Contribution of 4–1BBL on radioresistant cells in providing survival signals through 4–1BB expressed on CD8+ memory T cells in the bone marrow

Authors


Abstract

The persistence of memory lymphocytes is a critical feature of adaptive immunity. The TNF family ligand 4–1BBL supports the antigen-independent survival of CD8+ memory T cells. Here, we show that mice lacking 4–1BB only on αβ T cells show a similar defect in CD8+ T-cell recall responses, as previously shown in 4–1BBL-deficient mice. We show that 4–1BB is selectively expressed on BM CD8+ but not CD4+ memory T cells of unimmunized mice. Its ligand, 4–1BBL, is found on VCAM-1+ stromal cells, CD11c+ cells, and a Gr1lo myeloid population in unimmunized mice. Adoptive transfer of in vitro generated memory T cells into mice lacking 4–1BBL only on radioresistant cells recapitulates the defect in CD8+ T-cell survival seen in the complete knockout mice, with smaller effects of 4–1BBL on hematopoietic cells. In BM, adoptively transferred DsRed CD8+ memory T cells are most often found in proximity to VCAM-1+ cells or Gr1+ cells, followed by B220+ cells and to a much lesser extent near CD11c+ cells. Thus, a VCAM-1+CD45 stromal cell is a plausible candidate for the radioresistant cell that provides 4–1BBL to CD8+ memory T cells in the BM.

Introduction

Immunological memory is a key feature of our adaptive immune system. The persistence of memory lymphocytes affords the host long-term protection against reinfection. It is thought that lymphocytes must compete for space in defined cellular niches that are specific to a particular subset of lymphocytes [1, 2]. The cell types and key molecular components that make up the supportive niches for memory T cells are beginning to be defined [3-6]. These niches are expected to contain the chemokines that attract the lymphocytes to the site [3, 7], the adhesion molecules that provide retention signals at the site [5, 7], as well as the common γ-chain (γc) cytokines that provide homeostatic proliferative signals to the lymphocytes [8].

For CD8+ T cells, there is strong evidence that both IL-15 and IL-7 are required for their maintenance [8-17]. CD8+ CD44Hi memory phenotype T cells home to and are enriched in the BM [7, 18]. Moreover, the BM contains virus-specific memory T cells that can protect against reinfection [19], and CD8+ memory T cells in the BM show evidence of homeostatic proliferation [20, 21], independently of secondary lymphoid organs [22]. Thus, it has been proposed that the BM is a major site for homeostatic proliferation of CD8+ memory T cells [23]. However, there is limited evidence as to the nature of the BM niches that support the proliferation and survival of these cells.

In addition to a requirement for chemokines, γc cytokines, and adhesion molecules, emerging data also suggest that ligands of the TNF family are important players in maintaining immunological memory [24-27]. Previous studies have established that the TNF family ligand, 4–1BBL, provides an antigen-independent survival signal to CD8+ memory T cells [24, 28, 29]. Previous results using adoptive transfer of in vitro generated OT-I memory T cells into unimmunized mice revealed a two- to threefold defect in their maintenance after 3 weeks in 4–1BBL-deficient mice, under conditions where there was no defect in cell division [29]. 4–1BB engagement provides a survival signal to CD8+ effector and memory T cells that involves the TRAF1-dependent downmodulation of Bim [30, 31]. However, the cells that contribute 4–1BBL to the CD8+ memory T cells have not been identified.

In this report, we used BM chimeras to demonstrate that αβ T cells must express 4–1BB for maximal recall responses to influenza virus. In unimmunized mice, 4–1BB is preferentially expressed on CD8+ memory T cells in BM with minimal expression in the spleen or LN. We detected 4–1BBL expression on CD11c+ MHC class II (MHC II) cells, Gr1lo hematopoietic cells, as well as on VCAM-1+ CD45 stromal cells from the BM of unimmunized mice. Adoptive transfer of CD8+ memory T cells into radiation chimeras showed that 4–1BBL expressed on a radioresistant cell is important for maximal recovery of CD8+ memory T cells after parking the cells in the chimeric mice lacking antigen. The most abundant cell types found in close proximity to the transferred CD8+ memory T cells in the BM are VCAM-1+ stromal cells and Gr1+ cells, followed by B220+ cells. As the CD45 VCAM-1+ cells express 4–1BBL, a VCAM-1+ stromal cell is a plausible candidate for the radioresistant cell that provides 4–1BBL to sustain memory CD8+ T cells.

Results

4–1BB is required on αβ T cells for maximal recall responses of CD8+ T cells

Previous results have shown that 4–1BBL contributes signals to maintain CD8+ memory T cells in the absence of their specific antigen in vivo [29]. To address whether the effect of 4–1BBL requires that its receptor, 4–1BB, is expressed on the T cells, we first asked whether 4–1BB-deficient mice have the same decrease in CD8+ T-cell responses to influenza as previously determined for 4–1BBL-deficient mice [28]. We find that, similarly to results reported for 4–1BBL-deficient mice [28], the CD8+ T-cell response to influenza virus is unimpaired at the peak of the primary response in 4–1BB-deficient mice, but shows a statistically significant decline in the frequency of CD8+ T cells at 3 weeks post infection (Supporting Information Fig. 1A). This decline in CD8+ T cells late in the primary response correlates with a proportional decrease in secondary response upon rechallenge (Supporting Information Fig. 1A and B).

