Donor Lymphocyte–Derived Natural Killer Cells Control MHC Class I–Negative Melanoma

Natural killer (NK) cells are cytotoxic innate lymphoid cells (ILC) that kill tumor cells (1). In contrast to cytotoxic T lymphocytes (CTL), NK cells exert their cytotoxic functions independent of MHC-mediated antigen presentation (2, 3), therefore emerging as an alternative to CTLs in eliciting an immune response against tumors (4). Upon encountering MHC-I–negative solid tumors (e.g., melanoma), NK cells can be rapidly and effectively activated, generating cytotoxic activity that contributes to tumor control. This tumor control effect specifically relies on the infiltration and persistence of NK cells in tumor sites, as high NK-cell infiltration correlates with a positive outcome in patients and animal models (5, 6). However, NK-cell loss and defective NK cytotoxic activity correlates with high incidence of tumors (1). Consequently, multiple strategies aimed at the efficacy of NK-dependent tumor control are being employed in the clinic (7, 8).

The adoptive transfer of NK cells that are obtained from either the patient (autologous setting) or a healthy donor (allogenic setting, i.e., HLA-mismatched in human, MHC-mismatched in mice), including those with genetic modification and ex vivo expansion, is an approach that has been used in the clinic (9, 10). The transfer of expanded autologous NK cells is safe, but it has low efficacy in metastatic melanoma patients (UMIN000007527; refs. 11, 12). Conversely, the transfer of allogenic NK cells from HLA-mismatched donors is highly effective against acute myeloid leukemia (AML), but requires a substantial depletion of T cells to avoid graft-versus-host disease (GvHD; NCT00187096; refs. 13–15).

DLI after allo-BMT, including the enrichment of donor NK cells used in DLI, has become a well-established procedure to treat multiple tumors (16, 17). The transferred CD3⁻ NK cells are contained within the DLI, which are allografts and generate antitumor effects (14). The chimeric state of the host after allo-BMT is required for the DLI graft acceptance; however, this leads to GvHD (15). It has never been tested that the safety and feasibility of adoptive transfer of DLI (no prior allo-BMT) against solid tumors (e.g., melanoma), especially the expansion of tumor-infiltrating NK cells.

Here, we employed well-established experimental models of melanoma to study the antitumor efficacy of transferred NK cells after DLI without prior allo-BMT. We found that the transfer of allogenic DLI, but not enriched donor NK cells, lead to the infiltration and persistence of transferred NK cell in a tumor-induced environment, where tumor-elicited regulatory T cells (Treg) and sublethal radiation facilitated NK allograft acceptance. The transferred NK cells became active and mature enough to control melanoma, a process that required the presence of host CD4⁺ T cells producing IL2. Together, these studies will help to define the role of DLI-derived NK cells in controlling melanoma.

B16F0 and B16F10 melanoma cells were originally obtained from the ATCC in 2003 and, were cultured in DMEM (Gibco) supplemented with 10% heat-inactivated FCS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 g/mL streptomycin (Gibco) under standard conditions and were not grown for more than 2 weeks of culture. All cell lines were routinely tested negative for Mycoplasma, but cell line authentication was not routinely performed.

Female C57BL/6J and BALB/cJ mice were purchased from Janvier Labs and 8-week-old mice were used for the experiments. All experiments were performed in accordance with the approval of Institutional Animal Care and Use Committee (IACUC) obtained from KU Leuven (P159-2017).

For the generation of subcutaneous tumors, the flank of C57BL/6J mice was shaved and 5 |\times $| 10⁵ B16F0 or B16F10 cells in 50 μL of PBS were injected subcutaneously (s.c.). In some experiments, B16F10 cells incubated with IFNγ (20 μg/mL) overnight (referred to as IFNγ – B16F10 cells; ref. 18) or IFNγ – B16F10 cells followed by stripping with 50 mmol/L citric acid (pH 3; refs. 19, 20) were used for establishment of a subcutaneous MHC-I–positive or negative melanoma model, respectively. PBS-injected mice served as controls. Tumor growth was monitored twice weekly with a caliper and expressed as volume (mm³) by calculating the (length × width²)/2. Mice whose tumors exceeded 2,000 mm³ were euthanized.

