Study population Between May 10, 1985, and August 8, 1999, 572 patients received a marrow transplant at our centre from unrelated donors for the treatment of chronic myeloid leukaemia (CML) in chronic phase, accelerated phase or blast phase in remission. In an effort to reduce the effects of additive HLA disparities and of known risk factors for GVHD and survival (Hansen et al, 1998; Petersdorf et al, 1998), we restricted the current analysis to consecutive transplants who satisfied the following criteria (n = 205): (i) donors and recipients were allele-matched for graft-vs.-host recognition at HLA-A, -B, -C, -DRB1 and -DQB1; (ii) patients were prepared for transplantation with the use of a myeloablative regimen consisting of cyclophosphamide and total body irradiation; (iii) patients received T-replete marrow grafts using cyclosporin and methotrexate for GVHD prophylaxis; and (iv) DNA was available for retrospective DPB1 sequencing for both the recipient and the donor.
HLA typing HLA-A, -B and -C alleles were sequenced using direct automated fluorescence sequencing methods as described (Petersdorf et al, 1998). HLA-DRB1 and -DQB1 alleles were typed using sequence-specific oligonucleotide probe-based methods (Petersdorf et al, 1996). HLA-DQA1 and HLA-DPA1 were not given consideration in this study owing to the strong positive linkage disequilibrium with DQB1 and DPB1 respectively (Begovich et al, 1992; Spurkland & Vartdal, 1992; Sage et al, 1994; Hurley et al, 2000). Typing for HLA-DPB1 alleles in the current study was performed using direct automated fluorescence sequencing. HLA-DPB1 genes were amplified using 100–200 ng of genomic DNA, 0·75 U of AmpliTaq DNA Polymerase, and either 22·5 μl of HLA-DPB1 reverse polymerase chain reaction (PCR) ready mix (kind gift from Dr Malcolm McGinnis, Applied Biosystems) or buffer containing 10 mmol/l Tris-HCl (pH 8·3), 50 mmol/l KCl, 1·5 mmol/l MgCl2, 1·5 μl of 2·5 mmol/l dATP, dCTP, dGTP and dTTP, and 25 pmol of each primer DPB5 and DPB3 (Rozemuller et al, 1996). Initial denaturation at 98°C for 2 min was followed by 32 cycles of denaturation at 96°C for 45 s, annealing at 65°C for 1 min and extension at 72°C for 2 min. The final extension at 72°C for 10 min was followed by 4°C hold.
Amplified products were purified using Multiscreen-PCR filter plates according to manufacturer's protocols (Millipore, Bedford, MA, USA) with the following modification: PCR products were eluted with 75 μl of ddH2O and concentrated to 5–10 μl volume using Speed Vac Plus SC110A Concentrator (Savant Instruments, Holbrook, NY, USA). Purified PCR products were analysed using a direct fluorescence sequencing method as previously described (Petersdorf et al, 1999). The primer DPB5F (Rozemuller et al, 1996) was used to sequence exon 2 of the DPB1 gene in the 5′ orientation.
Sequencing data were manually reviewed, and HLA-A, -B and -C alleles were assigned by using alignments with known sequences. HLA-DPB1 sequence data were analysed using MatchTools PPC program (Applied Biosystems, Foster City, CA, USA) and by manual review.
When more than one possible combination of two DPB1 alleles could have resulted in the same chromatogram, allele-specific amplification was performed as follows: primer pair GSAP107C/DPB3 for heterozygous combinations DPB1*0401,0402/2301,5101; DPB5/GSAP205C for heterozygous combinations DPB1*02012,0401/0402,3301, DPB1*02012,2001/0402,0601, DPB1*02012,3501/0402,0901, DPB1*0301,1001/0901,2501/1401,3701, DPB1*0401,0601/ 20011,3301 and DPB1*0401,0901/3301,3501; primer pair DPB5/GSAP226C for heterozygous combinations DPB1*02012,1401/4501,4601, DPB1*02012,0301/0402, 2901/2501,4601, DPB1*02012,0901/1001,4601, DPB1* 0301, 0601/20011,2901, DPB1*0301,1601/0801,2001, DPB1*0301,1701/0601,3501/0901,2001 and DPB1*0202, 1001/4801,5401; and primer pair DPB5/GSAP256G for DPB1*0401,0501/2401,6301. Direct fluorescence sequencing was performed using primers DPB5F or DPB3F. Primer sequences have previously been described in detail (Versluis et al, 1995; Rozemuller et al, 1996). One new allele-specific primer DP03/26 (5′cagagaattacgt gtaccagttac3′, located at exon 2 position 11–34) was designed to resolve two heterozygous combinations DPB1*01011,0301/DPB1*26012,5001 and DPB1*01011, 20011/DPB1*2701,5001.
Transplant procedure All patients were prepared for transplantation with intravenous cyclophosphamide (60 mg/kg recipient body weight) administered on each of two successive days followed by total body irradiation (900–1575 cGy determined by phase of disease) from dual opposing 60Co sources. Prophylaxis for cytomegalovirus (CMV), fungal and Pneumocystis carinii infection was administered as described (Hansen et al, 1998). All protocols and consent forms were reviewed and approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center. The severity of acute GVHD was assessed according to criteria previously described (Petersdorf et al, 1998).
Statistical methods The major objective of this retrospective study was to test the hypothesis that HLA-DPB1 disparity increases the risk of clinically significant grades III–IV acute GVHD after unrelated donor stem cell transplantation for CML. Mismatching at HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1 was defined as the presence of recipient alleles not shared by the donor (recipient disparity). The probability of grades III–IV GVHD was estimated among patients mismatched for 0, 1 and 2 HLA-DPB1 alleles, and odds ratios were obtained using logistic regression. In addition, the distribution of grades of GVHD was compared among groups using the Kruskal–Wallis test. The hazard of mortality and appropriate hazard ratios were estimated using proportional hazards regression. As HLA-DPB1 was not considered during donor selection, potential imbalances in non-HLA risk factors between groups defined by number of mismatched HLA-DPB1 alleles should have occurred only through chance. To account for this possibility, however, multivariable regression models were fitted by considering non-HLA risk factors previously shown to be associated with increased GVHD and mortality. Estimates of the probability of survival were obtained using the method of Kaplan and Meier (1958), and the probability of GVHD was summarized using cumulative incidence estimates (Gooley et al, 1999), in which death without GVHD was regarded as a competing risk for GVHD. Global tests were conducted to examine the difference in distribution of GVHD grades (using the Kruskal–Wallis test), probability of severe GVHD (score test) and the hazard of mortality (score test). If the global tests yielded significant or suggestive differences, selected pairwise comparisons were made. Two-sided P-values resulting from regression models appropriate to these comparisons were derived from the Wald test and no adjustments were made for multiple comparisons.