Incidence of HLA Loss in a Global Multicentric Cohort of Post-Transplantation Relapses: Results from the Hlaloss Collaborative Study

Introduction. Genomic loss of an HLA haplotype encoding incompatible alleles ("HLA loss") has been described in previous single-center studies as a mechanism by which leukemic cells evade the graft-versus-leukemia effect mediated by alloreactive donor T cells and outgrow into a clinically evident relapse. HLA loss accounts for up to 30% of relapses after HLA-haploidentical transplants (Crucitti, Leukemia 2015), but the actual frequency and clinical relevance of this phenomenon in unrelated donor HSCTs, including cord blood transplants, are largely unknown. Here we present the first global collaborative study to investigate the incidence of HLA loss across different transplant settings.

Methods. Twenty transplant centers from Europe (n=16), North America (n=3) and Asia (n=1) joined to form the HLALOSS consortium. To date, we collected a total of 619 cases of hematologic relapse from adult patients with acute myeloid leukemia (78.5%), acute lymphoblastic leukemia (13.9%), myelodysplastic syndromes (4%) or myeloproliferative neoplasms (1.1%) after allogeneic HSCT from HLA-haploidentical relatives (31.7%), HLA-mismatched unrelated donors (MMUD, 21.3%), 10/10-matched unrelated donors (MUD, 37.2%), or unrelated cord blood units (UCB, 9.8%). Where available, the donor and patient germlines and the patient pre-transplant disease were collected in parallel. Until today, 476 cases were analyzed using conventional HLA typing of sorted leukemic blasts, the recently developed HLA-KMR assay (Ahci and Toffalori, Blood, 2017) or a novel Next-Generation Sequencing (NGS) method. The latter was developed adapting the HLA typing strategy in use at the DKMS (Lange, BMC Genomics 2013) to the study of chimeric samples, and allowing to cover all possible HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1 alleles and to analyze at least 48 different cases in a single run.

Results. Out of the 476 relapses analyzed to date, 396 (83.2%) were informative for the study of HLA loss. Of these, 155 occurred after haploidentical HSCT, 101 after MMUD HSCT, 93 after 10/10-matched, HLA-DPB1 mismatched MUD, and 47 after UCB HSCTs. Three-hundred-two (76.2%) of cases were analyzed using the NGS platform. This method resulted particularly robust, reliable and sensitive in analyzing large sample series: the mean coverage across the 6 sequenced loci was over 8500x, up to 0.5% of the HLA allele of interest could be detected in artificial chimerism curves, and relapse samples tested in parallel via the sequencing platform and HLA-KMR (n=10) showed remarkable concordance between the two methods (R²=0.86, p<0.0001). In total, we detected 51 HLA loss post-transplantation relapses out of the 396 cases analyzed (12.8%). Of these, 35 occurred after haploidentical HSCT (22.6% of relapses in this setting), 12 after MMUD HSCT (11.9%), 4 after 10/10 MUD HSCT (4.3%) and, notably, none after UCB HSCT.

Conclusions. The present data, obtained from the largest collaborative study on the immunobiology of relapse to date, confirm the clinical relevance of HLA loss as a major mechanism of immune evasion and post-transplantation relapse after allogeneic HSCT, with an incidence which is proportional to the number of donor-recipient HLA mismatches. The only exception is represented by UCB HSCT which, despite being often performed across multiple major HLA incompatibilities, does not appear to be associated with this relapse modality. This finding might reflect the fact that in UCB HSCT, multiple HLA mismatches are often not encoded in cis on the same chromosome, thereby reducing the selective advantage for leukemic cells that undergo an HLA haplotype loss. This phenomenon might in turn contribute to the lower incidence of relapse reported for UCB HSCT compared to other stem cell sources.