In HLA-nonidentical bone marrow transplantation, we studied the characteristics of donor NK cells, recipient leukemia cells, and the cytokine environment that predict the antileukemia effects of allogeneic NK cells. We found that the risk of relapse in pediatric patients with hematologic malignancies was best predicted by a model taking into consideration the presence of inhibitory killer cell Ig-like receptors (KIRs) on the donor’s NK cells and the absence of corresponding KIR ligand in the recipient’s HLA repertoire (a receptor-ligand model). The risk of relapse was prognosticated less precisely by the Perugia donor-recipient KIR ligand-ligand mismatch model or by a natural cytotoxicity model. In contrast to the ligand-ligand model, we found that the new receptor-ligand model was accurate when analysis was applied to patients with lymphoid malignancy. These findings corroborate our observations that the recipient’s KIR repertoire, which was derived from highly purified, HLA-disparate CD34+ cells, resumed a donor-specific pattern within 3 mo of transplantation, but did not correlate evidently with the donor or recipient ligand repertoire. In an in vitro assay and an in vivo mouse model, human NK cell cytotoxicity toward human leukemia cells with 11q23 chromosomal rearrangement increased with the number of receptor-ligand mismatch pairs or prestimulation with IL-12 and IL-18. These findings provide new insights into the determinants of antileukemia effects of allogeneic NK cells and therapeutic strategies.

Transplantation of allogeneic bone marrow or blood hemopoietic stem cells is an established method for the treatment of high risk hematologic malignancies (1). Less than 30% of patients have a sibling donor who is HLA identical. Transplantation using HLA-nonidentical donors, such as unrelated or mismatched family donors, is associated with increased risk of graft rejection and graft-vs-host (GvH) 3 disease. Despite these disadvantages, transplantation using an HLA-nonidentical donor provides a unique setting in which donor NK cells may contribute to antileukemia effects (2).

Human NK cells are regulated by inhibitory receptors for HLA class I molecules (3). A family of NK receptors, the killer cell Ig-like receptors (KIRs), is specific for epitopes shared by groups of HLA class I molecules (4). To avoid autoimmunity, each NK cell usually expresses at least one inhibitory receptor that recognizes one or another self HLA molecule (4, 5). In the setting of transplantation of HLA-nonidentical cells, GvH reactions may occur when none of the KIRs in a subset of donor NK cells is engaged by the host HLA molecule, because the host’s HLA repertoire lacks the appropriate ligand (6). Similarly, graft rejection is possible on the rare occasion when the host’s NK cells survive the high dose cytoreductive therapy given before transplantation. Animal studies of NK cell-mediated GvH reactions have demonstrated a predominant lysis of host lymphohemopoietic cells in the absence of systemic GvH reactions to other nonhemopoietic tissues (7), suggesting that clinical graft-vs-leukemia response may occur in the absence of GvH disease (8). These studies (6, 7) led to a clinical practice of selecting a ligand-ligand incompatible donor with an HLA mismatch that favors NK cell-mediated GvH reactions (2, 8). The rationale for this clinical practice, however, has been challenged by other investigators (9) because our understanding of the biological determinants of antileukemia effects of NK cells remains incomplete. In this study we conducted clinical and laboratory investigations in parallel to obtain additional fundamental insights into the antileukemia response of allogeneic NK cells.

Routine procedures for the selection of donor, mobilization of stem cells, conditioning, and supportive care have been described previously (10). These patients (0.3–17.9 years of age) received an experimental haploidentical transplant because their estimated disease-free survival was ≤25% with chemotherapy, and they did not have an HLA-matched related or unrelated donor. The transplantations were performed in Memphis (n = 22) or Tuebingen (n = 14) for myeloid (n = 17) or lymphoid (n = 19) malignancy. All grafts were purified for CD34+ cells using CliniMACS systems (Miltenyi Biotec, Auburn, CA). Because all the grafts contained <3 × 104 CD3+ cells/kg, GvH disease prophylaxis was not given to any patient after transplantation. Monitoring of the KIR repertoire and the natural cytotoxicity of NK cells was performed on blood samples obtained from patients treated in Memphis at 1, 3, 6, 9, and 12 mo after transplantation.

Three models were used to dichotomize the risk of primary disease relapse after transplantation into high and low risk categories. A patient was categorized as being at low risk 1) by the ligand-ligand model if at least one group of donor KIR ligand was missing in the recipient’s ligand repertoire, a model proposed by Ruggeri et al. (6) of Perugia University (6, 7); 2) by the receptor-ligand model if at least one KIR gene expressed in the donor’s NK cell repertoire did not recognize any of the HLA molecules in the recipient’s ligand repertoire, a model defined a priori based on the known ligands reported for each KIR (4); 3) by the cytotoxicity model if the NK cell cytotoxicity against K562 cells was higher than the median 1 mo after transplantation. The difference between the first and the second model was that the ligand-ligand model assumed the expression of corresponding receptors for all groups of (donor) self KIR ligands (at least in some donor NK cells), whereas the receptor-ligand model directly measured the receptor repertoire of the donor before transplantation. Risk classification using these two models was determined a priori for each patient before transplantation. The third model measured the general cytotoxicity of engrafting NK cells against standard targets during the stage of minimal residual disease after high dose cytoreductive therapy. The 1 mo time point was chosen because previous studies demonstrated that NK cells usually engraft in 2–3 wk, and alloreactive clones are not found in any patients >3 mo after transplantation (6).

The probability of relapse was estimated and compared, as described by Ruggeri et al. (7), from the time of transplantation to the time of relapse of primary disease using the method of Kaplan and Meier and the log-rank statistics. Cox regression models were used for multivariate analysis with forward stepwise addition of covariate one at a time. The Wilcoxon rank-sum test or trend test was used to compare in vitro and in vivo cytotoxicities.

