Decreased iKIR-HLA C Pair Confers Worse Clinical Outcomes for Patients With Myeloid Disease Receiving Antithymocyte Globulin-Based Haploidentical Hematopoietic Stem Cell Transplantation

Hematopoietic stem cell transplantation (HSCT) is a curative therapy for patients with malignant hematologic diseases. Killer immunoglobin-like receptor (KIR) expressed by NK cells is closely associated with the transplant outcomes, and it has been widely explored and debated for a few decades. Recently published studies have revealed that inhibitory KIRs (iKIRs) are educated by their cognate human lymphocyte antigen (HLA) ligands, and that decreased iKIR-HLA pairs post-transplantation may indicate a reduced NK cell function and impaired control of the primary disease. However, this theory still needs to be validated by additional clinical studies. Here we conducted a retrospective analysis of 246 patients who received haploidentical (haplo)-HSCT at our treatment center between January 2015 and June 2018. Our data suggests that decreased iKIR-HLA C pair post-HSCT correlated with a significantly higher risk of relapse [hazard risk (HR) = 2.95, p = 0.019] and reduced overall survival (OS) (HR = 3.74, p = 0.001) and disease-free survival (DFS) (HR = 4.05, p = 0.0004) in patients with myeloid disease. In conclusion, decreased iKIR-HLA C pair should be avoided during anti-thymocyte globulin (ATG)-based haplo-HSCT, especially for patients with myeloid disease.


INTRODUCTION
Natural killer (NK) cells act as the first line of defense in the immune system. They can rapidly recognize autologous cells and eliminate non-self-components without prior antigen presentation (1,2). Multiple receptors expressed on NK cells have been implicated in the regulation of their function, with a particular focus on the activities of killer immunoglobin-like receptors (KIRs).
It is well accepted that KIR genes and receptors can be divided into inhibitory and activating functions based on their diverse activities (3). Inhibitory KIRs (iKIRs) bind human lymphocyte antigen (HLA) class I molecules in a specific manner, KIR2DL1 recognizes HLA-C2 group allies, KIR2DL2 and KIR2DL3 recognize HLA-C1 group allies, KIR3DL1 recognizes HLA-Bw4 group allies, and KIR3DL2 recognizes HLA-A3/A11 allies. Activating KIRs (aKIRs) such as KIR2DS1, KIR2DS2, and KIR2DS4 recognize HLA C2, HLA C1, and HLA A11, respectively, but the ligands of the remaining KIRs remain largely unknown. Based on their chromosomal locations, KIR genes can be further identified as centromeric (cen) or telomeric (tel) genes. In addition, KIR genotype AA is made up of only one aKIR gene: KIR2DS4, while KIR genotype B/x is made up of a number of more variable aKIR genes (4).
Normally, autoimmune activation is inhibited because autologous cells express at least one inhibitory HLA ligand; however, tumor transformed cells downregulate HLA expression and/or upregulate activating signals that may trigger NK cell activation (5,6). Following allogeneic hematopoietic stem cell transplantation (allo-HSCT), donor-derived NK cells may be activated as the recipients may not express the same inhibitory HLA ligands as the donor, preventing their association with the donor iKIRs. This has led to widespread speculation that NK cell alloreactivity in graft versus host (GVH) direction may provide additional benefits to tumor-killing strategies.
The Perugia group first established the KIR ligand-ligand model (also known as the KIR ligand model) based on HLA phenotype differences between donors and recipients. In this model, they assumed that donor-derived NK cells might kill recipient cells because the HLA ligands presented by the donor might be absent in the recipient. When they evaluated T celldepleted (TCD) transplants without post-transplant immunosuppression, they were able to show that KIR ligand mismatch between donor-recipient pairs provided some protective effect against relapse, especially in patients with acute myeloid leukemia (AML) (7). Further development of KIR-typing technology allowed researchers to develop the receptor-ligand model (also known as the missing ligand model), which was used to evaluate the compatibilities between donor iKIRs and recipient HLA ligands. Results using this model suggested that the receptor-ligand model was a more accurate predictor for relapse risk than the KIR ligand model in leukemia patients (8). Additionally, Cooley et al. reported that KIR B/x donors significantly improved the relapse-free survival (RFS) rates for recipients with AML when compared to donors with a KIR AA genotype, suggesting that aKIRs may play a critical role in reducing relapse (9). Following these observations, numerous clinical studies have explored the impact of KIR on transplant outcomes. However, a large variability was found in these results and several factors may be responsible for these discrepancies, including disease type, transplant regimen, donor-recipient relationship, graft source and graft composition, etc (10)(11)(12).
In the last few decades, our understanding of NK cell reconstitution and KIR education has evolved a great deal. Pioneer studies in this field have found that reconstituted NK cells are highly immature and exhibit compromised cytotoxicity against leukemia cells in the early phases following transplantation. Afterward, these NK cells gradually acquire receptors and KIR reconstitution can take between several months and even years (13,14). Importantly, a variety of data has suggested that NK cells acquire specific functionality only after engagement between the iKIRs and their cognate ligands. However, NK cells expressing iKIRs without cognate ligands (non-self KIR) are hyporesponsive and referred to "uneducated cells" (15,16). Further, the education process mediated by cognate ligands is not restricted to autologous NK cells, but has also been demonstrated in donor-derived reconstituted NK cells after HSCT (17)(18)(19).
Recently, the Nowak team proposed the iKIR-HLA model to explore the optimal donor. Since the HLA environment may be altered after transplantation (from donor to recipient), the variations in iKIR-HLA pairs could be divided into three groups (decreased group: cognate iKIR-HLA pairs present in donor but absent in recipient; unchanged group: cognate iKIR-HLA pairs present both in donor and recipient; increased group: cognate iKIR-HLA pairs present in recipient but absent in donor). Consistent results from their studies showed that decreased iKIR-HLA pairs post transplantation correlated with a higher risk of relapse and inferior overall survival (OS), indicating that poor NK cell education resulted in weaker graft versus leukemia (GVL) effects (20)(21)(22). To further investigate the effects of these KIR interactions on transplant outcomes, we designed a retrospective study to evaluate a cohort of 246 patients, and evaluated our clinical outcomes using the iKIR-HLA model, the receptor-ligand model and KIR gene content.