To determine whether this defect was T-cell intrinsic, we generated mixed BM chimeras, in which only the BM-derived αβ T cells lack 4–1BB and compared these with completely 4–1BB-sufficient mice (Fig. 1A). We used a ratio of 1:4 4–1BB−/− to TCRα-deficient BM, so that all the T cells would lack 4–1BB, but only 20% of the non-T cells would be 4–1BB-deficient. Consistent with the result obtained in the complete 4–1BB−/− mice (Supporting Information Fig. 1A), 4–1BB on αβ T cells is dispensable for the primary CD8+ T-cell response to influenza virus (Fig. 1B and Supporting Information Fig. 2 for gating strategy). Upon secondary challenge with influenza A/PR8, the absence of 4–1BB on αβ T cells results in a significant decrease in the nucleoprotein (NP)-specific CD8+ T-cell response in the spleen and BM (Fig. 1C). For the mice used in Figure 1C, we had also confirmed the absence of a defect in primary response based on analysis of blood T cells at day 7 following priming (data not shown). Thus, 4–1BB expression on the αβ T cells is required for the maximal CD8+ T-cell recall response to influenza virus.

Figure 1.

4–1BB is required on αβ T cells for maximal CD8+ T-cell recall response. (A) Mixed BM chimeras were generated by reconstituting lethally irradiated CD45.1 C57BL/6 WT mice with a 4:1 mixture of TCRα−/−. WT or 4–1BB−/− BM cells as illustrated. Thirteen weeks post reconstitution, the BM chimeras were infected intraperitoneally with influenza X31, and then challenged with influenza PR8 a month later. CD8+ T-cell responses were analyzed on (B) day 7 postprimary or on (C) day 7 postsecondary infection by Db/NP366–374 tetramers. CD8+ T-cell effector function was analyzed by intracellular staining for IFN-γ after 6-h restimulation of the splenocytes with NP peptide (right). Data shown are gated on the CD45.2+ BM donor derived CD8+ T cells (gating strategy shown is in Supporting Information Fig. 2). Data shown are the sum of two independent experiments; each with five mice per group and each symbol represents an individual mouse. The lines represent the median of the data set. *p < 0.05, **p < 0.01, by Student's t-test.

4–1BB expression in unimmunized mice

Our finding that 4–1BB is required on αβ T cells for maximal recall responses coupled with our previous findings that 4–1BBL is required for the maintenance of memory CD8+ T cells in the absence of antigen in vivo [29], suggests that 4–1BB on T cells binding to 4–1BBL in mice contributes to the maintenance of the memory CD8+ T cells. Thus, 4–1BB should be expressed on T cells in unimmunized mice. A recent study reported a low level of 4–1BB expression on CD44Hi CD8+ T cells in the BM of unimmunized mice [32]. Here, we extend this analysis to examine 4–1BB expression on CD8+ and CD4+ CD44Hi T cells from BM as well as the spleen and LN of unimmunized WT mice, using 4–1BB−/− mice as a staining control. We observed a small population of 4–1BB+ CD44Hi CD8+ T cells in the BM of WT mice, however, this population is considerably smaller in LN and splenic CD8+ CD44Hi cells (Fig. 2A).

Figure 2.

4–1BB is expressed on CD8+ CD44Hi cells predominantly in the BM relative to spleen or LN of unimmunized mice. (A) 4–1BB expression on CD8+ T cells was analyzed directly ex vivo on cells from the LN, spleen, and BM of unimmunized mice. The 4–1BB+ gates were drawn based on the negative control obtained using cells from 4–1BB−/− mice stained with anti-4–1BB antibody. Unimmunized GFP FoxP3 knock-in mice were used to gate out the regulatory cells for analysis of 4–1BB expression on CD4+ FoxP3-negative (GFP-negative) nonregulatory T cells. (B) 4–1BB expression on CD4+ FoxP3+ (GFP+) cells recovered in the LN, BM, and spleen of unimmunized mice. (C) Gating strategies for the CD8+, CD4+ FoxP3, and CD4+ FoxP3+ T-cell populations. (D) Comparison of 4–1BB expression in the BM and spleen of C57BL/6 and BALB/c mice. (A–D) Results shown are representative of three to four mice per group in two independent experiments lines were drawn at the median. *p < 0.05, **p < 0.01 by one-way ANOVA.

A complication of analyzing 4–1BB on memory CD4+ T cells is that CD4+ Treg cells constitutively express 4–1BB [33, 34]. Thus, we used GFP-FoxP3 reporter mice to distinguish the CD4+ Treg population from the effector/memory CD4 T cells. As previously reported [34], 4–1BB is expressed on a significant proportion of GFP+ CD4+ Treg cells in spleen, LN, and BM (Fig. 2B). However, when the GFP-negative CD4+ CD44Hi cells were analyzed, little or no 4–1BB was detected compared with the CD8+ CD44Hi cells (Fig. 2A).

We also analyzed mice with a different genetic background, BALB/c, and found that similar to C57BL/6 mice, BALB/c mice have higher 4–1BB expression on CD8+ memory T cells in the BM compared with that in the LN and spleen of unimmunized mice (Fig. 2D). A similar trend of preferential 4–1BB expression in 129/SvImJ mice was also found in a separate experiment with three mice per group (data not shown). These results show that 4–1BB is selectively enriched on the CD8+ but not CD4+ memory T cells in the BM of unimmunized mice as compared with the LN and spleen, which show minimal 4–1BB expression.

