Increased peripheral IL-4 leads to an expanded virtual memory CD8⁺ population

Abstract

Memory-phenotype CD8⁺ T cells can arise even in the absence of overt antigen stimulation. Virtual memory (VM) CD8⁺ T cells are CD8⁺ T cells that develop a memory phenotype in the periphery of WT mice in an IL-15-depedent manner. Innate CD8⁺ T cells, in contrast, are memory-phenotype CD8⁺ T cells which develop in the thymus in response to elevated thymic IL-4. It is not clear whether VM cells and innate CD8⁺ T cells represent two independent T cell lineages or whether they arise through similar processes. Here, we use mice deficient in Nedd4-family interacting protein 1 (Ndfip1) to show that overproduction of IL-4 in the periphery leads to an expanded VM population. Ndfip1^−/− CD4⁺ T cells produce large amounts of IL-4 due to a defect in JunB degradation. This IL-4 induces a memory-like phenotype in peripheral CD8⁺ T cells that includes elevated expression of CD44, CD122 and Eomes, and decreased expression of CD49d. This is the first study to show that excess peripheral IL-4 is sufficient to cause an increase in the VM population. Our results suggest that VM and innate CD8⁺ T cells may be more similar than previously appreciated.

Introduction

Memory CD8⁺ T cells arise from naïve CD8⁺ T cells following antigen stimulation and effector differentiation. While several different subsets of memory cells have been described, they generally share certain phenotypic and functional similarities, such as high CD44 expression in the absence of recent activation (1, 2). In addition to the conventional pathway of bona fide memory cell development, CD8⁺ T cells can also acquire a memory-like phenotype driven primarily by exposure to cytokine and weak TCR signals rather than overt antigen stimulation. For example, naïve CD8⁺ T cells in a lymphopenic environment undergo homeostatic proliferation (HP) and acquire a memory phenotype even in the absence of cognate antigen (3-5). This HP is driven by the relative increase of IL-7 and IL-15 in lymphopenic hosts in concert with tonic TCR signaling from low-affinity self-ligands (6-8).

Similar cells also have been observed in immunosufficient mice. Virtual memory (VM) cells are CD8⁺ T cells which acquire a memory-like phenotype in the periphery identical to that of HP memory cells (9-11). Like HP memory cells, VM cells develop even in the absence of exposure to cognate antigen. VM cells arise naturally in unimmunized mice and their development is dependent on IL-15 and partly dependent on IL-4 (10, 11).

Memory phenotype CD8⁺ T cells have also been characterized in a variety of genetic models that result in increased thymic IL-4 production by PLZF⁺ cells (12-15). In these models, IL-4 acts in trans to induce a memory phenotype in bystander CD8 SP thymocytes. Some innate-like T-cell subsets such as NKT cells also acquire a memory-like phenotype in the thymus (16) and thus the bystander CD8⁺ T cells in the previously described models are often called “innate” CD8⁺ T cells (17). These innate CD8⁺ T cells arise naturally in BALB/c mice, which have a much larger population of PLZF⁺ thymocytes than C57BL/6 mice (12). The relationship between VM cells and innate CD8⁺ T cells is unclear.

Nedd4-family interacting protein 1 (Ndfip1) restricts IL-4 production in CD4⁺ T cells by facilitating degradation of the transcription factor JunB (18, 19). Ndfip1-deficient CD4⁺ T cells have increased JunB levels and consequently overproduce IL-4. This excess IL-4 impairs Th17 and iTreg differentiation (19, 20). Whether loss of Ndfip1 and/or exposure to IL-4 affect CD8⁺ T cell development or function is not known. In this study, we show that IL-4 in the periphery of Ndfip1^−/− mice is sufficient to induce an expanded population of memory phenotype CD8⁺ T cells. The cells are phenotypically identical to VM cells, despite arising in response to IL-4. These data suggest that the distinction between innate and VM CD8⁺ T cells is a result of particular experimental conditions that alter the relative amounts and locations of common gamma chain cytokines. Further, it raises the possibility that VM cells may be clinically relevant in diseases which are characterized by local increases in IL-4, such as asthma.

