Cancer-Induced Immunosuppression: IL-18 – Elicited Immunoablative NK Cells

, Abstract During cancer development, a number of regulatory cell subsets and immunosuppressive cytokines subvert adaptive immune responses. Although it has been shown that tumor-derived interleukin (IL)-18 participates in the PD-1 – dependent tumor progression in NK cell – controlled cancers, the mechanistic cues underlying this immunosuppression remain unknown. Here, we show that IL-18 converts a subset of Kit (cid:1) (CD11b (cid:1) ) into Kit þ natural killer (NK) cells, which accumulate in all lymphoid organs of tumor bearers and mediate immunoablative functions. Kit þ NK cells overexpressed B7-H1/PD-L1, a ligand for PD-1. The adoptive transfer of Kit þ NK cells promoted tumor growth in two pulmonary metastases tumor models and signi ﬁ cantly reduced the dendritic and NK cell pools residing in lymphoid organs in a B7-H1 – dependent manner. Neutralization of IL-18 by RNA interference in tumors or systemically by IL-18 – binding protein dramatically reduced the accumulation of Kit þ CD11b (cid:1) NK cells in tumor bearers. Together, our ﬁ ndings show that IL-18 produced by tumor cells elicits Kit þ CD11b (cid:1) NK cells endowed with B7-H1 – dependent immunoablative functions in mice. Cancer Res; 72(10); 1 – 11. (cid:1) 2012 AACR.


Introduction
Tumor progression subverts the adaptive arm of antitumor immune responses, either directly by compromising T-cell functions or indirectly by inhibiting antigen-presenting cells.
Tumor-infiltrating dendritic cells (DC)/myeloid suppressor cells tend to mediate metabolic immunosuppressive effects and promote the generation and local accumulation of suppressive regulatory T cells or Tr1 CD4 þ T cells.Tumor-associated regulatory T cells efficiently inhibit CD4 þ T-cell-dependent immune responses and suppress the proliferation and IFNg production of antigen-experienced CD8 þ T cells (1,2).Interleukin (IL)-10-producing natural killer (NK) cells can also dampen IL-12-dependent immune responses during systemic infections (3).Nonetheless, the clinical relevance and the molecular mechanisms underlying the concept of regulatory NK cells have remained largely unclear (4).
Driven by clinical observations associating high serum levels of IL-18 and NK cell defects (5,6) as well as the facilitating role of tumor-derived IL-18 in melanoma metastases (7,8), we investigated the potential immunosuppressive effects of IL-18 in NK cell-controlled tumors (9,10).We recently reported that low dosing of exogenous IL-18 or tumor-derived IL-18 promoted the dissemination of B16F10 and CT26 metastases in a PD-1-dependent manner (11).However, the precise mechanisms accounting for the IL-18/PD-1 immunosuppressive axis in these NK cell-controlled tumor models were not studied in this report.Now, we show that IL-18 secreted by tumor cells could stimulates tumor surveillance mediated by Kit À NK cells by preventing the accumulation of Kit þ CD11b À immunoablative NK cells.

Materials and Methods
Mice and tumors Female C57Bl/6 or BALB/c mice were obtained from Charles River Laboratories, and the Centre d'Elevage Janvier (Le Genest-St-Isle, France).C57Bl/6 nude mice were purchased from Taconic.Myeloid differentiation primary response protein (MyD88) À/À and IL-18 receptor (IL-18R)a À/À mice backcrossed on a C57Bl/6 background were kindly provided by B. Ryffel (CDTA, Orl eans, France).W/Wsh mice were kindly provided by S. M echeri (Institut Pasteur, Paris, France).RET þ/À mice were obtained from A. Pr evost-Blondel (Institut Cochin, Paris, France) with the permission of Prof. M. Kato (Chubu University, Aichi, Japan).All animals were maintained in Institut Gustave Roussy (Villejuif, France) animal facilities according to the Animal Experimental Ethics Committee Guidelines.B16F10 and CT26 tumor cells were obtained from American Type Culture Collection.

In vivo IL-18 neutralization
IL-18-binding protein (IL-18BP) was kindly provided by Prof. C. Dinarello (University of Colorado Denver, Denver, CO; ref. 13).At day 0, 3 Â 10 5 B16F10 or CT26 tumor cells were injected i.v.into C57Bl/6 or BALB/c mice.Mice received 20 mg of IL-18BP or saline buffer alone i.p. daily from days 0 to 8. The quantification of lymph nodes and splenic Kit þ NK cells and pulmonary metastases number was conducted on day 9.