Expression of 4–1BBL on BM cells from unimmunized mice

As 4–1BBL is required for the maintenance of CD8+ memory T cells in the absence of antigen [29], and 4–1BB is preferentially expressed on the BM CD8+ memory T cells, 4–1BBL should also be detected on cells from BM of unimmunized mice. However, it was difficult to detect 4–1BBL expression without reactivation of APCs ex vivo, possibly due to its low or transient expression in unimmunized mice, its down modulation or masking in the presence of its receptor, and/or its susceptibility to metalloproteinase cleavage [35]. To avoid the issue of in vivo masking, downregulation, or cleavage, we infused mice with biotinylated anti-4–1BBL antibody or control biotinylated rat IgG antibody and 1 day later tissues were harvested for analysis. We consistently observed expression of 4–1BBL on the CD11c+ population from the BM of unimmunized, biotinylated anti-4–1BBL infused mice, but not in mice that had received biotinylated rat IgG and not in biotinylated anti-4–1BBL treated 4–1BBL-deficient mice (Fig. 3A). Further analysis showed that the 4–1BBL-expressing CD11c+ populations are negative with respect to CD11b, CD4, and CD8 markers, and are enriched in the MHC-IIneg fraction (Fig. 3A and Supporting Information Fig. 3). 4–1BBL is absent on the CD11c+ CD4+, CD11c+ CD8+, and plasmacytoid DCs of unimmunized mice (Fig. 3A and data not shown). Thus, 4–1BBL is expressed on a population of CD11c+ CD11b CD4, 8 double-negative MHC-IIneg cells in the BM of unimmunized mice (Fig. 3A).

Figure 3.

4–1BBL is expressed on CD11c+ cells and CD45VCAM-1+ BM stroma from unimmunized mice. WT and 4–1BBL−/− mice were given 100 μg of biotinylated anti-4–1BBL antibody or control biotinylated rat IgG intravenously 1 day before the mice were sacrificed and BM was harvested. Expression of 4–1BBL on BM (A) CD11c+ cells and (B) CD45.2 Ter-119 cells was measured directly ex vivo. The black line represents the staining for 4–1BBL, overlayed with the histograms obtained after staining biotinylated anti-4–1BBL-treated 4–1BBL−/− mice as shown in gray. Data shown are representative of (A) seven and (B) three independent experiments with similar results.

We also detected 4–1BBL expression on CD45-negative Ter-119-negative “stromal” cells from WT but not 4–1BBL−/− mice immediately ex vivo in some experiments (Fig. 3B). It was possible that the harsh conditions required to disrupt the BM to obtain the stroma led to cleavage of 4–1BBL, as TNF family ligands are known to be subject to metalloproteinase cleavage [36]. Therefore, we cultured the stroma from BM and then analyzed 4–1BBL expression on the CD45-negative cells. The VCAM-1+ stroma consistently expressed 4–1BBL; whereas yields from VCAM-1 stroma were lower and 4–1BBL expression was not consistently detected (Fig. 4A).

Figure 4.

Characterization of ex vivo cultured stromal cells. (A) 4–1BBL expression on cultured stromal cells. The black line shows 4–1BBL staining on stromal cells from WT mice and the gray line shows staining on 4–1BBL-deficient mice. Samples were gated on the CD45 and either the VCAM-1+ or VCAM-1 population as indicated in the figure. Similar results for VCAM-1+ cells were obtained in three experiments. (B–C) CD45 VCAM-1+ and CD45 VCAM-1 cells were sorted using the gates shown in (A). RNAwas isolated, reverse transcribed, and analyzed for (B) IL-7 or (C) CCL19 expression by PCR. Actin is used as the loading control. Sorted CD45 VCAM-1+ cells were further analyzed for (D) CXCL12 and (E) 4–1BBL expression, and GAPDHwas used as the loading control. M, molecular weight markers; the numbers 1, 2, and 3 above the gel refer to the relative amount of cDNA loaded. Data shown are representative of (B–D) three and (E) two independent experiments with similar results.

Previous studies have shown that CD4+ memory T cells in the BM are found in close association with IL-7+ VCAM-1+ stromal cells [5]. In addition, CCR7 has been implicated in accumulation of CD8+ memory T cells in the BM, whereas CXCL12 has been shown to contribute to memory CD8+ T cells adhering to BM microvessels [7]. Therefore, we analyzed sorted VCAM-1+ stroma for 4–1BBL surface expression as well as for expression of IL-7, CXCL12, and the CCR7 ligand CCL19. PCR analysis of sorted CD45 VCAM-1+ and VCAM-1 cells showed that both VCAM-1+ and VCAM-1 stromal cells expressed IL-7 mRNA, whereas CCL19 mRNA was detected in the VCAM-1+ cells (Fig. 4B and C). VCAM-1+ cells were also found to express CXCL12 (Fig. 4D) and consistent with the flow cytometry result in Figure 4A, 4–1BBL transcripts were also detected in VCAM-1+ stromal cells (Fig. 4E).

4–1BBL on a radioresistant cell increases the number of CD8+ memory T cells in vivo

We next asked whether 4–1BBL on the CD11c+ cells or CD45 VCAM-1+ stromal cells could be important in providing survival signals to CD8+ memory T cells in the absence of antigen. As most CD11c+ MHC II cells are radiosensitive, whereas stromal cells are radioresistant, we generated radiation chimeras using WT or 4–1BBL-deficient BM to reconstitute lethally irradiated WT or 4–1BBL−/− mice such that 4–1BBL is absent on radiosensitive cells, radioresistant cells, or in the whole animal. The reconstitution efficiency of the chimeras was above 90% in the BM and spleen, and above 85% in the LNs (Supporting Information Fig. 4), thus the phenotype we observed was unlikely to be due to incomplete chimerism. The CFSE-labeled, in vitro generated CD8+ OT-I memory T cells were adoptively transferred into the radiation chimeras and OT-I cell recovery was analyzed a month later (Fig. 5A). The adoptively transferred cells were tracked by their CD45.1 and CD45.2 markers as well as by staining for TCR Vα2 and Vβ5 (Fig. 5B). The frequency (Fig. 5C) and total number (data not shown) of adoptively transferred CD8+ memory T cells recovered was reduced approximately twofold when 4–1BBL was absent from the host, recapitulating the defect seen in the complete knockout (Fig. 5C). There was a smaller defect in the recovery of OT-I memory T cells when 4–1BBL was absent only on radiosensitive cells (Fig. 5C).