For the induction of lung metastasis, 5 |\times $| 10⁴ B16F10 cells in 100 μL PBS were injected intravenously (i.v.) by tail vein and mice were sacrificed 15 days later. To calculate the metastatic burden, the area occupied by metastatic foci was divided by the total surface area.

Total splenocytes (i.e., DLI) and NK cells were harvested from donor BALB/cJ mice. NK cells from donor spleens were enriched by negative selection using EasySep Mouse NK Cell Isolation Kit (StemCell Technologies) according to the manufacturer's instructions. Recipients were sublethally irradiated with a single dose of 5.0 Gy delivered by an X-ray generator (General Electric). Twenty-two to 24 hours after irradiation, mice were injected intravenously by tail vein with a dose of 5 |\times $| 10⁶ NK cells in 200 μL of PBS from different NK preparations. In mice receiving DLI, 5 |\times $| 10⁷ total splenocytes in 200 μL of PBS from donor were used to contain the same dose of donor NK cells (https://assets.thermofisher.com). In the model of NK depletion, 5 |\times $| 10⁷ DLI inoculum without enriched NK cells were used. In the model of DLI lacking CD4⁺ or CD8⁺ T cells, 5 |\times $| 10⁷ DLI inoculum were used after CD4⁺ or CD8⁺ T-cell depletion by EasySep Mouse CD4⁺ or CD8⁺ T Cell Isolation Kit (StemCell Technologies) according to the manufacturer's instructions, respectively. In mice with mature NK cell transfer, tumor-infiltrating lymphocytes (TIL) in recipients given DLI were isolated as described (21), then NK cells from TILs were enriched by negative selection as described above. In some experiments, either donor NK cells or total splenocytes labeled with 5 μmol/L CFSE (CellTrace CFSE Cell Proliferation Kit, C34554, Thermo Fisher Scientific) were used, respectively. Recipient mice were monitored for survival and GvHD symptoms twice weekly using a published clinical scoring system (22).

Skin, liver, small intestine, and colon tissues were fixed in 4% formaldehyde solution for 24 hours at room temperature and embedded in paraffin. For each organ, 5-μm sections were stained with hematoxylin and eosin (H&E) at VIB-KU LEUVEN Center for Cancer Biology (Belgium). Slides were examined by microscope (Zeiss Axio Scan) and pathology scored using a semiquantitative scoring system (23).

Recipient mice were injected intraperitoneally (i.p.) with 200 μg of depleting antibodies diluted in 100 μL of PBS to CD4 (GK1.5, BioXCell) and CD8 (2.43, BioXCell), respectively. The injection was performed every third day, starting on one day after adoptive transfer of CFSE-labeled DLI (24). Control mice received equal amount of isotope control antibodies (rat IgG1, BioLegend) or equal volumes of PBS.

Recipient mice were injected intraperitoneally with 250 μg anti-CD25 (PC61.5.3, BioXCell) diluted in 100 μL of PBS on 5 and 7 days after B16F10 cells injection for early-stage Treg depletion, or twice weekly, starting on one day after DLI for late-stage Treg depletion, respectively. To deplete Th17 cells, recipient mice were injected intraperitoneally with 100 μg anti-IL17A (clone 17F3; BioXCell) diluted in 100 μL of PBS twice weekly (25), starting on one day after DLI. Control mice received equal amount of isotope control antibodies (rat IgG1, BioLegend) or equal volumes of PBS.

For IL2 blockade, recipient mice were injected intraperitoneally with a mixture of 100 μg anti-IL2 (S4B6; BioXCell) and 100 μg anti-IL2 (JES6-1A12; BioXCell) in 100 μL of PBS 2 days before adoptive transfer of CFSE-labeled NK cells and subsequently every third day (24).

