NK Cells Delay Allograft Rejection in Lymphopenic Hosts by Downregulating the Homeostatic Proliferation of CD8⁺ T Cells Free

Lymphopenia, defined as reduced number of mature lymphocytes in the blood and peripheral lymphoid tissues, is observed in experimental animals and humans after viral infection, treatment of malignancy, and lymphodepletion for bone marrow or solid organ transplantation (1–3). The re-establishment of the peripheral T cell pool after lymphopenia is accomplished by both thymic emigration of naive T cells and the spontaneous (homeostatic) proliferation of peripheral naive and memory T cells (4–7). A characteristic feature of homeostatic proliferation is the rapid generation of functional memory T cells from naive lymphocytes in the absence of cognate antigenic stimulation (8–12). Although beneficial to the host by providing immunity against infections and tumors (13–15), homeostatically generated memory T cells can cause autoimmunity (16, 17) and, in the setting of organ transplantation, precipitate allograft rejection and resist tolerance induction (18–22). Therefore, understanding the mechanisms that regulate homeostatic T cell proliferation is essential for avoiding the unwanted consequences of lymphodepletion.

The homeostatic proliferation of T cells in lymphopenic animals is driven by the same factors that govern the maintenance of naive and memory T cells under steady state conditions (1, 3). These are TCR interactions with peptide-MHC (pMHC) ligands and the actions of cytokines that bind the common cytokine receptor γ-chain (γc; or CD132), particularly IL-7 and IL-15 (23–26). Experiments in which naive T cells are transferred to lymphocyte-deficient mice have identified three forms of homeostatic T cell proliferation (3). “Slow” homeostatic proliferation, also known as lymphopenia-induced proliferation (LIP), is observed in wild-type (wt) mice rendered acutely lymphopenic by irradiation, cytotoxic agents, or anti-T cell Abs, a situation that mimics therapeutic lymphodepletion in humans and is driven by the increased availability of IL-7 and self-pMHC ligands. In chronically lymphopenic animals, such as genetically lymphocyte-deficient RAG^−/− mice, LIP coexists with a rapid form of homeostatic T cell proliferation driven by gut commensals (27). The latter resembles Ag-induced T cell division and is referred to as “fast” or chronic LIP. A third form of homeostatic proliferation, recently dubbed cytokine-induced proliferation, occurs in mice that lack either the IL-2Rβ– or γ–chain, whereby T cells proliferate in response to excess IL-2 and IL-15 (28). Therefore, the rate of homeostatic T cell proliferation is regulated to a large extent by competition for cytokines and pMHC ligands in the lymphopenic host.

NK cells are lymphoid cells with multiple effector functions. They play an important role in anti-viral and anti-tumor innate and adaptive immune responses and can cause the rejection of bone marrow allografts (29). It is increasingly recognized, however, that NK cells are also potent regulators of T cell immunity. They do so through inhibition of clonal expansion of Ag-stimulated CD4⁺ and CD8⁺ T cells by killing activated T cells (30–32), eliminating Ag-presenting dendritic cells (33–36), or producing inhibitory cytokines (37). Because both NK cells and CD8⁺ T cells are dependent on IL-15 for survival and self-renewal, it is also possible that NK cells regulate the size of the CD8⁺ T cell pool by competing for IL-15. In this study, we tested the hypothesis that NK cells regulate the homeostatic proliferation of CD8⁺ T cells in lymphopenic environments and studied the effect of NK cell depletion on allograft rejection in lymphocyte-deficient hosts.

C57BL/6 (B6), B6.SJL-Ptprc^a, BALB/c, B6 RAG2^−/−, B6 perforin^−/−RAG2^−/−, and B6 γc^−/−RAG2^−/− mice were purchased from Taconic Farms (Germantown, NY). B6.PL-Thy1a/CyJ mice were purchased from The Jackson Laboratory (Bar Harbor, ME). All mice were housed in a specific pathogen-free environment and were used between 6 and 12 wk of age. Experiments were performed in accordance with the Institutional Animal Care and Use Committee of the University of Pittsburgh.

