The GvHD Hub uses cookies on this website. They help us give you the best online experience. By continuing to use our website without changing your cookie settings, you agree to our use of cookies in accordance with our updated Cookie Policy

HLA-DPB1 mismatch models after unrelated donor stem cell transplantation

Oct 16, 2019

In patients with hematological malignancies, hematopoietic stem cell transplantation (HSCT) from unrelated donors (URD) is often used when a suitable related donor is not available. Human leukocyte antigen (HLA) matching is performed to find the closest matching donor to ensure the best possible outcomes in terms of overall survival (OS), non-relapse mortality (NRM) and graft- versus-host disease (GvHD). High-resolution HLA typing typically looks at specific loci (HLA-A, -B, -C, -DRB1) to determine the allele variant for donor and recipient, with mismatching of these adversely affecting outcomes. 1However, genotype matching for HLA-DPB1 (DPB1) has been found to have little impact on OS. Also, there is weak linkage disequilibrium between DPB1 and other HLA class II loci, which would make it difficult to implement. As a result, it is estimated that >80% of 8/8-matched URD-HSCT (HLA-A, -B, -C, -DRB1) are mismatched for DPB1.  There are several biological models used to determine which HLA-DPB1 mismatches are tolerable (permissive) or convey an increased risk (non-permissive) based on functional T-cell Epitope (TCE) prediction, and single nucleotide polymorphism (SNP) tags. 2

In a recent letter to Haematologica, Francesca Lorentino from the Hematology and Bone Marrow Transplantation Unit, IRCCS San Raffaele Scientific Institute,Milano, IT, and colleagues compared DPB1 permissiveness, based on five models, and how they are associated with outcomes in 422 HSCT patients. 2The objective was to identify a prediction model which is able to select permissive DPB1 mismatch combinations, associated with lower clinical risks compared to their high-risk, non-permissive, counterparts.

Study design and Patient characteristics

  • Models included
    • TCE prediction
      • TCE3 – Alleles are classified based on variation in the structure at the peptide antigen-binding domain
      • TCE4 – the same as the TCE3 model, but the classification is different in the assignment of DPB1*02
      • ΔFD – another derivative of TCE3 which classifies DPB1 alleles based on anti-DPB1 alloreactivity and differences between patient and donor
    • Prediction based on single nucleotide polymorphism (SNP) tags
      • Expression model – SNP tag applied to all DPB1 alleles to determine the expression level
      • DP2/DP5 – SNP tag applied to 19 alleles from the DP2 or DP5 groups of alleles
    • 422 Patients with available 2 ndfield DPB1 typing transplanted from 8/8 HLA-A, -B, -C and -DRB1 allele matched UDs were analyzed from 32 centers between 2012 and 2015 (patient characteristics are in table 1)
    • 382 patients had one or two DPB1 mismatches
Table 1:Patient, transplant and donor characteristics
AML; acute myeloid leukemia, ALL; acute lymphoblastic leukemia, MDS; myelodysplastic syndromes, MPN; myeloproliferative neoplasms, CLL; chronic lymphocytic leukemia, CMV; Cytomegalovirus, MAC; myeloablative conditioning, RIC; reduced-intensity conditioning, PB; peripheral blood, BM; bone marrow, ATG; anti-T-lymphocytic globulin, CSA; cyclosporine A, MTX; methotrexate.
  Population n= 422
Median follow-up for survivors, years (range) 3.2 (0.1—6)
Patient age, years, median (range) 49 (18 —70)
Patient gender, male, n (%) 244 (58%)

Type of diagnosis, n (%)




Lymphoma and Myeloma



168 (40%)

63 (15%)

69 (16%)

110 (26%)

12 (3%)

Disease status according to EBMT risk, n (%)





191 (45%)

111 (26%)

120 (29%)

HCT-CI score, median (range) 1 (0-7)
Karnofsky performance status, median % (range) 90 (50—100)
Donor gender, male, n (%) 306 (72%)
No of previous pregnancies for female donors, median (range) 0 (0-6)
Female donor/male recipient, n (%) 61 (14%)

Host/donor CMV serostatus, n (%)







157 (37%)

166 (39%)

36 (9%)

53 (13%)

10 (2%)

Type of conditioning, n (%)




271 (64%)1

11 (35%)

Source of stem cells, n (%)




343 (81%)

79 (19%)

ATG-based GvHD prophylaxis, n (%) 382 (91%)

GvHD prophylaxis details:


ATG + Sirolimus + MMF

Other ATG-based prophylaxis


Other prophylaxis


341 (81%)

26 (6%)1

5 (4%)

24 (5%)

16 (4%)


  • There was limited overlap between the five models
  • TCE4 was the most restrictive, classifying the fewest pairs as permissive (36%)
  • SNP models were only able to classify pairs with a single unidirectional DPB1 mismatch (towards GvHD), leaving 153/382 (40%) and 233/382 (61%) of pairs without classification with the expression model and the DP2/DP5 model respectively
  • Pairs in the permissive and non-permissive groups were comparable in terms of disease and transplant characteristics
  • Univariate analysis:
    • DPB1 mismatches were not associated with significant changes in OS
    • There was a higher risk of acute GvHD (aGVHD) but a lower risk of relapse with DPB1 mismatches
    • TCE4 was the only model which found a significantly superior OS in permissive versus non-permissive classified donors
  • Multivariate analysis:
    • Using the TCE4 model, permissive pairs had superior OS and GvHD relapse-free survival (GRFS), and were found to be at lower risk of NRM and chronic GvHD (cGvHD)
    • According to the three functional models (TCE3, TCE4 and ΔFD), permissive mismatches were associated with lower relapse risks than DPB1 allele matches
    • The SNP expression model and the DP2/DP5 model were able to show an association of high-risk, non-permissive mismatches with grade II — IV aGvHD, but not with NRM or OS

The team highlight that this study is the first to compare these five models of DPB1 permissiveness and that despite some of the models (TCE3, TCE4, ΔFD) describing the same interaction, they are not in complete agreement. The group suggest that this may be due to each of the models capturing only certain aspects of T-cell interaction with DPB1 molecules. They go on to discuss the functional TCE models as being better in predicting survival and cGvHD, while the SNP models being predictive of aGvHD. In terms of study limitations, Lorentino and colleagues felt that these included the small number of pairs included in the SNP tag models, and the lack of agreement with other studies around the TCE3 model associations, possibly due to the stem cell source (peripheral blood vsbone marrow) and different conditioning strategies. In conclusion, the authors highlighted the functional basis for TCE4 and its potential as a superior model for permissiveness in URD-HSCT.

  1. Flomenberg N. et al.,Impact of HLA class I and class II high-resolution matching on outcomes of unrelated donor bone marrow transplantation: HLA-C mismatching is associated with a strong adverse effect on transplantation outcome. blood. 2004 Oct 1;104(7):1923-30. DOI:10.1182/blood-2004-03-0803
  2. Lorentino F. et al., Comparative evaluation of biological HLA-DPB1 mismatch models for survival and graft versus host disease prediction after unrelated donor hematopoietic cell transplantation. Haematologica. 2019 Aug 30. DOI: 10.3324/haematol.2019.225177.[Epub ahead of print]