Evidence of a Novel IL-2/15Rβ-Targeted Cytokine Involved in Homeostatic Proliferation of Memory CD8+ T Cells1

The homeostasis of memory CD8+ T cells is regulated by cytokines. IL-15 is shown to promote the proliferation of memory CD8+ T cells, while IL-2 suppresses their division in vivo. This inhibitory effect of IL-2 appears to occur indirectly, through other cell populations including CD25+CD4+ T cells; however, the details of this mechanism remain unclear. In this study, we show that 1) both Ag-experienced and memory phenotype CD8+ T cells divided after the depletion of IL-2 in vivo; 2) this division occurred normally and CD44highIL-2/15Rβhigh CD8+ T cells generated after IL-2 depletion in IL-15 knockout (KO) and in IL-7-depleted IL-15 KO mice; 3) surprisingly, the blockade of IL-2/15Rβ signaling in IL-2-depleted IL-15 KO mice completely abolished the division of memory CD8+ T cells, although the only cytokines known to act through IL-2/15Rβ are IL-2 and IL-15; and 4) the expression of IL-2/15Rβ molecules on memory CD8+ T cells was required for their division induced by IL-2 depletion. These results demonstrate that the depletion of IL-2 in vivo induced memory CD8+ T cell division by an IL-15-independent but by an IL-2/15Rβ-dependent mechanism, suggesting the existence of a novel IL-2/15Rβ-utilizing cytokine that acts directly on memory CD8+ T cells to promote cell division.

L ong-lived memory CD8 ϩ T cells expand more vigorously and produce effector molecules more rapidly than do naive cells upon re-encountering a pathogen and contribute to rapid clearance of the pathogen. It is thought that a constant pool of memory CD8 ϩ T cells is maintained through a balance of slow basal proliferation and death (1,2). This homeostasis of memory CD8 ϩ T cells was shown to be established in the absence of MHC class I molecules (3). Subsequent studies revealed that cytokines, the IL-2 family in particular, are the principal mediators for the homeostasis of memory CD8 ϩ T cells (4 -6). At present, IL-2family cytokines include IL-2, -4, -7, -9, -15, and -21 (7,8). They have respective ligand-binding receptors (␣-chain), and share the common ␥-chain (CD132, ␥ c ) 4 for signal transduction. In addition, IL-2 and IL-15 require the IL-2/15R␤ chain (CD122) for their signaling. IL-2R␣ (CD25), IL-2/15R␤, and ␥ c form a high-affinity receptor complex for IL-2 with K a ϭ 10 Ϫ11 M (7). Similarly, the IL-15R complex consists of IL-15R␣, IL-2/15R␤, and ␥ c . But in contrast to IL-2R␣, IL-15R␣ alone is capable of binding IL-15 with high affinity (K a ϭ 10 Ϫ11 M) in the absence of other receptor subunits (7).
We recently reported that endogenous IL-2 suppresses the division of memory CD8 ϩ T cells in the homeostatic condition. A depletion of IL-2 leads to a significant increase in the basal homeostatic division of CD44 high memory phenotype CD8 ϩ T cells, but not of CD44 low naive CD8 ϩ T cells (9,10). This inhibitory effect of IL-2 on memory phenotype CD8 ϩ T cells in vivo appears to be indirect, given that an in vitro study revealed IL-2 alone stimulates the proliferation of memory CD8 ϩ T cells and that it does not inhibit the IL-15-induced proliferation of nor induce apoptosis in these cells (10). A possible mediator cell for the in vivo IL-2 action is CD25 ϩ CD4 ϩ regulatory T cells (10). Thus, the endogenous level of IL-2 negatively controls the homeostasis of memory CD8 ϩ T cells in vivo. However, the detailed mechanism of IL-2 action in vivo remains obscure.
In contrast, the positive regulation of the cytokine-mediated homeostasis of memory CD8 ϩ T cells has been revealed to be largely dependent on IL-15 (11,12). For instance, IL-15 is a potent and selective growth factor for memory CD8 ϩ T cells both in vitro and in vivo (13)(14)(15)(16), and numbers of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells are significantly decreased and their basal homeostatic proliferation is completely impaired in IL-15 knockout (KO) and IL-15R␣ KO mice (17)(18)(19)(20)(21)(22)(23). Moreover, microbial components such as LPS and poly (I:C) indirectly induce the division of memory CD8 ϩ T cells in vivo, in a process called "bystander proliferation," through IL-15 (24,25). Interestingly, the expression of IL-15R␣ on memory CD8 ϩ T cells is dispensable for their bystander proliferation (24) and basal homeostatic proliferation (22,23). A study using bone marrow-chimeric mice showed that IL-15R␣ expression on radiosensitive hemopoietic cells is important for the proliferation of memory CD8 ϩ T cells (22)(23)(24). Transpresentation of IL-15 by IL-15R␣-bearing non-T cells to CD44 high memory CD8 ϩ T cells was observed in vitro (22,26). It was also shown that memory CD8 ϩ T cell proliferation after IL-15 injection in vivo requires the expression of IL-15R␣ molecules on non-memory CD8 ϩ T cells but requires that of IL-2/15R␤ molecules by the memory CD8 ϩ T cells (23), suggesting that the transpresentation of IL-15 to memory CD8 ϩ T cells could occur in vivo. Another possibility that could account for the dispensability of IL-15R␣ expression on CD8 ϩ T cells is that the proliferation of memory CD8 ϩ T cells may be mediated by an undefined growth factor released from IL-15R␣-bearing non-T cells, as proposed by Ma and colleagues (27).