Fluorochrome- or biotin-conjugated mAbs against CD11b (clone M1/70), CD11c (HL3), Ter119 (Ly-76), pan-NK (DX5), NK1.1 (PK136), Gr-1 (RB6-8C5), I-A^b (AF6-120.1), B220 (RA3-6B2), CD8a (53-6.7), CD4 (RM4-5), CD62L (MEL14), and CD44 (IM7) as well as streptavidin-conjugated Pacific Blue were from eBioscience (San Diego, CA) or BD Pharmingen (San Jose, CA). Purified anti-CD16/32 (Fc-Block, 2.4G2) was from BD Pharmingen. PE-conjugated polyclonal anti-NKp46, biotin-conjugated anti–IL-15Rα, recombinant human IL-15, and recombinant mouse IL-15Rα–Fc chimera were from R&D Systems (Minneapolis, MN). Ethidium monoazide and CFSE were from Invitrogen (Carlsbad, CA). NK cell (PK136)-, CD4 (GK1.5)-, and CD8 (2.4)-depleting Abs, the corresponding isotype control Abs, and anti-CD154 (MR1) were from BioXCell (West Lebanon, NH).

Full- thickness BALB/c tail skin grafts (1 cm²) were transplanted to the dorsal trunk area of B6 mice as described (38) and secured with a bandage for 7 d. Grafts were then monitored every other day by visual inspection. Rejection was defined as complete loss of viable graft tissue. Heterotopic transplantation of primarily vascularized cardiac allografts was performed as described (39). Mice were monitored daily by abdominal palpation. Rejection was defined as cessation of palpable heartbeat and was confirmed histologically.

NK cell-depletion of B6 RAG^−/− mice was accomplished by injecting 250 μg anti-NK1.1 mAb i.p. on days −1, 1, and 4 relative to the adoptive transfer of T cells. IL-15/IL-15Rα–Fc was administered i.p. as previously described (40). T cells were depleted from wt mice on the indicated days by administering two doses of anti-CD4 and anti-CD8 mAb i.p. (100 μg each) 5 d apart. A subgroup of wt mice also received 250 μg anti-NK1.1 mAb i.p. every 5 d until the end of the experiment. Donor-specific transfusion (DST) consisted of 1 × 10⁷ allogeneic splenocytes that were irradiated (1000 Rad) and administered i.v. on the day of skin grafting. A total of 250 μg anti-CD154 was administered i.p. on days 0, 2, 4, and 6 relative to transplantation.

T cells were negatively enriched from pooled spleens and lymph nodes (LNs) (peripheral and mesenteric) of naive mice by incubating single-cell suspensions (prepared by manual mincing and hypotonic RBC lysis) with biotin-conjugated mAbs to CD16/32, B220, CD11b, CD11c, NK1.1, and Ter119, followed by binding to streptavidin-coated microbeads and separation by AutoMACS (both from Miltenyi Biotech, Auburn, CA). Resulting T cell purity was >85%. Where specified, naive (CD44^lo), memory (CD44^hi) CD4⁺, and CD8⁺ T cells, or central (CD44^hiCD62L^hi) and effector (CD44^hiCD62L^lo) memory CD4⁺ and CD8⁺ T cells were high-speed sorted using a FACS ARIA (BD Biosciences, San Jose, CA). Purity of sorted T cells was >98%. T cells were then labeled with 0.5 μM CFSE according to the manufacturer’s instructions (Invitrogen) and adoptively transferred into recipient mice by tail vein injection. For reconstitution of B6 γc^−/−RAG^−/− mice with NK cells, B6 RAG^−/− splenocytes were negatively enriched for NK cells (>75% purity) using an NK cell enrichment kit (Stemcell Technologies, Vancouver, Canada) and ~5 × 10⁵ cells were injected i.v. per mouse 4 wk prior to T cell transfer.

