Non-MHC-restricted cytotoxic cells: their roles in the control and treatment of leukaemias

Authors


Dr Mark W. Lowdell, Department of Haematology, Royal Free and University College Medical School (RF-Campus), London NW3 2PF, UK. E-mail: m.lowdell@rfc.ucl.ac.uk

The role of T lymphocytes in the MHC-restricted control and eradication of leukaemia after allogeneic bone marrow transplantation is established, but there is increasing evidence that the non-MHC-restricted cytotoxic lymphocytes have the capacity to lyse allogeneic or, in some cases, autologous leukaemic blasts. Recently, the field of tumour immunotherapy has been dominated by T cell-mediated approaches, probably as a result of the rapid and significant advances in our understanding of the mechanisms involved in the generation of responses from naïve T cells. During the last 5 years, our knowledge of the activation and control mechanisms involved in non-MHC-restricted cell-mediated immunity has advanced enormously and we are now seeing a concomitant increase in the understanding of these cell types in tumour immunotherapy and in the immunotherapy of haematopoietic malignancies in particular.

The term ‘non-MHC-restricted cytotoxic lymphocytes’ encompasses natural killer (NK) cells, NK-T cells and γδ T cells, all of which derive from a common lymphoid precursor cell but differentiate along separate pathways. NK-T and γδ T cells express surface CD3 and the T-cell receptor (TCR); they are thus true T cells. NK-T cells express the α and β chains of the TCR while, as their name suggests, γδ T cells express the γ and δ chain rearrangement. In contrast, NK cells lack both CD3 and TCR expression, but mostly (> 95%) express CD56 and/or CD16. Each cell type will be described in more detail including, when known, the mechanisms of antigen recognition, but in all cases this does not require MHC compatibility between effector and target cell, hence the term ‘non-MHC-restricted’, although NK cells do recognize MHC molecules and this interaction is involved in their control of cytolysis. Expression of appropriate MHC molecules on target cells is inhibitory to NK cell-mediated lysis. Thus, NK cells are permanently ready to lyse potential target cells and can be triggered in a variety of ways. Cells showing downregulation of MHC molecules, such as some tumours and some virally infected cells, are innate targets owing to the lack of inhibition of NK-mediated lysis. Other cells with normal MHC expression can still be NK targets if they provide appropriate stimulatory signals, for example by ligation of the Fcγ receptor (CD16), that are capable of overcoming the inhibition signals.

While each of the cell types discussed here are distinct entities, they share several characteristics with respect to differentiation or cell surface antigen expression (Table I), but their relationships in the ontogeny of the cellular immune system have yet to be unravelled. All these cell types are involved in primary immune responses to tumours and intracellular pathogens, but lack the ability to generate a ‘memory’ subset to mediate secondary cellular responses to recall antigens.

Table I.   Summary of basic characteristics of the three mediators of non-MHC-restricted cell mediated cytotoxicity.
 γδ T cellsNK-T/CIKNK
αβ TCR+
γδ TCR+
CD3+++
CD56+++
CD16+Majority +
Known reactivity
 CLLNot reportedNot reportedYes
 CMLYesYesYes
 ALLYesNot reportedLittle/None
 AMLYesYesYes

The lack of requirement for MHC matching by these cells, indeed the potential benefit of mismatching in the case of natural killer cells, means that many clinical applications of these non-MHC-restricted cytotoxic cells are based upon allogeneic donor systems. However, the role of autologous natural killer cells in the cure of leukaemia is beginning to be unravelled and may lead to better defined trials of passive immunotherapy in the near future.

γδ t cells

Between 2% and 10% of T lymphocytes in normal peripheral blood bear the γδ receptor (Raulet, 1989). Although these cells are similar to αβ T cells in many ways, there are important differences. One is that most γδ T cells do not co-express CD4 or CD8. They have also been shown to develop normally in the absence of MHC class II antigens in a knock-out mouse model (Bigby et al, 1993). It is difficult to elicit a response by γδ T cells against allogeneic MHC class I or class II antigens (Lanier, 1995) and they do not require presentation of antigens in the context of MHC class I or class II for activation (Schild et al, 1994). They do require CD28-mediated co-stimulation and, following activation, show some autocrine interleukin 2 (IL-2) production (Sperling et al, 1993). They can also be activated and expanded by anti-CD2 (Wesselborg et al, 1991).

The dynamics of graft–host interactions mediated by γδ T cells are now being elucidated. Ellison et al (1995) reported an increase in peripheral γδ T cells in murine studies of acute graft-vs.-host disease (GvHD) following allogeneic non-T cell-depleted (TCD) bone marrow transplantation (BMT). In this study, depletion of γδ T cells resulted in a significant decrease in GvHD-related mortality. However, studies in humans have, to this point, not shown γδ T cells to be primary effectors of GvHD. Two recent laboratory studies have shown γδ T cells to be poor responders in the allogeneic mixed lymphocyte culture (Schilbach et al, 2000; Lamb et al, 2001), even with the addition of exogenous IL-2. Biopsies of cutaneous lesions of GvHD show the epidermal layer to be devoid of γδ T cells, although they were found in the perivascular regions of the dermis (Norton et al, 1991). These data suggest that γδ T cells are probably not primary effectors of epidermal damage in cutaneous GvHD. Tsuji et al (1996) showed that, although γδ T cells were not able to induce GvHD on their own, host γδ T cells could be recruited into experimentally induced donor αβ T-cell lesions in which they were activated and induced to proliferate. Neipp et al (1999) did not show increased severity of GvHD in αβ-TCD grafts, while Cela et al (1996) and Lamb et al (1996, 1999) showed no association between increased circulating γδ T cells and GvHD in patients receiving TCD BMT grafts from a phenotypically matched unrelated donor (MUD) or a partially mismatched related donor (PMRD).