Materials and Methods

Mice

Ndfip1^−/−, Ndfip1^−/− IL4^−/−, and Ndfip1^fl/fl CD4-Cre⁺ mice have been described previously (18, 20, 21). MHCII^−/− (B6.129S2-H2^dlAb1-Ea/J) and CD45.1+ (C57BL6.SJL-Ptprc^a Pepc^b/BoyJ) mice were purchased from The Jackson Laboratory. MHCII^−/− mice were bred to Ndfip1^+/- mice in our lab to generate MHCII^−/− Ndfip1^−/− mice. All mice were used at 5-16 weeks of age unless otherwise noted. Ndfip1^−/− mice were bred from heterozygous parents and WT littermates were used as controls, with the exception of data presented in Figure 3. For these experiments, mice were bred with one heterozygous and one KO parent, and Ndfip1^+/- littermates served as controls. In some cases, Ndfip1^−/− mice were also Rag1^+/-. No differences in T-cell phenotype were observed in Rag1^+/- vs. Rag1^+/+ mice. All mice were maintained in a barrier facility at the Children’s Hospital of Philadelphia. All animal experiments were approved and followed the guidelines set by the Institutional Animal Care and Use Committee at the Children’s Hospital of Philadelphia.

Figure 3.

Memory-like phenotype in Ndfip1^−/− CD8⁺ T cells first arises in the periphery. (A) Representative histograms of CD44 expression on CD8⁺ T cells from spleen and thymus in Ndfip1KO and control Ndfip1+/-mice at 5, 8, and 28 days old. (B) CD44, CD122, and CD124 expression on CD8⁺ T cells from spleen and thymus based on flow cytometric analysis. (C) Spleen and thymus cells were stimulated 4hr with PMA & ionomycin and culture supernatants were analyzed for IL-4 by ELISA. p values, unpaired Student’s t-test. Error bars = SEM. n = 4-8 per timepoint. *, p < .05; **, p < .01; ***, p < .001

Fetal-liver chimeras

Livers were isolated from embryonic WT (CD45.1⁺) or Ndfip1^−/− (CD45.2⁺) mice and single-cell suspensions were prepared by mashing through a 70μm filter. Cells were resuspended in freezing media (90% FCS, 10% DMSO) and kept at −80°C until used. Thawed cells were resuspended in sterile PBS and injected i.v. into sub-lethally irradiated Rag1^−/− recipients, ~5×10⁵ cells/mouse. The chimeras were used in experiments six weeks later to allow reconstitution of the T cell compartment.

Antibodies and reagents

The following fluorochrome-conjugated antibodies were purchased from Biolegend, eBioscience, or BD Pharmingen: anti-mouse CD3 (17A2), CD4 (RM4-5), CD8α (53-6.7), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD49d (R1-2), CD122 (TM-β1 and 5H4), CD124 (mIL4R-M1), Eomes (Dan11mag), and IL-4 (11B11). Unlabeled anti-CD16/32 (Fc block, 2.4G2) was purchased from BD Pharmingen. Unlabeled PLZF antibody (D-9) was purchased from Santa Cruz. SIINFEKL-H2K^b tetramer conjugated to Alexa Fluor 647 and PBS57-CD1d tetramer conjugated to APC were obtained from the NIH Tetramer Facility. Dead cell staining was performed using Molecular Probes LIVE/DEAD Fixable Dead Cell Stain Kit (Invitrogen).

