The role of natural killer cells in hematopoietic stem cell transplantation

Abstract

Natural killer (NK) cells are important elements of innate immunity, and a large body of evidence supports the significant role of NK in immune surveillance against infections and tumors. Regulation of cytotoxic activity is mediated through activating and inhibitory receptors expressed on the cell surface. NK cells are key players of allogeneic hematopoietic stem cell transplantation (allo-SCT), and previous studies showed the beneficial effect of NK alloreactivity in prevention of relapse, especially in the setting of haploidentical SCT. Biology of human NK cells is an area of active research. Exploitation of the molecular mechanisms regulating NK maturation, tolerance to self, and NK-mediated cytotoxicity will help in the development of innovative NK cell immunotherapy methods.

Key words::

• Alloreactivity
• autoreactivity
• haploidentical stem cell transplantation (haplo-SCT)
• leukemia
• natural killer cells (NK)
• relapse
• tolerance

Abbreviations
ADCC	=	antigen-dependent cellular cytotoxicity
AML	=	acute myeloid leukemia
ALL	=	acute lymphoblastic leukemia
allo-SCT	=	allogeneic hematopoietic stem cell transplantation
D	=	domains
FasL	=	Fas ligand
Flt3	=	FMS-like tyrosine kinase 3
GM-CSF	=	granulocyte-macrophage colony-stimulating factor
GVH	=	graft-versus-host
GVH	=	Dgraft-versus-host disease
haplo	=	haploidentical
HVG	=	host-versus-graft
IFN-γ	=	interferon-γ
IL-10	=	interleukin-10
IL-13	=	interleukin-13
IL-15	=	interleukin-15
ITAM	=	immunoreceptor tyrosine-based activating motif
ITIM	=	immunoreceptor tyrosine-based inhibition motif
KIR	=	killer immunoglobulin-like receptors
LFS	=	leukemia-free survival
LIR	=	leukocyte Ig-like inhibitory receptors
MIC-A, MIC-B	=	MHC class I chain-related peptides type A and B
MisMUD	=	mismatched unrelated donor
MRD	=	matched related donor
MUD	=	matched unrelated donor
NCR	=	natural cytotoxicity receptors
NK	=	natural killer
NKR	=	natural killer cell receptors
OS	=	overall survival
PFS	=	progression-free survival
pro-NK	=	NK progenitors
RFS	=	relapse-free survival
TAP	=	peptide transporter-associated antigen processing
TNFα, TNFβ	=	tumor necrosis factor-α and -β
TRAIL	=	TNF-related apoptosis-inducing ligand
TRM	=	treatment-related mortality

Key messages

• The decision ‘to kill or not’ is dependent on the net sum of the different positive and negative signals transmitted to natural killer (NK) cells through the contact with the target.

• Human killer immunoglobulin-like receptor (KIR) gene expression is a stochastic process, occurs during NK development, and results in a highly diverse KIR repertoire.

• NK cells receive the ‘license to kill’ only after transmission of inhibitory signals through the contact with autologous cells during the maturation process.

Introduction

Natural killers (NK) are cellular elements of innate immunity. They are the progenies of a common lymphoid progenitor that gives rise to all lymphocyte subsets. Rare cases of patients with selective deficiency of NK cells have been reported, and most of them died from overwhelming fatal infections during childhood (1–3). Clinical and laboratory data support the significant role of NK in immune surveillance against infections and tumors (4–6). Morphologically they have the appearance of large granular lymphocytes. Approximately 10%–15% of peripheral blood lymphocytes are NK cells. NK cells migrate from peripheral blood to secondary lymphoid organs, placenta, etc., and vice versa (7).

Immunophenotype of NK cells

NK cells are defined by the expression of a unique combination of surface antigens. Immunophenotype using flow-cytometry is a simple tool for NK cell characterization. However, no single marker with 100% specificity and sensitivity has been discovered so far. Instead, for NK identification, two different antigenic markers have been used. The NK characteristic expression profile is: negative surface CD3, and positive CD56 expression (CD3⁻/CD56⁺) (Figure 1) (8,9).

Figure 1. An example of immunophenotypic analysis of peripheral blood lymphocytes from a healthy donor. NK cells lack expression of CD3 and express CD56 (D1, left upper quadrant).

The population of NK cells is phenotypically and functionally heterogeneous. Two different subsets of NK cells have been identified, depending on CD56 and CD16 antigen expression density. The vast majority (90%) of circulating NK cells have high CD16 and low CD56 surface density expression respectively (CD56^dim/CD16^bright). CD56^dim NK cells display strong cytotoxic activity against tumor cell lines and have the capacity to kill targets through the mechanism of antigen-dependent cellular cytotoxicity (ADCC). Approximately 10% of circulating NK cells (and nearly 100% in secondary lymphoid organs and placenta) have high CD56 and low CD16 surface density expression respectively (CD56^bright/CD16^dim). CD56^bright NK cells produce abundant amounts of various cytokines after stimulation, while cytotoxic activity and ADCC are minimal or absent (10). However, cytotoxic activity can be induced after in-vitro culture in the presence of interleukin-2 (IL-2). Moreover, in-vitro data showed that activated CD56^bright cells secrete a vast array of immune-modulatory cytokines, such as interferon-γ (IFN-γ), tumor necrosis factor-α and -β (TNFα, TNFβ), interleukin-10 (IL-10), interleukin-13 (IL-13), granulocyte-macrophage colony-stimulating factor (GM-CSF), etc. (11,12). The immune-modulating function of the CD56^bright subset is not well established, and for many years it was debatable whether these cells represent a distinct subset of NK or if they are just in an immature stage of NK differentiation (13).

Development and maturation of NK cells

New insights regarding the nature of CD56^bright and CD56^dim NK cells arise from recent data. The results of recent studies showed that CD56^bright and CD56^dim are NK subsets in different stages of maturation, and under certain conditions CD56^bright differentiate into completely functional CD56^dim NK cells (13). The NK developmental pathway from CD34⁺ progenitors to mature circulating NK cells has not been fully elucidated yet.

