A high resolution HLA class I and class II matching method for bone marrow donor selection

Transcription

1 Bone Marrow Transplantation, (1998) 22, Stockton Press All rights reserved /98 $ A high resolution HLA class I and class II matching method for bone marrow donor selection JR Argüello 1, A-M Little 1,2, E Bohan 1, D Gallardo 1, J O Shea 1, IA Dodi 1, JM Goldman 1,3 and JA Madrigal 1,2,3 1 Anthony Nolan Research Institute and 2 Department of Haematology, The Royal Free Hospital School of Medicine; and 3 The Imperial College School of Medicine, London, UK Summary: HLA matching in bone marrow transplantation has an important role in determining successful outcome. However HLA typing of both potential related and unrelated donors can be both time-consuming and laborious, and does not always resolve accurately the true level of histocompatibility. We have utilised a method, reference strand mediated conformation analysis (RSCA), which is technically simple and allows high resolution matching for all HLA loci, for the typing of 48 patients and their potential 120 donors. The results indicate that RSCA can detect many mismatches that are not routinely identified by conventional HLA typing methods. In addition, RSCA can be applied for the simultaneous analysis of multiple potential BM donor samples in order to quickly identify the best match for each patient. Keywords: HLA matching; bone marrow donor selection; RSCA Bone marrow transplants performed with unrelated donors have been shown to result in higher rates of graft failure, increased incidence and severity of GVHD and increased susceptibility to infections, when compared with sibling donor transplants. These differences between related and unrelated donors are likely the result of genetic differences including HLA disparities which are not identified prior to transplantation due to limitations in the HLA typing methods utilised. 1 It is now widely accepted that high resolution HLA typing is required for the identification of the best HLA-matched donor. However, it is becoming increasingly difficult to employ a single HLA-typing technique for the identification of the currently defined 825 HLA class I and II alleles (SGE Marsh, personal communication). There are therefore significant problems in the selection of the best potential bone marrow donor from the world-wide pool of over 4.6 million registered volunteer donors who have been HLA typed, albeit to varying degrees of resolution using a variety of different methods. 2 The ideal procedure would be to type all potential donors simultaneously using Correspondence: JA Madrigal, The Anthony Nolan Research Institute, The Royal Free Hospital, Pond Street, London NW3 2QG, UK Received 3 February 1998; accepted 22 April 1998 high resolution DNA typing with speed, accuracy and low cost. We have recently described a high resolution HLA class I typing method: reference strand mediated conformation analysis (RSCA), which utilises simple DNA manipulations, together with automated instrumentation and computer software, to detect differences in DNA conformation reproducibly, allowing the identification of alleles which differ by as little as one nucleotide. 3 Here we present the application of RSCA, for the quick and accurate assessment of the level of both HLA class I and II matching between patient and donor. Using RSCA, a population of potential donors either related or unrelated can quickly be screened for compatibility for eight polymorphic HLA loci. This method substantially reduces the amount of time spent on HLA typing potential bone marrow donors individually and in addition offers matching results at high resolution for all eight loci targeted. Materials and methods DNA samples DNA was extracted by conventional techniques from PBMC of 48 patients and 120 potential related or unrelated bone marrow donors. Eighty-six of the donors were selected on the basis of either being HLA-A, B serological matches or were related to the patient. Thirty-four of the potential donors were in the final stage of donor selection and had been matched for HLA class I and class II by various techniques. PCR PCR amplification was achieved for HLA-A, B, C, DRB, DQA1, DQB1, DPA1 and DPB1 loci using primers and conditions as previously described. 