|Home | About | Journals | Submit | Contact Us | Français|
Of the cancers treated with allogeneic hematopoietic stem-cell transplantation (HSCT), acute myeloid leukemia (AML) is most sensitive to natural killer (NK)–cell reactivity. The activating killer-cell immunoglobulin-like receptor (KIR) 2DS1 has ligand specificity for HLA-C2 antigens and activates NK cells in an HLA-dependent manner. Donor-derived NK reactivity controlled by KIR2DS1 and HLA could have beneficial effects in patients with AML who undergo allogeneic HSCT.
We assessed clinical data, HLA genotyping results, and donor cell lines or genomic DNA for 1277 patients with AML who had received hematopoietic stem-cell transplants from unrelated donors matched for HLA-A, B, C, DR, and DQ or with a single mismatch. We performed donor KIR genotyping and evaluated the clinical effect of donor KIR genotype and donor and recipient HLA genotypes.
Patients with AML who received allografts from donors who were positive for KIR2DS1 had a lower rate of relapse than those with allografts from donors who were negative for KIR2DS1 (26.5% vs. 32.5%; hazard ratio, 0.76; 95% confidence interval [CI], 0.61 to 0.96; P = 0.02). Of allografts from donors with KIR2DS1, those from donors who were homozygous or heterozygous for HLA-C1 antigens could mediate this antileukemic effect, whereas those from donors who were homozygous for HLA-C2 did not provide any advantage (24.9% with homozygosity or heterozygosity for HLA-C1 vs. 37.3% with homozygosity for HLA-C2; hazard ratio, 0.46; 95% CI, 0.28 to 0.75; P = 0.002). Recipients of KIR2DS1-positive allografts mismatched for a single HLA-C locus had a lower relapse rate than recipients of KIR2DS1-negative allografts with a mismatch at the same locus (17.1% vs. 35.6%; hazard ratio, 0.40; 95% CI, 0.20 to 0.78; P = 0.007). KIR3DS1, in positive genetic linkage disequilibrium with KIR2DS1, had no effect on leukemia relapse but was associated with decreased mortality (60.1%, vs. 66.9% without KIR3DS1; hazard ratio, 0.83; 95% CI, 0.71 to 0.96; P = 0.01).
Activating KIR genes from donors were associated with distinct outcomes of allogeneic HSCT for AML. Donor KIR2DS1 appeared to provide protection against relapse in an HLA-C–dependent manner, and donor KIR3DS1 was associated with reduced mortality. (Funded by the National Institutes of Health and others.)
Natural killer (NK) cells are lymphocytes that are critical for innate immunity against malignant or virally infected cells. Acute myeloid leukemia (AML) is the most common indication for hematopoietic stem-cell transplantation (HSCT) from unrelated donors, and the alloreactivity of donor NK cells can exert a potent antileukemic effect in this context.1,2 NK-cell function is controlled by an array of inhibitory and activating signals that are processed by cell-surface receptors, including the inhibitory and activating killer-cell immunoglobulin-like receptors (KIRs), whose genes vary in number and content from person to person. Inhibitory KIRs recognize the HLA class I ligand groups HLA-C1, C2, and Bw4 and mediate tolerance to self when they encounter self-HLA molecules on putative target cells.3
Earlier models of NK alloreactivity in HSCT focused on the interactions between inhibitory KIRs and HLA class I ligands, in which the alloreactivity of donor NK cells is triggered by a lack of self-HLA class I engagement of inhibitory KIRs.1,2,4,5 This relationship is most evident in HLA-haplotype–disparate HSCT, in which recipients who lack the HLA ligand present in the stem-cell donor have a decreased rate of relapse.1,2,5,6 An HLA-haplotype–disparate donor, however, is typically not preferred, because the high degree of HLA mismatching is associated with an increased risk of graft-versus-host disease (GVHD) and increased mortality.7 Therefore, better selection of donors to capture NK alloreactivity and prevent relapse without a high degree of HLA mismatching will probably improve the success of allogeneic HSCT.