Flow cytometry

Single-cell suspensions of lymphocytes isolated from spleen or thymus were stained for 10 min. on ice with live/dead stain, blocked for 5 min. with Fcblock, then stained a further 25 min. with antibodies directed against surface antigens. Cells were then washed twice with PBS plus sodium azide plus FCS (FACS buffer). For NKT-cell identification, CD1d tetramer was included in the surface antigen stain. For H2K^b tetramer staining, single-cell suspensions of combined spleen and lymph nodes were enriched for CD8⁺ T cells through negative selection using rat anti-mouse I-A/I-E (M5/114.15.2) and goat anti-rat IgG magnetic beads (New England Biolabs) according to the manufacturer’s protocol. Tetramer was added after blocking with Fcblock. Cells were then incubated on ice for 1 hour, washed twice with FACS buffer and stained 25 min. for surface antigens. For intracellular cytokine staining, after surface staining cells were fixed for 20 min. on ice with BD CytoFix/CytoPerm and washed twice with 1× BD perm/wash buffer. Cells were then resuspended in perm/wash buffer plus antibodies directed against cytokines and stained on ice 1 hour. For intracellular transcription factor staining, after surface staining cells were fixed 1-16 hours with eBioscience Foxp3 fix/perm solution and washed twice with 1× permeabilization buffer. Cells were then resuspended in FACS buffer plus antibodies directed against transcription factors and stained on ice 1 hour. For PLZF staining, cells were then washed and stained a further 30 minutes with anti-mouse IgG₁-FITC. Flow cytometry was performed on an LSRII or LSRFortessa (BD Biosciences) and results analyzed using Flow-Jo software (Treestar USA). Dead cells were excluded from analysis.

ELISA

Spleen cultures were set up at 1×10⁶ or 2×10⁶ cells/ml in complete DMEM. T cells were activated with 5 mg/ml soluble anti-CD3 and supernatants were collected 22 hrs after stimulation. NKT cells were activated with 200 ng/ml PBS44 and supernatants were collected 4 days after stimulation. All supernatants were kept at −80°C until used. ELISA was performed using the eBioscience Ready-Set-Go IL-4 kit according to the manufacturer’s instructions.

Results

Innate-like CD8⁺ T cells develop in Ndfip1-deficient mice

While Ndfip1 has a well-characterized role in activating ubiquitin complexes that prevent IL-4 production in CD4⁺ T cells, its function, if any, in CD8⁺ T cells is unknown. We therefore examined the phenotype of Ndfip1^−/− CD8⁺ T cells ex vivo by staining splenocytes from WT or Ndfip1^−/− mice for markers of activation and then analyzed expression by flow cytometry. We found that Ndfip1^−/− mice have an increased percentage of memory-phenotype CD8⁺ T cells (Fig. 1). Ndfip1^−/− CD8⁺ T cells are significantly more likely than WT CD8⁺ T cells to express high levels of CD44, CD122 (IL-2Rβ), and the transcription factor Eomesodermin (Eomes) (Fig. 1 A-B). Notably, Ndfip1^−/− CD8⁺ T cells express less CD49d (α4 integrin) than WT CD8⁺ T cells (Fig. 1 A-B). Low CD49d expression on CD44^hi cells has been described in HP memory and VM populations and CD49d has thus been suggested as a potential marker to distinguish true antigen-experienced memory cells from other memory-phenotype cells (8, 9, 11).

Figure 1.

Cell-extrinsic memory phenotype of Ndfip1^−/− CD8⁺ T cells. (A) Representative histograms of CD44, CD122, CD124, Eomes, and CD49d expression on splenic WT and Ndfip1KO CD8⁺ T cells. (B) CD44, CD122, CD124, Eomes, and CD49d expression on splenic WT and Ndfip1KO CD8⁺ T cells based on flow cytometric analysis. (C-F) WT (CD45.1⁺) or Ndfip1KO (CD45.2⁺) fetal liver was transferred singly or as a 50:50 mix into irradiated Rag1^−/− recipients. Six week later, splenic CD8⁺ T cells were isolated and analyzed by flow cytometry. (C) Representative histograms of CD44 and CD122 expression on CD8⁺ T cells from straight WT or Ndfip1KO chimeras (top) or from within one mixed chimera (bottom). (D) CD44 and CD122 expression in straight and mixed chimeras based on flow cytometric analysis. (E-F) Spleen cells were stimulated with anti-CD3 for 4hrs. (E) Representative histograms of IL-4 expression on CD4⁺ T cells from straight WT or Ndfip1KO chimeras (left) or from within one mixed chimera (right) after stimulation with anti-CD3 for 4 hrs. (F) IL-4⁺ cells as a percentage of total CD4⁺ T cells in straight and mixed chimeras after stimulation with anti-CD3 for 4 hrs. p values determined by unpaired Student’s t-test, except for comparisons within individual mixed chimeras in (D) and (F), which used paired Student’s t-test. Data are representative of three-five (A-B) or two (C-F) independent experiments. NS, not significant