Conversely, γδ T cells may be important mediators of alloengraftment. Blazar et al (1996) reported that murine host γδ+ T cells can reject repopulating donor cells and donor γδ T cells can facilitate the alloengraftment of TCD bone marrow. The mechanism of both rejection and facilitation of engraftment is through the recognition of a non-classic MHC class I H-2T gene product. Neipp et al (1999) were also able to show improved engraftment in murine αβ-TCD bone marrow. Kawanishi et al (1997) showed a significant association between improved alloengraftment in patients who received TCD BMT grafts with an increased γδ T-cell dose.

Trafficking of γδ T cells has been studied to discern their function in immune recovery and anti-leukaemia activity. Viale et al (1992) observed an increase in the ratio of Vδ1:Vδ2 cells in patients with acute GvHD, but the significance of this finding remains undetermined. Transitory increases in γδ T cells have been reported during the first 4 weeks post BMT in patients treated with granulocyte macrophage colony-stimulating factor (GM-CSF), but the cells return to normal levels within 8 weeks post BMT (Yabe et al, 1994). Increased γδ T cells have also been found to be associated with viral and fungal infections during the first year after TCD BMT in patients receiving either PMRD or MUD grafts in a single centre study (Cela et al, 1996). Lamb et al (1998) reported a subset of patients who received αβ-TCD grafts and spontaneously developed increased numbers of γδ T cells during the first year after PMRD BMT. Further study showed these patients to have a reduced incidence of relapse (Fig 1) than patients who did not increase their γδ T-cell compartment (Lamb et al, 1996). Patients who receive αβ-TCD grafts were also significantly more likely to develop spontaneous increases in γδ T cells than those who received a pan-TCD graft (Lamb et al, 1999), although incidence of relapse was similarly decreased in both groups.

Figure 1.

 Reduced incidence of relapse in patients who received αβ-T cell-depleted (TCD) grafts and spontaneously developed increased numbers of γδ T cells. (A) Disease-free survival in the T10B9 TCD group. (B) Relapse in the T10B9 TCD group. All patients survived for at least 100 d.

The anti-neoplastic role of γδ T cells has been known since Esslin and Formby (1991) showed that in vitro-activated peripheral blood γδ T cells have cytolytic activity against some human tumour cell lines compared with similarly activated αβ T cells. This reactivity is not MHC restricted, but is dependent on an lymphocyte function-associated antigen (LFA)-1b/intercellular adhesion molecule (ICAM)-1 interaction. Known targets of γδ T cells include some leukaemic cell lines (Duval et al, 1995), Epstein-Barr virus (EBV)-transformed B cells (Hacker et al, 1992), Burkitt lymphoma cells (Hacker et al, 1992), Daudi lymphoma cells (Marx et al, 1997), autologous haematopoietic cells in aplastic anaemia (Melenhorst et al, 1999) and neuroblastoma (Schilbach et al, 2000). Known antigens recognized by γδ T cells include heat shock proteins (Fisch et al, 1990; Kaur et al, 1993; Battistini et al, 1995), which preferentially activate Vγ9/Vδ2-expressing cells, the most common γδ subset in humans and known effectors against several tumours.

Interestingly, however, the clone generated by disease-free survivors with increased γδ T cells reported above was exclusively Vδ1+ (Lamb et al, 1999) with substantial cytolytic activity against K562 targets (Fig 2). Lamb et al (2001) showed that a Vδ1+ clone from BMT donors could be raised against acute lymphoblastic leukaemia (ALL) from the respective BMT recipient. The expanded Vδ1+ cells could bind and lyse the primary ALL and some lymphoid cell lines (Fig 3A and C), but failed to lyse third party mononuclear cells (MNCs) and selected myeloid cell lines, indicating lymphoid specificity. Specific anti-lymphoid cytotoxicity is further supported by data showing that leukaemia-specific γδ cells could not be expanded against AML (Fig 3B and D). Additionally, eight out of eight patients reported who received PMRD grafts for refractory ALL and subsequently developed increased numbers of γδ T cells remained disease-free for between 1 and 7 years.

Figure 2.

 Cytotoxicity of patient γδ cells vs. normal control γδ cells.

Figure 3.

 (A and C) Expanded Vδ1+ cells bound and lysed primary acute lymphoblastic leukaemia and some lymphoid cell lines, but failed to lyse third party mononuclear cells and selected myeloId cells lines. (B and D) Leukaemia-specific γδ cells were not expanded against acute myeloid leukaemia cells, thus further indicating lymphoid specificity.

Further support for γδ T cells as effectors of the immune response against ALL comes from Duval et al (1995), who reported cytotoxic activity of Vδ1+ cells obtained from a patient with B-ALL against B-cell lines. However, the ALL-associated antigen recognized by γδ T cells has not yet been determined. Vδ1+ cells have been generated against TCT.1 (Blast-1/CD48), an antigen broadly distributed on haematopoietic cells (Chouaib et al, 1991) as well as the non-classic human MHC antigens MICA and MICB (Steinle et al, 1998). However, preliminary experiments have shown that these antigens are not present on the ALL cells lysed by Vδ1+ cells. Further work will be necessary to determine the mechanism of recognition and killing of ALL by γδ T cells, although they are a promising novel cellular therapy agent for refractory or relapsed ALL.

