The spleen contains numerous NK cells whose differentiation profile is characterized by a preponderance of mature elements located mainly in the red pulp. In contrast, lymph nodes (LNs) contain few NK cells and they are sited mostly in T cell zones and skewed toward immature developmental stages. We show that, in mice, naturally occurring CD4+Foxp3+ regulatory T (Treg) cells are both necessary and sufficient to repress accumulation of NK cells in resting LNs. Moreover, we present evidence that Treg cells hamper generation of mature NK cells through short-range interactions with NK precursors. In turn, mature NK cells specifically regulate the amount of CD8α+ phenotypically immature dendritic cells present in LN T cell zones. We propose that the dominant influence of Treg cells on NK cell precursors and CD8α+ immature dendritic cells explains why “quiescent” LNs in the absence of infection function as privileged sites for induction and maintenance of tolerance to peripheral Ags.
Natural killer cells originate from bone marrow (BM)3 precursors (1, 2). What is not clear, however, is whether BM is unique in its capacity to assure complete NK cell maturation. An alternative would be that BM generates NK precursors that may complete their maturation in other organs (3). This is the case for a recently described population of NK cells that is generated by a GATA-3- and CD127-dependent pathway in the thymus (4). Provocative evidence suggests that lymph nodes (LNs) may represent a prominent site for NK cell differentiation in vivo. Indeed, NK precursors and NK cell development intermediates are present in human LNs and tonsils (5, 6). In line with this, mice lacking LNs due to knockout of the helix-loop-helix inhibitor Id2 gene (Id2−/−) are deficient in NK cells but not T or B lymphocytes (7). Moreover, lymphotoxin α (Lta)-deficient mice show impaired LN organogenesis and defective NK cell development (8). However, none of these observations provides incontrovertible evidence that LNs are significant sites of NK cell differentiation or development. Thus, it is formally possible that NK cell lineage populations do not arise within, but merely traffic through, the LNs (6). Furthermore, the concept that LNs are an important site of NK cell differentiation is difficult to reconcile with the fact that, under steady-state conditions, LNs contain very few NK cells (9, 10). In the absence of infection, murine NK cells represent an average of 2–8% of total lymphocytes in the blood and spleen, but only 0.5% in the LNs (11, 12). In addition, in vitro culture studies suggest that failure of NK cell production in Id2−/− mice is due, at least in part, to a defect in NK cell precursors rather than the microenvironment (7). Finally, injection of exogenous IL-15 can rescue the NK cell deficiency in Lta−/− mice suggesting that lack of LNs does not account for failure of NK cell development in those mice (13).
The initial objective of our work was to determine whether LNs normally contribute to NK cell development and homeostasis under steady-state conditions. We report that absence of LNs and Peyer’s patches in aly/aly mice has no impact on NK cell subset phenotypes in the BM, blood, and spleen. Resting LNs contain few NK lineage cells that are skewed toward immature developmental stages. Because naturally occurring CD4+Foxp3+ regulatory T (Treg) cells have a critical role in dampening NK cell immune responses (14, 15), we postulated that this cell type might influence the development and distribution of NK lineage cells in secondary lymphoid organs. We show for the first time that under steady-state conditions, CD4+CD25+ Treg cells are necessary and sufficient to repress accumulation of mature NK cells in LNs. We present evidence that CD4+CD25+ Treg cells directly hinder in situ generation of NK cells through short-range interactions with NK precursors in the LN paracortex. Accordingly, production of mature NK cells by LN-resident NK precursors was increased in the absence of CD4+CD25+ Treg cells but was hampered following adoptive transfer of CD4+CD25+ Treg cells. In turn, the frequency of mature NK cells selectively regulates the abundance of CD8α+ immature dendritic cells (iDCs) in the LN. Together, these data support an emerging model in which Treg cells directly control the maturation of NK cells and indirectly modulate the proportion of CD8α+ iDCs in the LN.
Materials and Methods
Mice and cells
C57BL/6 and B cell-deficient (B6.129S2-Igh-6tm1Cgn) mice were obtained from The Jackson Laboratory; female HY-TCR-transgenic Rag2−/− mice (B10.Cg-Rag2tm1Alt-Tg(TcrH-Y)Vbo N11) and P14 TCR-transgenic Rag2−/− mice ((C57BL/6 × C57BL/10SgSnAi)–(Tg)TCR LCMV P14-(KO)rag2) were obtained from Taconic Farms; aly/aly and aly/+ mice were obtained from CLEA Japan; and 2826 Foxp3-transgenic were a gift from Dr. F. Ramsdell (Celltech R&D, Bothell, WA). Mice were housed under specific pathogen-free conditions in the animal care facilities of the Institute of Research on Immunology and Cancer, and Lyman Duff Medical Building, and used between 7 and 15 wk of age. All work involving mice was conducted under protocols approved by the Comité de Déontologie de l’Expérimentation sur des Animaux of the University of Montreal. OP9-GFP cells were cultivated in α-MEM medium supplemented with 10% FBS, penicillin, streptomycin, glutamine, 2-ME, HEPES, and sodium pyruvate (16).
