|Home | About | Journals | Submit | Contact Us | Français|
Natural Killer (NK) cells are important in graft versus leukemia responses following hematopoietic cell transplantation (HCT). A variety of surface receptors dictate NK cell function, including killer immunoglobulin receptor (KIR) recognition of HLA-C. Previous single center studies show that HLA-C epitopes, designated C1 and C2, were associated with allogeneic-HCT outcomes; specifically recipients homozygous for the C1 epitope (C1/C1) experienced a survival benefit. Additionally, mismatching at HLA-C was beneficial in recipients possessing at least one C2 allele, while the opposite was true for homozygous C1 (C1/C1) recipients where HLA-C mismatching resulted in worse outcomes. In this analysis we aimed to validate these findings in a large multicenter study. We also set out to determine whether surface expression of recipient HLA-C, determined by polymorphism in a microRNA (miR-148a/b) binding site within the 3′-region of the HLA-C transcript, was associated with transplant outcomes. In this large, registry cohort, we were unable to confirm the prior findings regarding recipient HLA-C epitope status and outcome. Additionally, HLA-C surface expression (i.e., surface density), as predicted by the miR148a/b binding SNP, was also not with associated transplant outcomes. Collectively, neither HLA-C surface expression, as determined by miR148a/b, nor recipient HLA-C epitopes (C1, C2) are associated allo-HCT outcomes.
Natural killer (NK) cells play a critical role in graft versus leukemia (GVL) responses and have previously been linked to allogeneic hematopoietic cell transplant (allo-HCT) outcomes.1–5 NK cells are CD3−CD56+ and are the first donor lymphocytes to recover following allo-HCT. NK cells display a number of surface activating and inhibitory receptors that dictate function upon engagement with a potential target cell. The killer immunoglobulin receptors (KIR) are a key family made up of 15 distinct NK receptors, some with known specificity. Inhibitory KIR recognize polymorphic epitopes of HLA class I (also known as KIR ligands) and binding prevents NK activation. Examples include the inhibitory receptor, KIR2DL1 which recognizes HLA-C alleles that contain an Lysine (Lys) at position 80 (i.e., HLA-C2), while KIR2DL2 and KIR2DL3 bind to HLA-C that contain a Asparagine (Asn) epitope at position 80 (i.e., HLA-C1)6, 7. Despite the considerable genetic variation in HLA-C, the amino acid identity at position 80 dichotomizes HLA-C alleles into either C1 or C2, and these epitopes correspond to which inhibitory KIR recognizes HLA-C6. The binding of inhibitory KIR with its ligand (HLA class I) prevents NK cell activation (cytotoxicity and cytokine production); while the lack of this interaction tips the balance in favor of NK activation8.
Based on the above, recipient HLA-C (i.e., C1 or C2) may drive donor NK cell functionality following HCT and this may be associated with clinical outcomes9–11. The source of circulating NK cells early after HCT is both from the NK cells contained within the graft, as well as stem cell-derived NK cells, with some studies providing convincing evidence that the latter rapidly predominate12. For the NK cells that differentiate from stem cells, KIR acquisition is a late process and for unknown reasons, NK cells expressing KIR that recognize C1 (KIR2DL2 and KIR2DL3) emerge earlier than those that express KIR that recognize C2 (KIR2DL1)10. This may result in higher numbers of functional NK cells in C1/C1 recipients, due to a process called “NK cell licensing.” Briefly, licensing describes the acquisition of NK functionality upon interaction of inhibitory receptors with their MHC class I ligands13, 14. Fischer et al previously reported that C1/C1 recipients had better overall survival than C1/C2 and C2/C2 recipients.9, 10 Furthermore, they showed that HLA-C allele mismatching was beneficial in recipients that were heterozygous (C1/C2) or homozygous for C2 (C2/C2), while HLA-C mismatching was disadvantageous to patients homozygous for C1 (C1/C1)9. Other large registry studies have used both recipient HLA type and donor KIR gene content to show improved disease free survival and relapse protection in AML patients when the donor expressed a particular KIR gene(s) and the recipient lacked the ligand (HLA-B or -C)15, 16. However, clinical implementation of this approach is potentially cumbersome and requires real time typing of KIR genes during donor selection. While this approach is currently being tested for feasibility (NCT01288222), based on the above single center data, we set out to study whether recipient HLA C was sufficient to predict outcomes using a large registry cohort.
