Natural killer (NK) cell function can be modulated by the killer cell immunoglobulin-like receptors (KIR) which interact with human leukocyte antigen (HLA) class I molecules on target cells. KIR-ligand mismatching has recently been shown by van Bergen et al. (American Journal of Transplantation 2011; 11(9): 1959–1964) to be a significant risk factor for long-term graft loss in HLA-A, -B and -DR compatible kidney transplants. To verify this potentially important finding, we performed genotyping of 608 deceased-donor kidney graft recipients and their HLA-A, -B and -DR compatible donors for KIR and HLA, using samples and clinical data provided by the Collaborative Transplant Study. Graft survival of KIR-ligand-matched and -mismatched transplants was compared. We found no impact of KIR-ligand mismatching on 10-year graft survival in HLA-A, -B, -DR compatible kidney transplants. Further analysis did not reveal a significant effect of recipient activating/inhibitory KIR or KIR genotypes on graft survival. Our data do not support the concept that KIR-HLA matching might serve as a tool to improve long-term renal allograft survival.
polymerase chain reaction using sequence-specific primers
restriction fragment length polymorphism
Natural killer (NK) cells are lymphocytes that play an important role in innate immunity . They express a variety of inhibitory and activating receptors including the killer cell immunoglobulin-like receptors (KIR). Several KIR are known to bind to human leukocyte antigen (HLA) class I molecules, such as HLA-A3, -A11, HLA proteins carrying a Bw4 motif, HLA-C and HLA-G. HLA-C molecules can be grouped into two groups based on their KIR specificity: HLA-C group 1 (HLA-C1) and HLA-C group 2 (HLA-C2) . Upon engagement of HLA class I ligands, KIR can mediate inhibition or activation of NK cell responses [4-6].
There have been conflicting reports on whether NK cell alloreactivity influences outcomes in kidney transplantation [1, 7-13]. Most recently, van Bergen et al.  reported that, in a cohort of 137 HLA-A, -B and -DR compatible kidney transplants, KIR-ligand mismatches were associated with an approximately 25% reduction in 10-year death-censored graft survival. This impressive result, which if confirmed would have substantial clinical implications, prompted us to investigate the association between KIR-ligand mismatches and graft outcome in the kidney transplant cohort of the Collaborative Transplant Study (CTS). We also reassessed the impact of the presence of activating or inhibitory KIR and of KIR genotypes on graft survival as suggested by previous studies [9, 10, 13].
Materials and Methods
Kidney donors and recipients
Samples (blood or spleen tissue) and clinical data were obtained from transplant centers participating in the Collaborative Transplant Study (CTS) (see Acknowledgments). Only kidney transplants which fulfilled the following selection criteria were included: no multiorgan transplants, deceased donor, recipient and donor of Caucasian origin, transplants performed between 1988 and 2009 in Europe or North America, HLA-A, -B and -DRB1 zero mismatch (serology or DNA-based typing for HLA-A and -B, 2-digit DNA-based typing level for HLA-DRB1). 608 recipient/donor pairs of whom sufficient DNA was available were tested. Demographic characteristics are listed in Table 1.
Table 1. Demographic characteristics of the study cohort (n = 608)
No KIR-ligand mismatch (n = 415)
≥1KIR-ligand mismatch (n = 193)
Cold ischemia time (hr, mean±SD)
Pretransplant antibodies (PRA)
No IMPDH inhibitor
HLA and KIR genotyping
DNA was extracted from peripheral blood or spleen tissue by the salting-out method . HLA-A, -B and -DRB1 typing data were reported by the CTS-participating centers and 91.4% of all recipient/donor pairs were retyped in our Heidelberg laboratory for HLA class I by PCR-SSP (HLA-A/B CTS-PCR-SSP Kit, CTS, Heidelberg, Germany) and 100% were retyped for HLA class II using the RFLP method [15, 16] or the HLA-DRB1 CTS-PCR-SSP Kit (CTS, Heidelberg, Germany) according to the manufacturer's instruction (www.ctstransplant.org). HLA-C1 and -C2 groups were determined by PCR-SSP with primers which differentiated asparagine/lysine at position 80 . In addition, the HLA-C1/C2 group assignment was validated by HLA-C genotyping with PCR-SSP (HLA-C CTS-PCR-SSP Kit, CTS, Heidelberg, Germany) or PCR-SBT (CTS-SEQUENCE HLA-C Kit, CTS, Heidelberg, Germany) according to the manufacturer's instruction. Genotyping of 16 KIR genes (2DL1, 2DL2, 2DL3, 2DL4, 2DL5, 2DS1, 2DS2, 2DS3, 2DS4/2DS4 variant, 2DS5, 2DP1, 3DL1, 3DL2, 3DL3, 3DS1 and 3DP1/3DP1 variant) was performed in our department in Heidelberg with PCR-SSP using primer sequences as published by Ashouri et al. . A discrimination of 2DS4 alleles with full sequence (membrane-bound protein) and 2DS4 alleles with a 22 base pair-deletion (putative soluble protein) ensured the determination of carriers with a functional 2DS4 receptor on the cell surface . Patients who possessed only the deleted version of 2DS4 were not considered as having an activating 2DS4 receptor.