Abs and reagents
Abs against the following molecules were used: CD11c-FITC (HL3), CD122-FITC (TM-β1), purified CD3ε (145-2C11), CD3ε-PE Cy5 (145-2C11), CD4-allophycocyanin (RM4-5), CD4-PE Cy5 (H129.19), CD43-FITC (S7), CD45R/B220-allophycocyanin (RA3-6B2), CD45R/B220-allophycocyanin Cy7 (RA3-6B2), CD49b-PE (DX5), CD49b-allophycocyanin Cy7 (HMα2), CD8α-PE Cy5 (53-6.7), CD8α-PE Cy7 (53-6.7), TCRβ-PE Cy5 (H57-597), I-Ab-PE (AF6-120.1), CD103-biotin (M290), streptavidin-PE, and streptavidin-allophycocyanin obtained from BD Pharmingen; CD11b-allophycocyanin Cy7 (M1/70), CD11c-biotin (N418), CD19-PE Cy5 (6D5), NKG2D-biotin (CX5), CD3ε-Alexa 647 (17A2), Foxp3-biotin, and Foxp3-PE (FJK-16s) obtained from eBioscience; anti-rat-Alexa 488 and streptavidin-Alexa 568 obtained from Invitrogen Life Technologies-Molecular Probes; CD11b-PE (M1/70) and CD25-PE (PC61.3) obtained from Cedarlane Laboratories; CD86-allophycocyanin (PO3) obtained from Biolegend; NKG2D (191004) obtained from R&D Systems. PK136 (anti-NK1.1) and PC61 (anti-CD25) Abs were purified from hybridoma supernatants. Isotype control mouse IgG and rat IgG were purchased from Sigma-Aldrich. Collagenase type IV and DNase I were obtained from Sigma-Aldrich. Murine rIL-2, IL-15, stem cell factor (SCF), and FLT-3-L were purchased from PeproTech.
Isolation of NK cells and DCs
Single-cell suspensions from LNs (inguinal, brachial, axillary, and cervical; mesenteric were also harvested where mentioned) and spleen were prepared by mechanical disruption of the organs. BM cells were obtained by flushing the femurs and tibiae with PBS with 2% FBS. For phenotypic analysis, cells were incubated at 37°C in the presence of 0.5 mg/ml collagenase and 100 μg/ml DNase for 30 min and then inactivated with 5 mM EDTA. In functional studies, the latter step was omitted to prevent cell damage. Cell suspensions were passed through a 22-G needle to reduce aggregates and then filtered through a cell strainer and resuspended in PBS with 2% FBS.
Flow cytometry and cell sorting
Cells were stained with appropriate Abs and subjected to four-way sorting on a FACSAria cell sorter (BD Biosciences). TCRβ+CD4+CD25+ and TCRβ+CD4+CD25− as well as TCRβ−CD19−CD4−CD8α−CD122+NK1.1− and TCRβ−CD19−CD4−CD8α−CD122+NK1.1+CD49b− populations were sorted with >95% purity (data not shown). Cell surface staining was performed according to standard methods (17). Foxp3 intracellular staining was performed using the Foxp3 staining kit (eBioscience) according to manufacturer’s instructions. Analyses were done on a FACSAria flow cytometer using the biexponential axis (18) with the DIVA software.
In vitro culture
For NK cell generation assays, 104 stage A or stages B and C NK cells were cultured in 24-well plates (Costar) containing a confluent layer of irradiated (25 Gy) OP9-GFP cells and OP9 culture medium supplemented with 20 ng/ml IL-15, FLT-3-L, and SCF. Medium was changed every 4 days and cells were transferred on a new layer of irradiated OP9-GFP cells after 6 days. In Treg cell suppression assays, 105 preactivated CD4+CD25− or CD4+CD25+ cells were added to each well on day 0. In the latter case, 105 FACS-sorted lymphocytes (either CD4+CD25− or CD4+CD25+) were preactivated by incubation for 72 h in 96-well plates (Costar) with 10 μg/ml plate-bound purified anti-CD3ε and 100 U/ml IL-2.
In vivo assays
For CD4+ T cell subsets, LN single-cell suspensions were FACS sorted after labeling with anti-TCRβ, -CD4, and -CD25 Abs. Sorted cells were washed, resuspended in RPMI 1640 (Sigma-Aldrich), and injected i.v. by the tail vein into recipient mice (106 cells/mouse).
