(116) Impact of HLA-DPB1 and DPA1-DPB1 Linkage Mismatch on Survival in HLA-14/14 Matched Unrelated Donor HSCT: Establishment and clinical validation of a predictive model based on NGS technology

Aim: While HLA-DPB1 mismatches significantly influence unrelated donor hematopoietic stem cell transplantation (URD-HSCT) outcomes, the role of HLA-DPA1-DPB1 linkage mismatches remains unclear. This study evaluates their clinical impact and establishes a predictive model to optimize donor selection.

Methods: A retrospective study was conducted on 250 HLA-14/14 matched URD-HSCT cases using next-generation sequencing (NGS) for HLA-A, -B, -C, -DRB1, -DQB1, -DQA1, -DRB3/4/5, -DPA1, and -DPB1 typing. DPA1 and DPB1 mismatches were categorized, and their impact on 2-year overall survival (OS) and non-relapse mortality (NRM) was evaluated. A DPA1-DPB1 linkage prediction model was developed using an artificial neural network algorithm and validated it using clinical and follow-up data from 250 donor-recipient pairs.

Results: Only 10.4% of cases had matched DPA1 and DPB1 alleles. Among approximately 90% of cases with mismatches, we identified three key findings. For allele-specific effects, prevalent DPB1 allelic mismatches, such as DPB1*02:01/DPB1*03:01, were associated with decreased 2-year OS and increased NRM, while DPB1*02:02/DPB1*05:01 and DPB1*02:01/DPB1*05:01 mismatches had no significant impact. For DPA1-DPB1 linkage patterns, high-risk DPA1*02:02-DPB1*05:01/DPA1*02:01-DPB1*17:01 mismatch decreased 2-year OS, especially in AML/MDS recipients, while DPA1*02:02-DPB1*05:01/DPA1*01:03-DPB1*02:01 mismatch showed no impact. We established 47 types of DPA1-DPB1 linkage prediction models and categorized them into four types (Types I-IV) based on amino acid hydrophilic and hydrophobic properties in the anchoring position of peptides P1 to P9. Patients with P1 mismatches with donors had lower 2-year OS and higher NRM compared to fully matched cases, whereas there was no significant difference in the 2-year OS rate and NRM rate in P1 match patients. The predictive model outperformed the TCE model in predicting 2-year OS, with a larger area under the ROC curve.

Conclusion: HLA DPA1-DPB1 linkage mismatches significantly impact URD-HSCT outcomes, with specific allelic pairs and structural features predicting survival and NRM. Our linkage prediction model provides superior donor selection guidance compared to TCE, emphasizing the importance for comprehensive HLA-DP assessment in transplantation strategies.