, and FACS sorted from a female CD45.1/CD45.2 P14 mouse. Sorted P14 cells were then transferred into tamoxifen-pretreated congenically distinct wild-type (WT) and Tim-3 Treg KO mice (1500 P14/mouse, i.v.). One day after transfer, mice were infected with LCMV Cl-13 and analyzed at indicated timepoints.

To measure the functional activity of gp33-specific CD8⁺ T cells, we employed the xCELLigence system, which allows for real-time and label-free monitoring of cell viability. Briefly, this technology uses microtiter plates embedded with gold microelectrodes to noninvasively measure the viability of target cells attached to the plate by using electrical impedance as the readout (23). To test T cell functional capacity, we plated 20,000 B16-gp33 melanoma cells (target cells) per well on an E-Plate VIEW 96 (Agilent, catalog no. 3000-601-020) and added RPMI 1640 supplemented with 10% FBS. Cell cultures were maintained at 37°C and 5% CO₂. After 8 h, target cells were 100% confluent, at which time effector T cells were added to the culture. For effector T cells, we sorted live CD44^hiCD62L^lo CD8⁺ T cells from WT and Tim-3 Treg KO mice at 8 d postinfection. As a negative control, we sorted CD44^hiCD62L^lo CD8⁺ T cells from an uninfected WT mouse. A total of 60,000 effector T cells were added into each well to achieve a 3:1 E:T ratio, and the assay ran uninterrupted for 25 h, after which T cells were added. Light field microscopy images at a ×10 original magnification were taken with each electrical impedance readout every 15 min. Finally, the percent cytolysis was calculated by the xCELLigence system using the following formula: percentage of cytolysis = [(cell index, no effectors − cell index, plus effectors)/cell index, no effectors] × 100.

After initial quality control and adapter trimming, sequence data were quantified using Kallisto (24) to obtain transcript level abundances, using mm10 (UCSC) as the reference genome. After quantification, differentially expressed genes and transcripts between the WT and KO mice were identified using Sleuth (25). Significant differentially expressed transcripts were defined using a q value (Benjamini–Hochberg adjusted p value) threshold of <0.2. Overrepresentation analysis was performed using the clusterProfiler (26, 27) tool to identify enrichment of the genes in specific pathways described in the Kyoto Encyclopedia of Genes and Genomes (28) and Reactome (29) databases. RNA sequencing (RNA-seq) data are available under GEO accession number GSE254211.

We based the sample size on data obtained from previous work with the LCMV model (21, 30) and performed a power calculation. Based on that, we determined that a sample size of eight mice each (for a two-group experiment) is sufficient to detect an effect size of 2.8 (with 10% SDs), with an α value of 0.05 and power of 0.95. Each such in vivo experiment was performed at least twice. Two-way comparisons were analyzed using a Student t test. A p value <0.05 was considered significant for all tests. All statistical analyses were performed using GraphPad Prism (version 9.4.1).

To examine the role of Tim-3 in Tregs we first evaluated the changes in the frequency of Tregs and Tim-3⁺ Tregs during acute (Armstrong) versus chronic (clone 13) LCMV infection. As expected, Treg frequency decreased at the early stage of chronic infection, and then became persistently elevated (Supplemental Fig. 1A). In contrast, during acute infection the frequency of total Tregs remained unaltered at day 5 and decreased at later time points, compared with uninfected mice (Supplemental Fig. 1A). We next assessed the phenotype of these Tregs. Our data revealed a significant increase in the frequency of Tim-3⁺ Tregs at 5, 8, and 15 d postinfection with LCMV Cl-13 (Fig. 1A). We also observed an increase in Tim-3⁺ Treg frequency on day 5 postinfection with LCMV-Armstrong, relative to uninfected mice, which normalized to 2.5% by day 8 (Fig. 1A). Despite the changes in the frequency of Tregs and Tim-3⁺ Tregs during acute and chronic infection, we did not detect any significant differences in the proportion of CD4⁺Foxp3⁻ conventional T (Tconv) cells (Supplemental Fig. 1B) nor in Tim-3⁺ Tconv cells (Supplemental Fig. 1C). Therefore, these data indicate that the percentage of Tim-3⁺ Tregs increases in conditions of high viral load and that these cells persist during chronic infection.

