KIR-Ligand Mismatches Are Associated With Reduced Long-Term Graft Survival in HLA-Compatible Kidney Transplantation


Jeroen van Bergen,


Natural killer (NK) cells are cytotoxic lymphocytes of the innate immune system with the ability to detect HLA class I disparities via killer-cell immunoglobulin-like receptors (KIR). To test whether such KIR-ligand mismatches contribute to the rejection of human solid allografts, we did a retrospective cohort study of 397 HLA-DR-compatible kidney transplantations and determined the KIR and HLA genotypes of recipients and the HLA genotypes of donors. In transplantations compatible for HLA-A, HLA-B and HLA-DR (n = 137), in which a role for T cells and HLA antibodies in rejection was minimized, KIR-ligand mismatches were associated with an approximately 25% reduction in 10-year death-censored graft survival (p = 0.043). This effect was comparable to the effect of classical HLA-A and HLA-B incompatibility, and in HLA-A,-B-incompatible transplantations (n = 260) no significant additional effect of KIR-ligand mismatches was observed. Multivariate Cox regression analysis confirmed the effect of KIR-ligand mismatching as an independent risk factor in HLA-A,-B,-DR-compatible transplantations (hazard ratio 2.29, range 1.03–5.10, p = 0.043). This finding constitutes the first indication that alloreactive NK cells may thwart the success of HLA-compatible kidney transplantations, and suggests that suppression of NK-cell activity can improve the survival of such kidney grafts.


anti-thymocyte globulin


BK virus


group 1 HLA-C (HLA-CN80)


group 2 HLA-C (HLA-CK80)




collaborative transplant study


Epstein-Barr virus-transformed lymphoblastoid cell line


JC virus


killer cell Immunoglobulin-like receptor(s)

NK cell

natural killer cell


polymerase chain reaction using sequence-specific primers


HLA-A, HLA-B and HLA-DR incompatibility hampers successful human organ transplantation. CD8+ and CD4+ T cells respond to nonshared donor HLA class I and II molecules. They play a pivotal role in graft rejection, by direct cellular cytotoxicity and by providing help to B cells specific for these antigens. Pre-existing antibodies mediate hyperacute rejection, while early and chronic rejections are believed to be mediated by T cells and by antibodies that appear after transplantation. Natural killer (NK) cells also have the ability to respond to HLA class I disparities. Like CD8+ T cells, NK cells are cytotoxic cells that can produce IFNγ when they encounter allogeneic cells in vitro, and they might therefore also contribute to the rejection of solid allografts (1).

Transplantations that are classically classified as HLA compatible can still be mismatched from an NK-cell perspective. This is due to the unique ability of NK cells to respond to ‘missing self’ HLA antigens. NK cells expressing an inhibitory NK receptor binding ‘self’ HLA can be activated when confronted with allogeneic cells lacking a ligand for that inhibitory receptor. Humans use killer-cell immunoglobulin-like receptors (KIR) to discriminate between HLA-A, -B and -C allotypes. For example, the inhibitory KIR2DL1 binds HLA-C molecules carrying a lysine (K) at position 80 of the heavy chain (group 2 HLA-C or C2), but not HLA-C molecules carrying an asparagine (N) at this position (group 1 HLA-C or C1). Therefore, in an HLA-Cw4/HLA-Cw3 (C2/C1) recipient carrying the KIR2DL1 gene, KIR2DL1+ NK cells are self-tolerant because they are inhibited via the interaction between KIR2DL1 and HLA-Cw4 (C2). However, these NK cells can lyse cells from a HLA-Cw3 homozygous (C1/C1) donor, as 2DL1 does not bind HLA-Cw3, releasing the brake on NK-cell activation. In addition, and similar to T cells, NK cells can respond to ‘nonself’ C2 via the activating KIR2DS1 (2). The same principles apply to other KIR-ligand interactions, and allow the reliable prediction of NK-cell alloreactivity on the basis of such KIR-ligand mismatches defined by patient KIR and HLA and donor HLA genotyping (3–5).