Most persons have multiple activating KIRs, of which KIR2DS1 is the only one known to play a role in both NK-cell activation and tolerance, through its recognition of HLA-C2 (Asn 77 and Lys 80) molecules.8-10 KIR2DS1-positive NK cells isolated from HLA-C1–positive persons (with HLA-C1/C1 or C1/C2) secrete interferon-γ and are cytotoxic to target cells, in particular those that express HLA-C2. In HLA-C2 homozygous persons (with HLA-C2/C2), in whom high levels of the ligand are expressed as a self-molecule, NK cells that exclusively express KIR2DS1 are hyporesponsive.11-13 These findings are consistent with those of studies in transgenic mice, which have shown that the function of NK cells expressing activating receptors specific for self-ligands is diminished.14-16 These studies suggest that in allogeneic HSCT in humans, the functionally competent KIR2DS1-expressing NK cells from HLA-C1–positive donors could mediate leukemic cytotoxicity in the recipient, whereas the KIR2DS1-expressing NK cells from donors with HLA-C2/C2 would be poorly responsive in the setting of self HLA-C2 and would not mediate leukemic cytotoxicity. However, this hypothesis has not been directly tested either in vitro with the use of leukemia cells obtained from patients or in outcome studies of patients who undergo HSCT from unrelated donors.
We evaluated the effect of activating KIR2DS1 from donors on the outcome of HSCT. To minimize the contribution of the inhibitory KIR to NK alloreactivity,2,4,17-19 we evaluated HSCT pairs in which donors and recipients were matched for 9 or 10 of the possible 10 alleles at the HLA-A, B, C, DR, and DQ loci. We also sought to confirm our previous observation that activating KIR3DS1 is associated with increased survival.20
This retrospective study was designed to test the hypothesis that allografts from donors with KIR2DS1 and favorable HLA-C genotypes would improve the outcomes of HSCT from unrelated donors for patients with AML. Clinical data, HLA genotyping, and donor cell lines or genomic DNA for KIR genotyping were provided by the Center for International Blood and Marrow Transplant Research (CIBMTR). The CIBMTR is a research affiliate of the International Bone Marrow Transplant Registry, the Autologous Blood and Marrow Transplant Registry, and the National Marrow Donor Program (NMDP), which together account for more than 450 centers worldwide that contribute data on HSCT to a statistical center at the Medical College of Wisconsin in Milwaukee and the NMDP coordinating center in Minneapolis. Patients are followed longitudinally, with yearly follow-up. Studies conducted by the CIBMTR are performed in compliance with the Privacy Rule of the Health Insurance Portability and Accountability Act and with federal regulations pertaining to the protection of human research participants, as determined by the institutional review boards of the NMDP and the Medical College of Wisconsin. The principal investigators made the decision to submit the manuscript for publication. All authors vouch for the accuracy and completeness of the data and analyses; there was no writing assistance from anyone who is not listed as an author. There were no confidentiality agreements between funding agencies and the authors. No commercial support was involved in the study.
We evaluated 1277 patients with AML and 427 patients with acute lymphoblastic leukemia (ALL) who received stem-cell transplants between 1989 and 2008 from unrelated donors matched for 9 or 10 of the possible 10 alleles at HLA-A, B, C, DR, and DQ loci. All transplantations were facilitated by the NMDP, and the CIBMTR provided clinical data, HLA genotyping, and cell lines or genomic DNA from donors. HLA genotyping with allele-level resolution was verified through the NMDP. Consent from patients and donors was obtained through the CIBMTR. Both surviving and deceased patients were included in the analysis. The inclusion of deceased patients did not require consent, and approximately 4% of surviving patients did not provide consent. A random subset of surviving patients was excluded from the analysis to adjust for potential bias due to the failure to obtain consent from all surviving patients, as previously described.21
KIR genotyping was performed on genomic DNA from unrelated donors of hematopoietic stem cells with the use of polymerase-chain-reaction amplification with sequence-specific primers, as described previously,22 or with the use of a KIR genotyping kit (Invitrogen), according to the manufacturer's instructions.