Patients
This retrospective study was comprised of 246 patients with hematological malignancies. All transplants were performed between January 2015 and June 2018 and all methodologies applied during this study were consistent with the Declaration of Helsinki. The protocol was approved by the Ethics Review Committee of the First Affiliated Hospital of Zhejiang University and informed consent was obtained from each patient before transplantation. The authors had full access to the data and assume responsibility for its authenticity.

KIR and HLA Typing
Peripheral blood mononuclear cells were collected from recipients and their donors prior to transplantation and used for HLA and KIR testing. Alleles in the HLA-A, -B, and -C loci were determined using high-resolution HLA typing and KIR gene analysis was performed using the PCR-SSO method (KIR SSO Genotyping Test; OneLamda, Canoga Park, CA, USA).

Definitions
Relapse was defined as disease reoccurrence in bone marrow and/or extramedullary sites. Non-relapse mortality (NRM) was defined as death from any cause apart from relapse. Overall survival (OS) was defined as the time from transplant until death or last follow up, and disease-free survival (DFS) was defined as survival without relapse. Patients were classified as low/ intermediate risk or high/very high risk based on the refinement of the disease risk index (DRI) (25). Diagnosis of acute and chronic GVHD (aGVHD and cGVHD) was made using established criteria (26,27). The viral loads for Epstein-Barr virus (EBV) and cytomegalovirus (CMV) were monitored weekly for the first 3 months after transplantation, biweekly from the fourth to the sixth month post-transplant, and monthly from the seventh to the twelfth month post-transplant. Viremia was defined as a viral load in excess of 5 × 10 2 copies/ml.

Statistical Analysis
All clinical data were analyzed using SPSS 22.0 (IBM, Armonk, NY, USA) and R project 3.6.1 software (http://www.r-project. org). The clinical features for the samples were presented as median or percentage values. OS and DFS were calculated using the Kaplan-Meier method and compared using the log-rank test. The cumulative incidences of EBV viremia, CMV viremia, aGVHD, cGVHD, relapse, and NRM were estimated via the competing-risks model and compared using the Gray test. All variables with a p-value of <0.10 in the univariate analysis were then included in the multivariate analysis. Results were considered statistically significant when p < 0.05.

