Influence of KIR and NK Cell Reconstitution in the Outcomes of Hematopoietic Stem Cell Transplantation

Natural killer (NK) cells play a significant role in immune tolerance and immune surveillance. Killer immunoglobin-like receptors (KIRs), which recognize human leukocyte antigen (HLA) class I molecules, are particularly important for NK cell functions. Previous studies have suggested that, in the setting of hematopoietic stem cell transplantation (HSCT), alloreactive NK cells from the donor could efficiently eliminate recipient tumor cells and the residual immune cells. Subsequently, several clinical models were established to determine the optimal donors who would exhibit a graft-vs. -leukemia (GVL) effect without developing graft-vs. -host disease (GVHD). In addition, hypotheses about specific beneficial receptor-ligand pairs and KIR genes have been raised and the favorable effects of alloreactive NK cells are being investigated. Moreover, with a deeper understanding of the process of NK cell reconstitution post-HSCT, new factors involved in this process and the defects of previous models have been observed. In this review, we summarize the most relevant literatures about the impact of NK cell alloreactivity on transplant outcomes and the factors affecting NK cell reconstitution.

As the first reconstituted lymphocyte subset after transplantation (32,33), NK cells play a critical role in controlling early relapse and infections. They also possess the ability to eliminate recipient T cells and antigen-presenting cells (APCs), to prevent graft failure and GVHD (34)(35)(36)(37)(38) (Figure 2B). Three models were established historically in an attempt to optimize donor selection for HSCT based on KIR (Figure 2A). The Perugia group in Italy firstly proposed the donor-recipient KIR ligand-ligand model (also known as KIR ligand model) solely based on the HLA phenotype of the donor and recipient. The KIR ligand incompatibility in the GVH direction was defined as the absence in recipients of donor class I allele group(s) recognized by KIRs. Those authors observed that the HLA haplotype-mismatched transplants reduced the rejection and relapse rate and prevented GVHD in patients with acute myeloid leukemia (AML) (36). Subsequently, the second model (named receptor-ligand model or missing ligand model) was raised by Leung et al. based on the compatibilities between the recipient HLA and donor inhibitory KIR. This model focused on donor KIR instead of donor HLA and could, therefore, be used in both HLA-matched and HLA-mismatched transplants. The results of that study suggested that the receptor-ligand model better predicted the risk of primary disease relapse, especially for lymphoid malignancies, compared with the ligand-ligand model (39). Subsequently, with a deeper understanding of KIR haplotypes, the third model analyzed and compared the KIR genotypes of different donors. Cooley et al. showed that unrelated donors with KIR-B haplotypes conferred a significant relapsefree survival (RFS) benefit to patients with AML undergoing T cell-replete HSCT (40). Based on the three models described above, numerous studies have been carried out to explore the impact of NK cell alloreactivity. Clinical results obtained from KIR ligand model, receptor ligand model and KIR haplotype and gene model were summarized in Tables 2-4, respectively. Nevertheless, the results were controversial, and several key questions remained regarding NK cell biology post-HSCT. What are the exact effects of NK cell alloreactivity on patients after HSCT? How do NK cells reconstitute post-HSCT and which factors may interfere with the reconstitution process? This review summarizes the latest literature on this important topic and offer some instructive hypothesis.