Figure 5.

4–1BBL is predominantly required on radioresistant cells for CD8+ memory T-cell maintenance in vivo. (A) BM chimeras were generated by reconstituting lethally irradiated WT or 4–1BBL−/− mice with WT or 4–1BBL−/− BM cells as illustrated. At 13 weeks post reconstitution, 3 × 106 in vitro generated CFSE-labeled CD8+ memory T cells were adoptively transferred into the BM chimeras. Recovery of adoptively transferred OT-I cells in LN, spleen, and BM was tracked at 30 days post transfer. (B) Gating strategy is shown and (C) results are summarized as % OT-I of CD8+ T cells. (D) CFSE dilution of the adoptively transferred OT-I memory cells measured in the different chimeric mice. The histograms show the CFSEstaining of the OT-I CD8+ memory T cells obtained from the BM of the chimeras, and similar results were found for spleen and LN (data not shown). The data in (B) and (C) are pooled from two independent experiments for the spleen, with BM and LN monitored in the second experiment, with each symbol representing an individual mouse. Lines were drawn at the median. *p < 0.05, **p< 0.01, ***p < 0.001, by one-way ANOVA.

The adoptively transferred CD8+ T cells had a similar CFSE profile among all four groups (Fig. 5D), and defects in CD8+ T-cell recovery were similar in all three organs examined, arguing that differences in the frequency of recovered CD8+ memory T cells after 30 days in the absence of 4–1BBL were not due to differential homeostatic proliferation or homing, but likely due to differences in survival, as concluded previously using the complete 4–1BBL knockout mice [29]. Thus, 4–1BBL on radioresistant cells contributes to the recovery of CD8+ memory T cells after adoptive transfer in vivo, with smaller effects from 4–1BBL on radiosensitive cells.

CD8+ memory Tcells in BM are found near VCAM-1+ stromal cells as well as Gr1+ cells

We next used immunohistochemistry to identify the cells that are the nearest neighbors of CD8+ memory T cells in the BM. To this end, we generated Red fluorescent OT-I memory T cells by crossing OT-I mice with ACTB-DsRed transgenic mice. This transgene leads to expression of Red fluorescent protein under control of the β-actin promoter. Although Red fluorescent protein is a foreign protein in mice, initial experiments showed similar recovery of in vitro generated CD45.1 OT-I memory T cells or Red fluorescent CD8+ memory T cells for at least 6 days post transfer (data not shown). We transferred 6 million OT-I-DsRed CD8+ memory T cells into WT mice and 1 day later analyzed their location by immunofluorescence microscopy. This time point was chosen based on initial kinetic experiments showing the highest numbers of Red OT-I T cells in the BM at 1 day post transfer followed by a gradual decline. This is the same time frame analyzed by previous investigators to identify interactions of CD4 memory T cells in the BM [5].

The transferred memory T cells were found randomly scattered in the BM, with no obvious overall distribution pattern at low magnification (Fig. 6A). To gain insight into their local environment, we used costaining with other markers to assess which cell types were in close proximity to the transferred memory T cells. More than 70% of OT-I-DsRed memory T cells were found in close contact with VCAM-1+ cells in contrast to <5% in contact with CD31+ endothelial cells or 13% with CD11c+ cells (Fig. 6B). VCAM-1 can be found on inflamed endothelial cells [37] as well as on stromal cells [38]. However, the finding that there was minimal association of the CD8+ memory cells with CD31+ cells argues that the VCAM-1-positive stromal cell is the most abundant cell to be found in close proximity to the transferred red memory T cells.

Figure 6.

CD8+ memory T cells are found in proximity with VCAM-1+ stromal cells and Gr1+ cells in the BM. In vitro generated DsRed CD8+ memory T cells were adoptively transferred into C57BL/6 WT mice and (A) 1 day later the location of DsRed memory T cells in the host BM was analyzed microscopically. Red squares represent locations where a DsRed CD8+ memory T cell was found. Two representative pictures from three independent experiments are shown. (B) DsRed cells next to VCAM-1+, CD31+, CD11c+, F4/80+, IL-15+, B220+, and Gr1+ cells were enumerated and shown as frequencies of DsRed cells in contact with any such cell types in the BM of WT mice. Eighty DsRed cells were counted in order to determine the frequency of contact with VCAM-1+, 84 for CD31+, 94 for CD11c+ and F4/80+, 64 for IL-15+, 84 for B220+, and 91 for Gr1+ in the bones from three independent adoptive transfer experiments. (C) Frequencies of DsRed cells in contact with VCAM-1+, B220+, and Gr1+ cells in WT and 4–1BBL-deficient mice. Data from WT mice are repeated from panel B. a total of 190 DsRed cells were analyzed to determine the frequency of contact with B220+, 91 for Gr1+ cells, and 106 for VCAM-1+ in 4–1BBL−/− mice. (D) Representative pictures of three independent experiments are shown for CD8+ memory T cells (red) in close proximity with VCAM-1+ cells (blue, upper left), with a CD11c+ (blue, upper right), with an F4/80+ cell (green, middle left), with a B220+ cell (blue, middle right), with Gr1+ cells (green, lower left), and IL-15+ cells (green, lower right) and CD45+ cells (blue, lower right). Scale bars are 11 μm, pictures were taken at 20× or 40× magnification.