In vivo lymphocyte depletion.
For NK cell depletion, C57BL/6 and HY-TCR-transgenic Rag2−/− mice were injected with 200 μg of anti-NK1.1 mAb (PK136), or isotype control, i.p. every 4 days. For depletion of CD25+ cells, C57BL/6 mice were injected with 500 μg of anti-CD25 mAb (PC61), or isotype control, i.p. every 3 days.
LNs and spleens were collected and immediately snap frozen in Tissue-Tek (Canemco-Marivac). Frozen sections (5–7 μm) were stained with anti-NKG2D or anti-Foxp3-biotin with background staining with anti-CD3-Alexa Fluor647 or anti-CD3-biotin followed by streptavidin-A568. NKG2D staining was revealed with anti-rat-Alexa 488. For Foxp3 staining, sections were fixed and permeabilized using the Fix/Perm and permeabilization buffers (eBioscience) as stated in the manufacturer’s protocol. Staining with Foxp3-biotin was revealed with secondary streptavidin-Alexa 488. Immunofluorescence confocal microscopy was performed with a Zeiss Axiovert 200M confocal microscope. Individual composite fields of the organs were juxtaposed to reconstitute the image of a whole organ cross-section (using Adobe Photoshop).
The means of normally distributed data were compared using the Student t test, with a p value of <0.05 considered significant. Data are presented as mean ± SD. For analysis of cell distribution by immunohistofluorescence, the χ2 test was used.
LNs are not necessary for NK cell development
To determine whether LNs were required for NK cell development, we studied aly mice that carry a point mutation in the NF-κB-inhibiting kinase and have no LNs nor Peyer’s patches (19). We used the model of Di Santo (3) to define the following NK cell-lineage developmental stages: A (CD122+NK1.1−); B and C (CD122+NK1.1+CD49b−); D (CD122+CD49b+CD11blow); E (CD122+CD11bhighCD43low); and F (CD122+CD43high). Stage A–C cells are considered immature, while stage D–F cells are mature. Analysis of BM, spleen, and blood cells revealed no significant differences in the number of NK lineage cells nor in the representation of the various NK lineage development stages in aly/aly and aly/+ mice (Fig. 1). Because aly is a recessive allele, there were no differences between aly/+ and wild-type mice (data not shown). We conclude that LNs and Peyer’s patches are not necessary for development of murine NK cells. As a corollary, these results demonstrate that lack of NK cells in Id2−/− and Lta−/− mice cannot be ascribed to defective LN development.
LN-resident NK lineage cells are skewed toward immature developmental stages
Presence of cells with the phenotype of NK precursors in LNs (6) raises the question of their functionality, that is, of their ability to generate mature NK progeny. To address this, we determined the proportion of NK lineage cells at various development stages in the LNs, BM, spleen, and blood of C57BL/6 mice (Fig. 2,A). In accordance with previous studies, we found very low numbers of NK lineage cells in the LNs (10, 11, 12). The mean percentage of lymphoid cells expressing NK markers was 0.3% in the LNs, 0.9% in the BM, 3.2% in the spleen, and 6.4% in the blood (Fig. 2 B).
Analyses of NK cell subsets in various hematolymphoid compartments revealed two distinct profiles: one skewed toward immature stages in the BM and LNs and one skewed toward mature stages in the spleen and blood (Fig. 2,C). Peripheral blood and spleen contained mostly (75–85%) mature (stages D–F) NK cells, the vast majority of which were stage F cells. In contrast, LNs contained on average 40% immature (stages A–C) NK cells and only 19% stage F NK cells (Fig. 2 C). The profile of the BM was similar to that of the LNs. Notably, we found the same representation of NK developmental stages in mesenteric, cervical, and inguinal LNs (data not shown).
Poor generation of mature NK cells by LN-derived NK precursor cells
The high proportion of immature NK cells in LNs (Fig. 2,C) could be explained by two mechanisms: preferential attraction of immature as opposed to mature NK cells in the LNs, or defective development of LN-resident immature NK cells. To evaluate their ability to generate NK cells, we isolated NK precursors (stage A) by FACS sorting and cultured them in the presence of IL-15, FLT-3-L, SCF, and irradiated OP9 stromal cells (20). On a per-cell basis, the yield of CD122+NK1.1+ cells after 12 days of in vitro culture was ∼4-fold less for NK precursors from the LNs compared with NK precursors from the spleen or BM (Fig. 3 A). NK precursors from BM and spleen generated similar numbers of NK cells. Thus, the paucity of mature NK cells in LNs correlates with the fact that LN NK precursors display, on a per-cell basis, a sharp reduction in their ability to generate NK cells in vitro compared with NK precursors from the BM and spleen.