Next, we characterized the phenotype of Tim-3⁺ Tregs at 15 d postinfection, the peak of Tim-3⁺ Treg frequency during LCMV Cl-13 infection. Strong Treg immunosuppression has been associated with an increase in Treg proliferation and survival. We found that the frequency of Ki67⁺ cells (indicative of active proliferation) was elevated in Tim-3⁺ Tregs compared with Tim-3⁻ Tregs (Fig. 1B). These cells also showed a higher frequency of KLRG-1⁺ Tregs (Supplemental Fig. 1D), a marker of short-lived effector T cells, and downregulation in the expression of Bcl-2 (Supplemental Fig. 1E), an antiapoptotic molecule. Thus, the phenotype of Tim-3⁺ Tregs is consistent with these cells having an enhanced immunosuppressive capacity and perhaps an accelerated turnover.

Flow cytometry analysis also revealed upregulation in the expression of several activation and effector markers such as CD44, CTLA-4, ICOS, PD-1, and CD39 on Tim-3⁺ Tregs (Fig. 1C). In addition, we found elevated effector Treg signaling in Tim-3⁺ Tregs, as indicated by increases in p-STAT3 (suggesting IL-10 signaling) and p-STAT5 (suggesting IL-2 signaling) (Fig. 1D, 1E). Altogether, these findings demonstrate that Tim-3 expression correlates with a robust immunosuppressive phenotype in Tregs during chronic viral infection.

To evaluate whether Tim-3⁺ Tregs are derived from the thymus or induced in the periphery upon viral infection and persistence, we assessed the expression of Helios in Tregs during LCMV Cl-13 infection. Helios, a member of the Ikaros family, is expressed early in the development of all thymic-derived Tregs but is absent in peripherally induced Tregs (31). Consistent with Foxp3⁺ Tregs being thymically derived, we found that Tim-3⁺ Tregs have higher expression of Helios (Supplemental Fig. 1F) than Tim-3⁻ Tregs. We also noted a significant increase in the expression of Foxp3 in Tim-3⁺ Tregs compared with Tim-3⁻ Tregs, suggesting that expression of Tim-3 enhances, or is at least associated with, Treg stability (Supplemental Fig. 1G).

We next assessed the kinetics of the phenotypic changes in Tregs during chronic LCMV infection. Thus, there was a significant increase in the proportion of Tim-3⁺ Tregs with high levels of the proliferation marker Ki67⁺, whereas this metric remained unchanged in Tim-3⁻ Tregs (Supplemental Fig. 1H). Expression of Tim-3 also marked Tregs with a more profound change in the upregulation of CD44 and loss of CD62L, that is, activated/effector Tregs (Supplemental Fig. 1I); these Tregs also had the highest proportion of the effector/activation Treg markers ICOS (Supplemental Fig. 1J) and PD-1 (Supplemental Fig. 1K). Taken together, these data suggest that Tim-3⁺ Tregs are poised for a stronger response than Tim-3⁻ Tregs during chronic infection, at least based on the expression of commonly used activation markers.

To better understand the function of Tim-3⁺ Tregs during chronic LCMV infection, we used conditional Treg-specific Tim-3 KO mice previously described by our group, breeding mice carrying a floxed exon 4 of Havcr2 (encoding Tim-3) with mice expressing a Foxp3-driven, tamoxifen-inducible, Cre (Foxp3^{eGFP-Cre-ERT2}). These mice are referred to as Tim-3^Treg KO; mice expressing the Cre alone will are to as Cre-only or WT (12). These mice were treated with tamoxifen and infected with LCMV Cl-13 (experimental outline in Fig. 2A). Validation of the efficiency of Tim-3 KO after tamoxifen administration, with or without infection, is shown in Supplemental Fig. 2A and 2B. Thus, we observed that Tim-3^Treg KO mice lost less weight and had a significantly faster recovery in weight than did WT mice throughout the course of infection (Fig. 2B). Tim-3^Treg KO mice also recovered fully to their starting weight.