Nk-t cells and cytokine-induced killer (cik) cells

CD3+CD56+ cells, also referred to as NK-T cells, are abundant in the liver in which they make up approximately 40% of lymphocytes and localize to the sinusoidal walls under the endothelium. NK-T cells are also found in the bone marrow and smaller numbers in the spleen, lymph nodes and thymus. NK-T cells increase with age in all sites. The role of NK-T cells in mice and humans has been reviewed by Abo et al (1995). They may represent an intermediate stage in phylogenic development between NK cells and conventional T cells. Although their exact role is not clearly defined, it is thought that they are important for the surveillance of abnormal self cells, including rapidly dividing cells, cells infected with intracellular pathogens and malignant cells (Abo et al, 1995). In some cases, NK-T cells may play a role in regulating autoreactive lymphocytes and it has also been speculated that they constitutively eliminate abnormal cells associated with ageing. NK-T cells constitutively express the IL-2Rβ chain and respond rapidly to cells infected with intracellular pathogens, such as Listeria (Ohtsuka et al, 1995), herpes simplex (Hasegawa et al, 1996) and plasmodium (Abo et al, 1995).

Cytokine-induced killer (CIK) cells are NK-T cells expanded from peripheral blood mononuclear cells (PBMCs) by the timed addition of gamma interferon (γ-IFN) on d 1 and IL-2 and the monoclonal antibody OKT3 on d 2. After 21 d of continuous culture of these cells from patients with chronic myeloid leukaemia (CML), there is up to a 70-fold expansion of αβ TCR+CD3+CD56+ NK-T cells which show marked non-MHC-restricted cytotoxicity against a variety of cell lines and autologous and allogeneic CML progenitors (Schmidt-Wolf et al, 1991; Lu & Negrin, 1994; Hoyle et al, 1998). The bulk cultures of T cells have been termed CIK cells to differentiate them from IL-2-activated NK cells (LAK) cells and cytotoxic T lymphocytes (CTLs).

Cells with the phenotype CD3+CD56+ were first described in NK cell clones by Hercend et al (1984). These cells were further shown to be present in peripheral blood of normal donors and to lyse the NK-sensitive cell line K562 (Schmidt et al, 1986a,b). CD3+CD56+ cells were further characterized by Ortaldo et al (1991) who showed that these cells expressed HLA-DR, CD57, CD11b, CD8 (60%), CD4 (40%) and the αβTCR, but not CD16 or the IL-2 receptors. Southern blot analysis using a TCR β probe showed the cells were polyclonal.

Antibodies to CD3 act as a mitogenic stimulus for all T cells including NK-T, which can then be expanded by culturing in medium containing IL-2 (Ochoa et al, 1987; Anderson et al 1988; Kimoto et al, 1992; Ueda et al, 1993). The addition of γ-IFN on d 1 before the addition of IL-2, results in increased cytotoxicity (Schmidt-Wolf et al, 1991). IFN-γ stimulates monocytes to produce IL-12 (Hayes et al, 1995; Ma et al, 1996), which drives the cells to express a ‘Th 1’ phenotype. Removal of monocytes from the starting PBMCs led to poor expansion and culture with monocytes in a trans-well experiment also gave less expansion, suggesting that cell contact was also necessary. IL-12 synergizes with monoclonal antibody (mAb)-CD3 in inducing T-cell proliferation, which is additive to the effects of IL-2 (Zeh et al, 1993). IL-12 has been shown to enhance NK cytotoxicity within a few hours of incubation and it also increases and stabilizes transcription of IFN-γ mRNA by human T cells (Chan et al, 1991).

Precursors of CIK cells

CD3+CD56+ cells are derived from T cells and not NK cells (Lu & Negrin, 1994). NK cells (CD3CD56+) or T cells (CD3+CD56) were isolated from the peripheral blood and cultured under CIK conditions. The NK cells retained their phenotype after 30 d in culture. In contrast, the T cells differentiated into two populations; 43% were CD3+CD56+ and the remaining 57% were CD3+CD56. The percentage of CD3+CD56+ cells present in these cultures varies among individuals but is usually in the range of 20–50% using normal donors. Maximal generation of CD3+CD56+ cells was obtained from CD3+CD4CD8 cells, although both CD3+CD4+CD8+ and CD3+CD8+ cells could also generate CIK cells to a lesser extent. Interestingly, CD3+CD4+ cells were not suitable CIK precursors, but mature CIK cells may express either CD4 or CD8 (Lu & Negrin, 1994). CIK cells do not express the Fcγ receptor CD16 and, thus, do not mediate antibody-dependent cellular cytotoxicity.

Cytotoxicity mediated by CD3+/56+ cells

CIK and LAK cells have similar in vitro cytotoxicity against a variety of cellular targets including K562 and B-lymphoma cell lines such as SU-DHL-4 and OCI-Ly8. CIK cultures generate a greater number of lytic units than LAK cultures owing to the high proliferation rate of CIK cells (Lu & Negrin, 1994). Using d 4 LAK cell cultures (the day of maximal cytotoxicity), virtually all the cytotoxicity was mediated by cells expressing CD56. Cells expressing the α/β T-cell receptor (TCR) and the γ/δ TCR had no cytotoxic activity. In sharp contrast, when CIK cultures were studied after 14 d in culture, cytotoxicity was observed in cells expressing CD56 and the α/β TCR but not the γ/δ TCR (Schmidt-Wolf et al, 1993). In subsequent experiments, purified T cells were isolated from peripheral blood leucocytes (PBLs) and grown under CIK conditions. After 30 d, the cells were separated into CD3+CD56 and CD3+CD56+ populations of cells and tested for cytotoxicity. Virtually all the cytotoxic effect against OCI-Ly8 and K562 targets was mediated by the CD3+CD56+ cells. Further phenotypic analysis was performed which demonstrated that the CD3+CD56+ cells expressed CD8 (70%) and CD4 (20%). However, CD8 expression did not distinguish the cytotoxic population (Lu & Negrin, 1994). CIK cells efficiently kill a variety of tumour cell lines as well as fresh tumour cells, yet have only a minor effect on normal haematopoietic progenitor cells (Schmidt-Wolf et al, 1991; Hoyle et al, 1998).