Interestingly, Ndfip1^−/− CD8⁺ T cells also express increased levels of CD124 (IL-4Rα) (Fig. 1 A-B). CD124 is upregulated on T cells in response to IL-4 exposure (22) and high CD124 levels have been observed on innate-like CD8⁺ T cells that develop in an IL-4-rich environment (12). This suggested that exposure to cytokine may drive Ndfip1^−/− CD8⁺ T cells to acquire a memory-like phenotype. To test this, we generated chimeras by injecting WT (CD45.1⁺) or Ndfip1^−/− (CD45.2⁺) fetal liver alone (“straight chimeras”) or mixed in equal measure (“mixed chimeras”) into sub-lethally irradiated Rag1^−/− recipients. After reconstitution, we analyzed splenocytes from the chimeras by flow cytometry. As expected, CD8⁺ T cells from the straight Ndfip1^−/− chimeras expressed higher levels of both CD44 and CD122 than CD8⁺ T cells from the straight WT chimeras (Fig. 1 C-D). However, there was no significant difference between WT and Ndfip1^−/− cells that developed in the mixed chimeras, and this was largely due to increased expression of CD44 and CD122 on WT cells that developed in the presence of Ndfip1^−/− cells (Fig. 1 C-D). Thus, the memory-like phenotype of Ndfip1^−/− CD8⁺ T cells is cell-extrinsic. In contrast, upon stimulation with anti-CD3, Ndfip1^−/− CD4⁺ cells but not WT CD4⁺ cells produced IL-4 even in the mixed chimeras, suggesting that IL-4 production from Ndfip1^−/− cells drives acquisition of a memory phenotype in bystander CD8⁺ T cells (Fig. 1 E-F). CD8⁺ T cells, on the other hand, did not produce any IL-4 (data not shown).

Loss of Ndfip1 could potentially affect many cell types. To confirm that the memory phenotype of Ndfip1^−/− CD8⁺ T cells does not require loss of Ndfip1 outside the T-cell compartment, we crossed Ndfip1^fl/fl mice to mice expressing the Cre recombinase under the control of the CD4 promoter (CD4-cre⁺). The resulting mice (cKOs) lack Ndfip1 only in T cells. CD8⁺ T cells from these cKOs have a memory-like phenotype similar to that of CD8⁺ T cells from Ndfip1^−/− mice (Sup. Fig. 1 A-B; Fig. 1 A-B). As expected, the CD4⁺ T cells from cKO mice are more likely than control cells to produce IL-4 in response to stimulation (Sup. Fig. 1 C). These data indicate that loss of Ndfip1 in T cells only is sufficient to induce increased frequency of memory-like CD8⁺ T cells. To determine whether Ndfip1^−/− mice have an increased frequency of NKT or TCRγδ T cells, as has been observed in models of innate CD8⁺ T cells (15), we stained Ndfip1^−/− and WT spleen and thymus using CD1d tetramer loaded with PBS-57 (an α-galactosylceramide analogue) and anti-TCRγδ antibody. As shown in Supplementary Figure 2A and B, Ndfip1^−/− mice contain normal frequencies of TCRγδ T cells and a slightly lower frequency of NKT cells than WT mice. We also analyzed PLZF expression by intracellular staining and flow cytometry and determined that neither NKT cells nor TCRγδ T cells in Ndfip1^−/− mice express elevated PLZF levels compared to WT (Sup. Fig. 2 C). Finally, to compare IL-4 production by WT and Ndfip1^−/− NKT cells, we stimulated splenocyte cultures in vitro for 4 days with α-galactosylceramide analogue PBS44, then collected supernatants and analyzed with ELISA. We observed increased IL-4 production from Ndfip1^−/− NKT cells, although this was not statistically significant (Sup. Fig. 2 D).

