HLA-A3 increases and HLA-DR1 decreases the risk of acute graft-versus-host disease after HLA-matched sibling bone marrow transplantation for chronic myelogenous leukaemia


Dr Richard E. Clark, Department of Haematology, Royal Liverpool University Hospital, Prescot Street, Liverpool L7 8XP, UK. E-mail: clarkre@liverpool.ac.uk


Frequencies of human leucocyte antigens (HLA)-A, -B and -DR were determined in 751 patients with chronic myelogenous leukaemia (CML) reported to the European Group for Blood and Marrow Transplantation after bone marrow transplantation from HLA-identical family donors and related to the occurrence of graft-versus-host disease (GVHD). HLA-A3 and DR1 were significantly associated with acute GVHD, the first with a higher risk (44% in HLA-A3+ versus 34% in HLA-A3 patients) and the latter with a lower risk (28% in HLA-DR1+ versus 38% in HLA-DR1 patients) for developing acute GVHD grade II–IV. Both factors were independent of known variables for GVHD as shown in a multivariate analysis. The results show that MHC alleles independently influence the incidence of GVHD in bone marrow transplantation from an HLA-identical donor for first chronic-phase CML. Possible mechanisms might include an HLA antigen-specific allele-associated effect, and/or non-specific allele-associated immune hypo- or hyper-responsiveness.

Graft-versus-host disease (GVHD) remains a serious complication determining the morbidity and mortality of allogeneic bone marrow transplantation (BMT) (Thomas et al, 1975; Hansen et al, 1981; Storb et al, 1983a; Sullivan et al, 1989; Martin et al, 1990; Ferrara & Deeg, 1991). Recently, considerable progress has been made in identifying the mechanisms and structures involved in the pathogenesis of GVHD (Niederwieser et al, 1993; den Haan et al, 1995). Three phases can be identified. First, the cumulative effects of the underlying disease, its previous treatment and the transplant-conditioning chemoradiotherapy may all confer a greater risk of GVHD. This may be mediated by upregulation of human leucocyte antigens (HLA) and other molecules on the surface of host cells. Second, transplanted donor-derived immunocompetent cells recognize HLA differences on host cells. In mouse models, CD8+ T cells induce GVHD to class I differences, and CD4+ T cells induce GVHD to class II disparities (Korngold, 1992); both subsets also appear important in producing GVHD after allografting in humans (Schlegel et al, 1994; Lazarus et al, 1997). Third, extensive cytokine release may recruit large granular lymphocytes or natural killer cells, and these may all augment local tissue injury caused by cytotoxic T cells (Barrett & Malkovska, 1996).

The initiating events in GVHD are the recognition of relevant target antigens on host cells by donor-derived immunocompetent cells. In mismatched and unrelated donor transplantation, HLA disparity between host and donor may trigger GVHD. However, in the setting of HLA-identical sibling transplants, the recipient must express other non-HLA tissue antigens not present in the donor, termed minor histocompatibity antigens (MiHA). Human MiHA are polymorphic antigens that are inherited independently from HLA, show broad or restricted tissue distribution, and are recognized by alloreactive T cells (Niederwieser et al, 1993; den Haan et al, 1995). The MiHA HA-1 and HA-2 are expressed exclusively on haemopoietic cells (Niederwieser et al, 1993; Mutis et al, 1999), and the peptide sequence of these antigens has been defined (den Haan et al, 1995). GVHD-inducing MiHA are peptides derived from naturally processed intracellular proteins, enclosed in the groove of MHC antigens, which biochemically bind and carry the peptide from inside the cell to the surface, making them accessible to T lymphocytes. At present, only a few peptides, restricted predominantly by one MHC allele (HLA-A2), have been sequenced. In chronic myelogenous leukaemia (CML), certain peptides may bind to some HLA molecules more avidly than others. On theoretical grounds, 21 peptides, 8–11 amino acids in length, spanning the b3a2 and b2a2 BCR-ABL junction have been predicted to bind well to common HLA class I binding motifs, and four of these (all spanning the b3a2 junction) have been shown to bind well to HLA A3, A11 and B8 alleles (Bocchia et al, 1995, 1996). These observations may be relevant to clinical CML because HLA-B8 expression is associated with a decreased incidence of CML, especially if HLA-A3 is co-expressed (Posthuma et al, 1999), and HLA-DR3, DR4 and DR15 also lower the risk of CML, although HLA-DR3 is in linkage disequilibrium with HLA-B8 (Posthuma et al, 2000).