All samples were HLA-typed by the serologic and DNA methods used to match unrelated donors, as previously described (11). The nomenclature for inhibitory KIRs followed the proposal of the Seventh HLDA Workshop (Harrogate, U.K., June 2000) (12), and that used for KIR ligands followed the proposal of Vilches and Parham (4): CD158e1 (3DL1) is specific for HLA-B allotypes expressing the serologic Bw4 epitope (HLA-BBw4), CD158a (2DL1) is specific for HLA-C allotypes with lysine at position 80 (HLA-CLys80), and CD158b1/b2 (2DL2/2DL3) are specific for B*4601 and HLA-C allotypes with asparagine at position 80 (HLA-CAsn80) of the α-1 helix.

Surface expression of KIRs was determined by flow cytometric analysis using mAbs (13): CD158e1 was detected by DX9 alone (BD Immunocytometry Systems, San Jose, CA) (14), CD158a by EB6 (Immunotech, Marseilles, France) (15, 16) and HP-3E4 (BD Immunocytometry Systems) (17), and CD158b1/b2 by CH-L (BD PharMingen, San Diego, CA) (18), GL183 (Immunotech) (15, 16), and DX27 (BD PharMingen) (19). Analysis were gated on cells positive for CD56-allophycocyanin (Immunotech) and negative for CD3-ECD (Immunotech) and was performed on a BD LSR using CellQuest software (BD Immunocytometry Systems). The patterns of KIR expression were considered similar if the difference in Ab staining was within 50% of the average of the two samples, a cutoff estimated from data for identical twins (20). DX9 alone was used to detect CD158e1 expression, because no cross-reactivity with other KIRs has been demonstrated (19, 21). In contrast, multiple combinations of Abs conjugated with either PE or FITC in six separate tubes were always used to phenotype CD158a and CD158b1/b2. In addition, RT-PCR analysis, as described by Uhrberg et al. (22), was performed on samples from individuals whose repertoire of KIR expression was difficult to define by flow cytometry alone (e.g., ≤1% Ab reactivity or low fluorescent intensity; Fig. 1) and on random samples throughout the study for quality assurance (n = 8). Genotyping of KIRs was performed using the method described by Gagne et al. (23). Genotyping and phenotyping for other HLA class I inhibitory receptors were not performed, because either the receptor was present ubiquitously in each person (CD158d, CD158k, CD158z, and CD94:NKG2), or their ligand specificity was not certain (CD158f, CD158k, and CD158z) (4).

FIGURE 1.

Evaluation of KIR repertoire by flow cytometry. Analysis of KIR expression was performed on cells that were CD56+ and CD3 (R1). Samples that were difficult to assess by flow cytometry alone (e.g., low fluorescent intensity and low frequency of CD158a, as shown) were subjected to RT-PCR analysis to determine more accurately the KIR repertoire.

FIGURE 1.

Evaluation of KIR repertoire by flow cytometry. Analysis of KIR expression was performed on cells that were CD56+ and CD3 (R1). Samples that were difficult to assess by flow cytometry alone (e.g., low fluorescent intensity and low frequency of CD158a, as shown) were subjected to RT-PCR analysis to determine more accurately the KIR repertoire.

Close modal

When highly purified NK cells were needed for experiments, the CliniMACS system was used to first deplete a mononuclear cell population of CD3+ cells and then enrich for CD56+ cells. The cytotoxicity of NK cells in vitro was determined by a standard europium release assay, as previously described (13, 24). Target cells included standard K562 cells, 721.221 cells transfected with Cw03 or Cw04 genes, and human leukemia cell lines with t(4,11)(q21;q23) MLL rearrangement (provided by Drs. R. Greil (University of Tuebingen, Tuebingen, Germany) and D. Campana (St. Jude Children’s Research Hospital, Memphis, TN)). Three MLL-rearranged lines were chosen, because one expressed HLA-CAsn80 alone (MLL-1), another expressed HLA-CAsn80 and HLA-CLys80 (MLL-2), and the third expressed HLA-CAsn80, HLA-CLys80, and HLA-BBw4 (MLL-3): MLL-1 (B8, B62, Cw03, Cw07), MLL-2 (B14, B18, Cw08, Cw15), and MLL-3 (B14, B37, Cw06, Cw08). We are interested in studying leukemias with MLL rearrangement because they are common in infant and secondary leukemias that are highly resistant to conventional chemotherapy (25, 26) and may benefit from novel therapeutic strategies (27). We investigated the determinants of NK cell cytotoxicity toward these leukemia cells to develop a new treatment approach.

The cytotoxicity of NK cells against the MLL cell lines was also assessed in vivo using an nonobese diabetic/SCID mouse model. Control mice received only leukemia cells (20 × 106) via tail vein injections; treated mice received leukemia cells and NK cells in a 1:1 cell ratio (20 × 106 each). When leukemia cell burden was the outcome of interest, an NK cell-treated mouse was sacrificed on the same day that a control mouse was moribund with signs of leukemia (or vice versa if an NK-treated mouse was sick first). The numbers of cells in the spleen and bone marrow of the two femurs were counted (Z1 cell counter; Coulter Diagnostics, Miami, FL), and the percentages of leukemia cells were determined by flow cytometry using mAbs specific for the leukemia cells that had been injected into the mouse. When survival was the outcome of interest, each mouse was sacrificed on the day when it became moribund, and the cells from their marrow and spleen were examined by flow cytometry to confirm that the cause of illness was leukemia.

We followed prospectively the outcomes of 36 consecutive pediatric patients with high risk hematologic malignancy who received an HLA-haploidentical peripheral blood stem cell transplant from a family member. We used three models to dichotomize a priori the risk of primary disease relapse for each patient into high and low risk categories. HLA testing revealed that 28 of the 36 donor/recipient pairs were KIR ligand-ligand compatible and were therefore classified according to the ligand-ligand model as being at high risk for relapse. Analysis of the KIR expression on the donor’s NK cells and HLA testing of the recipient’s cells demonstrated that 13 recipients had appropriate KIR ligands for all KIRs in the donor’s repertoire. These 13 patients were classified as being at high risk of relapse according to the receptor-ligand model. For the cytotoxicity model, patients whose NK cell cytotoxicity against K562 cells was less than the median 1 mo after transplantation were considered at high risk of relapse. Among these three models, relapse of primary disease was best predicted by the receptor-ligand model (Fig. 2). Although the accuracy of the ligand-ligand model was comparable to that of the receptor-ligand model for their low risk category, the ligand-ligand model misclassified 13 patients who were classified correctly as low risk by the receptor-ligand model as being at high risk. The NK cell cytotoxicity model was the least accurate. When analysis was limited to patients with lymphoid malignancy, the receptor-ligand model remained the only accurate model statistically (Fig. 3).