Next, we assessed (by qPCR) the viral burden in the liver and kidney on day 30 postinfection. Dramatically, we found that Tim-3^Treg KO mice had a significant reduction of ∼500-fold in viral copies in the liver and 100-fold in the kidney, compared with WT mice (Fig. 2C). Kidney viral titers at day 30 were validated by plaque assay, for which six of seven of the Tim-3 Treg KO kidney samples were below the limit of detection (Supplemental Fig. 2C). In contrast, these mice did not show a difference in viral burden at day 15 postinfection (Supplemental Fig. 2D). Tim-3^Treg KO mice were also mostly protected from severe lymphopenia (Fig. 2D) and had an increase in the CD8/Treg ratio (Fig. 2E). Treg-specific Tim-3 deletion also led to a decrease in the frequency of total Tregs (Fig. 2F) and an elevated frequency of effector T cells in the spleen (Fig. 2G). Based on tetramer staining, Tim-3 Treg KO mice had an increase in the total number of gp33-specific T cells (Fig. 2H) compared with WT mice, although there was no change in the frequency of gp33-specific CD8⁺ T cells (Fig. 2I).

The development of exhausted T cells (Tex) among the pool of virus-specific T cells impairs the clearance of certain chronic viral infections, including LCMV (32, 33). To test the changes in Tex in Tim-3^Treg KO mice during chronic LCMV infection, we focused on the expression of PD-1 and Tim-3, well-described markers of T cell dysfunction during persistent viral infection (11, 34–36). We found that late during chronic infection (day 30), Ag-experienced T cells (CD44^hiCD8⁺) in Tim-3⁺ Treg KO mice exhibited a significant decrease in the frequency of terminally exhausted (PD-1⁺Tim-3⁺) T cells (Fig. 3A). We also measured the expression of Tox, a transcription factor required for the development of dysfunctional T cells (37). Thus, Ag-experienced CD8⁺ T cells from Tim-3^Treg KO mice also had a lower proportion of PD-1⁺Tox⁺ T cells (Fig. 3B). These findings were recapitulated by Tex analysis of gp33-specific T cells. Thus, gp33-specific T cells also showed a decrease in the frequency of terminally exhausted PD-1⁺Tim-3⁺ (Fig. 3C) and PD-1⁺Tox⁺ T cells (Fig. 3D). Taken together, these data suggest that the expression of Tim-3 by Tregs contributes to the development of gp33-specific Texs during the later stages of chronic infection, perhaps due to changes in the levels of virus.

The significant decrease in viral burden, along with the enhanced gp33-specific response and decrease in Tex cells late during chronic infection, motivated us to evaluate whether these changes are established in the early phase of chronic infection. We focused our analysis on day 8 postinfection, as this is a timepoint at which there is partial T cell exhaustion (Fig. 4A). Our data revealed that at this early stage of infection, Tim-3^Treg KO mice were protected from severe lymphopenia (Fig. 4B). However, Tim-3^Treg KO mice did not have a decrease in viral burden by qPCR at this time point (Fig. 4C). Tim-3^Treg KO mice did have a selective increase in total CD8⁺ T cell count, with no change in the number of CD4⁺ T cells (Fig. 4D). Deletion of Tim-3 on Tregs also did not change the number of total Tregs at this early timepoint (Fig. 4E), rather the increase in the CD8/Treg ratio was driven mainly by the expansion of CD8⁺ T cells (Fig. 4F). Tim-3^Treg KO mice did show a reduction in Treg frequency (Fig. 4G), with a corresponding elevated frequency of effector T cells (Fig. 4H). Furthermore, Tim-3^Treg KO mice had increases in both total cell count and the frequency of gp33-specific CD8⁺ T cells (Fig. 4I, 4J). Finally, although we do not currently have the means to track Tim-3 “wannabe” Tregs, we assessed the total Treg pool to determine whether selective loss of Tim-3 from these cells led to a corresponding loss of effector Tregs. Indeed, total Tregs in Tim-3^Treg KO mice contained a smaller population of CD39⁺ Tregs (Fig. 4K) but maintained normal levels of CD25 and PD-1 (Supplemental Fig. 3A), compared with WT Tregs.