While HLA-C has been widely accepted as a transplantation antigen and a ligand for KIR receptors, little consideration has been given to whether or not variations in HLA-C surface expression (i.e., the surface density) impact allo-HCT outcomes. The regulation of HLA-C surface density is controlled, in part, by post-transcriptional regulation via microRNA (miRNA) binding17, 18. miRNAs are a class of non-coding RNAs that bind to the 3′-untranslanted region (3′UTR) of coding transcripts and consequentially repress translation, cleave, or destabilize the respective mRNA transcript19. Stable binding between miRNA and the 3′UTR binding site is associated with a decreased mRNA level, leading to reduced expression of the associated protein. Recent evidence points to an essential role of HLA-C in the progression of HIV-1. The HIV-1 Nef protein down-regulates HLA-A and -B molecules, causing the infected cells to rely on HLA-C to maintain class I expression for antigen presentation and avoidance of NK cell-mediated killing20. A single nucleotide polymorphism (SNP) at position 263 of the HLA-C mRNA 3′-UTR, designated rs67384697, affects the binding stability of miR-148a/b and affects viral loads in the setting of HIV-117,21. Since the surface density of HLA-C can be predicted based on polymorphisms at position 263 of the HLA-C 3′UTR, we tested whether recipient polymorphism at rs67384697 is also associated with allo-HCT outcomes. We hypothesize that recipient HLA-C alleles that contained the miR-148a/b binding domain would have lower surface density, leaving the inhibitory KIR disengaged and enhance GVL response, leading to lower relapse rates and improved DFS.
In this study, we sought to determine whether recipient HLA-C epitope expression (C1 and C2) and the surface density of HLA-C, as determined by miR-148a/b binding SNP, are associated with allo-HCT outcomes using registry data from the Center for International Blood and Marrow Transplant Research (CIBMTR).
The CIBMTR registry includes a voluntary working group of more than 450 transplantation centers worldwide that contribute detailed data on consecutive allogeneic and autologous hematopoietic cell transplantations to a statistical center at the Medical College of Wisconsin in Milwaukee and the National Marrow Donor Program (NMDP) Coordinating Center in Minneapolis. Participating centers are required to report all transplants consecutively; patients are followed longitudinally and compliance is monitored by on-site audits. Computerized checks for discrepancies, physicians’ review of submitted data and on-site audits of participating centers ensure data quality. Observational studies conducted by the CIBMTR are performed in compliance with all applicable federal regulations pertaining to the protection of human research participants. Protected Health Information used in the performance of such research is collected and maintained in CIBMTR’s capacity as a Public Health Authority under the HIPAA Privacy Rule22.
The study population consisted of patients receiving their first bone marrow or peripheral blood stem cell unrelated donor transplant for acute myelogenous leukemia (AML), acute lymphoblastic leukemia (ALL), chronic lymphocytic leukemia (CML) or myelodysplastic syndrome (MDS) facilitated by the National Marrow Donor Program (NMDP) and reported to the CIBMTR between 1988 and 2009. All HLA-typing was verified using high resolution, DNA-based methods via PCR and sequence-specific oligonucleotide probes (PCR-SSOP) as previously described23. Early stage disease was defined as AML or ALL in first complete remission, CML in first chronic phase, and MDS subtype refractory anemia. Intermediate stage disease was defined as AML or ALL in second or subsequent complete remission, and CML in accelerated phase or second chronic phase. Advanced phase disease was defined as AML in first or higher relapse or primary induction failure, CML in blast phase and MDS subtypes refractory anemia with excess blasts or in transformation.