To rescue IL2, 0.25 μg of murine recombinant IL2 (212-12, PeproTech) in 100 μL of PBS was injected intraperitoneally daily between days 10 and 19 (24).

Mouse blood was retro-orbitally collected in heparinized capillary tubes followed by RBC lysis (BioLegend). Freshly collected mouse tumors were cut into 1-mm³ pieces and digested with the mouse tumor dissociation kit in C-tubes (Miltenyi Biotec) according to the manufacturer's instructions. Mouse spleens, lungs, and livers were collected after sacrifice and dissociated in C-tubes (Miltenyi Biotec) by the gentleMACS Dissociator (Miltenyi Biotec). At the end of dissociation, the cells were filtered through 70-μm strainer to obtain the single-cell suspensions. The tumor single-cell suspensions were then isolated with LymphoPrep (StemCell Technologies) to yield TILs. Subsequently, these single-cell suspensions were assessed for viability with Trypan Blue and manually counted, then incubated with Fc receptor blocking solution followed by the flow cytometry staining with the following antibodies: CD25 (PC61), CD4 (GK1.5), CD8 (53-6.7), CD3 (17A2), CD49b (DX5), NKp46 (29A1.4), NK1.1 (PK136), CD11c (N418), CD45 (30-F11), CD11b (M1/70), CD27 (LG.7F9), CD43 (eBioR2/60), KLRG (2F1), CD69 (H1.2F3), MHC-II (M5), H-2Kd (SF1-1.1), and H-2Kb (AF6-88.5; eBioscience). Intracellular staining of Foxp3 (FJK-16s) was performed using the Foxp3 Transcription Factor Staining Buffer Set (eBioscience). Intracellular staining of IL2 (JES6-5H4) was performed using Cytofix/Cytoperm buffer set (BD Biosciences). All procedures were performed according to the manufacturer's instruction. For intracellular staining of Granzyme B (NGZB) and IFNγ (XMG1.2), cells were ex vivo restimulated with phorbol 12-myristate 13-acetate (PMA, 20 ng/mL) and ionomycin (500 ng/mL; Sigma-Aldrich) in a humidified incubator with 5% CO₂ at 37°C for 16 hours. For quantification of the absolute number of cells, a defined number of fluorescent beads (BD Trucount Tubes, BD Biosciences) was added to each sample before acquisition and used as a counting reference. Stained cells were performed with a FACSCanto II cytometer (BD Biosciences) and analyzed with FlowJo software (version 10.5; TreeStar).

CD4⁺ T cells obtained from spleens of naïve C57BL/6 mice were stimulated using PMA (10 ng/mL) and ionomycin (500 ng/mL) for 72 hours and cultured for 1 hour either alone or with an equal number of B16F10. Next, CD4⁺ T cells (2 |\times $| 10⁵ cells/200 μL) were harvested and cocultured with an equal number of NK cells obtained from spleens of naïve BALB/c mice in 96-well U-bottom plates. As a negative control, NK cells alone were cultured in medium. All cells were cultured in RPMI1640 (Gibco) supplemented with 10% heat-inactivated FCS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 g/mL streptomycin (Gibco), and incubated at 37°C in a humidified incubator with 5% CO₂. After 15 hours, maturation or IFNγ production by NK cells were measured.

Data were expressed as mean ± SEM. A two-tailed Student t test was used when only two variables were present in the analysis. Two-way ANOVA with multiple comparisons was used to compare tumor growth over time. Survival curves were estimated using the Kaplan–Meier method. Difference was considered significant for P values < 0.05 (noted in figures as *), P < 0.01 (**), P < 0.001 (***), and P < 0.0001 (****). Analyses were conducted using GraphPad Prism software. The number of repeat experiments and mice per group are indicated in the respective figure legends.