KIR AND TRANSPLANT OUTCOMES NK Cell Alloreactivity and GVHD
GVHD is an important complication of HSCT with high morbidity and mortality in which allogeneic donor immune cells are activated by APCs and then recognize and attack the host tissue (105). Removing donor T cells from grafts reduces the occurrence of GVHD, while it also elevates the risk of graft failure and disease relapse (106)(107)(108).
As another component of immune cells, previous murine studies suggested that adoptive transfer of interleukin-2 (IL-2)-activated SCID NK cells with donor bone marrow cells FIGURE 1 | Simplified genomic maps of KIR. Inhibitory KIR genes are color-coded in blue, activating KIR genes in orange, and pseudogenes in gray. KIR haplotype A has only one activating KIR gene: KIR2DS4, KIR B haplotype has fixable content of activating KIR genes. KIR haplotype could be further determined as Cen haplotype and Tel haplotype.
promoted engraftment in allogenic hosts with no signs of GVHD (109). Later, Asai et al. reported that hosts receiving MHCincompatible bone marrow and spleen cells (as a source of T cells) rapidly succumbed to acute GVHD, while hosts who additionally received IL-2-activated donor NK cells on day 0 experienced a significant improvement in survival because of the lower incidence of severe GVHD. They further demonstrated that that the protective effect on GVHD was dependent on the transforming growth factor-beta (TGF-β) and could be abrogated by an anti-TGF-β antibody (35). Moreover, Ruggeri et al. showed that pre-transplant alloreactive Ly49 (Ly49 receptors recognize major histocompatibility complex (MHC) class I molecules in mice, which is analogous to KIR in humans) ligand-mismatched donor NK cell transfusion successfully eliminated host tumor cells and protected against GVHD by depleting host APCs. In contrast, hosts receiving bone marrow grafts without NK cell infusion died of GVHD, and non-alloreactive Ly49 ligand matched NK cell infusion did not provide protection against GVHD (36). Consistently, subsequent studies also found that donor alloreactive NK cells suppressed GVHD by inhibiting T cell proliferation and activation (37,110). However, the protective role of NK cells in GVHD pathogenesis has also been challenged. Pre-clinical evidence from a xenogeneic model showed that an in vitro IL-2-activated human NK cell infusion promoted GVHD in SCID mice via the production of cytokines such as IFN-γ and tumor necrosis factor-α (TNF-α) (111,112). Accordingly, GVHD was inhibited after the administration of anti-IFN-γ and depletion of Poly I:C-activated NK cells in murine studies (113,114).
In patients with hematological malignancies, a purified (115,116) or cytokine-induced (117)(118)(119)(120)(121) donor NK cell transfusion was also well tolerated and seldom induced severe GVHD (grade III-IV acute GVHD or moderate-to-severe chronic GVHD). More recently, a pilot study suggested that, after haplo-HSCT, patients with refractory AML who received a donor NK cell infusion experienced a significantly lower grade II-IV GVHD than did those without NK cell infusion (122). In contrast, Shah et al. observed that patients who received a donor FIGURE 2 | KIR models (A) and NK cell-mediated killing (B). APC, antigen presenting cell. (A). Donor NK cell is tolerant to self because donor inhibitory KIR is inhibited by its cognate HLA ligand; donor NK cell might kill recipient cell because HLA ligand for donor inhibitory KIR presents in donor but absents in recipient (KIR ligand model); donor NK cell could kill recipient cell because recipient HLA ligand does not inhibit donor inhibitory KIR (receptor ligand model); donor NK cell could kill recipient cell because donor activating KIR is activated by recipient (KIR B haplotypes and KIR B genes). (B). Alloreactive donor NK cell could kill recipient leukemia cell to prevent relapse; it could kill recipient T cell to prevent graft rejection; and it could kill recipient APC to prevent GVHD.
It is not entirely clear why the reconstituted alloreactive NK cells were unable to prevent GVHD as the adoptively transferred NK cells. Studies have indicated that this discrepancy was probably attributable to the impaired function of early reconstituted NK cells. Shilling et al. first observed that a period of several months or even years was required for the recipient to reconstitute an NK cell repertoire resembling that of the donor (124). Vago et al. also suggested that the NK cells that were reconstituted early after transplantation were immature and exhibited compromised cytotoxicity (125). In addition, NK cell reconstitution is affected by graft composition. Patients receiving more T cells in grafts experience a faster T cell reconstitution (126,127), while the absolute number of reconstituted NK cells and KIR expression are impaired by the co-grafted T cells (127)(128)(129)(130). Other than NK cells, nearly 5% of CD8 + T cells, 0.2% of CD4 + T cells, and 10% of γδ T cells in the peripheral blood also express KIRs (131)(132)(133). Therefore, it is possible that the potential beneficial effects of alloreactive NK cells are overwhelmed by the strong alloreactive T cell response. In addition, it was observed that NK cells generated more IFN-γ in the presence of T cells in grafts, leading to a higher occurrence of acute GVHD (aGVHD) (130). Moreover, post-transplant immune suppression also exerted negative effects on NK cell reconstitution (134,135).
Regarding specific genotypes, some studies have reported that KIR haplotype B donors afforded a significantly reduced risk of GVHD (60, 63, 86,96). Consistent with these findings, Sivori et al. suggested that donor NK cells expressing KIR2DS1 were efficient in killing allogenic dendric cells in the setting of haplo-HSCT, thus leading to a better GVHD control (136). However, several studies also found that donors with KIR-B/x led to higher GVHD occurrence in recipients compared with donors with A/A, probably because of the more potent production of IFN-γ by alloreactive NK cells (40,45,76,77,94,95).
Other factors, such as HLA mismatch, disease type, patient age, GVHD prophylaxis, and graft source, were also reported to interfere with GVHD occurrence in these studies (44, 45, 63, 66, 87,92,93,104). Collectively, the manner in which the reconstituted NK cells affect the risk of GVHD remains largely unknown, and the relationships between NK and T cells during the initiation and process of GVHD warrant further investigation.

