**5. Association between natural killer cell alloreactivity and graft-versus-leukemia effect**

#### **5.1. The complexity of hematopoietic stem cell transplantation may affect the graft-versus-leukemia effects of alloreactive natural killer cells**

HSCT is well-established curative treatment for hematologic malignances. The major limitation of HSCT is the absence of HLA-matched donor. Therefore, the HLA-haploidentical relatives are increasingly used as donor sources of HSCs. During the haplo-HSCT, donor and recipient share one identical HLA haplotype, while other haplotype is fully mismatched [99]. The GvL effect of NK cells was first observed in a landmark study which demonstrated that NK cells rapidly reconstituted and displayed potent anti-leukemia activity after extensive T-cell depleted haplo-HSCT [6]. After inhibitory receptor KIR was identified as a mediator of missing-self response, the GvL effect of NK cells was assumed to be mainly due to mismatches between donor KIR and recipient HLA class I which increase NK cell function thereby effectively lyse residual leukemia cells in the recipient. This hypothesis was first verified by Ruggeri L et al. [100]. They isolated donor-derived NK cells from haplotype-mismatched HCT recipients and tested their cytotoxicity toward recipient lymphocytes. Their results demonstrated donor alloreactive NK cells indeed effectively killed recipient myeloid leukemia cells. The follow-up study of this group demonstrated therapeutic utility of NK cell alloreactivity. They found transplants with KIR ligand mismatch were associated with decreased probability of relapse and increases in disease-free survival in the absence of GvHD [6].

Series studies subsequently tested NK cell alloreactivity-mediated GvL effect in the KIR ligand incompatibility model. However, not all studies supported the benefit role of alloreactive NK cells. During haplotype-HCT with less T cell depletion, KIR ligand mismatch patients even developed more acute GvHD (aGvHD), and the overall survival was poor. In a cohort using unrelated donors, no effect of predicted NK cell alloreactivity was observed, whereas in another similar cohort with T-cell depletion, KIR ligand mismatch prolonged overall survival [101, 102]. Frag et al. examined the clinical impact of NK cell alloreactivity in over 1500 unrelated transplants and found KIR ligand mismatch was not associated with relapse prevention [103]. In a retrospective study, data from 2062 unrelated transplant recipients were analyzed for acute myelocytic leukemia (AML), chronic myeloid leukemia (CML), and myelodysplastic syndrome (MDS) [104]. Missing one or more KIR ligands in the recipient were associated with less relapse in patients with early myeloid leukemia and in CML patients during first chronic phase. Nevertheless, KIR ligand mismatch was also associated with an increased risk of chronic GvHD (cGvHD) in CML patients.

significantly influence functional reconstitution of NK cells rather than numerical reconstitution. One study observed donor-derived NK cells reconstitution following HSCT with three different kinds of donor source including T-cell-replete adult grafts, T-cell-deplete adult grafts, and UCB [95]. Both of the T-cell-replete group and UCB group were treated with immunosuppression. They found KIR expression in T-cell-replete group was the highest among these three groups indicated donor-derived NK cells in this group may have better effector function. Accordingly, in an earlier study by the same group, they found IFN-γ production was more potent following HSCT with T-cell-replete grafts compared to T-cell-deplete transplantation [98]. These results were consistent with the observations by Nguyen et al. which demonstrated that without immunosuppression, recovery NK cells in the partial T-cell-depleted group had lower cytotoxicity than NK cells in the full T-cell-depleted group [92]. In summary, graft T cell content trends to enhance functional reconstitution of NK cells irrespective of immunosuppression. This promotion effect may be due to activation from T-cell-derived IL-2, or stimulation from other inflammatory cytokine such as IL-12 and IL-18, which are increased following donor T cell engraftment. Finally, CMV infection and GvHD may also have potent

effect on NK cell reconstitution, and we will discuss this in the following sections.