Cox regression was used to examine the association of the hazard of failure for various time-to-event outcomes of HSCT (death from any cause, relapse, disease-free survival, and death without relapse) with donor KIR genotype. Failure for disease-free survival was defined as relapse or death, whichever occurred first. Logistic regression was used to assess the association of donor KIR genotype with the probability of acute GVHD. Two-sided P values were derived from adjusted regression models and were estimated by means of the Wald test. All models were adjusted for the patient's age, level of risk (low, intermediate, or high), cytomegalovirus (CMV) serostatus (positive or negative), and cytogenetic profile (no abnormalities, good risk, intermediate risk, or poor risk); use of T-cell depletion (yes or no); conditioning regimen (ablative with total-body irradiation, ablative without total-body irradiation, or nonablative); and HLA match (10 of 10 alleles vs. 9 of 10 alleles). No adjustments were made for multiple comparisons. Estimates of overall survival were obtained with the use of the Kaplan–Meier method, and cumulative incidence estimates were used to summarize the probability of relapse. Death without relapse was regarded as a competing risk for purposes of estimating the probability of relapse.
Characteristics of the 1277 patients with AML and their donors are listed in Table 1. The frequencies of activating KIR genes in the predominantly white population were consistent with gene frequencies in previously published studies.22,23 Specifically, the frequency of KIR2DS1 was 33%, of KIR3DS1 was 34%, and of KIR2DS2 was 49%. The KIR2DS1-positive and KIR2DS1-negative groups were similar with respect to risk level, age, GVHD prophylaxis, degree of HLA matching, sex of patient–donor pairs, and patient's race or ethnic group, graft type, and CMV serostatus (Table 1) and with respect to cytogenetic profile. A total of 427 patients with ALL were evaluated in parallel; the characteristics of these and their donors are detailed in Table S1 in the Supplementary Appendix, available with the full text of this article at NEJM.org.
We found a significantly reduced risk of relapse among patients with AML whose donors were positive for KIR2DS1, as compared with patients whose donors were negative for K IR2DS1 (26.5% vs. 32.5%; adjusted hazard ratio, 0.76; 95% confidence interval [CI], 0.61 to 0.96; P = 0.02) (Fig. 1A). We hypothesized that this association would be observed only for donors with HLA-C1, not for donors with HLA-C2/C2, because the presence of high levels of the HLA-C2 self-ligand in the latter donors would tolerize KIR2DS1-bearing NK cells and diminish their activity. Allografts from KIR2DS1-positive donors with HLA-C2/C2 were associated with an increased rate of AML relapse, as compared with all allografts from KIR2DS1-negative donors (37.3% vs. 32.5%; hazard ratio, 1.51; 95% CI, 0.94 to 2.41; P = 0.09) (Table 2). In contrast, allografts from KIR2DS1-positive donors with HLA-C1/C1 were associated with a decreased rate of relapse, as compared with all allografts from KIR2DS1-negative donors (24.7% vs. 32.5%; hazard ratio, 0.71; 95% CI, 0.51 to 0.99; P = 0.05). Patients with AML who received allografts from donors who were heterozygous for HLA-C1 (i.e., with HLA-C1/C2) also had a decreased rate of relapse, as compared with all allografts from KIR2DS1-negative donors (25.1% vs. 32.5%; hazard ratio, 0.67; 95% CI, 0.49 to 0.92; P = 0.01). Allografts from KIR2DS1-positive donors with HLA-C1/C1 or C1/C2 were associated with a reduced relapse rate, as compared with allografts from KIR2DS1-negative donors (24.9% vs. 32.5%; hazard ratio, 0.69; 95% CI, 0.54 to 0.88; P = 0.003) (Table 2) and especially as compared with allografts from KIR2DS1-positive donors with HLA-C2/C2 (24.9% vs. 37.3%; hazard ratio, 0.46; 95% CI, 0.28 to 0.75; P = 0.002) (Fig. 1B). Decreased rates of relapse with allografts from KIR2DS1-positive donors were seen in low-, intermediate-, and high-risk AML groups; a formal test of interaction did not provide definitive evidence that the KIR2DS1 effect differed across the three groups.
