Natural Killer Cell Mediated Missing-Self Recognition Can Protect Mice from Primary Chronic Myeloid Leukemia In Vivo

Background Natural Killer (NK) cells are thought to protect from residual leukemic cells in patients receiving stem cell transplantation. However, multiple retrospective analyses of patient data have yielded conflicting conclusions regarding a putative role of NK cells and the essential NK cell recognition events mediating a protective effect against leukemia. Further, a NK cell mediated protective effect against primary leukemia in vivo has not been shown directly. Methodology/Principal Findings Here we addressed whether NK cells have the potential to control chronic myeloid leukemia (CML) arising based on the transplantation of BCR-ABL1 oncogene expressing primary bone marrow precursor cells into lethally irradiated recipient mice. These analyses identified missing-self recognition as the only NK cell-mediated recognition strategy, which is able to significantly protect from the development of CML disease in vivo. Conclusion Our data provide a proof of principle that NK cells can control primary leukemic cells in vivo. Since the presence of NK cells reduced the abundance of leukemia propagating cancer stem cells, the data raise the possibility that NK cell recognition has the potential to cure CML, which may be difficult using small molecule BCR-ABL1 inhibitors. Finally, our findings validate approaches to treat leukemia using antibody-based blockade of self-specific inhibitory MHC class I receptors.


Introduction
Chronic myeloid leukemia (CML) is a myeloproliferative disorder characterized by a reciprocal translocation between chromosome 9 and 22, the so-called Philadelphia (Ph) chromosome. This translocation juxtaposes the genes encoding the ABL1 tyrosine kinase and BCR (Breakpoint cluster region), resulting in a BCR-ABL1 fusion protein with constitutive tyrosine kinase activity. This activity is critically involved in the initial chronic phase of CML disease and the subsequent disease progression. Indeed, the BCR-ABL1 inhibitor imatinib has become the standard therapy in newly diagnosed CML patients. Based on multiple clinical studies, a majority of patients (52-69%) achieve a complete cytogenic response (i.e. no Ph+ metaphases in 20/20 cells) but only a minority of patients (12-40%) achieve a major molecular response (i.e. a 3-log reduction in BCR-ABL1 mRNA) by 12 months of treatment [1]. Nilotinib and desatinib are secondgeneration inhibitors that exhibit considerably higher activity against BCR-ABL1 and that show further increased response rates [1]. Despite the impressive ability to control disease, there are CML patients that do not respond to BCR-ABL1 inhibitors or in which the disease progresses, some times based on mutations in BCR-ABL1. Finally, recurrence has been observed in a significant fraction of patients when BCR-ABL1 inhibitor treatment is discontinued [2], suggesting that leukemia initiating cells may persist and be refractory to inhibitor treatment. Thus additional treatment options, which are able to target leukemia-propagating cells, are needed to treat certain CML patients.
Haematopoietic stem cell transplantation has the potential to cure CML [3]. This is in part due to immune cells present in the graft and/or developing from grafted stem cells, which mediate a graft versus leukemia (GvL) effect to eliminate residual leukemic cells. Unrelated HLA-matched and partially HLA-mismatched transplants may contain T cells, which recognize minor histocompatibility antigens or HLA molecules on residual leukemic cells, respectively. However, such T cell recognition bears the significant risk of graft versus host disease (GvHD), a life threatening complication of (partially) HLA mismatched stem cell transplantation, in which donor-derived T cells attack non-haematopoietic, healthy tissues of the recipient. A partial HLA mismatch can also be recognized by NK cells and it has been suggested that alloreactive NK cells can prevent leukemia relapse following stem cell transplantation [4]. In contrast to T cells, NK cells do not seem to cause GvHD [5,6].