NK progenitors (pro-NK) have been found in both peripheral blood and bone-marrow, as well as in secondary lymphoid tissues. Pro-NK cells under the influence of certain cytokines, such as c-Kit ligand, Flt3 ligand, interleukin-15 (IL-15), and ‘cell to cell contact’ differentiate into mature NK cells (7,14–16). Although recent data showed that NK cell maturation occurs in the parafollicular region of lymph nodes, it is still a matter of debate if the bone-marrow or lymph nodes is the main site where NK development and maturation takes place.

Natural killer cell receptors (NKRs)

Regulation of immune function is mediated through receptors expressed on the surface of effector cell receptors expressed on the surface of NK cells. Natural killer cell receptors (NKRs) are either activating or inhibitory, depending on the functional effect of the transmitted signal. The decision to kill is regulated by the interplay between activating and inhibitory signals received throughout receptor–ligand interaction. Finally, it is important to note that NKRs are not a unique feature of NK cells. NKRs are expressed on different subsets of T lymphocytes, and therefore they cannot be used as markers specific for NK cell identification (7). Human NK receptors are classified into four groups (Table I):

Table I. Natural killer cell receptors and cognate ligands.

Receptor
Inhibitory	Activating	Ligand
1: Killer immunoglobulin-like receptors (KIR)
KIR2DL1	KIR2DS1	HLA-C type 2 (Asn77-Lys80)
KIR2DL2/DL3	KIR2DS2	HLA-C type 1 (Ser77-Asn80)
	KIR2DL4^a	HLA-G
	KIR2DS4	Not known
KIR2DL5	KIR2DS5	Not known
	KIR3DS1	Not known
KIR3DL1		HLA-Bw4 type alleles
KIR3DL2		HLA-A3, A11 type alleles
2: C-type lectin receptors
CD94/NKG2A/B	CD94/NKG2C	HLA-E
	CD94/NKG2E/H	Not known
NKG2D		MIC-A, MIC-B, UL16 proteins
3: Leukocyte Ig-like inhibitory receptors (LIR)
LIR-1		HLA-G
Other LIRs		Not known
4: Natural cytotoxicity receptors
	NKp30	Not known
	NKp44	Not known
	NKp46	Not known

^aKIR2DL4 is an activating receptor (the only exception to the rule).

Killer immunoglobulin-like receptors (KIR)

Killer immunoglobulin-like receptors (KIR) are proteins encoded by a gene complex located on chromosome 19. Thirteen different KIR genes have been identified, encoding for the relevant receptors. The extracellular chain is composed by immunoglobulin-like domains (D). The functional status of each KIR is determined by the length of the intracytoplasmic chain. KIRs with a long intracytoplasmic chain (L, long) transmit inhibitory signals, whereas KIRs with a short intracytoplasmic chain (S, short) transmit activating signals. Inhibitory KIRs contain an immunoreceptor tyrosine-based inhibition motif (ITIM) located in the intracytoplasmic chain. KIR–cognate ligand interaction is followed by ITIM phosphorylation, signal transduction into the nucleus, and modulation of gene expression resulting in NK inhibition (17). By analogy, activating KIRs contain an immunoreceptor tyrosine-based activating motif (ITAM) (17). The number of extracellular domains (D) and the length of the intracytoplasmic chain (L or S) are used for KIR characterization, e.g. KIR2DL1 (Figure 2).

Figure 2. Graphic depiction of human KIR receptors. Inhibitory receptors contain ITIMs in their long intracytoplasmic (L) chains. Activating KIRs lack ITIMs and have short intracytoplasmic (S) chains. KIR2DL1 (two extracellular domains (two D), and long intracytoplasmic chain (L)) is an inhibitory receptor with specificity for HLA-C type 1 alleles. KIR3DL1 (three D, L chain) is an inhibitory receptor specific for HLA-Bw4 type alleles. Activating KIRs contain ITAMs in their short intracytoplasmic chain (S). KIR2DS1 is an activating receptor with specificity for HLA-C type 1 alleles.

Inhibitory KIRs recognize common antigenic epitopes present on HLA class I. Depending on the presence of specific epitopes, each HLA-C allele can be classified either as type 1 or type 2. HLA-C1 alleles (Ser77 and Asn80) are the ligands for KIR2DL2 and KIR2DL3 receptors, while HLA-C2 alleles (Asn77 and Lys80) are the ligands for KIR2DL1. Similarly each HLA-B allele can be classified either as Bw4 or Bw6 type. HLA-Bw4 alleles (Arg79, Ile80, Arg83 or Arg79, Thr80, Arg83) are the ligands for inhibitory KIR3DL1 receptor, whereas no receptor specific for HLA-Bw6 alleles has been identified. HLA-A alleles are characterized as HLA-A3 or HLA-11, and these types of alleles serve as the ligands for inhibitory KIR3DL2 receptor (18,19).

For each inhibitory KIR there is a homologous activating receptor that differs mainly in the length of the intracytoplasmic chain and the corresponding motif (ITAM versus ITIM). The vast majority of the ligands specific for activating KIRs are unknown. Except for KIR2DS1 (ligand: HLA-C2) and KIR2DS2 (ligand: HLA-C1), the ligands of other activating KIRs have not been identified yet (19). KIRs are highly expressed on the surface of cytotoxic CD56^dim/CD16^bright NK cells, while KIR expression on the CD56^bright/CD16^dim NK cell subset is absent or low.

It is interesting that at least a subset of NK cells express both inhibitory and activating receptors with specificity for the same ligand (HLA-C1, 2, HLA-E). Under normal circumstances, the activating signal is always weaker in comparison to the inhibitory signal received through the same ligand, thus protecting the host from autoimmunity. However in cases of certain infections, pathogen-derived peptides bind to HLA antigens and alter their molecular structure. It is possible that the altered ligand (HLA peptide complex) displays a different binding affinity for NK cell receptors favoring activation rather than inhibition, thus resulting in pathogen elimination.