Microarray procedure using Illumina whole genome arrays
RNA was isolated from purified Kit þ NK and Kit À NK cells as described (14).Biotin-labeled cRNA preparation was conducted using the Illumina TotalPrep RNA Amplification Kit (Applied Biosystems).cRNA (1.5 mg) was hybridized to the Illumina Sentrix BeadChip Mouse-6 V1 Array (Illumina) and scanned with an Illumina BeadStation 500Â.For data collection, we used Illumina BeadStudio 3.1.1.0software.Quantile normalization, statistical and bioinformatics analysis was conducted using the IlluminaGUI R-package (15).

MCMV infection
The Ly49H þ congenic strain BALB.B6-CT8 (H-2 d , I-A d , NK1.1 þ , Ly49H þ ) and the congenic Ly49H À strain BALB.B6-CT6 (H-2 d , I-A d , NK1.1 þ , Ly49H À ; ref. 16) were bred in the Animal Services Facility of the University of Western Australia (Nedlands, Western Australia, Australia).All mice were maintained in this facility during the course of infection.All animal experimentation was carried out with the approval of the Animal Ethics and Experimentation Committee of the University of Western Australia, according to the guidelines of the National Health and Medical Research Council.Mice were infected by intraperitoneal administration of 5 Â 10 3 plaqueforming units (pfu) of salivary gland propagated MCMV-K181-Perth diluted in PBS-0.05%fetal calf serum (FCS).Control mice received PBS/0.05%FCS.Single-cell suspensions were prepared from spleen and cervical lymph nodes and cells preincubated on ice for 30 minutes with EDTA-SS-FCS containing 20% normal goat serum before staining with the indicated antibodies.

Statistical analyses
Results are expressed as means AE SEM or as ranges when appropriate.Aberrant values were excluded using Dixon test.Two groups were compared by the Mann-Whitney test.When the variables studied were not normally distributed, nonparametric statistical methods were used.When 3 or more groups were compared, the Kruskal-Wallis test was used.We evaluated statistical significance with Prism software (GraphPad Software, Inc.).P values <0.05 were considered as significant.