NK cells can react to allogeneic cells based on various recognition events. First some NK cells can be activated using receptors, which are specific for allogeneic MHC-I [7,8]. In addition, many NK cells express inhibitory receptors specific for MHC-I [9,10]. MHC-I receptors counteract NK cell activation by receptors specific for ligands that are constitutively expressed on healthy host cells. This dual receptor system allows the killing of diseased host cells, which display aberrantly low levels of MHC-I molecules (missing-self recognition) [11]. Since inhibitory MHC-I receptors (KIR (Killer Immunoglobulin-like Receptors) in human and Ly49 family receptors in mice) do not recognize all MHC-I alleles, the dual receptor system can confer reactivity to allogeneic cells that express the wrong MHC-I (KIR ligand mismatch). NK cell alloreactivity is further dependent on NK cell education [12] i.e. activation receptors on NK cells, which express a KIR/Ly49 specific for self-MHC-I respond more efficiently to stimulation [13,14,15,16]. Consequently, NK cell alloreactivity depends on the expression of a KIR/Ly49 and its MHC-I ligand in the donor (for education) and the absence of MHC-I ligand in the recipient (for the relieve from inhibition). Conversely, the activation receptors on NK cells that do not express a KIR/Ly49 specific for self-MHC-I respond poorly stimulation [13,14,15,16]. However, the function of these activation receptors can improve when these uneducated NK cells are exposed to inflammatory cytokines [14,17]. Consequently, it is possible that uneducated NK cells acquire reactivity due to the peculiar inflammatory environment during stem cell transplantation [18]. In addition to NK cell alloreactivity, there is evidence that the upregulation of stress induced self ligands, such as those engaging the NKG2D activation receptor [19], represents an important NK cell recognition strategy for transformed cells [20]. The importance of this pathway in the context of GvL is not known.
A role of NK cells to protect from leukemia has been inferred from the retrospective analysis of data from patients receiving unrelated HLA matched or partially HLA mismatched hematopoietic stem cell transplantation. A reduced relapse and improved disease free survival of leukemia patients receiving partially mismatched transplants has been attributed to NK cell alloreactivity mediated by the absence of KIR ligand in the recipient (KIR-ligand mismatch i.e. missing-self recognition) [4]. However, certain studies have either found no beneficial effect of KIR ligand mismatch or reported even worse outcomes following partially HLA mismatched stem cell transplantation [21,22]. Moreover, other studies have proposed protective roles for NK cell recognition of allogenic HLA using activating KIR [23,24] or they suggested a role for uneducated NK cells [25,26,27] (for review see [28,29]). Some of these discrepancies may be due to distinct conditioning regimens, differences in the preparation, the source and the dose of the transplanted stem cells and/or the fact that the cohorts included patients with distinct hematological malignancies that are based on different primary genetic lesions.
Despite some of the above correlations, it has not been shown directly that NK cells control primary leukemic cells arising in vivo and, if so, it is not clear which NK cell recognition strategy would be most effective against leukemic cells. Finally it is not known whether NK cells can target leukemia initiating stem cells and thus whether NK cells have the potential to contribute to curing leukemia.
Here we have addressed whether NK cells do have the potential to control primary CML in vivo. To this end, we used a well-established mouse model, in which bone marrow (BM) precursor cells are transduced with the BCR-ABL1 oncogene. Infected BM cells are transplanted into lethally irradiated recipient mice, which develop a fatal CML disease [30,31]. This disease model was combined with the classical model of BM graft rejection mediated by host NK cells [32], which allowed us to determine whether there was a difference in the efficacy of NK cells against normal and leukemic cells. Our analyses reveal that NK cell-mediated missing-self recognition, but none of the other NK cell recognition strategies, significantly impacts the outgrowth of CML cells in vivo. The NK cell mediated effect is based at least in part on the targeting of leukemia initiating stem cells, which suggests that NK cell recognition may contribute to cure CML disease.

