Genomic Loss of the Mismatched HLA Locus in Leukemia Is a Major Mechanism of in Vivo Escape from T Cell Immunosurveillance Following Haploidentical HSCT

Hematopoietic Stem Cell Transplantation (HSCT) from haploidentical family donors is a promising therapeutic option for nearly all patients suffering from high-risk leukemia. Until now, its application has been limited by the prolonged immunodeficiency that patients suffer as a consequence of graft T cell depletion, used to prevent severe Graft versus Host Disease (GvHD). When efficient strategies to control GvHD are applied, adoptive immunotherapy with donor T cells grants a significant advantage for immune reconstitution. However, direct evidence for the role of haploidentical donor T cells in controlling leukemia relapse is still missing. Here we report on the in vivo selection of de novo mutant variants of acute myeloid leukemia (AML), accounting for relapse after haploidentical HSCT and adoptive transfer of donor T cells. These novel variants of AML were observed in 5 out of 17 (29%) patients suffering from disease relapse in a series of 43 patients transplanted at the San Raffaele Hospital in Milan from 2002 to 2008. All patients received a myeloablative conditioning regimen and high doses of haploidentical donor stem cells (median 10.2×10⁶ CD34⁺ cells/kg, range 4.6–15.5). Donor T lymphocytes were infused as part of the graft (n=21, median 438×10⁶ CD3⁺ cell/kg, range 179–796) or as post-transplant add-backs (n=22, median 111×10⁵ CD3⁺ cell/kg, range 1–900). Human Leukocyte Antigen (HLA) genomic typing was routinely used for post-transplant donor-recipient chimerism assessment. The five patients with de novo mutant variants of the original leukemia came to our attention because patient-specific HLA alleles could not be detected in bone marrow samples harvested at disease relapse, nor in subsequently sorted AML blasts. A Loss of Heterozygosity (LOH) study was performed on purified blasts from these patients, and demonstrated that patient-specific HLA alleles were lost due to extensive events of homologous recombination, encompassing a region of chromosome 6 comprising the entire HLA locus. We show that donor T cells capable of recognizing the original, HLA-heterozygous, leukemia were efficiently transferred from the haploidentical donor to the patient, granting an in vivo cytotoxic, cytokine and proliferative anti-tumor response by specific recognition of the mismatched HLA molecules. However, consistent with genomic loss of the patientspecific HLA locus in disease recurrence, the same alloreactive T cells were unable to recognize the mutant variant of the leukemia, harvested at the time of relapse. This observation strongly suggests that the genomic rearrangements we identified granted the disease an in vivo selective advantage in escaping from an established donor T cell response. Taken together, our data show that adoptive transfer of alloreactive donor T cells in haploidentical HSCT is efficient in providing a patient-specific antileukemic effect, and that the loss of this effect is an important mechanism underlying the outgrowth of relapsing disease. The frequency we documented for this phenomenon calls for routine assessment of the leukemia HLA genotype in the post-transplant follow-up and for careful consideration in the choice of a putative second haploidentical donor in case of leukemia relapse. Ultimately, our data provide the first direct evidence for the role of donor T cell alloreactivity in controlling minimal residual disease after haploidentical HSCT, favoring the use of donor T cell-based immunotherapeutic strategies to exploit alloreactivity for the cure of high-risk leukemia.