The second most abundant interaction of the memory T cells was with Gr1+ cells (50% of CD8+ memory T cells and this was not significantly different from the number found in proximity to VCAM-1+ cells). B220+ cells were found in close proximity with 35% of memory T cells and this was significantly lower than the number associated with VCAM-1+ cells. F4/80-positive cells were associated with 25% of the CD8+ memory T cells. We also showed that the Gr1+ and B220+ cells located in proximity to the OT-I-DsRed memory T cells did not coexpress the Gr1 and B220 markers (Supporting Information Fig. 5). Thus, these cells are not plasmacytoid DCs (which coexpress Gr1 and B220), but myeloid cells or granulocytes (Gr1+) and B cells.

As Gr1+ and B220+ cells were found near the red memory T cells, we next asked whether these cells could express 4–1BBL. The B220 cells from BM are 4–1BBL negative (Supporting Information Fig. 6A) as are Gr1hi cells (Supporting Information Fig. 6B). However, 4–1BBL is present at low levels on a population of cells that express lower levels of Gr1 (Gr1lo), likely a myeloid population in the BM (Supporting Information Fig. 6B). Further analysis of the Gr1lo cells shows that they express Ly-6C, CD11b, F4/80, and a low level of MHC-II but lack CD11c (Supporting Information Fig. 6C). On the other hand, we were unable to detect 4–1BBL by immunofluorescence on the sections of unimmunized mouse BM, even with prior infusion of biotinylated anti-4–1BBL and amplification (data not shown).

We also asked whether the absence of 4–1BBL in the mouse affected localization of the OT-I-DsRed memory T cells relative to other cells. A similar number of CD8+ memory T cells found were found in the BM sections of 4–1BBL-deficient BM 1 day post transfer (data not shown) and the absence of 4–1BBL did not change the percentages of CD8+ memory T cells associating with the VCAM-1+, B220+, or Gr1+ cells (Fig. 6C).

In sum, these data show that transferred CD8+ memory T cells can most often be found in close proximity to VCAM-1+ stromal cells and Gr1+ cells. As VCAM-1+ stroma can express 4–1BBL and the VCAM-1+ stromal cells are radioresistant, but Gr1+ cells are normally radiosensitive, VCAM-1+ stromal cells are a plausible candidate for the radioresistant cells that provide a 4–1BBL signal to maintain CD8+ memory T cells.

Discussion

Immunological memory induced by natural infection can last for decades even in the apparent absence of the inducing antigen in the environment [39]. Understanding the mechanisms that maintain immunological memory should provide insights into how one could manipulate the immune system to enhance long-term memory as we age. There has been much interest in understanding the factors required for the maintenance of immunological memory. The cellular and molecular nature of the immunological niches required for the maintenance of CD4 T cells and plasma cells in the BM is beginning to emerge. A CXCL12 and VCAM-1-positive, IL-7-negative mesenchymal cell in the BM interacts with long-lived plasma cells [3, 4], whereas CD4 memory T cells interact with a CXCL12-negative IL-7+ VCAM-1+ stromal cell [5]. The equivalent stromal cell for CD8+ memory T cells in the BM has yet to be defined [4]. In this study, we show that CD8+ memory T cells, like CD4 memory T cells, are found in the BM in close proximity with VCAM-1+ stromal cells. Moreover, we find that 4–1BBL on a radioresistant cell contributes to the maintenance of CD8+ memory T cells by 4–1BB. Our finding that 4–1BBL is expressed on CD45 VCAM-1+ stromal cells points to the VCAM-1+ stromal cell as a plausible candidate for the radioresistant cell that provides 4–1BBL to CD8+ memory T cells in the BM to support their maintenance. We also show that 4–1BBL on radiosensitive cells contributes to CD8+ memory T-cell recovery in unimmunized mice, albeit to a lesser extent than the radioresistant cells. In unimmunized mice, 4–1BBL is expressed on CD11c+ MHC II cells, however, only a small fraction of adoptively transferred CD8+ memory phenotype cells were found in contact with CD11c+ cells, making it difficult to evaluate their importance. We also detected 4–1BBL on Gr1lo CD11b+ F4/80+ MHC-IIlo CD11c cells from the BM of unimmunized mice, thus this population could be the radiosensitive cell that contributes 4–1BBL to the CD8+ T cells.

Previous studies have established that 4–1BBL is required for the maintenance of influenza-specific CD8+ T cells between 3 and 6 weeks post infection with influenza A/HKx31 virus, a time when this virus has been fully cleared from the host [28]. Further studies, using adoptive transfer of TCR transgenic CD8+ OT-I memory T cells confirmed this role for 4–1BBL in the antigen-independent maintenance of memory CD8+ T cells and inferred that this was likely due to effects of 4–1BB signaling on survival rather than trafficking or cell division [29]. Here, we have provided evidence that an αβ T-cell must express 4–1BB for maximal recovery of CD8+ memory T cells. As 4–1BBL affects the CD8+ but not the CD4 response to influenza virus [28, 40] and 4–1BB is expressed on resting CD8+ memory but not CD4+ memory T cells in the BM of unimmunized mice (Fig. 2), these data argue that the effects of 4–1BBL are likely through direct effects on CD8+ T cells in the BM.

The association of transferred Red fluorescent memory T cells with the stromal cells was not affected by 4–1BBL-deficiency. Thus, although 4–1BBL affects the number of T cells recovered in the BM when assayed after 3 weeks [29], it does not appear to affect the positioning of the memory T cells in these short-term assays. This is not surprising, as 4–1BBL is not known as a cell adhesion molecule, and its effects on T-cell survival would not be expected to affect T-cell recovery within the 24 h of our microscopy study.