Definition of KIR-ligand mismatch/match, activating/inhibitory KIR and KIR genotypes
KIR-ligand mismatch/match was defined according to van Bergen et al.  (Table 2). To investigate a potential effect of the number of activating or inhibitory KIR, the following known KIR characteristics were taken into account: Activating KIR are encoded by 2DS1, 2DS2, 2DS3, 2DS4 (the membrane-bound version), 2DS5 and 3DS1, whereas inhibitory KIR are products of 2DL1, 2DL2, 2DL3, 2DL5, 3DL1 and 3DL2. 2DL4 was not included in the analysis because it has been reported as having both activating and inhibitory functions [3, 4, 19, 20]. Likewise, 3DL3 was not evaluated since its expression and function remain unclear [6, 21]. To study the impact of KIR genotypes, three KIR genotypes (AA, BB and AB) were differentiated. The KIR genotype was classified as AA if none of the following KIR genes was present: 2DL2, 2DL5, 3DS1, 2DS1, 2DS2, 2DS3 and 2DS5 (according to http://www.allelefrequencies.net/). Although the distinction between AB and BB genotypes is considered difficult , we applied the following definition of the BB genotype for more detailed subgroup analysis of KIR genotypes [9, 22]: If 2DS4 was absent and the AA genotype was excluded, the BB genotype was assigned; all remaining KIR gene combinations (other than AA and BB) were designated AB. An additional analysis of the AA group versus the combined AB and BB groups (classified as Bx genotype) was also carried out.
Table 2. KIR-ligand mismatch/match definition
Number of transplants
KIR2DL2+ or KIR2DL3+
KIR2DL2+ or KIR2DL3+
A3+ or A11+
A3– and A11–
A3+ or A11+
A3+ or A11+
Comparisons of continuous variables between two groups were performed by a nonparametric Mann–Whitney U-test and of categorical variables by an exact Fisher test. Graft, death-censored graft and patient survival rates (mean ± standard error) were computed according to the Kaplan-Meier method and compared between subpopulations using the Mantel Cox log rank test. Multivariate Cox regression analysis was performed to determine hazard ratios (HR) and 95% confidence intervals (95% CI) of KIR-ligand mismatch, KIR genotype, number of activating and number of inhibitory KIR. The following confounders were considered: geographical region, transplant number, transplant year, recipient and donor sex and age, original disease leading to transplantation, pretransplant lymphocyte antibodies (panel-reactive antibodies, PRA), cold ischemia time, type of initial immunosuppressive treatment and antibody induction therapy. p-values below 0.5 were considered significant. A back-step elimination algorithm was used to exclude confounders with p > 0.2. The software package IBM SPSS Statistics 20 (SPSS Inc., Chicago, IL, USA) was used for all statistical analyses.