Considering that the spleen and LNs are secondary lymphoid organs that contain similar cell subsets, what might explain that LN NK precursors give rise to less NK cells than spleen NK precursors? As Treg cells are potent inhibitors of various aspects of NK cell responses (14, 15), we hypothesized that inhibition of NK precursor development in the LNs might be mediated by CD4+Foxp3+ Treg cells. Of note, there is currently no direct evidence that Treg cells regulate NK cell development or interact with NK precursors. To evaluate our hypothesis, we first asked whether Treg cells could inhibit generation of NK cells by NK precursors in vitro. More specifically, we evaluated whether addition of preactivated LN CD4+CD25+ or CD4+CD25− cells would influence in vitro generation of CD122+NK1.1+ NK cells by NK precursors. FACS-sorted CD4+CD25− and CD4+CD25+ LN cells were preactivated by culture for 3 days in the presence of IL-2 and plate-bound anti-CD3. It is well-established that preactivated Treg cells mediate potent Ag nonspecific suppressive activity in vitro in the absence of TCR re-engagement (21, 22). On day 3, preactivated CD4+ T cells were mixed with FACS-sorted BM NK precursors (stage A) at a 10:1 ratio. The latter ratio was selected because LNs contain 10 times more CD4+CD25+ T cells than NK precursor cells. Cells were cultured in 24-well plates in the presence of IL-2, IL-15, SCF, FLT-3-L, and a confluent layer of irradiated OP9-GFP cells. On day 12, the salient finding was that NK precursors gave rise to three times more CD122+NK1.1+ cells in the presence of preactivated CD4+CD25− cells relative to CD4+CD25+ cells (Fig. 3 B). There was no significant outgrowth of contaminant cell populations that might have been present at trace amounts in plated cells. Indeed, at the term of the culture period, the proportion of cells that were neither TCRαβ+CD4+ nor CD122+ among GFP-negative cells (that is, excluding OP9-GFP cells) was ≤5% (data not shown). These data demonstrate that preactivated CD4+CD25+ cells have a direct inhibitory effect on NK precursors and are sufficient to suppress NK cell development in vitro.
NK cells are located mostly in the T cell zones in the LNs but not the spleen
It is generally viewed that Treg cells interact with other cells through cell-cell contact or short-range interactions to mediate their effects (23, 24, 25). To explain discrepancies (in the number of NK cells and function of NK precursors) between the spleen and LNs, we therefore evaluated the propitiousness of the environments of the LN and the spleen to short-range interactions between Treg cells and NK lineage cells. Cytometric analyses revealed a modest (1.2- to 1.4-fold) but significant (p < 0.02) increase in the proportion of CD4+ T cells that were Foxp3+ or CD25+ in the LNs relative to the spleen (Fig. 3,C). Several studies suggest that CD103 may identify a particularly potent subpopulation of CD4+CD25+ Treg cells (26, 27). Interestingly, we found that the proportion of CD103+ cells among Foxp3+CD4+ cells was four times greater in LNs (20–23%) than the spleen (5–6%) (Fig. 3 D).
The most salient findings emerged when we sought to determine whether Treg cells and NK cells reside in similar microenvironments in the LN and spleen. We stained LN and spleen sections with anti-NKG2D or anti-Foxp3 and used anti-CD3 as background staining. This allowed us to identify T cell-rich zones and to exclude NKT cells from our analyses. Foxp3 is the most reliable marker of natural Treg cells, while NKG2D is present on stages B-F NK cells (3) and ∼50% of stage A NK cells (data not shown). In agreement with other studies (9, 10), LN NK cells (NKG2D+CD3−) localized primarily in the paracortex (T cell rich) zone (Fig. 4, B and G) whereas, in the spleen, the majority of NK cells were found in the red pulp (Fig. 4, D and G). As expected (28), Foxp3+ cells were found in T cell zones of the LNs (Fig. 4, A and G) and spleen (Fig. 4, C and G). To determine whether NK lineage cells found in the LN paracortex included NK precursors, we assessed the distribution of NKG2D+CD3− cells in the LNs of mice injected with anti-NK1.1 Ab. Because NK precursors (stage A) are NK1.1−CD3− and about half of them express NKG2D (data not shown), we reasoned that, in theory, the sole NKG2D+CD3− cells present in NK1.1-depleted mice should be NKG2D+ NK precursors. There were very rare NKG2D+ cells in the secondary lymphoid organs of NK1.1-depleted mice. Nonetheless, we were able to positively identify 25 NKG2D+ cells in 21 LN sections and 61 NKG2D+ cells in 10 spleen sections. The key finding was that in NK1.1-depleted mice, most LN NKG2D+ cells were in the T cell-rich areas, whereas most spleen NKG2D+ cells were outside T cell-rich areas (data not shown). These data suggest that in secondary lymphoid organs, the distribution of NK precursors inside compared with outside T cell-rich areas is similar to that of more mature NK1.1+ cells (Fig. 4 G). Thus, in the LNs but not the spleen, most NK precursors colocalize with Foxp3+ Treg cells in T cell-rich areas. Though depletion with anti-NK1.1 is very effective (≥98% in our hands), our conclusion on the localization of NK precursors must nevertheless remain tentative because of two caveats: we cannot formally exclude that residual stage B–F NK cells may have persisted in NK1.1-depleted mice (e.g., by down-regulating NK1.1 expression) and we could only identify 50% of NK precursors (those that were NKG2D+).