NK Cell Alloreactivity and Infection
Infections are especially challenging for patients after HSCT because of the immunological derangement caused by multiple factors, including an intensive conditioning regimen, immunosuppressive agents, and other complications, such as GVHD (137,138).
Several studies have reported that patients receiving KIR ligand-mismatched transplants are more vulnerable to infections. Schaffer et al. first reported that KIR ligand mismatch was associated with an increased infection-related mortality (43). Similarly, results from Zhao et al. showed that recipients from the KIR ligand-mismatched group experienced a significantly higher cytomegalovirus (CMV) reactivation rate. Moreover, the percentage of interferon-gamma (IFN-γ)-expressing NK cells in the peripheral blood was significantly higher in the KIR ligand matched group 30 and 100 days post-HSCT compared with the KIR ligand-mismatched group (53). The higher level of IFN-γ secretion from the NK cells might trigger Th1 immune responses, antigen presentation cell activation, and macrophage killing (7,8), leading to lower infection rate. While, KIR ligand mismatch may increase the risk of infection by eliminating recipient APCs by donor alloreactive NK cells (36).
Many studies have found that KIR-B genes protect patients with HSCT against infections and most of them were predominantly T cell replete (TCR) transplants (81,84,87,96,139,140  of KIR2DS1 to HLA-C2 triggered pro-inflammatory cytokine production by alloreactive NK cells (51). Moreover, without a cognate ligand (HLA-C1) in recipients, donor KIR2DS2 was associated with a higher CMV reactivation rate after HLAidentical sibling HSCTs (81). Apart from CMV reactivation, the incidence of bacterial infections was also reduced when patients had KIR-B/x donors (87). In contrast with previous results, KIR2DS2 gene and Cen-B/x donors related to a higher incidence of CMV reactivation and infection-related mortality in TCD transplants (53, 100). The reasons for these differing results may be due to the different graft composition. As previously described, NK cells generate more IFN-γ in TCR transplants, which may benefit the infection control (130). Of notice, the activating KIR targets outside of HLA are largely unknown, and these clinical observations still need to be confirmed by definitive functional analysis in the future.