Heterodimer CD94/C-type lectin receptors

These receptors are heterodimers and consist of two subunits: 1) CD94 protein chain which is common to all receptors, and 2) a C-type lectin NKG2 molecule. There are six different NKG2 proteins (NKG2A, NKG2C, NKG2E, NKG2F, NKG2B, NKG2H), and all of them, including CD94 protein, are encoded by genes located on chromosome 12. CD94/NKG2A and CD94/NKG2B are inhibitory receptors, while the rest of the receptors are activating. Non-classical HLA-E has been identified as the ligand for inhibitory receptors. HLA-E presents leader peptides from degraded HLA class I molecules, and the expression density of HLA-E functions as a sensor of the total HLA class I expression. The ligands for activating receptors have not been identified yet (20).

Activating receptor NKG2D is a member of C-type lectin family: NKG2D is expressed as a homodimer on the surface of NK cells. NKG2D is an activating receptor, and two different families of cognate ligands have been identified: 1) MHC class I chain-related peptides type A and B (MIC-A and MIC-B), and 2) UL16 binding protein family. NKG2D ligands are not expressed on normal cells. Expression is induced in cells under the influence of ‘stress’ (inflammation, neoplastic transformation) (20–23). MIC-A and MIC-B ligands are over-expressed on the surface of blast cells from patients with acute myelogenous leukemia (24). These data provide further proof of the concept that NK cells are among the key players of immune surveillance against cancer.

Leukocyte Ig-like inhibitory receptors (LIRs)

Leukocyte Ig-like inhibitory receptors (LIRs) are inhibitory receptors encoded by genes located on chromosome 19. The best known receptor is LIR-1, which interacts with HLA-G molecule (20). Recent studies showed that collagen molecules serve as ligands for LIR-1. Inhibitory signals transmitted through LIR-1 help in maintaining tolerance to self, especially in tissues rich in collagen (25).

Natural cytotoxicity receptors (NCRs)

Natural cytotoxicity receptors (NCRs) are activating receptors expressed on the surface of NK cells. NKp46, NKp30, and NKp44 are the three known receptors encoded by genes located on chromosomes 19, 6, and 6, respectively. NCRs transmit signals responsible for cytotoxic activity against virus-infected and transformed cells. NKp46 and NKp30 are expressed on the surface of all NK cells, while NKp44 is expressed on the surface of activated NK cells. Pathogen-derived structures are the known ligands for NCRs. Mandelboim et al. have shown that the NKp46 and NKp44 proteins but not the NKp30 recognize virally derived proteins such as hemagglutinin and hemagglutinin-neuraminidase of influenza and Sendai virus, respectively (26). In another study, Pogge von Strandmann et al. have shown that the nuclear factor HLA-B-associated transcript-3 (BAT-3) is a cellular ligand for NKp30 (27). Also plasmodium protein-1 and heparin sulfate structures have been postulated as ligands for NCRs (28).

However, most of the cellular ligands of the NCRs have not been identified yet. A large body of data suggests that unidentified cellular ligands for NCRs are present on the surface of malignant cells and their expression is critical for NK-mediated immune surveillance against tumors. Previous studies have shown that malignant cells from different tumors, such as acute myeloid leukemia (AML), melanoma, and prostate cancer, express ligands for NCRs (29–31). These findings are supported by data from AML studies showing an association between low levels of ligand expression and poor outcome (32).

Regulation of NK cytotoxic activity: the missing self hypothesis revisited

NK and CD8 T lymphocytes share the same cytotoxic machinery (perforin/granzymes, expression of TRAIL (TNF-related apoptosis-inducing ligand), and FasL (Fas ligand)).

Early studies showed that NK-inhibitory receptors recognize common epitopes on HLA class I antigens. Virus-infected or transformed cells with absent or low expression of HLA antigens are sensitive to NK cell-mediated cytotoxicity, while HLA expression on the surface of normal cells results in inhibition of lysis. The ‘missing self’ hypothesis was based on these observations. According to ‘missing self’ hypothesis, NK cells kill cells lacking HLA expression because the target cells cannot engage an inhibitory NK cell receptor for HLA class I (33). Prevention of autoreactivity is mediated through expression on each mature NK cell of at least one inhibitory receptor specific for self-HLA.

However, the ‘missing self’ hypothesis is currently considered as an oversimplification of NK activity regulation. A large body of data suggests that missing of self inhibition is not sufficient for NK activation. A vast array of inhibitory and activating receptors are expressed on the surface of each NK cell. The NK cell receives a large number of opposite stimuli, both inhibitory and activating, depending on the type of ligands expressed on the surface of the target cell. The decision ‘to kill or not’ is finally dependent on the net sum of the different positive and negative signals (Figure 3) (34,35).

Figure 3. ‘Missing self’ revisited. NK cells against potential targets and possible outcomes. A: NK against a normal autologous epithelial cell; inhibition from self-HLA, and absence of ligands for activating receptors; no lysis of target. B: NK against a malignant hematopoietic cell; absence of inhibition due to down-regulation of self-HLA, and presence of ligands for activating receptors; lysis of target. C: NK against a malignant hematopoietic cell; absence of inhibition due to down-regulation of self-HLA, and absence of ligands for activating receptors; no lysis of target. D: NK against an autologous virus-infected cell; inhibition by self-HLA, and high expression of activating ligands on the surface of the target; when NK encounter target cells expressing ligands for both inhibitory and activating receptors, the outcome is determined by the net sum of the strength of signals. In this case the result is lysis of the target.

Experimental data in agreement with the new concept are given below:

1. The absence of inhibitory signals is not sufficient for activation. In animal models of mismatched allogeneic bone-marrow transplantation, infusion of donor's NK cells to the recipient results in significant hematopoietic tissue ablation without the development of graft-versus-host disease (GVHD) (36). Donor's NK cells lacking inhibitory receptors specific for recipient MHC are alloreactive and exert significant cytotoxic activity against hematopoietic cells of recipient origin. Surprisingly, alloreactive NKs are not active against recipient's epithelial cells (no GVHD). The difference between normal epithelial and hematopoietic cells is the expression of ligands specific for activating NK receptors in the latter. In this scenario, absence of inhibitory in the presence of activating stimuli results in NK activation.