Results
Exogenous or tumor-derived IL-18 facilitates the accumulation of immature NK cells in vivo We compared different schedules of recombinant mouse endotoxin-free IL-18 (rIL-18) administration for their ability to interfere against B16F10 lung metastases.rIL-18 boosted the establishment of B16F10 metastases when it was injected twice a week (11), whereas daily administration of rIL-18 reduced tumorigenesis (11,17).Because B16F10 lung metastases are controlled by NK cells, we analyzed the phenotypic profile of immature and mature NK cell subsets in various lymphoid organs as defined previously (18).Biweekly administration of IL-18 augmented the proportions of Kit (CD117)-expressing cells among all CD3 À NK1.1 þ NK cells in all lymphoid organs (Fig. 1A and B, top) and also in lungs (Fig. 1C) as well as their absolute cell numbers (Fig. 1A-C, bottom) in both naive (Fig. 1) and tumor-bearing mice (not shown).It is of note that the IL-18-induced expansion of Kit þ NK cells in lymphoid organs was inversely correlated with intratumoral Kit À mature NK cells (11).
Because bioactive IL-18 is released from growing tumors and recirculates in the serum at low concentrations (11), we carried out the phenotypic characterization of NK cells from tumorbearing mice, which revealed a consistent expansion of Kitexpressing NK cells (Kit þ NK) in lymphoid tissues.This phenomenon was observed in different genetic backgrounds (C57Bl/6 in Fig. 1D and E, and BALB/c in Fig. 1G), in 3 different tumor models (B16F10 melanoma in Fig. 1D-F) in spontaneous melanomas arising in transgenic mice that express the RET proto-oncogene driven by the metallothionein promoter (ref.11; Supplementary Fig. S1A and S1B), in BCR/ABL-induced chronic myeloid leukemia in Supplementary Fig. S1C, in CT26 colon carcinoma in Fig. 1G and Supplementary Fig. S1D, which were inoculated either intravenously (Fig. 1D, E, and G) or subcutaneously (Supplementary Fig. S1E), while applying different phenotypic definitions for NK cells (CD3 À NK1.1 þ cells in C57Bl/6 mice and CD3 À CD49b þ cells in BALB/c mice).The absolute numbers of Kit þ NK cells increased by approximately 2-fold during tumor progression.This increase concerned primary and secondary lymphoid tissues with an early increase in draining or nondraining lymph nodes at days 6 to 9 after intravenous injection of tumor cells (Fig. 1D) or in draining lymph nodes in subcutaneous tumor models (Supplementary Fig. S1E).The distribution of Kit þ NK cells among distinct mouse NK cell subsets defined by the differential expression of CD27 and CD11b (18) revealed that approximately 70% of the Kit þ NK cells accumulating in tumor-draining lymph nodes were CD27 þ CD11b À (Fig. 1F).Indeed, the establishment of CT26 lung metastases was accompanied by an accumulation of Kit þ CD11b À NK cells in the spleen (Fig. 1G) and in the peripheral lymph nodes (data not shown), a phenomenon also observed in naive mice treated with a biweekly schedule of rIL-18 (Supplementary Fig. S1G).
To establish a causal link between IL-18 released by tumor cells and lymphoid organ-resident Kit þ CD11b À NK cells, we carried out the transfection of B16F10 tumor cells with 2 distinct IL-18 siRNAs, which reduced the capacity of tumor cells to produce bioactive IL-18 as previously described (11).Upon inoculation into mice, B16F10 engineered to express low levels of IL-18 failed to expand Kit þ NK cells (Fig. 2A) and compromised B16F10 tumorigenesis (11).To further substantiate the decisive contribution of IL-18 to the tumor-driven differentiation of Kit þ NK cells in vivo, we used saturating amounts of IL-18BP, a naturally occurring IL-18 antagonist (13), that successfully abolished the development of Kit þ NK cells in lymph nodes and spleens in both B16F10 and CT26 lung metastases models (Fig. 2B and C and not shown) as well as compromised tumor progression (11).
We recently reported that the numbers of B16F10 metastases were decreased in IL-18R À/À or MyD88 À/À mice, suggesting that B16F10 progression was dependent on the IL-18R/ MyD88 signaling pathway of the host (11).When B16F10 tumor cells were injected into IL-18R À/À or MyD88 À/À mice, they failed to increase the number of Kit þ NK cells in lymph nodes (Fig. 2D) and spleens (not shown).
We further aimed at investigating the mechanism of Kit þ NK cell expansion.In vitro, purified Kit À NK cells from tumor-free mice acquired surface expression of c-Kit within 24 hours after stimulation with recombinant mouse IL-18 (rIL-18), but not with stem cell factor (SCF), rIL-2, or rIL-12 (Fig. 2E and not shown).IL-18 stimulated the transcription of c-kit in bulk cultures of NK cells (not shown).As we failed to detect a significant increase in bromodeoxyuridine (BrdUrd) incorporation into the expanding Kit þ NK cell subset upon administration of IL-18 (Supplementary Fig. S2), these cells likewise arise from differentiation rather than proliferation.Indeed, IL-18 promoted predominantly the differentiation of the immature Kit À CD27 þ NK cell subset into Kit þ NK cells (Fig. 2F).
It is noteworthy that SCF, the ligand for the KIT receptor, did not synergize with IL-18 for the ex vivo conversion of Kit À into Kit þ NK cells (Supplementary Fig. S3A) and was not required for the IL-18-mediated differentiation of Kit þ NK cells in vivo because IL-18 can induce Kit þ NK cell differentiation in Kitdeficient (W/Wsh) mice (Supplementary Fig. S3B).Kit signaling did not appear to be involved in survival of Kit þ NK cells in tumor-bearing mice because imatinib mesylate, a tyrosine kinase inhibitor that targets Kit, did not modulate the antiapoptotic protein Bcl2 (Supplementary Fig. S3C).
In conclusion, the IL-18/IL-18R/MyD88 axis is involved in the expansion of Kit þ -expressing NK cells associated with tumor progression.

