Cytomegalovirus (CMV) infection remains a major source of morbidity and mortality in solid organ transplant recipients. Killer immunoglobulin-like receptors (KIR) are genetically polymorphic natural killer (NK) cell receptors important in antiviral responses. A retrospective, single-center cohort study was performed to study the interaction of KIR genotype and primary control of CMV infection after transplantation. Time to first CMV viremia was determined for a cohort of 531 CMV serology donor positive/recipient negative solid organ transplant recipients. Of the KIR genes, KIR2DL3 and KIR2DS2 were most strongly associated with time to CMV viremia in random survival forest analysis. As KIR2DL3 and KIR2DS2 both interact with HLA-C1, these interactions were evaluated. Seventy-six recipients were found to be positive for both KIR2DL3 and KIR2DS2 and expressed only HLA-C1 antigens in both recipient and donor. These patients had a substantially reduced hazard of CMV viremia in the first year after solid organ transplantation (hazard ratio 0.44, 95% CI 0.27–0.72, p = 0.0012). In KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients who received an organ from a non-C1/1 donor, this protective effect was not observed. These results improve our understanding of human NK cell function in primary CMV infection after transplant.
donor positive/recipient negative
electronic medical record
killer immunoglobulin-like receptor
solid organ transplantation
sequence-specific oligonucleotide probe
Outcomes after solid organ transplantation (SOT) continue to improve. However, acute and chronic forms of rejection remain common and almost all solid organ transplant recipients remain on immunosuppressive agents. The main risks associated with these treatments include increased malignancy rates as well as risk for a variety of opportunistic infections . Transplant recipients differ from other immunocompromised patients as they are at risk for donor-derived infections in addition to their immunosuppressed state. The most common clinically relevant donor-derived infection is cytomegalovirus (CMV). CMV infection may cause a myriad of clinical symptoms and syndromes, including asymptomatic viremia, CMV syndrome and tissue invasive disease. In addition to direct effects, CMV infection after SOT is a risk factor for graft rejection, as well as for other infections . In SOT, those recipients who are seronegative for CMV, but receive a CMV-positive allograft are at the highest risk for disease . Immunologically, these infections are of interest as they represent a primary infection with a defined time of exposure, in which the innate immune system is thought to play an important role.
Natural killer (NK) cells are a part of the innate immune response to viral infections [3, 4]. Specifically, both murine models as well as studies in patients suggest that NK cells play a central role in CMV infections [5, 6]. Whether an NK cell kills its target cell depends on a complex interaction between inhibitory or activating receptors on NK cells and self-antigens on target cells. Of the genes encoding NK cell receptors, the killer immunoglobulin-like receptor (KIR) family is the most polymorphic. The interaction between KIR and HLA plays a major role in determining NK-mediated immunity toward pathogens . Of the various HLA antigens that serve as ligands for KIR, HLA-C seem to play the most important role in the regulation of NK cells, and all individuals carry these alleles . Interestingly, many of the widely used immunosuppressive agents such as tacrolimus and mycophenolic acid relatively preserve NK innate immune effector cells and their cytotoxic potential . This suggests that the study of NK cell immunology in solid organ transplant recipients may be of interest. The relevance of KIR and their interaction with HLA ligands has been implicated in influencing the clinical outcomes of CMV infection postsolid organ and hematopoietic stem cell transplantation [10, 11]. In this study, we focus on the impact of KIR and their interaction with HLA ligands on the risk of CMV viremia after transplantation in the highest risk donor/recipient serology combination (D+/R−).