2. The presence of inhibitory stimuli does not necessarily result in prevention of cytotoxicity. Virus-infected cells under the influence of ‘stress’ express high levels of activating ligands. Self-NK cells, despite not being autoreactive (inhibition through self-HLA), are still capable of killing autologous virus-infected cells (37). It has been suggested that activating signals, of high magnitude, are able to overcome weaker inhibitory signals resulting in NK activation and killing of the target cells. In agreement with these observations, previous studies have shown that autologous NK cells, collected from the peripheral blood of patients with acute myeloid leukemia in complete remission, exert significant cytotoxic activity against leukemic blasts collected at the time of diagnosis (38).

The diversity of NK receptor repertoire

The KIR gene family is extremely polymorphic, with differences in the number and type of genes present. KIR genes are clustered in the same genetic locus and are inherited as a haplotype (39). Each individual inherits two haplotypes, one maternal and one paternal. Depending on the number and type of KIR genes, KIR haplotypes can be divided into two general subcategories. The haplotypes A are characterized by the total absence of genes encoding for activating receptors except KIR2DS4. The haplotypes B are more diverse and are characterized by the presence of more than one activating KIR. Moreover, each gene encoding for a specific receptor is polymorphic, with many different alleles present in the general population (40).

Human KIR gene expression is a stochastic process, occurs during NK development, and results in a highly diverse KIR repertoire. The most important factor that regulates the generation of the KIR repertoire is the KIR genotype itself. KIR gene expression is controlled at the transcriptional level (41). Non-expressed KIR alleles are methylated, while active KIR genes are demethylated. Similarly to KIR genes the expression of C-type lectin heterodimers occurs in a random manner. Activating receptors, such as NKG2D, NKp46, and NKp30, are expressed in the vast majority of circulating mature NK cells, while NKp44 is expressed on the surface of activated NK cells.

NK cell cytotoxicity and tolerance to self: license to kill

The genes that encode for HLA and KIR proteins are located in different chromosomes and are inherited independently. NK cell receptor expression is a stochastic event and occurs during NK development (42). According to the ‘at least one’ model, autoreactivity is prevented through acquisition by each mature NK of at least one inhibitory receptor specific for self-HLA. During NK development, NK precursors that fail to express inhibitory receptors specific for self-HLA undergo apoptosis, and the mature NK repertoire contains only non-autoreactive cells (43,44).

However, more recent data do not support the concept of the ‘at least one’ model. Anfossi et al. analyzed the NK repertoire of healthy individuals homozygous for HLA-C alleles type C1 (C1/C1). Surprisingly, they found that 10% of circulating NK cells did not express any of the inhibitory KIRs, nor the inhibitory receptor NKG2A (KIR-/NKG2A−). Moreover, they showed that 15% of NK cells expressed KIR2DL1 (specific for type C2) as the only inhibitory receptor (KIR2DL1+/NKG2A−).

The (KIR−/NKG2A−) NK and (KIR2DL1+/NKG2A−) NK clones do not receive inhibitory signals when they encounter autologous cells, and theoretically these NK clones are considered as potentially autoreactive (45). In the context of new data an old question re-emerged: how is tolerance to self maintained?

Anfossi et al. examined the functional status of ‘autoreactive’ NK clones present in the peripheral blood of healthy donors. Interestingly, (KIR−/NKG2A−) NK cells and (KIR2DL1+/NKG2A−) NK cells were minimally reactive (hyporesponsive) when tested against K562 (HLA-deficient cell line) targets. In contrast, self-tolerant KIR2DL2+NK cells (non-autoreactive, inhibition mediated through KIR2DL2–HLA-C type 1 interaction) were highly reactive against K562 targets.

The ‘licensing’ hypothesis was based on the above findings in an effort to explain the mechanisms of tolerance to self: immature NK cells acquire NKR receptors in a stochastic manner, resulting in the production of a large number of NK clones expressing different combinations of receptors. NK cells that did not receive inhibitory signals during maturation remain in a state of hyporesponsiveness (without license to kill). ‘Unlicensed’ NK cells are not cytotoxic against normal autologous cells. On the contrary NK cells that received inhibitory signals during maturation remain in a state of ‘ready-to-respond’ and exert significant cytotoxic activity under the appropriate circumstances (with license to kill). ‘Licensed’ NK cells receive inhibition from self-HLA and are not cytotoxic against normal autologous cells (46,47).

Prevention of NK-mediated autoimmunity in patients with ‘peptide transporter-associated antigen processing’ (TAP) deficiency is a further proof of the concept of ‘licensing’ (48). TAP is a protein system required for presentation of antigenic peptides in the context of HLA class I molecules. Cells from patients with defects in TAP system express negligible amounts of HLA class I molecules on the cell surface (49). Deficient HLA class I expression renders cells susceptible to NK-mediated cytotoxicity (50). However, most of the patients with TAP deficiency do not experience NK-mediated autoimmune disorders, because NK cells remain in a state of hyporesponsiveness. In agreement with clinical observations, previous experiments have shown that NK cells from TAP-deficient patients are not cytotoxic against autologous HLA class I-negative B lymphoblastoid cell lines (51).

However, it is important to note that ‘unlicensed’ NK cells are not just a biological waste, because the state of hyporesponsiveness is potentially reversible. Indeed, these cells regain cytotoxic activity in cases of infection or after in-vitro culture in the presence of IL-2 (49).

The molecular mechanisms regulating the process of maturation and ‘licensing’ are not known yet. Similarly, the origin of the cells that present ligands (self-HLA) to immature precursors during NK development has not been identified. In preclinical studies using a mouse model, Sykes et al. showed that the cells presenting self-HLA to NK precursors were of hematopoietic origin (52,53).