The mechanisms by which CD3+CD56+ cells (NK-T/CIK cells) recognize and kill tumour targets have not been elucidated. Blocking studies with monoclonal antibodies to CD2, 3, 4, 6, 8, 28, 56, VLA-4, TCR-αβ, MHC class I and MHC class II have failed to inhibit killing, but antibodies against LFA-1 (α chain-CD 11a and β chain CD 18) and ICAM-1 (CD54) have shown substantial inhibition (Schmidt-Wolf et al, 1993). Earlier studies had, however, shown that anti-CD3 antibodies partially blocked killing of K562 by CD3+CD56+ cells (Schmidt et al, 1986b), although it is difficult to hypothesize a mechanism for this action. Later studies (Mehta et al, 1995) were unable to repeat the blocking effect of anti-CD3 but confirmed the role of LFA-1 in CIK-mediated lysis.

Cold target inhibition studies suggest that CIK cells may recognize targets using at least two pathways (Schmidt-Wolf et al, 1993). Killing of the CIK-sensitive cell line SU-DHL-4 could be inhibited by K562 and OCI-Ly8, whereas lysis of human umbilical vein endothelial cells (HUVECs) could not be inhibited by SLJ-DHL-4 or OCI-Ly8, suggesting different mechanisms.

CIK cells contain cytolytic granules, with perforin and granzyme activity, which are released to the extracellular space on binding with susceptible target cells or following cross-linking CD3 with an anti-CD3 mAb) adhered to plastic. This same antibody did not block cytolysis of susceptible tumour lines. Treatment of the CIK cells with db-cAMP, which inhibits the conversion of LFA-1 from a low avidity to a high avidity receptor for binding to ICAM-1, inhibited degranulation and cytotoxicity induced by both anti-CD3 mAb and target cells (Mehta et al, 1995). Cyclosporine A and FK506 inhibited anti-CD3-mediated degranulation of CIK cells but did not affect degranulation or cytotoxicity of CIK cells against tumour targets, indicating that LFA-1/ICAM-1 interactions are essential for killing of tumour targets, whereas T-cell receptor activation is not (Mehta et al, 1995).

Mouse models

To explore the in vivo capability of CIK cells, an animal model utilizing mice with severe combined immunodeficiency (SCID) engrafted with human lymphoma cells was used (Schmidt-Wolf et al, 1991; Lu & Negrin, 1994). Animals were first injected i.v. with 1 × 106 SU-DHL-4 tumour cells, which is a human B-cell lymphoma cell line with a t(14;18) chromosomal translocation. One day later, either LAK or CIK cells were injected i.v. or i.p. In control animals injected with the tumour cells alone, macroscopic evidence of disease was evident by approximately 35–40 d and the animals survived for a median of approximately 60 d (Fig 4). Animals injected with 4 × 107 LAK cells and tumour cells survived slightly longer yet there were no long-term survivors. In contrast, following the injection of 4 × 107 CIK cells and the same number of tumour cells, the median survival was 92 d (P < 0·001 compared with controls; P < 0·002 compared with NK cells) and 45% of the animals survived for > 100 d. CIK cells were effective whether given i.v. or i.p., but irradation (15 cGy) abrogated their in vivo activity, suggesting that proliferation is required.

Figure 4.

 Survival of transplanted mice given the BCL 1 cell line +/– cytokine-induced killer (CIK) cells.

The in vivo activity of CIK cells was not dependent upon exogenous IL-2 administration to the animals. Indeed, the co-injection of IL-2 and/or IL-12 with the CIK cells did not result in improved efficacy. Importantly CIK cells have been demonstrated to be effective in the autologous setting. Autologous CIK cells, given a month after injection of CP-CML cells, were able to inhibit the growth of CML and EBV lymphoma in SCID mice (Hoyle et al, 1998).

To test whether CIK cells mediate GvHD, a class I and class II mismatched BMT was established (Baker et al, 2001). In brief, Balb c (H2b) mice were lethally irradiated (8 cGy) and transplanted with bone marrow or purified stem cells (phenotype = CD34+, Thy1.1lo, Sca 1+, Lin) from BA (H2d) mice. All mice engrafted with no GvHD. If 1 × 107 spleen cells were also given on d 0, all mice died of severe GvHD by d 10, whereas 1 × 108 CIK cells infused on d 0 caused no GvHD. To demonstrate a graft-vs.-leukaemia (GvL) effect, the recipient mice were also given 1 × 103 BCL 1 cells (a mouse lymphoma cell line) i.p. on d 0. Mice given lymphoma cells alone without BMT died from disseminated disease after 5 weeks. Those receiving purified stem cells also died from disease after 5 weeks, whereas mice given whole BM (containing approximately 3% of T cells) survived considerably longer with some long-term survivors. Most notably, mice transplanted with purified stem cells and 1 × 107 CIK cells had the best overall survival (Fig 4).

