It is well established that the use of a donor matched for 9–10/10 alleles at HLA-A,-B,-C,-DRB1,-DQB1 significantly improves overall survival (OS) after unrelated donor (UD) haematopoietic stem cell transplantation (HSCT). Whilst the matching status for HLA-DPB1 alleles has been shown to influence transplant complications (relapse and graft-versus-host disease (GVHD), its impact on survival has not been well defined. The current unmet need in clinical practice is an approach to stratify selection criteria when a clinician is confronted with the choice between several 10/10 or 9/10 matched unrelated donors.
There is now considerable interest in exploring different types of matching criteria to define permissive HLA-DPB1 mismatches which may be associated with an improved outcome. We have previously shown that HLA-DPB1 permissiveness can be functionally defined by the characterization of shared T cell epitopes (TCE) recognized by alloreactive T cells. In this model, allelic HLA mismatches are classified as permissive if they do not involve TCE disparities, and as non-permissive if they do. Using this concept, we developed two overlapping algorithms of permissivity for allelic HLA-DPB1 mismatches, on the basis of 3 (TCE3) or 4 (TCE4) groups of DPB1 alleles encoding immunogenic TCE. Data from relatively small prospective studies has shown a worse outcome to be associated with non-permissive DPB1 TCE disparities.
Here, we present outcomes in 9123 UD-HSCT pairs, collected through the International Histocompatibility Working Group (IHWG). The cohort was comprised of 5809 10/10 matched transplant pairs and 3314 9/10 matched pairs. Within the 10/10 and 9/10 matched pairs three groups of patients were identified: 1. Zero DPB1 mismatches (i.e. allele matched), 2. Permissive DPB1 mismatch, 3. Non-permissive DPB1 mismatch. The model was adjusted for disease severity, source of stem cells, conditioning regimen, use of T-cell depletion, patient/donor gender and patient age.
In line with DPB1 allele frequencies in worldwide populations, the number of transplants scored as permissive was higher for TCE3 (4398/7270 [60.4%]) than for TCE4 (2577/7270 [35.4%]). Using the DPB1 permissive mismatch transplants as the reference group (either 10/10 or 9/10 matched), we showed that DPB1 allelic matches resulted in similar survivals to DPB1 permissive mismatches, both in the 10/10 (HR 0.96, p=0.498 for TCE3 and HR 0.99, p=0.85 for TCE4) and the 9/10 setting (HR 0.97, p=0.70 for TCE3 and HR 0.99, p=0.96 for TCE4).
In contrast, survival was significantly worse in the presence of a non-permissive TCE3 or TCE4 mismatch, both in the 10/10 (HR 1.15, p=0.0005 for TCE3 and HR 1.13, p=0.0035 for TCE4) and in the 9/10 matched setting (HR 1.13, p=0.0140 for TCE3 and HR 1.11, p=0.0448 for TCE4). The survival detriment appeared to be due to a significantly increased non-relapse mortality (TCE3: 10/10 HR 1.27, p<0.001 and 9/10 HR 1.21, p=0.0001; TCE4: 10/10 HR 1.24, p<0.001 and 9/10 HR 1.13, p=0.0514), as well as an increase in grades II-IV acute GVHD (TCE3: 10/10 HR 1.17, p<0.001 and 9/10 HR 1.29, p<0.001; TCE4: 10/10 HR 1.12, p=0.0035 and 9/10 HR 1.19, p<0.0001). There was no significant difference in disease relapse between permissive and non-permissive mismatched pairs. Finally, using the 10/10 DPB1 permissive mismatched group as a reference, we found survival to be similar for 10/10 DPB1 non-permissive (HR 1.15) and 9/10 DPB1 permissive (HR 1.20) or DPB1 allele matched (HR 1.17) transplants.
In conclusion, our results suggest that extending donor selection to include HLA-DPB1 both allelic and functional TCE matching may result in better prediction of survival for patients. These findings provide an attractive new algorithm to stratify donor choice when several well-matched UD are identified.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.