NK cells have memory: a bridge between innate and adaptive immunity

As already mentioned, NK cells have been classified as cells of innate immunity due to certain characteristics, such as absence of previous sensitization, lack of memory, etc. Immune cells of adaptive immunity (T cells) show a characteristic four-phase kinetic pattern after encountering a pathogen. During the first phase, known as the expansion phase, naive T cells with specificity against pathogen-derived peptides proliferate and give rise to thousands of effectors resulting in the elimination of the invader. The second phase, known as the contraction phase, is characterized by massive apoptosis of activated effectors. During the third ‘maintenance’ phase a number of previously activated immune cells escape from the apoptotic process and revert to a state of ‘quiescence’. These cells are the memory cells that protect the host from re-challenge with the same pathogen. Finally, during the secondary or recall phase, if the same pathogen invades the host for a second time, then memory cells expand rapidly and give rise to a large number of activated effectors resulting in the elimination of the pathogen.

However, the distinction between ‘innate’ and ‘adaptive’ is currently questionable because recent data showed that NK cells display some features of adaptive immunity (54). In an elegant experiment, Sun et al. examined the kinetics of specific NK cell response in mice infected with murine CMV (MCMV). Interestingly they identified a long-lived memory pool of NK cells that persist in host tissues. When these ‘memory’ NK cells are adoptively transferred in neonatal secondary recipients infected with MCMV, they expand rapidly and protect host mice much more efficiently than ‘naive’ NK cells. Thus, like memory T cells, ‘memory’ NK cells are long-lived, have the ability of self-renewal, and can mount a secondary response that protects the host in case of pathogen re-challenge (55). In light of these new data, Sun et al. proposed that NK cells should be classified as an ‘evolutionary bridge’ between innate and adaptive immunity (56).

It is obvious that the identification of such ‘memory’ NK cells in humans will be of great importance especially in the field of allo-SCT and may lead to the development of innovative immunotherapeutic approaches aiming at elimination of residual disease and prevention of infections post-transplant.

NK cells in the setting of haploidentical transplantation

The major limitation of allo-SCT is the lack of a matched donor, either related or unrelated. In these cases, unrelated cord blood and allo-SCT from a haploidentical family member (haplo-SCT) represent alternative options. In the case of haplo-SCT, donor and recipient share one common haplotype (57). To overcome the barrier of a mismatched haplotype it is necessary to: 1) treat the patient with a highly immunosuppressive conditioning in order to ablate the very last recipient's T lymphocyte reactive against donor and therefore minimize the risk of rejection; 2) infuse a graft containing a ‘megadose’ of CD34⁺ cells (previous animal studies performed by Reisner et al. showed that CD34⁺ cells act as veto cells and help in the elimination of recipient's immune cells that survived conditioning (58)); and 3) infuse a graft vigorously depleted of T lymphocytes in order to prevent GVHD that is otherwise inevitable in the setting of a haplotype mismatch. Positive selection of CD34⁺ cells, using immune-magnetic beads, is the most effective method for extensive T cell depletion.

Among the pioneers in the development of haplo-SCT are the members of the Perugia team. Clinical application of haplo-SCT not only revolutionized the field of allo-SCT but also shed light on the biology of NK cells and renewed the scientific interest in an immune cell subset that was neglected for years. A large number of haplo-SCT have been performed in Perugia mostly in patients with relapsed, refractory, or high-risk acute leukemias in remission. Intensive clinical and laboratory research revealed that the presence of NK alloreactivity in GVH (graft-versus-host) direction is associated with a favorable outcome after haplo-SCT in patients treated for acute myeloid leukemia (59). According to the Perugia team, NK alloreactivity is defined as KIR ligand-to-ligand mismatch. In more detail, NK alloreactivity to GVH direction is observed in a donor–recipient pair if a ligand (HLA) for a specific inhibitory KIR is present in the donor but absent in the recipient. Molecular typing for HLA class I allele identification in both donor and recipient and KIR-inhibitory phenotype analysis of donor's NK cells are required for prediction of NK alloreactivity. An example of NK alloreactivity in GVH direction is given below: Donor is heterozygous for HLA-C alleles (C1/C2), and recipient is homozygous for HLA-C alleles (C1/C1). Ligand C2 is present in donor but absent in recipient.

Ruggeri et al. performed a retrospective analysis of more than 100 patients with acute myeloid leukemia who underwent haplo-SCT in Perugia during the years 1996–2006. Multivariate analysis revealed disease status and NK alloreactivity as the two most significant factors, independently associated with treatment outcome (60). NK alloreactivity in GVH direction was associated with less rejection, less GVHD, and higher leukemia-free (LFS) and overall survival (OS). Ruggeri et al., using elegant experiments, showed that alloreactive NK clones of donor origin were present in the peripheral blood of patients (59). Experimental data showed that alloreactive NK cells helped in: 1) elimination of recipient's immune cells that survived conditioning and are responsible for graft rejection (36,61); 2) elimination of recipient's dendritic cells that present host antigens to donor T cells, resulting in the initiation of GVHD process (62); and 3) elimination of residual leukemic cells resulting in improved LFS.

The beneficial impact of NK alloreactivity has been reported for myeloid malignancies but has been questionable for B lineage acute lymphoblastic leukemia (ALL). Interpretation of clinical data showed that the beneficial effect of NK alloreactivity was absent in adults with ALL (63). Previous studies supported the concept that leukemic lymphoblasts escape from NK-mediated immune surveillance, probably due to the lack of expression of ligands for activating receptors (64). However, a recent clinical study in children with ALL showed the existence of a NK-mediated graft-versus-leukemia effect after haplo-SCT (65). In accordance with clinical data, Pfeiffer et al. in an experimental study showed that the intensity of HLA expression on the surface of leukemic lymphoblasts is the key parameter that determines the degree of NK-mediated cytotoxicity. In this study, KIR ligand mismatch between donor and recipient gained statistical significance only when the intensity of HLA expression was taken into consideration. In more detail, the effect of KIR mismatch was obscured in cases with low HLA expression, while it was more pronounced in cases with high HLA expression (66).

