Natural killer cells: alloreactive effects in hematopoietic stem cell transplantation


  • 5D-S44-01

Katharine C. Hsu, Adult Allogeneic Bone Marrow Transplantation Service, Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY 10065, USA


Background and Objectives  Natural Killer (NK) cells are capable of recognizing target cells altered by pathologic states such as viral infection and malignant transformation. The NK cell’s capacity to recognize and kill abnormal cells while sparing normal cells relies on the integration of activating signals and tolerance to cells expressing self-MHC molecules. Our understanding of how NK receptors mediate these signals and control NK cell function has advanced to the point where models of NK biology can be tested and revised in the clinical setting of hematopoietic stem cell transplantation.

Materials and Methods  This review article highlights several of the seminal contributions to the field of NK biology and their relevance to allogeneic and autologous hematopoietic cell transplantation.

Results and Conclusions  Inhibitory NK receptors are critical for achieving tolerance to cells expressing self-MHC molecules and recognizing target cells lacking self-MHC class I molecules. Activating NK receptors can circumvent NK tolerance and directly signal activation. Both inhibitory and activating receptors play an important role in hematopoietic stem cell transplantation outcome for hematologic and solid tumors. Incorporation of NK receptor genotyping in donor selection algorithms may benefit recipients of both HLA-matched and -mismatched allografts.

NK cell receptors mediate tolerance to self and control effector function

Natural killer (NK) cells are lymphocytes whose ability to identify and eradicate virally infected and transformed cells fulfill a critical role in the innate immune system. In contrast to T and B lymphocytes that employ rearranged antigen-specific receptors, NK cells express germline-encoded receptors with inhibitory and activating functions. The principal NK cell receptors in humans are the killer cell immunoglobulin (Ig)-like receptors (KIR), whose 14 genes and 2 pseudogenes are members of a polymorphic receptor family within the immunoglobulin superfamily. In mice, the main NK receptors are members of the Ly49 C-type lectin family. In both humans and mice, NK cells also express a lectin-like receptor heterodimer comprised of the CD94 molecule coupled with members of the NKG2 family. Several other receptor types with activating function, such as the natural cytotoxicity receptors and NKGD, are also expressed on the cell surface.

Several inhibitory NK cell receptors recognize MHC class Ia molecules, while CD94/NKG2 recognizes HLA-E, which presents signal peptides from MHC class Ia molecules, thereby achieving recognition of MHC class Ia expression indirectly [1]. In humans, the three inhibitory KIR groups with identified class I ligands include KIR2DL2/2DL3, which recognize HLA-CAsn80 alleles (HLA-C1 group), KIR2DL1 which recognizes HLA-CLys80 alleles (HLA-C2 group), and KIR3DL1 which recognizes HLA alleles with the Bw4 epitope. Constitutively expressed on healthy cells, class I ligands of inhibitory receptors are often downregulated during pathologic states such as viral infection or transformation, rendering infected or transformed cells more noticeable to NK cell surveillance through their lack of engagement of class I-specific receptors [2,3]. Critical to inhibitory receptor signalling is the immunoreceptor tyrosine-based inhibitory motif (ITIM), which resides in the cytoplasmic tails of the inhibitory receptors and associates with the cytoplasmic tyrosine phosphatase SHP-1, but may also recruit other signalling molecules [4].

Inhibitory human KIR receptors and mouse Ly49 receptors are functionally analogous, despite being structurally distinct, indicating that similar immune pressures resulted in convergent evolution of the two receptor families [5]. In addition to their recognition of MHC class I ligands and ITIM-mediated inhibition, common features to the inhibitory KIR and Ly49 receptor families include constitutive expression on resting NK cells, stochastic expression on overlapping NK subsets [6] with expression of none to several inhibitory receptors by an individual NK cell [7], encoding by germline clusters of polymorphic genes [8,9], and close sequence homology to ITIM-negative activating receptors. The activating KIR and Ly49 receptors share strong sequence homology in the extracellular regions to the corresponding inhibitory receptors but have short cytoplasmic tails and associate with DAP12 or other immunoreceptor tyrosine-based activation motif (ITAM)-containing signalling molecules [1]. The ligands for many activating receptors remain unknown, although it is known that the activating KIR2DS1 can recognize the same HLA class I molecules as its inhibitory KIR2DL1 counterpart. Other non-KIR activating receptors, such as NKG2D, recognize MHC class I-like molecules, whose constitutive expression can become upregulated in response to stress or other stimuli, resulting in enhanced NK cell activation [10].