**5. Association between natural killer cell alloreactivity** 

**the graft-versus-leukemia effects of alloreactive natural killer cells**

**5.1. The complexity of hematopoietic stem cell transplantation may affect** 

HSCT is well-established curative treatment for hematologic malignances. The major limitation of HSCT is the absence of HLA-matched donor. Therefore, the HLA-haploidentical relatives are increasingly used as donor sources of HSCs. During the haplo-HSCT, donor and recipient share one identical HLA haplotype, while other haplotype is fully mismatched [99]. The GvL effect of NK cells was first observed in a landmark study which demonstrated that NK cells rapidly reconstituted and displayed potent anti-leukemia activity after extensive T-cell depleted haplo-HSCT [6]. After inhibitory receptor KIR was identified as a mediator of missing-self response, the GvL effect of NK cells was assumed to be mainly due to mismatches between donor KIR and recipient HLA class I which increase NK cell function thereby effectively lyse residual leukemia cells in the recipient. This hypothesis was first verified by Ruggeri L et al. [100]. They isolated donor-derived NK cells from haplotype-mismatched HCT recipients and tested their cytotoxicity toward recipient lymphocytes. Their results demonstrated donor alloreactive NK cells indeed effectively killed recipient myeloid leukemia cells. The follow-up study of this group demonstrated therapeutic utility of NK cell alloreactivity. They found transplants with KIR ligand mismatch were associated with decreased probability of relapse and increases in

Series studies subsequently tested NK cell alloreactivity-mediated GvL effect in the KIR ligand incompatibility model. However, not all studies supported the benefit role of alloreactive NK cells. During haplotype-HCT with less T cell depletion, KIR ligand mismatch patients

**and graft-versus-leukemia effect**

134 Natural Killer Cells

disease-free survival in the absence of GvHD [6].

The effect of NK alloreactivity in recipients of UCB grafts has also been tested. Willemze et al. evaluated the impact of KIR ligand incompatibility in recipients of UCB grafts with T-cell depletion and found KIR ligand mismatch was associated with relapse prevention and improved overall survival [105]. However, Brunstein et al. reported different results [106]. They investigated the impact of KIR ligand mismatch in patients of UCB grafts after different intensity conditioning regimens. After myeloablative conditioning, KIR ligand mismatch did not exert protective effect on GvHD, relapse, transplantation-related mortality, or survival. Following reduced intensity conditioning, KIR ligand mismatch patients developed more aGvHD and poorer overall survival.

The different results among these studies using either adult grafts or UCB grafts may be explained by the complexity of HCT, such as variable preparative regimens, donor sources, and degree of HLA mismatch with the donor. For instance, the degree of T-cell depletion was different among these studies, the grafts were extensively T-cell depleted in some studies, whereas only were partially depleted in others. T cell content may outcompete NK cells for cytokines and thereby influence the beneficial effect of alloreactive NK cells. Additionally, different intensity of preparative regimens also affects numeral and functional reconstitution of alloreactive NK cells after HSCT thus influences HSCT outcomes. Finally, the origin of leukemic blasts also affects HSCT outcomes. No beneficial effects of alloreactive NK cells in adult acute lymphoblastic leukemia (ALL) patients were observed probably due to the lack of expression of activating ligands [100]. However, allorecative NK cells have been proved to positively affect the outcome of HSCT in children with ALL [107]. Accordingly, the intensity of HLA expression on the leukemic lymphoblasts was shown to be critical for NK-mediated cytotoxicity in an experimental study [108].

#### **5.2. Donor killer immunoglobulin-like receptors affect beneficial effects of alloreactive natural killer cells and outcome after hematopoietic stem cell transplantation**

As we have mentioned above, KIR genes can be simplified into two main haplotypes: haplotype A and haplotype B. A hapltotype consists of main inhibitory KIR and single-activating KIR, KIR2DS4. B haplotype contains variable KIR gene encoding both inhibitory and activating KIRs which can be further divided into either centromeric (Cen) or telomeric (Tel) regions. The role of these haplotypes in HSCT is also evaluated. Mc Queen et al. first demonstrated grafts from KIR A/A genotype donors into KIR B recipients resulted in poorer survival in HLAmatched T-cell-replete sibling transplants [109]. In contrast, Stringaris et al. reported transplants from KIR B haplotype donors including genes for KIR2DS1, KIR3DS1, and KIR2DL5 improved overall survival and prevented relapse specific for AML patients with T-cell depletion during HLA identical siblings HSCT [110]. A larger cohort was performed to further evaluate the effect of KIR B haplotype in AML patients. The results indicated that grafts from donors with KIR B haplotype significantly reduced the risk of relapse and improved overall survival [111]. A follow-up study expanded numbers of AML patients with T-cell-depleted unrelated donor transplants and found that donor KIR B haplotype-mediated relapse prevention was enhanced in recipients who have HLA-C1 alleles rather than C2 homozygous recipients [36]. Michaelis SU et al. reported transplants from KIR B haplotype donors resulted in better HSCT outcome in ALL patients [112].