We herein report that the MHC class I-independent division of memory CD8 ϩ T cells induced by IL-2 depletion occurred normally even in the absence of IL-15. In addition, we provide evidence that this process was mediated by a putative novel IL-2/ 15R␤-utilizing cytokine, since treatment with anti-IL-2/15R␤ Ab completely abrogated the anti-IL-2-mediated division of memory CD8 ϩ T cells in IL-15 KO mice. Furthermore, we show that the expression of IL-2/15R␤ molecules on memory CD8 ϩ T cells was essential for action of the putative cytokine. These results demonstrate the existence of a previously undefined mechanism for the homeostasis of memory CD8 ϩ T cells in vivo.

Mice
Retired C57BL/6 (B6) mice ϳ5-7 mo of age were obtained from CLEA Japan (Osaka, Japan) or SLC (Shizouka, Japan). IL-15 KO, K b D b double KO, and B6.SJL (CD45.1)-congenic mice were obtained from Taconic Farms (Germantown, NY). IL-2 KO and B6.PL (Thy-1.1)-congenic mice were purchased from The Jackson Laboratory (Bar Harbor, ME). OT-1 TCR-transgenic and IL-7R␣ KO mice were kindly provided by Drs. W. R. Heath (Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia) and K. Ikuta (Kyoto University, Kyoto, Japan), respectively. All strains used in this study were a B6 background. All of these mice except IL-2/15R␤ KO mice were kept at the Institute of Experimental Animal Sciences at Osaka University Medical School (Osaka, Japan). IL-2/15R␤ KO mice were kept at the Institute of Experimental Animal Sciences at Nagoya University Medical School (Nagoya, Japan).

Adoptive transfer of donor T cells
Spleen and lymph nodes were harvested and homogenated by cell strainers (100 m; BD Biosciences). The T cell-enriched fraction was obtained by passing the mixture of splenocytes and lymph node cells through nylon wool columns (Wako Biochemicals, Osaka, Japan and Polysciences, Warrington, PA). The T cell fraction was then sorted based on the CD4, CD8␣, CD44, and IL-2/15R␤ levels using a cell sorter (MoFlo; DakoCytomation, Carpinteria, CA). The purity of the sorted populations was typically 97-99%. The purified populations were labeled with 5 M CFSE (Invitrogen Life Technologies, Grand Island, NY) at 37°C. The CFSE-labeled donor cells (0.5-2 ϫ 10 6 cells/mouse) were transferred i.v. into nonirradiated mice. For Fig. 5B, the T cell-enriched fraction obtained from wild-type (WT) or IL-2 KO mice, which contained 3-5% of CD44 high IL-2/15R␤ high CD8 T cells, were used for donor cells.

Generation of Ag-experienced memory CD8 ϩ T cells
OT-1 CD8 ϩ T cells (CD45.2) were transferred into B6.SJL (CD45.1) mice, followed by immunization i.p. and i.v. with extensively washed LPS-stimulated, OVA-pulsed bone marrow-derived dendritic cells. The immunized mice were then rested for at least 8 wk. Ag-specific OT-1 memory CD8 ϩ T cells were sorted based on the expression of CD8 ϩ CD45.2 ϩ V␣2TCR ϩ CD44 high . These cells were then labeled with CFSE and transferred into recipient mice.

Ab or poly(I:C) treatment
The hybridoma for anti-IL-2-neutralizing mAb (S4B6) was grown in hybridoma serum-free medium (Invitrogen Life Technologies). Anti-IL-2 mAb was then purified from the culture supernatant using a protein G column. The purified mAb was dialyzed against PBS and then sterilized by 0.2-m pore filtration. Mice that received donor cells were given anti-IL-2 mAb i.p. five to seven times at 1 mg/mouse per day. Rat IgG (Sigma-Aldrich, St. Louis, MO) was used as a control Ab (9,10). In some experiments, anti-IL-2R␣ (PC61), anti-IL-2/15R␤ (TM-␤1), anti-IL-7 (M25), anti-IL-7R␣ (A7R34) mAb, or anti-␥c mAbs (4G3 and 3E12) was coinjected i.p. with anti-IL-2 mAb. These hybridomas except 4G3 and 3E12 (BD Pharmingen, San Diego, CA) were grown in RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% heat-inactivated FCS (Equitech-Bio, Kerrville, TX), and each mAb was purified as described above. The splenocytes and/or lymph node cells from treated animals were passed through nylon wool columns and the resulting T cell fraction was subjected to flow cytometry. The F(abЈ) 2 of anti-IL-2/15R␤ mAb (TM-␤1) was prepared by pepsin digestion of the Ab at pH 4.0, followed by purification of the F(abЈ) 2 fraction using a Mono S cation exchange column or a gel filtration column (Takara Bio, Otsu, Japan). Contaminating whole IgG molecules in the F(abЈ) 2 preparation were below the detection limit by ELISA using an anti-rat Fc fragment-specific Ab (Valeant Pharmaceuticals, Costa Mesa, CA). Western blot analysis under nonreducing conditions using anti-rat IgG (H ϩ L)-HRP (Zymed Laboratories, San Francisco, CA) was also performed. The intensity of the band corresponding to whole IgG (150 kDa) in the F(abЈ) 2 preparation was equivalent to a 1/1000 dilution of the original anti-IL-2/15R␤ mAb. Poly(I:C) (Amersham Biosciences, Piscataway, NJ) was injected at 50 -100 g/mouse i.p. after donor T cell transfer.