4,5 The PCR cycling conditions for all loci were: 95 C, 4 min; 95 C, 30 s; 65 C, 50 s; 72 C, 30 s; for 33 cycles, and extension at 72 C for 8 min. Locus-specific fluorescently labelled references (FLR) Locus-specific FLRs were prepared from homozygous B- LCLs by PCR as above except the 5 primer for each was labelled with the fluorochrome Cy5 (Pharmacia Biotech

2 528 A high resolution HLA-matching method for donor selection Ltd, St Albans, UK). HLA class I: HLA-A: A*0101 (STEINLIN) and A*0217 (AMALA); HLA-B: B*4402 (SP0010) and B*4201 (RSH); HLA-C: Cw*0701 (STEINLIN) and Cw*0303 (AMALA). HLA class II: HLA-DRB1: DRB1*08021 (SPL) and DRB1*0101 (plasmid pbactinneo DRB1*0101 construct); HLA- DQA1: DQA1*0101 (KAS116) and DQA1*05011 (VAVY); HLA-DQB1: DQB1*0402 (BTB); HLA-DPA1: DPA1*0103 (T5 1) and DPA1*02011 (SAVC); HLA- DPB1: DPB1*02012 (SPOO10) and DPB1*1001 (SAVC). The names in parentheses are the International Histocompatibility cell lines from which the DNA was extracted for use as a template for FLR preparation. 6 Two FLRs were utilised for each locus. For DQB1 analysis, one FLR is the DQB1 allele and the second is the DQB2 allele from the same sample. Duplex formation Duplexes were formed by the addition of 1 l of FLR PCR product ( ng/ l) to 3 l of sample PCR product ( ng/ l) followed by denaturation at 95 C for 4 min, in order to separate the sense and antisense strands of the DNA fragments present in the mixture. The mix was allowed to cool to 55 C for 5 min and 15 C for at least 3 min, thus allowing the complementary sense and antisense strands to re-anneal. In addition, annealing occurs between the sense and antisense strands of the different DNA species present in the mixture to form heteroduplexes, eg the sense strand from the FLR PCR product anneals with the antisense strands from the two alleles present in the sample PCR product (Figure 1). 6X Ficoll (0.8 l) loading buffer (15% Ficoll, 0.25% bromophenol blue) was added prior to electrophoresis. Electrophoresis Duplexes were separated by electrophoresis in an ALFexpress automated sequencer (Pharmacia Biotech). Two microlitres of each duplex sample was loaded in a nondenaturing 6% polyacrylamide gel: 9.6 ml Long Ranger gel solution (JT Baker, Milton Keynes, UK) UK; 8 ml 10 TBE buffer (Biowhittaker, Wokingham, UK) 48 l TEMED and 480 l 10% ammonium persulphate (both Pharmacia Biotech), in a total volume of 80 ml. 3 Electrophoresis was performed at 30 W constant power. Gels were 21 cm long and 0.5 mm thick. The gel temperature was maintained at 40 C during electrophoresis by an external PAGE and laser detection Electropherogram FLR x y Hybridisation FLR x y Hybridisation Donor 1 FLR Hybridisation x y Donor 2 Figure 1 Diagrammatic representation of the RSCA method. The locus specific FLR PCR product, contains a fluorescent Cy5 label on its sense strand indicated by a star. The FLR is hybridised with the locus-specific PCR product from the sample to be tested. For a heterozygous loci, two alleles are present, indicated as x and y. During hybridisation, the sense and antisense strands of the DNA fragments present are initially separated by denaturation followed by re-annealing where the sense and antisense strands can cross-hybridise generating in addition to the starting homoduplexes, heteroduplexes, two of which possess the Cy5 labelled sense strand of the FLR. The duplexes formed are separated by PAGE, and those duplexes possessing a Cy5 label are detected with the laser. In this diagram the patient and donor 1 have alleles with identical mobility, whereas donor 2 has one allele which differs.