Inhibitory and activating receptor signals jointly control the NK cell response to a target cell. Detection of missing self-MHC on a target cell leads to a decreased inhibitory signal, allowing domination of the activation signal and triggering cytokine production (IFN-γ, TNFα, and GM-CSF) and/or a cytotoxic response through granule exocytosis or through the Fas-Fas ligand pathway. Less commonly, NK recognition of a highly expressed activating ligand on a target cell can override inhibitory signals and result in NK activation despite normal MHC expression on the target cell.

The ability of the NK cell to identify and kill virally infected and malignant cells while sparing normal cells remained poorly understood until the late 1980s and the introduction of the ‘missing self’ hypothesis [11], which proposed that downregulation of self-MHC class I molecules during viral infection or malignant transformation triggers NK activation. Recent findings have illuminated how NK cell recognition of self-MHC molecules dictates NK response while avoiding auto-aggression. In the ‘NK cell licensing’ model, an NK cell that has engaged self-MHC via MHC-specific inhibitory receptors is ‘licensed’ to be responsive to stimuli via activation receptors, while NK cells that do not engage self-MHC are termed ‘unlicensed’ [12,13]. By this model, there are two general populations of NK cells, both tolerant to cells expressing self-MHC: NK cells licensed for effector function through self-specific MHC class I inhibitory receptors but which maintain self-tolerance through direct inhibition via these same receptors by self-MHC; and unlicensed NK cells, which lack self-specific MHC class I inhibitory receptors, cannot engage self-MHC and are functionally hyporesponsive.

Studies examining the responses of murine NK cells upon target cell-free antibody cross-linking provided initial evidence for the licensing model [12]. Naïve NK cells stimulated with plate-bound antibodies against the NK cell activation receptor NK1.1 expressed on all NK cells in C57BL/6 (H2b) mice were assessed at the single cell level for activation. Only NK cells expressing an inhibitory Ly49 receptor specific for self-MHC produced IFNγ upon ex vivo plate-bound antibody stimulation [12]. Therefore, Ly49A+ NK cells from MHC congenic or MHC transgenic mice expressing H2Dd, a known ligand for Ly49A, were “licensed” and produced IFNγ upon stimulation. In contrast, in mice lacking MHC ligand for Ly49A, such as mice with the H2b haplotype, Ly49A+ NK cells were unlicensed and produced significantly less IFNγ. Evaluation of NK cytotoxicity provided consistent results. Taken together, these data suggest that engagement of an inhibitory receptor with its cognate MHC class I ligand, expressed as self, endows an NK cell with functional competence.

A similar model of NK licensing also applies to human NK cells. Human CD56dim NK cells lacking inhibitory receptors for self-MHC display a hyporesponsive phenotype, exhibiting reduced cytokine production and cytotoxicity in response to stimulation with MHC class I-negative target cells as well as anti-CD16 cross-linking and antibody-dependent cell-mediated cytotoxicity (ADCC) [14,15]. In contrast, NK cells expressing inhibitory receptors for self-HLA-B and self-HLA-C ligands exhibit robust response to class I-negative target cells. Moreover, a dose effect of inhibitory receptors and their cognate class I ligands exists: NK cells expressing more than one inhibitory receptor for self-MHC demonstrate higher effector function [15], and inhibitory KIR-expressing NK cells from individuals homozygous for the cognate HLA class I ligands demonstrate more potent effector function [16]. It should be noted here that the ‘disarming’ model of NK education is an equally viable alternative to the NK licensing model and proposes that in the setting of chronic stimulation through activating receptors, the lack of MHC class I inhibition of non-self MHC-specific receptors directs NK cells to an anergic or exhausted state [17]. For the purposes of this review, however, “licensed” is used to describe those NK cells that exhibit an inhibitory receptor for self-MHC class I.