Materials and Methods
All CMV serology donor positive/recipient negative (D+/R−) recipients who underwent SOT between January 3, 1999 and January 6, 2009 were included in this retrospective cohort study. The study cohort included kidney, liver, lung, heart, pancreas, intestine and combined organ transplants as detailed in Table 1. For sequential transplantation, only the first transplant was evaluated. This study was approved by the Institutional Review Board of the Cleveland Clinic (CCF IRB#09-528).
|Age at transplant—year median (range)||51 (1–76)|
|Transplanted before 2005||173 (32.6)|
|Organ (category for analysis)|
|Other or combined1||37 (7.0)|
|Basiliximab and thymoglobulin||4 (0.8)|
|Rejection in first year after transplant||196 (36.9)|
|≤3 months duration||339 (63.8)|
|>3 months duration||192 (36.1)|
|CMV viremia in first year after transplant||232 (43.7)|
|Time to CMV—days median (range)2||107.5 (1–360)|
For each patient, the following variables were collected from the electronic medical record (EMR): age at the time of transplantation, gender, race/ethnicity, organ transplanted, date of transplant, days from transplant to first documented CMV viremia, days from transplant to death, days from transplant to first rejection episode, type of induction, use of CMV prophylaxis and the duration of CMV prophylaxis. If more than one organ was transplanted nonsimultaneously, the date of the first CMV D+/R− transplant was used. If patients were retransplanted during the study period, only the first episode was analyzed. Organ categories were used for multivariable analyses. Patients with combined or sequential transplants or with isolated pancreas or small bowel transplants were assigned to categories based on a priori CMV risk as shown in Table 1. Date of death was determined by combined review of the transplant database, the EMR and the Social Security Death Index. All biopsy confirmed rejection episodes were included regardless of grading. For analysis purposes, in patients with CMV viremia, only those episodes of rejection which occurred prior to CMV viremia were included in multivariable models. Regarding CMV prophylaxis, patients were dichotomized based on whether or not they had received antiviral prophylaxis with in vitro anti-CMV activity. Those on anti-CMV prophylaxis were further subdivided by duration of prophylaxis.
CMV DNA measurement
CMV viremia was defined as any level positive CMV DNA in peripheral blood. The quantitative Digene hybrid capture assay (Digene Corp., Gaithersburg, MD) was used during the entire study period. Results are expressed as copies per milliliter of whole blood. All CMV D+/R− recipients were monitored every 1–2 weeks in the first year after transplantation for the development of CMV viremia since January 1, 2005. Prior to 2005, monitoring for CMV viremia was performed as clinically indicated. For this reason, era of transplant was included as a covariable in all multivariate statistical analyses. In either era, patients who presented with signs and symptoms of CMV infection to clinic or during hospitalization were tested for CMV viremia. All CMV DNA testing was performed in a central virology laboratory.
KIR and HLA genotyping
The presence or absence of 16 KIR genes (KIR2DL5A and KIR2DL5B were considered together) was determined in the recipients using commercial reverse Polymerase Chain Reaction Sequence-Specific Oligonucleotide Probe (SSOP) reagents (One Lambda, Inc., Canoga Park, CA) and/or Sequence Specific Primers reagents (Life Technologies, Carlsbad, CA). Recipients and donors were genotyped for HLA-C prospectively during pretransplant evaluation using SSOP (One Lambda, Inc.). Genotypes were retrospectively reviewed to determine the presence/absence of KIR ligands C1 and C2.
KIR haplotype group assignment
Detection of at least one of the KIR B haplotype-defining loci (KIR2DL5, KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS5 or KIR3DS1) in an individual predicted that the genotype contained at least one B haplotype. Such individuals were assigned the genotype designation B/x. Since haplotype A may not be readily distinguished in presence of the longer haplotype B based on gene content, samples in which KIR2DS4 was the only detected activating KIR gene were assigned the genotype A/A. The recipients were also stratified according to the number of activating KIR (1 through 6). KIR genotypes were further classified on the basis of the presence and absence of centromeric and telomeric KIR gene clusters and KIR B-content score of (0 through 4) were analyzed as a categorical variable as proposed by Cooley et al .