Analysis of the effect of KIR-ligand mismatching on transplant outcomes
In HLA-A, -B and -DRB1 compatible kidney transplants from deceased donors (n = 608), KIR-ligand mismatch had no impact on 10-year graft survival in univariate analysis (Figure 1). This was confirmed in a multivariate Cox regression model (HR 0.85, 95% CI 0.63–1.16, p = 0.31). Similar results were obtained for 10-year death-censored graft survival (HR 0.75, 95% CI 0.50–1.13, p = 0.17, Table 3 and Figure 2) and patient survival (HR 0.90, 95% CI 0.61–1.33, p = 0.58). We also analyzed the effect of each individual KIR-ligand mismatch separately and in combinations (0, 1 or more than 1 mismatch) and could not identify any significant influence of single or multiple KIR-ligand mismatches on graft survival (data not shown); however, the numbers of cases available for analysis of individual KIR-ligand match/mismatch groups were small (Table 2).
Table 3. Results of 4 Cox regressions for death-censored graft survival during first 10 posttransplant years (n = 608 transplants)
Number of activating KIR
Number of inhibitory KIR
KIR genotype AB
A subgroup analysis of nonsensitized recipients (panel-reactive antibodies PRA ≤ 5%) revealed no significant differences in 10-year death-censored graft survival rates between KIR-ligand-matched (72.9 ± 3.3%, n = 287) and KIR-ligand-mismatched (78.1 ± 4.8%, n = 117) transplants (p = 0.22). A similar negative result was obtained for sensitized recipients with PRA > 5% (KIR-ligand match, 75.9 ± 4.5%, n = 123, versus KIR-ligand mismatch, 70.3 ± 7.5%, n = 73, p = 0.67).
Subcohorts of the 608 HLA-A, -B and -DRB1 compatible transplants were additionally typed for HLA-C (n = 558), -DQB1 (n = 604) and -DPB1 (n = 153). No impact of HLA-C, -DQB1 or -DPB1 mismatches on graft survival could be detected. Furthermore, comparison of graft survival rates between KIR-ligand match and mismatch within each of the subgroups of HLA-C-matched (n = 416) and -mismatched (n = 142), HLA-DQB1-matched (n = 564) and -mismatched (n = 40) as well as HLA-DPB1-matched (n = 41) and -mismatched (n = 112) transplants showed no influence of KIR-ligand mismatching (data not shown).
Analysis of the effect of the recipient KIR repertoire and KIR genotype on graft survival
In a previous publication, increased NK cytotoxicity after kidney transplantation (measured in a cytotoxicity assay) was related to a higher number of recipient activating KIR that recognized donor HLA class I (13). Another publication suggested that a high number of inhibitory KIR genes in the recipient genotype may provide protection against acute renal allograft rejection . We therefore investigated the impact of the number of recipient activating or inhibitory KIR genes. Multivariate analysis failed to reveal any influence of the number of recipient activating (HR 1.00, 95% CI 0.89–1.11, p = 0.95) or inhibitory (HR 0.94, 95% CI 0.75–1.17, p = 0.57) KIR genes on 10-year death-censored graft survival (Table 3).
Nowak et al.  specifically investigated a potential role of the activating receptor KIR2DS5 in acute renal allograft rejection. They found that presence of the 2DS5 gene in the recipient appeared to protect the graft from rejection. Our comparison of 10-year death-censored graft survival rates showed no significant difference between 2DS5 positive (n = 180) or 2DS5 negative (n = 428) transplant recipients (75.9 ± 4.2% and 72.7 ± 2.7%, respectively, p = 0.25).
Since the presence of KIR haplotype B (containing multiple activating KIR genes) has been reported to affect outcomes in hematopoietic stem cell transplantation [23, 24], we investigated whether recipient KIR genotypes might have an impact on kidney allograft survival. Cox regression analysis did not show significant differences for 10-year death-censored graft survival among the three transplant recipient groups with AA (HR 1.26, 95% CI 0.85–1.86, p = 0.26, n = 178), AB (reference group, n = 408) or BB (HR 1.56, 95% CI 0.67–3.63, p = 0.30, n = 22) KIR genotype (Table 3). A similarly negative result was obtained when the AB and BB groups were combined (called Bx genotype) and compared with the AA group (data not shown).