Therefore, it is possible that the incidence of GVHD may vary according to HLA type because HLA molecules differ in their ability to present relevant MiHA to the incoming donor-derived T cells. Similarly, if a leukaemic cell expresses a leukaemia-specific peptide on the cell surface, certain HLA molecules may present this more avidly than others, resulting in differing graft-versus-leukaemia (GVL) effects. We have examined this in the present study by looking at the effect of HLA type on the incidence of GVHD among patients with chronic-phase CML. Only transplants from an HLA-identical sibling were included in the analysis, to minimize the effect of alloreactive T cells on ′foreign′ HLA molecules. We have also analysed the effect of HLA alleles on leukaemia-free survival (LFS), transplant-related mortality (TRM) and relapse incidence (RI) in this patient cohort. Using univariate and multivariate analyses, we report that MHC alleles influence the incidence of GVHD, independently from other GVHD risk factors.

Patients and methods

Study design and data collection The present analysis is a retrospective study based on transplants performed between 1980 and 1994. The data were collected using questionnaires by the European group for Blood and Marrow Transplantation (EBMT) and contained information on donor and recipient identity, sex, age and histocompatibility. The vast majority of the transplants were carried out before the advent of molecular techniques for HLA typing. Data on primary disease, transplant procedure, conditioning, GVHD prevention method and outcome were also available, collected annually and updated on 1 January 1995. For logistic reasons, some teams did not report all their patients: some teams ceased to report, whereas some only began reporting at a later stage. The participating centres are listed in the Appendix.

Patients The present analysis concentrated on 751 patients with CML, transplanted in the first chronic phase of their disease with bone marrow from an HLA-identical related donor. Table I summarizes the clinical characteristics of the patients transplanted in 62 bone marrow transplant centres with regard to age, year of transplant, recipient and donor sex, recipient–donor sex combinations and GVHD prevention methods. The median age of the patients was 34 years, ranging from 0·5 to 62 years, and the majority of the 751 patients (456, 61%) were transplanted after 1987. Fifty-nine per cent of the 751 patients were men, and 23% of them received a graft from a female donor. The majority of the patients (65%) received either cyclosporine (CyA) alone or in combination with methotrexate (MTX) as GVHD prophylaxis, and in 27% of patients, donor marrow was T cell-depleted. HLA-A typing was available in 737, HLA-B in 735 and HLA-DR in 531 of the 751 patients.

Table I.  Clinical characteristics of patients with CML receiving a matched related BMT between 1980 and 1994, reported to the EBMT.
Total number of patients751 
BMT centres62 
Age (median; range in years)34 (0–62) 
Year of transplant:
Recipient sex:
Donor sex:
Recipient–donor sex combination:
 Donor female; Recipient male17423
 Other permutations57777
GVHD prophylaxis:
 CyA ± MTX48965
 T-cell depletion and other20027

Grading of acute GVHD The participating teams were asked to enter the grade of acute GVHD according to the Seattle criteria (Thomas et al, 1975). GVHD grade 0–1 was considered as absent, GVHD grades II–IV as present.

Statistical analysis Screening for associations between HLA-A, HLA-B and HLA-DR antigens and acute GVHD incidence was carried out using chi-square statistics with Yates correction. HLA alleles, correlating to GVHD in the chi-square statistics with a P value < 0·10, were then included in a stepwise multivariate logistic regression analysis to investigate the predictive power for acute GVHD for some combinations of these HLA alleles. The HLA alleles found predictive in this logistic regression were examined, again with a multivariate logistic regression, for their predictive power for acute GVHD, additional to some known acute GVHD risk factors.


