KIR-HLA Interactions Lack Clinical Utility in Matched Unrelated Donor Transplantation for AML: An Analysis of the CIBMTR and DRST Registries

Background: The interaction between donor killer immunoglobulin-like receptor (KIR) and recipient HLA has been postulated to enhance the graft-versus-leukemia effect in allogeneic hematopoietic cell transplantation (HCT) for acute myeloid leukemia (AML). Historically, analyses of individual interactions between single KIR and their respective HLA ligands have yielded conflicting findings, and the clinical importance of these interactions in the matched unrelated donor (MUD) setting remains controversial. Here, we applied a systematic approach, studying both a wide range of KIR and class I HLA interactions at the single-receptor level as well as the most prevalent KIR genotypes in a large cohort of AML patients undergoing MUD transplantation.

Methods: We included adult AML patients in complete remission transplanted from an 8/8-HLA MUD between 2010 and 2016 and reported to the Center for International Blood and Marrow Transplant Research (CIBMTR). Donor-KIR and respective recipient-HLA ligand interactions were assessed in multivariable Cox proportional hazard models for standard transplantation outcomes. To account for the compound effect of simultaneous KIR/HLA interactions, we applied a combinatorial approach to identify aggregate KIR genotypes based on combinations of individual KIR genes. The most frequently observed donor-KIR genotypes, in combination with recipient ligands, were evaluated for association with relapse using multivariable regression. Those associated (p < 0.01) with relapse risk were evaluated for differential relapse in a DRST (German stem-cell registry)/Collaborative Biobank cohort of donors/patients with similar inclusion criteria.

Results: A total of 2,036 transplantations from the CIBMTR were included. Most patients were treated in first complete remission (78%) and received myeloablative conditioning (59%). We first studied eight known interactions between donor KIR and their respective HLA ligands (Figure A). Only donor-KIR-2DL2+/recipient-HLA-C1+ was associated with reduced relapse (compared to donor-KIR-2DL2-/recipient-HLA-C1+, hazard ratio [HR] 0.80 [95% confidence interval 0.67-0.94], p=0.008). However, no difference was found when comparing HLA-C group pairs among KIR-2DL2+ recipients, suggesting this finding is confounded by co-occurrence of other receptors.

There are hundreds of possible KIR gene combinations (i.e. genotypes), which are typically clustered into two primary haplotypes, A and B. To study the cumulative effect of donor KIR, we investigated nine prevalent KIR genotypes (Figure B) and identified three significantly associated with relapse risk. (1) Donor KIR genotype 5 in all recipients irrespective of their HLA (Figure C, n = 138/2,036) and (2) genotype 3 in HLA-Bw4/x recipients (Figure D, n = 51/1,198) had significantly decreased relapse risk (HR 0.53 [0.37-0.78], p=0.002 and 0.34 [0.15-0.75], p=0.008, respectively). (3) KIR genotype 2 was associated with greater relapse in HLA-C1-homozygous recipients (Figure E, n = 87/836, HR 1.62 [1.14-2.30], p=0.007). These findings were not confirmed in the external European dataset (n = 796, Figure 1C-E); however, this cohort differed in ways that might affect the importance of KIRs, such as the higher frequency of reduced intensity conditioning (74% vs. 41%) and in-vivo T-cell depletion (79% vs. 37%).

Conclusion: Our systematic investigation in two large AML cohorts receiving MUD allogenic HCT did not validate any association between individual KIR-HLA interactions and clinical outcomes. A combinatorial approach identified combinations potentially protective against relapse, however these could not be confirmed in a second dataset. Overall, our findings do not support KIR-informed donor selection using the approaches outlined here.

Figure 1

View largeDownload slide

Figure 1

View largeDownload slide

Close modal