In vivo BrdU labeling
Mice were given 0.8 mg/ml BrdU (Sigma-Aldrich) in their drinking water or it was administered i.p. at 1 mg/mouse during the anti-IL-2 treatment. BrdU staining for flow cytometry was performed using the BrdU Flow kit (BD Biosciences).

Results
In vivo depletion of IL-2 selectively induced the homeostatic division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells Our previous reports revealed that IL-2 depletion induced the division of CD44 high memory phenotype CD8 ϩ T cells, but not that of CD44 low naive CD8 ϩ T cells (9,10). To extend this finding, we first examined the effect of IL-2 depletion on the CD4 ϩ as well as CD8 ϩ T cell compartments. Memory CD8 ϩ T cells can be distinguished from naive CD8 ϩ T cells by the expression levels of several surface markers, such as CD44, IL-2/15R␤, and Ly6C (28). Similarly, the expression level of CD44 defines the naive and memory phenotype CD4 ϩ T cell subsets. Naive CD4 ϩ (CD44 low ), memory phenotype CD4 ϩ (CD44 high ), naive CD8 ϩ (CD44 low IL-2/15R␤ low ), or memory phenotype CD8 ϩ (CD44 high IL-2/ 15R␤ high ) T cells were sorted from the splenocytes and lymph node cells of retired mice, in which a large number of memory phenotype T cells accumulates (10,29). These T cell populations bearing Thy-1.2 molecules were labeled with CFSE and transferred into nonirradiated Thy-1.1-congenic mice. The recipient mice were then given anti-IL-2-neutralizing mAb (S4B6). The Ab treatment in this experiment was designed to be a longer injection period (11 days) compared with our standard protocol (6 -7 days) to perform a detailed examination whether CD4 ϩ T cells have an ability to divide after anti-IL-2 treatment. Eleven days after the transfer, the division of donor T cells was detected by the dilution of CFSE. As shown in Fig. 1A, IL-2 depletion did not affect the division of CD44 low naive CD4 or CD44 low naive CD8 T cells. In contrast, the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells was significantly induced, as reported previously (9,10). Consistent with this result, DNA synthesis, as judged by BrdU incorporation, was selectively increased by the anti-IL-2 treatment in the CD44 high fraction of CD8 ϩ T cells compared with control IgG (rat IgG) treatment (Fig. 1B). The cell division and BrdU incorporation of CD44 high memory phenotype CD4 ϩ T cells were not affected by the anti-IL-2 treatment protocols tested here (Fig. 1, A and B). Furthermore, the division of CD44 high IL-2/ 15R␤ high memory phenotype CD8 ϩ T cells was also induced in IL-2 KO mice even without Ab treatment (data not shown).
We next examined changes in the expression levels of cell surface markers upon cell division induced by anti-IL-2 mAb and compared them with those induced by poly(I:C), a well-known inducer of bystander proliferation of memory CD8 ϩ T cells (30). As shown in Fig. 1C, the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells induced by anti-IL-2 treatment or by poly(I:C) injections did not associate with the up-regulation of an activation marker, IL-2R␣ (CD25), as reported previously (9,30). In addition, no change was observed for the expression levels of TCR␤, CD8␣, CD44, CD45, CD62L, CD69, Thy-1, IL-2/15R␤, and IL-7R␣ by anti-IL-2 treatment ( Fig. 1C and data not shown). We also examined the effect of anti-IL-2R␣ treatment, which led to a blockade of IL-2 signaling and depletion of IL-2R␣ high cells like CD25 ϩ CD4 ϩ regulatory T cells on the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells. The host mice that had received CFSE-labeled memory phenotype CD8 ϩ T cells were treated with rat IgG, anti-IL-2 mAb, anti-IL-2R␣ mAb (PC61), or a combination of anti-IL-2 and anti-IL-2R␣ mAbs. In contrast to IL-2 depletion by anti-IL-2 mAb, anti-IL-2R␣ treatment rarely promoted the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells compared with rat IgG treatment (data not shown), suggesting neutralization of IL-2 molecules themselves may be critical for induction of memory CD8 ϩ T cell division by this short-term experiment. As reported previously (9), however, treatment with the combination of anti-IL-2 and anti-IL-2R␣ mAbs resulted in more enhanced division of these cells than anti-IL-2 treatment alone.
Taken together, these results suggest that the endogenous level of IL-2 selectively suppresses the homeostatic division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in vivo.