Lack of NK cell recognition of MHC-I matched BCR-ABL1+ myeloid cells in vivo
To determine whether NK cells can protect against primary BCR-ABL1+ cells in vivo, BM progenitor cells were infected with a retrovirus encoding the BCR-ABL1 (p210) oncogene plus green fluorescent protein (GFP) or with a control retrovirus expressing only GFP. Transduced BM precursor cells (usually around 10% GFP+) were transplanted into lethally irradiated recipient mice. Polymorphisms in the CD45 molecule were utilized to discriminate hematopoietic cells of donor (CD45.1) and recipient (CD45.2) origin.

NK cell mediated missing-self recognition reduces the expansion of BCR-ABL1+ myeloid cells
We next addressed whether a partial MHC-I mismatch, which results in NK cell mediated missing-self recognition, impacts the expansion of BCR-ABL1 expressing cells. Control infected B6 BM precursors yield myeloid cells in B6 hosts (10 5 cells) ( Fig. 1B) but not in B6 hosts expressing a H-2D d transgene (H-2 b D d )(,10 4 cells) (Fig. 1D), illustrating NK cell mediated rejection of normal BM grafts based on missing-self recognition [33,34]. The transplantation of BCR-ABL1 B6 BM progenitors into B6D d hosts yielded a large number of BCR-ABL1+ myeloid cells (.10 6 cells) when NK1.1+ cells had been depleted (Fig. 1C, D). When NK cells were present, the number of BCR-ABL1+-CD11b+ cells was approximately 30 fold lower (Fig. 1C, D). Thus a partial MHC-I mismatch significantly diminished the expansion of BCR-ABL1+ cells in vivo due to the presence of NK cells. This is likely mediated by NK cells expressing inhibitory Ly49A+ and Ly49G2+ receptors, which are inhibited by H-2D d but not H-2 b molecules [35].
We next tested whether complete MHC-I deficiency further accentuated the protective effect. Indeed, the abundance of BCR-ABL1+ myeloid cells that developed from b2m-deficient (b2m-ko, MHC-I low ) BM precursors was reduced 500 fold in the presence of NK cells in B6 recipient mice (H-2 b ) (Fig. 1E, F). We conclude that NK cell mediated missing-self recognition can efficiently control myeloid expansion driven by the BCR-ABL1 oncogene in vivo. The improved protective effect is likely due to a higher number of mismatches between inhibitory receptors and MHC-I ligands as observed for the NK cell mediated rejection of MHC-Idifferent normal cells [36].
NK cell mediated ''non-self'' or ''induced-self'' recognition fails to control BCR-ABL1+ cells To address whether NK cell activation by non-self MHC-I conferred a protective effect, we transplanted BCR-ABL1 expressing B6D d BM precursors into B6 recipients. In this donor/recipient combination, host NK cells partially reject normal B6D d BM allografts based on the H-2D d -specific Ly49D activation receptor [37]. However, NK cells in B6 recipients failed to impact the expansion of BCR-ABL1+ B6D d myeloid cells ( Fig. 2A), indicating that NK cell mediated non-self recognition is not effective against primary BCR-ABL1+ cells in vivo.
There is considerable evidence that the engagement of the NKG2D activation receptor by stress-induced ligands plays an important role for the recognition of transformed cells by NK cells [20]. While B6-background BCR-ABL1+ myeloid cells do not express the NKG2D ligands Rae-1 (Fig. 2B) or Mult1 (not shown), myeloid cells from certain mouse strains upregulate Rae-1 following BM transplantation and this results in graft rejection in specific recipient mouse strains [38]. Accordingly, we used the MHC-I matched, unrelated strain combination of BALB.B (H-2 b )2.B6 (H-2 b ) to address whether induced-self recognition can control BCR-ABL1+ cells. However, despite significant expression of Rae-1 (but not MULT1 or H60) ( Fig. 2B and data not shown) we failed to observe a significant NK cellmediated effect on BALB.B BCR-ABL1+ myeloid cells (Fig. 2C). Of note, though, we observed a significant NK cell-dependent rejection of normal BALB.B myeloid cells (Fig. 2C), in agreement with [38]. Thus, despite the expression of NKG2D ligand, induced-self recognition of BCR-ABL1+ cells in vivo is impaired.