Why are NK alloreactive clones of donor origin present in the peripheral blood of patients?

Circulating alloreactive NK clones of donor origin were present in the vast majority of patients from donor–recipient pairs with NK alloreactivity. On the contrary, no NK alloreactive clones were found in patients from donor–recipient pairs without NK alloreactivity. Of importance is the fact that alloreactive NK clones were present in peripheral blood only during the first months post-transplant and were lost thereafter.

Alloreactive NK clones were not passively transferred through graft infusion, because positive selection of CD34 cells results in an extensive depletion not only of T cells but also of other immune cell subsets, such as NK cells. Of importance are previous data supporting the concept that lymphopenic environment favors NK generation and expansion. Moreover, it is well established that interleukin-15 (IL-15) is required for differentiation, expansion, and survival of NK cells (67). Miller et al. observed a marked rise of endogenous IL-15 in patients treated with highly immunosuppressive regimens. The degree of induced lymphopenia was inversely associated with serum levels of IL-15 and the persistent expansion of allogeneic NK cells infused (68). The specific conditions of the Perugia protocol, such as the extensive T cell depletion, the infusion of a ‘megadose’ of CD34⁺ cells, and the omission of cyclosporine from GVHD prophylaxis, result in a ‘wave’ of NK cell generation from the large number of CD34⁺ NK precursors present in the graft (69).

Understanding the mechanisms of alloreactive NK cell generation is of paramount importance, because it might open new avenues in NK cell immunotherapy. Similar to T cell development, post-transplant generation of NK from NK precursors mimics the normal NK cell ontogeny. The generation of alloreactive NK clones can be explained by applying the ‘licensing hypothesis’ in the same way as in the case of normal NK cell development. An example of a donor–recipient pair with NK alloreactivity in GVH direction is given below: Donor is heterozygous for HLA-C alleles (C1/C2), and recipient is homozygous for HLA-C alleles (C1/C1). In the early post-transplant period, CD34 progenitors of donor origin seed in bone-marrow and give rise to generation not only of myeloid cells but also of NK progenitors. Pro-NK cells follow the developmental pathway in a microenvironment characterized by the presence of hematopoietic cells of donor origin (abundant presence of C2 ligands). As already mentioned, KIR2DL1 is the only inhibitory receptor expressed in a subset of immature NK cells (KIR2DL1+/others−). The (KIR2DL1+/others−) immature NK receive inhibitory signals through interaction with C2 ligands (present on donor hematopoietic cells) during development and become fully responsive (licensed). Residual normal and leukemic cells of recipient origin do not express ligand C2 and therefore are susceptible to lysis by (KIR2DL1+/others−) ‘licensed’ NK present in the peripheral blood of recipient post-transplant (Figure 4).

Figure 4. The ‘licensing’ hypothesis in the setting of haploidentical SCT: Donor/recipient pair with NK alloreactivity to GVH direction according to ‘KIR ligand–ligand mismatch’ model (detailed description of the events is given in the text).

Why do NK alloreactive NK clones not permanently persist in the peripheral blood of recipients?

Ruggeri et al. in the same experiments observed that alloreactive NK clones did not persist in the peripheral blood of recipients beyond the first year post-transplantation (59). Pende et al. studied the presence of donor-derived alloreactive NK cells in 21 children with high-risk acute leukemias after haploidentical SCT. In contrast to previous results, they observed that alloreactive NK cells persisted in the majority of patients for a long time (28 months) post-transplant (70). The generation and persistence of alloreactive NK cells in patients after allo-SCT need to be further investigated.

Models for prediction of NK alloreactivity after stem cell transplantation

In cases of haplo-SCT, a donor with NK alloreactivity in GVH direction is considered as the most suitable. According to the Perugia team, NK alloreactivity in GVH direction between donor and recipient is predicted using the KIR ligand–ligand mismatch model. On the contrary, other studies showed that NK alloreactivity is better predicted using the receptor–ligand mismatch model proposed by Leung et al. (65) (Table II). According to this model NK alloreactivity is observed if the recipient lacks a ligand (HLA), while the cognate receptor (KIR) is present on donor's NK repertoire: for example a recipient homozygous for C1/C1, and KIR2DL1 expressed on the donor's NK cells. A very interesting issue regarding the ‘receptor–ligand mismatch’ model is that HLA mismatch is not a prerequisite for the generation of NK alloreactive clones. Moreover, NK alloreactivity can be observed even after SCT from matched family members or even after autologous stem cell transplantation. It is obvious that the ‘licensing’ hypothesis cannot be applied for the explanation of NK alloreactivity in the setting of a HLA match.

Table II. Clinical trials evaluating the effect of NK (allo)-reactivity in different transplant settings.