The memory phenotype of Ndfip1KO CD8⁺ T cells is largely IL-4-dependent

The previous experiments suggested that increased IL-4 in Ndfip1^−/− mice causes CD8⁺ T cells to develop a memory-like phenotype. To test whether the CD8⁺ T-cell phenotype is really dependent on IL-4, we compared splenocytes from Ndfip1^−/−IL4^−/− mice and control Ndfip1^+/+IL4^−/− mice. Loss of IL-4 almost completely ablated the increase in memory-phenotype cells observed in Ndfip1^−/− mice (Fig. 2). Expression of CD122, CD49d, and CD124 was not statistically different in Ndfip1^−/−IL4^−/− CD8⁺ T cells compared to IL4^−/− controls (Fig. 2 A-B). Further, the percentage of cells expressing high levels of CD44 or Eomes was drastically reduced in Ndfip1^−/− IL4^−/− CD8⁺ T cells compared to Ndfip1^−/− cells from IL-4-sufficient mice, although a small but statistically significant increase compared to IL4^−/− controls remained (Fig. 2 A-B; Fig.1 A-B). Therefore, the overwhelming majority of memory-phenotype CD8⁺ T cells in Ndfip1^−/− mice require IL-4 for their development or maintenance.

Figure 2.

IL-4 deficiency ablates the memory phenotype of Ndfip1^−/− CD8⁺ T cells. (A) Representative histograms of CD44, CD122, CD124, Eomes, and CD49d expression on splenic IL4KO and IL4KO Ndfip1KO (DKO) CD8⁺ T cells. (B) CD44, CD122, CD124, Eomes, and CD49d expression on splenic IL4KO and DKO CD8⁺ T cells based on flow cytometric analysis. p values, unpaired Student’s t-test. Data are representative of three independent experiments. NS, not significant

Because CD4⁺ T cells are major producers of IL-4 in Ndfip1^−/− mice, we attempted to generate mice lacking CD4⁺ T cells by crossing Ndfip1^−/− mice to mice lacking all conventional MHC class II genes (MHCII^−/− mice). Surprisingly, we observed that Ndfip1^−/− MHCII^−/− CD8⁺ T cells had a phenotype similar to Ndfip1^−/− CD8⁺ T cells (Sup. Fig. 3 A-B; Fig. 1 A-B). MHCII^−/− mice have previously been described as containing a small percentage of mature peripheral CD4⁺ T cells which may be selected on nonconventional MHC molecules (23). We analyzed splenocytes from Ndfip1^−/− MHCII^−/− mice and confirmed that ~2-3% of lymphocytes were CD4⁺ T cells (Sup. Fig. 3 C). Nearly 100% of these cells were CD44^hi (data not shown). To test the possibility that this small but apparently activated population of cells could be producing IL-4, we stimulated total splenocytes with anti-CD3 overnight and analyzed IL-4 secretion by ELISA. As expected, neither WT nor control MHCII^−/− cells produced detectable levels of IL-4, while Ndfip1^−/− cells produced a great deal of this cytokine (Sup. Fig. 3 D). Notably, although IL-4 was greatly reduced in Ndfip1^−/− MHCII^−/− cells, it was still present at levels approximately 50 times the limit of detection (Sup. Fig. 3 D). Although this is much more IL-4 than was observed when invariant NKT cells were specifically stimulated with glycolipid for 4 days (Sup. Fig. 2 D), it is possible that this result represents IL-4 production by NKT cells. Next, we analyzed intracellular cytokine production after stimulation with PMA and ionomycin. Ndfip1^−/− MHCII^−/− CD4⁺ cells were significantly more likely to be IL-4⁺ than control MHCII^−/− CD4⁺ cells (Sup. Fig. 3 E). Taken together, this data suggests that Ndfip1 restricts IL-4 production even in non-classical CD4⁺ T cells (including, potentially, NKT cells), and the IL-4 produced in Ndfip1^−/− MHCII^−/− mice is sufficient to induce a memory-like phenotype in CD8⁺ T cells.