The reduced rate of AML relapse is consistent with the observation in vitro that KIR2DS1-mediated NK function is dependent on HLA-C background.11-13 We therefore examined the effect of donor KIR2DS1 across the donor HLA-C ligand groups. In contrast to the protective effect of KIR2DS1 against relapse in patients who received allografts from donors with HLA-C1/C1 or C1/C2 (relapse rate, 24.9%, vs. 31.2% among patients who received allografts from KIR2DS1-negative donors with HLA-C1/C1 or C1/C2; hazard ratio, 0.71; 95% CI, 0.55 to 0.92; P = 0.009), no protective benefit of KIR2DS1 against relapse was seen for patients who received allografts from donors with HLA-C2/C2. The high levels of activating HLA-C2 ligands for KIR2DS1 in donors with HLA-C2/C2 tolerize an otherwise potent antileukemic NK population, leading to similarly high relapse rates among recipients of allografts from KIR2DS1-positive and KIR2DS1-negative donors with HLA-C2/C2 (37.3% and 41.1%; hazard ratio, 1.17; 95% CI, 0.66 to 2.08; P = 0.58) (Fig. 1C).
Recipients of allografts from KIR2DS1-positive donors with HLA-C1/C1 or C1/C2 had lower mortality than recipients of allografts from KIR2DS1-negative donors (60.3% vs. 67.0%; hazard ratio, 0.85; 95% CI, 0.73 to 1.00; P = 0.04), whereas recipients of allografts from KIR2DS1-positive donors with HLA-C2/C2 and recipients of allografts from KIR2DS1-negative donors had similar mortality (57.7% and 67.0%, respectively; hazard ratio, 1.01; 95% CI, 0.69 to 1.46; P = 0.98) (Table 2). The findings with respect to disease-free survival (i.e., relapse or death) were similar (Table 2). There was no association between allografts from KIR2DS1-positive donors and the incidence of grade II to IV acute GVHD (51.0%, vs. 51.4% for allografts from KIR2DS1-negative donors; odds ratio, 0.97; 95% CI, 0.77 to 1.24; P = 0.83) or death without relapse (32.6% vs. 34.0%; hazard ratio, 0.99; 95% CI, 0.80 to 1.21; P = 0.90). Among patients with ALL, the KIR2DS1 status of donors was not associated with relapse or survival, a finding that is consistent with reports that NK reactivity is strongest against myeloid leukemias.1,24,25
Patients who are homozygous for HLA-C2 represent 13 to 19% of the HSCT population, and in studies of smaller cohorts, such patients have been reported to have high relapse rates.26-28 We hypothesized that a lack of KIR2DS1-induced NK alloreactivity from donors with HLA-C2/C2 is responsible for this observation. In our cohort, transplant recipients with HLA-C2/C2 indeed had an increased risk of relapse, as compared with recipients with HLA-C1/C1 or HLA-C1/C2 (38.4% vs. 29.3%; hazard ratio, 1.33; 95% CI, 1.01 to 1.75; P = 0.05) (Fig. 2A). In recipients with HLA-C2/C2, allografts from KIR2DS1-positive donors did not provide protection from relapse (relapse rate, 37.7%, vs. 38.7% with allografts from KIR2DS1-negative donors; hazard ratio, 0.98; 95% CI, 0.56 to 1.73; P = 0.95) (Fig. 2B). In contrast, patients with HLA-C1/C1 or C1/C2 benefited significantly from receiving an allograft from a KIR2DS1-positive donor (relapse rate, 24.8%, vs. 31.5% with allografts from KIR2DS1-negative donors; hazard ratio, 0.72; 95% CI, 0.56 to 0.93; P = 0.01) (Fig. 2B). The increased rates of relapse associated with allografts from donors with HLA-C2/C2, both KIR2DS1-positive and KIR2DS1-negative (Fig. 1C and and2B),2B), however, indicate that in addition to tolerizing KIR2DS1-bearing NK cells in the donor, HLA-C2/C2 homozygosity is associated with an increased rate of relapse owing to causes that are unrelated to KIR2DS1.