FIGURE 2.

Prediction of relapse. Three models (the cytotoxicity, ligand-ligand, and receptor-ligand models) were used to dichotomize the risk of relapse of primary disease. The risk of relapse was best predicted by the receptor-ligand model. Samples for cytotoxicity assay were obtained only from patients treated in Memphis (n = 22), and four had insufficient cells.

FIGURE 2.

Prediction of relapse. Three models (the cytotoxicity, ligand-ligand, and receptor-ligand models) were used to dichotomize the risk of relapse of primary disease. The risk of relapse was best predicted by the receptor-ligand model. Samples for cytotoxicity assay were obtained only from patients treated in Memphis (n = 22), and four had insufficient cells.

Close modal
FIGURE 3.

Relapse in patients with lymphoid malignancy. Patients were dichotomized using the receptor-ligand model. Analysis was limited to patients with lymphoid malignancy (n = 19) and excluded patients with myeloid malignancy.

FIGURE 3.

Relapse in patients with lymphoid malignancy. Patients were dichotomized using the receptor-ligand model. Analysis was limited to patients with lymphoid malignancy (n = 19) and excluded patients with myeloid malignancy.

Close modal

Multivariate analyses were performed to examine whether the statistical significance of the receptor-ligand model held up when adjusted for potential confounders. We found that receptor-ligand incompatibility remained the only significant prognostic factor (all p < 0.042), not age (p = 0.20), disease (myeloid vs lymphoid; p = 0.89), remission status (p = 0.10), or number of CD34+ cells (p = 0.29) or CD3+ cells (p = 0.09) in the graft. The median time to relapse for the patients in the high risk category of the receptor-ligand model was only 99 days.

A concurrent laboratory investigation was performed to evaluate the ligand-ligand model and receptor-ligand model by studying the KIR repertoire of 50 normal donors to determine the relationship between KIR repertoire and self HLA. CD158a, CD158b1/b2, or CD158e1 was found in the repertoire of many donors (24–36%) with no corresponding ligand, whereas the HLA-BBw4, HLA-CLys80, or HLA-CAsn80 allotype was present in a few donors (0–8%) with no corresponding KIR (Fig. 4). The former observation provided an explanation for why the KIR ligand-ligand model might misclassify many ligand-ligand-compatible patients who were not at high risk of relapse, and the latter observation provided an explanation for why the ligand-ligand model was quite accurate for the ligand-ligand-incompatible patients in its low risk category. Misclassification may involve either CD158a, CD158b, or CD158e1.

FIGURE 4.

Disparity between KIR repertoire and self ligand repertoire. Percentages of persons in a group of 50 donors who either expressed an inhibitory KIR but not the corresponding ligand or expressed a KIR ligand but not the corresponding KIR.

FIGURE 4.

Disparity between KIR repertoire and self ligand repertoire. Percentages of persons in a group of 50 donors who either expressed an inhibitory KIR but not the corresponding ligand or expressed a KIR ligand but not the corresponding KIR.

Close modal

To further investigate whether the donor or recipient HLA affects the level or frequency of expression of each KIR on NK cells, we assessed KIR expression on the donor’s NK cells before and after transplantation. Among the 50 potential donors, we observed no significant relationship between KIR expression and self HLA allelic groups. The presence or the absence of a group of ligands did not predictably alter the mean fluorescent intensity (MFI) or the frequency of expression of its corresponding KIR on the NK cells. The expression of one KIR also did not affect the MFI or frequency of expression of another KIR on the NK cells (Fig. 5). Subsets of NK cells in some of our donors did not express any inhibitory KIRs for any self classical-HLA molecules.

FIGURE 5.

KIR expression was not affected by ligand repertoire or the expression of another KIR on the same cell. Illustrative patterns of KIR expression in three donors with three different ligand repertoires are shown (CD158a is specific for HLA-CLys80, CD158b1/b2 for HLA-CAsn80, and CD158e1 for HLA-BBw4). The presence of a ligand did not predict either the fluorescent intensity or the frequency of expression of its corresponding receptor on NK cells. The expression of one KIR also did not affect the MFI of another KIR on NK cells.

FIGURE 5.

KIR expression was not affected by ligand repertoire or the expression of another KIR on the same cell. Illustrative patterns of KIR expression in three donors with three different ligand repertoires are shown (CD158a is specific for HLA-CLys80, CD158b1/b2 for HLA-CAsn80, and CD158e1 for HLA-BBw4). The presence of a ligand did not predict either the fluorescent intensity or the frequency of expression of its corresponding receptor on NK cells. The expression of one KIR also did not affect the MFI of another KIR on NK cells.

Close modal

KIR expression was monitored in the developing NK cells of our patients after they received a highly purified CD34+ cell graft from an HLA-haploidentical donor. This was a unique setting for the biological study of KIR development in an HLA-disparate recipient, because the stem cell graft contained <0.01% contamination of mature CD56+ or CD3+ cells, and the patients did not receive any immunosuppressants for GvHD prophylaxis and did not develop GvHD. Within 3 mo of transplantation, the KIR expression pattern of all 22 study patients resumed a donor-specific pattern recognized in the donor NK cells before transplantation (Fig. 6). Thereafter, KIR expression was stable in all samples examined for up to 12 mo and did not appear to be affected by the recipient’s HLA groups (data not shown).

FIGURE 6.

Donor-specific pattern of KIR expression resumed within 3 mo after transplantation. Illustrative patterns from two donor/recipient pairs are shown; NK cells from donor 1 constitutively expressed CD158e1; those from donor 2 did not. One month after transplantation, NK cells derived from the graft of donor 1 did not express CD158e1, whereas those from donor 2 did. The patterns of KIR expression reverted back to a pretransplant pattern within 3 mo and remained stable thereafter. RT-PCR confirmed that cells from donor 2 did not constitutively express CD158e1 transcripts; however, the CD158e1 gene was detectable by PCR using two different sets of primers for exon 3 (shown) (20 ) and exon 4 (not shown), confirmed by sequencing the PCR products. Both donors and recipients are HLA-Bw4 positive.