Treg cells regulate the number of mature NK cells in LNs
We next sought to determine whether lack of Treg cells would alter NK cell development in the LNs. We reasoned that a forthright approach to this question would require studies on mice that are characterized by absence of Treg cells, no chronic inflammation and LNs containing well-developed T cell zones. This excluded Scurfy mice (Foxp3−/−) that die of autoimmune disease around 3 wk of age (29), as well as non-TCR-transgenic Rag2−/− mice that are devoid of T cell zones. In contrast, our criteria were met in HY-TCR-transgenic Rag2−/− mice. As expected, we found that Treg cells were absent from HY-TCR-transgenic LNs (Fig. 4, E and G). The salient finding was that the absolute number of mature (stage E and F) NK cells was increased 2-fold in LNs of HY-TCR-transgenic mice relative to C57BL/6 mice (p < 0.01) (Figs. 4,F and 5,A). Moreover, representation of the various NK developmental stages in LNs of HY-TCR-transgenic mice reproduced that found in the spleen of wild-type mice (Figs. 2,C and 5,B). Notably, mature NK cells in HY-TCR-transgenic LNs accumulated almost exclusively in the T cell-rich paracortex (Fig. 4,F). Rag2−/− mice lack B cells. However, the proportion of mature NK cells was similar in LNs from B cell-deficient and wild-type mice (Fig. 6 A). Thus, the increased proportion of mature NK cells in LNs of HY-TCR-transgenic Rag2−/− mice cannot be ascribed to lack of B cells.
In wild-type mice, low numbers of NK cells in the LNs (Fig. 2,B) correlated with low yields of NK cells upon in vitro culture of LN-derived NK precursors (Fig. 3,A). We therefore asked whether NK precursors would display a different in vitro behavior when they were obtained from Treg-deficient LNs compared with wild-type LNs. FACS-sorted stage A NK precursors from LNs and spleen of C57BL/6 and HY-TCR-transgenic Rag2−/− mice were cultured in vitro under the same conditions as in Fig. 3,A. At the term of in vitro culture, the key finding was that in contrast with C57BL/6 NK precursors, precursors from HY-TCR-transgenic mice generated at least as many NK1.1+ cells as spleen-derived NK precursors (Fig. 5 C). Hence, in the Treg-deficient TCR-transgenic mouse, the spleen and LNs harbored similar percentages of NK cell subsets, and LN-resident NK precursors were as effective as spleen NK precursors in generating NK cells in vitro.
To further evaluate the effect of Treg cells on LN NK cells, we depleted/neutralized Treg cells by i.p. injection of 500 μg of anti-CD25 mAb (PC61) every 3 days to wild-type mice (Fig. 5,D). In preliminary experiments, we had found that lower doses of PC61 Ab did not induce a significant depletion of Foxp3+ cells and had no effect on NK cells (data not shown). After 12 days, we observed a 60% decrease in the percentage of Foxp3+ cells in LNs and a 27% increase in the percentage of mature NK cells. We next asked whether Treg cell addition would decrease the frequency of mature NK cells in LNs. To this end, HY-TCR-transgenic Rag2−/− mice were injected either with 106 CD4+CD25− cells or with 0.75 × 106 CD4+CD25− cells plus 0.25 × 106 CD4+CD25+. No mice were injected with 100% CD4+CD25+ cells because Treg cells need IL-2-producing CD4+CD25− T cells to survive and expand following adoptive transfer (Ref. 30 and our unpublished observations). On day 12 postinjection, we found a 20% decrease in mature NK cells in the LNs of mice that received CD4+CD25+ compared with those that received only CD4+CD25− cells (Fig. 5 E). In contrast, we observed no differences in the percentage of mature NK cells in the spleen (data not shown).