Using the subcutaneous B16F10 melanoma model, we performed an analysis of host versus transferred NK cells after adoptive transfer of MHC-mismatched donor splenic NK cells–either DLI (i.e., total splenic cells) or control NK cells (i.e., enriched NK cells from donor spleens), in mice with or without a sublethal dose of irradiation (Fig. 1A). The effect of irradiation was confirmed by a decrease of lymphocytes in spleens, excluding Tregs in melanoma-bearing mice (Supplementary Fig. S1A and S1B).

We established a flow cytometry staining protocol that allows distinction between the transferred and host NK cells. DX5⁺ from the CD3⁻ population expressed NKp46 (Fig. 1B), validating them as NK cells, but not NKT cells (1). Host NK cells expressed NK1.1, whereas transferred NK cells did not. We focused initial analyses on 4 days after the transfer, before the onset of any donor T-cell–mediated immune control–T-cell reconstitution. Whereas DLI resulted in a slight increase of the transferred NK cells, the addition of irradiation markedly increased the frequency and numbers of transferred NK cells in spleens, tumors, and blood. This increase was not seen when using control NK-cell transfer, irrespective of the irradiation (Fig. 1B and C; Supplementary Fig. S2A).

The presence of tumor also facilitated NK allograft acceptance, as the high infiltration of transferred NK cells was not seen in spleens in tumor-free mice, regardless of the irradiation (Fig. 1D). Next, we asked whether this allograft acceptance conferred by the tumors was mediated by Tregs. Depletion of Tregs in host mice before DLI almost abolished the high infiltration of the transferred NK cells in tumors (Fig. 1E), confirming that tumor-elicited Tregs are critical for graft acceptance (21, 26).

Eleven days after transfer, when T-cell reconstitution of DLI origin started to occur, the transferred NK cells still remained present at high numbers in blood, spleens, and tumors of irradiated B16F10 melanoma–bearing mice given DLI (Fig. 2A–D; Supplementary Fig. S2B). We hypothesized that transferred NK cells may persist for a long time, thus contributing to tumor control (1). Indeed, the transferred NK cells, specifically in spleen, blood, lung, and liver, were still readily observed when tumors became invisible (e.g., 99 days after DLI), despite being at lower numbers (Fig. 2E and F). This suggests that the high infiltration of the transferred NK cells was positively correlated with their persistence.

The persistence of transferred NK cells also depended on the presence of tumor before irradiation, as the transfer of DLI into irradiated tumor-free mice resulted in the GvHD-related animal death, with high clinical GvHD scores and histologic GvHD scores, which were not seen in B16F10 melanoma–bearing mice (Supplementary Fig. S3A–S3C). This phenomenon was in line with our previous finding that solid tumors protect against GvHD after allo-BMT in mice (21), suggesting that low risk of GvHD after DLI results in increased animal survival, thus providing sufficient time for the persistence of the transferred NK cells.

In the subcutaneous B16F10 melanoma model, without irradiation, no significant differences were observed in tumor growth and overall survival after DLI versus control NK-cell transfer, even when compared with no transfer (Fig. 3A and B). Irradiation alone inhibited tumor development to some extent, versus no treatment. However, in combination with irradiation, DLI resulted in a significant decrease of tumor growth and increase of overall survival, versus control NK transfer or irradiation alone (Fig. 3C and D). This pattern of tumor control was positively correlated with NK allograft acceptance.

To demonstrate that tumor control was not restricted to a particular MHC-I–negative melanoma cell line, we used another melanoma cell line B16F0 also lacking MHC-I. Similar results were obtained (Fig. 3A–D). However, when using IFNγ – B16F10 cells expressing MHC-I (Supplementary Fig. S4; ref. 18), a less aggressive tumor model, no measurable benefits were observed (Fig. 3A–D). After stripping surface MHC-I on IFNγ – B16F10 cells by low pH citric acid buffer (Supplementary Fig. S4; refs. 19, 20), measurable benefits were regained (Fig. 3A–D), suggesting that the tumor control was especially efficient in MHC-I–negative melanoma. Next, we confirmed this notion by using B16F10 melanoma metastases model (27). DLI, but not control NK transfer, in combination with irradiation, significantly restricted the lung metastases (Fig. 3E and F).

