Introduction
Mice with 4T1 breast cancer can be cured by transfer of 5 million alloreactive NK cells after a non-myeloablative dose of total body irradiation and cyclophosphamide (“chemo-irradiation”) [
1]. NK cell alloreactivity is present when tumor cells do not express the appropriate major histocompatibility (MHC) alleles for one or more inhibitory receptors of donor NK cells (i.e., “missing self” concept [
2]), a situation that typically may occur when using MHC-mismatched donors. Similar anti-cancer effects exerted by transferred alloreactive NK cells have been observed in mouse models for murine [
3‐
6] and human acute myeloid leukemia [
7], and for human multiple myeloma [
8]. Indications that alloreactive NK cells may be able to kill solid non-hematological tumor tissue comes from a preclinical study where it had been demonstrated that freshly isolated solid tumor tissue is only killed by NK cells from alloreactive and not from non-alloreactive donors [
9]. Unequivocal evidence that transfer of alloreactive NK cells exerts an anti-cancer effect in patients does not exist, either because the chemotherapy and/or irradiation applied before the administration may have resulted in increased progression-free survival or because transferred NK cells are not, transiently or only in limited numbers detectable in recipients [
10]. It is for this reason that many efforts are currently performed to produce large amounts of NK cells for clinical application, and the success of these approaches is yet to be awaited. As a dose–response relation of the number of transferred alloreactive NK cells and the anti-4T1 breast cancer effect had not been demonstrated yet, we wanted to proof this formally in the current study.
An alternative and clinically applicable source for alloreactive NK cells are MHC-mismatched hematopoietic cells. The advantages of this source are the guaranteed and continuous production of NK cells that are alloreactive toward the patient in case the patient does not express one or more ligands for NK cell inhibitory receptors that are present in the donor. In this setting, donor NK cells mature under the influence of the MHC of the donor’s hematopoietic system which directs the licensing of all NK cells that bear inhibitory receptors for self-MHC resulting in NK cell alloreactivity toward tumor cells of patients that lack the appropriate MHC alleles [
11‐
13]. The possible benefit of this treatment is not hypothetical, as results of retrospective clinical studies show that relapse rates in patients with acute myeloid leukemia after MHC-mismatched hematopoietic stem cell transplantation (HSCT) from NK-alloreactive donors are lower compared to the results with non-NK-alloreactive donors under the condition of a low incidence of acute graft-versus-host disease (GVHD) [
14‐
19]. MHC-mismatched HSCT was, however, until recently a very risky procedure due to a high treatment-related mortality from opportunistic infections due to a prolonged T cell deficiency state when deep T cell depletion was applied to prevent GVHD [
15,
20,
21]. However, the now widely used application of post-HSCT cyclophosphamide (PT-CY) prevents prolonged T cell deficiency, high infection rates, and GVHD, and this has made MHC-mismatched HSCT a safe and feasible procedure [
22,
23]. NK-alloreactive MHC-mismatched HSCT has never been tested for patients with breast cancer or any other type of non-hematological cancer, and hence we wanted to study the curative potential of alloreactive NK cells that had matured in 4T1 breast cancer-bearing mice. As 4T1 is a rapidly growing tumor and because it takes several months for NK cells to mature and become fully functional after HSCT [
13,
24], we employed a surrogate model to mimic treatment of breast cancer-bearing mice with HSCT from a NK-alloreactive MHC-mismatched donor. In this model, MHC-mismatched mice that are NK alloreactive toward 4T1 breast cancer were injected either subcutaneously (s.c.) or intravenously (i.v). with 4T1 cells representing primary and metastasized disease, respectively, and followed for several weeks for tumor development. In some experiments, a non-myeloablative dose of cyclophosphamide and total body irradiation were applied to study if these would increase NK cell-activating cytokine plasma levels, NK cell activation, and the anti-tumor effect, and in vivo NK cell depletion by antibodies was used to demonstrate the indispensability of NK cells for the anti-tumor effect.