Ndfip1KO CD8⁺ T cells are IL-4-dependent virtual memory cells

Ndfip1^−/− CD8⁺ T cells displayed characteristics of both innate and VM CD8⁺ T cells in the previous experiments. One major difference between these two cells types is that VM cells typically arise in the periphery, while innate CD8⁺ T cells arise in thymus (11). To determine the origin of memory-phenotype Ndfip1^−/− CD8⁺ T cells, we sacrificed mice shortly after birth and compared CD8⁺ T cells isolated from thymus and spleen. Because Ndfip1 inheritance is non-Mendelian (unpublished observations), in order to generate more knockouts we bred in Ndfip1^+/-mice to Ndfip1^−/− mice and used Ndfip1^+/- littermates as controls. In Fig. 3 A-B it is apparent that splenic CD8⁺ T cells in Ndfip1^−/− mice begin to take on a memory phenotype as early as 8 days after birth. In contrast, Ndfip1^−/− CD8 SP thymocytes do not display a memory phenotype even at 4 weeks of age. This is consistent with results of IL-4 ELISA (Fig. 3 C) which indicate very little IL-4 production in the thymus compared to spleen. Thus, it appears that both IL-4 production and appearance of memory-phenotype CD8⁺ T cells occur primarily in the periphery, not the thymus, of Ndfip1^−/− mice.

To distinguish VM cells from antigen-experienced memory cells, it is helpful to use tetramer staining to identify cells specific for nominal antigen in unimmunized mice. To do this, we stained Ndfip1^−/− spleen and LN cells with SIINFEKL-H2K^b tetramer to identify endogenous OVA-specific CD8⁺ T cells. Tetramer⁺ CD8⁺ T cells from Ndfip1^−/− mice displayed elevated levels of CD44, CD122, CD124 and Eomes, and decreased expression of CD49d (Fig. 4 B-C). This phenotype was absent in Ndfip1^−/−IL4^−/− cells (Fig. 4 D-E). Thus, OVA-specific CD8⁺ T cells in naïve Ndfip1^−/− mice have the same memory phenotype observed in bulk CD8⁺ T cells, and this phenotype is dependent on IL-4. To further confirm that the memory phenotype can be induced in the absence of overt antigen stimulation, we next sorted for naïve (CD44^lo CD62L^hi CD25⁻) CD8⁺ T cells from WT mice expressing the congenic marker CD45.1. Naïve CD8⁺ T cells were then transferred to cKO (Ndfip1^fl/fl CD4-Cre⁺) or control (Ndfip1^fl/fl CD4-Cre⁻) recipients. Mice were bled nine days after transfer, at which point WT CD45.1 CD8⁺ T cells isolated from cKO recipients expressed higher levels of CD124 and Eomes than T cells isolated from control recipients, but did not yet display statistically significant differences in CD44 or CD49d expression (Fig. 5 A-B). At 14 days post transfer, mice were sacrificed and lymphocytes were isolated from spleens. At this timepoint, WT CD45.1 CD8⁺ T cells that had been transferred to cKO recipients not only clearly expressed more CD124 and Eomes than WT cells transferred to control recipients, but also expressed more CD44 and slightly less CD49d (Fig. 5 A-B). Expression of CD122 was not significantly different between T cells transferred to cKO and control recipients at either timepoint (data not shown). This experiment demonstrates that naïve, WT CD8⁺ T cells in the presence of Ndfip1^−/− T cells acquire a phenotype similar to VM cells. This suggests that the increase in memory-phenotype CD8⁺ T cells in Ndfip1^−/− mice could be driven at least in part by increased phenotypic conversion of naïve cells, rather than solely by proliferation of existing memory-like cells.