3 temperature cooling system (Grant, PLS, Forest Row, UK). Running times were 600 min for HLA class I and 300 min for class II analysis. Gels were reused up to five times with fresh running buffer (1 TBE). Each FLR duplex combination was analysed separately in an individual lane of the gel. Analysis of results The mobility of each labelled duplex was analysed using Fragment Manager software (Pharmacia Biotech). The arbitrary value of 1 was assigned to the high mobility fluorescent primer peak from the PCR reaction which generated the FLR and the value of 1000 was assigned to the fastest duplex which in all cases is the FLR homoduplex. The values of 1 and 1000 set the lower limits of a scale which was extrapolated to encompass the slower fluorescent heteroduplex allowing the assignment of values for the heteroduplex signals. The values of 1 and 1000 were also used for alignment of tracks across each gel. Results RSCA, a recently described method for mutation detection and typing of polymorphic loci, has been utilised for the identification of HLA-matched related and unrelated bone marrow donors. Twenty-six patients and 86 related or unrelated potential bone marrow donors were analysed for HLA class I and II matching by RSCA, and 22 patients and 34 unrelated potential donors for HLA class I matching only. The level of HLA typing which had been previously performed for both patients and donors varied between the loci typed; for example, most individuals were typed to medium/high resolution by sequence specific oligotyping (SSO) for HLA-DRB1 and DQB1 loci, whereas the level of resolution achieved for HLA-A, B and C loci varied between low to medium resolution achieved by either serological methods or sequence-specific PCR (SSP) (Tables 1 and 2). No typing for HLA-DPB1, DQA1 and DPA1 was A high resolution HLA-matching method for donor selection performed except for two patient/donor pairs who were typed for HLA-DPB1 loci by SSP (Table 1). Using RSCA, we have confirmed matching for alleles encoded by HLA-A, B, C, DRB1, DQA1, DQB1, DPA1 and DPB1 loci, for patients and potential donors who had been previously typed using medium to high resolution typing methods. However, for loci which had been typed using low to medium level typing methods, the application of RSCA allowed the identification of additional mismatches which were not detected with the HLA-typing methods used. Table 1 illustrates the HLA-typing results obtained for patient and donor pairs for which both class I and class II loci were analysed. This group of potential donors was considered as they were either related to the patients or they were unrelated HLA-A, B serological matches. Table 2 presents results from the patient and donor pairs for which only HLA class I loci were analysed by RSCA. These patient donor pairs were at the last stage of donor selection and had been typed previously by conventional methods for HLA class I and II loci. Figure 2 illustrates the side by side comparison of HLA- A locus alleles of a patient and three potential HLA-A serologically identical donors by RSCA. Donor 1 possesses a variant HLA-A3 allele not identified by serology. This allele was confirmed to be HLA-A*0302 by direct DNA sequencing (data not shown). Using the electrophoretic conditions described in Materials and methods, which allow the simultaneous analysis of up to 40 samples, as many as 39 potential donors can be compared with a patient in one gel run for a single HLA locus, using one of the two FLRs. Alternatively multiple loci can be simultaneously analysed for fewer donors in one gel run as identical electrophoretic conditions are utilised for all loci so far studied. In Figure 3, HLA-A locus matching between a patient and 13 potential unrelated donors is demonstrated. All donors were typed by serology as being matches with the patient (HLA-A2,3), whereas analysis by RSCA demonstrates that the patient possesses a variant A2 allele which is shared by only one of the 529 Table 1 HLA typing results for 86 potential bone marrow donors and the 26 patients awaiting BMT Typing method HLA loci Match by Mismatch by Match by Mismatch by methods 1 3 methods 1 3 RSCA RSCA A B C DRB1 DQA1 DQB1 DPA1 DPB1 1 Serology 112 a 112 a 329 a 15 b 306 b 38 b (86) (86) 2 SSP 10 a 4 a 16 b 0 15 b 1 b (6) (2) 3 SSO 112 a 112 a 191 b 153 b 185 b 159 b (86) (86) Samples not tested 102 a 112 a 112 a 108 a 468 b 204 b by 1 3 (80) (86) (86) (84) RSCA 112 a 112 a 112 a 112 a 112 a 112 a 112 a 112 a 974 b 402 b (986) (86) (86) (86) (86) (86) (86) (86) a No. of individuals tested (or untested), ie for HLA-A, 86 donors and 26 patients were each typed for HLA-A antigens giving a total of 112. b No. of allele pairs matched or mismatched for each donor/patient pair. For the 26 patients there are 86 potential donors, thus the total number of donor/patient pairs considered is 86. For SSP results, a total of eight donor/patient combinations were studied (six for HLA-C and two for HLA-DPB1) therefore 16 allele pairs were analysed. Homozygous loci counted as two identical antigens or alleles. The numbers in parentheses indicate the total number of donor/patient combinations included in the analysis, ie for HLA-C SSP analysis, six potential donors for four patients were analysed, giving a total of six donor/patient combinations.