It is unclear whether NK-cell alloreactivity influences solid organ transplantation. NK cells do infiltrate kidney allografts within days after transplantation (6), and this is accompanied by antidonor NK-cell alloreactivity detectable in blood samples (7). Yet, in the largest genetic association study performed to date (n = 2757), no effect of KIR ligand incompatibility on 10-year kidney graft survival was observed (8). However, all studies in kidney transplantation thus far included transplantations incompatible for HLA-A, HLA-B and/or HLA-DR and alloreactive T cells may therefore have dominated the alloimmune response, possibly obscuring NK-cell effects (7–11). Therefore, we here examined the role of KIR-ligand mismatches in kidney transplantation in a large number (n = 397) of HLA-DR-compatible kidney transplantations, a third of which were also HLA-A and HLA-B compatible.

Material and Methods

Study design and participants

This retrospective cohort study included kidney transplantations performed in the Leiden University Medical Center (LUMC, Leiden, the Netherlands, n = 204) and in the Erasmus Medical Center (EMC, Rotterdam, the Netherlands, n = 193) between 1990 and 2004. For both transplant centers, the HLA-A, HLA-B and HLA-DR type of donors and recipients had been determined in a single tissue typing laboratory (LUMC), through serology and/or two-digit DNA-based genotyping. The patients received standard post-transplant immune suppression consisting of calcineurin inhibitors and corticosteroids. Immune suppression was never tapered, but was supplemented with ATG to treat any rejection episodes.

KIR and HLA-C typing

For this study, HLA-C and KIR genotyping was performed without knowledge of clinical outcomes. For HLA-C genotyping, a PCR-SSP typing protocol was used that discriminated between HLA-C group 1 and 2 allotypes, and that had been validated using a large panel of EBV-LCL and material from HLA-C typed donors. KIR genotyping was performed by PCR-SSP as well. Detailed methods, primer sequences and raw genotyping results can be found in the online Supporting Information (Supporting Methods, Tables S1 and S2).

Clinical data

Graft loss was defined as graft failure requiring hemodialysis and/or retransplantation or resulting in recipient death. Recipients who died with a functioning graft were excluded from the analysis from the time of death onward (death censored). Initial disease diagnoses and causes of graft failure are listed in supplementary data tables S4 and S5.

Statistical methods

The primary endpoint was death-censored graft survival. Graft survival rates were computed according to the Kaplan–Meier method, comparing transplantations with and without KIR-ligand mismatches (Log-rank test). Cox regression analysis was subsequently performed to determine hazard ratios considering the following covariates: recipient sex, donor sex, recipient age, donor age, recipient blood group (ABO), donor blood group (ABO), donor type (deceased or living), transplantation period (1990–1996 or 1997–2004), current panel reactivity, highest panel reactivity, cold ischemia period (in hours), repeat transplant and donor CMV serostatus. Patient CMV serology and patient and donor EBV, BKV and JCV serology data were generally not determined in this cohort. Each variable was first tested in univariate analysis, and variables significantly associated with graft survival rate were subsequently tested in multivariate analysis. p values below 0.05 were considered significant. All statistical analyses were performed using the software package SPSS (version 16.0, SPSS Inc., Chicago, IL).


We reasoned that KIR-ligand mismatching effects on kidney transplant survival would be most apparent in HLA-compatible transplantations, where minimal T- and B-cell alloreactivity was expected. For this reason, we analyzed separately transplantations compatible for HLA-A, HLA-B and HLA-DR (n = 137), and less well-matched transplantations compatible for HLA-DR but not HLA-A or HLA-B (n = 260). NK-cell alloreactivity toward the donor was expected if there was at least one KIR-ligand mismatch based on the recognition of ‘missing self’ via KIR2DL1, KIR2DL2, KIR2DL3, KIR3DL1, KIR3DL2 or the recognition of ‘nonself’ via KIR2DS1 (3,5). This algorithm (Table 1) included not only the presence of an HLA difference between recipient and donor (3), but also took into account the requirement for the presence of the corresponding recipient KIR to detect that HLA difference (5).

