Evaluation of KIR ligand incompatibility in mismatched unrelated donor hematopoietic transplants

One of the important functions of HLA class I alleles is interaction with natural killer (NK) cells. Receptors termed killer immunoglobulinlike receptors (KIRs) on NK cells recognize groups of HLA class I alleles.1,2 Failure to recognize the appropriate KIR ligand on a mismatched cell triggers NK cell cytotoxicity. The consequences of NK cell alloreactivity, based on KIR ligand matching as determined by HLA class I typing, have recently been explored in haploidentical transplantation.3 That study analyzed NK function, using donor NK clones and recipient cells as targets, and correlated this activity with HLA-typing. A highly predictable algorithm for NK cell function was devised based on these HLA-typing results, which was then used for analysis of clinical end points. The previously published data demonstrated that donor-versus-recipient NK cell alloreactivity could eliminate leukemia relapse and graft rejection and protect patients against graft-versus-host disease (GVHD). In this study we have applied the same paradigm to recipients of unrelated donor transplants with mismatch at class I loci to start to address the generalizability of this algorithm in recipients of unrelated donor bone marrow transplants (BMTs) in the United States, transplanted at a single institution.

The analysis included 175 recipients of unrelated donor BMTs at the University of Minnesota between 1989 and 1999, who had allele level molecular typing at HLA-A, -B, -C, and -DRB1, and had at least one allele mismatch at a class I locus. Cases were divided into those with and those without KIR ligand incompatibility, as described by Ruggeri et al,3 based on known KIR ligands (HLA-C alleles with Asn77-Lys80; HLA-C alleles with Ser77-Asn80; HLA-Bw4 alleles). Characteristics of the transplant recipients and procedure are shown in Table 1.

No significant differences were identified between those with and without KIR ligand incompatibility in recipient age, diagnosis, stage of disease, preparative regimen, or GVHD prophylaxis (Table 1). There was no difference between the 2 groups in the year of the transplantation procedure (data not shown).

Acute GVHD was diagnosed on clinical criteria with histopathologic confirmation when possible. Overall staging used published criteria4 and was assigned by an independent review team retrospectively.

Data regarding transplantation patient characteristics, posttransplantation complications, and outcomes were prospectively collected by the Biostatistical Support Group using standardized collection procedures. Cumulative incidence rates and 95% confidence intervals (CIs) were estimated for engraftment, grades II to IV and grades III to IV acute GVHD, and relapse (for patients with malignant diseases).5 Kaplan-Meier methods were used to estimate survival.6 Log-rank test statistics were used to evaluate the univariate effects of KIR ligand incompatibility on outcome. To evaluate the independent effect of KIR ligand incompatibility, Cox regression was performed.7Factors included in the models were recipient and donor age, diagnosis, GVHD prophylaxis, conditioning regimen, and HLA mismatch.

KIR ligand status, according to the algorithm generated by Ruggeri et al,3 was determined for each donor-recipient pair. Incompatibility was assigned if a donor KIR ligand class I allele was not present in the recipient. Thus, donor NK clones are not inhibited by a ligand expressed on recipient cells. The algorithm provided a high degree of correlation between NK functional activity and HLA typing in haploidentical donor-recipients pairs.3 The presence of donor alloreactive NK cells should eliminate host antigen-presenting cells and host tumor cells, thereby diminishing rejection, GVHD, and residual tumor burdens. However, in our population of unrelated donor patients, no significant differences were seen in survival or acute GVHD grades III to IV (Table 2).

There was a trend toward an increased frequency of grades II to IV GVHD in those with KIR ligand incompatibility (61% versus 50%,P = .09). These data differ from the report of Ruggeri et al3 in which the presence of KIR ligand incompatibility was associated with protection from graft failure, GVHD, and relapse.3 Much of the benefit of KIR ligand incompatibility reported by Ruggeri et al3 was seen in those with acute myelogenous leukemia. We reanalyzed the data, including only those with myeloid disease (acute myelogeneous leukemia [AML], n = 14; chronic myelogeneous leukemia [CML] = 58). No substantial differences were found between those with and without KIR ligand incompatibility in recipient age, diagnosis, stage of disease, preparative regimen, or GVHD prophylaxis (data not shown). Analysis of outcomes showed no differences between those with and without KIR ligand incompatibility in the end points of graft failure, GVHD, or relapse (Table 2). Survival was better in those without KIR ligand incompatibility (38% [95% CI, 24-52] versus 13% [95% CI, 0-26] at 5 y, P < .01).

This difference in survival was surprising, as an advantage for KIR ligand incompatibility was expected.3 The present data might reflect an increased number of HLA mismatches in the KIR ligand mismatched group. Analysis of the total number of class I mismatches shows a significantly higher number of class I allele level mismatches in myeloid cases with KIR ligand incompatibility (46% versus 12% with > 2 allele level mismatches, P < .01). However, this does not explain the survival difference seen, as inclusion of the number of HLA mismatches in a Cox regression model did not substantially change the effect of KIR ligand incompatibility on mortality. The relative risks of KIR ligand incompatibility on mortality with and without HLA mismatch included were 2.2 (95% CI, 1.3-4.0) and 2.7 (95% CI, 1.3-5.4), respectively.

The full haplotype-mismatched transplantations reported by Ruggeri et al3 were performed by using a high CD34⁺ cell dose for engraftment, combined with extensive T-cell depletion to prevent GVHD, such that no postgrafting immunosuppression was needed. Such transplantations are typically associated with rapid recovery of NK cells and slow T-cell recovery.8-10 In contrast, two thirds of the unrelated donor transplantations we report were not T-cell depleted. In addition, the transplantations that were T-cell depleted were treated with elutriation that typically yields a 2-log T-cell depletion.11 This finding is much less than the 4-log depletion used in haploidentical transplantation. The different schedules of immune reconstitution might have masked any potential role for NK cell alloreactivity in influencing outcome. Because the donor grafts in our study contain a higher number of T cells than those of Ruggeri et al,3 the NK cell effects may have been obscured by donor T-cell effects and/or the immune suppression necessary to prevent/control GVHD caused by higher numbers of donor T cells. The report by Ruggeri et al3 focused on acute leukemia, and the important influence of KIR ligand incompatibility on relapse rates was seen in those patients with AML. In this study we examined CML and AML, but if the role of alloreactivity is clinically evident only in AML, we may have had too few AML cases to detect the NK effect.

The current study suggests that the beneficial effects of KIR ligand incompatibility in the GVHD direction observed in haploidentical transplantations are not generalizable to unrelated donor BM transplantations. The study of haploidentical transplantations provided direct evidence that KIR ligand incompatibility in the GVH direction is, with no exception, predictive of donor-versus-recipient NK cell alloreactivity. In the present study no functional assessment of the donor's NK cell repertoire was performed. It is therefore possible that stimulatory receptors, KIRs yet to be identified, or NK cell receptors from other families (eg, NKG2a/CD94 heterodimers) may play a role in the outcome of unrelated donor transplantations or confound known KIR/KIR ligand interactions.12 13 However, analyses of 32 unrelated pairs of healthy individuals indicate KIR ligand mismatches are predictive of NK cell mismatches even among genetically unrelated individuals (A.V. and L.R., unpublished data, April, 2002). An AML-specific effect might be observed in unrelated donor transplantations only if strategies from haploidentical transplantations are used: (1) transplantation of high doses of stem cells, (2) extensive T-cell depletion, (3) no postgrafting immunosuppression, and (4) donors selected for the “perfect mismatch” at HLA loci to drive favorable NK cell–mediated effects. Prospective studies are needed to address these possibilities.