All surviving recipients included in this analysis were retrospectively contacted and provided informed consent for participation in the NMDP research program. Research was approved and conducted under the supervision of the NMDP Institutional Review Board. A modeling process was used as previously described to adjust for any bias introduced by exclusion of non-consenting survivors24.
Histocompatibility testing of patient-donor pairs was performed and pairs were classified according the degree of HLA match (8/8 vs. 7/8). If HLA mismatched, it was determined if the mismatch occurred at HLA-C, and the recipient’s HLA-C encoded epitope expression status was identified (i.e., C1/C1 vs. C2/-). This grouping produces four 7/8-matched patient subsets (Table 1) that were defined as “poor risk” or “good risk” based on prior study findings. More specifically, HLA-C mismatching was considered beneficial in patients possessing at least one C2 allele, but detrimental in patients homozygous for C1.9, 10
HLA-C surface expression was predicted by polymorphisms in the 263 position of the HLA-C 3′UTR (Table 2). The site of this polymorphism is encompassed within the miR-148a and miR-148b binding site and affects the relative binding stability of miR-148a/b, in turn leading to down regulation of the mRNA transcript17, 18. An insertion at this site is associated with high affinity binding and consequential low HLA-C expression, while a deletion is associated with low affinity binding and high HLA-C surface expression (Table 2)17.
The primary outcome is overall survival (OS), defined as time to death from any cause. The secondary outcomes include disease relapse, acute graft versus host disease (GVHD, grades II–IV and III–IV), chronic GVHD, disease-free survival (DFS), time to relapse, or death from any cause other than relapse: treatment-related mortality (TRM, death in continuous complete remission of primary disease).
Descriptive statistics include medians and ranges for continuous variables and frequencies for categorical variables. Chi-square or Fisher’s exact tests were used to compare frequencies for categorical variables, and t-tests were used to compare means for continuous variables. Univariate probabilities for OS and DFS were calculated using the Kaplan-Meier estimator26. Comparison of survival curves was made using the log-rank test. The cumulative incidences of acute GVHD II–IV, acute GVHD III–IV and TRM were calculated using a Taylor series approximation to estimate the variance27. Multivariate models were built for OS, DFS, Relapse, TRM, aGVHD (Grades II–IV), aGVHD (Grades III–IV) and cGVHD using the Cox proportional hazards model. All the clinical variables were tested for the affirmation of the proportional hazards assumption (p<0.01). Factors violating the proportional hazards assumption were adjusted through stratification. Then a stepwise model selection procedure was used to develop a multivariate model for each outcome with a threshold of 0.05 for both entry and retention in the model. The association of HLA-C matching and recipient C1 and C2 epitope status with clinical outcomes was tested with adjustments for the selected clinical variables. Interactions between HLA-C matching and recipient C1/C2 epitope status and the adjusted clinical variables were tested and no significant interactions were detected. SAS version 9.3 (SAS Institute, Cary, NC) was used for all the analyses.
A cohort of 7,327 patients undergoing a first allo-HCT was included in this analysis. As shown in Table 3, 70% were 8/8 HLA matched, while the remainder (30%), received transplantation with a 7/8 HLA matched donor. Approximately 40% received peripheral blood stem cells (PBSC) and ~80% of patients underwent myeloablative conditioning. As described in the methods, HLA mismatched patients were divided into four risk groups based on HLA C mismatching and recipient C1 and C2 epitope status (Table 1). These groups did not differ significantly with respect to disease or transplant conditioning intensity, disease risk status, but patients receiving HLA-C mismatched grafts were significantly younger (Table 3 and not shown).