To test whether NK allograft acceptance would directly affect tumor growth, we utilized NK-depleted DLI. NK depletion from DLI almost completely abolished the tumor control (Fig. 3G and H). Of note, depletion of CD4⁺ or CD8⁺ T cells contained in the DLI did not impair tumor control (Fig. 3G and H), suggesting that T cells of DLI origin did not contribute to the protective effect of DLI against MHC-I–negative melanoma.

We then asked whether expansion of transferred NK cells contributed to their high infiltration in tumors. To address this, DLI or control NK cells were labeled with CFSE before transfer, allowing for tracing cell expansion by CFSE^lo in vivo. Using the same gating strategy to define the transferred NK cells (Fig. 1C), a significant increase of CFSE^lo cells was observed in B16F10 tumors of irradiated tumor-bearing mice given CFSE-DLI versus CFSE-NK (Fig. 4A). These data suggested that the expansion directly affected the infiltration of NK cells in tumors. NK cell expansion can be induced by T cells (28), thus we questioned whether T cells were also involved in this expansion in our model. Depletion of CD4⁺ T cells, but not CD8⁺ T cells, significantly decreased the expansion (Fig. 4A), indicative of a requirement of CD4⁺ T-cell help.

We next examined whether these NK cells also produced cytotoxic cytokine such as IFNγ (1). Indeed, IFNγ production in NK cells, which was restricted to donor origin, was significantly increased in tumors of irradiated B16F10 melanoma–bearing mice given DLI versus control NK transfer (Fig. 4B). This effect depended on CD4⁺ T-cell help, as depletion of CD4⁺ T cells, but not CD8⁺ T cells, reduced IFNγ production (Fig. 4B). Similar results were observed concerning Granzyme B production (Fig. 4C), a signature of NK-cell activation (1). Of note, this process was efficiently induced in vivo after transfer, as NK cells of donor origin before transfer failed to express IFNγ and Granzyme B, and other classic activation markers (i.e., CD25 and CD69; Supplementary Fig. S5A–S5C). As expected, the tumor control effects were significantly suppressed by depletion of CD4⁺ T cells, instead of by depletion of CD8⁺ T cells (Fig. 4D and E).

Given the CD4⁺ T-cell help in this model, we extended the analysis to other CD4⁺ T-cell subsets, such as Tregs and Th17 cells. After DLI, depletion of Tregs, but not Th17 cells, significantly increased the expansion and activation of transferred NK cells (Fig. 4A–C), resulting in better tumor control (Fig. 4D and E).

NK cells develop effector functions after a four-stage maturation program from CD11b^loCD27^lo |\to $| CD11b^loCD27^hi |\to $| CD11b^hiCD27^hi |\to $| CD11b^hiCD27^lo, with the latter being fully mature cytotoxic NK cells (29). FACS staining of the transferred NK cells on day 11 showed an increase of fully matured NK cells (CD11b^hiCD27^lo) in tumors of irradiated B16F10 melanoma–bearing mice given DLI versus control NK cells (Fig. 5A). A similar pattern was also observed in the expression of Killer cell Lection-like Receptor G1 (KLRG1) and CD43, markers for NK-cell cytotoxic activity and terminal maturation (ref. 29; Fig. 5B). Again, this NK-cell maturation was restricted to donor origin (i.e., the transferred NK cells), as the above maturation markers were not observed in the corresponding NK cells of host origin in tumors (Supplementary Fig. S6A and S6B).

The maturation of transferred NK cells in tumors was significantly diminished in CD4⁺ T-cell–depleted host mice (Fig. 5A and B), suggesting that CD4⁺ T cells also impacted the maturation of transferred NK cells. Accordingly, the extent of this NK-cell maturation also involved Tregs (at a later stage), but not Th17 cells (Fig. 5A and B).