These data demonstrate the potential clinical utility of CIK cells and support the hypothesis that CIK cell-based immunotherapy may prove to be more efficacious than LAK+ IL-2-based treatment. This is based on the observations that CIK cells expand readily, are more potent in vivo than LAK cells and do not require the exogenous administration of cytokines such as IL-2 for in vivo activity, the use of which has been associated with considerable toxicity in some studies. CIK cells may be very effective as ‘donor lymphocyte’ infusions following allogeneic stem cell transplants as, in common with classic NK cells, they do not appear to mediate GvHD.

Natural killer cells

Cure of leukaemia by allogeneic haematopoietic transplantation is achieved through the concerted action of two mechanisms; a lethal myeloablative and immune suppressive radio- and/or chemotherapy-based conditioning regimen prior to transplantation, and the ability of the immune cells in the graft to recognize and eliminate the leukaemia cells that survive the conditioning regimen (GvL effect). Although specific immune responses have been documented against CML cells and haematopoietic tissue-specific minor antigens (Bonnet et al, 1999; Molldrem et al, 2000), as a rule the donor immune system exerts its GvL effect through T cell-mediated alloreactions directed against host MHC antigens which have a broad tissue distribution. Because graft-vs.-host T-cell reactivity does not distinguish between the MHC antigens displayed on host leukaemia cells and tissues, it may cause GvHD. Nonetheless, encouraging results are obtained with transplant strategies in which part of the standard conditioning regimen prior to transplantation is replaced by T-cell alloreactivity in order to (i) establish mixed chimaerism and host-vs.-graft tolerance, (ii) exert a GvL effect at the time of transplant (by T cells in the graft), and (iii) exert a GvL effect by later donor lymphocyte infusions (DLI) (Carella et al, 2000). Even in genotypically HLA-matched transplants, GvHD is the major obstacle to the full success of these strategies.

In unmanipulated mismatched stem cell transplants, such as those from haploidentical family members, owing to the high frequency of alloreactive donor T cells recognizing major MHC antigens, T-cell alloreactions are unmanageable as they would usually cause lethal GvHD. Indeed, transplants from full HLA-haplotype mismatched family donors for the ≈40% of leukaemia patients who cannot find a matched graft have long been impossible because of lethal GvHD. In the two principal centres for haploidentical donor transplants, extensive T-cell depletion has been used to prevent GvHD (Henslee-Downey et al, 1997; Aversa et al, 1998; Reisner & Martelli, 1999). Moreover, the Perugia group used high doses of peripheral blood stem cells and no immune suppressive treatment was given to the patients after transplant. Extensive depletion of donor T cells from the graft is essential to the success of these transplants, as it is the only measure that prevents GvHD. The impossibility of exploiting T-cell alloreactivity in a mismatched transplant setting, which must by definition be extensively T cell-depleted, raises the question, ‘Is there any chance that such T cell-depleted mismatched transplants can exert a GvL effect?’

NK cells are regulated by both inhibitory and activatory signals. A number of molecules that mediate NK cell inhibition have been cloned over the past 10 years and their ligands are almost exclusively class I MHC molecules (Moretta & Moretta, 1997; Lanier, 1998). Some of these receptors (‘KIRs’ from Killer cell Ig-like receptors) are specific for determinants shared by certain class I alleles, and each KIR is expressed by a subset of NK cells. Therefore, in the NK repertoire some NK cells recognize and are blocked by specific class I alleles (Fig 5). There are three major NK inhibitory receptors that are able to recognize specific class I alleles and, therefore, three major NK cell subsets which are capable of specific recognition. Each of these NK cells is blocked by the recognition of an epitope shared only by certain alleles. The first NK cell type (bearing KIR2DL2/3) recognizes the motif Asp77-Lys80 and is therefore blocked by HLA-C group 1 alleles (Cw1, 3, 7, 8, 9), the second NK cell type (bearing KIR2DL1) recognizes Ser77-Asn80 and is therefore blocked by HLA-C group 2 alleles (Cw2, 4, 5, 6), the third NK cell type (bearing KIR3DL1) is blocked only by HLA-B alleles sharing the Bw4 supertypic specificity. In addition, KIR3DL2 is specific for a determinant expressed by HLA-A3/A11. Thus, NK cells have a limited view of class I polymorphism but cells can be responsible for alloreactions when the mismatched target cells do not express the class I alleles that block every NK cell in the repertoire. In other words, in order to be killed, the allogeneic target (eg. the leukaemic or T cells in a transplant recipient) must fail to express at least one of the class I allele groups expressed by the donor cells. In this way, one of the NK cells in the donor's repertoire will not find its inhibitory class I ligand and its lytic pathway will be activated (Valiante & Parham, 1997).

Figure 5.

 The basis of natural killer (NK) cell alloreactivity. 1. NK cell lysis is negatively regulated by receptors for MHC class I. 2. Some receptors (killer cell Ig-like receptors; KIRs) are specific for determinants shared by certain class I alleles. 3. Each KIR is expressed by a subset of NK cells.

Impact of NK cell alloreactivity on clinical haematopoietic transplantation

The mismatched transplant combination in which the donor NK cells can exert alloreactions against the host is relatively frequent. Under these conditions, the Perugia group noticed that the engrafted stem cells gave rise to a transient (1–3 months) wave of reconstituting NK cells whose repertoire was identical to that originally displayed by the donor, including high-frequency donor-vs.-recipient alloreactive NK clones (Fig 6) (Ruggeri et al, 1999). This is different from the physiological situation in which only NK cells bearing the correct inhibitory receptors that recognize all self MHC class I ligands are allowed to develop. The data imply that, at least in a transplant situation, the educational process leading to deletion of self-reactive NK cells takes at least 1–3 months. Strikingly, such post-grafting regeneration of donor-vs.-recipient alloreactive NK cells does not cause clinical GvHD, leading one to conclude that non-lymphohaematopoietic tissues lack ligands able to activate NK cell lysis. This conclusion is supported by the so-called ‘hybrid resistance’ phenomenon observed in murine transplantation models (Yu et al, 1996).