One of the consequences of wholesale Treg depletion is upregulation of the inhibitory ligand PD-L1 on innate immune cells and other infected cells (1). Therefore, we carried out a broad characterization of the innate immune compartment early during chronic infection of WT versus Tim-3^Treg KO mice. Interestingly, Tim-3^Treg KO mice had a modest but significant reduction in the frequency of dendritic cells (DCs), and these cells also expressed lower levels of PD-L1 (Supplemental Fig. 3B, 3C). Additional innate populations examined, including monocytes, neutrophils, and macrophages, exhibited no difference in their proportions or in PD-L1 expression in Tim-3^Treg KO versus WT mice (Supplemental Fig. 3B, 3C). None of the innate cells that we examined exhibited a change in the expression of the costimulatory molecule CD80 (Supplemental Fig. 3D), but there was an increase in CD86 expression by DCs and monocytes (Supplemental Fig. 3E). Altogether, these findings suggest that Tim-3⁺ Tregs limit the functional capacity of gp33-specific T cells in part by promoting the upregulation of PD-L1 and limiting the costimulation potential of certain innate cells early during chronic infection.

We next aimed to determine the changes in Tex cells early during chronic infection, at day 8 postinfection. Thus, Ag-experienced T cells did not show any difference in PD-1 and Tim-3 frequencies at day 8 postinfection (Fig. 5A). Similarly, there was no detectable change in the proportion of Tox-expressing cells in this population (Fig. 5B). In contrast, Tim-3^Treg KO mice had significant reductions in LCMV-specific PD-1⁺Tim-3⁺ (Fig. 5C) and PD-1⁺Tox⁺ gp33⁺ T cells (Fig. 5D), as well as an increase in the stem-like PD-1⁺Tim-3⁻ gp33⁺ T cells (Fig. 5C). These findings suggest that Tim-3 expression by Tregs promotes the early development of T cell exhaustion and limits the generation of the stem-like PD-1^int population.

Next, we sought to extend our results by evaluating the effects of Tim-3⁺ Tregs on gp33-specific T cells with a fixed TCR, using adoptive transfer of WT P14 T cells into WT and Tim-3^Treg KO mice. Reminiscent of our observations with polyclonal LCMV-specific T cells, there was a significant reduction in the frequencies of terminally exhausted PD-1⁺Tim-3⁺ (Fig. 5E) and PD-1⁺Tox⁺ P14 T cells (Fig. 5F) in Tim-3^Treg KO mice early during chronic infection. However, in contrast to polyclonal gp33-specific CD8⁺ T cells, there was no difference in the proportion of transferred P14 T cells with a PD-1⁺Tim-3⁻ phenotype (Fig. 5E). Nonetheless, overall, these data suggest that selective deletion of Tim-3 from Tregs is sufficient to limit the development of terminally exhausted WT P14 T cells.

The improved gp33-specific T cell response observed at 8 d postinfection in Tim-3^Treg KO mice led us to further assess whether these cells were more functionally capable of killing gp33⁺ target cells and producing inflammatory cytokines. To test this, we cultured flow-sorted CD44^hiCD8⁺ Ag-experienced polyclonal T cells with B16.gp33 target cells, a murine melanoma cell line that expresses gp33. Representative images showed a notable decrease in the confluency of B16.gp33 target cells adhered to the cell culture plate 25 h after adding Ag-experienced T cells (Fig. 6A). Quantification revealed an increase in the cytotoxicity of Ag-experienced T cells from Tim-3^Treg KO mice compared with WT mice (Fig. 6B). These data suggest that Treg-specific Tim-3 deletion enhances the functional capacity of gp33-specific T cells early during chronic infection.