Kit þ CD11b À NK cells mediate immunosuppressive functions in tumor settings
To determine whether Kit þ NK cells might stimulate the metastatic dissemination of NK-controlled tumors, we comparatively assessed the potential of Kit þ (mostly composed of the CD11b À fraction; Fig. 1F) compared with Kit À NK cells to influence lung metastases upon adoptive transfer into C57Bl/6 mice injected with B16F10 melanoma cells (Fig. 3A, left) or BALB/c mice injected with CT26 colon carcinoma cells (Fig.   3A, right).In both cases, Kit þ (but not Kit À ) NK cells promoted the establishment of lung metastases (Fig. 3A).
Interestingly, the adoptive transfer of Kit þ CD11b À NK cells (but not Kit À NK cells) could eliminate part of the DC pool in vivo, depleting both CD8a þ and CD11b þ spleen-or lungresident DCs (Fig. 3B).Considering the critical role of DCs in maintaining the NK cell homeostasis (19), it was not surprising to observe that upon adoptive transfer into athymic C57Bl/6 nude mice, Kit þ NK lymphocytes also reduced the absolute number of Kit À NK cells both in the spleen (Fig. 3C) and in the lungs (data not shown).Because DCs play a key role in controlling NK cell activation (20), a reduction of the DC/NK cell pool could have functional consequences during stimulation with Toll-like receptor (TLR)9 agonists.We monitored the effects of a systemic adoptive transfer of Kit þ NK cells in lymph nodes, draining a CpG ODN-driven inflammation.Kit þ (but not Kit À ) NK cells induced a significant decrease in the proportion of IFNg-producing NK cells in the lymph node draining the TLR9L-induced inflammation (Fig. 3D).
Altogether, Kit þ CD11b À NK cells are endowed with immunosuppressive functions.
Kit þ CD11b À NK cells mediate their immunoablative functions in a B7-H1-dependent fashion We showed that the adoptive transfer of Kit þ CD11b À NK cells in naive mice was accompanied by a reduction of the DC pool (Fig. 3B).The Kit þ NK-mediated DC depletion appeared to be a direct effect, as Kit þ NK cells could kill immature but not maturing/activated bone marrow-derived DCs in vitro (Fig. 5A and B).Killing of immature DCs was observed only after cocultures of DCs with Kit þ NK cells but not Kit À NK counterparts (AnnexinV þ DC in DC þ Kit À NK cells: 10.7% AE 3.6% vs. AnnexinV þ DC in DC þ Kit þ NK cells: 34% AE 4.1%, P ¼ 0.02) and was abrogated in the presence of neutralizing anti-B7-H1 but not anti-Lag3 antibodies (Fig. 5C).In line with these observations, the contraction of DCs induced by Kit þ NK cells (Fig. 3B) was abolished by injection of an anti-PD-1-blocking antibodies in vivo (Fig. 3C, right), suggesting that B7-H1 present on Kit þ NK cells depleted PD-1-expressing DC, compromising indirectly the DC-mediated mature NK cell homeostasis in lymphoid organs (Fig. 3C, left).Importantly, preincubation of Kit þ NK cells with neutralizing anti-B7-H1 antibody before adoptive transfer into B16F10 tumor-bearing mice abolished the metastasis-enhancing effect of Kit þ NK cells (Fig. 5D).

