Viral infection is a common complication after kidney transplantation. The role of natural killer cells (NK cells) in this setting remains unknown. NK cells express activating and inhibitory killer cell immunoglobulin-like receptors (KIR). We analyzed whether activating KIR genes carried by kidney transplant-recipients influence the rate of viral infection during the first year after transplantation. In patients with a KIR A/A genotype (n = 40, KIR2DS4 only activating KIR) the rate of cytomegalovirus (CMV) infection and reactivation was 36%, as compared to 20% in transplant recipients with more than one activating KIR gene (KIR B/X genotype, n = 82, p = 0.04). Adjusting for other risk factors in Cox regression, the relative risk of B versus A genotype patients was 0.34 (95% CI 0.15–0.76, p = 0.009). The degree of protection increased with the number of activating KIR genes. Symptomatic CMV disease was only observed in four individuals, all carrying a KIR A/A genotype. As for viral infections other than CMV, and for bacterial infections, no KIR-linked protective effect could be detected. Also, graft function and the rate-rejection episodes were similar in KIR A/A and KIR B/X genotype individuals. This study supports a role for activating KIR in the control of CMV infection after kidney transplantation.
‘Nonself’ or ‘altered self’ triggers natural killer (NK) cell activity. Cells modified by viral infection and cells from allogeneic donors thus both represent potential targets for NK cells.
In hematopoietic stem cell transplants (HSCT) the role of allo-reactive NK cells has been well established (1). Particularly, in human leukocyte antigen (HLA) mismatched HSCT, NK cells have been demonstrated to exert powerful graft-versus-leukemia effects (2). NK cells may also contribute to rejection of HLA-mismatched hematopoietic grafts, a property that led to their initial description (3). More recently, also in HSCT, a role for NK cells in the control of infection has been suggested. Specifically, patients at high risk for infection have been identified by genotyping killer cell immunoglobulin-like receptors (KIR), a family of inhibitory and activating receptors expressed on NK cells and a small subset of T cells. A simple dichotomization distinguishes Group A haplotypes (carrying as their only activating receptor KIR2DS4) from Group B haplotypes (containing additional activating KIR genes) (4). Based on the presence of activating KIR genes other than KIR2DS4, genotypes may be divided into AA (homozygous A, KIR2DS4 only activating KIR) and BX (BB or BA, presence of activating KIR other than KIR2DS4). This grouping has proven useful in the risk assessment of individuals undergoing HSCT as patients receiving hematopoietic grafts from group BX genotype individuals have a reduced risk of infection and transplant-related mortality (5–8).
Similar to HSCT recipients, in patients after solid organ transplantation T-cell immunity is compromised (9). Little is known about the role of NK cells with regards to both the control of viral infection and alloreactivity in the context of solid organ transplantation.
Here we investigated, in kidney transplant recipients grouped according to their KIR genotype, the association between the recipient's KIR genotype and (i) the frequency of posttransplant infectious complications and (ii) graft rejection episodes and transplant function.
Patients and Methods
Patients receiving a kidney transplant between July 2004 and May 2007 were enrolled in this study (n = 122). Written informed consent was obtained from all study participants and the study was IRB approved. Seventy-four study participants received an organ from a deceased donor, 48 participants received an organ from a living related donor. Main underlying diagnoses were diabetic/hypertensive/vascular nephropathy (n = 36), polycystic kidney disease (n = 31), glomerulonephritis (n = 22) and other/unknown (n = 33).
The large majority of patients was initiated on a triple-agent immunosuppressive regimen containing a calcineurin inhibitor (tacrolimus or cyclosporine A); azathioprin or mycophenolate; and either prednisone or rapamycin. Patients with a high-risk CMV constellation (i.e. donor positive/recipient negative), and patients treated with antithymocyte globulin (ATG) as induction therapy or for early rejection received prophylactic treatment with valgancyclovir for 4 months after transplantation or rejection, respectively. Patient characteristics and therapeutic regimens are summarized in Table 1.