Regardless of which model most accurately reflects the forces shaping the functional repertoire, a hierarchy of effector function clearly exists among the NK repertoire, where NK cells lacking all inhibitory receptors are largely functionally hyporesponsive; cells expressing the inhibitory heterodimer CD94/NKG2A alone display modest functional competence compared to cells with a single inhibitory Ly49 or KIR receptor for self-MHC, which display slightly higher functional competence; and cells with more than one inhibitory for self-MHC display still higher functional potency [15]. In humans, the resting NK repertoire achieves a state of ‘intermediate inhibition’, with cells of lowest and highest functional potency present in lower frequecies compared to cells of intermediate potency [15,18]. Data from mouse models also demonstrate a correlation between various Ly49-MHC combinations and NK functional potencies, leading investigators to describe a repertoire ‘tuned’ for diversity of response [19,20].

Most recently, it has become clear that the significant population of ‘unlicensed’ cells in humans and mice are capable of response under specific circumstances. While responding poorly to stimulation with MHC class I-negative target cells or plate-bound antibody, they can still produce IFNγ when stimulated with PMA + ionomycin with similar frequency to self-MHC-specific NK cells, indicating that they retain the capacity for effector function. In vitro culture with cytokines can circumvent the effects of licensing, inducing unlicensed populations to respond to stimulation [12,21]. Similar effects on NK cells occur in vivo during the inflammatory states of viral infection, where NK cells in mixed wild type: beta(2)-microglobulin-deficient [β2m(−/−)] chimeric mice are initially tolerant of MHC class I-deficient host cells but readily reject host β2m(−/−) bone marrow after infection with murine CMV [22]. Engagement of specific activating receptors can also circumvent the licensing effects: in mice, unlicensed NK cells expressing the Ly49H receptor but lacking self-specific inhibitory receptors are selectively capable of clearing CMV-infected cells [23]. In humans, NK cells expressing the activating KIR2DS1 are capable of cytokine response and cytotoxicity to target cells expressing the specific HLA-C2 ligand, if the same NK cell lacks a self-MHC-specific inhibitory receptor KIR2DL1 [24,25]. Taken together, these studies provide evidence that NK effector function need not necessarily require engagement of self-MHC by self-MHC-specific inhibitory receptors on the NK cell.

NK cell alloreactivity and hematopoietic stem cell transplantation

NK cells have long been recognized to play an important role in allogeneic hematopoietic stem cell transplantation (HSCT). NK cells are the first lymphocyte subset to reconstitute the peripheral blood immediately following transplantation and are important in the suppression of graft-versus-host disease, promotion of bone marrow engraftment and mediation of a graft-versus-leukaemia effect. Infusion of higher CD56dim/CD56bright NK ratios in the stem cell allograft and higher NK cell counts immediately following transplant correlate with improved HSCT outcomes [26,27]. Our increasing understanding of how inhibitory KIR interaction with MHC class I ligands drives NK function has clarified the molecular mechanisms underlying previously described clinical correlations in the HLA-mismatched transplant setting. Meanwhile, studies correlating KIR immunogenetics with transplant outcomes continue to provide additional insights into the impact of activating NK receptors, about which comparatively little is known and whose ligands remain largely obscure.

The HLA-mismatched or more specifically KIR-ligand mismatched allogeneic HSCT provides a suitable clinical setting for the purposes of examining the clinical effects of ‘missing self’ by licensed NK cells from the stem cell donor. In this setting, donor-derived NK cells infused into patients with leukaemia lacking the HLA class I KIR ligands present in the donor recognize the absence of these donor-defined ‘self’ ligands on patient target cells, leading to a lack of engagement of inhibitory receptors on the NK cells and therefore lack of inhibition of otherwise licensed NK cells, enhanced leukaemic clearance and improved patient outcomes. In seminal studies of high-risk leukaemia patients undergoing HLA haplotype-disparate HSCT, a missing-self relationship between donor and recipient HLA genotypes predictive of donor NK alloreactivity was associated with lower incidence of relapse, and improved event-free survival [28–30]. Importantly, the authors noted that donor-derived alloreactive NK clones could only effectively kill myeloid leukaemia target cells, such as acute myelogenous leukaemia (AML) and chronic myelogenous leukaemia (CML), but could not reliably kill acute lymphoblastic leukaemia (ALL) cells, identifying a disparity in NK sensitivity between the leukaemia types. Studies of the missing-self effect in other HLA-mismatched transplant populations have yielded variable results [31,32], however, indicating that the impact of missing self on HSCT outcome may be most evident under specific treatment conditions, such as in vivo T-cell depletion.