FIGURE 6.

Donor-specific pattern of KIR expression resumed within 3 mo after transplantation. Illustrative patterns from two donor/recipient pairs are shown; NK cells from donor 1 constitutively expressed CD158e1; those from donor 2 did not. One month after transplantation, NK cells derived from the graft of donor 1 did not express CD158e1, whereas those from donor 2 did. The patterns of KIR expression reverted back to a pretransplant pattern within 3 mo and remained stable thereafter. RT-PCR confirmed that cells from donor 2 did not constitutively express CD158e1 transcripts; however, the CD158e1 gene was detectable by PCR using two different sets of primers for exon 3 (shown) (20 ) and exon 4 (not shown), confirmed by sequencing the PCR products. Both donors and recipients are HLA-Bw4 positive.

Close modal

To further characterize the receptor-ligand model, we investigated whether a larger number of receptor-ligand mismatch pairs would result in greater antileukemia effects in vitro and in vivo in a mouse model. NK cells from donors who expressed all the inhibitory KIRs (CD158a, CD158b1/b2, and CD158e1) were used. In the in vitro NK cell cytotoxicity assay (Fig. 7 a), the sensitivity of MLL-1 leukemia cells expressing the HLA-CAsn80 ligand only (with two receptor-ligand mismatch pairs) was similar to that of LCL cells transfected with Cw03 (HLA-CAsn80) and was lower than that of K562 cells that expressed a very low level of HLA class I molecules (equivalent to three receptor-ligand mismatch pairs in this assay). In MLL-2 cells the expression of HLA-CLys80 in addition to HLA-CAsn80 resulted in only one receptor-ligand mismatch pair and provided additional protection against NK cell cytotoxicity, lysis that was less than that of LCL cells transfected with either Cw03 (HLA-CAsn80) or Cw04 (HLA-CLys80) alone. The MLL-3 cells that expressed all three KIR ligands (HLA-CAsn80, HLA-CLys80, and HLA-BBw4) with no receptor-ligand mismatch pairs were most resistant to NK cell-mediated lysis (MLL-1 vs MLL-2, p = 0.004; MLL-2 vs MLL-3, p = 0.05). The susceptibility of these cell lines to NK cell cytotoxicity was independent of their expression of adhesion molecules, including members of the β1 integrin family (CD49d), β2 integrin family (CD11a,b,c), L-selectin family (CD62L), and Ig superfamily (CD54, CD58; data not shown).

FIGURE 7.

Susceptibility to NK cell-mediated lysis increases with the number of receptor-ligand mismatches. NK cells from donors who expressed all the inhibitory KIRs were used. a, Percentages of NK cell-mediated lysis in vitro (E:T cell ratio, 40:1). b, Leukemia burdens in the spleens of the mice injected with leukemia cells and NK cells (in a low 1:1 cell ratio) are shown as percentages of those of mice injected with leukemia cells alone. c, Survival of mice after injection with leukemia cells (dashed lines) or in combination with NK cells (solid lines). NK cells from the same donor were used for each set of experiments. Data for all panels were combined from independent experiments using NK cells from different donors (n = 5 for 5a, n = 3 for 5b, n = 3 for 5c). The p values of comparisons among the MLL-1, MLL-2, and MLL-3 groups in a and b were obtained by trend test.

FIGURE 7.

Susceptibility to NK cell-mediated lysis increases with the number of receptor-ligand mismatches. NK cells from donors who expressed all the inhibitory KIRs were used. a, Percentages of NK cell-mediated lysis in vitro (E:T cell ratio, 40:1). b, Leukemia burdens in the spleens of the mice injected with leukemia cells and NK cells (in a low 1:1 cell ratio) are shown as percentages of those of mice injected with leukemia cells alone. c, Survival of mice after injection with leukemia cells (dashed lines) or in combination with NK cells (solid lines). NK cells from the same donor were used for each set of experiments. Data for all panels were combined from independent experiments using NK cells from different donors (n = 5 for 5a, n = 3 for 5b, n = 3 for 5c). The p values of comparisons among the MLL-1, MLL-2, and MLL-3 groups in a and b were obtained by trend test.

Close modal

Using an in vivo mouse system (Fig. 7,b), leukemia burden in bone marrow and spleen of mice transplanted with NK cells and MLL cells (in a 1:1 cell ratio) was lowest in mice injected with MLL-1 cells, higher in those injected with MLL-2 cells, and highest in those injected with MLL-3 cells. Subsequent experiments demonstrated prolongation of survival (median, 7 days) in the mice injected with NK cells and MLL-1 cells (1:1 cell ratio), but not in those injected with NK cells and MLL-3 cells (Fig. 7 c).

Leukemias such as MLL-3 cells that express all groups of ligands will probably not be cured with unmanipulated NK cell therapy, because they are inherently resistance to NK cell cytotoxicity regardless of the donor KIR repertoire. We investigated whether this inherent resistance could be overcome by prestimulating the NK cells with IL-12 and IL18 and whether these cytokines changed the expression of KIR. These two NK-stimulatory cytokines were selected for study, because they might paradoxically reduce GvHD and therefore be useful in the clinic (28, 29). We found that these two cytokines or high dose IL-12 alone overcame the inherent resistance of MLL-3 cells to NK cell-mediated lysis (Fig. 8), but did not change the expression of inhibitory or activating KIRs, CD94, and CD161 (data not shown).

FIGURE 8.

IL-12 and IL-18 reduce the resistance of leukemia cells with no receptor-ligand mismatch to NK cell-mediated lysis. NK cells from donors who expressed all the inhibitory KIRs were prestimulated for 24 h with IL-2 (1000 IU/ml), IL-12 (10 ng/ml), IL-18 (1 μg/ml), IL-12 plus IL-18 (10 ng/ml and 1 μg/ml, respectively), IL-12(100) (100 ng/ml), or IL-12(1000) (1000 ng/ml). a, Cytokine effects on cytotoxicity toward NK cell-sensitive K562 cells. b and c, Cytokine effects on cytotoxicity toward NK cell-resistant MLL-3 cells with no receptor-ligand mismatch. For each panel, the mean percentages of cell lysis were shown for each E:T cell ratio of three independent experiments using NK cells from different donors.