To further address the functional importance of CD4⁺ T-cell help, tumor-infiltrating NK cells (i.e., DLI-derived NK cells) were enriched from CD4⁺ T-cell–replete or CD4⁺ T-cell–depleted host mice 11 days after DLI plus irradiation (Fig. 5C and D), which contained mature and immature NK cells (Fig. 5A and B), respectively. In the context of irradiation, transfer of the mature NK cells inhibited the subcutaneous B16F10 melanoma development, compared with the transfer of control NK cells (Fig. 5C and D). In contrast, immature NK cells did not impact tumor growth (Fig. 5C and D), confirming a critical role for CD4⁺ T cells in the maturation of transferred NK cells, which involves their cytotoxic activity in the control of melanoma.

Rapid secretion of IL2 by CD4⁺ T cells triggers NK-cell responses, including expansion, activation, and maturation (24). We questioned whether IL2 produced from CD4⁺ T cells in this model could control NK-cell function. In tumors of irradiated host mice given DLI, CD4⁺ T cells, but not other cells (CD4⁻), produced IL2 (Fig. 6A), suggesting that CD4⁺ T cells were the main source of IL2. To exclude the possibility that CD4⁺ T-cell help would function independently of IL2 production, experiments were performed using an anti-IL2 antibody – IL2 blockade. When compared with isotype control, IL2 blockade suppressed transferred NK-cell expansion (Fig. 6B), fully recapitulating the effect of CD4⁺ T-cell depletion (Fig. 4A). Accordingly, IL2 blockade increased B16F10 melanoma growth and decreased overall survival (Fig. 6C and D), suggesting that CD4⁺ T-cell–produced IL2 was responsible for the NK-dependent tumor control.

The availability of IL2 can be limited by Tregs via direct consumption and suppressing CD4⁺ T-cell activation (24). Eleven days after DLI, when donor-type T-cell reconstitution was low (<3%; Supplementary Fig. S2B), host-type CD4⁺ T cells in tumors preferentially developed into an activated phenotype (i.e., CD25⁺Foxp3⁻), but not Tregs (i.e., CD25⁺Foxp3⁺), both in frequency and numbers (Fig. 6E). This suggested that host CD4⁺ T-cell–produced IL2, which did not facilitate Tregs accumulation, was likely available for the activation of transferred NK cells. However, IL2 blockade almost completely abolished the activated phenotype, preferentially leading to Tregs instead (Fig. 6E). While we cannot exclude that the other cytokines (i.e., TGFβ, IL10, IL35) may induce Tregs (30), our data suggested that both IL2 blockade and/or Tregs accumulation may have limited the activation of transferred NK cells, thus explaining the dampened tumor control (Fig. 6C and D).

To further address whether IL2 exerted additional antitumor effects beyond a CD4⁺ T-cell–mediated effect, we performed a rescue experiment, in which murine recombinant IL2 (rIL2) was administered daily from days 10 to 19 (Fig. 7A). rIL2 decreased the tumor development in CD4⁺ T-cell–replete mice, versus vehicle-injected controls (Fig. 7B and C). These data indicated that IL2 plays a role in this model. In contrast to the observation in CD4⁺ T-cell–replete mice, rIL2 administration did not alter the tumor development in CD4⁺ T-cell–depleted mice (Fig. 7B and C), suggesting no direct impact of IL2 on the transferred NK cells.

Finally, to test whether host CD4⁺ T cells directly affected the activation and maturation of transferred NK cells, we conducted an in vitro coculture assay. Host-type CD4⁺ T cells were activated and then exposed to B16F10 cells. By culturing these purified CD4⁺ T cells with donor-type NK cells, we could rule out the possibility of NK-cell activation and phenotype changes were induced by the contaminating tumor cells. Both IFNγ production by NK cells and fully mature NK phenotype (CD11b^hiCD27^lo) were higher in the presence of tumor-experienced versus tumor-free CD4⁺ T cells or control NK cells alone (Fig. 7D and E), indicating that the observed effect was mediated directly by tumor-experienced host CD4⁺ T cells.

