Molecular HLA mismatching for prediction of primary humoral alloimmunity and graft function deterioration in paediatric kidney transplantation

Introduction: Rejection remains the main cause of allograft failure in paediatric kidney transplantation and is driven by donor-recipient HLA mismatching. Modern computational algorithms enable assessment of HLA mismatch immunogenicity at the molecular level (molecular-mismatch, molMM). Whilst molMM has been shown to correlate with alloimmune outcomes, evidence demonstrating improved prediction performance against traditional antigen mismatching (antMM) is lacking. Methods: We analysed 177 patients from the CERTAIN registry (median follow-up 4.5 years). molMM scores included Amino-Acid-Mismatch-Score (AAMS), Electrostatic-Mismatch-Score (EMS3D) and netMHCIIpan (netMHC1k: peptide binding affinity ≤1000 nM; netMHC: binding affinity ≤500 nM plus rank <2%). We stratified patients into high/low-risk groups based on risk models of DSA development. Results: DSA for all HLA loci; the EMS3D algorithm had particularly consistent performance (area under the receiver operating characteristic curve (AUC) >0.7 for all HLA loci vs. 0.52-0.70 for antMM). ABMR (but not TCMR) was associated with HLA-DQ molMM scores (AAMS, EMS3D and netMHC). Patients with high-risk HLA-DQ molMM had increased risk of graft function deterioration (50% reduction in baseline eGFR (eGFR50), adjusted HR: 3.5, 95% CI 1.6-8.2 high vs. low EMS3D). Multivariable modelling of the eGFR50 outcome using EMS3D HLA-DQ stratification showed better discrimination (AUC EMS3D vs. antMM at 2 years: 0.81 vs. 0.77, at 4.5 years: 0.72 vs. 0.64) and stratified more patients into the low-risk group, compared to traditional antMM. Conclusion: Molecular mismatching was superior to antigen mismatching in predicting humoral alloimmunity. Molecular HLA-DQ mismatching appears to be a significant prognostic factor for graft function deterioration in paediatric kidney transplantation.