IL-2 depletion induced the division of Ag-experienced memory as well as CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells CD44 high IL-2/15R␤ high memory-phenotype CD8 ϩ T cells have many properties of memory cells (5,31), but it is uncertain whether they have experienced Ag stimulation or not. To examine whether IL-2 also controls Ag-experienced memory CD8 ϩ T cells, OVA-specific TCR-transgenic (OT-1) memory CD8 ϩ T cells were generated (see Materials and Methods), CFSE-labeled, and transferred into Ag-inexperienced recipient mice. Similar to the CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells, the Agexperienced OT-1 memory CD8 ϩ T cells underwent several cell divisions after the anti-IL-2 treatment (Fig. 2). Given that both CD44 high IL-2/15R␤ high memory phenotype and Ag-experienced memory CD8 ϩ T cells were found to be regulated by IL-2 in a FIGURE 1. IL-2 depletion selectively induces the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in vivo. A, CD44 low or CD44 high CD4 ϩ T cells and CD44 low IL-2/15R␤ low or CD44 high IL-2/ 15R␤ high CD8 ϩ T cells from mice expressing Thy-1.2 molecules were sorted and labeled with CFSE. The CFSE-labeled donor cells were transferred into nonirradiated syngeneic Thy-1.1 mice on day 0. Anti-IL-2 mAb (S4B6, 1 mg) or control Ab (rat IgG, 1 mg) was given i.p. on days 1, 3, 4, 5, 6, 7, and 8. CFSE levels were examined on day 11. Histograms shown are gated on the Thy-1.2-expressing donor population. B, Mice were given BrdU in their drinking water for 8 days. The anti-IL-2 treatment was performed for the first 6 days. The BrdU-positive population in the T cell compartments was examined. C, CFSE-labeled CD44 high IL-2/15R␤ high CD8 ϩ T cells were transferred into nonirradiated hosts on day 0. The anti-IL-2 treatment was conducted on days 1-5. Poly(I:C) was injected i.p. at 100 g on days 1, 3, and 5. The CFSE levels and IL-2R␣, CD8␣, and TCR␤ levels were examined on day 7. The plot of CFSE vs IL-2R␣ was gated on lymphocyte (low side scatter but not dead cells) and TCR␤ ϩ population and those of the others were gated on the lymphocyte population. The CFSE-negative population is derived from the host mouse. The results are representative of more than two to three independent experiments. similar way, we used CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells as donor cells in the following experiments.

IL-2 depletion induced the division of memory CD8 ϩ T cells in the absence of IL-15
It has been demonstrated that the maintenance of memory CD8 ϩ T cells can be established independent of MHC class I molecules (3). To examine whether the division of memory CD8 ϩ T cells induced by IL-2 depletion requires MHC class I molecules, CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells were transferred into MHC class I KO (K b D b double KO) mice. Consistent with our previous report, in which memory phenotype CD8 ϩ T cells lacking ␤ 2 -microglobulin molecules could divide upon IL-2 depletion in ␤ 2 -microglobulin KO mice (9), donor CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells exhibited multiple divisions after IL-2 depletion in the recipients lacking MHC class I molecules (Fig. 3A). We also previously reported evidence suggesting that the inhibitory effect of IL-2 in vivo does not appear to be due to a direct action on memory phenotype CD8 ϩ T cells (10). These observations led us to hypothesize that endogenous IL-2 suppresses the activity of a soluble factor, such as a cytokine, that triggers memory phenotype CD8 ϩ T cell-specific proliferation in vivo. The most likely candidate for this agent was IL-15, because IL-15 is a potent growth factor for memory CD8 ϩ T cells and currently known mechanisms for the Ag-independent division of these cells in vivo are largely dependent on IL-15 (13-23, 27, 31). To test this hypothesis, CFSE-labeled CD44 high IL-2/ 15R␤ high memory phenotype CD8 ϩ T cells isolated from WT mice were transferred into IL-15 KO mice, and the host mice were then treated with anti-IL-2 mAb. Surprisingly, the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells was induced normally in the IL-15 KO hosts (Fig. 3B), even though we could not see any division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in IL-15 KO mice after poly(I:C) treatment as described previously (Fig. 3B) (25).
IL-15 is known to be produced by a variety of cell types and T cells are thought not to be a major IL-15 producer (32, 33). However, to rule out the possibility that IL-15 derived from WT donor T cells mediated the proliferation after IL-2 depletion shown in Fig. 3B, WT or IL-15 KO mice were given BrdU during the anti-IL-2 treatment and endogenous BrdU-positive proliferating cells were observed. IL-15 KO mice contained only a marginal population of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells, as reported previously (18,25), and they were mostly BrdU negative (Fig. 3C, IL-15 KO treated with rat IgG). IL-2 depletion in IL-15 KO mice increased the BrdU-positive CD8 ϩ T cells bearing the CD44 high phenotype in the spleen compared with IL-15 KO mice treated with the control Ab (Fig. 3C, top). This increased population of CD44 high CD8 ϩ T cells in the IL-2-depleted IL-15 KO mice expressed a high level of IL-2/15R␤ molecules (Fig. 3C,  bottom). The increase in the CD44 high IL-2/15R␤ high CD8 ϩ T cell population in the IL-2-depleted IL-15 KO mice was also observed in nonlymphoid organs such as the liver and lungs (data not shown). It is intriguing that IL-2 depletion could restore the phenotype of IL-15 KO mice, which usually show a marked reduction in CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in both lymphoid and nonlymphoid organs.