PCR analysis of sorted VCAM-1+ and VCAM-1 stroma showed preferential expression of CCL19 on the VCAM-1+ as compared with VCAM-1 stroma, consistent with a role for chemokines in attracting the CD8+ T cells to the VCAM-1+ stroma in the BM [7]. We also found CXCL12 in the cultured stromal cells. The association of the memory T cells with the VCAM-1+ cells in the BM is also consistent with the observation that memory T cells express three to four times the level of VLA-4 as compared with that of naïve T cells [41]. A caveat to these experiments is that VCAM-1+ cells are highly abundant in the BM and we have not shown that the proximity of the VCAM-1+ cells to the adoptively transferred memory T cells results in a productive interaction. Nevertheless, these data indicate that it is plausible that 4–1BBL+ VCAM-1+ cells could provide a signal to the CD8+ 4–1BB+ memory cells found in the BM.

CD8+ memory T cells rely on IL-15 presented by macrophages or DCs for their maintenance [8, 10, 12, 17, 42]. In particular, DCs can transpresent IL-15 in complex with the IL-15Rα-chain to central memory T cells and IL-15 transpresented by macrophages can support both effector and memory CD8+ T cells [17]. In our study, about 40% of the transferred memory T cells are in close proximity to either an F4/80+ or a CD11c+ cell. Recent studies show that human BM memory T cells are in close contact with cells expressing IL-15 message [43]. With our system, we did not observe enrichment of IL-15-expressing cells in proximity to the CD8+ memory T cells, as we found less than 2% of memory T cells in contact with IL-15+ cells. This might be due to the limited sensitivity of the IL-15 antibody stain, resulting in us only detecting cells with the highest IL-15 expression.

It has been reported that adoptively transferred leukemic cells as well as DCs and B cells populate perivascular regions in cranial bones of mice [44, 45]. In contrast to those studies, we did not observe enrichment of the transferred memory T cells to sub-regions within the BM, rather they were found randomly scattered throughout the BM. A reason for this difference in results might be the different T-cell types analyzed and/or differences in cellular organization in long bones as compared to the cranium.

We also detected other cell types located in close proximity to the transferred CD8+ memory T cells. The most abundant of these were the Gr1+ cells, whose proximity to the CD8+ memory T cells was not statistically different than that of the VCAM-1+ stromal cells. Based on flow cytometry, the Gr1hi granulocytes do not express 4–1BBL, whereas, 4–1BBL was detected on Gr1o MHC II+, CD11b+ F480+ cells in the BM of unimmunized mice (Supporting Information Fig. 6). We do not know if our microscopy is only detecting the abundant Gr1hi granulocyte population or also includes this 4–1BBL+ Gr1lo population.

About 35% of the memory T cells were found near B220+ cells. However, B220+ cells from the BM do not express 4–1BBL (Supporting Information Fig. 6A) and moreover, B cells are not essential for CD8+ T-cell memory [46] making it unlikely that the B cells make nonredundant contributions to the support of CD8+ memory T cells. It is also possible that these tangencies (with VCAM-1+, Gr1+, or B220+ cells) are merely coincidental, as we observed memory T cells touching up to eight cells in one section. Additionally, the cells could also be competing for similar stromal cell factors as the CD8+ T cells.

In conclusion, this study begins to define the cells that contribute to the maintenance of CD8+ memory T cells by 4–1BB and 4–1BBL. We demonstrate that 4–1BB on an αβ T-cell allows increased recall responses of CD8+ T cells. We further show that 4–1BBL on a radioresistant cell with lesser effects of 4–1BBL on a radiosensitive cell allows increased recovery of memory CD8+ T cells after parking in mice without antigen. The finding that 4–1BB is preferentially expressed on a subset of CD8+ memory T cells in the BM relative to the spleen and LN suggests that the BM is the likely site of the 4–1BB-4–1BBL interaction that contributes to the maintenance of CD8+ memory T cells. The finding that VCAM-1+ stroma express 4–1BBL, CCL19, CXCL12, and IL-7 and that adoptively transferred CD8+ memory T cells are often found in proximity to VCAM-1+CD45 cells in the BM demonstrates the plausibility of the VCAM-1+ stromal cell as the radioresistant cell that provides 4–1BBL to memory CD8+ T cells in the BM. These data support a model in which a radioresistant VCAM-1+ stromal cell attracts the VLA-4+ CD8+ memory T cells via CCL19, where they can receive 4–1BB-4–1BBL induced survival signals. As the VCAM-1-positive stromal population is very abundant in the BM, there may be heterogeneity in the VCAM-1+ stroma with respect to 4–1BBL, cytokines, and chemokines that contribute to CD8+ T-cell memory maintenance. Further analysis will be required to definitively identify the 4–1BBL-expressing radioresistant cell that contributes to CD8+ T-cell memory.