NK Cell Alloreactivity and Relapse/Survival
Primary disease relapse remains the main obstacle that hampers the long-term survival of patients with hematological malignancies. Previous experience showed that adoptive transfer of autologous NK cell for patients with tumors was safe but inefficient (141)(142)(143)(144)(145), probably because autologous NK cells could not overcome the inhibition mediated by   tumor cells expressing self-HLA. In contrast, allogenic (117), especially haploidentical, donor NK cell infusion demonstrated wide prospects in the salvage treatment (115,120,121) and prophylactic treatment (118,119) of patients with hematological malignancies. In allo-HSCT, whether the reconstituted alloreactive NK cells prevent the disease relapse remains controversial. In HLA-mismatched transplants, the Perugia group first observed that, in the context of T cell depletion, high stem cell dose, and absence of post-transplant immune suppression, KIR ligand mismatch reduced the risk of relapse and markedly improved survival in patients with AML, but not in those with acute lymphoblast leukemia (ALL) (36). This protective effect on relapse or survival was supported by many clinical studies (42, 44, 47, 51, 54), especially in myeloid disease (44, 47, 51) and transplants with TCD grafts (42, 47, 51, 54). However, conflicting results stemmed from many studies that failed to replicate these results (39, 46, 58, 102), and some even reached the opposite conclusions (41, 43, 45, 48, 50, 55).
Studies using the receptor-ligand model including HLAmatched donor-recipient pairs also reported conflicting results. Leung et al. first reported that the receptor-ligand model was more accurate than the KIR ligand model when predicting the risk of relapse, especially for lymphoid malignancies. Moreover, the potency of the relapse protection positively correlated with the number of receptor-ligand mismatch pairs (39). Subsequently, the protective effect of receptor-ligand mismatch has been confirmed by many investigations (58, 59, 66, 69, 71,73,76,79,80). Moreover, a survival advantage was also observed in patients with receptor-ligand mismatch compared with receptorligand matched pairs (59, 66, 69-71, 73, 76, 79). However, several other studies described opposite results (63, 64, 75, 77, 78). Of notice, two studies from Japan observed that the lack of the HLA-C2 ligand for donor inhibitory KIR afforded relapse protection in patients with AML and chronic myeloid leukemia, but increased the relapse rate in patients with ALL (73,75). To date, no plausible explanation has been put forward for this disparity in relapse.
In contrast to the controversial results described above, transplantations from KIR haplotype B donors achieved greater agreement. Cooley (97). At the genetic level, the KIR2DS2 gene, which is located on the Cen-B motif (72,76,79,90,92,96), and the KIR2DS1 gene, located on the Tel-B motif (51, 72,82,98), were found to be related to a decreased relapse rate or an improved survival. However, several studies found that Cen-B donors indicated a lower OS (56, 70, 100). Meanwhile, Verneris et al. did not find any association between transplant outcomes and NK cell alloreactivity or KIR gene content in pediatric patients with acute leukemia (46).
Recently, Krieger et al. developed a scoring system, in which interactions of multiple KIR genes and HLA ligands were quantitatively analyzed. This comprehensive method raised an improved strategy to select a donor and exhibited great potential in the future (146).
Collectively, it is still controversial to determine an optimal donor who exhibits the best NK cell function using the three established KIR models. A better knowledge of NK cell reconstitution after HSCT may promote a better understanding of how NK cells affect the transplant outcomes in these patients. More in-depth studies focusing on "functional changes in NK cells" rather than "match or mismatch" may help us get closer to an optimal donor.

Maturation and Differentiation of NK Cells
NK cells are derived from the CD34 + hematopoietic stem and precursor cells in the bone marrow, which then migrate to the periphery (147). Recent evidence suggested that not only the bone marrow, but also secondary lymphoid tissues contribute to the development of NK cells (148). According to the surface expression of CD56, NK cells could be divided in two main subtypes: CD56 bright and CD56 dim NK cells. CD56 bright NK cells exist mainly in lymph nodes and tonsils, while CD56 dim NK cells, the more mature subset transformed from CD56 bright NK cells, are dominant in the peripheral blood (7,147,149,150). CD56 bright and CD56 dim NK cells are equipped with distinct functions. The former population responds rapidly to interleukin-mediated stimulation with proliferation and cytokine secretion, while the latter population displays higher cytolytic capacity and lower proliferation (7,8,149). During the process of maturation, CD94/NKG2A is the first receptor that is expressed on immature NK cells. Together with the downregulation of CD56 expression, NK cells upregulate CD16 expression, lose NKG2A, and acquire KIR receptors. Finally, a subset of CD56 dim cells continue to differentiate and express CD57, together with an increased KIR expression and a completely abolished proliferative ability (150,151).
In HSCTs with post-transplant cyclophosphamide (PT-Cy) as GVHD prophylaxis, NK cells experience two waves of expansion. After graft infusion, peripheral NK cells and T cells (mainly mature cells from the donor) were detectable at very low levels. PT-Cy administration results in a further decrease in T cells and NK cells, and NK cells are barely detectable in the peripheral blood. Subsequently, the reconstituted NK cells gradually recover and express high levels of CD56 and NKG2A. Around 60 days after transplantation, the KIR expression returns to normal. The expression of CD56 and NKG2A gradually decreases and becomes stable at 9-12 months post-transplantation. Other receptors expressed on NK cells, such as DNAM-1and 2B4, also require several months to return to normal (152). In summary, post-transplantation NK cell reconstitution is a longterm process (124,125,152).