We first screened for an association between low-resolution HLA-A, -B and -DR alleles and acute GVHD (Table II). Cross-tabulating each HLA allele with the occurrence of GVHD, four alleles were found to be associated either with the presence or absence of GVHD with P ≤ 0·10. These HLA alleles were A3, A11, B7 and DR1 as shown in Table IIA. Applying a stepwise logistic regression, two of these four variables remained associated with GVHD: HLA-A3 (relative risk 1·52; 95% CI 1·02–2·28; P = 0·03) and HLA-DR1 (relative risk 0·59; 95% CI 0·35–0·98; P = 0·03). Therefore, HLA-A3 was a risk factor and HLA-DR1 a protective factor for acute GVHD (Table IIB).

Table II.  (A) Results of screening for association (i.e. P ≤ 0·10) with occurrence of acute graft-versus-host disease. (B) Results of stepwise logistic regression of the four variables found from the screening in Table IIA. : (A)

HLA allele

present (%)

present (%)


Combining all possible combinations of HLA-A3 and DR1 with GVHD, A3 was found to be the strongest risk factor, irrespective of HLA-DR1 status. The probability to develop GVHD grade II–IV in patients having HLA-A3 with or without DR1 was 42·0% and 41·2% respectively. Patients showing HLA-DR1 without A3 had a GVHD incidence of only 18·6%, while those lacking both alleles had a GVHD incidence of 36% (Table III).

Table III.  Association of combinations of HLA-A3 and HLA-DR1 with GVHD (n = 531) (P = 0·02).
   GVHD grade II–IV
HLA-A3HLA-DR1Patients nN%

To determine whether the association of HLA-A3 and DR1 with the occurrence of GVHD was an independent variable, a multivariate analysis was performed using factors that previously had been identified as predictors for acute GVHD (recipient age, GVHD prevention method, sex combination and year of BMT) (Gale et al, 1987). The stepwise logistic regression confirmed recipient age, T-cell depletion and BMT year as predictive variables for GVHD. Then HLA-A3 and -DR1 were added as variables to the model, with the finding that each of these alleles conferred an additional prognostic value (Table IV).

Table IV.  Multivariate analysis for acute GVHD using recipient age, T-cell depletion, year of BMT, presence/absence of HLA-A3 and DR-1 as prognostic variables.
 nRR95% CIP
  1. Abbreviations as for Table II.

Recipient age (years)
 < = 20611·00  
 > 204702·001·06–3·78< 0·05
T-cell depletion
 Yes1600·370·24–0·58< 0·05
BMT year
 1988–943190·560·38–0·83< 0·05
 Positive1461·571·04–2·38< 0·05
 Positive930·570·34–0·98< 0·05

The effect of HLA alleles on overall survival (OS), LFS, RI and TRM was investigated. As shown in Table V, HLA variables were not associated with significant changes in any of these endpoints at 5 years. Only patients with HLA-DR1 showed a trend towards better OS, LFS and less TRM than patients without this allele, but the difference was not statistically significant.

Table V.  Effect of HLA-A3 and DR1 on overall survival, leukaemia-free survival, relapse incidence and transplant-related mortality at 5 years.
Total cases5431804329659472
Overall survival (%)515251626651
Number at risk873672201379
P0·42 0·16 0·13 
Leukaemia-free survival (%)404239494840
Number at risk731158171164
P0·39 0·18 0·26 
Relapse incidence (%)333136313335
Number at risk733158171164
P0·76 0·73 0·55 
Transplant-related mortality (%)404139302939
Number at risk733158171164
P0·42 0·18 0·1 


The data presented here suggest that MHC alleles are independent determinants for acute GVHD in adults undergoing allogeneic BMT, even if the donor is an HLA genotypically matched sibling. In our homogeneous patient population with CML, HLA-A3 was associated with a higher, and HLA-DR1 with a lower, risk of developing GVHD in multi as well as univariate analyses.

