Abstract 1957


In spite of the considerable progress achieved during the last decade in the field of allogeneic Hematopoietic Stem Cell Transplantation (HSCT), disease relapse remains a crucial unsolved issue, due to our limited insights into the biology of the interplay between leukemia and donor immunity. We and others have shown that genomic loss of patient-specific mismatched HLA is a frequent mechanism by which residual leukemic cells can evade donor T cell-mediated immune control and determine a clinical relapse. HLA loss relapses have been reported both after family haploidentical and after volunteer unrelated donor HSCT, but the actual incidence and risk factors for these peculiar relapses are to date unknown.


We retrospectively evaluated 230 consecutive transplants performed over the last ten years in a single institution. All donors were partially HLA mismatched (family mismatched: 170; volunteer unrelated: 60). All patients were affected by high-risk myeloid malignancies (Acute Myeloid Leukemia (AML): 179; Myelodysplastic Syndrome (MDS): 27; Myeloproliferative Neoplasms: 17; Others: 7), received a fully myeloablative conditioning regimen, and infusion of donor T cells, either as part of the graft or as post-transplantation add-back. In addition to standard bone marrow morphological examination, post-transplantation follow-up comprised genomic HLA typing of bone marrow aspirate samples, aimed to identify HLA loss relapses.


A total of 83 relapses occurred, 72 and 11 after related and unrelated donor HSCT, respectively. We documented 21 relapses with genomic loss of the patient-specific mismatched HLA (25% of relapses). HLA loss occurred predominantly in patients with AML (n=19), but was also observed in patients with MDS (n=1) and primary myelofibrosis (n=1). None of the 11 cases of relapse after volunteer unrelated donor HSCT displayed HLA loss, whereas it was evident in 21 out of 72 (29%) relapses after family mismatched donor HSCT. Accordingly, the presence of multiple (more than 4) HLA mismatches between donor and recipient represented a significant risk factor for HLA loss (HR: 4.17, 95% CI: 0.88–19.82, p=0.07). None of the disease-related factors tested (morphological subtype, presence of dysplasia, hyperleukocytosis, cytogenetic and molecular profile) correlated significantly with HLA loss. In line with the hypothesis that selection of these leukemic variants may be elicited more frequently in the presence of a strong alloreactive immune pressure, HLA loss occurred more frequently in patients with clinical Graft-versu-Host Disease (GvHD), either acute (p=0.02) or chronic (p=0.007). Interestingly, HLA loss relapses occurred later than their “classical” counterparts (median time to relapse 307 vs 116 days, range 56–784 vs 10–579, p<0.0001), suggesting the existence of a long phase of equilibrium between leukemic cells and donor T cells before the occurrence of a de novo mutation and the outgrowth of immune escape variants. Of notice, HLA loss relapses shared several features with post-transplantation extramedullary relapses, suggesting a common biological drive leading to the two immune escape mechanisms.


Loss of the mismatched HLA is a frequent mechanism of relapse for patients with high-risk myeloid malignancies, and is associated with donor-versus-host alloreactivity, as demonstrated by its significant correlation with the number of donor-recipient HLA mismatches and with the occurrence of GvHD. Ultimately, given the documented ineffectiveness of lymphocyte infusions from the original stem cell donor in patients with HLA loss relapses, efforts should be aimed towards the optimization of early detection tools and targeted therapeutic strategies specific for these peculiar and frequent relapse variants.


Bordignon:MolMed SpA: Employment. Bonini:MolMed S.p.A.: Consultancy.

Author notes


Asterisk with author names denotes non-ASH members.