Table 1.  KIR-ligand mismatch definitions
 KIR-ligand mismatch1
RecipientDonor HLA
  1. 1KIR-ligand mismatches were defined according to Ruggeri et al. (3,4), taking into account the fact that such a mismatch requires the presence of the corresponding KIR to detect it (5). For example, a KIR2DL1-ligand mismatch requires the presence of the KIR2DL1-gene and a HLA-C group 2 allele (C2+) in the recipient, together with the absence of HLA-C group 2 alleles from the donor's genotype (C2-). KIR2DS1 is an activating receptor that binds HLA-C group 2, allowing NK cells to respond to allogeneic C2-positive target cells (2,5). For this reason, KIR2DS1 was also included in the algorithm. Transplantations were classified as KIR-ligand mismatched if at least one of these mismatches applied, in which case NK-cell mediated alloreactivity toward the donor was expected.

KIR2DL2/3KIR2DL2+ or KIR2DL3+C1+C1-
KIR3DL2KIR3DL2+A3+ or A11+A3- and A11-

In the transplantations compatible for HLA-A, HLA-B and HLA-DR (for baseline characteristics, see Table 2), univariate analysis showed that the presence of one or more KIR-ligand mismatches was associated with a reduced 10-year graft survival, from 81% to 59% (log-rank: p = 0.043, Figure 1A). The presence of a KIR B-haplotype, characterized by the presence of multiple activating KIR, in the patients was not associated with graft survival (not shown). Separate analyses of the effects of individual KIR-ligand mismatches were not informative due to the low number of cases available. HLA-A and HLA-B incompatibility is known to induce antibodies and T-cell alloreactivity and is associated with lower graft survival (12). In line with this, KIR-ligand mismatches did not significantly affect graft survival in HLA-A and/or HLA-B-incompatible (HLA-DR compatible) transplantations (Figure 1B). Irrespective of KIR-ligand mismatches, the 10-year survival of these transplants was approximately 59%. Therefore, in HLA-DR-compatible kidney transplantations, the impact of KIR-ligand mismatches on 10-year graft survival was comparable to the impact of classical HLA-A and HLA-B incompatibility.

Table 2.  Baseline characteristics of HLA-DR-compatible transplantations
  HLA-A,-B compatibleHLA-A,-B incompatible
  KIR-ligand mismatchesTotalChi-square valueKIR-ligand mismatchesTotalChi-square value
  1. Data are presented as number (%). n.s., not significant. * These p-values are unreliable, as there are too many cells with expected values lower than 5.

  2. Panel reactivity <5% defines nonimmunized patients. The cut-off for the transplantation period was set at 1996 as the Eurotransplant kidney allocations system (ETKAS) was implemented in that year. Furthermore, this date divides the cohort into two equally sized groups.