At the indicated time points after adoptive T cell transfer or after lymphocyte depletion, mice were sacrificed and peripheral blood, spleen and peripheral LNs (pooled inguinal, axillary, and cervical) were harvested and single-cell suspensions prepared. T cell surface phenotype and proliferation (measures by CFSE dilution) were analyzed after staining for CD4, CD8, CD44, and CD62L surface markers and excluding dead (ethidium monoazide⁺) cells as well as Gr-1⁺, CD11b⁺, and NK1.1⁺ cells. Measurement of IFN-γ and TNF-α production by CD4⁺ and CD8⁺ T cells after ex vivo stimulation was carried out by flow cytometry as previously described (41). Briefly, spleen and lymph node cells were incubated for 16 h with either syngeneic (B6) or allogeneic (BALB/c) stimulator splenocytes in the presence of brefeldin A. Cells were then stained for surface markers, fixed with Cytofix/Cytoperm (BD Biosciences), permeabilized with 0.25% saponin for 1 h at room temperature, and intracellular cytokine staining was performed using PE-conjugated anti–IFN-γ (clone XMG1.2) and FITC-conjugated anti–TNF-α (clone MP6-XT22) Abs (both from eBiosciences). All analyses were performed using a LSR-II flow cytometer (BD Biosciences) and FlowJo software (TreeStar, Ashland, OR). Where indicated, division index of CD8⁺ T cells was calculated using the FlowJo CFSE proliferation analysis tool. Division index is defined as the average number of cell divisions that the responding cells underwent (i.e., ignores peak 0).

Recovered T cell numbers were compared using unpaired Student t test. Differences in allograft survival were analyzed using a log-rank test. Significance was set at p < 0.05. All data are presented as mean ± SEM. GraphPad Prism software (GraphPad Software, San Diego, CA) was used for plotting data and performing statistical analyses.

To test whether NK cells regulate the homeostatic proliferation of T cells, we compared the expansion of adoptively transferred polyclonal T cells between two genetically lymphocyte-deficient hosts: RAG^−/− mice (which lack T and B cells) and common cytokine receptor γc^−/−RAG^−/− mice (which lack T, B, and NK cells). The 1 × 10⁷ CFSE-labeled T cell-enriched splenocytes were transferred from naive B6 mice to either B6 RAG^−/− or γc^−/−RAG^−/− mice and, 7 d later, the CD4⁺ and CD8⁺ T cells that repopulated the host spleen and LNs were quantitated and analyzed by flow cytometry. Although equal numbers of CD4⁺ T cells were retrieved from RAG^−/− and γc^−/−RAG^−/− hosts, ~8-fold more CD8⁺ T cells were retrieved from the latter (Fig. 1). Flow cytometric analysis of CFSE dilution showed that CD8⁺ T cells had proliferated faster in the γc^−/−RAG^−/− group (Fig. 1), indicating that the increased number of retrieved CD8⁺ T cells was due to accelerated homeostatic proliferation. The majority of CD8⁺ T cells that had undergone homeostatic proliferation in γc^−/−RAG^−/− mice exhibited a central memory (CD44^highCD62L^high) phenotype, whereas those that divided in RAG^−/− hosts had a predominantly effector memory (CD44^highCD62L^low) phenotype (Fig. 1). These data suggest that the homeostatic proliferation of CD8⁺ T cells in genetically lymphocyte-deficient hosts is dysregulated in the absence of NK cells.

In addition to the absence of NK cells, γc^−/−RAG^−/− mice lack peripheral LNs and GALTs (42). Therefore, dysregulated homeostatic proliferation of CD8⁺ T cells in these mice could have resulted from either the lack of NK cells or altered sites of T cell proliferation. To directly test the role of NK cells, we used two complementary approaches. First, we transferred polyclonal T cells to either unmanipulated or NK-depleted RAG^−/− hosts and compared CD8⁺ T cell recovery, proliferation, and phenotype between the two groups. As shown in Fig. 2A, twice as many CD8⁺ T cells were recovered from NK-depleted RAG^−/− hosts 6 d after cell transfer. Flow cytometry performed on blood, spleen, and LN cells 3 and 6 d after cell transfer showed accelerated proliferation of CD8⁺ T cells in the NK-depleted group and an increased proportion of the central memory phenotype (CD4^highCD62L^high) among the divided cells (Fig. 2A). In the second approach, T cells were transferred to either unmanipulated or NK-reconstituted γc^−/−RAG^−/− mice. Six days after adoptive transfer, half as many CD8⁺ T cells were retrieved from NK-reconstituted γc^−/−RAG^−/− hosts (Fig. 2B). Flow cytometric analysis showed reduced CD8⁺ T cell homeostatic proliferation in the NK-repleted group as well as a shift toward lower proportion of divided cells with a central memory phenotype (Fig. 2B). These results indicate that NK cells downregulate homeostatic proliferation of CD8⁺ T cells in genetically lymphocyte-deficient (chronically lymphopenic) hosts and inhibit the accumulation of central memory-phenotype T cells.