Table 1. Patient characteristics
A/A Genotype (n = 40)
B/X Genotype (n = 82)
Patient age at transplantation
Type of transplant (n,%)
HLA mismatches (A/B/DR)
Tacrolimus, MMF, prednisone
Tacrolimus, MMF, sirolimus
Tacrolimus, azathioprin, prednisone
Antibody induction (n,%)
CMV serology (n,%)
Don neg/rec neg
Don neg/rec pos
Don pos/rec neg
Don pos/rec pos
EBV Serology (n,%)
Don neg/rec neg
Don neg/rec pos
Don pos/rec neg
Don pos/rec pos
Don n.a./rec pos
Valgancyclovir prophylaxis (n,%)
Endpoints and definitions
Details on the clinical transplant outcome were obtained from reviewing patients’ charts. Evaluated outcomes (major endpoints) were predefined and included the incidence of (i) viral infections/reactivations (defined as a combined endpoint by documented replication assessed by polymerase chain reaction [PCR] and/or antigenemia), and (ii) bacterial infections (defined as clinical evidence for bacterial infection plus either microbiological confirmation or documented systemic antibacterial treatment). Patients were screened for cytomegalovirus (CMV) (by pp65 antigenemia or quantitative realtime PCR) and Epstein-Barr virus (EBV) (by quantitative realtime PCR) regularly during follow-up (at least once/month), other viral infections were searched for only in case of clinical suspicion.
Minor endpoints (also predefined) were rejection-free survival (defined as days from transplantation until biopsy proven clinical or subclinical rejection ≥ grade Banff Ia (10)) and kidney function assessed by estimating glomerular filtration rate according to the MDRD formula at 1, 3, 6 and 12 months posttransplantation (11).
KIR and HLA genotyping
Donors and patients were typed for HLA-A, HLA-B and HLA-DR1 by serology at the time of transplantation, and results were confirmed by sequence-specific primer PCR. Additionally, peripheral blood mononuclear cell-samples of transplant recipients were collected at time of transplantation and stored in liquid nitrogen until use. All further genotyping was performed retrospectively, i.e. at the end of follow-up. Cryopreserved recipient cells (peripheral blood mononuclear cells [PBMC]) were used for KIR genotyping applying a reverse sequence-specific oligonucleotide method according to the manufacturer's instructions (Onelambda Inc., Canoga Park, CA). Briefly, three separate PCR reactions (exons 3, 5 and 7–9) were conducted for each sample using biotinylated KIR exon-specific primer sets. Each PCR product was denaturated, hybridized to complementary DNA probes coupled to fluorescently coded microspheres, and stained with phycoerythrin-conjugated streptavidin. Binding of PCR product to the microspheres was then assessed using a LABScan 100 flow analyzer (Luminex Corporation, Austin, TX). The assignment of genotypes was based on the reaction pattern compared to patterns associated with published KIR gene sequences. The same technology was used for retrospective intermediate resolution typing of HLA-C in all recipients.
The presence of KIR ligands in recipients was assessed by grouping each patient's HLA class I antigens according to defined specificities: patients were considered to have C1 group ligands if they possessed an HLA-C molecule with an asparagine at position 80 (e.g. HLA-Cw1, Cw3, Cw7, Cw8), C2 group ligands if they possessed an HLA-C molecule with a lysine residue at position 80 (e.g. HLA-Cw2, Cw4, Cw5, Cw6), and Bw4 ligands if their HLA-B antigens included at least one antigen with the Bw4 specificity (e.g. HLA-B5, B13, B17, B27). As inhibitory and activating KIR share extensive homology in their extracellular domain (and may therefore share ligands) we determined the number of potential activating KIR/HLA interactions based on the presence of activating KIR/HLA-antigen pairs as follows: patients received 1 point if they possessed both the KIR2DS1 gene and an HLA-C antigen belonging to the C2 group (its putative ligand), another point if they possessed either KIR2DS2 and an HLA-C antigen belonging to the C1 group and a third point if they possessed both KIR3DS1 and an HLA-B antigen belonging to the Bw4 group.
Cumulative incidences were calculated for viral infection/reactivation (combined endpoint) and compared using the Gray test. Cox models were used to compare the impact of KIR genotypes on infection-associated morbidity in multivariate analyses. Covariates analyzed in the Cox models included the number of activating KIR genes, the number of activating KIR–HLA interactions, the presence of activating KIR genes, the type of immunosuppression, donor and recipient pretransplant virus serological status, CMV-prophylaxis with valgancyclovir (as a time-dependent covariate) and immunosuppressive treatment for transplant rejection (as a time-dependent covariate). KIR genotype was forced into all models, other covariates were added in a forward-stepwise conditional fashion with a statistical cutoff value set at 0.05. Rejection-free survival was estimated using the Kaplan–Meier estimator and compared by log rank test. Pretransplant characteristics of group A and group B patients were compared using the Mann–Whitney U-test or Pearson's Chi-square test where appropriate. All analyses were performed using the R statistical package, version 2.6.0 (http://www.R-project.org). All p-values were two-sided and were considered significant if < 0.05 unless otherwise stated. When analyzing single-activating KIR genes (KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS4, KIR2DS5, KIR3DS1) as predictors of infectious complications, a significance level of 0.01 was used to compensate for multiple testing.