Consistent with a number of prior reports, multivariable analysis showed that, HLA mismatched recipients (7/8) had inferior outcomes when compared to those receiving fully HLA matched grafts, with increased TRM and aGVHD, reduced DFS and OS, and no impact on relapse (Table 4)25, 26. When examining the subgroups of HLA mismatched recipients (as shown in Table 1), we found that similar to prior study findings9, 10, HLA-C mismatched (7/8) patients who were homozygous for C1 experienced a higher relative risk for aGVHD and TRM, resulting in inferior DFS and OS when compared to fully matched recipients (Table 4). However, contrary to these previous reports, the present study did not show significantly improved outcomes for the HLA-C (7/8) mismatched recipients who expressed a C2 ligand (Table 4). This group had equally inferior DFS and OS as other mismatched groups, again due to an increase in aGVHD and TRM (Figure 1A–D).
Since the prior studies9, 10 were conducted only in patients with acute and chronic myeloid leukemia who received PBSCs, we performed a planned subset analysis restricting only to these diseases and this stem cell source. This analysis did not reveal any differences in outcomes (not shown). Based on prior studies15, 16 suggesting the presence of at least one C1 allele (C1/C1, C1/C2), rather than one or more C2 alleles (C1/C2, C2/C2) dictate transplant outcome, we also analyzed the three HLA C genotypes separately, but this did not reveal any differences in outcomes (not shown). Therefore, these results suggest that it is the single mismatched HLA status that accounts for the worse outcomes, not consideration of the recipient C1/C2 status in the context of HLA-C match status.
To determine whether a novel algorithm to predict HLA-C surface density is associated with outcomes following transplant, we retrospectively classified recipients according to a single nucleotide polymorphism (SNP) at position 263 of the HLA-C 3′-untranslated region (rs67384697) which is the binding site of miR-148a/b17. We elected to restrict this analysis to patients with AML due to the well-established association between NK cells and AML27, 28. This subset included 2,910 patients from the larger study cohort described above (Table 3). Patients were divided into three HLA-C expression groups (i.e. low/low (n=1012), low/high (n=1412), high/high (n=486)), predicted by a SNP insertion or deletion at rs67384697 (described in the methods and Table 2). In brief, these three groups did not differ based on age at transplantation, race, sex, performance, disease status, HLA matching (8/8 vs. 7/8), or stem cell source (Table 3). However, these expression groups did vary by HLA-C ligand status: the high/high group predominately included C1 homozygous patients (67%), the high/low group included 61% C1/C2 HLA-C ligand patients, and the low/low group was more evenly distributed between C2/C2 (43%) and C1/C2 (40%). Using multivariable analysis, controlling for relevant clinical factors, there were no significant differences in OS, DFS, relapse, NRM, or aGVHD III–IV (Table 5). In an additional exploratory analysis, the cohort was further divided into a pediatric subset (n=397), a homozygous C2 group (n=401), and a group restricted to patients with early to intermediate stage AML who received myeloablative conditioning and T replete grafts (n= 1417). Similarly, no significant associations were evident (not shown). Further, a subset of 8/8 HLA-matched patients were divided into nine subgroups based on HLA-C ligand status and HLA-C expression to determine if the interplay of both factors have an effect on OS. However, there were no significant differences between the groups in OS (data not shown). These results do not support our hypothesis that HLA-C expression, as predicted by the miR-148a/b binding SNP, is associated with outcomes after allo-HCT.
In an attempt to improve outcomes in allo-HCT, investigators have sought to understand the effect of HLA-C beyond simply donor-recipient HLA matching. We first set out to determine the effect of recipient HLA-C epitope status (C1/C1 vs C1/C2 vs C2/C2) on outcomes following allo-HCT. Based on prior findings we had two hypotheses: 1) that C1/x patients would have better OS, due to earlier NK cell licensing and reduced relapse and 2) that HLA-C mismatching would be beneficial in patients that possessed at least one C2 allele9. We also tried to understand whether recipient HLA C surface density was associated with transplant outcomes. Thus, we hypothesized that recipient HLA-C alleles that contained the miR-148a/b binding domain would have lower surface density, which would leave inhibitory KIR disengaged and enhance GVL responses, resulting in lower relapse rates and improved DFS.