FIGURE 8.

IL-12 and IL-18 reduce the resistance of leukemia cells with no receptor-ligand mismatch to NK cell-mediated lysis. NK cells from donors who expressed all the inhibitory KIRs were prestimulated for 24 h with IL-2 (1000 IU/ml), IL-12 (10 ng/ml), IL-18 (1 μg/ml), IL-12 plus IL-18 (10 ng/ml and 1 μg/ml, respectively), IL-12(100) (100 ng/ml), or IL-12(1000) (1000 ng/ml). a, Cytokine effects on cytotoxicity toward NK cell-sensitive K562 cells. b and c, Cytokine effects on cytotoxicity toward NK cell-resistant MLL-3 cells with no receptor-ligand mismatch. For each panel, the mean percentages of cell lysis were shown for each E:T cell ratio of three independent experiments using NK cells from different donors.

Close modal

The missing self hypothesis of the antilymphoma effects of NK cells was proposed by Karre et al. (30) over a decade ago. Subsequent murine and clinical studies of adult leukemias conducted by Ruggeri and Velardi et al. (6, 7) led to the novel concept of perfect mismatch to deliberately search for a donor with an HLA mismatch that will induce favorable NK cell-mediated antileukemic effects (2). In this study we showed that 1) the accuracy of the prediction of relapse could be improved by directly evaluating the KIR repertoire of the donor; 2) the pattern of KIR expression was donor-specific, even after stem cell transplantation into an HLA-disparate recipient; 3) the potency of the antileukemia effects increased with an increasing number of receptor-ligand mismatch pairs; and 4) the resistance of leukemia cells to lysis in the setting of no receptor-ligand mismatch could be overcome by prestimulating the NK cells with IL-12 and IL-18.

In agreement with the findings of Velardi et al. (6, 7), we found that the KIR ligand-ligand prediction model was quite accurate for the patients in its low risk category, an accuracy comparable to that of the receptor-ligand model (Fig. 2). However, the ligand-ligand model misclassified 13 of the 15 patients in the low risk category of the receptor-ligand model as being at high risk. The high accuracy of the ligand-ligand model for (low risk) patients with ligand-ligand incompatible donor may be explained by our findings that the presence of a KIR ligand in a donor usually correlated with the expression of its receptor (Fig. 4). The improved precision of the receptor-ligand model for our patients with a ligand-ligand-compatible donor may be explained by the observation that CD158a, CD158b1/b2, and CD158e1 were found in the KIR repertoire of many donors who had no corresponding ligand in the donor (Fig. 4). Moreover, we observed that the NK cells derived from highly purified CD34+ cells (<0.01% contamination of CD56+ or CD3+ cells) always acquired a donor-specific pattern of KIR expression independent of self HLA within the first 3 mo of transplantation. Thus, during this period of receptor acquisition, subsets of NK cells may express only one of the KIRs and could, therefore, be alloreactive if the corresponding ligand was absent in the recipient’s cells. This response appears to occur only during the first few months after transplantation; alloreactive clones could not be generated beyond 3 mo after transplantation even when their generation was possible earlier (6). This latter finding may explain why autoimmunity is not found in our adult donors with a specific KIR but no corresponding ligand. At present, the molecular mechanism is still poorly understood for the apparent presence of a potent antileukemia effect immediately after transplantation but the lack of a long term, NK cell-mediated autoimmunity in our patients with receptor-ligand mismatch pairs, although a dominant activity of ligand-negative cells in inducing tolerance has been demonstrated by neighboring hemopoietic cells and nonhemopoietic stromal cells during NK cell development in mice (31, 32, 33). Nonetheless, our finding that the KIR expression on NK cells derived from donor’s CD34+ cells always adopts a donor-specific pattern before the end of the 3-mo window suggests that a perfect mismatch donor could be selected on the basis of a single evaluation of the donor’s KIR repertoire before transplantation.

CD158k was not considered a priori in both the ligand-ligand model and the receptor-ligand model, because its specificity is uncertain. First, NK cells expressing CD158k were inhibited by both HLA-A3 and A11 in one study (34), but not by HLA-A11 in another study (35). Second, if HLA-A3 is the only ligand for the ubiquitously expressed CD158k, one would predict an unrealistically low relapse rate (∼10%), because the overall frequency of HLA-A3 is only ∼10% in the general population (36). In fact, both the ligand-ligand model and the receptor-ligand model were found to be less accurate if the CD158k and its ligand (HLA-A3 with or without A11) were included in our prediction models (data not shown). Taken together, these findings suggest that CD158k may have additional ligands not yet identified. When the complete set of ligands is known, incorporation of CD158k-ligand matching may perhaps be useful for donor selection.

Prior investigations in adult patients have demonstrated the NK cell-mediated antileukemia effect toward AML, but not ALL (6, 7). In this pediatric study we confirmed the effect on myeloid leukemia, but also extended the observation to lymphoid malignancy (Fig. 3). Unlike adult ALL that frequently do not express adhesion molecules that are essential for NK target conjugation and activation (6), blasts from all cases of childhood ALL have been consistently found to express at a high rate adhesion molecules such as members of the β1 integrin family (CD29, CD49d), the β2 integrin family (LFA-1), and the Ig superfamily (ICAM-1, LFA-3) (37).