Kit þ NK cells accumulated not only in the context of a tumor but also in MCMV infection
Accumulation of Kit þ NK cells was also observed during the NK cell-controlled MCMV infection.The frequency and number of Kit þ NK cells were monitored over time after MCMV infection in MCMV-susceptible (BALB.B6-CT6) and -resistant (BALB.B6-CT8) mouse strains (16).Kit þ NK cell accumulation was maximal in MCMV-resistant BALB.B6-CT8 mice, in both the spleens and cervical lymph nodes (Fig. 6).Although Ly49H þ NK cells are essential to limit MCMV during the acute stage of infection, their expansion occurs later during infection and those cells are Kit À (25,26).Like cancer-induced Kit þ NK cells, MCMV-induced Kit þ NK cells also overexpressed Lag3 and 2B4 Quantification of lung metastases was carried out at day 12.Data pooled from 2 experiments with 3 mice per group are shown.The Kruskal-Wallis multiple comparison test was used for statistical analyses.Ã , P < 0.05.NS, not significant.B, transient depletion of spleen resident DC subsets after adoptive transfer of Kit þ CD11b À NK cells in WT mice.About 10 5 purified Kit þ CD11b À or CD11b þ NK and Kit À NK cells were adoptively transferred into C57Bl/6 mice.The absolute numbers of DCs (CD11c þ /I-Ab þ cells; left) and among those the CD8a þ DC subset (middle) or the CD4 þ CD8 a À DC subset (right) residing in spleens or lungs (not shown) were determined at 48 hours by enumerating cells in trypan blue exclusion assays combined to 6 color staining and flow cytometry.The data pooled from 3 independent experiments are shown.Kruskal-Wallis test was used for statistical analyses and significant effects are indicated with asterisks.C, Kit þ NK cells eliminate DCs and Kit À NK cells in vivo in a PD-1-dependent manner.A total of 3 Â 10 5 purified Kit þ or Kit À NK cells was adoptively transferred into nude C57Bl/6 mice.Blocking anti-PD-1 (or isotype control) antibody was injected i.p. as described in Materials and Methods.Absolute numbers of DC (CD11c þ /CMH Class II þ ) and Kit À NK (CD3 À NK1.1 þ CD117 À ) cells were determined in spleen at 48 hours by flow cytometry.The graph shows the data from 2 independent experiments with 2 to 4 mice per group.D, Kit þ NK cells keep in check IFNg-producing NK cells during TLR9L-induced inflammation.A total of 5 Â 10 5 Kit þ or Kit À NK cells were adoptively transferred i.v.into naive C57Bl/6 mice, 6 hours after footpad inoculation of 10 mg of CpG ODN 1668.Mice were sacrificed at 18 hours after NK cell transfer to assess the proportion of IFNg-producing CD3 À NK1.1 þ NK cells in the draining (DLN) and controlateral (CLN) lymph nodes.The data pooled from 3 independent experiments are shown (each dot representing one mouse).ÃÃ , P < 0.01; ÃÃÃ , P < 0.001.Ctl, control; FSC, forward scatter.and were CD11b À (not shown).B7-H1 expression on Kit þ NK cells increased slightly, peaking at day 6 postinfection in the Kit þ fraction of NK cells especially in the lymph nodes (Fig. 6).Therefore, Kit þ NK cells may represent a novel innate lymphoid subset exerting immunoregulatory functions in NK celldependent cancers but also in infections.

Discussion
Although the counter-regulation of the T-cell arm of immunity by suppressor T cells has been widely investigated, the regulatory activity of innate effectors has not been thoroughly explored in the context of tumors (27).NK cells can produce immunosuppressive cytokines that modulate immune responses (28).Even more overlooked is the potential counter-regulation of NK cells by NK cells (29).Here we show for the first time that tumor-derived IL-18 can expand a subset of NK cells that express the c-kit receptor.This NK cell subset circulates in the blood stream and resides in primary and secondary lymphoid organs (Fig. 1, Supplementary Fig. S1 and data not shown) and can deplete the peripheral pool of DCs.Because DCs are involved in the NK cell activation and homeostasis, this mechanism can, at least in part, contribute to NK cell-based immunosuppression and tumor outgrowth.We showed that this NK cell subset regulates innate NK cell functions (as shown in the manuscript in cancer and CpG ODN-mediated inflammation).
The data presented here suggest that an NK cell subset could act as a negative regulator of the DC/NK cell responses (Fig. 3A,  C, and D).Kit þ NK cells represent about 3.3% AE 0.7% of blood, 1.23% AE 0.5% of lung, 4.8% AE 1.4% of spleen, 13.2% AE 2.4% of lymph node, and 5.4% AE 1.8% of bone marrow NK cells in naive  Among IL-18-induced Kit þ NK cells, the median of B7-H1 expression of CD27 þ CD11b À and of CD27 þ CD11b þ is 1,512 and 648, respectively.For statistics in all graphs, ÃÃ , P < 0.01; ÃÃÃ , P < 0.001 according to the Mann-Whitney test.
animals (Fig. 1).In tumor-bearing animals, the percentages and absolute numbers of Kit þ NK cells increased (Fig. 1) to levels that mediate inhibitory effects on the DC/NK cell crosstalk.The adoptive transfer of Kit þ NK cells could promote the outgrowth of B16F10 melanoma and CT26 colon carcinoma cells (Fig. 3A) and reduce CpG-induced NK cell IFNg secretion (Fig. 3D).In the context of a tumor, Kit þ NK cell accumulation is dependent on tumor-derived IL-18 that induces a conversion of Kit À NK cells into Kit þ B7-H1 þ NK cells.Kit þ NK cells did not rely on the c-Kit pathway for their conversion (Fig. 2E; Supplementary Fig. S3A and S3B), B7-H1 expression (Supplementary Fig. S3D), or survival (Supplementary Fig. S3C).These Kit þ NK cells expressed higher levels of B7-H1 than in Kit À NK cells.Saudemont and colleagues described the capacity of Cxcl10 to induce B7-H1 expression on spleen NK cells, which was associated with the NK cell-dependent control of minimal residual disease in leukemia (30).In tumor-or IL-18-dependent experimental settings, Cxcl10 is not upregulated in spleen (11), A and B, direct killing of immature DCs by Kit þ but not Kit À NK cells in vitro.In vitro cocultures of Kit þ NK cells with immature bone marrow-derived DC (derived in granulocyte macrophage colony-stimulating factor), or lipopolysaccharide-or CpG-matured BM DCs at a 1/1 or 5/1 DC/NK ratio were examined in flow cytometry for AnnexinV/propidium iodide (PI) expression.Analysis was conducted on CD11c þ /I-Ab þ cells.Representative dot plots (A) or histograms (B) are shown in each condition of coculture.The experiments have been conducted 3 times with identical results.C, same experimental setting as in A, but adding neutralizing anti-B7-H1 (10 mg/mL) or anti-Lag3 antibodies (10 mg/mL) in the coculture.Percentage of killed DCs was determined by flow cytometry using AnnexinV/7-aminoactinomycin D (AAD) staining.Percentage of DC lysis was calculated as follows: [(% killed DCs in test) À (% killed DC in DC alone)]/(% killed DC in DC alone) Â 100.Means AE SEM of percentages obtained in 3 experiments are indicated.D, B7-H1-mediated prometastatic activity of Kit þ NK cells on the B16F10 melanoma.A total of 3 Â 10 5 Kit þ or Kit À NK cells were preincubated with saturating amounts of neutralizing anti-B7-H1 or isotype (iso) control (ctl) antibody (Ab) and washed before adoptive transfer in B16F10-bearing mice.Ã , P < 0.05 according to the Kruskal-Wallis test.
whereas CXCR3 expression (at the transcriptional and protein levels) is superior in Kit þ than in Kit À NK cells.These observations suggest that the functional outcome of IL-18 compared with Cxcl10 on NK cell triggering may be different.
The identification of the human ortholog of mouse Kit þ CD11b À NK cells and of causal links between IL-18 and the B7-H1/PD-1 pathways in human malignancies remains important challenge of the future development of antibodies blocking immune tolerance checkpoints.

Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.

Figure 1 .Figure 2 .
Figure 1.Low dose of IL-18 and tumor progression promote the accumulation of Kit þ CD11b À NK cells in lymphoid organs.A-C, expansion of Kit þ NK cells during various schedules of rIL-18 administration.C57Bl/6 mice received rIL-18 i.p. (1 mg/inoculation) once on day 1 (1Â), compared with twice on days 1 and 3 (2Â), compared with 5 times (5Â) on days 1 to 5. Percentages and/or absolute numbers of Kit þ NK cells were determined in the lymph nodes (LN; A), the bone marrow (BM; B), and in the lungs (C) at day 7.The graph depicts the results for 6 animals in 2 independent experiments.The Kruskal-Wallis multiple comparison test was used for statistical analyses.ÃÃ , P < 0.01; ÃÃÃ , P < 0.001 as compared with PBS group.D and E, accumulation of Kit þ NK cells in lymphoid organs from C57Bl/6 mice bearing B16F10 metastases (mets).Single-cell suspensions were prepared from LN (D), spleen, bone marrow (BM), and tumorbearing lungs (E) at various time points after i.v.inoculation of B16F10 cells.The percentage of Kit þ NK cells (CD3 À NK1.1 þ Kit/CD117 þ ) among all NK cells was determined.Flow cytometric analyses on LNs are shown in D for 1 representative experiment of 5 yielding similar results (means indicated in the top quadrant).The percentages of Kit þ NK cells are shown in D as mean AE SEM of 5 independent experiments (including >3 mice per group).The Kruskal-Wallis multiple comparison test was used for statistical analyses.Ã , P < 0.05; ÃÃÃ , P < 0.001 as compared with day 0. F, CD27/CD11b expression on Kit þ NK cells.Identical experimental setting as in C. Kit þ NK cells were analyzed for their expression of CD27 and CD11b in lymph nodes of naive or tumor-bearing hosts at day 12. G, Kit þ CD11b À NK cells expand in CT26-bearing BALB/c mice.Proportion and absolute number of Kit þ CD11b À NK cells in spleens from CT26 metastases-bearing BALB/c mice.Data pooled from 3 independent experiments are presented.ÃÃÃ , P < 0.001 according to the Mann-Whitney test.