Recipient sexFemale34 (35.8%)21 (50.0%)55 (40.1%)n.s.58 (46.0%)50 (37.3%)108 (41.5%)n.s.
Male61 (64.2%)21 (50.0%)82 (59.9%) 68 (54.0%)84 (62.7%)152 (58.5%) 
Donor sexFemale45 (47.4%)12 (28.6%)57 (41.6%)0.04049 (38.9%)54 (40.3%)103 (39.6%)n.s.
Male50 (52.6%)30 (71.4%)80 (58.4%) 77 (61.1%)80 (59.7%)157 (60.4%) 
Recipient age (years)≤5056 (58.9%)20 (47.6%)76 (55.5%)n.s.70 (55.6%)74 (55.2%)144 (55.4%)n.s.
>5039 (41.1%)22 (52.4%)61 (44.5%) 56 (44.4%)60 (44.8%)116 (44.6%) 
Donor age (years)≤5065 (68.4%)28 (66.7%)93 (67.9%)n.s.79 (62.7%)86 (64.2%)165 (63.5%)n.s.
>5030 (31.6%)14 (33.3%)44 (32.1%) 47 (37.3%)48 (35.8%)95 (36.5%) 
ABO recipientO44 (46.3%)18 (42.9%)62 (45.3%)*51 (40.5%)51 (38.1%)102 (39.2%)*
A43 (45.3%)17 (40.5%)60 (43.8%) 48 (38.1%)53 (39.6)101 (38.8%) 
B5 (5.3%)6 (14.3%)11 (8.0%) 20 (15.9%)23 (17.2%)43 (16.5%) 
AB3 (3.2%)1 (2.4%)4 (2.9%) 7 (5.6%)7 (5.2%)14 (5.4%) 
ABO donorO50 (52.6%)20 (47.6%)70 (51.1%)*60 (48.0%)67 (50.0%)127 (49.0%)*
A42 (44.2%)17 (40.5%)59 (43.1%) 47 (37.6%)48 (35.8%)95 (36.7%) 
B1 (1.1%)5 (11.9%)6 (4.4%) 15 (12.0%)17 (12.7%)32 (12.4%) 
AB2 (2.1%)0 (0.0%)2 (1.5%) 3 (2.4%)2 (1.5%)5 (1.9%) 
Donor typeDeceased77 (81.1%)41 (97.6%)118 (86.1%)0.010121 (96.0%)128 (95.5%)249 (95.8%)n.s.
Living18 (18.9%)1 (2.4%)19 (13.9%) 5 (4.0%)6 (4.5%)11 (4.2%) 
Donor CMV positiveNo39 (47.6%)23 (50.0%)62 (48.4%)n.s.58 (49.6%)58 (44.3%)116 (46.8%)n.s.
Yes43 (52.4%)23 (50.0%)66 (51.6%) 59 (50.4%)73 (55.7%)132 (53.2%) 
Transplantation period≤199629 (30.5%)19 (45.2%)48 (35.0%)n.s.59 (46.8%)72 (53.7%)131 (50.4%)n.s.
>199666 (69.5%)23 (54.8%)89 (65.0%) 67 (53.2%)62 (46.3%)129 (49.6%) 
Current panel reactivity (%)<562 (65.3%)33 (78.6%)95 (69.3%)n.s.92 (73.0%)97 (72.4%)189 (72.7%)n.s.
≥533 (34.7%)9 (21.4%)42 (30.7%) 34 (27.0%)37 (27.6%)71 (27.3%) 
Highest panel reactivity (%)<526 (27.4%)11 (26.2%)37 (27.0%)n.s.35 (27.8%)37 (27.6%)72 (27.7%)n.s.
≥569 (72.6%)31 (73.8%)100 (73.0%) 91 (72.2%)97 (72.4%)188 (72.3%) 
Cold ischemia period (hour)≤2459 (62.8%)27 (67.5%)86 (64.2%)n.s.79 (66.9%)70 (56.0%)149 (61.3%)n.s.
>2435 (37.2%)13 (32.5%)48 (35.8%) 39 (33.1%)55 (44.0%)94 (38.7%) 
Repeat transplantNo79 (83.2%)37 (88.1%)116 (84.7%)n.s.100 (79.4%)111 (82.8%)211 (81.2%)n.s.
Yes16 (16.8%)5 (11.9%)21 (15.3%) 26 (20.6%)23 (17.2%)49 (18.8%) 
Rejection(s) in first 6 months post-transplantNo75 (78.9%)28 (66.7%)103 (75.2%)n.s.87 (69.6%)90 (69.8%)177 (69.7%)n.s.
Yes20 (21.1%)14 (33.3%)34 (24.8%) 38 (30.4%)39 (30.2%)77 (30.3%) 
ATGNo81 (85.3%)32 (76.2%)113 (82.5%)n.s.101 (80.8%)108 (83.7%)209 (82.3%)n.s.
Yes14 (14.7%)10 (23.8%)24 (17.5%) 24 (19.2%)21 (16.3%)45 (17.7%) 
Figure 1.

Effect of KIR-ligand mismatching on death-censored graft survival in (A) HLA-A,-B,-DR-compatible and (B) HLA-DR-compatible, but HLA-A,-B-incompatible transplantations. Vertical bars indicate censored events. For the definition of KIR-ligand mismatches, see Table 1; for the complete overview of KIR ligand mismatches in these transplantations, see Table S3; for causes of graft failure, see Table S5. p-Values were based on the log-rank test on Kaplan–Meier plots.

Multivariate Cox regression analysis considering risk factors that are known to influence transplant outcome (Table 3) confirmed the effect of KIR-ligand mismatching as an independent risk factor for graft loss on HLA-A,-B,-DR-compatible transplantations (HR 2.29, range 1.03–5.10, p = 0.043). Potential confounders such as donor type (deceased or living) and donor sex, which were unequally distributed between KIR-ligand matched and mismatched transplantations (Table 2), did not significantly influence graft survival. The only other independent risk factor was transplantation number: in line with previous studies (13), first transplants survived longer than subsequent transplants (HR 3.26, range 1.39–7.64, p = 0.006).