Donor–recipient HLA class I mismatching may not be necessary to achieve an NK cell-mediated effect in HSCT. Several studies have demonstrated that NK effects can be beneficial even in HLA-matched HSCT, specifically in the circumstance where donors possess inhibitory KIR for which neither the donor nor the recipient exhibits the relevant class I ligand [33–36]. In a study of 178 patients with leukaemia who received T-cell-depleted allografts from HLA-identical siblings, AML and myelodysplastic patients lacking HLA class I ligand for donor inhibitory KIR had significantly lower relapse and improved overall and disease-free survival [33]. In this clinical setting, donor-derived licensed NK cells by definition could not become activated by the recognition of ‘missing self’ in the HLA-matched recipient. Instead, the clinical data suggested that NK cells expressing non-self-specific KIR (classically defined as ‘unlicensed’) were responsible for the improved clinical outcome, by responding to ‘missing ligand’ in the patient. Yu et al. have demonstrated that in the first 3 months following HSCT, unlicensed NK cells with inhibitory KIR for non-self HLA do indeed exhibit effector function, displaying cytokine response and cytotoxicity towards tumour target cells lacking the class I ligand for the inhibitory KIR [37]. It is possible that HSCT may provide an in vivo inflammatory cytokine milieu conducive for the recruitment of unlicensed NK cells to the functional repertoire. The cytokines responsible for inducing activation of unlicensed cells in vivo remain unknown but may be elaborated only in specific transplant conditions, accounting for inconsistencies identifying the missing-ligand effect in other allogeneic HSCT populations [30,38,39].

If normally hyporesponsive NK cells can circumvent the effects of licensing and become reactive against targets lacking the HLA class I ligand in HLA-identical allogeneic HSCT, comparable cytokine conditions in patients undergoing autologous HSCT may have similar outcomes. Indeed, recent studies investigating the ‘missing ligand’ NK effect in autologous HSCT have revealed a significant NK effect against solid tumours, specifically neuroblastoma [40,41]. Taken together, these studies underscore the importance of innate immunity in both haematologic and solid tumour control and imply that KIR-HLA immunogenetics represents a prognostic marker for specific malignancies.

Finally, activating KIR genotypes may help to guide donor selection, not only for allogeneic HSCT but also for adoptive NK cell therapy. Patients with donors exhibiting genotypes with the activating receptors KIR2DS1 and 2DS2 have been reported to have decreased relapse [42], and KIR2DS1+ NK cells can mediate direct lysis of HLA-C group 2 homozygous leukaemic blasts [43]. Moreover, patients with AML who receive allografts from donors with a KIR B-haplotype, where KIR B-haplotypes are generally characterized by more than one activating KIR, experience higher relapse-free survival [44]. Furthermore, KIR3DS1 in the donor is associated with protection from grade II–IV GVHD [45]. Identification of the molecular ligands for the activating KIR will clarify the mechanisms underlying many of these correlative studies.


The function of an NK cell can be predicted by its inhibitory and activating receptors in the context of self-MHC and non-self MHC class I molecules. The clinical effects of NK activation are most evident in allogeneic HSCT, where licensed NK cells can behave according to ‘missing self,’ and unlicensed NK cells according to ‘missing ligand’. As more information accumulates on the importance of specific NK receptors on transplant outcome, NK receptor genotyping will become increasingly relevant for donor selection, not only for allogeneic HSCT but also for adoptive NK therapies in the treatment of malignancies.


None declared.