The method of Random Survival Forests was used to identify the variables most strongly associated with time from transplant to first episode of CMV viremia . All genetic variables and clinical variables were considered in the building of the Random Survival Forests. Models were fit using the R package randomSurvivalForest . Random forests were run with 1000 trees and default splitting rules. The variable importance measure, which was calculated using permutation, was used to rank the strength of association of variables with time to CMV detection. Cox proportional hazards models were also fit using SAS/STAT(R) 9.2 (SAS Institute Inc., Cary, NC) to provide semi-parametric estimates of risk associations and aid in interpretation.
For the Cox proportional hazards model, a combined KIR/HLA term was included. This term was based on previously reported biologic pathways of KIR signaling. For the clinical variables, backwards selection was employed with an alpha of 0.1 required to stay in the model. In addition, the following clinical variables were forced in the model based on previously reported associations: rejection and CMV prophylaxis (both included as time-dependent variables). Death and graft loss were treated as competing risks in this model.
In the study period, 531 CMV D+/R− SOTs were performed (Table 1). Twenty-nine patients received no anti-CMV prophylaxis, 339 patients received 3 months or less, and 192 patients received more than 3 months of prophylaxis. In the first year after transplantation, 43.7% of patients developed CMV viremia. In patients with CMV viremia, the median time to viremia was 107.5 days. We first tested the univariate association with KIR gene combinations that have been previously associated with risk for CMV infection. No association between time to CMV viremia and KIR haplotype (AA vs. Bx), number of activating KIR receptors, centromeric or telomeric KIR gene content or number of “missing ligands” was found (data not shown).
We then employed a random survival forest analysis to obtain a ranking of factors associated with time to CMV viremia. In this model, all clinical variables were included, as well as individual KIR genes and HLA ligands (Figure 1). Of the KIR genes, KIR2DL3 and KIR2DS2 were most strongly associated with time to viremia. Based on this analysis, we explored the interactions of KIR2DL3. Although KIR2DL2 and KIR2DL3 have separate names, they are known to be allelic . Both KIR2DL3 and KIR2DL2 are inhibitory receptors for HLA-C group 1 (C1) antigens that have asparagine at position 80 [8, 15]. The interaction between KIR2DL2 and HLA-C1 has been reported to result in a stronger inhibitory signal as compared to the KIR2DL3/HLA-C1 interaction [7, 8]. KIR2DS2 is the activating receptor for HLA-C1 antigens, and is in linkage disequilibrium (LD) with KIR2DL2 (Table 2).
Given the LD between these genes, we dichotomized the groups into patients who had both KIR2DL3 and KIR2DS2 present and all other patients. Out of 531 recipients, 241 (45%) patients were positive for both KIR2DL3 and KIR2DS2. We then evaluated the role of HLA-C antigens. As HLA-C antigens are the primary target for KIR2DL3 and KIR2DS2, we further divided our population by recipient HLA-C corresponding ligands.
Seventy-six (14%) of recipients were found to be positive for both KIR2DL3 and KIR2DS2 and only had HLA-C1 antigens in both recipient and donor. This genetic combination would theoretically be associated with the greatest protection against CMV viremia. These 76 recipients were indeed found to be at a significantly decreased hazard of CMV viremia, as shown in the cumulative incidence plot in Figure 2A (log rank p < 0.001). In multivariate Cox proportional analysis, a hazard ratio of 0.39 was found (95% CI 0.24–0.61, p < 0.0001), when correcting for clinical variables (Table 3). In a stratified analysis, the same trend toward protection against CMV viremia was seen for KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients who received an organ which was HLA-C1/1 for lung, heart, kidney and liver transplantation (Figure S1).