There have been conflicting reports on the impact of KIR and KIR-ligand matching in kidney transplantation [1, 7-13]. Several years ago, we analyzed and found no effect of ‘KIR ligand mismatch’ on renal allograft survival , using the algorithm described by Ruggeri et al. . At that time, however, ‘KIR ligand mismatch’ was defined only as mismatch at the level of HLA epitopes (e.g. C1, C2 and Bw4) which were known to bind to KIR; no KIR genotyping was performed. In the present study, “KIR ligand mismatch” was defined under consideration of recipient KIR genotype, taking into account a potential impact of the putative interaction between the KIR receptor and its ligand (HLA class I molecules). Our study was prompted by the remarkable report by van Bergen et al.  who demonstrated that KIR-ligand mismatching was an independent risk factor for graft loss in HLA-A, -B and -DR compatible kidney transplants. We followed the definition of KIR-ligand mismatch used by Van Bergen at al. based on the recognition of ‘missing self’ via the inhibitory receptors KIR2DL1, KIR2DL2, KIR2DL3, KIR3DL1 and KIR3DL2; in addition, the recognition of ‘nonself’ via the activating receptor KIR2DS1 was also considered for the analysis. The a priori assumption was that KIR-ligand mismatch might induce recipient NK alloreactivity towards the allograft, thereby exerting a detrimental effect on graft survival. Indeed, the presence of one or more KIR-ligand mismatches was found by van Bergen et al. to be associated with a striking reduction of approximately 25% in the 10-year death-censored graft survival rate . Based on their observation, the authors suggested that when allocating donor organs, when there are several HLA-compatible potential recipients, KIR-ligand matching might be appropriate for choosing the best-matched recipient. This impressive finding prompted us to validate the relevance of KIR-ligand mismatching in the kidney transplant cohort of the Collaborative Transplant Study (CTS). All 608 HLA-A, -B and -DR compatible deceased-donor kidney transplants of whom sufficient recipient and donor DNA was available for testing were included in our study. We could not confirm the findings of van Bergen et al. In our study cohort, KIR-ligand mismatch did not have a significant impact on 10-year allograft survival. Up to 5 years posttransplant, KIR ligand mismatch paradoxically appeared to be associated with better graft survival compared with KIR ligand match, a trend was however not sustained with prolonged follow up to 10 years (Figures 1 and 2). One could speculate that a potentially protective effect of KIR-ligand mismatch during the early posttransplant period might be due to recipient NK cell lysis of graft-derived donor antigen presenting cells, which could lead to inhibition of the direct pathway of antigen recognition. However, even if such an effect occurred, it had no clinical impact on long-term outcome. There are some notable differences between our study and the study by van Bergen et al. The total number of transplants investigated by us was 608 as compared with the 137 transplants tested by van Bergen et al. Whereas the latter cohort included transplants which had been performed at only two centers (Leiden and Rotterdam, both in the Netherlands), the CTS analysis was based on transplants from 48 centers worldwide. 4 CTS centers contributed 60 to 80 transplants each, 11 centers submitted samples and data on 11–30 transplants each, and the remaining 33 centers provided materials of less than 10 transplants each to this study. Separate center-specific analysis for each of the 4 major centers and the remaining groups of 11 and 33 centers combined consistently showed no difference in graft survival between KIR-ligand-matched and -mismatched transplants. Additional subanalysis of the 550 transplants performed in European countries also failed to show an effect of KIR ligand mismatching on graft survival (data not shown). Rather than the optimism expressed by some authors , our data suggest that a more cautious view is indicated with respect to a potential clinical application of KIR genotyping and KIR ligand matching for improving renal allograft survival.
In line with our negative results, several groups found no correlation of the incidence of acute graft rejection with KIR genotypes [7-9, 11] or KIR ligand incompatibilities [7, 8]. However, there were three publications which reported statistically significant associations between KIR and NK cell alloreactivity posttransplant or increased risk for acute renal allograft rejection [9, 10, 13]. Remarkably, their results were contradicting: While Vampa et al.  observed an increase in NK cytotoxicity (in vitro assay) posttransplant in recipients who possessed more activating KIR genes specific for donor HLA class I, Nowak et al.  found a protective effect of the activating receptor KIR2DS5 on acute rejection. Kunert et al.  detected no influence of activating KIR, but noticed a protective effect of inhibitory KIR in the recipient genotype against acute graft rejection. In contrast to these authors, we could not see any association of the number of activating or inhibitory KIR, nor of the presence of KIR2DS5 with long-term graft survival. The study of Vampa et al.  contained a very small cohort (n = 21); in addition, the authors did not mention whether the increased NK cell activities measured in vitro impacted on clinical outcomes. Nowak et al.  compared patients with acute rejection and healthy controls instead of nonrejecting patients and similarly studied only a small number of cases (14 patients were KIR2DS5-positive), resulting in a p-value that was no longer significant after Bonferroni correction.