Characteristics of Patients and Donors
The clinical features of these 246 donor-patient pairs are summarized in Table 1. In this retrospective study, 142 (57.7%) patients with myeloid disease and 104 (42.3%) patients with lymphoid disease received haplo-HSCT at our center. Disease types included acute myeloid leukemia (AML, n = 1 1 5 ) , m y e l o d y s p l a s t i c s y n d r o m e ( M D S , n = 2 2 ) , myeloproliferative neoplasm (MPN, n = 5), acute lymphoblastic leukemia (ALL, n = 93), and lymphoma (n = 11). The median age of the patients and donors in these groups were 30 years (range, 9-50 years) and 35 years (range, 11-59 years), respectively. The median mononuclear (MNC) cell and CD34 + cell counts in the grafts were 15.34 × 10 8 /kg (range, 2.97-59.80 × 10 8 /kg) and 6.30 × 10 6 /kg (range, 0.27-34.37 × 10 6 /kg), respectively. A total of 233 (94.7%) patients received the MAC regimen and 13 (5.3%) received the RIC regimen. ATG-F was used in 205 (83.3%) patients while the other 41 (16.7%) received ATG-G as part of their conditioning regimen. One hundred eighty-one (73.5%) patients received haplo-HSCT during their first remission (CR1); 73 (29.7%) patients were defined as high or very high risk based on the refinement of DRI (49 in the myeloid cohort and 24 in the lymphoid cohort, 34.5 vs 23.1%, p = 0.053). Most patients expressed HLA C1C1 or HLA C1C2 and only 4.5% presented with a HLA C2C2 ligand.
The CI for CMV viremia within 180 days of transplant was 65.0% (78.1% in patients treated with ATG-G and 62.4% in patients treated with ATG-F, p = 0.003). Donor-patient pairs with R-L mismatch at HLA-C locus tended to experience a lower CI for CMV viremia than did donor-patient R-L C matched pairs (62.5 vs 71.4, p = 0.079). The multivariate analysis revealed that only ATG-G was an independent risk factor for CMV viremia (HR = 1.70, p = 0.008).

aGVHD and cGVHD
Following transplantation, 83 (33.7%) developed grade II-IV aGVHD (aGVHD 2-4 ). As expected, a significant reduction in aGVHD 2-4 occurrence was found in patients receiving RIC conditioning compared with patients receiving MAC conditioning (7.7 vs 35.2%, p = 0.041). Patients with low and intermediate risk also experienced a lower CI of aGVHD 2-4 (30.6 vs 41.1%, p = 0.080). However, none of these factors remained significant in the multivariate analysis. In lymphoid cohort, there was a trend that donor-patient pairs with R-L mismatch on HLA-C locus experienced a lower aGVHD 2-4 (33.3 vs 53.9%, P = 0.095).
Among patients who survived more than 100 days after transplantation, 100 (41.8%) patients developed cGVHD and 42 (17.6%) of them had moderate to severe cGVHD. Univariate analysis identified KIR2DS2 (p = 0.048) and KIR2DS3 (p = 0.083) as two potent protective factors for moderate to severe cGVHD. Nevertheless, no such correlations were found in the multivariate analysis.

Relapse and NRM
After a median follow up time of 3.0 years (yr) (range, 0.1-5.5 yr), 55 (22.4%) patients experienced relapse. Patients with lymphoid disease experienced a higher 3-yr relapse rate than patients with myeloid disease (26.2 vs 17.3%, p = 0.087). The CI for 3-yr relapse was also higher in patients with high/very high-risk disease (32.9 vs 16.1%, p = 0.002). Patients who received HSCT at CR1 experienced a lower 3-yr relapse rate than the other group (16.5

OS and DFS
The CI for 3-yr OS was 75.6% for all patients. Disease category (p = 0.016), disease status (p = 0.010), and disease risk index (p = 0.036) were all found to influence 3-yr OS in the univariate analysis. In addition, the 3-yr OS rate in transplants with decreased iKIR-HLA C pair was shown to be 65.1% [95% confidence interval (CI): 52.3-81.0%], which was lower than those with unchanged or increased iKIR-HLA C pair (77.9%, 95% CI: 72.3-83.9%, p = 0.093), and the negative impact of decreased iKIR-HLA C pair was more apparent in the myeloid cohort  (Figures 2A-C). In the lymphoid cohort, patients who received