The KIR repertoire in a normal individual who did not receive transplantation had been shown to be primarily genetically regulated and not linked to self HLA (20, 38, 39). In 18 patients who had received routine immunosuppression for GvHD prophylaxis after transplantation from an HLA-matched sibling or unrelated donor, Shilling et al. (39, 40) reported idiosyncratic patterns of KIR reconstitution associated with clinical acute or chronic GvHD. Our transplantation of highly purified CD34+ cells from an HLA-haploidentical donor allowed for the first time a biological study of KIR expression in NK cells developing in HLA-disparate hosts, without interference by prophylaxis or treatment for GvHD. Similar to (donor) self HLA, we found that the ligand repertoire of the HLA-disparate recipient did not appear to correlate with the expression of KIRs in the NK cells derived from the donor’s stem cells, a finding that is not compatible with the Ly49 receptor calibration model in mice (41). The expression of one KIR also did not appear to affect the frequency or level of expression of another KIR, a finding incompatible with the Ly49 selection or sequential models in mice (31, 42, 43). However, for all 50 donors in this study we always observed subsets of NK cells expressing at least one inhibitory KIR that recognized at least one of the self HLA groups, in agreement with the “at least one” hypothesis (5, 31). As the presence of a KIR ligand almost always correlates with the presence of its corresponding receptor in the NK cell receptor repertoire (Fig. 4), all our clinical donors had at least one inhibitory KIR in their receptor repertoire that recognized one of the HLA groups expressed in the recipient (the haplotype that is shared by the donor and recipient). Although the “at least one” hypothesis applied to our data at the repertoire level, four-color flow cytometric analyses demonstrated that subsets of NK cells in some of our donors did not express any inhibitory KIRs for any self or recipient classical HLA molecules at the individual cell level. A recent study demonstrated that Ly49 receptors in subsets of murine NK cells can inhibit killing both by those cells and, via a cell-cell interaction, by other subsets of NK cells that do not express that inhibitory receptor (44). This finding suggests that “at least one” match at the repertoire level may translate functionally to the cellular level in vivo. In the clinical perspective, one has to be cautious when selecting a donor with complete receptor-ligand mismatch (not “at least one”) in an attempt to maximize antileukemia activity, because the safety of such a transplant is not yet known, although autoimmunity was not observed in mice with complete absence of MHC class I ligands (45, 46, 47).

For the patients who have more than one potential donor, we asked whether a larger number of receptor-ligand mismatch pairs increases the antileukemia effects of the donor’s NK cells. We selected three cell lines with MLL rearrangement common in infants that have one, two, or three groups of KIR-ligands and tested their susceptibility to NK cells with all three corresponding KIRs in their repertoire. The results from our in vitro and in vivo assays suggest that the larger the number of receptor-ligand mismatch pairs, the higher the susceptibility of MLL cells to NK cell-mediated lysis. This finding leads to the speculation that a donor with a larger number of mismatch pairs may be a better donor, provided that the NK cells of the donor have at least one KIR in their repertoire that recognizes one of the recipient’s HLA, as discussed above.

Can patients with all three groups of KIR ligands still benefit from NK cell therapy? We found that the MLL-3 cell line, which had all three ligands, was very resistant to NK cell-mediated lysis, and this type of leukemia will probably not be cured with unmanipulated NK cell therapy. However, we found that prestimulation of NK cells with IL-12 and IL-18 overcome the leukemia cells’ resistance to lysis. Besides the known effects of IL-12 and IL-18 on cytokine production by NK cells (48), our results demonstrated for the first time that this cytokine-augmented natural cytotoxicity was KIR independent, thus suggesting that patients with all three groups of KIR ligands may still benefit from NK cell therapy. IL-12 and IL-18 are convenient and attractive candidates for ex vivo stimulation of NK cells, because they have been shown to paradoxically reduce T cell-mediated GvH disease in mice (28, 29), although their efficacy and toxicity in human are unknown (49). High dose IL-12 may also be used for the augmentation of NK cell cytotoxicity toward leukemia cells with no receptor-ligand mismatch.

In summary, this work provides additional fundamental insights into the antileukemia response of allogeneic NK cells. These results may be useful for the development of strategies for NK cell therapy in the 70% of patients who do not have an HLA-identical sibling donor.

We thank Dr. Janice Riberdy for critically reviewing the manuscript, Dr. Angela J. McArthur for scientific editing, and Annie Babarin-Dorner, Martha Holladay, James Houston, and Thasia Leimig for technical assistance.

1

This work was supported in part by National Institutes of Health Grant P30CA21765-24 (to W.L), the Assisi Foundation of Memphis, and the American Lebanese Syrian Associated Charities.

3

Abbreviations used in this paper: GvH, graft-versus-host; KIR, killer cell Ig-like receptor; MFI; mean fluorescent intensity.