KIR Education: From Anergic to Responsive
As described earlier, the random combination of KIR receptor and HLA ligand can exist in healthy individuals. However, the autoimmune attack is inhibited because each NK cell expresses at least one self-inhibitory receptor. To avoid autoreactivity, NK cells must undergo an education process: NK cells expressing inhibitory KIR for self-HLA ligand (self-KIR) are educated, which means that these cells can be inhibited by self-inhibitory signals and become alloreactive against self-HLA-deficient targets. In contrast, NK cells expressing an inhibitory KIR that lacks a self-HLA ligand (non-self KIR) are uneducated, which means that they are tolerant to the self but also to infected or malignant cells (19,21).
In the last decades, studies on KIR education have much extended our knowledge of NK cell function. After transplantation, most reconstituted NK cells express a donorlike KIR repertoire that is significantly different from that of recipient NK cells prior to transplantation (124,151). Therefore, reconstituted NK cells expressing donor KIR may exert alloreactivity in recipients, or become anergic, as recipients may not present the cognate HLA (Figure 3). Foley et al. and Björklund et al. observed that reconstituted NK cells with nonself KIR remained tolerant, while those with self KIR acquired better functions after transplantation (65, 153). However, Yu et al. reached the opposite conclusion that alloreactive NK cells broke the self-tolerance and displayed functional capacities in the first 3 months, then gradually acquired self-tolerance by day 100 post-transplantation (154). Rathmann et al. also suggested that alloreactive NK cells were increased in the peripheral blood and exhibited a GVL effect in the early period after transplantation (155). One possible explanation for this observation is that the infusion of a megadose of donor CD34 + cells may create a transient donor dominant HLA environment in recipient bone marrow, and the early reconstituted NK cells expressing nonself KIR for the recipient may become educated by donor HLA and acquire functions (156). After migration to a recipientdominant environment, reconstituted NK cells may gradually lose their responsiveness.
In murine studies, it was observed that mature NK cells from major histocompatibility complex (MHC) class I-sufficient mice become hyporesponsive after transfusion into MHC class I-deficient mice. Conversely, anergic NK cells from MHC class I-deficient mice acquired functions after exposure to the MHC class I-sufficient environment (157,158). Using a murine cells from KIR3DL1 + donors were transfused to B27 Tg + and Tg − mice, respectively. A functional analysis suggested that the most cytotoxic responsive cells were KIR3DL1 + NK cells from Bw4 + donors and developed in B27 Tg + mice (Bw4 + donors and Tg + mice), while the leastresponsive cells were KIR3DL1 + NK cells from Bw4 − donors and developed in Tg − mice (Bw4 − donors and Tg − mice). Recipients with the other two combinations (Bw4 + donors and Tg − mice and Bw4 − donors and Tg + mice) displayed a medium level of responsiveness. The stepwise escalation of NK cell responsiveness suggested that both the donor and recipient MHC environments are critical for the maintenance and adjustment of NK cell education (159).
Recently, the Nowak team proposed that inhibitory KIR (iKIR)-HLA pairs could predict the post-HSCT NK cell education status, i.e., donors presenting cognate HLA for donor iKIR and recipients lacking it predict a downward education level; in contrast, recipients presenting cognate HLA for donor iKIR and donors lacking it predict an upward education level. Those authors found that the decrease in iKIR-HLA pairs posttransplantation is associated with a higher relapse and poorer survival (160)(161)(162), indicating that reconstituted NK cells acquire better functions after interaction with more cognate HLA class I ligands in recipients. Zhao et al. also observed that, when the donors and recipients expressed three major HLA ligands (HLA-C1, C2, Bw4), patients with AML and myelodysplastic syndrome (MDS) experienced the lowest relapse rate, and NK cells expressing three inhibitory receptors exhibited the greatest cytotoxicity and cytokine responsiveness against K562 targets (163). Based on the findings described above, it is likely that three factors (donor KIR, donor HLA, and recipient HLA) all contribute to the variation in NK cell function. Therefore, the KIR ligand and receptor-ligand models, which only take two factors into account, may not accurately predict donors that exhibit the greatest NK cell function post-transplantation.