4 530 Table 2 A high resolution HLA-matching method for donor selection HLA typing results for 22 potential bone marrow donors and the 34 patients awaiting BMT Typing method HLA loci Match by Mismatch by Match by Mismatch by methods 1 2 methods 1 2 RSCA RSCA A B C 1 Serology 56 a 56 a 132 b 4 b 128 b 8 b (34) (34) 2 SSP 28 a 26 b 4 b 21 b 9 b (15) Samples untested 28 a 32 b 6 b (19) RSCA 56 a 56 a 56 a 181 b 23 b (34) (34) (34) a,b As for Table 1. Homoduplex A*0302 Donor 1 Donor 2 Donor Figure 2 Electropherogram with results for HLA-A locus matching for a patient and three potential unrelated bone marrow donors. The patient and three potential donors were matched by serology for HLA-A2 and A3. By RSCA analysis, donor 1 has an allele with a different mobility when compared with the patient s HLA-A alleles. Control DNA from the cell line CGM, typed as A * 0302 and A * 2902, was analysed in parallel with DNA from donor 1 and the mobility of the A * 0302 allele matched the mobility of the allele from donor 1 (data not shown). Direct sequencing for HLA-A locus confirmed the type of donor 1 (data not shown). Allele signals from donor 2 and 3 match the patient allele signals perfectly. potential donors. The typing of this variant A2 allele as A* 0205 was confirmed by comparison of the patient s DNA to a DNA control for A*0205 (data not shown). Figure 4 illustrates the application of RSCA to distinguish homozygous loci from heterozygous loci. The patient and two potential donors were typed as homozygous for HLA-B44 by serology. However, RSCA analysis demonstrates that neither of the two donors is HLA-B matched with the patient, who is homozygous for B*4402. Donor 1 is heterozygous for two common subtypes of the B44 serological specificity (B*4402 and B*4403) whereas donor 2 is homozygous for HLA-B*4403. RSCA can be used to screen potential related BM donors quickly. In Figure 5, two potential donors are compared with the patient for HLA-DRB loci. Without knowing the HLA type of this family, it is clear that the HLA-DRB products from the daughter of the patient are matched with the patient, whereas the son is mismatched for one of the four alleles. HLA typing analysis confirmed this result as indicated in Figure 5. Four HLA-DRB heteroduplex peaks are present in the electropherogram shown in Figure 5, as the PCR primers used to amplify DRB1 co-amplify DRB3, DRB4 and DRB5 alleles and thus duplexes are formed with the HLA-DRB1 FLR, for both DRB1 alleles and each DRB3, DRB4 and DRB5 allele if present in the DNA sample. There are no DRB3 alleles present in the haplotypes analysed in Figure 5. Similarly in Figures 6 and 7, analysis for HLA-DQB1 and DPB1 compatibility between a patient and three family members illustrates identity at these loci between the patient and the patient s father. For HLA-DQB1 analysis, two constant extra peaks were detected in addition to the homoduplex peak in all samples tested (Figure 6). These extra peaks resulted from the co-amplification of the DQB2 pseudogene with the PCR primers used for both the DQB1 FLR and sample amplifications. Two constant heteroduplexes were thus generated: one (K1) formed between the labelled sense strand of the DQB1 FLR with the anti-sense strand of the DQB2 pseudogene present in both the FLR and the sample. The second constant heteroduplex (K2) was formed between the labelled sense strand of the DQB2 pseudogene and the anti-sense unlabelled strand of the DQB1 FLR. Thus the number of variable heteroduplex signals was 0, 2, 3 or 4 depending on zygosity status and the sharing of alleles with the FLR. For example, a homozygous sample with an allele identical to the FLR produced no variable heteroduplex signals; a sample homozygous with an allele different to the FLR produced two variable heteroduplex signals; a heterozygous sample with one allele identical to the FLR produced three variable heteroduplex signals, etc. In our studies so far we have not found any examples