Additionally, in vitro restimulation of CD8⁺ cells with gp33 peptide revealed that Ag-experienced T cells from Tim-3^Treg KO mice had an elevated frequency of TNF⁻IFN-γ⁺ and TNF⁺IFN-γ⁺ T cells at 8 d postinfection (Fig. 6C) and a small increase in Gzmb expression (Fig. 6D). To confirm these results, we adoptively transferred congenically marked WT P14 cells into either WT or Tim-3^Treg KO mice, followed by LCMV Cl-13 infection, and harvested these cells from the spleen at 8 d postinfection. Following in vitro gp33 peptide stimulation, P14 recovered from Tim-3^Treg KO mice had an increase in the frequency of TNF⁺IFN-γ⁺ cells (Fig. 6E) relative to WT mice. However, the level of Gzmb expression of P14 T cells transferred into Tim-3^Treg KO mice was unchanged, compared with P14 cells transferred into WT mice (Fig. 6F). Overall, these findings support a model that Treg-specific Tim-3 deletion contributes to enhancing the effector functional capacity of LCMV-specific CD8⁺ T cells. Similarly, the lower level of Treg-mediated suppression in Tim-3^Treg KO mice is sufficient to preserve the functional capacity of transferred WT P14 cells early during chronic LCMV infection. Nonetheless, there may be differences in the sensitivity of individual T cell clones to this effect.

Next, we wanted to determine whether there are differences in the transcriptional identity of Tex cells when comparing the same Tex subsets between WT and Tim-3^Treg KO mice. Thus, we performed bulk RNA-seq of adoptively transferred WT P14 T cells into WT and Tim-3 Treg KO mice at 8 d postinfection (Fig. 7A). We sorted the P14 cells into three Tex subsets: PD-1⁻Tim-3⁻ (double negative [DN]), PD-1⁺Tim-3^- (PD-1^int), and PD-1⁺Tim-3⁺ (double positive [DP]). Principal component analysis revealed clear divergence between the three P14 Tex subsets from WT versus Tim-3^Treg KO mice (Fig. 7B). This analysis suggests that Tim-3^Treg KO drives (directly or indirectly) the development of Tex subsets with distinct transcriptional profiles during chronic infection.

Differential gene expression analysis of these RNA-seq data demonstrated a reduction in the transcript abundance of Tox in all three Tex groups from Tim-3^Treg KO mice, compared with WT mice (Fig. 7C). DN P14 T cells that had been transferred into Tim-3^Treg KO mice had elevated levels of transcripts encoding costimulatory molecules such as Cd226 and markers of effector T cells such as Id2, Gzma, and an increase in genes involved in T cell lymph node egress such as Cx3cr1 and S1pr5 (Fig. 7C). Analysis of the PD-1^int (precursor or stem-like exhausted) P14 population from Tim-3^Treg KO mice revealed a decrease in the inhibitory molecule Ctla4, along with an increase of the costimulatory molecule Cd86 (Fig. 7C). These cells also had an increase in genes associated with effector function (Fasl, Gzmb, Gzma, Id2, Klrg1, Tbx21 [T-bet]) and in genes encoding proteins that promote trafficking into the circulation (Cx3cr1 and S1pr1) (Fig. 7C). Finally, terminally exhausted DP P14 T cells from Tim-3^Treg KO mice showed lower expression of Lag3, in addition to elevated gene sets associated with T cell activation such as Gzmk, Gzmm, Id2, and Klrg1 (Fig. 7C). This population of DP P14 T cells also exhibited higher expression of genes important for migration, although they also upregulated Cd69, which antagonizes the function of S1pr1 (38–40) (Fig. 7C). Thus, these results demonstrate that deletion of Tim-3 in Tregs alters the identity of gp33-specific T cells at different stages of T cell exhaustion during chronic infection.

The goal of this study was to dissect the contribution of Tim-3⁺ Tregs to viral persistence and T cell responses during chronic viral infection. We used the LCMV Cl-13 chronic infection model, which establishes a long-term systemic high viral load, develops T cell exhaustion, and is widely used to model chronic infections in humans. As expected, we observed that Tim-3 is upregulated in Tregs during chronic LCMV infection. Strikingly, we found that conditional Treg-specific Tim-3 deletion was sufficient to ameliorate virus-specific T cell exhaustion and improve T cell function. Tim-3^Treg KO mice also had greater viral clearance at later time points.