We conclude that CD4+CD25+ Treg cells are necessary and sufficient to inhibit accumulation of mature NK cells in the LN paracortex. In addition, CD4+CD25+ T cells directly suppressed generation of NK cells by NK precursors in vitro (Fig. 3 B). This suggests that Treg cells repress NK cell accumulation in the LN by hampering in situ maturation of NK precursor. We cannot rule out that Treg cells also inhibit entry of mature NK cells in the LNs, but we are at a loss to explain how such an interaction might be mediated.
Negative correlation between proportion of mature NK cells and of CD8α+ iDCs in the LNs
During the course of infection, NK cells have pleiotropic effects on innate and adaptive immune responses (31, 32). However, there is no evidence that they play any role under steady-state conditions. Therefore, repression of NK development by Treg cells in the LNs raises one prime question: what could be the biological role of that repression in the “resting” LN (i.e., in the absence of infection)? Why would it be advantageous to minimize the number of mature NK cells in the LNs under steady-state conditions? In the absence of infection, LNs serve two main functions: they support peripheral homeostatic T cell proliferation and they maintain tolerance to peripheral self Ags (33, 34). DCs play a central role in both cases (35, 36, 37) and emerging evidence shows that they are engaged in a complex cross-talk with NK cells. These interactions can lead either to maturation and activation of iDCs or to TRAIL-dependent iDC killing (38, 39, 40, 41, 42, 43, 44, 45).
We therefore hypothesized that Treg cell-mediated repression of NK cell development in the LN paracortex might impinge on the survival or maturation of iDCs. To test this assumption, we first compared the percentage of DC (CD11chighI-Ab low/high) subsets in LNs of mice with a normal (C57BL/6) vs high (HY-TCR-transgenic Rag2−/−) proportion of mature NK cells (Fig. 7,A). We found that LNs of HY-TCR-transgenic mice showed a significant decrease in the percentage of iDCs (CD86lowI-Ab low) but not of phenotypically mature DCs (CD86highI-Ab high). The decrease affected specifically iDCs expressing CD8α (Fig. 7,B). Of note, depletion of CD8α+ iDCs in LNs of HY-TCR-transgenic mice cannot be ascribed to lack of B cells because B cell-deficient mice showed normal proportions of mature and iDCs (Fig. 6,B). We further analyzed the correlation between LN contents of mature NK cells and CD8α+ iDCs by adding two other mouse models: 1) P14-TCR-transgenic Rag2−/− mice that are deficient in Treg cells; and 2) Foxp3-transgenic mice that have LNs containing increased numbers of Treg cells (46). Altogether, there was a very strong negative correlation (R = −0.96) between LN contents of mature NK cells and CD8α+ iDCs in the four mouse models tested (Fig. 7 C).
Depletion of NK cells increases the proportion of CD8α+ iDCs in LNs
From the above data, it cannot be inferred whether the number of LN CD8α+ iDCs is regulated by Treg cells and/or by mature NK cells. To elucidate this conundrum, we depleted NK cells by injecting PK136 (anti-NK1.1) Ab to C57BL/6 and HY-TCR-transgenic mice. NK cell depletion was almost total in HY-TCR-transgenic mice and C57BL/6 mice (Fig. 8, B and D), and in both cases significantly increased the proportion of CD8+ iDCs in LNs (Fig. 8, A and C). We conclude that depletion of mature NK cells is sufficient, and consequently that Treg cells per se are not necessary to augment the proportion of CD8+ iDCs in LNs.
Addition or depletion of Treg cells modulates the proportion of CD8α+ iDCs in LNs
All these data support an emerging model in which Treg cells inhibit NK cell development in the LNs and thereby promote accumulation of CD8α+ iDCs. To further substantiate this model, we assessed whether depletion or addition of Treg cells would affect LN CD8α+ iDCs. First, we found that in vivo depletion/neutralization of Treg cells with PC61 Ab, shown to increase the proportion of mature NK cells in LNs (Fig. 5,D), resulted in a 40% decrease in the percentage of CD8α+ iDCs (Fig. 8,E). Of note, mice treated with PC61 Ab showed no signs of inflammation or disease at the time of sacrifice. Second, we found a significant increase in the percentage of LN CD8α+ iDCs in LNs of Treg-deficient mice (HY-TCR-transgenic Rag2−/−) that received CD4+CD25+ cells (Fig. 8,F). Thus, addition of CD4+CD25+ cells to a mouse deficient in Treg cells is sufficient to decrease the proportion of mature NK cells (Fig. 5,E) and thereby increase the proportion of CD8α+ iDCs in the LNs (Fig. 8 F).