After transfer of allogenic DLI, tumor-bearing host mice present relatively high numbers of the transferred NK cells, especially in tumors. This NK allograft acceptance relied on the synergy between sublethal radiation and tumor-induced environment and was supported by three independent lines of evidence. First, tumor-bearing host mice given sublethal radiation that showed evidence of lymphopenia had more infiltration of transferred NK cells, as compared with tumor-bearing host mice not given sublethal radiation. Second, tumor-free host mice countered the NK allograft acceptance. Third, depletion of Tregs, which were driven by tumors in this model, also counteracted the NK allograft acceptance despite the presence of sublethal radiation.

The NK allograft acceptance preferentially occurred at relative early (i.e., 4–11 days) after DLI transfer in our model, as compared with the relative absence of donor T-cell reconstitution. This paradigm of allograft acceptance in mice may be explained by the lack of establishment of a host NK-cell–mediated graft rejection (31, 32), recovery pattern that favors rapid NK-cell reconstitution, correlating with low risk of GvHD (33, 34), and/or the absence of a mixed chimeric state after allo-BMT that are required for the induction of T-cell tolerance (35–37).

NK allograft acceptance is critical to establish tumor control along with reduced GvHD (14). However, under mixed chimeric state conditions, transferred NK cells from allo-BMT inoculum do not mediate tumor control, due to the fact that these NK cells are specifically tolerant to host-derived tumor cells (38). Presumably, NK-cell tolerance in mixed allogenic chimeras is associated with a failure of NK-cell activation in vivo (39). The use of donor-derived activated NK-cell infusion (NK-DLI) after allogenic hematopoietic stem cell transplantation (HSCT) is associated with higher donor CD3 chimerism favors augmented T-cell alloreactivity in patients with cancer, leading to the acute GvHD (clinical trial NCT01287104; ref. 15). Therefore, the activity and long-term persistence of the transferred NK cells in vivo becomes essential to restrict the T-cell alloreactivity in the occurrence of GvHD (40) and to establish tumor control. Consistent with this notion, we found that tumor-infiltrating transferred NK cells rapidly expressed cytotoxic molecules, for example, IFNγ and Granzyme B, and cytokines associated with the powerful tumor control. This NK activation was CD4⁺ T-cell–dependent and occurred after transfer, which is consistent with previous findings showing that NK cells adapt to their environment and are regulated by CD4⁺ T cells (24), and that NK-cell activation depend on CD4⁺ T cells (41). Accordingly, either depletion of NK cells contained-in DLI or depletion of CD4⁺ T cells in host mice almost completely abolished the antitumor efficacy; this appeared to rely on the activation of tumor-infiltrating NK cells of DLI origin. Presumably, this also explained why NK cells of DLI origin persisted long-term (i.e., up to 99 days) in multiple organs when tumors were eliminated in our model. This was associated with the long-term survival of host mice that were protected against GvHD.

Whereas transfer of enriched donor NK cells resulted in less infiltration and poorer tumor control, this did not appear to be involved the expansion of transferred NK cells (24, 28). Instead, transfer of DLI was associated with the CD4⁺ T-cell–dependent expansion of transferred NK cells, presumably explaining the high infiltration of transferred NK cells in tumors. However, Tregs involvement in this model is complex. After DLI, Tregs may impair the functional significance of transferred NK cells, in contrast to their beneficial role before DLI. Therefore, Tregs at distinct stages appear to exert distinct roles, as reported previously (21). Unlike Tregs, Th17 cells' involvement was not observed, specifically after DLI, explaining the low risk of GvHD (42).