Table 3.  Multivariate Cox regression analysis
  1. Variables tested: recipient sex, donor sex, recipient age, donor age, ABO recipient, ABO donor, donor type (deceased/living), transplantation period (before or after 1996), current panel reactivity (%), highest panel reactivity (%), donor CMV status, cold ischemia period, repeat transplantation, KIR-ligand mismatch. Repeat transplants and the presence of a KIR-ligand mismatch were the only variables significantly associated with graft survival.

KIR-ligand mismatch2.
Repeat transplant3.261.397.640.006


Our data reveal that KIR-ligand mismatching has a significant negative impact on the long-term survival of HLA-A,-B,
-DR-compatible kidney grafts. To the best of our knowledge, this constitutes the first indication that NK-cell alloreactivity may hinder HLA-compatible human solid organ transplantation.

Previous genetic association studies in kidney transplantation examined the influence of KIR on the occurrence of rejection in the first year after transplantation (10,11). The number of activating KIR genes in the patient did not affect rejection-free survival within the first year after transplantation (10,11). One study observed that the presence of KIR2DL2 and KIR2DS2 in patients was associated with the absence of acute rejection within three months after transplantation, but only when the graft carried HLA-C group 1 alleles (10). Taken together, the evidence for a role for KIR in the short-term outcome of kidney transplantation is limited. This does not necessarily mean that NK cells have no role to play early after transplantation, as recent data indicate that NK cells can act as effector cells in antibody-mediated rejection in patients with donor-specific antibodies (14), which may include MICA-specific antibodies (15).

Since the studies described above did not take into account the patients’ HLA class I, they did not address a possible role for NK-cell alloreactivity directed against the donor. The presence in an individual of alloreactive NK-cell clones able to lyse cells from another individual can be predicted by HLA genotyping both individuals (3,4). NK-cell alloreactivity against donor cells is expected when the patient's NK cells detect the absence from the graft of one or more of the patient's HLA class I allele groups recognized by inhibitory KIR (i.e. C1, C2, Bw4, A3/11). The Collaborative Transplant Study (CTS) examined the effect of such KIR-ligand mismatches on long-term kidney graft survival, but found no effect in 2757 transplantations (8). However, the CTS data did not include KIR genotyping and covered a highly heterogenous set of transplantations from many centers on multiple continents, the large majority of which were HLA-A, -B and/or -DR incompatible, which may have diluted effects of KIR-ligand mismatches within the HLA-A,-B,-DR-compatible subgroup.

In this study, we used an improved definition of KIR-ligand mismatches (5) that takes into account the requirement for the presence of the corresponding recipient KIR to detect KIR-ligand mismatches (i.e. HLA allele group differences), as well as the recent observation that NK cells can respond to ‘nonself’ C2 via the activating KIR2DS1 (2). In HLA-A, -B, -DR-compatible transplantations derived from only two transplant centers in a single country, and served by a single histocompatibility laboratory, we report a significant reduction of long-term graft survival associated with the presence of KIR-ligand mismatches.

The presence of KIR-ligand mismatches, as determined by genotyping, is predictive of in vitro NK-cell alloreactivity between two individuals (3–5). This strongly suggests, but does not prove, that alloreactive NK cells mediate the effects of KIR-ligand mismatches we observe in a HLA-compatible transplantations. It is also possible that other KIR+-cell populations, such as memory T cells, mediate this effect, which would be in keeping with the observed effect on long- rather than short-term survival. In addition, NK cells need not directly lyse graft tissue to influence long-term graft survival, but could also exert their effect by influencing downstream immune responses damaging the graft. Along these lines, NK-cell activation shortly after transplantation, as observed by Vampa and colleagues (7), might boost low-level T-cell alloreactivity in HLA-A, -B, -DR-compatible (but HLA-C, -DQ, -DP incompatible) transplantations. Finally, it is possible that NK cells do directly affect long-term graft survival, as new alloreactive NK cells are continuously generated by the bone marrow, and are unlikely to be be tolerized by a transplanted kidney. To discriminate between these possibilities, it will be necessary to determine the presence, location and phenotype of KIR-expressing cells in kidney grafts at multiple time points after transplantation.