During the last decade, considerable progress has been made towards clarifying the molecular events involved in the processing and presentation pathways used by MHC molecules (Germain & Margulies, 1993, Howard, 1995). Crystallographic structures of several peptide-MHC molecules have been determined, and sequence motifs associated with naturally processed peptides and high-affinity binding peptides have been derived (Madden, 1995). Using direct binding assays with radiolabelled peptides and HLA-class I expressing cells, it has been shown that a given peptide may bind more avidly to certain HLA alleles than to others, although a certain degree of cross-reactivity is present (Sidney et al, 1996). Varying degrees of binding of MiHA-derived peptides (Loveland et al, 1990; Speiser et al, 1990; Niederwieser et al, 1993; den Haan et al, 1995) to HLA molecules might contribute to different frequencies of clinically significant (grade II–IV) GVHD between HLA types. In the present context, HLA-A3 might bind relevant minor immunogenic peptides more frequently than HLA-DR1, and there is some evidence that HLA-A3 molecules will strongly bind a peptide of sequence KQSSKALQR derived from the b3a2 BCR-ABL junction (Bocchia et al, 1995, 1996). One additional possibility is that HLA-DP (which is not in linkage disequilibrium with the alleles studied here) might be more highly mismatched in the context of the HLA-A3 allele, and less mismatched if associated with HLA-DR1. Techniques for the high-resolution study of HLA-DP have only recently been developed and are not available for the present data. This association merits further study.

HLA alleles have also been shown to influence immune responsiveness in a non-specific manner. MHC-linked genes differentially modulate non-specific inflammatory mediators such as histamine, prostaglandin, corticosteroids and cytokines, which are important in the pathophysiology of GVHD. One cytokine crucially involved in the pathogenesis of GVHD is tumour necrosis factor alpha (TNF-α), whose gene is located on chromosome 6 within the MHC loci (Holler et al, 1990; Jacob, 1992). The TNF-α gene is polymorphic and associated with different HLA alleles (Messer et al, 1991). The HLA-DR1 associated TNF haplotype has been described to be a low-secretor phenotype, and is associated with low TNF-αsecretion compared with other HLA alleles (Bouma et al, 1996). The reduced secretion of TNF-α might be responsible for the protective effect of HLA-DR1 on GVHD incidence. Conversely, certain class II alleles may confer increased immune reactivity; for example, the DR6 allele has been associated with an increased risk of renal transplant rejection (Hendriks et al, 1986).

Associations between HLA alleles and the occurrence of acute GVHD have previously been reported. Storb et al (1983b) found that in 130 patients engrafted with marrow from HLA-identical siblings for severe aplastic anaemia, HLA-B18 was associated with an increased risk of GVHD, and HLA-B8 or B35 was associated with a decreased risk of GVHD. In a separate study, HLA-A26 was reported to be associated with an increased and DR3 with a decreased risk of GVHD in 469 patients transplanted from HLA-matched siblings (Weisdorf et al, 1991). Smyth et al (1993) reported a decreased risk of GVHD in patients with HLA-B7 and an increased risk in HLA-B44 for 51 patients with different diseases. However, in each of these later reports, only 25–30% of patients suffered from CML; the present study of 751 patients is by far the largest study of HLA associations with GVHD in the first chronic phase of CML. It is now clear that the underlying disease and its treatment are risk factors for GVHD. The treatment of chronic-phase CML prior to BMT is such that most patients are unlikely to have received intensive or repeated courses of chemotherapy, unlike the situation for most other haematological illnesses treated by BMT. This may explain why the HLA associations described here for CML differ from previous studies focusing on other diseases.

In summary, we report that the HLA type can be added to the list of factors that influence the risk of GVHD after HLA-identical allogeneic BMT in the first chronic phase of CML. This effect might be caused by an antigen-specific allele-associated mechanism and/or a non-specific allele-associated immune hypo- or hyper-responsiveness. Further studies are required, focusing on the effect of individual MiHA on transplant outcome.


This work was financially supported by Grants of the Austrian Research Fund ‘Zur Förderung der Wissenschaftlichen Forschung’, project no. 10525.