1
Leung, W., C. I. Civin.
2000
. Use of purified stem and progenitor cells for transplantation. K. Atkinson, ed.
Clinical Bone Marrow and Blood Stem Cell Transplantation
2nd Ed.
1191
. Cambridge University Press, Cambridge.
2
Karre, K..
2002
. A perfect mismatch.
Science
295
:
2029
.
3
Natarajan, K., N. Dimasi, J. Wang, R. A. Mariuzza, D. H. Margulies.
2002
. Structure and function of natural killer cell receptors: multiple molecular solutions to self, nonself discrimination.
Annu. Rev. Immunol.
20
:
853
.
4
Vilches, C., P. Parham.
2002
. KIR: diverse, rapidly evolving receptors of innate and adaptive immunity.
Annu. Rev. Immunol.
20
:
217
.
5
Valiante, N. M., M. Uhrberg, H. G. Shilling, K. Lienert-Weidenbach, K. L. Arnett, A. D’Andrea, J. H. Phillips, L. L. Lanier, P. Parham.
1997
. Functionally and structurally distinct NK cell receptor repertoires in the peripheral blood of two human donors.
Immunity
7
:
739
.
6
Ruggeri, L., M. Capanni, M. Casucci, I. Volpi, A. Tosti, K. Perruccio, E. Urbani, R. S. Negrin, M. F. Martelli, A. Velardi.
1999
. Role of natural killer cell alloreactivity in HLA-mismatched hematopoietic stem cell transplantation.
Blood
94
:
333
.
7
Ruggeri, L., M. Capanni, E. Urbani, K. Perruccio, W. D. Shlomchik, A. Tosti, S. Posati, D. Rogaia, F. Frassoni, F. Aversa, et al
2002
. Effectiveness of donor natural killer cell alloreactivity in mismatched hematopoietic transplants.
Science
295
:
2097
.
8
Farag, S. S., T. A. Fehniger, L. Ruggeri, A. Velardi, M. A. Caligiuri.
2002
. Natural killer cell receptors: new biology and insights into the graft-versus-leukemia effect.
Blood
100
:
1935
.
9
Frohn, C., J. M. Brand, J. Luhm.
2003
. NK cells: a lesson from mismatched hematopoietic transplantation.
Trends Immunol.
24
:
8
.
10
Handgretinger, R., T. Klingebiel, P. Lang, M. Schumm, S. Neu, A. Geiselhart, P. Bader, P. G. Schlegel, J. Greil, D. Stachel, et al
2001
. Megadose transplantation of purified peripheral blood CD34+ progenitor cells from HLA-mismatched parental donors in children.
Bone Marrow Transplant.
27
:
777
.
11
Leung, W. H., V. Turner, S. L. Richardson, E. Benaim, G. Hale, E. M. Horwitz, P. Woodard, L. C. Bowman. Effect of HLA class I or class II incompatibility in pediatric marrow transplantation from unrelated and related donors.
Hum. Immunol.
62
:
399
.
12
Andre, P., R. Biassoni, M. Colonna, D. Cosman, L. L. Lanier, E. O. Long, M. Lopez-Botet, A. Moretta, L. Moretta, P. Parham, et al
2001
. New nomenclature for MHC receptors.
Nat. Immunol.
2
:
661
.
13
Handgretinger, R., A. Geiselhart, A. Moris, R. Grau, O. Teuffel, W. Bethge, L. Kanz, P. Fisch.
1999
. Pure red-cell aplasia associated with clonal expansion of granular lymphocytes expressing killer-cell inhibitory receptors.
N. Engl. J. Med.
340
:
278
.
14
Litwin, V., J. Gumperz, P. Parham, J. H. Phillips, L. L. Lanier.
1994
. NKB1: a natural killer cell receptor involved in the recognition of polymorphic HLA-B molecules.
J. Exp. Med.
180
:
537
.
15
Moretta, A., C. Bottino, D. Pende, G. Tripodi, G. Tambussi, O. Viale, A. Orengo, M. Barbaresi, A. Merli, E. Ciccone.
1990
. Identification of four subsets of human CD3CD16+ natural killer (NK) cells by the expression of clonally distributed functional surface molecules: correlation between subset assignment of NK clones and ability to mediate specific alloantigen recognition.
J. Exp. Med.
172
:
1589
.
16
Vitale, M., S. Sivori, D. Pende, R. Augugliaro, C. Di Donato, A. Amoroso, M. Malnati, C. Bottino, L. Moretta, A. Moretta.
1996
. Physical and functional independency of p70 and p58 natural killer (NK) cell receptors for HLA class I: their role in the definition of different groups of alloreactive NK cell clones.
Proc. Natl. Acad. Sci. USA
93
:
1453
.
17
Lanier, L. L., J. E. Gumperz, P. Parham, I. Melero, M. Lopez-Botet, J. H. Phillips.
1995
. The NKB1 and HP-3E4 NK cells receptors are structurally distinct glycoproteins and independently recognize polymorphic HLA-B and HLA-C molecules.
J. Immunol.
154
:
3320
.
18
Ferrini, S., A. Cambiaggi, R. Meazza, S. Sforzini, S. Marciano, M. C. Mingari, L. Moretta.
1994
. T cell clones expressing the natural killer cell-related p58 receptor molecule display heterogeneity in phenotypic properties and p58 function.
Eur. J. Immunol.
24
:
2294
.
19
Lanier, L. L., B. Corliss, J. H. Phillips.
1997
. Arousal and inhibition of human NK cells.
Immunol. Rev.
155
:
145
.
20
Gumperz, J. E., N. M. Valiante, P. Parham, L. L. Lanier, D. Tyan.
1996
. Heterogeneous phenotypes of expression of the NKB1 natural killer cell class I receptor among individuals of different human histocompatibility leukocyte antigens types appear genetically regulated, but not linked to major histocompatibililty complex haplotype.
J. Exp. Med.
183
:
1817
.
21
Soderstrom, K., B. Corliss, L. L. Lanier, J. H. Phillips. CD94/NKG2 is the predominant inhibitory receptor involved in recognition of HLA-G by decidual and peripheral blood NK cells.
J. Immunol.
159
:
1072
.
22
Uhrberg, M., N. M. Valiante, B. P. Shum, H. G. Shilling, K. Lienert-Weidenbach, B. Corliss, D. Tyan, L. L. Lanier, P. Parham.
1997
. Human diversity in killer cell inhibitory receptor genes.
Immunity
7
:
753
.
23
Gagne, K., G. Brizard, B. Gueglio, N. Milpied, P. Herry, F. Bonneville, M. L. Cheneau, N. Schleinitz, A. Cesbron, G. Follea, et al
2002
. Relevance of KIR gene polymorphisms in bone marrow transplantation outcome.
Hum. Immunol.
63
:
271
.
24
Uherek, C., T. Tonn, B. Uherek, S. Becker, B. Schnierle, H. G. Klingemann, W. Wels.
2002