Author, year (Ref)	No of patients	Diseases	Donor	Unmanipulated graft	ATG	Effect of NK (allo)-reactivity
NK alloreactivity to GVH direction predicted according to KIR ligand-to-ligand mismatch model
Ruggeri, 2007 (60)	112	High-risk AML	Haplo	T cell depletion	Yes	Reduced relapse, improved LFS
Pende, 2009 (70)	21	AML, ALL in children	Haplo	T cell depletion	Yes	Reduced relapse, improved LFS
Ciceri, 2008 (63)	266	AML, ALL	Haplo	T cell depletion	Yes	No beneficial effect
Kligenbiel, 2010 (94)	127	ALL in children	Haplo	T cell depletion	Yes (60%)	No beneficial effect
Giebel, 2003 (72)	130	Various hematological malignancies	MUD, MisMUD	Yes	Yes	Less relapse, better RFS and OS
Davies, 2002 (73)	175	Various hematological malignancies	MisMUD	T cell depletion (28%)	Not reported	No beneficial effect
Bornhauser, 2004 (95)	118	Myeloid malignancies	MUD, MisMUD	Yes	Yes	No beneficial effect
De Santis, 2005 (96)	104	Various hematological malignancies	MUD, MisMUD	Yes	Not reported	No beneficial effect
Schaffer, 2004 (97)	190	Various hematological malignancies	MUD, MisMUD	Yes	Yes	Increased TRM
Beelen, 2005 (98)	374	Myeloid malignancies	MRD, MUD, MisMUD	Yes	No	Less relapse, better DFS
Elmaagacli, 2005 (99)	236	CML	MRD, MUD, MisMUD	Yes	Not reported	Improved molecular CR
Farag, 2006 (74)	1571	Myeloid malignancies	MUD, MisMUD	T cell depletion (19%)	Not reported	No beneficial effect
Morishima, 2007 (75)	1790	Myeloid and lymphoid leukemias	MUD, MisMUD	Yes	No	Adverse effect
Yabe, 2008 (76)	1489	Various hematological malignancies	MUD, MisMUD	Yes	Only 9% of patients	Adverse effect
NK alloreactivity to GVH direction predicted according to receptor–ligand mismatch model
Leung, 2004 (65)	36	Myeloid and lymphoid leukemias	Haplo	T cell depletion	Yes	Reduced relapse, improved LFS
Cook, 2004 (81)	220	Various hematological malignancies	MRD	Yes	No	Reduced OS in patients homozygous for C2 and donor 2DS2+
Hsu, 2005 (79)	178	Various hematological malignancies	MRD	T cell depletion	No	Improved RFS in patients with AML and MDS
Sobecks, 2007 (101)	60	AML	MRD	Yes	No	Reduced relapse, improved LFS
Clausen, 2010 (80)	100	Various hematological malignancies	MRD	Yes	No	No beneficial effect
Hsu, 2006 (77)	1770	Myeloid and lymphoid leukemias	MUD, MisMUD	Yes	Not reported	Decreased relapse rate only in MisMUD
Miller, 2007 (78)	2062	Myeloid malignancies	MUD, MisMUD	Percentage of TCD not reported	Not reported	Decreased relapse rate
Leung, 2007 (84)	16	Lymphoma, solid tumors	Auto-SCT	Yes	No	Improved PFS, OS
Stern, 2008 (85)	67	Lymphoma, solid tumors	Auto-SCT	T cell depletion (30%)	No	No effect
Venstrom, 2009 (86)	169	Neuroblastoma	Auto-SCT	Yes	No	Improved PFS

ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; ATG = anti thymocytic globulin; Auto-SCT = autologous SCT; CR = complete remission; DFS = disease free survival; Haplo = haploidentical; LFS = leukemia-free survival; MDS = myelodysplastic syndrome; MisMUD = mismatched unrelated donor; MRD = matched related donor; MUD = matched unrelated donor; RFS = relapse-free survival; TCD = T cell depletion; TRM = transplant-related mortality.

The receptor–ligand mismatch model is based on the ‘abnormal maturation’ hypothesis: NK cell development in the early post-transplant period occurs in a disturbed microenvironment that favors the break of tolerance to self. Indeed in a previous study, authors examined the functional status of NK cells of donor origin generated in recipients after T cell-depleted HLA-identical SCT. Interestingly, they showed that during the first 3–5 months post-transplant ‘unlicensed’ NK cells had normal cytotoxic activity, while ‘licensed’ NK were hyperresponsive. NK function was restored to normal after the fifth month, indicating that breaking of self-tolerance was a transient phenomenon (71). In agreement with these data Ruggeri et al. observed that NK alloreactive clones were not present in any of the patients tested after the first year post-transplant.

NK alloreactivity after allogeneic SCT from matched and mismatched unrelated donors

The impact of NK alloreactivity has been examined in the setting of allo-SCT from partially mismatched unrelated donors (Table II). The KIR ligand–ligand mismatch model can be applied in donor–recipient pairs mismatched at HLA-C and/or B loci. Results from previous studies are contradictory (72,73). In the study performed by Giebel et al., KIR incompatibility in GVH direction was associated with decreased relapse rate, while three other studies did not reveal any association between KIR incompatibility and transplant outcome (72).

Finally this issue was addressed in three large retrospective studies (74–76). In the study conducted by Farag et al. (74), the impact of KIR ligand mismatch on transplant outcome was examined in 1,571 patients with myeloid malignancies who underwent unrelated donor transplantation. Patients were divided in four groups according to KIR ligand–ligand mismatch model: Group 1: HLA and KIR ligand matched (n = 1,004 patients); Group 2: HLA mismatched and KIR ligand mismatched in GVH direction (n = 137 patients); Group 3: HLA mismatched and KIR ligand mismatched in host-versus-graft (HVG) direction (n = 170 patients); and Group 4: HLA mismatched but KIR ligand matched (n = 260 patients). The outcomes between the four groups were compared using Cox regression analysis. Treatment-related mortality (TRM) was significantly decreased in recipients transplanted from HLA-matched donors (Group 1) in comparison with the three other groups. Relapse rates were not different between the four groups. Similar results were obtained from two other large retrospective studies performed by Morishima et al. and Yabe et al., respectively (75,76).

However, it is important to note that the patients included in these studies received a T cell-repleted graft. It has been suggested that the anti-leukemic effect of NK alloreactivity might have been obscured by the much stronger GVH effect mediated through donor T cells contained in the graft. It is believed that if a beneficial effect of NK alloreactivity really exists, then it will be unmasked if strategies used in haplo-SCT, such as high stem cell dose, extensive T cell depletion, and no post-grafting immune suppression, were employed in the setting of allo-SCT from partially mismatched unrelated donors.

The receptor–ligand mismatch model has been tested in the setting of allo-SCT from matched unrelated donors (MUD). In a large retrospective study, performed by Hsu et al., authors examined the effect of NK alloreactivity (predicted according to the receptor–ligand mismatch model) on transplant outcome in 1,770 patients who underwent allo-SCT from MUD. NK alloreactivity was associated with decreased relapse rates only in patients transplanted from mismatched unrelated donors (77). A similar study in a large cohort of 2,062 patients with myeloid malignancies, performed by Miller et al., showed that NK alloreactivity (receptor–ligand mismatch model) was associated with decreased relapse rates in the subset of patients without advanced myeloid leukemias (78).