The data presented in this study do not support previous reports showing an influence of recipient HLA-C epitopes on transplant outcomes9, 10. First, there was no observed survival benefit in any of the HLA C groups (C1/C1, C1/C2, C2/C2). In addition, we show that for recipients that were 7/8 HLA matched, HLA-C matching was neither beneficial nor detrimental, regardless of HLA-C encoded KIR ligand status (C1 vs. C2). When the analysis was restricted to AML patients we also were unable to discern differences in transplant outcomes (not shown). In contrast, Cooley et al15 have reported a protective effect of C1/x specifically for transplants in which donors express at least two KIR B motifs. As our study did not test for KIR, the population likely includes donors with various KIR gene expression, probably diluting the beneficial C1/x effect.
Regarding our hypothesis about HLA C surface expression, we were unable to find differences in allo-HCT outcomes based on the predicted recipient HLA-C surface levels. Ultimately, we conclude that in silico analysis of HLA-C surface levels using the approach here is not associated with transplant outcomes. Contrasting results have been reported for HLA-DP where a SNP at rs9277534 controls surface expression and is associated with aGVHD29. In HLA-DPB1 mismatched transplants, recipients with a SNP encoding high HLA-DPB1 expression (rs9277534G) have higher rates of GVHD29. This is presumably explained by high surface expression of HLA-DPB1 acting as a more visible target for alloreactive, minor antigen (miAg)-specific donor CD4+ T cells which drive acute GVHD reactions. Similar to this, the rs67384697 SNP, studied in this report, influences HLA-C surface expression in HIV and is strongly associated with the control of viremia17, 30. The variable HLA-C surface expression is mediated by miR-148a or miR-148b, which have the same binding site (at rs67384697)17,30. Stable miRNA binding leads to degradation of HLA-C mRNA transcripts and, thus, reduces the number of targets for viral specific CD8+ T cells. It is reasonable to assume that these same base pair variations control HLA-C levels in the transplant setting. However, we did not observe any associations between transplant outcomes and recipient HLA-C expression levels, based on the above miRNA binding site. One possible explanation for these seemingly discrepant findings may be the fundamental differences between T cell and NK cell target recognition methods. More specifically, while T cells rely solely on the T-cell receptor for recognition of viral and/or miAg peptides presented in the context of HLA, NK cells interact with their targets through a number of different receptor/ligand pairs, including KIR/HLA class I. Importantly, these receptor/ligand pairs are likely redundant, thus, NK cells may be less sensitive to perturbations in a single receptor-ligand pair, in this case, differences in the level of HLA class I molecules, such as HLA-C. Additionally, binding of miRNA 148a/b to the 3′UTR and the associated modulation of HLA-C surface expression has been previously confirmed in smaller studies (n=~300), however we did not reconfirm this association in our cohort and this remains a potential weakness31.
While the strength of this study is the large patient numbers provided by the CIBMTR, it is also a weakness since there are many nuanced variations in transplant protocols and the supportive care of patients at various centers that might have confounded the results from this multi-center study and any subgroup stratification based on geographic origin would result in inadequate statistical power. Importantly, there is moderate overlap in the subset of AML patients of the current study and two prior studies that did find an improvement in DFS and reduced relapse in patients that express C1/x15, 16. This highlights that the highly heterogenic study population must be considered in the interpretation and gravity of these results.
In summary, the mechanisms contributing to GVHD and relapse are more complex than simple HLA-mismatching and may involve base pair variations within HLA-C molecules. Our data shows that neither HLA-C encoded KIR ligand epitopes (C1, C2) nor HLA-C surface expression level, as predicted by miRNA binding polymorphisms, are associated with OS, DFS, relapse, or GVHD in allo-HCT.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.