of inter-allelic ambiguity which could be caused by two or more alleles (DRB and DQB analyses) having over-lapping mobilities. Two FLRs must be used for the analysis of each locus to minimise the potential for such ambiguities. In total 39 previously undetected mismatches were identified using RSCA (Table 3), with most mismatches being seen mainly for HLA-A and B loci which had been typed only by serological methods. Discussion Accurate HLA typing is required for selection of the best BM donor for transplantation. As at present more than 825 HLA class I and II alleles are defined, techniques which

5 A high resolution HLA-matching method for donor selection 531 Homoduplex A*0205 D1 D2 D3 D4 D5 A*0205 D6 D7 D8 D9 D10 D11 D12 D Figure 3 Screening of potential HLA-matched bone marrow donors. HLA-A locus-specific amplification was performed on DNA extracted from the patient and 13 potential donors PBMCs, 3 l of the PCR product was hybridised with 1 l ofa*0101 FLR, and 2 l of the mixture was loaded on an automated sequencer. Fragment Manager (Pharmacia) software was used for data analysis. The patient shares one allele () with all potential donors but only shares the second allele, A * 0205 with donor 6. Donor 1 Donor 2 B*4402 B*4402 B*4403 B* Figure 4 HLA-B locus matching. Electropherogram demonstrating matching at HLA-B locus for a patient and two potential donors. The patient is homozygous for B * 4402 whereas donor 1 is heterozygous for two subtypes of B44, one of which is shared with the patient. Donor 2 is homozygous for a different subtype of B44 compared with the patient. The FLR B * 4201 was used to generate heteroduplexes. target nucleotide or amino acid motifs require numerous reagents in order to assign HLA type accurately. However, not all of the 825 defined HLA alleles are present at equal frequencies in all populations therefore it can be considered wasteful to create reagents to define HLA types which may never be encountered in donor selection. Direct sequencing of HLA alleles can give a high resolution type without the need for numerous typing reagents, however as both alleles at heterozygous loci are sequenced together it can be difficult to assign an accurate type for some combinations of alleles. 7,8 With RSCA, the alleles are separated and analysed individually by forming duplexes with the FLR. Thus difficulties in assigning the cis or trans orientation of a sequence motif are not encountered. Conformational techniques such as single strand conformation polymorphism (SSCP) and heteroduplex analysis (HA) have also been used to aid the selection of donors for transplantation. 9 These methods are relatively easy to use and inexpensive. However, the most significant limitations of SSCP and HA are unreliable results due to variability

6 A high resolution HLA-matching method for donor selection 532 DRBI*1501 DRB5*0101 DRB4*0101 DRB1*0701 DRBI*1501 DRB5*0101 DRB4*0101 Daughter DRB1*0701 DRBI*1501 DRBI*0401 DRB5*0101 Son DRB4* Figure 5 HLA-DRB matching for a patient and two related potential bone marrow donors. The primers used for PCR co-amplify DRB1, DRB4 and DRB5 alleles in these DNA samples. The patient s daughter shares all DR alleles with the patient, whereas the son has only three DRB alleles in common. The homoduplex peak is not shown in the Figure, and a cloned DRB1 * 0101 allele was used as FLR. K1 DQBI*0301 K2 DQB1*0301 K1 K2 Cousin K1 DQBI*0301 K2 DQB1*0301 Father K1 DQBI*0604 DQBI*0604 K2 Mother Figure 6 Electropherogram of results for DQB analysis for a patient and three related potential donors. The PCR primers used co-amplify the pseudogene DQB2 with DQB1 alleles, resulting in the formation of two constant peaks (K1 and K2), thus four heteroduplex peaks are generated: two for each of the labelled references (DQB1 * 0402 and DQB2) individually hybridised to the two HLA-DQB1 alleles present in each sample. within and between gels and differences in the intensity of the DNA signals. Also, complex banding patterns are generated, which are difficult to interpret and do not give information on whether both alleles are mismatched or only one allele and if only one, which allele it is. In addition, the detection systems used more conventionally, such as ethidium bromide staining, limit considerably the resolution of bands with similar mobility. One important concern regarding the application of a conformation-based method for analysis of HLA matching