|Variable||HR (95% CI)||p|
|KIR2DL3/2DS2, R C1/1, D C1/11||0.44 (0.27–0.72)||0.0012|
|CMV prophylaxis2||0.69 (0.39–1.23)||0.20|
|Basiliximab induction||1.35 (0.96–1.91)||0.058|
We also investigated the respective roles of the donor and recipient HLA-C antigens. When comparing time to CMV viremia in KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients who received an organ that was HLA-C1/1 the hazard was significantly decreased versus KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients who did not (Figure 2B, log rank p < 0.01). Of note, a similar benefit of receiving an HLA-C1/1 organ was not observed in recipients who were not KIR2DL3+/KIR2DS2+/HLA-C1/1 (data not shown). Similarly, KIR2DL3+/KIR2DS2+ recipients who received an HLA-C1/1 organ who expressed any HLA-C2 antigens were at increased hazard of CMV viremia, as compared to KIR2DL3+/KIR2DS2+ recipients who received an HLA-C1/1 organ who expressed only HLA-C1 antigens (Figure 2C, log rank p < 0.001). To rule out that this observation was not simply due to HLA-C matching, we compared the influence of C1/1 matching on time to CMV viremia in the subset of patients who are not KIR2DL3+/KIR2DS2+ recipients. In this subset, no effect was seen (data not shown). As KIR2DS2 and KIR2DL2 are in LD, we sought to evaluate the likelihood that the protective effect observed in KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients who received an organ which was HLA-C1/1 was driven by KIR2DL2 rather than KIR2DS2. Six of 76 KIR2DL3/KIR2DS2/HLA-C1/1 who received an HLA-C1/1 organ were homozygous for KIR2DL3 and did not carry KIR2DL2. In these six patients, none had CMV viremia in the first year after transplant, suggesting that it is indeed the combination of KIR2DL3/KIR2DS2 and not KIR2DL3/KIR2DL2, which is associated with the protective effect.
Our results suggest an essential role for NK cell immunity in the innate immune response to primary CMV infection. In agreement with current understanding of KIR/HLA interactions, we found that HLA-C/KIR interactions were associated with the control of primary CMV infection as measured by the time to CMV viremia in CMV D+/R− solid organ transplant recipients. Not surprising, the presence of the activating receptor (KIR2DS2) in combination with a weaker inhibitory receptor (KIR2DL3 on at least one allele) was associated with protection against CMV viremia. However, this protective effect was only present if neither the donor nor the recipient expressed any HLA-C2 molecules. This finding could be explained at least in part by the fact that in the clinical context of CMV D+/R−, CMV infection is donor derived. This finding is also consistent with a previous report by Bohl et al  that the lack of specific KIR ligands in the donor genotype was associated with sustained BK viremia after kidney transplantation. The hypothesis of NK cell licensing through activating receptors may further explain these findings. Licensing is a proposed mechanism to account for the apparent paradox that while HLA antigens are the ligands for KIR receptors, both MHC and KIR gene content are highly polymorphic and are inherited independently on separate chromosomes. The licensing model suggests that two types of peripheral NK cells exist side by side; licensed and unlicensed. Licensed NK cells are fully competent, but have tolerance to self by inhibition through the same receptors that allowed for their licensing. If the expression of the ligands for these receptors is altered, for instance in the case of viral infection or malignant conversion, licensed NK cells become directly cytotoxic. Unlicensed NK cells, on the other hand, are hyporesponsive to stimulation and are less able to kill virally infected cells . Licensing of NK cells may occur through either inhibitory, activating or both types of receptors.
In KIR2DL3+/KIR2DS2+/HLA-C1/1 recipients, KIR2DS2 expressing NK cells would be predicted to be licensed. On the other hand, in patients with absent (HLA-C2/2) or diminished (HLA-C1/2) HLA-C1 antigen expression, these cells may be unlicensed and functionally less competent. This would also explain why donor cells—the carriers of CMV into the recipient—also must have the same HLA C1 antigen expression for optimal cytotoxicity toward CMV infected cells.