We sought to determine whether the association of donor KIR2DS1 and a reduced rate of relapse was more prominent in the HLA-matched or HLA-mismatched combinations, particularly because the latter may also promote inhibitory KIR-mediated NK alloreactivity.1,2 Although the relapse rate was not significantly lower among patients with allografts from KIR2DS1-positive donors, as compared with recipients of allografts from KIR2DS1-negative donors, when the allograft was matched for 10 of 10 alleles (29.3% vs. 33.5%; hazard ratio, 0.85; 95% CI, 0.63 to 1.16; P = 0.31) (Fig. 3A) or was mismatched at loci other than HLA-C (26.8% vs. 29.0%; hazard ratio, 0.84; 95% CI, 0.55 to 1.28; P = 0.42) (Fig. 3B), there was a significant effect of donor KIR2DS1 positivity for allografts with a single-allele mismatch at the HLA-C locus (relapse rate, 17.1%, vs. 35.6% for donor KIR2DS1 negativity; hazard ratio, 0.40; 95% CI, 0.20 to 0.78; P = 0.007) (Fig. 3C). Of patients with HLA-C2/C2, the 11 patients with a KIR2DS1-positive HLA-C–mismatched donor had a lower relapse rate than the 50 patients with a KIR2DS1-negative donor matched for HLA-C (hazard ratio, 0.38).
We therefore examined whether NK alloreactivity due to inhibitory KIRs contributed to these effects. Of the 216 HLA-C–mismatched pairs for which relapse data were available, 56 had the potential for donor-derived inhibitory KIR-mediated NK alloreactivity, with a fairly even distribution among the KIR2DS1-negative pairs (25%) and KIR2DS1-positive pairs (27%). Patients who had a relapse after receiving an HLA-C–mismatched transplant were evenly distributed between recipients of allografts from donors with the potential for inhibitory KIR-mediated NK alloreactivity (29%) and recipients of allografts from donors without this potential (30%). In aggregate, inhibitory KIR-mediated NK alloreactivity did not appear to contribute to KIR2DS1-related protection from relapse.
KIR2DS1 is in positive genetic linkage disequilibrium with other activating KIR genes22 — in particular, KIR3DS1 (r2 = 0.56 in this cohort). Although donor KIR3DS1 positivity was associated with a decreased rate of relapse (28.9%, vs. 31.3% with KIR3DS1 negativity; hazard ratio, 0.86; 95% CI, 0.69 to 1.07; P = 0.17), this protective effect was not sustained when donors were stratified according to KIR2DS1 status (Table 2). The association of donor KIR2DS1 positivity with a decreased rate of relapse, on the other hand, was similar regardless of donor status with regard to KIR3DS1 (Table 2, and Fig. S1A in the Supplementary Appendix). Others have reported a favorable, dose-dependent effect of the KIR2DS2-containing centromeric KIR B-partial haplotype (Cen-B) on the risk of relapse among patients with AML.24,25 We evaluated homozygosity for donor Cen-B positivity versus donor Cen-B negativity in this cohort and observed an association of Cen-B homozygosity with a reduced rate of relapse, although the association was not significant (28.1% vs. 33.7%; hazard ratio, 0.77; 95% CI, 0.52 to 1.13; P = 0.18). Adjustment for the presence of Cen-B did not change the association between donor KIR2DS1 and a decreased rate of AML relapse (hazard ratio, 0.76 and 0.77, before and after adjustment for Cen-B).
The previous observation that allografts from KIR3DS1-positive donors were associated with improved transplant-related and overall survival indicates that NK cell populations that express individual activating KIRs may play nonoverlapping roles in HSCT from unrelated donors.20 In the current study, the presence of donor KIR3DS1 was associated with decreased overall mortality (60.1%, vs. 66.9% in the absence of KIR3DS1; hazard ratio, 0.83; 95% CI, 0.71 to 0.96; P = 0.01) (Table 2, and Fig. S1B in the Supplementary Appendix), largely owing to reduced rates of death without relapse (Table S2 in the Supplementary Appendix). These reductions in mortality were influenced little by the presence of donor KIR2DS1 (Table 2, and Table S2 and Fig. S1C in the Supplementary Appendix).