NK cell mediated missing-self recognition can protect from CML disease in vivo
Since NK cell mediated missing-self recognition impacted myeloid expansion at d8 post transplantation, we next determined whether recipient mice were protected from CML disease in vivo.
Since NK cell rejection of normal BM allografts is lethal between d12 and d14 post transplantation we ensured the long-term survival of host mice by co-transplanting MHC-I matched rescue BM (Fig. 3A). Similar to the transplantation with a single type of BM, such mixed BM grafts rapidly induced CML disease, which is characterized by weight loss, increased numbers of peripheralblood cells (with a predominance of mature granulocytes), splenomegaly and pulmonary hemorrhage, owing to granulocyte infiltration into the lung (Fig. 3B) [31].
When BCR-ABL1+ BM was transplanted into MHC-I matched recipients, the presence of NK cells did not improve the survival of recipient mice, delay the onset of disease (Fig. 4A) or alter any of the symptoms associated with CML disease (data not shown). Corresponding observations were made when BCR-ABL1+ B6 BM was transplanted into NK cell-sufficient RAG-1-knock out mice (H-2 b ) and into NK cell-deficient RAG-1 common c chain double-knock-out mice (data not shown). Thus, as expected, NK cells do not impact MHC-I matched CML disease.
In contrast to the early time point after transplantation (d8) (Fig. 1C, D), animals with CML disease had comparable numbers of BCR-ABL1+ myeloid cells, irrespective of whether NK cells were deleted or not (Fig. 5A). The eventual failure to prevent disease progression was not due to a loss of NK cells. On the contrary, NK cells were actually slightly more abundant at later time points (Fig. 5B). Thus the initial control of leukemic cells is incomplete and BCR-ABL1+ myeloid cells eventually escaped NK cell-mediated missing-self recognition.
Finally we tested whether complete MHC-I mismatch provided an improved protection against CML disease. In the absence of NK cells, the majority of B6 recipients receiving b2m-ko BCR-ABL1+ BM developed disease symptoms (70%, 7/10) (Fig. 4C). In contrast, when NK cells were present most recipients of b2m-ko BCR-ABL1+ BM remained disease free (89%, 8/9) for .45d (Fig. 4C). Interestingly, the only recipient that developed disease, had symptoms characteristic of B-ALL (B cell acute lymphocytic leukemia) rather than CML. B-ALL is characterized by a moderate splenomegaly and enlarged lymph nodes due to an expansion of B220+ B cells and such mice do not have the extensive infiltration of the lung and liver characteristic of CML (data not shown). In the absence of NK cells, some B6 recipients of BCR-ABL1+ b2m-ko BM also developed B-ALL (2/7 mice that were diseased). However most recipients developed CML (5/7). These experiments formally show that following BM transplantation, NK cell mediated missing-self recognition can protect from CML disease.

NK cell recognition of BCR-ABL1+ leukemia initiating cells
Finally we addressed whether NK cells exerted their protective effect by targeting mature BCR-ABL1+ myeloid cells, immature myeloid precursors or the small population of leukemia initiating cells, which propagate CML disease [39]. As mentioned above, NK cell mediated missing-self recognition strongly reduced the abundance of mature BCR-ABL1+ myeloid cells (Fig. 1). Moreover, myeloid/erythroid progenitor cells (Lin-Kit+ Sca-1-(LKS-)) GFP+ were also strongly reduced (Fig. 6A, B, D), suggesting that NK cells efficiently target immature myeloid progenitors. Finally, the abundance of leukemia initiating cells, which are present in the haematopoietic stem cell compartment (Lin-Sca-1+ c-kit+) (LSK) expressing BCR-ABL1 [40], was significantly reduced based on a partial MHC-I deficiency of the graft relative to the host (Fig. 6A, C). Leukemia initiating cells were essentially absent when the graft lacked MHC-I molecules (Fig. 6E). These data show that NK cell-mediated missing-self recognition can target leukemia initiating cancer stem cells and that this correlates with protection from CML disease.

