An Immunogenetic Perspective on Bone Marrow Failure Syndromes: The KIR-HLA System

Acquired bone marrow failure (BMF) syndromes, chiefly aplastic anemia (AA) and related disorders such as paroxysmal nocturnal hemoglobinuria (PNH) are characterized by immune mediated hematopoietic suppression triggered by yet unknown inciting events. Clearly, the pathogenesis is a complex combination of traits and overshooting immune reaction mediated mainly by cellular immune system. Such an autoimmune reaction may be a result of overcompensation due to immunodeficiency and/or hyperreactivity traits summarized as putative immunogenetic predisposition. In addition to the well-studied HLA risk alleles, killer immunoglobulin-like receptor (KIR) genotype may be also involved, as both systems regulate reactivity of cytotoxic T and NK cells.

We previously described the landscape of KIR and HLA in T-LGLL serving as the extreme polarized model of hematopoietic immune suppression (1) and demonstrated that the 3DL2 KIR receptor and its HLA-A*03/A*11 ligand were significantly enriched in T-LGLL compared to controls.

To better define the immunogenetic landscape in this realm, we here extend our analysis to AA/PNH and MDS.

We examined the frequencies of individual KIR genes between the different diagnoses. In the AA/PNH cohort (n=128), the inhibitory KIR gene, 2DL1, was under expressed compared to controls (89% vs 96%, p=0.05). Similarly, 2DL3, another inhibitory receptor, was found at a lower frequency in the AA/PNH group (85% vs 95%, p=0.02). For the MDS cohort (n=82), the 2DS3 activating KIR gene was significantly enriched compared to controls (42% vs 23%, p=0.006).

Since KIRs interact with HLA ligands, we also analyzed the genotypic frequencies of various HLA class I alleles between diseases across known serological motifs. In AA/PNH, the HLA C1/C2 allotype was present at a lower frequency compared to controls (33% vs 49%, p=0.01). Similarly, the homozygous Bw6 motif (Bw6/Bw6) genotype was under expressed in AA/PNH patients (32% vs 45%, p=0.04). Conversely, the Bw4/Bw6 genotype was enriched in AA/PNH patients (55.5% vs 42.3%, p=0.05). For MDS, the C2/C2 allotype was more frequent than in controls (27 vs 15%, p=0.044). Also, the Bw4/Bw4 genotype was overrepresented (28% vs 12%, p=0.006) while the Bw6/Bw6 genotype was less frequent (18% vs 45%, p<0.001). In both cohorts, the proportions of individuals having one copy of HLA A*03 or HLA A*11 allele were significantly higher than controls.

We explored the frequencies of KIR-HLA configurations to uncover further differences in the study groups. In AA/PNH, the 2DL1/C2 configuration, an inhibitory match, was significantly under expressed (46% vs 62%, p=0.03). Additionally, the 2DL3/C1 combination was also lower in AA/PNH compared to controls (68% vs 81%, p=0.03). The 2DL3/C1 configuration was also under expressed in the MDS cohort (60% vs 80%, p=0.002). Conversely, 3DL1/Bw4 inhibitory combination was enriched in MDS patients (73% vs 53%, p=0.008). The frequency of patients presenting 3DL2 and at least one copy of the A*03 or A*11 ligands was higher in each of the diseased cohorts vs controls.

We also performed a random forest classification on each of the disease groups vs controls to identify potentially more complex patterns. The HLA C1/C1 allotype, 2DS3 KIR gene, and Bw6/Bw6 phenotype had the highest mean decrease in Gini scores, indicating their importance in correlating with the disease type.

Finally, we gauged the net KIR activity (NKA) for each patient by taking the number of inhibiting matches and subtracting that from the number of activating matches. The NKA for the control cohort was –1.65, indicating an increased inhibiting genotype but no significant differences were found compared to the disease cohorts, with NKA’s of –1.57 and –1.55 for AA/PNH and MDS respectively. We also mapped clinical phenotypes to the NKA scores of the T-LGLL cohort previously reported (1). Patients with an NKA above the median had a higher proportion of lymphocytosis and poorer outcomes vs controls (40% vs 22%, p=0.08 and 49% vs 30%, p=0.08).

In summary, both similarities and differences exist in the landscapes of AA/PNH and MDS compared to controls. Most notably, the combination of 3DL2 and HLA A*03/A*11 seems to be significantly enriched in both diseased groups, suggesting their possibility as risk factors.