Figure 4.

Memory-like phenotype in Ndfip1^−/− CD8⁺ T cells is independent of nominal antigen exposure. (A) Representative contour plots of WT sample stained for surface markers and SIINFEKL-loaded tetramer as described in the Methods section (left), WT sample stained for surface markers with no tetramer (center), used to set tetramer⁺ gate, and positive control cells from an OTI⁺ Rag1^−/− mouse stained for surface markers and SIINFEKL-loaded tetramer (right). Cells in black are gated on CD8⁺ T cells, while cells in gray are gated on CD4⁺ T cells from the same sample. (B) Representative histograms of CD44, CD122, CD124, Eomes, and CD49d expression on splenic WT and Ndfip1KO tetramer⁺ CD8⁺ T cells. (C) CD44, CD122, CD124, Eomes, and CD49d expression on splenic WT and Ndfip1KO tetramer⁺ CD8⁺ T cells based on flow cytometric analysis. (D) Representative histograms of CD44, CD122, CD124, Eomes, and CD49d expression on splenic IL4KO and DKO tetramer⁺ CD8⁺ T cells. (E) CD44, CD122, CD124, Eomes, and CD49d expression on splenic IL4KO and DKO tetramer⁺ CD8⁺ T cells based on flow cytometric analysis. p values, unpaired Student’s t-test. NS, not significant

Figure 5.

Naïve WT CD8⁺ T cells acquire a memory phenotype in the presence of Ndfip1^−/− T cells. (A) Representative histograms of CD124, Eomes, CD44, and CD49d expression on WT CD45.1⁺ CD8⁺ T cells transferred to cKO or control recipients, isolated from blood (d9) and spleen (d14). (B) CD124, Eomes, CD44, and CD49d expression on these T cells based on flow cytometric analysis. p values, unpaired Student’s t-test. NS, not significant

The previous results establish that memory-phenotype Ndfip1^−/− CD8⁺ T cells have the characteristics of VM cells. However, it was recently suggested that IL-4 plays a minor role in the development of the VM population of WT mice and that VM cells arise primarily in response to IL-15 (11). To determine whether IL-4 affected the VM cells in Ndfip1^−/− mice, we examined the CD44^hi CD49d^lo VM cell population in WT and Ndfip1^−/− mice and their IL-4-deficient counterparts. Consistent with previous findings (10, 11), we observed a small but significant decrease in the CD44^hi CD49d^lo VM cell population in IL-4-deficient mice compared to WT mice (Fig. 6 A-B). Importantly, the majority of VM cells in Ndfip1^−/− mice are absent in Ndfip1^−/− IL-4^−/− mice. When endogenous OVA-specific CD8⁺ T cells were isolated from unimmunized mice, we again observed a large increase in CD44^hi CD49d^lo VM cells in Ndfip1^−/− mice that was ablated in the absence of IL-4 (Fig. 6 C). In contrast, the presumptive antigen-experienced population (CD44^hi CD49d^hi) is a similar percentage of total T cells regardless of the presence of IL-4 (Fig. 6 A, data not shown). Together, this indicates that the VM cell population expands when exposed to increased IL-4 levels.

Figure 6.

Virtual memory cells in Ndfip1^−/− mice are IL-4-dependent. (A) Representative contour plots of CD44 and CD49d expression on splenic WT, Ndfip1KO, IL4KO, and IL4KO Ndfip1KO (DKO) CD8⁺ T cells. (B) VM cells (CD44^hi CD49d^lo) as a percentage of total CD8⁺ T cells based on flow cytometric analysis. (C) VM cells (CD44^hi CD49d^lo) as a percentage of total tetramer⁺ CD8⁺ T cells based on flow cytometric analysis. p values, unpaired Student’s t-test. Data are representative of three independent experiments.