We next examined the involvement of IL-7 in this process, because the overexpression of IL-7 in mice overcomes the IL-15 dependency of CD44 high memory phenotype CD8 ϩ T cells (34), and the irradiation-induced acute homeostatic proliferation of memory CD8 ϩ T cells is halted when both IL-7 and IL-15 are suppressed (20,35). Therefore, if IL-2 depletion up-regulates the production of IL-7, then CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells could proliferate even in the absence of IL-15. To examine this possibility, CFSE-labeled CD44 high IL-2/15R␤ high memory CD8 ϩ T cells were transferred into IL-15 KO mice, and the mice were then treated with the combination of anti-IL-7 and anti-IL-7R␣ mAbs to remove both IL-7 and IL-15 at the same time. They were then treated with anti-IL-2 mAb. As shown in Fig.  3D, inhibition of IL-7 and its signaling in IL-15 KO mice had no effect on the division of donor cells. B lymphopoiesis in the bone marrow was impaired in the anti-IL-7-plus anti-IL-7R␣ mAbtreated mice, verifying the efficacy of these Abs (data not shown). In addition, the CD44 high memory phenotype CD8 ϩ T cells from IL-7R␣ KO mice divided in IL-15 KO mice after the anti-IL-2 treatment (data not shown). These results suggested that IL-2 negatively controlled the homeostasis of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in vivo in an IL-15-independent fashion by a mechanism that is distinct from MHC class I-or IL-7-dependent processes.

A putative novel IL-2/15R␤-utilizing cytokine mediates the division of memory CD8 ϩ T cells
We demonstrated previously that anti-IL-2/15R␤ treatment, which is known to block both IL-2 and IL-15 receptor-mediated signaling, diminishes the basal homeostatic division of memory phenotype CD8 ϩ T cells in vivo (9). However, in the present study we found that CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells divided normally under the IL-2-depleted and IL-15-null condition (Fig. 3, B-D). Only IL-2 and IL-15 have been identified as IL-2/15R␤-utilizing cytokines. However, if there were a third IL-2/15R␤-utilizing cytokine or factor that induces the division of memory CD8 ϩ T cells, it would explain the discrepancy noted above. Namely, under the IL-2-depleted, IL-15-null condition created by the anti-IL-2 treatment in IL-15 KO mice, the hypothesized third IL-2/15R␤-utilizing cytokine could induce the division of memory CD8 ϩ T cells. To examine this hypothesis, IL-2-depleted IL-15 KO mice received CFSE-labeled CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells and were simultaneously treated with anti-IL-2/15R␤ mAb. As expected, this treatment resulted in complete inhibition of the anti-IL-2-induced division of FIGURE 2. IL-2 depletion induces Ag-experienced memory CD8 ϩ T cells in vivo. OVA-specific OT-1 memory CD8 ϩ T cells were generated in mice that received OT-1 CD8 ϩ T cells and were subsequently immunized with OVA-loaded bone marrow-derived dendritic cells. CD44 high OT-1 memory or CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells were labeled with CFSE and transferred into nonirradiated mice on day 0. The Ab treatment was performed on days 2-6. CFSE levels were examined on day 7. The CFSE-negative population is derived from the host mouse. The results are representative of three independent experiments. CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in IL-15 KO mice (Fig. 4A). In contrast to anti-IL-2/15R␤, treatment with anti-IL-2R␣ mAb had no such inhibitory effect, but rather intensified the efficacy of anti-IL-2 mAb alone in IL-15 KO mice (Fig.  4, A and C), indicating that the effect of the anti-IL-2/15R␤ treatment was not simply due to a blockade of signaling from the remaining IL-2 molecules in the IL-2-depleted IL-15 KO mice. The first peak of CFSE signals from the donor IL-2/15R␤ high CD8 ϩ T cells, which might be depleted by an anti-IL-2/15R␤ complement reaction, was clearly detectable in the anti-IL-2/ 15R␤-treated mice (Fig. 4A), and the F(abЈ) 2 of anti-IL-2/15R␤ mAb was also effective (Fig. 4B), excluding the possibility of the Fc-mediated depletion of donor cells. The F(abЈ) 2 of anti-IL-2/ 15R␤ mAb was less efficacious than the intact one in terms of blocking anti-IL-2-induced division of memory CD8 T cells (Fig.  4B). It was reported that a half-life of F(abЈ) 2 in vivo is generally shorter than that of whole IgG (36,37). We think this could be the case. In addition, when compared with the intact anti-IL-2/15R␤ mAb, at least five times more F(abЈ) 2 was needed to block the binding of a biotinylated anti-IL-2/15R␤ mAb on CD44 high IL-2/ 15R␤ high CD8 ϩ T cells (data not shown), suggesting that removal of the Fc portion of Abs might also affect their avidity to Ags. As shown in Fig. 4C, the anti-IL-2/15R␤ treatment also negated the emergence of the CD44 high IL-2/15R␤ high CD8 ϩ T cell population induced by IL-2 depletion in IL-15 KO mice.