7 A high resolution HLA-matching method for donor selection Homoduplex 533 DPB1*11011 DPB1*0402 DPB1*11011 Cousin DPB1*11011 DPB1*0402 Father DPB1*11011 Mother Figure 7 DPB1 matching for a patient and three related donors. The father and the patient match for both DPB1 alleles whereas the other two potential donors share only one allele (DPB1 * 11011) with the patient. Table 3 Number of mismatches detected at each loci HLA-loci Samples typed by serology, Total No. of SSO or SSP methods m/m identified by RSCA c Original No. Extra m/m of m/m a defined by RSCA b A B C DRB DQA1 NT 0 59 DQB DPA1 NT 0 15 DPB Total a No. of samples mismatched by serology, SSO or SSP methods. b No. of mismatches defined by RSCA in addition to those defined by footnote a. c No. of mismatches defined by RSCA for the total number of patient and donor pairs analysed in the study. NT = not tested. could be that perhaps not all alleles are distinguished. We have shown previously that using two different FLRs per sample allows the detection of alleles differing by a single nucleotide. However there is always a possibility that untested alleles differing by single nucleotides may not be resolved. This would depend on the location of the nucleotide substitution, the FLR utilised, the length of the gel and electrophoretic conditions. 3 Another potential concern may be that for HLA class I analysis, a DNA fragment containing intron 2 in addition to the polymorphic exons 2 and 3 is analysed. Thus any polymorphism within intron 2 will also affect the conformation of the DNA duplexes generated which could be misinterpreted as a coding polymorphism. To investigate this potential problem, we have performed over 3800 tests by RSCA in samples with welldefined HLA alleles on identical and different haplotypes from a range of ethnic groups. This study has demonstrated that intra-alleleic intron 2 polymorphism is a very rare event (unpublished data). For HLA class I, RSCA has been further developed as a typing methodology, 10 which allows the assignment of allelic level type, including the identification of alleles differing by synonymous substitutions. Although the protein products of alleles differing by synonymous substitutions will have identical functions, there are several examples of such alleles being in linkage disequilibrium with different alleles at other loci. For example we have found Cw*02024 only in linkage with B*1503, whereas Cw*02022 is found with different HLA-B alleles. Thus identification of synonymous polymorphisms can aid the definition of HLA haplotype through linkage disequilibrium between alleles at different loci. In the selection of allogeneic BM donors, the best donor is only found after the serial typing of potential donors which can be time-consuming, and unfortunately for some patients time is limited. We have shown here the application of RSCA as a one protocol technique for high resolution allelic level HLA class I and II matching for BM