Discussion

Our data demonstrate that IL-4 in the periphery can lead to an expanded VM population. Although thymic IL-4 has a well-characterized role in the generation of innate CD8⁺ T cells, to our knowledge this is the first study to link overproduction of IL-4 to an increase in VM cells. Our data is consistent with recently published data (10, 11) showing that the naturally-occurring VM population in WT mice is partially decreased in the absence of IL-4. One explanation for the comparatively minor role for IL-4 observed in those studies is that they primarily described VM cells in WT C57BL/6 mice, which have very low IL-4 levels. Indeed, it is possible that multiple cytokines contribute to the formation of the normal VM population, though clearly IL-15 has a crucial role as described in (11). However, our data show that when IL-4 is produced at high levels in peripheral compartments, it is an important driver of VM cells, leading to a fairly specific increase in VM cells compared to conventional antigen experienced cells (Fig. 6). Our results indicate that the IL-4-rich Ndfip1^−/− host environment has the potential to induce phenotypic conversion of naïve CD8⁺ T cells (Fig. 5), but it is likely that the large endogenous VM population in Ndfip1^−/− mice represents a combination of both increased conversion of naïve cells and increased proliferation of existing VM cells.

By establishing that high IL-4 levels can lead to increased VM cells, our data also suggests a possible link between innate CD8⁺ T cells and VM cells. The primary differences that have been described in these two memory-phenotype populations are their relative reliance on IL-4 and their origins in the thymus vs. the periphery. Both these differences may actually stem from differences in the experimental systems in which they were described. Unlike VM cells, innate CD8⁺ T cells have been described primarily in systems with elevated percentages of IL-4⁺ PLZF⁺ T cells. These IL-4⁺ PLZF⁺ T cells are, by and large, non-conventional T cells that become activated in and localize to the thymus. In contrast, Ndfip1^−/− mice have greatly increased IL-4 levels compared to WT mice, but the IL-4 is produced by cells in the periphery. It is thus not surprising that the expansion of memory-phenotype CD8⁺ T cells in Ndfip1^−/− mice occurs in the periphery in an IL-4-dependent manner. This result may suggest that innate CD8⁺ T cells and VM cells are not fundamentally different cell types, but rather reflect different pathways leading to acquisition of the same phenotype. Further characterizations of these cell types, including direct comparisons of functional properties, are necessary to address this possibility.

Development of both VM cells and innate CD8⁺ T cells is dependent on Eomes (11, 24). Sosinowski et al. point out that this links memory-phenotype cells with IL-15, as IL-15 increases Eomes expression, which in turn increases CD122 expression and therefore sensitivity to IL-15 (25, 26). However, IL-4 also induces Eomes expression in CD8⁺ T cells (12, 27, 28), as does IL-2 (29). Moreover, the timing of phenotypic changes to naive WT cells transferred to cKO recipients (Fig. 5 A-B) is consistent with a model in which IL-4 exposure, indicated by increased CD124, leads to an increase in Eomes expression, which then promotes acquisition of a memory phenotype. In light of this, it seems probable that memory-phenotype CD8⁺ T cells can arise in response to localized increases in any of several cytokines.

It has been speculated that unconventional memory-phenotype CD8⁺ T cells may be important in early responses to infection (15, 30). However, our data suggests that these cells may be especially relevant in autoimmune or allergic disorders, which frequently involve increased local production of IL-4. For example, CD8⁺ T cells can play a role in the pathology of asthma, and some of this may be due to bystander effects of IL-4. In mouse models of allergic airway disease, CD8⁺ T cells that develop in the presence of IL-4⁺ CD4⁺ cells worsen lung pathology (31, 32). Additionally, increased bronchial CD8+ cell infiltrate in human asthma patients is predictive of decreased lung function (33). Future research should address the question of whether memory-phenotype CD8⁺ T cells can be generated in a disease setting and, if so, whether they contribute to the pathology of inflammatory diseases.

Abbreviations used in this article

cKO

conditional knockout

Eomes

Eomesodermin

HP

homeostatic proliferation

Ndfip1

Nedd4-family interacting protein 1

PLZF

promyelocytic leukemia zinc finger protein

SP

single-positive

VM

virtual memory

WT

wild-type