Figure 4 .
Figure 4. Phenotypic characterization of Kit þ NK cells.A, transcriptional profile of Kit þ versus Kit À NK cells.The heatmap details 28 genes elevated in Kit þ NK cells [fold change (FC) ¼ 2.5, P < 0.05] and 28 genes elevated in Kit À NK cells (FC ¼ 2.5, P < 0.05).Shown are row-wise z-transformed intensity values (mean ¼ 0, SD ¼ 1) with high-intensity values shown in red and low-intensity values shown in green.Representative histograms of fluorescence intensities for B7-H1, CD127, CTLA4, and Lag3 in Kit þ versus Kit À NK cells from naive LNs are depicted.Microarray data have been submitted to GEO database under GSE9431.B, expression of B7-H1 molecules on NK cells.Medians of fluorescence intensities of B7-H1 on NK cells in various locations in naive mice (top) versus tumor bearers (bottom) are depicted as mean AE SEM of 3 independent experiments.C, IL-18 preferentially drives the CD27 þ CD11b À subset of Kit À NK cells into B7-H1 þ NK cells.Same experimental setting as in Fig. 2F but B7-H1 surface expression was analyzed.One representative analysis of 3 is shown.Among IL-18-induced Kit þ NK cells, the median of B7-H1 expression of CD27 þ CD11b À and of CD27 þ CD11b þ is 1,512 and 648, respectively.For statistics in all graphs, ÃÃ , P < 0.01; ÃÃÃ , P < 0.001 according to the Mann-Whitney test.

Figure 5 .
Figure 5. Kit þ NK cells can kill immature DCs through a B7-H1-dependent manner.A and B, direct killing of immature DCs by Kit þ but not Kit À NK cells in vitro.In vitro cocultures of Kit þ NK cells with immature bone marrow-derived DC (derived in granulocyte macrophage colony-stimulating factor), or lipopolysaccharide-or CpG-matured BM DCs at a 1/1 or 5/1 DC/NK ratio were examined in flow cytometry for AnnexinV/propidium iodide (PI) expression.Analysis was conducted on CD11c þ /I-Ab þ cells.Representative dot plots (A) or histograms (B) are shown in each condition of coculture.The experiments have been conducted 3 times with identical results.C, same experimental setting as in A, but adding neutralizing anti-B7-H1 (10 mg/mL) or anti-Lag3 antibodies (10 mg/mL) in the coculture.Percentage of killed DCs was determined by flow cytometry using AnnexinV/7-aminoactinomycin D (AAD) staining.Percentage of DC lysis was calculated as follows: [(% killed DCs in test) À (% killed DC in DC alone)]/(% killed DC in DC alone) Â 100.Means AE SEM of percentages obtained in 3 experiments are indicated.D, B7-H1-mediated prometastatic activity of Kit þ NK cells on the B16F10 melanoma.A total of 3 Â 10 5 Kit þ or Kit À NK cells were preincubated with saturating amounts of neutralizing anti-B7-H1 or isotype (iso) control (ctl) antibody (Ab) and washed before adoptive transfer in B16F10-bearing mice.Ã , P < 0.05 according to the Kruskal-Wallis test.

Figure 6 .
Figure 6.Accumulation of Kit þ NK cells during MCMV infection in resistant mice.A, spleen and cervical (CLN) lymph nodes were collected from MCMVresistant (BALB.B6-CT8) mice infected with 5 Â 10 3 pfu of MCMV-K181 for 0, 2, 6, and 10 days and absolute numbers of NK cells enumerated.The number of c-Kit þ NK cells was also determined.Data from spleen and cLNs are shown.B, numbers of splenic c-Kit þ NK cells within the total NK cell pool as well as the Ly49H À and Ly49H þ NK cell subpopulations were determined at the indicated times after infection.C, NK cells isolated from the spleen and cLNs were collected from uninfected control (ctl) mice or MCMV-infected (day 6 pi) mice and analyzed for c-Kit, PD-1, and B7-H1 expression.Isotype control antibodies were used to determine the fluorescence background, as indicated.Values indicate the percentage of cells in the indicated gates and quadrants.pi, postinjection.