. Retargeting of natural killer-cell cytolytic activity to ErbB2-expressing cancer cells results in efficient and selective tumor cell destruction.
Blood
100
:
1265
.
25
Faderl, S., H. M. Kantarjian, M. Talpaz, Z. Estrov.
1998
. Clinical significance of cytogenetic abnormalities in adult acute lymphoblastic leukemia.
Blood
91
:
3995
.
26
Leung, W., N. Pitts, K. Burnette, J. M. Cunningham, E. M. Horwitz, E. Benaim, G. Hale, P. Woodard, C. H. Pui, L. C. Bowman. Allogeneic bone marrow transplantation for infants with acute leukemia or myelodysplastic syndrome.
Bone Marrow Transplant.
27
:
717
.
27
Pui, C. H., P. S. Gaynon, J. M. Boyett, J. M. Chessells, A. Baruchel, W. Kamps, L. B. Silverman, A. Biondi, D. O. Harms, E. Vilmer, et al
2002
. Outcome of treatment in childhood acute lymphoblastic leukaemia with rearrangements of the 11q23 chromosomal region.
Lancet
359
:
1909
.
28
Yang, Y. G., J. J. Sergio, D. A. Pearson, G. L. Szot, A. Shimizu, M. Sykes.
1997
. Interleukin-12 preserves the graft-versus-leukemia effect of allogeneic CD8 T cells while inhibiting CD4-dependent graft-versus-host disease in mice.
Blood
90
:
4651
.
29
Okamoto, I., K. Kohno, T. Tanimoto, K. Iwaki, T. Ishihara, S. Akamatsu, H. Ikegami, M. Kurimoto.
2000
. IL-18 prevents the development of chronic graft-versus-host disease in mice.
J. Immunol.
164
:
6067
.
30
Karre, K., H. G. Ljunggren, G. Piontek, R. Kiessling.
1986
. Selective rejection of H-2-deficient lymphoma variants suggests alternative immune defence strategy.
Nature
319
:
675
.
31
Raulet, D. H., R. E. Vance, C. W. McMahon.
2001
. Regulation of the natural killer cell receptor repertoire.
Annu. Rev. Immunol.
19
:
291
.
32
Wu, M. F., D. H. Raulet.
1997
. Class I-deficient hemopoietic cells and nonhemopoietic cells dominantly induce unresponsiveness of natural killer cells to class I-deficient bone marrow cell grafts.
J. Immunol.
158
:
1628
.
33
Johansson, M. H., C. Bieberich, G. Jay, K. Karre, P. Hoglund.
1997
. Natural killer cell tolerance in mice with mosaic expression of major histocompatibility complex class I transgene.
J. Exp. Med.
186
:
353
.
34
Pende, D., R. Biassoni, C. Cantoni, S. Verdiani, M. Falco, C. di Donato, L. Accame, C. Bottino, A. Moretta, L. Moretta.
1996
. The natural killer cell receptor specific for HLA-A allotypes: a novel member of the p58/p70 family of inhibitory receptors that is characterized by three immunoglobulin-like domains and is expressed as a 140-kD disulphide-linked dimer.
J. Exp. Med.
184
:
505
.
35
Dohring, C., D. Scheidegger, J. Samaridis, M. Cella, M. Colonna.
1996
. A human killer inhibitory receptor specific for HLA-A.
J. Immunol.
156
:
3098
.
36
Mori, M., P. G. Beatty, M. Graves, K. M. Boucher, E. L. Milford.
1997
. HLA gene and haplotype frequencies in the North American population: the National Marrow Donor Program Donor Registry.
Transplantation
64
:
1017
.
37
Mengarelli, A., D. Zarcone, R. Caruso, C. Tenca, I. Rana, R. M. Pinto, C. E. Grossi, G. De Rossi.
2001
. Adhesion molecule expression, clinical features and therapy outcome in childhood acute lymphoblastic leukemia.
Leuk. Lymphoma
40
:
625
.
38
Gardiner, C. M., L. A. Guethlein, H. G. Shilling.
2001
. Different NK cell surface phenotypes defined by the DX9 antibody are due to KIR3DL1 gene polymorphism.
J. Immunol.
166
:
2992
.
39
Shilling, H. G., N. Young, L. A. Guethlein, N. W. Cheng, C. M. Gardiner, D. Tyan, P. Parham.
2002
. Genetic control of human NK cell repertoire.
J. Immunol.
169
:
239
.
40
Shilling, H. G., K. L. McQueen, N. W. Cheng, J. A. Shizuru, R. S. Negrin, P. Parham.
2003
. Reconstitution of NK cell receptor repertoire following HLA-matched hematopoietic cell transplantation.
Blood
101
:
3730
.
41
Olsson, M. Y., K. Karre, C. L. Sentman.
1995
. Altered phenotype and function of natural killer cells expressing the major histocompatibility complex receptor Ly-49 in mice transgenic for its ligand.
Proc. Natl. Acad. Sci. USA
92
:
1649
.
42
Fahlen, L., U. Lendahl, C. L. Sentman.
2001
. MHC class I-Ly49 interactions shape the Ly49 repertoire on murine NK cells.
J. Immunol.
166
:
6585
.
43
Hanke, T., H. Takizawa, D. H. Raulet.
2001
. MHC-dependent shaping of the inhibitory Ly49 receptor repertoire on NK cells: evidence for a regulated sequential model.
Eur. J. Immunol.
31
:
3370
.
44
Raziuddin, A., D. L. Longo, M. Bennett, R. Winkler-Pickett, J. R. Ortaldo, W. J. Murphy.
2002
. Increased bone marrow allograft rejection by depletion of NK cells expressing inhibitory Ly49 NK receptors for donor class I antigens.
Blood
100
:
3026
.
45
Bix, M., N. S. Liao, M. Zijlstra, J. Loring, R. Jaenisch, D. Raulet.
1991
. Rejection of class I MHC-deficient haemopoietic cells by irradiated MHC-matched mice.
Nature
349
:
329
.
46
Ljunggren, H. G., L. Van Kaer, H. L. Ploegh, S. Tonegawa.
1994
. Altered natural killer cell repertoire in Tap-1 mutant mice.
Proc. Natl. Acad. Sci. USA
91
:
6520
.
47
Dorfman, J. R., J. Zerrahn, M. C. Coles, D. H. Raulet.
1997
. The basis for self-tolerance of natural killer cells in β2-microglobulin and TAP-1 mice.
J. Immunol.
159
:
5219
.
48
Lauwerys, B. R., N. Garot, J. C. Renauld, F. A. Houssiau.
2000
. Cytokine production and killer activity of NK/T-NK cells derived with IL-2, IL-15, or the combination of IL-12 and IL-18.
J. Immunol.
165
:
1847
.
49
Carson, W. E., J. E. Dierksheide, S. Jabbour, M. Anghelina, P. Bouchard, G. Ku, H. Yu, H. Baumann, M. H. Shah, M. A. Cooper, et al
2000
. Coadministration of interleukin-18 and interleukin-12 induces a fatal inflammatory response in mice: critical role of natural killer cell interferon-γ production and STAT-mediated signal transduction.
Blood
96
:
1465
.