TM-␤1 has been used as a blocking mAb for IL-2/15R␤-mediated signaling in vivo (9,25). But if this anti-IL-2/15R␤ mAb induced an unexpected inhibitory signal in memory CD8 ϩ T cells, their division must be stopped. We also suggested that the endogenous level of IL-2 may induce apoptosis of memory phenotype CD8 ϩ T cells (9). We therefore examined the effect of the mAb on cell survival. With a 4-day culture in vitro, only Ͻ5% of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells were alive without cytokines, as judged by the 7AAD-negative population (Fig. 5). IL-7, which belongs to the IL-2-family cytokine but does not utilize IL-2/15R␤ for signaling (7), rescued ϳ50% of these cells from apoptosis. The addition of control IgG (rat IgG) to the culture had no effect (data not shown); however, anti-IL-2/ 15R␤ mAb (TM-␤1) only slightly inhibited IL-7-mediated survival of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells as shown in Fig. 5 (11.3% inhibition compared with IL-7 alone). A similar degree of inhibition was observed by the addition of anti-IL-2 mAb (S4B6; 11.2% inhibition compared with IL-7 alone), suggesting that IL-2 has some role in IL-7-mediated survival of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells in vitro and that the slight inhibition of cell survival by anti-IL-2/ 15R␤ mAb (TM-␤1) may be due to an indirect secondary effect (i.e., a blockade of autocrine IL-2) but not as a result of an inhibitory signaling through the IL-2/15R␤ molecules. In contrast, anti-IL-2/15R␤ mAb (TM-␤1), but not anti-IL-2 mAb (S4B6), completely suppressed IL-15-mediated survival of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells as expected (Fig. 5). These results indicate that anti-IL-2/15R␤ mAb (TM-␤1) acts as a specific blocking mAb and does not transduce an unexpected inhibitory signal that suppresses cell survival. These data strongly suggest the existence of a novel IL-2/15R␤-utilizing cytokine distinct from IL-2 and IL-15 that mediates the division of memory CD8 ϩ T cells in vivo.

Expression of IL-2/15R␤ molecules on memory CD8 ϩ T cells is essential for their division after IL-2 depletion
The results using anti-IL-2/15R␤ mAb suggested the presence of the novel IL-2/15R␤-utilizing cytokine (Fig. 4), but they did not reveal which cell type(s) can respond to the putative cytokine. To investigate whether the putative IL-2/15R␤-utilizing cytokine acts directly on memory CD8 ϩ T cells, CD44 low naïve or CD44 high memory phenotype CD8 ϩ T cells were isolated from WT or IL-2/15R␤ KO mice, labeled with CFSE, and transferred into IL-15 KO mice. These host mice were then treated with anti-IL-2 mAb. As shown in Fig. 6A, CD44 low naive CD8 ϩ T cells were not affected by IL-2 depletion, but CD44 high memory phenotype CD8 ϩ T cells from WT mice underwent several divisions as expected. However, the donor CD44 high memory phenotype CD8 ϩ T cells lacking IL-2/15R␤ molecules failed to divide in response to the anti-IL-2 treatment. This result indicates that the putative cytokine acts directly through the IL-2/15R␤ molecules on memory CD8 ϩ T cells to promote their cell division.
The result of the in vitro experiment (Fig. 5) implied that autocrine IL-2 might contribute to survival and/or proliferation of memory CD8 ϩ T cells. This result and the requirement of IL-2/ 15R␤ molecules on memory CD8 ϩ T cells shown in Fig. 6A suggested a possibility that even though anti-IL-2 treatment was done in vivo, a trace of IL-2 molecules could induce division of memory CD8 ϩ T cells in an autocrine manner. To exclude this possibility, CD8 ϩ T cells from IL-2 KO mice were transferred into WT hosts that subsequently received anti-IL-2 treatment. Since IL-2 KO mice spontaneously develop immune disorders (38), we used very young IL-2KO and age-matched WT mice (Ͻ3 wk old). As shown in Fig. 6B, anti-IL-2 treatment increased a population of IL-2 KO CD8 ϩ T cells that underwent cell divisions from 10.9 to 23.2%, which was a similar degree to that of WT donor T cells (14.5-24.3%). This result suggests that donor cell-derived IL-2 itself was not the factor that promoted CD8 ϩ T cell division after anti-IL-2 treatment.

The putative IL-2/15R␤-targeting factor is a member of the IL-2 family
It is well known that every member of the IL-2-family cytokines shares the ␥ c molecule as a receptor component (7,8). Therefore, we examined whether the division of memory CD8 ϩ T cell induced by IL-2 depletion could be inhibited by anti-␥ c blocking mAb. CFSE-labeled CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells isolated from WT mice were transferred into IL-15KO hosts. The host mice were then treated with the combination FIGURE 4. A putative IL-2/15R␤-utilizing factor mediates the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells induced by IL-2 depletion. A, CFSE-labeled CD44 high IL-2/15R␤ high CD8 ϩ T cells sorted out from WT mice were transferred into nonirradiated IL-15 KO mice on day 0. Treatment with anti-IL-2 mAb was performed simultaneously with anti-IL-2/15R␤ mAb (TM-␤1, 0.03 mg) or anti-IL-2R␣ mAb (PC61, 0.3 mg) on days 2-6. CFSE levels were examined on day 8. B, Nonirradiated IL-15 KO mice were transferred with CFSE-labeled CD44 high IL-2/15R␤ high CD8 ϩ T cells on day 0. The F(abЈ) 2 of anti-IL-2/ 15R␤ mAb (TM-␤1, 0.1 mg) was administered during the anti-IL-2 treatment on days 2-6. CFSE levels were examined on day 8. C, IL-15 KO mice were treated with control IgG alone (rat IgG, 1 mg), or with anti-IL-2/15R␤ mAb (TM-␤1, 0.03 g), or anti-IL-2R␣ mAb (PC61, 0.5 mg) during the anti-IL-2 treatment for 5 days. The endogenous CD44 high IL-2/ 15R␤ high population in CD8 ϩ T cells was examined. Another anti-IL-2/ 15R␤ mAb (5H4), which can bind to IL-2/15R␤ molecules covered with TM-␤1, was used for flow cytometry. A and B, The CFSE-negative population is derived from the host mouse. The results are representative of more than three independent experiments.