Materials and methods
Cells, animals, and tumor models
4T1 breast cancer cell line of Balb/cfC3H origin [
25] was cultured in RPMI1640. Harvest was after 2 min trypsinization, and 5 × 10
4 viable cells were injected either s.c. or i.v. Balb/c (H-2
d), C57BL/6 (“B6,” H-2
b), (CBA × C57Bl/6)F1 (“B6CBAF1,” H-2
b/k) and (Balb/c × C57Bl/6)F1 (“CB6F1,” H-2
b/d) mice were from Harlan Laboratories (Horst, the Netherlands) and housed under specified pathogen-free conditions. Mice were in follow-up for at least 100 days after tumor induction and underwent standard autopsy for the presence of lung and liver metastases which we never found in mice that were cured from their s.c. tumors. Balb/c mice that succumbed from breast cancer regularly had metastases in lungs and liver, while this was occasionally the case for B6CBAF1 mice with s.c. tumors. All B6CBAF1 mice that died from i.v. injected 4T1 tumor cells had pulmonary metastases.
For the study of a dose–response effect of transferred NK cells, 4T1-bearing Balb/c mice were used. In the second set of experiments where the anti-tumor effect of NK cells that had endogenously matured from bone marrow in the tumor-bearing host, we chose a model where MHC-mismatched B6CBAF1 mice served as recipients of 4T1 breast cancer cells instead of transplanting MHC-mismatched HSC into Balb/c 4T1-bearing mice. This was because the rapid growth of the 4T1 breast cancer cells, even after chemo-irradiation, would not allow sufficient time for the donor-derived alloreactive NK cells to become mature and active. S.c. injected 4T1 was used as a model for localized breast cancer, while i.v. injection of 4T1 cells mimics the process of metastasis.
Chemo-irradiation
Chemo-irradiation was performed by 2× 2 Gy TBI (PHILIPS X-ray unit, 225 kV, 10 mA, dose-rate 66 cGy/min) at 8 and 9 days after tumor induction combined with 200 mg/kg cyclophosphamide (Baxter Oncology GmbH, Halle, Germany) at day 9. This conditioning is non-myeloablative in mice [
26].
NK cell transfer
NK cell-enriched spleen cell batches were prepared from single cell suspensions from the spleens of donor mice by MACS negative selection (Miltenyi Biotec B.V., Utrecht, the Netherlands); a typical NK cell dose of 5 × 10
6 NK cells contained approximately 0.06 × 10
6 T cells. We made use of a difference in frequency of NK cells with alloreactive activity toward H-2
d target cells of various strains of mice as dictated by their H-2 type. Whereas all NK cells of B6 (H-2
b) and B6CBAF1 (H-2
b/k) mice are alloreactive toward H-2
d target cells (i.e., “full-alloreactive”), only half of the NK cells in CB6F1 (H-2
b/d) mice are (i.e., “half-alloreactive”). This difference has been shown to translate in faster in vivo elimination of H-2
d-positive target cells by full-alloreactive NK cells than by half-alloreactive NK cells [
1,
27,
28]. Besides, CB6F1 T cells are tolerant to Balb/c tissue and are therefore not able to act as alloimmune effector cells.
In vivo NK cell depletion
In vivo NK cell depletion was performed in two consecutive experiments by intraperitoneal (i.p.) injection with either anti-AsialoGM1 or anti-NK1.1. We chose for using both antibodies to be sure that the effect of antibody administration on the anti-tumor effect could most surely be attributed to the depletion of NK cells only, because subsets of T cells express AsialoGM1 and many NKT cells express NK1.1 [
29,
30]. 200 µl of mouse-specific polyclonal rabbit anti-AsialoGM-1 (Wako Pure Chemical Industries) [
31] was administered intraperitoneally at days 0, 5, and 10 after tumor induction. In the next experiment, 500 µg anti-NK1.1 (PK136, BD Pharmingen) [
29,
32] was administered intraperitoneally at day 0, 5, and then every other 5 days until the death of the mice or the end of the experiment, because we realized from the results of the experiment with anti-AsialoGM1 that this should have been administered during the whole experiment too. The NK cell depletion efficacy had been previously checked in three mice for each antibody by measuring the NK cell content in blood and spleen just before the time of second administration of either anti-AsialoGM1 or anti-NK1.1; NK cell depletion amounted approx. 1 log in all.
