The evaluation of the HLA-KIR interaction models may predict the NK alloreactivity in haplo-HSCT based on ptcy: Experience of a single spanish center

Abstract INTRODUCTION The reconstitution of NK cells and alloreactivity mediated by KIR receptors after allogeneic transplantation of hematopoietic progenitors from an HLA-haploidentical donor (Haplo-HSCT) gain relevance due to their involvement to the GvL effect. However, this contribution is still under debate in strategies with non-manipulated cells with PTCy The aim was to examine the clinical benefit of predicted KIR-alloreactivity on haplo-HSCT with PTCy using the composite variable graft-versus-host disease & relapse-free survival (GRFS). METHODS A total of 145 patients with hematological disorders receiving an haplo-HPCT with PTCy were included. HLA typing was performed Next-generation sequencing for recipients, while donors typing was developed by SSO-PCR when family-haplotype segregation was available. Second-field of the non-shared haplotype of donors was inferred based on the most common allele at the tool Haplostats. Pairs were genotyped for KIR genes KIR-SSO kit. HLA-B leader peptide was analyzed using the BEAT and IMGT/HLA algorithms. NK-alloreactivity was evaluated according different prediction models. Patients were follow-up to 36 months. RESULTS The most frequent diagnoses were AML and MDS (52%). Forty-eight individuals underwent a previous HSCT. Patients were conditioned predominantly by RIC (72%). HCT-CI was ≥3 in 30% and the DRI was high/very-high in 17% of case. The conditioning regimen resulted as an important driver of the NK alloreactivity. the incidence of cGVHD was higher in the MAC group (33% versus 13%, p = 0.013), while the RIC group tended to show a higher incidence of mortality (29% versus 20%, p = 0.091). MAC significantly diminished the impact of NK alloreactivity, since no statistically significant associations were observed for most clinical outcomes. However, when multiple iKIRmm was assessed using the missing-ligand model, a shortened GRFS was observed (HR: 3.86, p = 0.021). Conversely, in the setting of RIC, NK alloreactivity demonstrated a pronounced post-transplant detrimental effect, since it showed a significant impact on grade III-IV aGvHD, relapse, RFS and GRFS. The prediction of NK-KIR alloreactivity using HLA-KIR interaction models calculated by the presence of multiple inhibitory KIR mismatches (iKIRmm) was strongly associated with a shortened GRFS. The presence of multiple iKIRmm evaluated by the missing-ligand model showed a median GRFS of 5 months compared to the 25 months of the complete cohort (HR: 2.46, p=0.001). The same result was detected for the missing-licensing model (HR: 2.41, p=0.015). In both cases, the multivariate analysis including those pre-transplant parameters obtained from a previous univariate analysis showed that the missing-ligand (HR 2.16, p=0.008) and the missing-licensing (HR: 2.265, p=0.022) separately, with the HCT-CI≥3 (HR: 1.79, p=0.019; HR: 1.91, p=0.008, respectively) were independent risk factors for a diminished GRFS. The analysis of the HLA-B leader peptide did not show any significant results according to the different outcomes analysis, probably due to the size of the cohort. CONCLUSION The analysis of the HLA-KIR interaction models for the prediction of the NK-KIR alloreactivity could be a useful tool for the selection of the most suitable donor in haplo-HSCT.