The following institutions contributed data to the study. Austria: Universitätsspital, Innsbruck (D. Niedenwieser); Universitätsklinik fur Innere Medizin I, Vienna (W. Linkesch, P. Kalhs). Belgium: Clinique Universitaire St Luc, Brussels (A. Ferrant, C. Vermylen); University Hospital, Brussels (B. van Camp, A. Schors); University Hospital, Leuven (M. A. Boogaerts). Denmark: Rigshospitalet, Copenhagen (N. Jacobsen). Finland: University Hospital, Helsinki (T. Ruutu); University Central Hospital, Turku (K. Remes). France: Centre Hospitalier Régional, Caen (O. Reman); Hôpital Henri Mondor, Creteil (C. Cordonnier); Centre Hospitalier, Grenoble (J. J. Sotto, L. Molina, F. Nicolini); Hôtel Dieu, Paris (R. Zittoun, B. Rio); Hôpital du Haut-Léveque, Pessac (J. Reiffers, C. Fabéres); Hôpital Etienne, St Etienne (D. Guyotat, J. L. Stephan). Germany: Charité Virchow Klinikum, Berlin (W. Siegert); Medizinische Hochschule, Hannover (A. Ganser, B. Hertenstein); Christian-Albrechts-Universität, Kiel (N. Schmitz); Klinikum Grosshadern, Munich (H.-J. Kolb); Medizinische Universitäts-Klinik, Ulm (D. Bunjes). Hungary: National Institute of Haematology, Budapest (E. Kelemen, K. Palocz, R. Denes). Ireland: St James's Hospital, Dublin (S. R. McCann). Israel: Hadassah University Hospital, Jerusalem (S. Slavin). Italy: Ospedale Maggiore di Milano (G. Lambertenghi Deliliers); Ospedale di Niguarda, Milano (P. Marenco, R. Cairoli)); Clinica Oncoematologia Paediatrica, Padova (C. Messina); Policlinico S Matteo, Pavia (C. Bernasconi); Universita ‘La Sapienza’, Rome (W. Arcese, F. Mandelli, G. Meloni); Universita Cattolica, Rome (S. Cuore, S. Sica, G. Leone); Ospedale S. Camillo, Rome (A. De Laurenzi); Istituto per l'Infanzia Burlo Garofolo, Trieste (M. Andolina). Netherlands: University Medical Centre, Leiden (J. Vossen, R. Willemze, W. E. Fibbe, J. J. van Rood); University Hospital, Maastricht (H. C. Schouten); University Hospital, Nijmegen (A. Schattenberg, T. de Witte, J. Groot, L. Beek); Daniel den Hoed Cancer Centre, Rotterdam (J. Cornelissen); University Hospital, Utrecht (L. F. Verdonck). Norway: Rikshospitalet, Oslo (D. Albrechtsen). Poland: K. Dluski Hospital, Wroclaw (A. Lange). Portugal: Instituto Portugues de Oncologia, Lisbon (M. Abecasis). Saudi Arabia: King Faisal Hospital, Riyadh (P. Ernst). Spain: Santa Creu, Barcelona (I. Badell Serra, J. Cubells-Riero, A. Domingo Albos, C. Sola); Hospital Clinic, Barcelona (E. Montserrat, E. Carreras); Hospital Reina Sofia, Cordoba (A. Torres Gómez); Hospital Universitario M de Valdecilla, Santander (A. Iriondo, E. Conde). Sweden: Huddinge Hospital, Huddinge (P. Ljungman); University Hospital, Lund (A. N. Bekassy); University Hospital, Uppsala (B. Simonsson, K. Calson). Switzerland: Kantonsspital, Basel (A. Gratwohl); Hospital Cantonal Universitaire, Geneva (B. Chapuis, J. Humbert). Turkey: Ibni Sina Hospital, Ankara (H. Koc); Hacettepe University Hospital, Ankara (K. Oezerkan). UK: Queen Elizabeth Hospital, Birmingham (J. A. Holmes); Royal Infirmary, Edinburgh (A. C. Parker, M. J. Mackie, P. Johnson); Hammersmith and Charing Cross Hospital, London (J. M. Goldman, D. Samson); University College Hospital, London (A. H. Goldstone); The London Clinic (P. J. Gravett); Royal Marsden Hospital, London (R. Powles); Royal Free Hospital, London (H. G. Prentice, M. Potter); Royal Children's Hospital, Manchester (A. M. Will).