The transfer of T cells from naive mice to lymphocyte-deficient mice that also lack NK cells resulted in preferential recovery of central memory (CD44^highCD62L^high) CD8⁺ T cells (Figs. 1, 2). The origin of these cells, however, remained unclear as the starting (transferred) T cell population contained both naive (CD44^low) and “natural” memory (CD44^high) T cells. To determine which cell subpopulation divided more and was responsible for increased central memory in the absence of NK cells, we cotransferred sorted, CFSE-labeled naive (Thy1.1⁺CD44^low) and congenic memory (CD45.1⁺CD44^high) CD8⁺ T cells to either NK-sufficient or NK-depleted CD45.2⁺RAG^−/− mice. Six days later, T cell number, division, and phenotype were analyzed by gating on either the Thy1.1⁺ or CD45.1⁺ population. As shown in Fig. 3A, NK depletion increased the number and homeostatic proliferation of the “natural” memory CD8⁺ T cell population to a greater extent than that of the naive population. In the absence of NK cells, homeostatically divided naive CD8⁺ T cells predominantly differentiated into an effector memory (CD62^low) phenotype, whereas the CD8⁺ natural memory population preferentially generated central memory (CD62^high) cells. Using an analogous experimental approach whereby sorted central (CD44^hiCD62L^hi) and effector (CD44^hiCD62^lo) CD8⁺ “natural” memory T cells were cotransferred into either RAG^−/− or NK-depleted RAG^−/− hosts (1 × 10⁵ cells of each population per mouse), we found that only the homeostatic proliferation of the central memory CD8⁺ T cell population is dysregulated in the absence of NK cells (Fig. 3B). NK depletion did not have an effect on the homeostatic proliferation of either central or effector CD4⁺ memory T cells (data not shown). These data therefore indicate that excess central memory CD8⁺ T cells that arise in the absence of NK cells originate from the “natural” central memory T cell pool.

One mechanism by which NK cells regulate adaptive immunity is perforin-dependent lysis of Ag-activated T cells (30–32). To test whether NK cells limit homeostatic T cell proliferation by a perforin-dependent mechanism, we transferred CFSE-labeled T cells to either RAG^−/− or perforin^−/−RAG^−/− mice. Six days later, equal numbers of CD8⁺ T cells were recovered from both groups, but the numbers were significantly less than those recovered from NK-depleted RAG^−/− hosts (Fig. 4A). Analysis of CFSE dilution confirmed that proliferation of CD8⁺ T cells in perforin^−/−RAG^−/− hosts was similar to that in RAG^−/− mice (Fig. 4A), indicating that regulation of homeostatic proliferation by host NK cells is independent of perforin.

We next asked whether NK cells regulate CD8⁺ T cell homeostatic expansion by competing for IL-15 because both cell populations require IL-15 for survival and proliferation (43). IL-15 is trans-presented to CD8⁺ and NK cells in the context of the IL-15Rα–chain on a variety of cells (44), and administration of IL-15 bound to IL-15Rα (IL-15/IL-15Rα complexes) dramatically increases the cytokine’s half-life and potency in vivo (40, 45). We therefore treated γc^−/−RAG^−/− and NK-repleted γc^−/−RAG^−/− mice with a single dose of IL-15/IL-15Rα–Fc complexes 1 d after the adoptive transfer of CFSE-labeled T cells. CD8⁺ T cell number and proliferation were measured 4 and 6 d later to test whether excess IL-15 reverses the downregulatory effect of NK cells. As shown in Fig. 4B, administration of IL-15/IL-15Rα–Fc led to equal recovery of large numbers of CD8⁺ T cells (~1 × 10⁸/spleen) from either group compared with much smaller numbers recovered from γc^−/−RAG^−/− hosts that did not received IL-15/IL-15Rα–Fc complexes (Fig. 2B). Analysis of CFSE dilution confirmed that IL-15/IL-15Rα–Fc administration abrogated the differences in CD8⁺ T cell division profiles between NK-deficient and NK-repleted γc^−/−RAG^−/− hosts (Fig. 4B). The majority of CD8⁺ T cells generated in these mice had a central memory (CD44^hiCD62L^hi) phenotype (plots not shown). These results indicate that NK cells downregulate the homeostatic proliferation of CD8⁺ T cells by competing for IL-15. They are also consistent with our observation that, in the absence of NK cells, homeostatic CD8⁺ T cell proliferation leads to the preferential accumulation of central memory cells. Central memory CD8⁺ T cells express the highest level of the IL-2R/IL-15Rβ–chain (CD122) among all CD8⁺ T cell subsets (Fig. 4C) and thus, are expected to preferentially expand in an environment where IL-15 is present and competing NK cells are absent.