Figure 6.

 Post-transplant regeneratIon of same NK cell repertoire in HLA-mismatched recipient expressing HLA-C 1 and HLA-C 2, but lacking HLA-Bw4 leads to lysis by KIR3 clones (Ruggeri et al, 1999).

In vitro leukaemia killing assays have demonstrated that a minority of common lymphoblastic leukaemias are killed by alloreactive NK clones and that this is associated with failure of these leukaemias to express LFA-1, an adhesion molecule instrumental to the process of NK-to-target binding (and in contrast to all other NK-susceptible targets, which do express LFA-1). Remarkably, however, 100% of acute and chronic myeloid leukaemias are killed by all alloreactive NK clones (Ruggeri et al, 1999). This suggests that KIR epitope-mismatching in the graft-vs.-host direction could predict anti-myeloid leukaemic effects in vivo. Long-term follow-up of the recipients of haploidentical transplants in Perugia now provides evidence of the clinical impact of this phenomenon (Ruggeri et al, 2000a). Donor–recipient pairs were divided into two groups, those without KIR epitope incompatibility in the GvH direction and presumably no chance for donor-vs.-recipient NK cell reactivity, and those with KIR epitope incompatibility in the GvH direction. In this case, donor NK cells have the potential to exert anti-recipient alloreactivity. The analysis is done separately for myeloid and lymphoid leukaemias. To date, in spite of the intensity of the pretransplant conditioning regimen, myeloid leukaemia relapses were high in transplants without KIR epitope mismatch (17 out of 47 – 36%). This reflects the high relapse risk of patients entered in this protocol to date. Remarkably, transplants with KIR epitope mismatches in the GvH direction almost completely controlled relapses, as only one relapse was observed in this group of 28 patients (3·6%). As might have been predicted in light of the in vitro experiments, relapses of the NK-resistant target, ALL, were distributed equally in the two groups. Consequently, donor-vs.-recipient NK cell alloreactivity has become a major criterion for donor selection in mismatched haematopoietic stem cell transplants in Perugia and is being studied actively by other groups performing similar mismatched transplants.

The use of donor alloreactive NK cells for conditioning in ‘Mini’ mismatched transplantation – a murine model

KIR epitope incompatibility in the GvH direction is associated with lower rejection rates in clinical transplants (Ruggeri et al, 1999; and unpublished observations), indicating that a lethal conditioning combined with the spontaneous post-transplant generation of anti-host NK cells may favour engraftment. In the ‘hybrid resistance’ experiment (Yu et al, 1996), NK alloreactivity is exerted in the host-vs.-graft (HvG) direction and causes bone marrow rejection because specific NK cells in the host are not equipped with the correct inhibitory receptors to recognize and, hence, to be blocked by the MHC class I alleles displayed on donor bone marrow cells. However, hybrid resistance also shows that NK alloreactivity may not target tissues other than haematopoietic cells because the hybrid mouse readily tolerates organ grafts. In a reversal of the hybrid resistance transplantation partners (Fig 7), the hybrid H-2d/b F1 mouse was used as donor (of bone marrow and NK cells) and the H-2b/b parent was used as recipient (Ruggeri et al, 2000b). Hence, NK reactivity was in the GvH direction. Irradiation below 8·5 Gy was non-lethal, mice rejected bone marrow and recovered uneventfully. After 5–7 Gy irradiation, the infusion of hybrid H-2d/b mouse Ly49A+/G2+, H-2b/b reactive, NK cells into H-2b/b hosts killed the mice (spleen and bone marrow counts were greatly reduced). Adding donor bone marrow (10–20 million cells) rescued all mice that displayed full donor chimaerism in their bone marrow and spleens, without GvHD. The minimum effective Ly49A+/G2+ cell dose was 100 000/mouse (= 4 million cells/kg body weight, a feasible dose with human NK clones). Thus, donor-vs.-recipient NK cell alloreactivity can promote sufficient immune suppression and myeloablation to allow engraftment of MHC disparate bone marrow transplants, in combination with a non-lethal (‘mini’) preparative regimen, without any risk of causing GvHD.

Figure 7.

 NK alloreactivity for condItIoning prior to bone marrow.

In conclusion, approximately 10 years of intensive research in mice and humans has unravelled the biology of natural cytotoxicity. Thus, it is now known that NK cell lysis is negatively regulated by several, clonally distributed receptors that recognize MHC class I molecules. As a consequence, NK cells in any given individual's repertoire, when confronted with an allogeneic target that does not express the correct class I alleles that can block NK cell activation, give rise to specific alloreactions targeted at host haematopoietic tissues. Significantly, only major MHC disparities lead to NK cell alloreactions. In clinical haematopoietic transplantation across major MHC barriers, these reactions might have been responsible for worsening the risk of rejection and GvHD. In contrast, NK cell alloreactions are not only harmless but indeed carry several therapeutic advantages. No NK contribution to rejection has been observed. Alternatively, there is now very clear evidence from mouse preclinical models (Ruggeri et al, 2000b), and from in vitro and in vivo observations in > 100 transplanted patients (Ruggeri et al, 1999, 2000a), that NK cell alloreactions, when directed in the donor-vs.-recipient direction, exert beneficial effects. These are manifested in additional conditioning of the host peritransplant by ablation of the host immune system and in control of leukaemia relapse. Most importantly, several in vivo observations in mice and patients exclude the possibility that NK alloreactions mediate GvHD. Thus, NK cell alloreactions, in contrast to T-cell alloreactivity, are restricted to host lymphohaematopoietic targets and leukaemia cells, and spare other tissues. In addition, given the perfect concordance between specific host–donor MHC class I disparities and the in vitro and in vivo occurrence of NK alloreactions, one can select the one donor whose HLA disparity with the patient will guarantee the post-transplant benefits of NK alloreactions. Thus, predictability, safety and effectiveness are distinctive features of this system. In the future, one may envisage the use of low-toxicity conditioning regimens to transplantation which rely upon the infusion of alloreactive NK cells for engraftment and control of leukaemia relapse for the treatment of patients and diseases for which the hazards imposed by the current high-intensity conditioning regimens are not justified.