Materials and methods

Mice

C57BL/6 WT mice were obtained from Charles River Laboratories (St. Constant, QC, Canada). 4–1BB−/− mice [47] extensively backcrossed to the C57BL/6 (n = 10) background were bred in our facility. These mice were previously provided to us by Dr. Byoung S. Kwon (National Cancer Center, Ilsan, Korea). 4–1BBL-deficient (4–1BBL−/−) mice were originally obtained under a materials transfer agreement from Immunex (Amgen, Thousand Oaks, CA, USA) and further backcrossed to the C57BL/6 background in our facility (total n = 9). OT-I and CD45.1 congenic mice were obtained from Jackson Laboratories (Bar Harbor, ME, USA) and crossed to generate CD45.1+/+ or CD45.1+/− OT-I mice. TCRα−/– mice were kindly provided by Dr. Cynthia Guidos (Hospital for Sick Children, Toronto). FoxP3gfp knock-in mice on the C57BL/6 background were kindly provided by Dr. Mohamed Oukka (Harvard Medical School) [48]. ACTB-DsRed transgenic mice expressing DsRed protein under control of the β-actin promoter and backcrossed to B6 mice for five generations (B6.Cg-Tg (ACTB-DsRed*MST) 1Nagy/J) were obtained from the Jackson laboratories and crossed with OT-I mice to obtain OT-I ACTB-DsRed mice (OT-I-DsRed). Mice were maintained under specific pathogen-free conditions in sterile microisolators at the University of Toronto. All mouse experiments were approved by the University of Toronto animal care committee in accordance with the regulations of the Canadian Council on animal care (University of Toronto approved protocol #20007828).

Generation of memory T cells in vitro

CD8+ T cells with a central memory phenotype were generated by culture with Ag followed by IL-15 using a variation of a previous protocol [7, 29]. In brief, OT-I splenocytes were stimulated with 0.1 μg/mL SIINFEKL peptide and 1 μg/mL of LPS for 1 day, and then the nonadherent cells were rested for 2 days in fresh media (RPMI-1640 with 10% heat-inactivated FCS, 0.03% L-glutamine, antibiotics, and 2-mercaptoethanol). Live cells were enriched by lympholyte (Cedarlane) followed by culturing in media containing 20 ng/mL recombinant human IL-15 (R&D, Minneapolis, MN). Media was replaced every 2 days for a total culture time of 9 days. For CFSE (Molecular Probes, Eugene, OR, USA) labeling, 5 × 107/mL T cells were incubated in prewarmed PBS containing 1 μM CFSE for 10 min at 37°C followed by extensive washing and resuspension in PBS for adoptive transfer.

BM chimeras and adoptive transfer

Mixed BM chimeras in which only the αβ T cells lack 4–1BB were generated using TCRα−/− and 4–1BB−/− mice as described previously [49]. For the generation of the 4–1BBL−/− BM chimeras, 5 × 106 congenically marked 4–1BBL−/− or WT BM cells were used to reconstitute lethally irradiated WT or 4–1BBL−/− mice. All irradiated BM reconstituted mice were given water supplemented with 2 mg/mL of neomycin sulfate (Bio-Shop, Burlington, Ontario, Canada) during the first 4 weeks, and they were further rested for an additional 2 months before use. Three million OT-I T cells, prepared as above, were delivered i.v. to the mice and their recovery from spleen, LN, and BM analyzed 30 days later.

Influenza virus infection

Influenza A/PR8 and A/HKx31 viruses were grown in eggs and their tissue culture infectious dose determined by infection of MDCK cells [50]. Age- and sex-matched mice were used for infection. A dose of 100 HAU influenza A/X31 in 200 μL volume was used for primary intraperitoneal infection. Influenza A/X31 primed mice were rested for at least 30 days before challenge with influenza A/PR8 at a dose of 100 HAU in 200 μL intraperitoneally.

Flow cytometry and antibodies

Analysis of influenza NP366–374-specific CD8+ T cells using MHC tetramers as well as CD107a and intracellular cytokine staining following a 6-hour restimulation was carried out as previously described [28]. H-2Db/NP366–374 tetramers were provided by the National Institute for Allergy and Infectious Diseases tetramer facility (Emory University, Atlanta, GA, USA). Uninfected mice were used as negative controls for Db/NP366–374 tetramer staining. Isotype or fluorescence minus one controls were used as negative controls for cytokine staining. Congenically marked OT-I TCR transgenic cells were tracked using PE- or allophycocyanin-anti-mouse CD45.2 (eBioscience), Pacific Blue-anti-CD45.1 (Biolegend), FITC-anti-Vβ5.1 (BD Biosciences), biotin-anti-Vα2 and PerCP-anti-mouse CD8+ (BD Biosciences). Other antibodies used in this study included allophycocyanin-anti-mouse IFN-γ, FITC-anti-CD107a, PE- or PE-Cy5- or allophycocyanin-anti-CD44, FITC-anti-Ter119, Pe-Cy7- or PE-anti-mouse CD3, biotinylated-anti-mouse-4–1BB (clone 3H3), Alexa Fluor450- or PE-anti-B220, PE-, PE-Cy7- or allophycocyanin-anti-CD11c, Alexa Fluor488 anti-Gr1, PE-anti-Ly-6C, PE-anti-MHC-II, PE-Cy7-anti-F4/80, PerCP- or PE-Cy7-anti-CD11b, PE-Cy5.5-anti-mouse TCRβ, PE-Cy5.5-anti-mouse CD19, and FITC-anti-PDCA-1. The 4–1BB-deficient mouse was used as a negative control for analysis of 4–1BB expression on CD8+ T cells. Detection of 4–1BBL was done by i.v. infusion of 100 μg of biotinylated anti-4–1BBL Ab (clone TKS-I) or biotinylated Rat IgG 1 day before harvesting of the organs. Digested organs were further stained with biotinylated-anti-4–1BBL ex vivo, followed by Streptavidin-PE or Streptavidin-allophycocyanin amplification. Biotinylated anti-4–1BBL-treated 4–1BBL-deficient mice were used as a negative staining control for analysis of 4–1BBL expression. The samples were analyzed using FACScalibur, FACSCanto, or LSR II (BD Biosciences) with Cell-Quest or FACSDiva acquisition software. Data analysis was done using FlowJo software (TreeStar Inc., Ashland, OR, USA).