Aiming at a validation of the study by van Bergen et al. , we applied their algorithm for the definition of KIR-ligand mismatches, based on the assumption of a mutually exclusive interaction of KIR2DL1 with HLA-C2 and KIR2DL2/3 with HLA-C1. However, there have been reports on more complex interactions between KIR and HLA-C, showing that some KIR2DL2 and KIR2DL3 alleles (such as KIR2DL2*001 and KIR2DL3*001) interact not only with HLA-C1, but also with several HLA-C2 allotypes . These ‘cross-reactions’ make a division of the patients into KIR-ligand-matched and KIR-ligand-mismatched difficult and require KIR genotyping at allele level which was not performed by van Bergen et al. nor by us. Therefore, whether recipient KIR alleles play a role in renal allograft rejection has not conclusively been answered by these studies.
An issue that cannot be resolved in this study is whether donor-specific antibodies directed against mismatched HLA-C, -DQ or -DP antigens might have played a role in this cohort of HLA-A+B+DR compatible transplants. HLA-C, -DQ or -DP antibodies could theoretically be involved in NK cell antibody-dependent cell-mediated cytotoxicity (ADCC). Whereas no effect of KIR-ligand mismatching on graft survival was seen even in sensitized recipients (PRA > 5%), sera were not available to us for retrospective characterization of HLA antibody specificities. Further studies are needed to clarify whether donor-specific antibodies may influence transplant outcomes in the context of KIR-ligand mismatch.
In summary, we were unable to confirm the significant impact of KIR-ligand mismatching in HLA-A, -B and -DR compatible kidney transplants which was recently reported by van Bergen et al. . Reassessment of reports on an effect of activating/inhibitory KIR or KIR genotypes on graft survival also failed to show any significant associations. Our data do not support the concept of using KIR-ligand matching or KIR genotype as a potential tool for optimizing kidney allocation. The biological role of NK cells in kidney transplantation is complex and requires further investigation. Although theoretically plausible, the prediction of NK cell alloreactivity merely based on KIR-ligand matching does not appear to correlate with clinical outcomes.
We are very grateful to the following transplant centers participating in the kidney transplant DNA typing project of the Collaborative Transplant Study (CTS) for their invaluable support:
Basel (Switzerland), Berlin (Germany), Berlin, University Hospital Benjamin Franklin (Germany), Bern (Switzerland), Bochum (Germany), Bremen (Germany), Cambridge (United Kingdom), Cardiff (United Kingdom), Chicago (United States of America), Debrecen (Hungary), Dublin (Ireland), Edmonton (Canada), Erlangen-Nuernberg (Germany), Essen (Germany), Frankfurt (Germany), Freiburg (Germany), Giessen (Germany), Glasgow (United Kingdom), Hann.-Muenden (Germany), Hannover (Germany), Heidelberg (Germany), Innsbruck (Austria), Kaiserslautern (Germany), Kiel (Germany), Lausanne (Switzerland), Leicester (United Kingdom), Leuven (Belgium), Lisbon (Portugal), Louisville (United States of America), Luebeck (Germany), Malmo-Lund (Sweden), Mannheim (Germany), Marburg (Germany), Minneapolis (United States of America), Muenster (Germany), Nantes (France), Portland (United States of America), Prague (Czech Republic), Quebec (Canada), Rijeka (Croatia), Rostock (Germany), San Antonio (United States of America), St. Etienne (France), Szeged (Hungary), Toronto, General Hospital (Canada), Toulouse (France), Ulm (Germany), Zurich (Switzerland).
We thank Christiane Weis, Roswitha Kohler, Nicole Ballreich-Jung, Evelina Kudlek, Karin Kotelmann and Cornelia Klages for their excellent technical assistance.
The authors of this manuscript have no conflicts of interest to disclose as described by theAmerican Journal of Transplantation.