Immunotherapy with autologous natural killer cells

Patients who achieve complete remission after chemotherapy are potential beneficiaries of autologous immunotherapy, with or without concomitant stem cell transplantation. This concept is not new; the application of recombinant IL-2 in the 1980s led to a number of in vitro studies and clinical trials. Herberman's group in Pittsburgh were the first to report that IL-2 treatment of mononuclear cell preparations from leukaemic patients generated NK cells that were lytic to autologous and allogeneic leukaemic blasts (Adler et al, 1988), although it had already been reported that γ-IFN, a product of activated NK cells, could induce lysis of AML blasts (Price et al, 1987). The bone marrow transplant group at the Royal Free Hospital in London also demonstrated the in vitro lysis of autologous myeloma plasma cells (Gottlieb et al, 1990) and EBV-transformed B lymphocytes (Leger et al, 1987), and the suppression of AML growth in semisolid cultures (Gottlieb et al, 1989) by IL-2-activated NK cells.

Of the early clinical trials, some were done in recipients of autologous BMT (ABMT) (Hamon et al, 1993; Attal et al, 1995) while others involved patients who had received chemotherapy alone (Adler et al, 1988; Gottlieb et al, 1989). Despite initially encouraging reports, the overall results, including those from large randomized but unreported studies, have been disappointing and have been reviewed elsewhere (Goodman et al, 1998). However, in a trial from Seattle in 14 patients with AML in relapse or second complete remission (CR 2), the administration of IL-2 +/– LAK cells after ABMT (Benyunes et al, 1993) gave a 4-year survival probability of 71%, suggesting that the benefit of IL-2 therapy may require the presence of minimal residual disease to generate a leukaemia-specific immune response. Subcutaneous administration of low-dose IL-2 after ABMT was also shown to be well tolerated and to generate activated NK cells that lyse tumour cell lines in a study from the University of Minnesota (Miller et al, 1997). Twelve patients (six lymphoma, six breast cancer) were entered into this phase I dose-escalation study with a median time of commencement of d 94 post ABMT. Self-administered IL-2 infusions were conducted for 84 consecutive days with the best tolerated dose reported as 0·25 × 106 U/m2/d. Two out of six patients receiving this dose experienced toxicities resulting in cessation of administration, one thrombocytopenia and one mild neutropenia; both resolved without hospitalization. IL-2 infusions were associated with the in vivo generation of activated NK cells, as demonstrated by increased cell-mediated lysis of NK-resistant lymphoma (Raji) and breast cancer (MCF-7) cell lines.

In a recent report from the MD Anderson Cancer Center (Cortes et al, 1999), AML patients in first CR who received low-dose IL-2 by continuous infusion over 12 weeks plus weekly bolus doses showed improved disease-free (6 out of 18 versus 7 out of 36 at 3 years) and overall survival (9 out of 18 versus 10 out of 36 at 3 years) compared with matched historical controls who received the same chemotherapeutic regimen without IL-2. The small number of patients in this pilot study precluded the results from achieving statistical significance, but the increased survival and the fact that a tolerable dose regimen was established for IL-2 in these patients suggest that a further trial is warranted.

A role for autologous passive immunotherapy in relapsed ALL has recently been reported (Tewari et al, 1999). The authors described the treatment of a mother in CR 1 of ALL who relapsed during the third trimester of pregnancy. After reinduction chemotherapy she received radiotherapy and infusion of autologous IL-2-activated peripheral blood stem cells. In contrast to the four previous reports of ALL relapsing during pregnancy, this patient remained in CR beyond 22 months and was alive and disease-free at the time of the report.

CML is generally regarded as the leukaemia most susceptible to immune control, although this is typically thought to be mediated by T cells, largely as a result of the high incidence of relapse after T cell-depleted allogeneic BMT in this group of patients (Horowitz et al, 1990). However, it should be remembered that typical T cell-depletion strategies, such as sheep red blood cell rosetting, anti-lymphocyte globulin or Campath, also remove NK cells from the donor graft. Pierson and Miller (1997) demonstrated that NK cells derived from patients with CML showed a decreased proliferative capacity compared with NK cells from normal donors and the degree of abnormality increased with disease progression. They further demonstrated that NK cells from CML patients showed defective lytic function, but this could be overcome by in vitro culture with IL-2. These IL-2-stimulated patient NK cells were able to suppress colony growth of autologous CML progenitors in vitro but the levels of IL-2 required were greater than could be safely administered to patients. Pierson and Miller (1997) hypothesized that low-dose subcutaneous IL-2 combined with administration of ex vivo IL-2-activated NK cells (as used by Benyunes et al, 1993) might represent a safe alternative to in vivo high-dose IL2 for these patients. However, given the success of α-IFN therapy in some CML patients and observations of the effect of α-IFN on autologous NK cell activity in two cases of AML (Lowdell et al, 1997, 1999), the routine monitoring of NK cell responses in patients with CML may be informative.