In a large cohort, we found that specific activating KIR genes in donors were associated with distinct clinical outcomes of allogeneic HSCT for the treatment of AML. A KIR2DS1-dependent graft-versus-leukemia effect was modified by donor HLA-C2 in a manner consistent with in vitro findings of KIR2DS1-mediated NK function, in which the interaction of KIR2DS1 with high levels of HLA-C2 in persons with HLA-C2/C2 reduces NK reactivity.11-13 Thus, although KIR2DS1 was present in 33% of donors, the KIR2DS1-associated reduction in the rate of AML relapse was restricted to donors with HLA-C1/C1 or C1/C2, and the benefit was eliminated in transplants from donors with HLA-C2/C2. Capturing NK alloreactivity in HSCT from unrelated donors is therefore more complex than selecting donors with multiple activating KIR genes.24,25
Induction of tolerance in NK cells that express activating receptors through chronic exposure to the ligand has been described in mice14-16; we now show an association between clinical outcomes and NK tolerance in HSCT from unrelated donors. An interesting finding in this study is that KIR2DS1-positive donors heterozygous for HLA-C2 did not have evidence of diminished function, as assessed by control of leukemic relapse. This clinical observation is consistent with recent findings from our group (unpublished data) and others13 that in vitro–derived NK clones from KIR2DS1-positive donors with HLA-C1/C1 or C1/C2 genotypes have similarly high frequencies of cytotoxic activity against target cells, in contrast to NK clones from donors with HLA-C2/C2. To reduce the risk of AML relapse, the data support preferential selection of a KIR2DS1-positive donor if the donor has HLA-C1/C1 or C1/C2. Because of the strong positive linkage dysequilibrium between KIR2DS1 and KIR3DS1, the large majority of KIR2DS1-positive donors will also be positive for KIR3DS1, a favorable genetic marker for survival whose ligand and biologic function remain elusive.
Current practice dictates preferential selection of HLA-matched donors in order to minimize the risk of GVHD; however, the increased vulnerability of recipients with HLA-C2/C2 to relapse challenges the clinician to consider an HLA-C KIR ligand–mismatched donor in order to capture KIR2DS1-mediated NK alloreactivity and NK alloreactivity due to inhibitory KIR-mediated mechanisms, even in the face of a potentially increased risk of GVHD.29
An increased understanding of how KIR–HLA interactions dictate NK function could lead to more informed selection of stem-cell donors. We found that the potential benefit of KIR2DS1 is not simply predicted by the presence of the gene in the donor, but rather is modified by the presence of its ligand in the donor. We also found that donor KIR3DS1 may confer a survival advantage by reducing the risk of death without relapse. The effect of these and other activating KIR associations should be examined in a prospective manner in ethnically diverse cohorts, in which HLA and KIR genotype frequencies may vary.
The views expressed in this article do not reflect the official policy or position of the National Institutes of Health, the Department of the Navy, the Department of Defense, or any other agency of the U.S. government.
Supported in part by grants from the National Institutes of Health (U01 AI69197, KL2 RR024997, R01 HL088134, and P01 CA23766). The CIBMTR is supported by a Public Health Service grant/cooperative agreement from the National Cancer Institute (NCI), the National Heart, Lung, and Blood Institute (NHLBI), and the National Institute of Allergy and Infectious Diseases (U24-CA76518); a grant/cooperative agreement from the NHLBI and NCI (5U01HL069294); a contract with the Health Resources and Services Administration (HHSH234200637015C); grants from the Office of Naval Research (N00014-10-1-0204 and N00014-1-1-0339); and funding from Allos Therapeutics, Amgen, Angioblast Systems, Ariad, Be the Match Foundation, Blue Cross and Blue Shield Association, Buchanan Family Foundation, CaridianBCT, Celgene, CellGenix, Children's Leukemia Research Association, Fresenius Biotech North America, Gamida Cell–Teva Joint Venture, Genentech, Genzyme, GlaxoSmithKline, Kiadis Pharma, the Leukemia and Lymphoma Society, the Medical College of Wisconsin, Millennium Pharmaceuticals, Milliman USA, Miltenyi Biotec, National Marrow Donor Program, OptumHealth Care Solutions, Otsuka America Pharmaceutical, Seattle Genetics, Sigma-Tau Pharmaceuticals, Soligenix, Swedish Orphan Biovitrum, Therakos, WellPoint, and an anonymous donation to the Medical College of Wisconsin.
We thank Dr. Mary Horowitz for assistance in study design; Dr. Stephanie Lee for comments; and Ms. Clara Pinto, Ms. Reenat Hassan, Ms. Alice Yeh, and Ms. Zoe Harris for technical assistance.
Disclosure forms provided by the authors are available with the full text of this article at NEJM.org.