Discussion
A potential role of NK cells to prevent leukemia relapse has been suggested based on the retrospective analysis of data from patients receiving unrelated HLA matched or partially HLA mismatched bone marrow transplantation. However, direct evidence that NK cells can prevent leukemia relapse is lacking and it has remained unclear what NK cell recognition events might be effective. Here we provide direct evidence that NK cells can be effective against primary CML in vivo. Unexpectedly, a protective effect was only observed for NK cell mediated missingself recognition whereby an increased MHC-I receptor-ligand mismatch improved the protective effect in vivo. In contrast, the positive recognition of allogeneic MHC-I exerted no measurable effect on the abundance of BCR-ABL1-expressing myeloid cells in vivo. Further, it was possible that lethal irradiation/BM transplantation produced a cytokine milieu that reversed the reduced responsiveness of activation receptors on non-educated NK cells (i.e. Ly49A+ and Ly49G2+ NK cells in H-2 b mice), which are not inhibited by the MHC-I of the tumor. However, the absence of an effect of NK cells upon MHC-I matched transplantation provides circumstantial evidence that uneducated NK cells do not become reactive against BCR-ABL1+ cells. Finally, the recognition of stress-induced ligands by NKG2D also failed to mediate a significant protective effect against leukemia, even in a situation where the respective ligand(s) are expressed. Since normal BM cells that express NKG2D ligand are rejected, the data raise the possibility that BCR-ABL1 expression impairs the recognition of cells expressing NKG2D ligand. This defect was not based on a down-regulation or loss of the NKG2D receptor on NK cells (data not shown). Collectively, we have been unable to identify NK cell recognition events that allow the selective detection of BCR-ABL1+ as compared to normal host cells. Rather this study has identified NK cell recognition events that are impaired due to the expression of BCR-ABL1. Missingself recognition was the only NK cell recognition strategy, which was not impaired and which significantly impacted the course of CML disease.
The NK cell activation receptors (and their ligands) mediating missing-self recognition remain incompletely defined. It is clear, however, that NK cells are activated by multiple receptors specific for constitutively expressed ligands on healthy host cells. The predominant reactivity of NK cells for cells of haematopoietic origin is explained in part by the expression of SLAM family receptors/ligands by haematopoietic cells [41]. Such restricted expression of ligands may also explain the observation that NK cells do not mediate GvHD [42]. The identification of all the activating receptors and their ligands will be required to explain why missing-self recognition is particularly effective against CML and whether the recognition of CML is dependent on the same or partially distinct activation receptors. Moreover, this knowledge will be essential to understand how CML eventually escapes NK cell mediated missing-self recognition. In addition to specific receptor-ligand interactions, it is also possible that the particular cytokine milieu generated in the course of conditioning and BM transplantation favors missingself reactivity.