8 534 A high resolution HLA-matching method for donor selection donor selection. RSCA requires only one pair of PCR primers and two FLR DNAs per locus for the resolution of potentially all HLA alleles, with polymorphisms within exons 2 and 3 for class I and exon 2 for class II, including new alleles, without updating of reagents. One complete RSCA test takes 12 h, including PCR amplification, agarose gel electrophoresis to confirm presence of PCR product, FLR hybridisation and PAGE (Alfexpress instrument with 40 lanes; Pharmacia Biotech). Each gel can be re-used up to five times. Sample throughput can be increased by purification of the FLR PCR product to remove unincorporated labelled primer, thus allowing re-loading of gel every 2 h. With the exception of the initial cost of instrumentation, the cost of consumables required for RSCA is minimal, compared with other methods such as SSP and direct sequencing. RSCA can be used to assess quickly the level of matching between patient and donor, prior to the availability of HLA typing data. Those donors who best match with the patient by RSCA can then specifically be HLA typed by RSCA or other HLA-typing methods. For unrelated donor selection, potential donors can be selected from donor registries based on the available HLA-typing data. At present there are around 45 unrelated stem cell donor registries worldwide (including both volunteer BM donor and cord blood registries) which together provide a pool of over 4.6 million donors. 2 However most of these donors have only been typed for HLA-A and B loci by serology and the DRB1 locus by medium level molecular typing, and not typed at all for HLA-C, DQ and DP loci. It is likely that many unrelated BM transplants have been performed with undetected HLA mismatches and certainly it appears that not all mismatches are detrimental to transplant outcome. 11 In order to understand fully the contribution each HLA loci has on transplant outcome, accurate high resolution typing is required. This would allow an assessment of the true level of mismatching between donors and patients and thus define which HLA mismatches may be permissable in particular clinical situations. RSCA is a simple technique that permits accurate high resolution matching of all HLA loci and the simultaneous analysis of multiple donors allowing quick assessment of the level of HLA compatibility. Acknowledgements The authors thank Steven GE Marsh and Pel-Freez Clinical Systems for helpful discussion. This work was funded by The Anthony Nolan Bone Marrow Trust. RA is a recipient of a fellowship from the Consejo Nacional de Ciencia y Tecnologia, Mexico and Overseas Research Students Awards (CVCP) UK. DG is funded by a fellowship from Fondo de Investigacion Sanitaria BAE 96/5063, Ministerio de Sanidad, Spain. References 1 Madrigal J, Scott I, Argüello R et al. Factors influencing the outcome of bone marrow transplants using unrelated donors. Immunol Rev 1997; 157: Bone Marrow Donors Worldwide. leidenuniv.nl Argüello R, Little A-M, Pay A et al. Mutation-detection and typing of polymorphic loci through double-strand conformation analysis. Nat Genet 1998; 18: Cereb N, Maye P, Lee S et al. Locus-specific amplification of HLA class I genes from genomic DNA: locus-specific sequences in the first and third introns of HLA-A, -B, and C alleles. Tiss Antigen 1995; 45: Charron D, Fauchet R (eds). Twelfth International Histocompatibility Workshop Technical Handbook Marsh SGE, Packer R, Heyes JM et al. The 12th International Histocompatibility Workshop Cell Lines Panel. In: Charron D (ed). Proc Twelfth International Histocompatibility Workshop and Conference. EDK: Paris, 1997, pp Rozemuller EH, Tilanus MGJ. A computerized method to predict the discriminatory properties for class II sequencing based typing. Hum Immunol 1996; 46: Scheltinga S, Johnston-Dow L, White C et al. A generic sequencing based typing approach for the identification of HLA-A diversity. Hum Immunol 1997; 57: Clay T, Bidwell J, Howard M, Bradley B. PCR-fingerprinting for selection of HLA matched unrelated marrow donors. Collaborating Centres in the IMUST Study. Lancet 1991; 337: Argüello J, Little A-M, Bohan E et al. High resolution HLA class I typing by reference strand mediated conformation analysis (RSCA). Tiss Antigen 1998; 52: Petersdorf E, Smith A, Mickelson E et al. The role of HLA- DPB1 disparity in the development of acute graft-versus-host disease following unrelated donor marrow transplantation. Blood 1993; 81: