IL-15–Independent Maintenance of Tissue-Resident and Boosted Effector Memory CD8 T Cells

This article is featured in In This Issue, p.3497

After clearance of an acute infection, expanded populations of pathogen-specific memory CD8 T cells are maintained and can be broadly divided into three distinct subsets, each with discrete trafficking properties and roles in immunosurveillance (1). Central memory CD8 T cells (T_CM) are recirculating T cells that migrate between blood and lymph nodes (LN) by crossing high endothelial venules (HEV) and are typically poised to proliferate should they re-encounter Ag. Effector memory CD8 T cells (T_EM) patrol blood and certain nonlymphoid tissues and constitutively express markers of effector differentiation, but lack the ability to cross HEV. Resident memory CD8 T cells (T_RM) make up a third, more recently appreciated subset that does not leave and re-enter tissues (and thus does not cross HEV). T_RM can be found in nonlymphoid tissues, certain vascular compartments, and secondary lymphoid organs (SLO), and they accelerate control of local infections (2, 3). Recent work highlighted the collaboration between resident and recirculating CD8 T cells in the event of reinfection; however, the factors that regulate the size of these populations and how subsequent unrelated infections alter T cell numbers are not completely understood.

IL-15 is a common γ-chain cytokine that promotes homeostatic proliferation (HP) and survival of circulating T_CM and T_EM after primary infection (4–6). IL-15 is also required for the development and maintenance of CD103⁺ CD8 T_RM in the skin epidermis following HSV infection (7). Consequently, one hypothesis is that IL-15 defines the host’s carrying capacity for memory CD8 T cells within lymphoid and nonlymphoid tissues, suggesting that subsequent infection may result in the displacement of pre-existing memory CD8 T cells (8), in part because of competition for IL-15 (9). However, it is unclear whether all memory CD8 T cells require IL-15, and IL-15–independent CD8 T cells were observed after local lung infection, by T_RM in secondary lymphoid organs, and during persistent infection (10–13).

Heterologous prime-boost-boost (HPBB) vaccination involves immunizing a host with three serologically distinct vectors that all carry the same conserved CD8 T cell epitope. Iterative T cell stimulation with potent HPBB vaccines can generate amplified populations of recirculating memory CD8 T cells that undergo less contraction with each subsequent boost (14). Most memory CD8 T cells generated by HPBB are T_EM, as defined by their lack of CD62L expression (15, 16). However, T_EM generated by HPBB are phenotypically distinct from T_EM generated after a primary infection (17). Of note, we demonstrated previously that HPBB vaccination did not induce substantial attrition of pre-existing CD8⁺ T_CM specific for other pathogens; thus, HPBB immunization substantially increased the overall magnitude of the memory CD8 T cell compartment (15). How HPBB is able to introduce a large new pool of T_EM without causing coordinate erosion of pre-existing memory CD8 T cells is unknown.

In this study, we evaluated the role of IL-15 in different contexts of memory T cell differentiation, including HPBB vaccination and a primary viral infection that establishes broadly distributed T_RM.

C57BL/6 mice were purchased from The Jackson Laboratory. All mice were used in accordance with National Institutes of Health and the University of Minnesota Institutional Animal Care and Use Committee guidelines. For experiments analyzing CD8 T cells generated by the HPBB system, N_52–59-specific primary memory CD8 T cells were generated by i.v. infection of CD45.2⁺ naive C57BL/6 mice (age matched to tertiary C57BL/6 mice) with 1 × 10⁷ PFU vesicular stomatitis virus Indiana strain (VSVind). For the generation of tertiary memory CD8 T cells, CD45.1/CD45.1⁺ 8–10-wk-old naive C57BL/6 mice were primed i.v. with 5 × 10⁵ PFU vesicular stomatitis virus New Jersey strain, rested for 60–90 d, infected i.v. with 2 × 10⁶ PFU recombinant vaccinia virus expressing the N protein of VSV (18), rested for an additional 60–90 d, and challenged i.v. with 1 × 10⁷ PFU VSVind. For homeostatic maintenance experiments, 5 × 10⁵ primary P14 or H-2K^b-N_52–59–specific tertiary memory CD8 T cells were cotransferred into naive C57BL/6J or IL-15^−/− recipients. The number of donor cells in spleen was quantified 60 d posttransfer. To measure HP, donor cells from primary P14 or H-2K^b-N_52–59–specific tertiary memory CD8 T cells were labeled with CFSE or CellTrace Violet, according to the manufacturer’s protocol (Invitrogen). A total of 5 × 10⁵ labeled cells was transferred into naive C57BL/6J or IL15^−/− mice. Turnover was analyzed in spleen 60 d posttransfer.

For the generation of memory P14 CD8 T cells, 5 × 10⁴ naive Thy1.1 P14 CD8⁺ T cells were transferred i.v. into naive C57BL/6 or IL15^−/− recipients. The following day, mice were infected i.p. with 2 × 10⁵ PFU lymphocytic choriomeningitis virus (LCMV) Armstrong. For BrdU experiments, at 52 d postinfection, P14 memory wild-type (WT) or IL15^−/− chimeras were given 0.8 mg/ml BrdU in a solution of 2% sucrose/drinking water for 8 d.

Lymphocytes were isolated from peripheral tissues, as previously described (19). Isolated lymphocytes were stimulated with LCMV gp_33–41 peptide (1 μg/ml) for 4 h at 37°C, and cytokine secretion was evaluated as described earlier (20). Single-cell suspensions were stained with Abs against CD69 (H1.2F3), CD44 (IM7), Thy1.1 (OX-7), and CD62L (MEL-14) (all from BioLegend); CD8a (53-6.7), CD103 (M290), CD45.1 (A20), CD45.2 (104), IFN-γ (XMG1.2), TNF-α (MP6-XT22), and IL-2 (JES6-5H4) (all from eBioscience); and BrdU (B44) (BD Biosciences). Samples were acquired on a Fortessa flow cytometer (BD Biosciences). Immunofluorescence was performed as described (21). Abs against Thy1.1 (OX-7), CD45.1 (A20), and CD8β (YTS156.7.7) (all from BioLegend) were used to stain sections.

For in situ tetramer staining, mouse female reproductive tracts (FRT) and salivary glands (SG) were manually cut into ∼500-μm sections using a surgical blade. Tissue pieces were incubated with PE-conjugated tetramers in 2% FBS in PBS overnight at 4°C. The following day, tissues were washed with ice-cold PBS three times on ice. Tissue pieces were fixed in 2% paraformaldehyde in PBS for 2 h, washed in PBS, and placed in a 30% sucrose PBS solution overnight. The following day, tissue pieces were frozen in OCT.

HPBB vaccination was shown to establish abundant T_EM with minimal corresponding erosion in pre-existing T_CM, leading to a net increase in total host memory CD8 T cells (15). How additional memory CD8 T cells are accommodated is unknown, which led us to investigate whether HPBB CD8 T_EM are controlled by different homeostatic mechanisms compared with primary circulating memory CD8 T cells. We immunized CD45.1⁺ C57BL/6 mice on three occasions with serologically distinct viruses (vesicular stomatitis virus New Jersey strain, followed by recombinant vaccinia virus expressing the nucleoprotein from vesicular stomatitis virus and then VSVind, see 2Materials and Methods) that each stimulates an H-2K^b-N_52–59–specific CD8⁺ T cell response (HPBB vaccination). Mice were rested for 60–90 d between immunizations; HPBB vaccination resulted in abundant CD45.1/CD45.1⁺ tertiary memory CD8 T cells specific for H-2K^b-N_52–59 (14). We generated primary memory CD8 T cells by adoptively transferring naive Thy1.1⁺ P14 CD8 T cells that are specific for the glycoprotein of LCMV into naive C57BL/6 mice. Mice were infected with LCMV on the following day. Sixty-five days after primary LCMV or tertiary HPBB immunization, lymphocytes were isolated from the spleens of both mice, labeled with CFSE, and cotransferred into C57BL/6 recipients to monitor basal turnover within the same mice (Fig. 1A). Sixty days after transfer, splenocytes were isolated, and CFSE dilution was assessed on primary P14 and tertiary H-2K^b/N MHC I tetramer⁺ memory CD8 T cells by flow cytometry. As shown in Fig. 1B and 1C, tertiary memory CD8 T cells underwent fewer homeostatic divisions than primary CD8 T cells. Because CD62L⁻ memory CD8 T cells undergo less HP than CD62L⁺ memory CD8 T cells, this may be explained by the enrichment of CD62L⁻ cells in tertiary immune populations (Fig. 1D) (17, 22). Therefore, we segregated our analysis based on CD62L expression and found that, regardless of CD62L expression, tertiary memory CD8 T cells underwent less HP compared with their primary memory CD8 T cell counterparts (Fig. 1E, 1F). Thus, primary and tertiary memory CD8 T cells exhibited fundamental differences in their rate of HP that was not completely defined by CD62L expression.

The common γ−chain cytokine IL-15 is required for HP and maintenance of primary memory CD8 T cells after acute infection (23). To determine whether IL-15 is required for maintaining tertiary memory CD8 T cells, we cotransferred Thy1.1⁺ primary and CD45.1⁺ tertiary memory CD8 T cells into CD45.2⁺ WT or IL-15^−/− naive recipients. After 60 d, we enumerated Thy1.1⁺ P14 and H-2K^b/N_52–59 MHC I tetramer⁺ CD8 T cells in the spleen. As shown in Fig. 2, recovery of primary and tertiary CD62L⁺ memory CD8 T cells was decreased 60 d after transfer into IL-15^−/− mice compared with controls. In contrast, maintenance of tertiary CD62L⁻ N_52–59-specific CD8 T cells was not observably impaired in IL-15^−/− mice. These data indicate that tertiary CD8 T_EM are less stringently dependent on IL-15 for maintenance, which could explain how they exceed IL-15–defined carrying capacities that may dictate the size of CD8 T_CM and primary CD8 T_EM populations.

The presence of memory CD8 T cells within barrier tissues accelerates protection against local infections (24–26). It is unknown whether the total number of memory CD8 T cells within mucosal tissues is homeostatically regulated and, thus, fixed at a defined overall carrying capacity or whether mucosal memory CD8 T cell quantity is flexible in size. This issue is particularly important given the impetus to design vaccines that attempt to establish abundant memory CD8 T cells at mucosal surfaces. To address whether the FRT and SG had a defined carrying capacity for memory CD8 T cells, we compared total CD8β⁺ lymphocytes, as well as N_52–59-specific memory CD8 T cells, in the FRT and SG via in situ tetramer staining and immunofluorescence microscopy among naive, primary immune (VSVind only), and tertiary immune (HPBB, as above) mice. As shown in Fig. 3A and 3B, HPBB-vaccinated mice had ∼20-fold more N_52–59-specific memory CD8 T cells in the FRT and SG compared with mice that received a single immunization. Primary and tertiary memory CD8 T cells in the FRT expressed CD69 and low levels of CD62L, suggesting that they may include resident populations (Fig. 3C). In addition, HPBB immunization precipitated an ∼30–50-fold increase in the total number of CD8β⁺ T cells within the FRT and SG compared with naive mice (Fig. 3D, 3E). These data suggested that HPBB immunization increased the magnitude of recirculating T_EM in blood and spleen (15), as well as the magnitude of mucosal T_RM. Moreover, we also found that a single immunization increased the total number of CD8β⁺ T cells in the FRT ∼6-fold compared with naive mice. This suggested that the numerical size of the CD8β⁺ T_RM population within the mucosa was quite flexible and perhaps even primary memory CD8 T cells were not subject to the constraints of IL-15–mediated homeostatic regulation in this compartment.

We next tested whether primary CD8 T_RM that became established within mucosal tissues could persist in the absence of IL-15. To this end, we used a model system in which resident memory populations were characterized extensively via parabiosis (19). Naive Thy1.1⁺ P14 CD8 T cells were transferred into naive C57BL/6 WT or IL-15^−/− mice. The following day, mice were infected with LCMV Armstrong, and P14 CD8 T cells were enumerated at the peak of the effector response (day 7) and at a memory time point (day 60).

Seven days postinfection, there was no significant difference in the number of P14 CD8 T cells between WT and IL-15^−/− mice in any tissue examined (Fig. 4A), suggesting that IL-15 is not necessary for the generation of effector CD8 T cells or their migration to nonlymphoid tissues after LCMV infection. Consistent with the dependence of recirculating primary memory CD8 T cell populations on IL-15 for maintenance, we found a substantial reduction in memory CD8 T cells within LN and spleen at 60 d postinfection (Fig. 4B). However, in several tissues that are populated almost exclusively by T_RM, including the FRT and the small intestine (SI), we did not observe a reduction in memory CD8 T cells in the absence of IL-15 or a significant change in the expression of CD103 (Supplemental Fig. 1). Nevertheless, IL-15 dependence varied demonstrably among T_RM populations within different nonlymphoid tissues. Indeed, T_RM in the SG and kidney were almost completely lost in IL-15^−/− mice. These data reveal the existence of CD8 T_RM that are homeostatically maintained independently of IL-15, but they also highlight potential heterogeneity among T_RM that inhabit different tissues.

The long-term maintenance of the circulating memory CD8 T cell pool is intrinsically tied to IL-15–driven homeostatic self-renewal. Therefore, we posited that IL-15 dependence in nonlymphoid tissue would dictate whether CD8 T_RM underwent HP in different tissues. To test this hypothesis, we treated WT and IL-15^−/− P14 LCMV immune chimeras with BrdU in the drinking water for 8 d and then examined BrdU incorporation to measure CD8 T cell turnover. As shown in Fig. 5, ∼15% of circulating memory CD8 T cells in spleen or LN of WT mice incorporated BrdU after 8 d, whereas its incorporation by circulating memory CD8 T cells was dramatically reduced in IL-15^−/− mice. Additionally, SLO T_RM, a minority population that can be identified on the basis of CD69 expression in the LCMV infection model (11), also underwent less HP in IL-15^−/− hosts compared with WT hosts. Interestingly, the CD69⁺ resident population became enriched relative to their recirculating CD69⁻ counterparts in SLO, suggesting that SLO T_RM were dependent on IL-15 for HP but not for survival. CD8 T_RM in the thymus exhibited a similar decrease in BrdU incorporation in IL-15^−/− mice without a decrease in overall numbers (Fig. 4B). These data demonstrate that SLO and thymus T_RM persistence does not require HP.

Unlike CD8 T_RM in SLO and thymus, CD69⁺ CD8 T_RM in the SG and kidney were dependent on IL-15 for survival and HP. This suggests that T_RM are composed of distinct subsets with heterogeneous requirements for IL-15 and HP. In support of this interpretation, T_RM in the FRT, SI, and pancreas underwent significant levels of HP, even though these T_RM populations were maintained in the absence of IL-15. Moreover, FRT, SI, and pancreas T_RM incorporated equivalent BrdU in IL-15^−/− mice, demonstrating IL-15–independent HP. In summary, these data reveal the existence of T_RM that divide and depend on IL-15 for survival, that divide in an IL-15–dependent manner but do not depend on IL-15 for survival, and that divide and survive independently of IL-15. Thus, T_RM are a heterogeneous population composed of distinct subsets with unique maintenance requirements and patterns of self-renewal.

CD8 T cell–derived cytokines are important mediators of host protection, and the ability of CD8 T cells to secrete cytokines is often used as a metric for assessing T cell quality (27). Previous work demonstrated the ability of IL-15 to promote CD8 T cell secretion of cytokines (28–30); thus, we next tested whether IL-15 deficiency would have deleterious effects on CD8 T_RM function. P14 immune chimeras were established in WT and IL-15^−/− mice (as in Fig. 4). Thirty days after LCMV infection, T_RM were isolated from the SI epithelium, the pancreas, and the FRT, and circulating memory CD8 T cells were recovered from spleen. Isolated memory CD8 T cells were restimulated in vitro with gp_33–41 peptide for 4 h, and secretion of IFN-γ, IL-2, and TNF-α was assessed by flow cytometry. We specifically assessed these three cytokines because they were shown to mediate a previously reported T_RM sensing and alarm function (31, 32). In all tissues examined, we found that T_RM were able to secrete all three cytokines independently of IL-15 (Fig. 6). Further, circulating memory CD8 T cells isolated from the spleens of IL-15^−/− mice exhibited only a minor defect in the capacity to secrete IFN-γ and TNF-α, which is consistent with previous observations (33). Thus, we failed to detect a role for IL-15 in regulating CD8 T_RM cytokine polyfunctionality.

It was proposed that the abundance of memory CD8 T cells positioned within barrier tissues could be a critical determinant of host protection (34). One deterrent against developing vaccines based on this objective is the concern that each newly introduced T_RM would displace a T_RM specific for previously encountered infections. In other words, vaccination could compromise host immunity against commonly encountered pathogens. This remains an important consideration that demands further exploration, but our data may provide some perspective on this hypothesis. The introduction of new T_RM simply resulted in more CD8 T cells within the FRT, and the total number of CD8 T cells varied over a 50-fold range depending on the infectious experience. A previous report demonstrated that, like T_CM and T_EM, T_RM were IL-15 dependent (7). However, this analysis was based solely on CD103⁺ T_RM in the skin epidermis. Our data revealed that T_RM within different anatomic compartments varied considerably in their IL-15 dependence. Teleologically, regulating FRT T_RM via IL-15 may have the unwanted consequence of inducing lymphopenic-driven expansion of T_RM in primary infected hosts. This may not be problematic in the skin because T_RM precursors compete with a pre-established visibly saturating population of dendritic epidermal T cells that densely colonizes the skin during development (35). In summary, these data support the hypothesis that distinct tissues regulate different populations of T_RM, which may point to the existence of distinct T_RM subsets. Moreover, the total number of T_RM in nonlymphoid tissues may not be stringently regulated because they varied over a ≥50-fold range, although further studies will be needed to test whether an upper limit is achievable under physiological conditions.

Observations of a dynamic range of total memory CD8 T cells are not limited to T_RM. HPBB vaccination results in increased numbers of total memory CD8 T cells within the host, which are largely contained within the T_EM pool. Many natural iterative stimulation events, either through heterosubtypic infections or persistent infections, may also lead to uncommonly large CD8⁺ T_EM populations. Thus, it may be disadvantageous for the host to evacuate its pre-existing memory CD8 T cells to accommodate expanded T_EM populations. Consistent with this conjecture, natural infection of humans with EBV, which causes a persistent viral infection resulting in robust CD8 T cell responses and increased CD45RA⁺ T_EM (T_EMRA), did not induce a measureable reduction in pre-existing CD8 T cells specific for CMV or influenza virus (36). In this study, we demonstrate that, unlike T_CM and primary T_EM, tertiary T_EM were maintained after transfer to IL-15^−/− hosts. This provides a potential mechanistic explanation by which the iteratively stimulated T_EM population may result in increases in net memory CD8 T cells, as well as highlights heterogeneity in the regulation of circulating memory CD8 T cell subsets. Of course, these data do not exclude an important role for IL-15 in the generation of boosted memory CD8 T cells (33), nor observations that certain infections induce attrition of pre-existing memory CD8 T cells via acute induction of type I IFN, LN fibrosis, or other means (8, 37).

We thank other members of the Masopust and Vezys laboratories for helpful discussions.

The authors have no financial conflicts of interest.