To test whether NK cells regulate homeostatic T cell proliferation in a more clinically relevant setting, we studied T cell recovery in wt mice rendered acutely lymphopenic with anti-CD4 and anti-CD8 mAbs. B6 wt mice received two doses of anti-CD4 and anti-CD8 Abs on days 0 and 5, a regimen previously shown to induce 80–90% peripheral T cell depletion (20), and were either depleted of NK cells or not. Splenocytes were harvested on day 20 and analyzed by flow cytometry. As shown in Fig. 5, T and NK cell-depleted mice exhibited significant CD8⁺ T cell recovery, whereas T cell-depleted but NK cell-sufficient mice were still devoid of CD8⁺ T cells by day 20. NK cell depletion in the group that received anti-NK1.1 was confirmed by flow cytometry (Fig. 5). Unlike the CD8⁺ population, no significant differences in the recovery of CD4⁺ T cells were observed between the NK cell-depleted and NK cell-sufficient groups (data not shown).

Lymphodepletion with Abs that also deplete NK cells is commonly used in the clinic to prevent rejection of solid organ allografts. Homeostatic proliferation of residual T cells and the subsequent expansion of the memory T cell pool, however, often leads to undesirable late rejection episodes and, in experimental animals, interferes with tolerance induction (19–22). Having observed increased homeostatic proliferation of CD8⁺ T cells in the absence of NK cells, we asked whether NK cell deficiency would have detrimental effects on allograft survival in lymphopenic hosts. We first compared cardiac allograft survival in B6 RAG^−/− and γc^−/−RAG^−/− mice after the adoptive transfer of B6 wt T cells. To mimic the clinical situation whereby T cell recovery is observed after the transplanted organ had sufficient time to heal (46), T cells were transferred 50 d after transplantation. As shown in Fig. 6A, the transfer of 1 × 10⁷ T cells, which undergo greater homeostatic proliferation in γc^−/−RAG^−/− than RAG^−/− mice (Figs. 1, 3), precipitated cardiac allograft rejection at a significantly faster rate in the former than the latter (median survival time [MST] = 6.5 and 11 d, respectively; p < 0.05). The transfer of a much larger number of T cells (1 × 10⁸/mouse), which is sufficient to precipitate rejection without prior homeostatic proliferation, led to equivalent rejection rates in both groups (MST = 6.5 and 7.5 d, respectively; p > 0.05). These data provide evidence that faster allograft rejection in NK-deficient hosts is most likely due to dysregulated homeostatic proliferation of T cells. To further test this hypothesis, we analyzed skin allograft survival in wt mice treated with T cell-depleting Abs with or without concomitant NK cell depletion. Fully allogeneic skin grafts were rejected at a significantly faster rate by T cell- and NK cell-depleted than T cell-depleted but NK cell-sufficient recipients (MST = 16 and 23.5 d, respectively; p < 0.05) (Fig. 6B). Furthermore, an immunosuppressive regimen (donor specific transfusion and anti-CD154) that further prolonged skin allograft survival in T cell-depleted but NK-sufficient wt mice (MST = 50 d) failed to do so in mice that were both T cell- and NK cell-depleted (MST = 17 d; p < 0.01) (Fig. 6C).