Several studies have evaluated the association between KIR genotype and risk of CMV infection after SOT [6, 10, 18]. Of note, in these studies, patients with all recipient and donor CMV serology combinations were included. The number of activating receptors, telomeric KIR gene content and number of “missing ligands” were found to be associated with risk for CMV infection [6, 10, 18]. We believe that the main reasons that our conclusions differ are threefold. First and most importantly, we analyzed CMV D+/R− recipients only, assuring immunologic homogeneity. These recipients represent a unique quasi-experimental population in whom the first CMV infection after transplantation represents primary infection. Second, our sample size was substantially larger, 531 patients as compared to 339, 196 and 122 patients. Third, we employed random survival forest analysis, a sophisticated multivariate time-to-event analysis tool, which does not rely on the same assumptions as Cox proportional hazards analysis. All potentially relevant variables, both clinical and genetic may be entered into a random survival forest analysis, without concerns of multiple comparisons or proportionality assumptions. This is a powerful tool for time-to-event analyses that has been used successfully in other time-to-event analyses . This is a substantially different approach from the more commonly used approach of performing several univariate analyses followed by a multivariate model. In such an approach, it is difficult to establish how variables interact to influence the main outcome. The ability to allow for these unrestricted interactions is a major benefit of random forest analysis.
Recently, Behrendt et al  reported an intriguing finding in HLA-matched allogeneic hematopoietic stem cell transplant recipients. In that study, activating KIR2DS2 was independently associated with decreased likelihood of CMV reactivation only when the corresponding ligand was absent from the genotype shared by donor and recipient, an effect attributed by the investigators to non-licensed NK cell. The data presented here as well as the data presented by Behrendt et al  point toward the KIR-HLA-C pathway as a key pathway in antiviral risk. In contrast to Behrendt et al , who studied stem cell recipients who were CMV seropositive in donor and/or recipient—27% of patients were CMV D−/R+, the comparable high-risk group in the stem cell recipient population—we studied only high-risk solid organ transplant recipients who were CMV seronegative prior to transplant and who received an organ from a CMV seropositive donor. This may account for the differences in our findings.
Combined with the current understanding of NK cell biology that has been accumulated mostly through ground-breaking murine studies, our findings support the model of HLA-guided licensing of NK cells and add some insights on how NK cells might function during viral infections. The innate immune system has been previously linked to posttransplant CMV infection. Kang et al  reported a large study of over 700 liver transplant recipients in whom an association between a polymorphism in Toll-like receptor-2 (TLR-2) and risk of CMV infection was found. This is particularly interesting as TLR-2 is one of the pattern recognition receptors involved in the immune responses of NK cells to CMV .
Our study has several limitations. It is a retrospective study that was performed at a single center. Both the measurement of co-variables that may confound the outcome, as well as the outcome measure itself (first CMV viremia) may have suffered from the retrospective design. However, we obtained all variables through a combination of chart review, transplant database query and automated queries of the EMR to minimize any bias. Furthermore, the overwhelming majority of posttransplant CMV DNA determinations are performed in our reference laboratory, even if the patient is physically not in our region, in which case the blood is mailed in. Another potential limitation of this study is that we studied a mixture of transplanted organs. This does increase the external validity of our findings, but may be considered a threat to the internal validity. However, when performing a stratified analysis by single organ group, the same trend toward protection against CMV viremia was seen in all organ groups.
In summary, we propose a synergistic model in which a combination of an activating KIR, a weak inhibitory KIR and homozygosity for the corresponding ligand (KIR2DL3+/KIR2DS2+/HLA-C1/1) in solid organ transplant recipients is associated with decreased risk for CMV infection if they receive an organ that only expresses HLA-C1 antigens. This is an illustration for the role of NK cell immunity in the control of primary CMV infection.
No external funding was received for this study.
The authors of this manuscript have conflicts of interest to disclose as described by the American Journal of Transplantation. D. van Duin has served on the speaker's bureau for Astellas, on a data safety monitoring board for Pfizer, and has received research funding from Steris, Inc.