In vitro culture and analysis of BM stromal cells

Marrow was flushed from the femurs and tibias of 10 C57BL/6 mice and digested for 45 min at 37°C with 0.2 mg/mL Collagenase P (Roche) and 0.2 mg/mL DNase I (Sigma). The single cells were seeded in Petri dishes at a density of 1–2 × 106 cells/cm2. One day later, nonadherent cells were washed away and the remaining adherent cells were expanded in medium (see Generation of memory T cells in vitro) for up to 25 days. Half of the medium was replaced with fresh medium once a week. On day 25, adherent cells were removed by trypsinization, stained with anti-CD45.2 (ebioscience), anti-VCAM-1 (ebioscience), biotinylated anti-4–1BBL (19H3), and secondary streptavidin, and assessed by flow cytometry. In addition, CD45-negative cells were sorted into the VCAM-1+ and VCAM-1 population. Yields of CD45VCAM-1+ cells were approximately 500,000 in all three experiments, whereas yields of CD45VCAM-1 cells ranged 13,000- 165,000 cells. Total RNA was extracted and purified using RNeasy Micro kit (Qiagen). Random hexamer primers and purified RNA were used for the reverse transcription reaction (Invitrogen). PCR of the cDNA was done using following primers: 4–1BBL forward: 5′-CTT GAT GTG GAG GAT ACC-3′, 4–1BBL reverse: 5′-GCT TGG CGA ACA CAG GAG-3′, CCL19 forward: 5′-GCC TCA GAT TAT CTG CCA T-3′, CCL19 reverse: 5′-AGA CAC AGG GCT CCT TCT GGT-3′, IL-7 forward: 5′-TCC TCC ACT GAT CCT TGT TC-3′, IL-7 reverse: 5′-TTG TGT GCC TTG TGA TAC TG-3′, CXCL12 forward: 5′-GTC CTC TTG CTG TCC AGC TC-3′, CXCL12 reverse: 5′-TAA TTT CGG GTC AAT GCA CA-3′, actin forward: 5′-GGG AAT GGG TCA GAA GGA-3′, actin reverse: 5′-AAG AAG GAA GGC TGG AAA-3′, GAPDH forward: 5′-AAC TTT GGC ATT GTG GAA GG-3′, and GAPDH reverse: 5′-GGA GAC AAC CTG GTC CTC AG-3′.

Immunohistochemistry

CD8+ memory+ T cells were generated in vitro from OT-I DsRed splenocytes as described [29]. A total of 6 × 106 cells were adoptively transferred into C57BL/6 mice. One day later, femurs were harvested, fixed for 4 h in 4% paraformaldehyde at 4°C, dehydrated in 10, 20, and 30% sucrose solution for 24 h, respectively, and frozen in SCEM embedding medium (Section-Lab Co. Ltd., Yokohama, Japan). Bone cryosections (7 μm) were prepared using Kawamoto's Film Method [51]. Sections were stained with the following Ab from eBioscience if not indicated otherwise: VCAM-1 (429), CD31 (MEC13.3), CD11c (HL3) (BD Bioscience), F4/80 (A3–1) (Serotec), B220 (RA3–6B2), Gr1 (RB6–8C5), and rabbit-anti- IL-15. Bound anti-IL-15 was visualized by anti-rabbit antibody (Invitrogen). Antibodies were labeled with Alexa Fluor 488, Alexa Fluor 647, FITC, or allophycocyanin. BM was analyzed on a Quorum Spinning Disk Confocal Microscope, equipped with an ASI motorized XY stage. Data were analyzed using Volocity software (http://www.perkinelmer.ca/en-ca/pages/020/cellularimaging/products/volocitydemo.xhtml), which allowed individual pictures to be linked together to reconstruct the entire femur. Then, after identifying red fluorescent T cells at low magnification, the direct contacts of each transferred memory T cells were enumerated for each set of stains.

Statistical analysis

Where indicated, for comparison of two groups, p-values were obtained using the Student's t-test (unpaired, two-tailed, 95% confidence interval). One-way ANOVA was used to compare multiple groups, and statistical significant differences with p < 0.05, p < 0.01, and p < 0.001 were indicated as *, **, and ***, respectively.

Acknowledgments

We thank Byoung Kwon, National Cancer Center, Korea, for 4–1BB−/– mice; Robert Mittler, Emory University, for provision of the 3H3 anti-4–1BB and 19H3 anti-4–1BBL hybridomas, Hideo Yagita of Juntendo University for provision of the TKS-1 hybridoma; Peter Doherty and Paul Thomas, St. Jude Children's Research Hospital, for providing influenza A/HKx31-OVA; the National Institute of Allergy and Infectious Disease tetramer facility for MHC I tetramers, and Birinder Ghumman and Thanuja Ambagala for technical assistance. This research was funded by grant number MOP 84419 from the Canadian Institutes of Health Research (CIHR) to T.H.W. T.H.W. holds the Sanofi Pasteur chair in Human Immunology at the University of Toronto; G.H.Y.L. was funded by a CIHR doctoral award. F.E. was funded by a research fellowship of the German Research Foundation (DFG). A.E.H. was supported by research grant HA5354/4–1 from the German Research Foundation (DFG).

Conflict of Interest

The authors declare no financial or commercial conflict of interest.

Abbreviations
γc

common γ-chain

NP

nucleoprotein

OT-I-DsRed

OT-I TCR transgenic mice expressing the DsRed gene under the control of the β-actin promoter

Ancillary