Collectively, it is logical to hypothesize that the presence of activating KIR in donor KIR B haplotype is favorably associated with better transplant outcome. However, it is difficult to evaluate the contribution of each activating KIR to relapse protection and better overall survival due to their unknown ligands. KIR2DS1 has been proved to interact with HLA-C group 2 alleles and involved in the killing of leukemic blasts from HLA-C2 patients. The interaction between KIR2DS1 and HLA-C2 was found to overcome NKG2A mediated inhibitory signals in vitro [113, 114]. These results suggested KIR2DS1-mediated activating signal may break the barrier of NKG2A and exert positive effect during HSCT. Venstrom et al. investigated the role of KIR2DS1 from donor in unrelated HSCT in 1277 patients with AML and found that KIR2DS1 expression in donor mediated a significant GvL effect [114]. Notably, no beneficial effect was observed when donors were HLA C2 homozygous as in those donors, NK cells expressing KIR2DS1 cannot be educated and acquire effector function. Therefore, at least one copy of HLA-C1 in donor is needed for KLR2DS1-dependent GvL effect. In addition, KIR3DS1 was also found to associate with lower rate of relapse and infection. Mancusi A et al. reported KIR3DS1 was associated with reduced infection mortality [115]. Further studies are still needed to identify ligands for activating KIR so as to confirm the effect of each activating KIR gene on allo-HSCT outcome.

The intensity of interaction between KIR and HLA may also influence the alloreactivity of NK cells during allo-HSCT. HLA alleles have been shown to bind NK cells with different affinities. The HLA class I molecules are highly polymorphic, and HLA-B alleles can be divided into HLA-B alleles with an isoleucine at position 80 (80I) and with a theonine (80T). The affinity between certain HLA-B alleles with an 80I and KIR3DL1 was found to be much stronger than those with an 80T [116, 117]. The impact of this different affinity was tested in HIV patients [118]. Presence of 80I but not 80T in HIV patients associated with slower AIDS progression. These results indicated strong interaction with KIR3DL1 may promote NK cell education and enhance their effector function. HLA-C1 alleles can also bind to NK cells with different affinity [119]. Donors who were homozygous for HLA-C\*07 produce higher level of IFN-γ compared to donors expressing either HLA-C\*01, 03\*,\*08, \*1402, or 1403\*. Since the strength of the interaction between KIR and HLA is critical for NK cell education and function, it is tempting to speculate that during allo-HSCT certain HLA combination may enhance NK cell activity and contribute to better outcome of HSCT. The differing KIR/HLA affinity also can be generated by KIR allelic differences. KIR3DL1 is one of the most polymorphic KIR genes with more than 60 alleles [120]. KIR3DL1\*01052 binds to Bw4 more firmly than KIR3DL\*007 due to their high-expression level [119]. Although the surface expression of KIR3DL1\*002 is similar to KIR3DL\*007, the former generates stronger interaction with Bw4 than the latter [121]. KIRrecognizing HLA-C alleles also have varying interactions. KIR2DL1 has been shown to be the strongest KIR for HLA-C2 followed by KIR2DL2 then KIR2DL3 [122]. These differing affinities may explain the beneficial effect of haplotype B from donors as these donors are homozygous for KIR2DL2 which may account for the generation of NK cells with enhanced function.