Plasma NK cell-activating cytokine level measurements
Plasma was prepared from blood drawn from just priorly euthanatized mice. IL-2, IL-15, IL-18, and IL-21 plasma levels were determined by standard ELISA (R&D Systems).
Flowcytometry
All flowcytometry analyses were performed using a BD FACS CantoII flow cytometer using BD DIVA software. In vivo NK cell depletion efficiency was by staining spleen single-cell suspensions with the mouse-specific antibodies CD3e PerCP, and CD49b APC after incubation with NMS. NK cell activation was determined with antibodies specific for CD69, CD107a, TRAIL, and FasL. 7-AAD or Pi was for excluding non-viable cells, and all flowcytometry analyses were exclusively on viable cells. All monoclonal antibodies and 7-AAD were from BD Biosciences, and Pi was from Invitrogen.
Ethical approval
The local animal ethical committee had approved all the mouse experiments.
Statistics
Survival curves were composed using the Kaplan–Meier method and compared with the Mantel–Cox log-rank test. Differences in cytokine levels and expression of NK cell activation markers after chemo-irradiation were compared with non-chemo-irradiated mice using the two-tailed Wilcoxon’s signed-rank sum test. Values are presented as mean ± SEM. In all cases, differences were considered statistically significant when probability (p) values were less than 0.05.
Discussion
In this study, we demonstrated a dose–response relation between adoptively transferred NK cells from NK-alloreactive donors and the anti-tumor effect as well as the dispensability of alloreactive T cells in the 4T1 mouse breast cancer model. The human equivalent of the minimally required number of full-alloreactive NK cells per mouse (5 million for a mouse weighing 20 g amounts 0.25 × 10
9/kg) would be 18.75 × 10
9 for a patient weighing 75 kg. This number can never be harvested from a donor in a single procedure and necessitates in vitro NK cell expansion. Each individual mouse and man bears NK cell subsets expressing different inhibitory and activating receptors. Two preconditions determine if a given donor NK cell is alloreactive: (1) membrane expression of iKIR specific for a ligand that is present in the donor and absent in the patient (i.e., certain MHC class I alleles) and (2) no NKG2A expression (inhibitory receptor binding ubiquitously expressed HLA-E that is not subject to allelic differences with respect to binding to NKG2A). Additional prerequisites for successful clinical application of expanded NK cells are sufficient numbers and absence of donor T cells causing severe GVHD. At present, the vast majority of the laboratories working on clinical grade expansion of NK cells do not unequivocally demonstrate that their NK cell products meet all four prerequisites [
37‐
49]. Only recently a report was published on a successful though laborious expansion procedure in the presence of membrane-bound IL-21, which resulted in preserved KIR expression and NKG2A absence [
50]. Feasibility of the clinical application of this NK cell product is yet to be awaited. It remains, in general, also to be seen, if a single administration of alloreactive NK cells which results in only a transient engraftment is as effective in patients like in our mouse model. This then justifies the exploration of alternative ways to apply NK cells, especially when resulting in the permanent presence of alloreactive NK cells in the patient.
One such an alternative strategy is to apply HSCT from NK-alloreactive MHC-mismatched donors as a permanent source of NK cells, which is nowadays a fairly safe procedure when PT-CY is applied [
22,
23]. Reports on MHC-mismatched HSCT from NK-alloreactive donors all point toward a benefit with respect to leukemia-free survival in settings with a low incidence of acute GVHD [
15‐
17,
51,
52], and the role of alloreactive NK cells in this setting is underscored by the fact that these cells are fully functional within a few months after HSCT [
11‐
13,
53,
54]. Advantages of NK cell therapy by MHC-mismatched HSCT are that it results in a permanent production of alloreactive NK cells, as the latter quality hinges on the origin of the hematopoietic cells (in this case the donor’s) among which they mature [
11‐
13]. In contrast, adoptively transferred NK cells will probably lose their alloreactive quality within days due to exposure to the MHC-disparate patient’s hematopoietic system [
55‐
57].