In LCMV Cl-13 infection, the increase in Tim-3 expression seems to be unique to Tregs rather than Tconv cells. The kinetics for the proportion of total Tconv cells and Tim-3⁺ Tconv cells for acute and chronic LCMV infection did not show a notable difference compared with uninfected mice. Interestingly, this is also a phenotype that we have reported in PWH-ART (18). In addition, thymic development might be the source of most Tim-3⁺ Tregs that expand during chronic infection. In both infected and uninfected mice, splenic Tim-3⁺ Tregs had elevated expression of Helios and Foxp3 in comparison with the Tim-3⁻ Tregs. Therefore, these data demonstrate that during chronic infection, there is an increase of thymically derived Tim-3⁺ Tregs, but not in CD4⁺ Tconv cells.

Tim-3 expression by Tregs may promote a robust but short-lived effector Treg phenotype during chronic infection. Similar to our previous findings in cancer (12) and HIV (18), Tim-3⁺ Tregs upregulated important activation and immunosuppression markers such as CD25, CTLA-4, and IL-10 during LCMV Cl-13 infection. In addition, Tim-3⁺ Tregs in PWH-ART have an increase in the frequency of annexin V staining and active caspase-3, suggesting that these cells are prone to apoptosis (18). Interestingly, in chronic infection, these cells displayed a dramatic increase in expression of KLRG-1 and a decrease in Bcl-2. These findings indicate that Tim-3 expression identifies Tregs with a strong effector phenotype, but which may also have a relatively short lifespan. Previous studies evaluating Tim-3⁺PD-1⁺ Tregs found that the frequency of these cells increases at the peak of graft rejection and that ex vivo stimulation with galectin-9, a known ligand for Tim-3, induced cell death (17). In this same study, adoptive transfer of Tim-3⁺ Tregs failed to prolong graft survival, and these cells had elevated annexin V staining at 5 d posttransfer. In addition, Tim-3 seems unlikely to drive Treg instability because Tim-3⁺ Tregs maintained robust Foxp3 expression during infection. In conclusion, these findings demonstrate that Tim-3⁺ Tregs possess a robust immunosuppressive phenotype during persistent LCMV infection.

In our experiments, deletion of Tim-3 in Tregs led to rapid recovery from wasting disease that was independent of viral load during chronic infection. Previous work has shown that it is the CD4⁺ T cell compartment, and not necessarily the proinflammatory cytokines IFN-γ and TNF, that promotes wasting disease (41). Our findings with Tim-3^Treg KO mice indicate that Tim-3⁺ Tregs are major contributors to morbidity during chronic infection. The fact that Tim-3^Treg KO mice exhibited no change in the number of total Tregs is consistent with the idea that Tim-3⁺ Tregs are a key regulator of wasting disease.

Tim-3 has been described to have both inhibitory and stimulatory functions, depending on the cellular context (42–45). Our data demonstrating a positive association between Tim-3 expression and Treg effector status are more consistent with Tim-3 function on Tregs falling under the latter category. This is also supported by a decrease in CD39, a major marker of immune suppression, in total Tregs from Tim-3^Treg KO mice compared with their WT counterpart. Thus, even at the points of infection when Tim-3 did not seem to affect Treg number, these cells exhibited a phenotype indicative of reduced suppression. In addition, uninfected Tim-3^Treg KO mice did not experience any signs of wasting disease during the 40 d following tamoxifen administration, suggesting that lack of Tim-3⁺ Tregs is not enough to destabilize Tregs and thus unleash the development of spontaneous autoimmunity. Consistent with these findings, germline deletion of Tim-3 also does not appear to lead to autoimmunity (46). Therefore, our study suggests that Tim-3 is not required for the development of all effector Tregs, including those that suppress self-reactivity in specific tissues. Alternatively, the duration of our experiments was possibly insufficient for development of autoimmunity, and/or the genetic background of the mice (C57BL/6) was not conducive to the development of autoimmunity.

We also demonstrate that Tim-3⁺ Tregs control viral persistence and gp33-specific responses during chronic infection. This finding may result from the improved gp33-specific T cell function and the reduction in the development of T cell exhaustion in Tim-3^Treg KO mice. However, we did not assess the potential contribution of a simultaneous improvement in Ab-mediated clearance of virus. A previous study revealed that cessation of viremia in LCMV Cl-13 depends on the formation of viral envelope-specific Abs, and that mice with impaired B cell responses fail to resolve chronic infection (47). This study also demonstrated that neutralizing Ab titers start to increase after days 40–50 postinfection. Therefore, although it is possible that improved B cell responses contribute to the decrease in viral burden in Tim-3^Treg KO mice, the fact that we observed changes in viral burden at 10–20 d before the expected appearance of neutralizing Abs, together with the early improved T cell responses, supports a model where most of the improved viral control is driven by T cells.