NK alloreactivity after allogeneic SCT from matched related donors

According to the ‘receptor–ligand mismatch’ model, NK alloreactivity can be observed even in cases of allo-SCT from HLA-matched siblings. Indeed, in a previous study, Hsu et al. examined the impact of KIR and HLA genotype on transplant outcome in a cohort of 178 patients with hematological malignancies who underwent T cell-depleted HLA-identical sibling allo-SCT (79). Approximately 60% of recipients lacked an HLA ligand for an inhibitory KIR present on the donor's NK cells. Missing KIR ligand was associated with decreased incidence of relapse and improved LFS and OS in patients with AML and myelodysplastic syndrome (MDS), while no difference was observed in patients with chronic myeloid leukemia (CML) and ALL. Yu et al., from the same team, showed that this anti-leukemic effect was due to the presence of autoreactive NK cells in the peripheral blood of patients during the first 4 months after transplantation (71). The generation of NK alloreactive clones in the setting of HLA match is a further proof of the concept of the ‘receptor–ligand mismatch’ model.

However, contradictory results were observed in studies including patients who underwent T cell-repleted allo-SCT from HLA-matched family members (80–82) (Table II). In these studies no beneficial effect of missing KIR ligand on transplant outcome was observed. In accordance with clinical results, Bjorklund et al. showed that ‘unlicensed’ NK cells remained tolerant to self at all time points after transplantation (83).

Based on the above data, it has been suggested that the bone-marrow microenvironment after T cell-depleted allo-SCT favors breaking of tolerance to self during NK development, while the opposite is true in the case of T cell-repleted allo-SCT.

NK autoreactivity after autologous SCT

If NK alloreactivity can be observed in the setting of allo-SCT from HLA-matched donors, then it is reasonable to assume that NK autoreactivity might be present even after autologous SCT (Table II).

According to the ‘receptor–ligand mismatch’ model, NK maturation occurs in a disturbed microenvironment resulting in breaking of tolerance. If this phenomenon really exists, then the same pathogenetic mechanisms might take place after auto-SCT. Breaking of tolerance to self will give rise to autoreactive NK clones during the early post-transplant period. Autoreactive NK cells might help in eradication of minimal residual disease, resulting in decreased relapse rates after auto-SCT. In a previous study including 16 patients with lymphoma or solid tumors who underwent auto-SCT, authors examined the impact of missing KIR-ligand on relapse rate (84). Survival analysis showed that missing KIR ligand was associated with increased progression-free survival (PFS) and OS. On the contrary, Stern et al. did not observe any beneficial effect of missing KIR ligand on the outcome of 67 patients with solid tumors and lymphoma who underwent auto-SCT (85). Finally in a more recent study, Venstrom et al. examined the effect of missing KIR ligand in a subset of 169 children with neuroblastoma after auto-SCT. Interestingly they showed that missing KIR ligand was independently associated with improved PFS (86).

The role of activating KIRs

The role of activating KIRs on transplant outcome is currently an area of active research. NK cells express numerous activating receptors that potentially interact with cognate ligands expressed not only on virus-infected and transformed cells but also on the surface of normal autologous cells. An important question that needs to be answered is how NK autoreactivity is prevented. Recent data from animal studies strongly support the concept that expression of activating receptors during NK cell maturation contributes to the process of ‘licensing’. Normal cells from adult mice do not express significant amounts of the Rae family of NKG2D ligands. NK cells from immunocompetent mice exert significant cytotoxic activity against malignant or virus-infected cells expressing Rae molecules. These findings indicate that NK cells are able to kill normal autologous cells expressing NKG2D ligands. On the contrary, NK cells from Rae-transgenic mice are not cytotoxic against normal or malignant Rae-expressing cells. These data indicate that when NK cell maturation takes place in the presence of certain activating ligands, such as Rae, then NK cells become ‘hyporesponsive’ (they do not get the license) to activating signals transmitted through the NKG2D receptor (87).

Studies evaluating the role of activating receptors in the setting of hematopoietic SCT are briefly described: Giebel et al., in a prospective study, examined the effect of donor and recipient-activating KIR genes on transplant outcome in a cohort of 100 patients with hematological malignancies who underwent allo-SCT (88). The authors showed that the presence of KIR2DS1 and KIR2DS3-activating KIRs in the donor was associated with increased incidence of acute and chronic GVHD, respectively. In a recent study Cooley et al. analyzed the effect of donor KIR haplotype on transplant outcome. A total of 444 patients with AML who underwent allo-SCT from unrelated donors were included in the study. In multivariate analysis, the presence of KIR haplotype type B (rich in activating receptors) in the donor produced less relapse and improved OS in patients (89).

NKG2D, an activating receptor expressed on all NK cells and on a subset of T cells, plays a significant role in immunity against infectious agents, as well as in cancer immune surveillance. Recently, two different haplotypes (LNK1 and HNK1) of the NKG2D gene were identified (90). A previous study showed that among Japanese individuals, the HNK1 haplotype is associated with increased cytotoxic activity of peripheral blood NK cells and a reduced incidence of epithelial neoplasms (91,92). Espinoza et al. examined the impact of donor and recipient NKG2D polymorphisms on outcome in a group of 145 patients with hematological malignancies who underwent allo-SCT from a matched unrelated donor (93). OS was significantly improved in recipients with standard-risk disease who were transplanted from donors carrying the NKG2D-HNK1 haplotype.

Conclusions

NK cells are important elements of innate immunity and have a significant contribution to immune surveillance against infectious agents and transformed cells. Moreover, NK cells are key players of allo-SCT, and a large body of evidence supports the beneficial effect of NK cells in prevention of relapse. Understanding the mechanisms regulating NK maturation, tolerance to self, and NK-mediated cytotoxicity will help in the development of innovative NK cell immunotherapy methods.

Declaration of interest: The authors state no conflict of interest and have received no payment in preparation of this manuscript.