of anti-␥ c blocking mAbs (4G3 and 3E12) during the anti-IL-2 treatment. As shown in Fig. 7, the division of donor cells in response to IL-2 depletion was totally abrogated when ␥ c signaling was blocked. This result suggests that the putative IL-2/15R␤utilizing cytokine shares the ␥ c and belongs to the IL-2 family of cytokines.

Discussion
Previously described mechanisms for the positive regulation of memory CD8 ϩ T cells in homeostasis are largely dependent on IL-15 (13-23, 27, 31). In contrast to those observations, our present study shows that anti-IL-2 treatment induced the division of CD44 high IL-2/15R␤ high memory phenotype CD8 ϩ T cells, even in IL-15 KO mice and in IL-7-depleted IL-15 KO mice (Fig. 3,  B-D). These findings imply a novel process for the Ag-independent homeostasis of memory CD8 ϩ T cells in vivo. We found that the treatment of IL-2-depleted IL-15 KO mice with anti-IL-2/15R␤ mAb, but not with anti-IL-2R␣ mAb, abrogated the division of the CD44 high IL-2/15R␤ high memory-phenotype CD8 ϩ T cells (Fig. 4A). In addition, we showed that memory phenotype CD8 ϩ T cells from IL-2/15R␤ KO mice failed to divide in response to IL-2 depletion (Fig. 6A). Only IL-2 and IL-15 have been identified as IL-2/15R␤-utilizing cytokines, suggesting the existence of a putative third IL-2/15R␤-utilizing cytokine that acts on memory CD8 ϩ T cells to induce IL-15-independent division of these cells in vivo. Since a blockade of ␥ c signaling inhibited the division of memory CD8 ϩ T cells induced by IL-2 depletion (Fig.  6B), we suggest that the putative IL-2/15R␤-utilizing cytokine is a member of the IL-2-family of cytokines.
Several lines of evidence suggest that the putative factor is not IL-2 itself. First, it is reasonable that there is almost no IL-2 in the mice treated with anti-IL-2 mAb, because we hardly saw CD25 high CD4 ϩ regulatory T cells and clearly detected an excess amount of the Ab molecules in serum in the mice (data not shown). Second, a blockade of IL-2/IL-15R␤ or ␥ c signaling was inhibitory, whereas that of IL-2R␣ signaling was stimulatory for the cell division induced by anti-IL-2 treatment (Figs. 1D, 4, and 7). Third, there is a direct evidence that autocrine IL-2 is not required for the memory CD8 ϩ T cell division by an adoptive transfer of the cells from IL-2KO mice (Fig. 6B). Finally, if IL-2 acts in a paracrine manner even after anti-IL-2 treatment, memory CD4 ϩ T cell division must have been observed (Fig. 1).
Theze and colleagues (39) reported that a chemically synthesized 30-aa fragment of the N-terminal human IL-2 (p1-30) acts as an agonist for the IL-2/15R␤ signaling). In addition, several alternative forms of mouse IL-2 were listed in the expressed sequence tags databases, although there were no N-terminal short fragments similar to the p1-30 in the databases and S4B6, the anti-IL-2 mAb we used, recognizes an N-terminal portion of IL-2 molecules (p26 -45) that overlaps with the p1-30 (40). However, it might still be possible to speculate that in vivo IL-2 depletion by S4B6 increases the activity of the alternative forms that do not contain the S4B6 determinant (e.g., p1-26). We hypothesize that such natural variants might act as agonists for IL-2/15R␤-induced signaling on memory CD8 ϩ T cells, especially if the concentration of such variants increases after IL-2 depletion in vivo. To examine this possibility, we performed RT-PCR experiments using splenic FIGURE 6. The expression of IL-2/15R␤ on memory CD8 ϩ T cells is essential for the division of these cells after IL-2 depletion. A, CD44 high memory phenotype CD8 ϩ T cells were sorted from WT or IL-2/15R␤ KO mice (CD45.2), then labeled with CFSE and transferred into nonirradiated IL-15 KO mice (CD45.1) on day 0. Treatment with anti-IL-2 mAb was performed on days 1-5. The CFSE labels were examined on day 7. Histograms shown are gated on the CD45.2-expressing donor population. B, T cell-enriched fraction from WT or IL-2 KO mice (Thy-1.2) within 3 wk old was labeled with CFSE and transferred into nonirradiated WT hosts (Thy-1.1) on day 0. Anti-IL-2 treatment was performed on days 1-7. CFSE labels were examined on day 8. Histograms shown are gated on the Thy-1.2-expressing donor CD8 ϩ population. The results are representative of two independent experiments.