Mouse studies support a role for alloreactive NK cells in solid organ transplantation. In cardiac allograft models where little to no T- and B-cell alloreactivity can occur, NK cells contribute to acute rejection and chronic graft vasculopathy (16,17). Of note, chronic vasculopathy is one of the main causes of long-term failure of organ transplants, and may therefore be one the underlying mechanisms behind the putative NK-cell effects we observed. Furthermore, Rag−/− mice, which do not possess functional B or T cells, can reject allogeneic skin grafts provided their NK cells are preactivated by IL-15 (18). In settings where T and B cells play a significant role, NK cells inhibit alloreactive T-cell responses by removing graft antigen-presenting cells (APC) from lymph nodes (19). Thus, in mouse transplantation models, NK cells positively and negatively regulate the rejection of solid allografts, depending on the magnitude of the adaptive allo-immune response. In HLA-incompatible human kidney transplantations, the net result of these opposing effects may be neutral: in line with a previous report (8), we observed no effect of KIR-ligand mismatching in HLA-A and/or HLA-B-incompatible transplantations. As the impact of HLA-DR incompatibility is larger than the impact of HLA class I incompatibility, it is conceivable that in HLA-DR-incompatible transplantations, KIR-ligand mismatches may in fact improve graft survival. In short, the results obtained in experimental mouse models correspond well with our results and help reconcile our current data with previous data obtained in humans (8).

Even though HLA-A, -B, -DR-compatible transplants are deemed fully compatible, other polymorphic loci in the HLA region can also trigger B- and T-cell mediated immune responses and thereby contribute to graft loss. Even though we did not address this issue directly, this is unlikely to explain our data. First, HLA-C, HLA-DQ and MICA mismatches are unlikely to segregate with the presence of KIR-ligand mismatches. Since HLA-C encodes many alleles but only two types of KIR-ligands, HLA-C-incompatible transplantations can be KIR-ligand matched. Inversely, due to the nature of missing-self recognition, HLA-C-compatible transplantations can be KIR-ligand mismatched. Second, since MICA is in strong linkage dis-equilibrium with HLA-B, HLA-B-compatible transplantations are likely to be MICA-compatible as well. The same applies to HLA-DR and HLA-DQ. Finally, the effects of these mismatches on graft survival are too small to explain the size of the KIR-ligand mismatching effect we observed.

Currently, one in five Eurotransplant kidney transplantations is HLA-A, HLA-B and HLA-DR-compatible, while in the United States about 6–20% of the patients receive a fully compatible graft depending on the recipient's race. In graft allocation, if there are multiple potential recipients compatible for HLA-A, -B and -DR, KIR-ligand matching may help choose the best-matched recipient. Alternatively, our findings may help identify recipients who would benefit from immunosuppressive treatment to suppress NK-cell activity. Studies in humans and rats indicate that cyclosporine and FK506, the dominant immunosuppressive agents in this study, do not impair NK-cell function (20). Viable alternatives may be rapamycin or MPA, as these drugs greatly reduce human NK-cell cytotoxicity (20).

In summary, our data suggest that not only alloreactive T and B cells, but also alloreactive NK cells can contribute to the loss of solid allografts. These findings support the idea that NK cells contribute to graft loss (1) and may help optimize the results of kidney transplantation.


We are grateful to Arno van der Slik (Leiden, the Netherlands) and Dr. Carlos Vilches (Madrid, Spain) for advice in KIR genotyping and sharing unpublished data, and to prof. J. Trowsdale and prof. R. Willemze for critical reading of this manuscript. This work was supported by a grant from the Dutch Kidney Foundation to J.v.B. and I.I.N.D. The Dutch Kidney Foundation had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.


Authors and contributors

JB and IIND designed the study concept, AT performed the KIR and HLA-C genotyping, GH did the data collection and statistical analysis, JIR and JWF contributed clinical data, and JB, FHJC, FK and IIND supervised the project. JB wrote the paper. All authors helped to interpret the data and all read and approved the final version. No part of the manuscript was prepared or funded by a commercial organization.

Conflict of interest

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.