Frequency distribution of KIR A/A versus KIR B/X genotypes
Based on the absence of activating KIR other than KIR2DS4, 40 patients (32.8%) were assigned a KIR A/A genotype, the remaining 82 patients (67.2%) were assigned a KIR B/X genotype, reflecting a frequency distribution in accordance with the published literature (12).
Patients with a KIR A/A versus a KIR B/X genotype did not significantly differ with regards to age, HLA mismatches of organs received, type of immunosuppressive regimen and CMV and EBV serological status (Table 1). By contrast, significantly more KIR A/A genotype than KIR B/X genotype individuals received an organ from a living donor (p = 0.04), and were treated with valgancyclovir (p = 0.04).
Patients with a KIR B/X genotype were further grouped according to the number of activating KIR (i.e. KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS4, KIR2DS5, KIR3DS1). Thirteen patients (16%) possessed between one or two activating KIR, 47 patients (57%) had three or four activating KIR genes and the remaining 22 patients (27%) five or six activating KIR genes. The number of potential activating KIR/HLA interactions in patients with a KIR B/X haplotype was 0 in 12 patients (14.6%), 1 in 40 patients (48.8%), 2 in 27 patients (32.9%) and 3 in the remaining 3 patients (3.7%).
KIR genotype and rate of viral and bacterial infections/reactivations posttransplantation
In univariate analysis, patients possessing a KIR B/X genotype had a significantly lower rate of CMV infections/reactivations than individuals with a KIR A/A genotype (cumulative incidence at 1-year posttransplantation 20 ± 5% vs. 36 ± 8%, p = 0.04, Figure 1A). No effect of the KIR genotype on the rate of other viral infections/reactivations was detected (EBV 8 ± 3% vs. 8 ± 5%, p = 0.96, Figure 1B; BK virus 22 ± 5% vs. 24 ± 7%, p = 0.70, Figure 1C; Herpes simplex 14 ± 4% vs. 10 ± 5% p = 0.62, Figure 1D). The incidence of other viral infections was low: one case of parvovirus B19 (in KIR A/A genotype individual), and two cases of Varizella zoster virus infections (one in KIR A/A and one in KIR B/X genotype individual). There was no correlation between KIR genotype and the rate of bacterial infections (KIR B/X genotype 46 ± 6% vs. KIR A/A genotype 41 ± 9%, p = 0.76, Figure 1E).
Multivariate Cox models confirmed the results of the cumulative incidence analyses (Table 2). After correcting for donor and recipient serology, type of immunosuppression and antiviral prophylaxis, the relative risk of CMV infection/reactivation in individuals with a KIR B/X genotype was 0.34 (95% CI 0.15–0.76, p = 0.009, comparator KIR A/A genotype [RR 1.00]). Protection from CMV reactivation (in patients being CMV IgG positive at transplantation) was greater (RR KIR B/X genotype versus KIR A/A genotype: 0.31, p = 0.05) than protection from de novo infection (patients CMV IgG negative at transplant, RR KIR B/X genotype versus KIR A/A genotype: 0.56, p = 0.30). The degree of protection from CMV increased with the number of activating KIR genes: Compared to individuals with a KIR A/A genotype (RR 1.00), individuals with a KIR B/X genotype with 1–2 activating KIR had a relative risk of 1.03, those with 3–4 activating KIR genes a relative risk of 0.40 and those with 5–6 activating KIR genes a relative risk of 0.23. Analysis of single activating KIR genes as prognostic factors for CMV infection/reactivation revealed that KIR2DS5 had the strongest predictive power (relative risk if present 0.26, 96% CI 0.09–0.74). However, the p-value (0.015) did not meet the predefined level of significance adjusted for multiple testing. Moreover, if single activating KIR genes were included in a model in addition to KIR A/B haplotype, none of the activating KIR genes remained a statistically significant predictor. No effect of KIR haplotype or number of activating KIR genes was observed for EBV or BKV infections/reactivations in multivariate analysis.
Table 2. Multivariate analysis of risk factors for viral infection
n.a. = not available; n.s. = not significant; not significant for any endpoint: Cyclosporine versus tacrolimus, mycophenolate versus azathioprin, induction containing ATG or rituximab.