The ability of the transferred NK cells' infiltration correlates with their maturation state, suggesting that the antitumor cytotoxic effect of the transferred NK cells dependent on their mature state. In keeping with this notion, transfer of the tumor-infiltrating mature NK cells of DLI origin expressing high levels of cytotoxic marker (e.g., KLRG1 and CD43), leads to a powerful tumor control response, that is, delayed tumor growth and prolonged survival. Loss of CD4⁺ T cells in host mice suppresses the transferred NK cell maturation and its tumor control response.

It is well established that IL2 produced by CD4⁺ T cells can modulate NK-cell tolerance or responses (24). In support of this notion, NK-cell responses are impaired in IL2-deficient mice (43) or mice with high numbers of tumor-elicited Tregs that limit the availability of IL2 (44, 45). In our model, after DLI into the irradiated tumor-bearing mice, CD4⁺ T cells produced relatively high amounts of IL2, which conferred antitumor efficacy via the expansion of transferred NK cells and restriction of Tregs in tumors. Conversely, IL2 blockade led to the opposite effects (e.g., NK-cell tolerance) and dampened antitumor effects. Similar to other NK-cell tolerance mechanisms (38), NK-cell responses mediated by IL2 did not act beyond a CD4⁺ T-cell–mediated effect, that is, exogenous IL2 failed to protect against tumors without CD4⁺ T cells in this model. This phenomenon was confirmed by our in vitro coculture data where tumor-experienced host CD4⁺ T cells directly induced the activation and maturation of donor NK cells. These data highlighted the importance of activated host CD4⁺ T-cell help as an efficient in stimulating DLI-derived NK-cell responses for the treatment of MHC-negative melanoma.

It was not clear whether IL2 produced by CD4⁺ T cells in recipients given DLI allowed for transferred NK cells expansion directly competes with tumor-elicited Tregs accumulation. The presence of tumor-elicited Tregs before transfer preferentially initiated NK allograft acceptance that were of DLI origin. In the case of DLI, this was associated with donor-type dendritic cells (DC) maturation (Supplementary Fig. S7A and S7B) in response to the host tumor milieu. This type of DC–NK-cell interplay can activate CD4⁺ T cells in the tumor microenvironment (TME; refs. 46–48). Recipients given DLI developed relative high numbers of activated CD4⁺ T cells of host origin. Presumably, this explained why CD4⁺ T cells produced IL2. It is therefore possible that local IL2 production by CD4⁺ T cells in the TME was necessary and sufficient for enabling the dynamic circuit between the expansion of transferred NK cells' expansion and host Tregs decline (49).

A considerable proportion of the transferred NK cells developed after DLI, and NK cells were able to mediate efficient tumor control without inducing GvHD. The observation that these transferred NK cells established resistance to MHC-I–negative melanoma (47, 50), in which host NK cells poorly infiltrated into the tumors even when donor-type DCs became mature (Supplementary Fig. S7), confirms that MHC-I–negative melanoma were susceptible to NK-cell–mediated killing.

In conclusion, without prior allo-BMT, the use of DLI, but not enriched donor NK cells, led to NK allograft acceptance in the melanoma milieu, where IL2-producing host CD4⁺ T cells regulated the expansion and cytotoxic activity of transferred NK cells. This paradigm may pave the way to new antimelanoma treatments specifically targeting allogenic NK cells.

No potential conflicts of interest were disclosed.

Conception and design: N. Dang, M. Waer, B. Sprangers

Development of methodology: N. Dang, Y. Lin, B. Sprangers

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Dang, Y. Lin, M. Waer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Dang, Y. Lin, B. Sprangers

Writing, review, and/or revision of the manuscript: N. Dang, Y. Lin, M. Waer, B. Sprangers

Study supervision: N. Dang, M. Waer, B. Sprangers

This work was supported by the Olivia Hendrickx Research Fund (http://www.olivia.be). The authors thank past and present members of the Waer lab for their assistance and feedback. B. Sprangers is a senior clinical investigator for The Research Foundation Flanders (to Fonds Wetenschappelijk Onderzoek — Vlaanderen; 1842919N).

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.