grafts from KIR A/A genotype donors into KIR B recipients resulted in poorer survival in HLAmatched T-cell-replete sibling transplants [109]. In contrast, Stringaris et al. reported transplants from KIR B haplotype donors including genes for KIR2DS1, KIR3DS1, and KIR2DL5 improved overall survival and prevented relapse specific for AML patients with T-cell depletion during HLA identical siblings HSCT [110]. A larger cohort was performed to further evaluate the effect of KIR B haplotype in AML patients. The results indicated that grafts from donors with KIR B haplotype significantly reduced the risk of relapse and improved overall survival [111]. A follow-up study expanded numbers of AML patients with T-cell-depleted unrelated donor transplants and found that donor KIR B haplotype-mediated relapse prevention was enhanced in recipients who have HLA-C1 alleles rather than C2 homozygous recipients [36]. Michaelis SU et al. reported transplants from KIR B haplotype donors resulted

Collectively, it is logical to hypothesize that the presence of activating KIR in donor KIR B haplotype is favorably associated with better transplant outcome. However, it is difficult to evaluate the contribution of each activating KIR to relapse protection and better overall survival due to their unknown ligands. KIR2DS1 has been proved to interact with HLA-C group 2 alleles and involved in the killing of leukemic blasts from HLA-C2 patients. The interaction between KIR2DS1 and HLA-C2 was found to overcome NKG2A mediated inhibitory signals in vitro [113, 114]. These results suggested KIR2DS1-mediated activating signal may break the barrier of NKG2A and exert positive effect during HSCT. Venstrom et al. investigated the role of KIR2DS1 from donor in unrelated HSCT in 1277 patients with AML and found that KIR2DS1 expression in donor mediated a significant GvL effect [114]. Notably, no beneficial effect was observed when donors were HLA C2 homozygous as in those donors, NK cells expressing KIR2DS1 cannot be educated and acquire effector function. Therefore, at least one copy of HLA-C1 in donor is needed for KLR2DS1-dependent GvL effect. In addition, KIR3DS1 was also found to associate with lower rate of relapse and infection. Mancusi A et al. reported KIR3DS1 was associated with reduced infection mortality [115]. Further studies are still needed to identify ligands for activating KIR so as to confirm the effect of each activating

The intensity of interaction between KIR and HLA may also influence the alloreactivity of NK cells during allo-HSCT. HLA alleles have been shown to bind NK cells with different affinities. The HLA class I molecules are highly polymorphic, and HLA-B alleles can be divided into HLA-B alleles with an isoleucine at position 80 (80I) and with a theonine (80T). The affinity between certain HLA-B alleles with an 80I and KIR3DL1 was found to be much stronger than those with an 80T [116, 117]. The impact of this different affinity was tested in HIV patients [118]. Presence of 80I but not 80T in HIV patients associated with slower AIDS progression. These results indicated strong interaction with KIR3DL1 may promote NK cell education and enhance their effector function. HLA-C1 alleles can also bind to NK cells with different affinity [119]. Donors who were homozygous for HLA-C\*07 produce higher level of IFN-γ compared to donors expressing either HLA-C\*01, 03\*,\*08, \*1402, or 1403\*. Since the strength of the interaction between KIR and HLA is critical for NK cell education and function, it is tempting to speculate that during allo-HSCT certain HLA combination may enhance NK cell activity and contribute to better outcome of HSCT. The differing KIR/HLA affinity also can be

in better HSCT outcome in ALL patients [112].

136 Natural Killer Cells

KIR gene on allo-HSCT outcome.

Besides using "missing self" model to predict alloreactivites of NK cells as described above, "missing ligand" model has also been used to speculate alloreactivites. According to this model, NK cell alloreactivity may also be observed when the KIR on the surface of donor NK cells cannot recognize the ligands either from donor or recipient [123, 124]. As we have mentioned, these "unlicensed" NK cells are hyporesponsive in situ but can have potential to acquire effector function in recipients. Thus, NK cell alloreactivity can occur even after autologous HSCT. This model is supported by the observation that most individuals have 3 inhibitory KIR, but only 1 or 2 corresponding HLA KIR ligands are expressed on the surface of their own cells [125, 126]. Several comparative analyses have been performed to test the prediction rate of this model. Some groups found the selection of donors based on this model indeed was associated with better outcome, while others reported opposite results [127–129]. These variable results may be due to the complexity and variables of HSCT.