Our study highlights unexpected discrepancies in NK cell subsets present in secondary lymphoid organs. The spleen contains numerous NK cells whose differentiation profile is similar to that of blood NK cells and is characterized by a preponderance of mature elements. The vast majority of spleen NK cells are located in the red pulp. In contrast, LNs contain few NK cells, most of which are immature and located in the T cell zone. The paucity of LN NK cells correlates with the fact that LN NK precursors display a defective ability to generate NK cells in vitro compared with NK precursors from the BM and spleen. In the absence of Treg cells, we observed an increase both in the number of mature NK cells in the LN paracortex and in the progenitor activity of LN NK precursors. Addition of Treg cells at a 10:1 ratio inhibited in vitro generation of NK cells by NK precursors. Furthermore, adoptive transfer of Treg cells to Treg-deficient mice decreased the proportion of mature NK cells in the LNs. We conclude that Treg cells are necessary and sufficient to inhibit accumulation of mature NK cells in the LN paracortex. Our conclusion is coherent with a recent study showing that diphtheria toxin-induced ablation of Foxp3+ cells in Foxp3DTR mice leads to accumulation of NK cells in LNs (47). Our data strongly suggest that blockade of NK precursor development in situ is instrumental in this process. We cannot formally exclude that Treg cells also regulate entry of mature blood NK cells in the LN. Caution is advised when extrapolating data on mouse NK cell development to humans because different markers have been used to study NK cell ontogeny in these two species. Nevertheless, it is remarkable that, relative to blood and spleen NK cell populations, NK cells in human LNs also represent a much lower proportion of lymphoid cells and are mostly noncytolytic cells with an immature phenotype (6, 48).
Why do CD4+CD25+ Treg cells repress development of NK precursors in the LN but not the spleen? Our data suggest that location is the key explanation. In the LN, NK cells in general and NK precursors in particular are primarily found in T cell zones, where most CD4+CD25+ Treg cells are located (Fig. 4 and data not shown). In the absence of Treg cells, accumulation of mature NK cells was found specifically in LN T cell zones (Fig. 4). Conversely, most spleen NK cells are found in the red pulp, that is, outside T cell zones. In addition, LN Treg cells might be more potent than spleen Treg cells. Indirect support for that assumption is that the proportion of CD103+ cells among Foxp3+CD4+ was higher in LNs than spleen (Fig. 3 D), and that CD103 expression correlates with Treg cell-suppressive activity in some models (26, 27).
Although CD4+CD25+ Treg cells are known to inhibit in vitro proliferation of CD4+ T cells (49) and CD8+ T cells (50), our report is the first to demonstrate that the presence of Treg cells has a negative impact on the development of NK cell precursors in LNs. These data further reinforce the concept that functional interactions between Treg cells and NK cells are multifaceted. Indeed, CD4+CD25+ Treg cells have been shown to have many effects on mature NK cells: inhibition of NK cell cytotoxicity, proliferation, IFN-γ secretion, and NKG2D expression (14, 15, 51). In addition, TGF-β and IL-10, two important cytokines produced by Treg cells, were shown to suppress NK cell activation (52), proliferation and homeostasis (53), cytolytic activity and IFN-γ secretion (54), and to reduce expression of NKp30 and NKG2D both of which are required for in vitro killing of DCs by NK cells (55). One issue to be addressed in further studies is how Treg cells may hinder development of NK precursors. Because several Treg-cell effects are mediated by TGF-β, one clue may be that NK precursors express particularly high levels of TGF-βR (M. Giroux, unpublished observations).
In view of the early recruitment of NK cells to LNs during the course of infection (11), it has been postulated that elimination of iDCs by NK cells ensures maximal activation of adaptive immunity by allowing only mature DCs to be involved in Ag presentation (43, 56, 57). However, in the absence of infection, LN iDCs are important for induction and maintenance of T cell tolerance to peripheral Ags (36, 58, 59). Supporting this view, selective depletion of iDCs (and a reciprocal increase in the number of mature DCs) is a hallmark of a classical human autoimmune disease: systemic lupus erythematosus (60, 61). We found that in LNs of wild-type mice and three models of transgenic mice, the presence of mature NK cells negatively correlated with the proportion of CD8α+ iDCs. CD8α+ iDCs represented 1 in 3 DCs in the wild-type LN but only 1 in 12 DCs in the LNs of TCR-transgenic mice (whose LNs are enriched in mature NK cells). Moreover, the proportion of LN CD8α+ iDCs increased following depletion of NK1.1+ cells or after adoptive transfer of CD4+CD25+ T cells but decreased after depletion of CD25+ cells. We therefore deduce that, at least in the LNs, mature NK cells selectively reduce the frequency of CD8α+ iDCs. Because Treg cells can prevent DC maturation in vitro (62), we cannot exclude that Treg cells might also have a direct NK-independent effect on LN DC subsets in vivo. Nevertheless, our in vivo data are the first to show that depletion of mature NK cells is sufficient to increase the proportion of CD8α+ iDCs. Interestingly, a substantial body of evidence suggests that CD8α+ iDCs are particularly important for induction of peripheral self tolerance (63, 64). At least two factors may explain the key role of CD8α+ DCs in peripheral tolerance: their production of indoleamine 2,3-dioxygenase (63, 65) and their unique ability to cross-present tissue-specific Ags to T cells in vivo (66, 67, 68). We therefore propose that NK-mediated decrease of CD8α+ iDCs in LNs of uninfected animals would perturb peripheral T cell tolerance.