Despite the long interest in the therapeutic potential of NK/LAK cells in haematological malignancies, few groups have looked for a role for these cells in the control of leukaemia in the absence of IL-2 therapy. Recently, reports have appeared of patients who spontaneously develop immunity to residual leukaemia after ABMT and even after chemotherapy alone. Robertson et al (1996) from the MD Anderson Cancer Center demonstrated that NK cells can arise after conventional treatment of CLL with fludarabine which lyses autologous CLL blasts in vitro but does not lyse autologous remission bone marrow mononuclear cells. In a study of recipients of ABMT for acute leukaemias, we reported the spontaneous generation of leukaemia-lytic NK cells post transplant (Lowdell et al, 1997). The loss of these cells in two patients (T-ALL 1; AML M4eo 1) was associated with rapid relapse, while the maintenance of the response in one other case (AML M4) has been associated with continued remission beyond 6 years. One of the relapsing patients (AML-M4 eo) was returned to CR with further chemotherapy and, following treatment with α-IFN, recovered her NK-mediated leukaemia-lytic activity and has remained in molecular CR for the past 6 years.

Alpha-interferon has also been used for the in vivo generation of NK-mediated leukaemia cytolytic activity (LCA) in CR after chemotherapy alone (Lowdell et al, 1999). This report described a patient with poor-risk AML who relapsed from first CR and was unwilling to undergo high-dose chemotherapy with stem cell rescue. In second chemotherapy-induced CR, the patient had no evidence of LCA in an in vitro assay and she commenced IFN-α (Roferon). She subsequently developed high levels of leukaemia-specific cytotoxicity and has remained in second CR for over 4 years.

These findings led to the hypothesis that long-term disease-free survival after chemotherapy with or without autologous BMT for acute leukaemias might require the generation of NK-mediated LCA. The Royal Free group has recently found that a cut-off level of NK-LCA in peripheral blood samples of patients in remission after chemotherapy can be determined that predicts disease-free survival beyond 2 years with high specificity and sensitivity (Fig 8) (Lowdell & Koh, 2000). Furthermore, the cells which mediate this activity have been identified as a subset of NK cells that express the CD8α homodimer through which they can receive signals, leading to mobilization of intracellular calcium and upregulation of the activation antigen CD69. In support of the non-MHC-restricted action of these cells, they have been shown also to lyse allogeneic AML blasts (Fig 9). The molecular interactions involved in the activation of these NK cells remains elusive and, certainly in the autologous setting in which the target cells express normal levels of class I MHC molecules, it appears that specific activatory signals rather than simple lack of ligands for inhibitory signalling molecules are required. Cloning of NK cells that mediate autologous LCA will facilitate the identification of these activatory molecules and their ligands on leukaemic blasts.

Figure 8.

 Determination of cut-off level of NK cell leukaemia cytolytic activity (LCA) level that predicts disease-free survival (DFS) beyond 2 years.

Figure 9.

 Lysis of allogeneic AML blasts by NK cells illustrating their non-MHC-restricted action.

Plainly autologous immune responses to acute leukaemias can arise and may be sufficient for the eradication of residual disease that has escaped first-line treatment. In future, it may be possible to identify patients in whom this activity is absent and who would therefore benefit from the immunotherapy of an allogeneic stem cell transplant or admission into a trial of autologous or allogeneic non-MHC-restricted immunotherapy. However, the idea that immunotherapy is the panacea for chemotherapy-resistant leukaemias is unrealistic. As shown in the case of the AML M4 sample in Fig 9, some leukaemic blasts are relatively or totally resistant to immune-mediated lytic mechanisms. It has been reported that susceptibility of colon carcinoma cells to NK-mediated lysis is dependent upon ICAM-1 and LFA-1 expression (Oblakowski et al, 1991; Rivoltini et al, 1991) and we have similar observations in ALL blasts. Recently, Uharek's group in Kiel have shown that resistance to perforin binding by leukaemic blasts is a mechanism to evade immune-mediated lysis (Lehmann et al, 2000).

Conclusions

The earlier disappointing results associated with attempts to induce non-MHC-restricted cellular responses by cytokine therapies 10–15 years ago led to the abandoning of NK/LAK strategies and the study of MHC-restricted forms of cellular immunity. With the notable exception of melanoma and, given the success of donor leucocyte infusion and the recent data from Molldrem et al (2000), perhaps also CML, a specific T-cell response has not been demonstrated for any tumour. In contrast, many solid tumours and haematological malignancies are known targets of non-MHC-restricted cytolytic cells, both autologous and allogeneic. Clinical trials with CIK cells in CML are underway in the USA and the commercialization of ex vivo expansion systems for γδ T cells for allogeneic immunotherapy after αβ-T cell-depleted allogeneic stem cell transplantation is under discussion. In HLA-haploidentical donor stem cell transplants, donor decisions are already being made on the basis of KIR mismatching to maximize allogeneic NK cell-mediated GvL. Evidence that patients receiving non-transplant-based treatments for acute and chronic leukaemias can make immune responses to their original malignancy suggests that we may have been inadvertently using immunotherapy for many years and encourages us to believe that we will be able to enhance this for therapeutic benefit in the future.

Given our increasing understanding of the activating ligands of all the cell types described, together with improvements in our ability to manipulate cells ex vivo, the prospect of real advances in the immunotherapy of leukaemias with non-MHC-restricted cytotoxic cells has never been better. We believe that the use of allogeneic or autologous non-MHC-restricted cytotoxic cells will substantially improve the outcome for recipients of conventional chemotherapy for leukaemia.

Ancillary