FIGURE 7.
A blockade of the ␥ c signaling abrogates the division of memory CD8 T cells after IL-2 depletion. CFSE-labeled CD44 high IL-2/ 15R␤ high CD8 ϩ T cells from WT mice were transferred into nonirradiated IL-15 KO mice on day 0. Treatment with control IgG (rat IgG, 0.2 mg) or anti-␥ c mAbs (4G3 and 3E12, 0.1 mg each) was performed during the anti-IL-2 treatment (days 1, 2, and 4 -6). CFSE levels were examined on day 7. The CFSE-negative population is derived from the host mouse. The results are representative of two independent experiments. cDNA with or without the anti-IL-2 treatment by several primer sets for the murine IL-2 gene. However, we did not observe a consistent result showing that the transcripts for the alternative forms significantly increased after anti-IL-2 treatment. Additionally, we have not seen the existence of the short N-terminal fragments of the IL-2 mRNA (data not shown). Furthermore, we observed that the blockade of ␥ c signaling abrogated the division of memory CD8 ϩ T cells after IL-2 depletion in IL-15 KO mice (Fig. 6B). This result shows a sharp contrast to the report by Eckenberg et al. (39), since they demonstrated that IL-2/15R␤mediated, but not ␥ c -derived, signaling is critical for the action of p1-30. Therefore, we suggest that the putative cytokine described here is not the natural variant(s) of IL-2 molecules that are agonistic to the IL-2/15R␤ signaling, similar to p1-30.
Since a deficiency of IL-15 or IL-15R␣ in mice leads to significant impairment of the maintenance of CD44 high IL-2/15R␤ high memory CD8 ϩ T cells (17)(18)(19), we hypothesize that the putative IL-2/15R␤-utilizing factor would act downstream of or in parallel to IL-15. In this regard, it is intriguing that the expression of IL-15R␣ on memory CD8 ϩ T cells is dispensable, whereas its expression on the other cells is shown to be important for their IL-15-mediated division in vivo (22)(23)(24). It has also been reported that IL-2/15R␤, another component for IL-15R complex, is required to be expressed on memory CD8 ϩ T cells to respond to IL-15 injection (23). Transpresentation of IL-15 by the IL-15R␣-bearing cells to other IL-2/15R␤-and ␥ c -expressing cells such as memory CD8 ϩ T cells (IL-15R␣ low or IL-15R␣ neg CD44 high ) has been reported in vitro (22,26), which could explain the dispensability of IL-15R␣ and the necessity of IL-2/15R␤ expression on memory CD8 ϩ T cells in vivo. Additionally, another possibility proposed by Ma and colleagues (27) is that the proliferation of memory CD8 ϩ T cells may be induced by an undefined growth factor secreted from IL-15R␣-bearing non-T cells after IL-15 stimulation. The in vivo function of the putative IL-2/15R␤-utilizing cytokine suggested here might be consistent with their idea, if IL-15 induces and if IL-2 suppresses secretion of the cytokine or expression of its receptor component(s).
Blattman et al. (41) recently reported that the injection of rIL-2 increases the proliferation of virus-specific resting memory CD8 ϩ T cells in mice. The administration of exogenous IL-2 is also beneficial for the proliferation of activated CD8 ϩ T cells in vivo (42). One interpretation for the inconsistency between our results (i.e., IL-2 depletion induces the division of CD44 high IL-2/15R␤ high memory CD8 ϩ T cells shown in this study and in Refs. 9 and 10) and theirs is that IL-2 acts as a two-edged sword: the endogenous, relatively low concentration of IL-2 is inhibitory, but the exogenous IL-2 supplement could raise the concentration to levels that are much higher than in the normal in vivo state, thereby promoting the division of memory CD8 ϩ T cells. In fact, a high concentration of IL-2 stimulates the proliferation of CD44 high memory phenotype CD8 ϩ T cells in vitro (10). It is likely that the rIL-2 treatment would have a direct action on memory CD8 ϩ T cells even in vivo. In contrast, our previous data strongly suggest that the inhibitory effect of IL-2 on the division of memory phenotype CD8 ϩ T cells is the result of an indirect action through other cell populations, including CD25 ϩ CD4 ϩ T cells (10). Therefore, these distinct mechanisms of IL-2 action, i.e., direct vs indirect, on memory CD8 ϩ T cells could cause the different outcomes.
In summary, our present study demonstrates that IL-2 negatively regulates the homeostatic division of memory CD8 ϩ T cells through an IL-15-independent mechanism. In addition, we provide evidence that this process is completely dependent on IL-2/15R␤ molecules expressed on memory CD8 ϩ T cells, suggesting the existence of a third IL-2/15R␤-utilizing cytokine that regulates the homeostasis of memory CD8 ϩ T cells in vivo, although we understand the possibility that differential efficacies of different Abs that may influence the results might be retained along with multiple indirect effects of Abs on various cell types in vivo. Molecular identification of the putative IL-2/15R␤-utilizing cytokine, one of our current research efforts, is thus an issue of great importance for understanding the regulation of memory CD8 ϩ T cell homeostasis.