Here we have aimed at determining the potential of NK cells to protect from CML following BM transplantation. The advantage of our set-up is that the efficacy of NK cells against BCR-ABL1+ cells can be compared to that against normal cells, which is well established based on classical BM graft rejection experiments. Indeed, as detailed above, we did identify significant differences between the control of normal and BCR-ABL1+ BM progenitors. It is clear, however, that GvL is mediated by donorderived and not by host-derived NK cells. Thus future studies will verify whether missing-self recognition by donor-derived NK cells is equally effective. While missing-self recognition has the potential to protect from CML disease, the efficacy of host NK cells against partially mismatched CML was limited. Despite a delay in disease onset, the majority of recipient mice (12/14; 86%) eventually succumbed to CML disease. While NK cells largely controlled BCR-ABL1+ myeloid cells at an early timepoint after transplantation (day 8), in diseased animals the abundance of CML cells and their precursors was comparable, independent of the presence of NK cells. The eventual lack of control was not due to a complete loss host NK cells. On the contrary, host NK cells were slightly more abundant in the spleen of diseased mice. In agreement with these data, normal allogeneic BM induces the expansion of adoptively transferred NK cells [43]. Irrespective of this, BCR-ABL1+ cells eventually escape NK cell mediated control. The limited efficacy of NK cells may be explained by irradiation damage to host NK cells. Indeed, we have noted that lethal and even sub-lethal irradiation significantly reduces the abundance of splenic NK cells (5 fold). Moreover, within the first few days after transplantation residual host NK cells do not proliferate [44] (and our unpublished data). As donor-derived or adoptively transferred NK cells should not be damaged, it will be of interest to see whether these NK cell populations are more effective against CML. On the other hand, recent data suggest that the systemic exposure of mature NK cells to a large number of MHC-I-deficient normal cells profoundly reduces NK cell function [45]. Induction of NK cell hyporesponsiveness seems to be the outcome when mature NK cells are persistently stimulated under non-inflammatory conditions. Indeed, when NK cells are stimulated in the context of an inflammatory environment their function can stably improve [46,47]. If so, approaches to keep stimulated NK cells functional in a non-inflammatory environment may be needed to obtain protective effect against leukemia. The mouse models used here should be useful to address these issues.
Significantly we provide evidence that NK cells can impact the course of CML disease by targeting leukemia initiating stem cells that are refractory to the control by BCR-ABL1 tyrosine kinase inhibitors. These data suggest that NK cell recognition has the potential to cure CML. Finally, our findings validate approaches to treat leukemia using mAb-based blockade of self-specific inhibitory MHC-I receptors [48,49] and they may influence the selection of optimal donor/recipient combinations for stem cell transplantation to treat leukemia.

Retroviral infection and bone marrow transplantation
Retroviral infection and BM transplantation was performed essentially as described in [31]. Briefly, plasmids containing MSCV IRES GFP and MSCV BCR-ABL1 (p210) IRES GFP were transiently transfected into 293T cells for high titer virus production using standard procedures. BM donor mice were injected with 5-Fluorouracil (5-FU) (0.15 mg/ml) and BM cells were harvested 4 days later. BM cells were cultured in DMEM (plus 10% FCS, 50 mM 2-ME, 50 U/ml penicillin/50 mg/ml streptomycin, 10 mM HEPES) supplemented with IL-3 (10 ng/ ml), IL-6 (10 ng/ml) and SCF (50 ng/ml) for 48 h and infected with retrovirus in the presence of polybrene (8 mg/ml). After 24 h, 10 5 cells (usually around 10% GFP+) were injected i.v. into lethally irradiated recipient mice (2 doses of 480rad 4 h apart from a 137 Cs source). One day prior to irradiation, some recipient mice were injected i.p. with 200 mg of mAb PK136 (anti-NK1.1) to deplete NK1.1+ cells. Recipient mice were sacrificed at day 8 after transplantation and analyzed by flow cytometry (see below). Alternatively, for leukemia induction experiments, irradiated recipient mice were transplanted with a mixture of infected BM and non-infected (MHC-I-matched) rescue BM (10 5 each), both derived from 5-FU treated donors. Recipient mice were monitored daily for weight loss and failure to thrive. Pre-morbid animals were sacrificed and relevant tissues were harvested and analyzed visually and by flow cytometry (see below). For continuous depletion of NK1.1+ cells, recipient mice were injected i.p. with 200 mg of mAb PK136 (anti-NK1.1) one day prior to irradiation and mAb injection was repeated every 2 weeks.

Statistical analysis
A two-tailed student's t-test was used to determine significant differences between data sets and a Gehan-Wilcoxon test was used to compare survival curves. Data sets were considered significantly different when p,0.05.