DISCUSSION
There has been a longstanding debate about the impact of KIR alloreactivity on clinical outcomes. It was only recently revealed that reconstituted KIR are educated by HLA ligands and that the loss of the cognate ligands dampens NK cell functions (17)(18)(19). This means that searching for donors who exhibit the greatest NK cell function in recipients rather than "match or mismatch" would be a more reliable measure for predicting transplant success.   Previously, Nowak et al. proposed the iKIR-HLA model that could be used to predict transplant outcomes (20)(21)(22). Among the multiple interactions between the iKIRs and HLA ligands, we identified that only decreased iKIR-HLA C pair post transplantation was a negative indicator for relapse and survival, especially in patients with myeloid disease. Nevertheless, variations in iKIR-HLA A3/A11 pair and iKIR-HLA Bw4 pair did not influence the treatment outcomes.
It is widely accepted that almost all HLA C molecules are recognized by iKIRs. However, only a minority of HLA B and HLA A epitopes act as KIR ligands (29)(30)(31). Similarly, all patients in our cohort expressed at least one HLA C ligand, while the HLA Bw4 and A3/A11 ligands were expressed at a frequency of 60.2 and 46.7%, respectively. This suggests that the HLA C ligands play a dominant role in KIR education (32). Given this, reconstituted NK cells with decreased iKIR-HLA C pair may exhibit impaired anti-tumor effects (18,19). In addition, the expression levels of HLA A and B ligands on normal cells are more than tenfold higher than that of the HLA C (33), this means that when cancerous cells downregulate HLA class I antigens to escape immune surveillance, the stability of the selftolerance mediated by iKIR-HLA C interactions is more vulnerable to be broken. In other words, HLA-C may play a major role in missingself recognition and modulate NK cell activation. Moreover, Pende et al. found that lymphoblastic leukemias express a higher surface density of HLA class I molecules than myeloid leukemias (34). Verheyden et al. went on to test the expression of HLA ligands in normal T cells, AML cells, B-ALL cells, and B-chronic lymphoid leukemic (B-CLL) cells. Interestingly, only HLA C were dramatically downregulated on all types of leukemic cells as compared with their healthy control, with this downregulation being the most apparent in AML cells (35). Makanga et al. demonstrated that CD57 + and KIR + NK cells from healthy individuals exhibited the highest degree of cytotoxicity against AML blasts, while ALL targets were less susceptible to KIR + NK subsets compared with NKG2A + NK subsets (36). On the basis of previous studies, we hypothesize that KIR may have a minor impact on the elimination of lymphoblastic leukemias, and patients with myeloid disease are more likely to benefit from well KIR-educated NK cells.
In many European studies, aKIRs, especially KIR2DS1 (37-39) and KIR2DS2 (40,41), have been shown to be associated with improved survival or reduced relapse. Yet, as reported in several studies from East Asia (42)(43)(44)(45), aKIRs were not found to grant any survival advantage or relapse protection to the patients in our cohort. One reason for this may be the genetic differences between these different ethnic groups. Single et al. revealed that almost 46.7% Europeans express the KIR2DS2 gene, and 66.5% present the HLA C2 ligand for KIR2DS1 (46). However, both the KIR2DS2 gene (18.7%) and the HLA C2 ligand (32.1%) were expressed at much lower frequencies in this study. The KIR2DS1 gene frequency in our cohort was also a bit lower than those of the European populations (33.7 vs 37.8%). Thus, we speculate that KIR2DS1 may have a reduced chance of activation resulting from the absence of its cognate ligand,  and that the beneficial impact of KIR2DS2 on transplant outcomes may be more apparent in a larger cohort of Chinese patients. Additionally, we could not find evidence of any significant association between receptor ligand mismatch and clinical outcomes. Since mature donor lymphocytes are mostly eliminated following ATG treatment, the transient expression of alloreactive NK cells in the recipients may not be sufficient to influence GVHD (47)(48)(49). After which the reconstituted NK cells expressing non-self KIRs may not exhibit enough cytotoxicity to eliminate the remaining leukemic cells (17)(18)(19).
In summary, we conclude that when using ATG-based haplo-HSCT, deceased iKIR-HLA C pair should be avoided during donor selection, especially for patients with myeloid disease. The exact role of the aKIRs in the Chinese population still needs to be explored in future studies.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Ethics Review Committee of the First Affiliated Hospital of Zhejiang University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

AUTHOR CONTRIBUTIONS
HH designed the study and supervised the analyses and manuscript preparation. YZ, FG, and YW collected and analyzed the data, YZ and FG wrote the manuscript. All authors discussed and interpreted the results. All authors contributed to the article and approved the submitted version. YZ and FG contributed equally to this work and should be considered as co-first authors.