Factors That Affect NK Cell Reconstitution
Although CMV reactivation suggests an immune-compromised state, patients who experienced CMV reactivation had a lower relapse rate or better survival (70,98,99,164). This protective effect might be attributed to the rapid maturation of NK cells. During CMV reactivation, NK cells that express NKG2C rapidly expand and continue to increase for 1 year (165). The number of CD56 dim NK cells in the peripheral blood, their KIR expression, and IFN-γ production in response to K562 cells were also elevated in patients who developed CMV reactivation (165)(166)(167)(168)(169)(170)(171)(172)(173). Furthermore, nearly 60% of NKG2C + NK cells achieved complete differentiation and expressed CD57 after CMV reactivation. These cells were termed memory-like NK cells and could be detected long after the primary CMV infection, offering a long-lasting protection (147,166). In contrast, for non-CMV-infected patients, a higher proportion of NKG2A + NKG2C − KIR − NK cells in the peripheral blood indicates a slow NK cell maturation. Interestingly, CMV antigen exposure to recipients also leads to an increased frequency of NKG2C + NK cells, accompanied by increased KIR expression and decreased NKG2A expression (174).
As mentioned above, T cells in the graft impair the recovery of NK cells and KIR reconstitution (127)(128)(129)(130). A possible explanation for this observation is that T cells compete with NK cells for IL-15, a cytokine that regulates immune cell survival and development (175,176). Unlike ex-vivo TCD grafts, pretransplant anti-thymocyte globulin (ATG) administration results in partial T cell depletion. Two recent studies found that ATG administration promoted NK cell recovery and delayed the reconstitution of CD4 + and CD8 + T cells, while sparing the effector memory T and regulatory T cells (Tregs) (177,178). Compared with ATG, PT-Cy is more efficient in eliminating NK cells, with a higher residual ratio of CD4 + T cells and Tregs (179). Of note, several studies showed that T cells in the graft may contribute to a better NK cell function (153,180). Several studies reported that CD56 bright NK cells in lymph nodes could be stimulated by IL-2-producing T cells, resulting in NK cell maturation with higher IFN-γ secretion and cytotoxic functions (181,182).
The relationship between GVHD and NK cell reconstitution remains controversial. Previous studies demonstrated that GVHD correlated with an impaired NK cell reconstitution and KIR expression (183)(184)(185). Ullrich et al. found that CD56 bright NK cells were dramatically decreased in patients with GVHD, while CD56 dim NK cells, the more mature subtype, did not show significant changes (185). In addition, Hu et al. found that the NKG2A subset of CD56 dim NK cells was significantly decreased in patients with GVHD. Remarkably, a functional analysis showed that NKG2A + NK cells from GVHD and non-GVHD patients exhibited a comparable GVL effect. Furthermore, the co-culture of donor T cells with NKG2A + cells from non-GVHD patients suggested that NKG2A + NK cells inhibit T cell proliferation and activation, indicating that the decreased number of NKG2A + NK cells might be a cause, rather than a consequence, of GVHD (186). In addition, the administration of immunosuppressive agents could also affect immune recovery. Both Ullrich et al. and Giebel et al. suggested that steroid treatment, rather than GVHD, was related to the delayed NK cell reconstitution (184,187).

FUTURE DIRECTIONS
Numerous studies have found that alloreactive NK cells affect treatment outcomes. Although great progress has been made through both pre-clinical and clinical investigations based on the three KIR models, the controversy remains, especially regarding the benefits of KIR alloreactivity on relapse control. Recent findings showed that donor KIR, donor HLA, and recipient HLA environment all contribute to the variation of NK cell function. The newly proposed iKIR-HLA pair model needs to be further examined in the future.
NK cells, the lymphocytes that are reconstituted first after transplantation, could be negatively affected by the T cells in the graft. However, NK cell function could also be promoted through T-cell-mediated activation. The exact interactions between NK and T cells, as well as the strategy to trigger a potential synergistic NK and T cell effect remains to be investigated.
It is noteworthy that the protective role of NK cell alloreactivity in relapse protection mostly exists in myeloid disease; in fact, some studies even found that NK cell alloreactivity increased the risk of relapse for patients with lymphoid disease. The discrepancy between expressing ligands among different diseases and their binding affinity to KIR should raise more attention. In this way, we might identify which patients would benefit from the KIR-based donor selection.

CONCLUSION
In the early period after transplantation, reconstituted alloreactive NK cell may not directly influence GVHD occurrence, as it is immature and it could be affected by T cells and immunosuppressive agents. The compatibility between donor KIR and the recipient HLA ligand may protect patients from infection. In the late period after transplantation, the iKIR-HLA pair model may reflect the variation in NK cell function, and quantitative analysis of KIR-HLA interactions may provide more convincing results regarding relapse and survival.

AUTHOR CONTRIBUTIONS
YZ and HH designed. FG and YY wrote this paper. All authors revised and approved the final manuscript.