NK cells have been shown to suppress donor-specific T cell responses in lymphocyte-deficient mice by mechanisms other than regulation of homeostatic CD8⁺ T cell proliferation (36). To further explore the mechanisms responsible for accelerated allograft rejection in NK and T cell-depleted mice, we studied the donor-specific responses of CD4⁺ and CD8⁺ T cells in an additional cohort of wt B6 mice that were either T cell-depleted or T cell- and NK cell-depleted, transplanted with BALB/c skin allografts, and treated with DST plus anti-CD154. Twenty-four days after skin transplantation (33 d after T cell depletion and before any allograft rejection had occurred), the mice were sacrificed and their spleen and LN cells were analyzed for cytokine production in response to donor-specific allostimulation. As shown in Fig. 6D, equal numbers of CD4⁺ T cells were recovered from NK-depleted and NK-replete mice, but almost twice as many CD8⁺ T cells were recovered from NK-depleted than the NK-replete group. Recovered T cells were predominantly of the memory (CD44^hi) phenotype (>70%) and the majority of those were central memory (CD62L^hi) (plots not shown). Analysis of IFN-γ and TNF-α production after ex vivo stimulation of host LN or spleen cells with donor (BALB/c) splenocytes showed significantly enhanced cytokine production by CD8⁺ T cells recovered from the NK-depleted group (Fig. 6D). In contrast, no differences in cytokine production by CD4⁺ T cells were observed between NK-depleted and NK-replete mice. These findings indicate that NK depletion of lymphocyte-deficient animals leads to exaggerated expansion of memory CD8⁺ T cells with enhanced alloresponsiveness to donor Ags. In contrast, NK depletion does not influence CD4⁺ T cell homeostatic proliferation nor alter their alloresponsiveness. These findings support the conclusion that accelerated allograft rejection in lymphopenic animals that also lack NK cells is likely the result of rapid accumulation of homeostatically generated memory CD8⁺ T cells.

We investigated in this study whether NK cells regulate lymphopenia-driven, homeostatic proliferation of T cells. We found that NK depletion augments the homeostatic proliferation of CD8⁺ T cells in both genetically lymphocyte-deficient RAG^−/− and acutely lymphodepleted wt mice. Enhanced homeostatic CD8⁺ T cell proliferation in the absence of NK cells had important biological consequences as it expanded the memory CD8⁺ T cell pool, preferentially increasing the proportion of central memory phenotype cells, and accelerated the rejection of heart and skin allografts. We also demonstrated that regulation of homeostatic T cell proliferation by NK cells is not dependent on perforin but is due to competition for IL-15, the principal cytokine required for maintaining the NK and CD8⁺ memory T cell pools.

Our finding that NK cells predominantly regulate the homeostatic proliferation of the “natural” central memory rather than the naive CD8⁺ T cell pool is consistent with the rules that govern the maintenance of these lymphocyte populations (3). Both NK and CD8⁺ memory T cells, particularly central memory T cells, express high levels of the IL-15Rβ–chain (CD122) and rely on IL-15 for survival and proliferation (44, 47). Naive T cells and CD4⁺ memory T cells, in contrast, have very low expression of CD122 and instead are heavily dependent on IL-7 for their maintenance (3). Moreover, peripheral NK cell homeostasis is independent of IL-7 (47). Therefore, it is not surprising that in our experiments central memory CD8⁺ T cells exhibited the most homeostatic proliferation in the absence of IL-15 consumption by NK cells. Even in γc^−/−RAG^−/− mice, where IL-15 levels are elevated (28), cytokine amounts are still limiting as they are expected to control the overall sizes of the T cell and NK pools. Thus, competition between cell populations that use the same cytokine for survival and/or homeostatic division is inevitable (48, 49). Our finding that administering excess IL-15 to γc^−/−RAG^−/− mice abrogates differences in CD8⁺ T cell homeostatic proliferation between the NK-deficient and NK-repleted groups supports this hypothesis. At the same time, our findings do not rule out homeostatic proliferation of naive T cells is an important source of memory T cells in lymphopenic mice, particularly if NK cells are present (50).

Recent work by Ramsey et al. (28) showed that T cells transferred to CD132^−/− (γc^−/−) mice divide very rapidly and lead to significantly higher recovery of central memory CD8⁺ T cells than that observed in RAG^−/− hosts. The authors also demonstrated that increased accumulation of central memory T cells in CD132^−/− mice was primarily due to elevated basal levels of IL-15 that in turn drove the homeostatic proliferation of naive CD8⁺ T cells. Our results confirm that IL-15 is crucial for generating central memory in lymphopenic environments but also extend the observations made by Ramsey et al. (28) in four important ways. First, we demonstrated that in genetically lymphocyte-deficient mice, NK cells downregulate homeostatic CD8⁺ T cell proliferation by competing for IL-15. Second, we provided evidence that excess central memory CD8⁺ T cells that arise in the absence of competing NK cells derive primarily from the homeostatic proliferation of the IL-15-responsive “natural” central memory T cell population, a population that expresses significantly higher levels of CD122 than naive T cells or effector memory T cells. Third, we established that NK regulation of homeostatic CD8⁺ T cell proliferation also applies to acutely lymphodepleted wt mice. Fourth, we showed that increased homeostatic CD8⁺ T cell proliferation in the absence of NK cells has important biological implications, namely, increased IFN-γ and TNF-α production and the accelerated rejection of heart and skin allografts. Therefore, our results complement those of Ramsey et al. (28), whereas further elucidating the subtleties of immune regulation in lymphopenic environments by underscoring the role of NK cells in this process. Whether NK cells regulate the size and phenotype of the memory T cell pool in lymphocyte-replete animals remains to be determined.