In this study, treatment with NK-alloreactive MHC-mismatched HSCT of a breast cancer patient was mimicked using a MHC-mismatched host as recipient of 4T1 breast cancer, because the 4T1’s far more rapid tumor growth than human breast cancer results in death weeks before the donor bone marrow would have produced enough functional NK cells. We demonstrated that both s.c. and i.v. injected 4T1 breast cancer is eradicated by AGM1- and NK1.1-positive cells. As only NK cells share both characteristics [
58,
59] and because NK cells alone are sufficient for cure in the NK cell transfer model [
1] [and this manuscript], it is apparent that the anti-tumor effect resulted from the MHC-mismatched host’s activated NK cells.
The elimination of i.v. injected tumor cells (including tumorigenic cells) by the NK cells of the NK-alloreactive recipient was efficient and did not require further treatment. This indicates that breast cancer metastasis in patients after NK-alloreactive MHC-mismatched HSCT may be prevented by the circulating donor-derived NK cells. For extra-vascular tumor sites, additional treatment with low-dose chemo-irradiation seems warranted for NK cell-mediated elimination. This indicates that chemo-irradiation may be required in patients with metastasized breast cancer when functional donor-derived NK cells have been produced several months after NK-alloreactive MHC-mismatched HSCT. During the time of NK cell repertoire constitution, tumor growth should be prevented by the application of myeloablative conditioning before MHC-mismatched HCSCT [
60] which has shown to be feasible even in heavily pretreated patients [
61], and by hormonal and/or anti-HER2neu therapies when appropriate.
Puzzling is that adoptive transfer of sufficient numbers of CB6F1 semi-MHC-mismatched NK cells cures Balb/c mice from 4T1 breast cancer, while s.c. and i.v. injection of 4T1 in CB6F1 mice invariably results in high tumor-induced mortality even when chemo-irradiation is applied in contrast to the results in fully MHC-mismatched B6CBAF1 mice. This discrepancy may be a matter of the number of actual alloreactive NK cells toward the Balb/c-type tumor. The fact that half-MHC-mismatched HSCT from NK-alloreactive donors is very effective in preventing leukemia relapse in patients [
15‐
17,
51,
52] brings promise that the number of alloreactive NK cells several months after half-MHC-mismatched HSCT will also be sufficient for the elimination of residual breast cancer after the conditioning before HSCT and the subsequent application of chemo-irradiation.
Various mechanisms may contribute to the potentiation by chemo-irradiation of the NK cell-mediated anti-tumor effect. One may be related to the ascertained NK cell activation either by the observed increase in plasma levels of pro-inflammatory and NK cell-activating cytokines, possibly related to the chemo-irradiation-induced reduction in lymphocyte number [
62‐
65], and/or by translocation of commensal gut flora and LPS [
63]. Although the number of NK cells also decreases after chemo-irradiation [
63], their numbers may still have remained sufficiently high to eradicate 4T1 breast cancer upon their cytokine-induced activation, analogous to the observation that i.v. injected MHC-mismatched hematopoietic stem cells are rejected by host NK cells even in lethally irradiated mice (the so-called “hybrid resistance model” [
66]). Another effect may be a reduction of 4T1 tumor-induced NK cell inhibitory elements like regulatory T cells and myeloid-derived suppressor cells [
63,
67‐
73], similar to its relevance for disease-specific T-cell therapy [
74].
What may the implications of our findings be for the treatment of patients with metastasized breast cancer? Adaptive transfer of alloreactive NK cells may at first glance seem the most preferable therapy as it evades the possible risks resulting from haploidentical HSCT like GVHD and infections. The toxicity of adaptive NK cell therapy in human patients is anticipated to be low based on present-day experience, although the numbers of transferred NK cells were not as high as we would aim at based on our dose–response experiment [
75‐
77]. It remains to be seen if the application of higher numbers of alloreactive NK cells resulting in short-term engraftment will be as curative as in our mouse model. Haploidentical HSCT from NK-alloreactive donors and the subsequent application of low-dose chemo-irradiation after NK cell repertoire maturation is then an intriguing and promising treatment approach as it results in the continuous production of alloreactive NK cells in women with metastasized breast cancer. When applicable, it may serve as a platform for subsequent application of low-dose CY+TBI and/or other immune-modulating treatments to enhance the alloreactive NK cell effect.