Don pos/Rec pos
Don neg/Rec neg
Don neg/Rec pos
Don pos/Rec neg
Rapamycin versus prednisone
The number of putative KIR/HLA interactions had no impact on the rate of CMV infection/reactivation. Compared to patients with a KIR A/A genotype (RR = 1.00), patients with 1 interaction had a relative risk of 0.47 (95% CI 0.18–1.25), those with two interactions a relative risk of 0.29 (95% CI 0.09–0.91). No episodes of CMV infection/reactivation were recorded in the three patients with 3 putative KIR-HLA interactions. When both the number of activating KIR and the putative KIR/HLA interactions were taken into account in a Cox model, the impact of KIR/HLA interactions was nonsignificant (p = 0.73).
Four patients in the cohort had symptomatic CMV disease (colitis in all four). Two patients had a serological high-risk constellation (donor positive/patient negative), in the other two cases the constellation was donor positive/patient positive, with one patient having received induction treatment with ATG. Intriguingly, all four patients had a KIR A/A genotype (p = 0.004 compared to individuals with a KIR B/X genotype). Two non-CMV symptomatic viral infections were recorded: one patient with a KIR B/X genotype developed Varizella zoster pneumonia, another patient with a KIR A/A genotype had an episode of pancytopenia due to parvovirus B19 infection. No episodes of posttransplant lymphoproliferative disease (PTLD) were recorded during follow-up, although one patient with a KIR A/A genotype had a history of PTLD after a previous kidney transplantation.
KIR genotype and graft rejection and function
The KIR genotype had no influence on the incidence of graft rejection episodes (both clinical and subclinical). Rejection-free survival after 12 months was 77 ± 7% in individuals with a KIR A/A genotype, compared to 78 ± 5% in individuals with a KIR B/X, p = 0.74, Figure 2A). Similarly, MDRD estimated glomerular filtration rates were not significantly different at 1, 3, 6 and 12 months posttransplantation in individuals with a KIR A/A versus a KIR B/X genotype (Figure 2B).
The key finding of this study was that—in a dose dependent manner—activating KIR of kidney allograft recipients were associated with a decreased incidence of CMV de novo infection and reactivation. By contrast, no such association with (i) the rate of other viral or nonviral infections, and (ii) transplant function and the number of clinical or subclinical rejection episodes were observed.
As for CMV, a reduction was seen for both the total number of infectious episodes—defined as documented viral replication—and for the rate of symptomatic CMV disease. Importantly, this protection was ‘dose-dependent’, i.e. the more KIR genes a given individual possessed the lower was the likelihood for CMV-related infectious episodes. Intriguingly, no similar effect was observed for other herpes viruses, although the power of our study might have been too small to detect differences. Also—and not unexpectedly—the rate of bacterial infections was uninfluenced by the number of activating KIR genes. These results are in line with data from HSCT patients and show—to our knowledge for the first time—that the recipient's KIR genotype is associated with the frequency of infectious complications (specifically CMV-related complications) after transplantation of a solid organ. The mechanism for this protective effect remains to be determined since the ligands for activating KIR have not been defined. Although extensive homology exists in the extracellular domains of inhibitory and activating KIR—suggesting that they might share ligands—functional studies have shown only a weak binding affinity of activating KIR to HLA class I molecules (13). Recent evidence has suggested that binding of activating KIR to HLA class I might depend on presentation of viral protein by the given MHC molecule (14). However, we were not able to show an influence of the number of putative activating KIR/HLA interactions on the rate of viral infections. Alternatively, the true ligands for activating KIR may not be HLA class I antigens as, in mice, a homologue of the human KIR receptor has been shown to directly engage the virally encoded protein m157, thereby conferring resistance against murine CMV (15).
In our study no effect of the recipient's KIR genotype on the incidence of graft rejection and graft function was observed. Our results, obtained in a population undergoing regular protocol biopsies to detect subclinical rejection, therefore confirm and extend a recent report which also failed to show a correlation between the recipient's KIR genotype and the incidence of graft rejection (16).
In summary, by studying a cohort of 122 kidney allograft recipients, we provide evidence for a protective effect of activating KIR on the rate of posttransplant CMV infections/reactivations. While our data are in line with recent evidence from hematopoietic stem cell transplant recipients, results of this retrospective study must be interpreted as associations that need to be validated. Unraveling the biology that confers this—to some extent virus-specific—protective effect may provide important insight into the functioning of NK cells in the context of CMV infection.
We thank Gabriela Zenhaeusern, Ineke Oehri, Bojana Durovic and Denise Bielmann for technical assistance. M. S. is supported by foundations from AstraZeneca and the University Hospital Basel (VFWAWF-Fonds). C. H. is supported by the Swiss National Science Foundation (SNF professorship grant PP00B-114850). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the article.