Another potential mechanism for the observed regulation of homeostatic proliferation by NK cells is NK-mediated lysis of activated CD8⁺ T cells (30, 31). This regulatory mechanism has been described in the setting of foreign Ag-induced T cell proliferation but does not appear to have a significant role in lymphopenia-driven homeostatic proliferation. We observed in the current study that the homeostatic proliferation of CD8⁺ T cells is equivalent in RAG^−/− and perforin^−/−RAG^−/− mice. This is in contrast to CD8⁺ T cell responses to foreign histocompatibility Ags in lymphocyte-replete hosts where the absence of perforin leads to exaggerated T cell proliferation and accelerated allograft rejection (31, 51). Alternatively, NK cells could regulate homeostatic T cell proliferation indirectly through interactions with dendritic cells. Dendritic cells trans-present IL-15 in the context of IL-15Rα to NK and CD8⁺ T cells and contribute to the homeostatic proliferation of CD8⁺ T cells by presenting self- or foreign pMHC (44, 52, 53). Therefore, it is conceivable that by modulating these functions of dendritic cells NK cells could influence homeostatic T cell proliferation in lymphopenic mice.

Enhanced recovery of central memory CD8⁺ T cells in lymphodepleted animals that have also been rendered NK deficient has important clinical implications. In some cases, skewing of homeostatically dividing T cells to a central memory phenotype may be beneficial. For example, central memory CD8⁺ T lymphocytes confer superior anti-tumor immunity than their effector memory counterparts (54, 55), and removal of NK cells enhances the efficacy of tumor-specific CD8⁺ T cells transferred to lymphopenic mice (56). In contrast, increased generation of memory T cells in lymphopenic hosts can cause autoimmunity (16, 17), precipitate allograft rejection, and prevent tolerance induction (18–22). In these situations, concomitant NK and T cell depletion is detrimental to allograft survival as shown in our experiments. An important point to add is that agents that deplete NK cells (for example, PK136 Abs in mice) also deplete NKT cells. Therefore, immune dysregulation observed after administering such agents could also be due to depletion of NKT cells, which are known to have immunoregulatory functions (57).

NK cells appear to play a dual role in solid organ transplantation (29). They contribute to the pathogenesis of acute and chronic allograft rejection (58–60), or they facilitate tolerance induction (31, 36). Beilke et al. (31) showed that costimulation blockade-induced prolongation of islet allograft survival is dependent on the presence of perforin-competent NK cells. Yu et al. (36) compared alloantigen-driven T cell proliferation and skin allograft survival between RAG^−/− and γc^−/−RAG^−/− mice and found increased T cell proliferation and accelerated graft rejection in the latter. This observation was attributed to a necessary regulatory role for NK cells in killing donor dendritic cells that migrate to secondary lymphoid organs. It is possible that the same mechanism is operational in our model, but our data provide evidence that homeostatic competition between NK and CD8⁺ T cells in lymphopenic hosts directly influences allograft survival.

In summary, we have shown that NK cells control the homeostatic proliferation of CD8⁺ T cells in lymphopenic environments by competing for IL-15, a cytokine required for both NK and CD8⁺ T cell maintenance. Absence of NK cells resulted in accelerated allograft rejection under these conditions. These findings point to an immunoregulatory role for NK cells that could limit the generation of undesirable autoreactive or alloreactive memory T cells during recovery from lymphopenia. Despite their putative role in promoting rejection responses, NK cells should perhaps be spared in bone marrow and solid organ transplant recipients receiving induction therapy with lymphoablative agents.

Disclosures The authors have no financial conflicts of interest.