How do mature NK cells regulate the frequency of LN CD8α+ iDCs? In vivo, prolonged cell-to-cell contacts between NK cells and DCs have been shown to occur in the paracortical T cell-rich zone of the LNs (9, 69, 70). According to in vitro studies, the outcome of NK/iDC interactions is either DC maturation or demise. At a low NK:DC ratio (1:5), NK-DC interactions stimulate DC-mediated cytokine secretion (IL-12 and TNF-α) and DC maturation. But, the converse NK-DC ratio (5:1) will result in killing of iDCs (38) through TRAIL-mediated apoptosis (45). iDC decrease, either by elimination or maturation, would increase Ag-presentation efficiency by reducing nonresponsive DCs to the ongoing inflammation and thereby freeing up resources for reactive, maturing DCs. We therefore surmise that by abrogating generation of mature NK cells in the LN paracortex, Treg cells support persistence of immature CD8+ iDCs. As a corollary, absence of Treg cells would lead to accumulation of mature NK cells and thereby beget either maturation or apoptosis of CD8α+ iDCs. The respective importance of CD8α+ iDC maturation and killing has to be further explored. To the best of our knowledge, we are the first to provide evidence that CD8α+ iDCs may be more susceptible to NK-induced maturation or killing than CD8α− iDCs. The reason for the differential susceptibility of iDC subsets to NK cells is not inherently obvious because hundreds of genes are differentially expressed in CD8α+ relative to CD8α− DCs (71). It is nevertheless logical to assume that three classes of genes may be of prime relevance here: MHC Ia and Ib molecules; ligands recognized by activating NKRs; and intracellular mediators of apoptosis. Expression of those gene families in iDC subsets needs to be assessed in future studies.
Together, data presented herein suggest that Treg cells tip the balance toward tolerance in the quiescent LN by inhibiting NK cell development in the LN paracortex and thereby protecting CD8α+ iDC presence (Fig. 9). In accordance with this idea, it has been reported that CD62L+ but not CD62L−CD4+CD25+ Treg cells efficiently protect against lethal graft-vs-host-disease, graft rejection, and diabetes (72, 73, 74, 75). Moreover, the population of CD4+CD25+ Treg cells that can delay diabetes in prediabetic NOD mice expresses high levels of CD62L and CCR7, and accumulates in draining LNs but not the spleen (74, 75). The fact that expression of CD62L and CCR7 characterizes LN-homing cells is coherent with the concept that LN-homing Treg cells have a critical role in tolerance induction. In line with this idea, LN (but not spleen) occupancy by Treg cells is mandatory for allograft acceptance (76).
Our studies have been done in uninfected mice housed in a specific pathogen-free environment. During infection, high numbers of NK cells accumulate in the LN paracortex (9). Further studies are required to evaluate to what extent accumulation of NK cells during infection is due to in situ maturation of NK precursors and/or to importation of mature NK cells in LNs. Accumulation of NK cells in LNs is advantageous to the infected host because NK cells can kill infected cells, produce IFN-γ and help to launch adaptive T cell responses. We speculate that accumulation of NK cells in LNs may be empowered by the transient inactivation of Treg cells induced by TLR signaling during the course of infection (77, 78, 79, 80). In this way, interactions between Treg cells and innate immune cells (NK cells and DCs) would determine whether LNs are geared toward inducting tolerance or immunity.
We are grateful to Dr. Sylvie Lesage for sound advice on analysis of DC subsets, to the personnel of the Institute of Research in Immunology and Cancer animal facility for skilled support, and to J. A. Kashul for editorial assistance.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants from the Canadian Institutes for Health Research (CIHR) to C.A.P. (MOP 67211) and C.P. (MOP 42384). M.G. and J.S.-P. were supported by the Cole Foundation and the CIHR, respectively. E.Y. is a recipient of a fellowship from the CIHR Training Grant in Neuroinflammation. C.A.P. and C.P. hold Canada Research Chairs in “Regulatory Lymphocytes of the Immune System” and “Immunobiology,” respectively.
Abbreviations used in this paper: BM, bone marrow; DC, dendritic cell; iDC, immature DC; LN, lymph node; Treg, regulatory T; SCF, stem cell factor.