Early interactions between Tim-3⁺ Tregs and DCs may also alter the priming phase of T cell activation and promote viral persistence. Our study revealed that Tim-3^Treg KO mice led to a reduction in PD-L1 expression by DCs. PD-L1 is an important molecule in chronic infection, as blocking this inhibitory ligand reduces chronic LCMV viral burden (34, 48). Upregulation of PD-L1 by DCs and infected cells restricts clearance of a chronic LCMV infection after Treg ablation (1). Our findings also support the idea that Tim-3⁺ Tregs limit the expression of CD86 in DCs and monocytes during infection, leading to a reduction in costimulatory capacity. Therefore, interactions of Tim-3⁺ Tregs with DCs and monocytes may induce a tolerogenic state that dampens T cell responses and promotes viral persistence.

Conditional deletion of Tim-3 in Tregs resulted in a decrease in exhausted T cells throughout chronic infection. This is a phenotype that we have also observed in terminally exhausted, tumor-infiltrating CD8⁺ T cells (12). Thus, Tim-3⁺ Tregs limit antitumor immunity and promote tumor progression, another context where there is chronic Ag exposure of T cells. In viral chronic infection, we observed these changes in gp33-specific Tex cell subsets as early as 8 d postinfection and they were even more pronounced later during infection. Interestingly, Ag-experienced T cells showed a significant difference in the proportion of T cell exhaustion markers (including both surface markers and Tox) only during late infection. These findings suggest that a decrease in the proportion of immunodominant gp33-specific Tex cells may contribute to the control of viral persistence in Tim-3^Treg KO mice.

Our findings further suggest that Tim-3⁺ Tregs may control viral persistence by promoting Tox-dependent gene programming in gp33-specific T cells. Bulk RNA-seq analysis of transferred P14 T cells early during the establishment of chronic infection revealed transcriptionally distinct profiles in PD-1⁻Tim-3^- (DN), PD-1^intTim-3⁻ (PD-1^int), and PD-1⁺Tim-3⁺ (DP) exhausted T cells in Tim-3^Treg KO versus WT mice. Common changes observed in all three populations included upregulation of transcripts related to T cell activation and effector function. Surprisingly, the only major Tex-related molecule downregulated in all three populations from Tim-3^Treg KO mice was Tox, a transcription factor that is required for the early commitment of dysfunctional T cells during chronic infection. In fact, a recent study demonstrated that enforced Tox expression is sufficient to induce a T cell exhaustion program (37). Therefore, early during chronic infection, Tim-3⁺ Tregs may promote strong epigenetic and transcriptional changes driven by Tox in T cells, which may promote exhaustion and thus limit viral clearance.

One limitation of this study is that we are currently not able to specifically track “ex–Tim-3⁺” Tregs. The ability to track such cells would help address the question of whether they lose the effector phenotype after Tim-3 deletion and/or whether Tim-3 is required for Treg survival and long-term maintenance. Nonetheless, based on previous work from our group, Tregs from people with HIV have reduced suppressive capacity in the presence of a putative blocking mAbs to Tim-3 (18), further supporting the model that Tim-3 is required for robust immunosuppressive responses. Finally, future studies are necessary to address whether changes in the spatial distribution of Tim-3⁺ Tregs within tissues are needed to achieve such profound effects on viral persistence.

In summary, this work supports a model whereby Tim-3⁺ Tregs control viral persistence by dampening gp33-specific T cell function and promoting the development of T cell exhaustion. Therefore, defining the mechanism by which Tim-3 functions on Tregs may provide insights into new therapeutic avenues for the management of chronic infections.

The authors have no financial conflicts of interest.

We thank Dr. E.J. Wherry’s laboratory for providing us with LCMV Cl-13 viral stock and BHK cells for further virus propagation in our lab. We also thank the University of Pittsburgh Unified Flow Core for help with flow cytometry. The visual abstract was created using BioRender.com.