Evidence that an HLA-DQA1-DQB1 haplotype influences susceptibility to childhood common acute lymphoblastic leukaemia in boys provides further support for an infection-related aetiology.

Comparison of DQA1 and DQB1 alleles in 60 children with common acute lymphoblastic leukaemia (c-ALL) and 78 newborn infant control subjects revealed that male but not female patients had a higher frequency of DQA1*0101/*0104 and DQB1*0501 than appropriate control subjects. The results suggest a male-associated susceptibility haplotype in c-ALL and supports an infectious aetiology.

obtained from the umbilical cords of 78 normal full-term newborn infants (38 boys and 40 girls) delivered at St Marys Hospital.

HLA-DQA1 molecular typing
Genomic DNA was extracted from patient and control blood samples as previously described (Dearden et al. 1996). and DQAI typing carried out as described by Noreen et al (1992). A 228-bp exon 2 fragment of DQAI was amplified by the polymerase chain reaction (PCR) using the primers DQAAMP-A (5'-ATG GTG TAA ACT ACC AGT-3') and DQAAMP-B (5'-TTG GTA GCA GCG GTA GAG T1G-3'). obtained from the British Society for Histocompatibility and Immunogenetics (BSHI). PCR mixtures consisted of 50 no of genomic DNA. 0.5 Em of each primer and 0.3 mi dNTPs in 20 jl of PCR buffer. Amplifications (32 cycles) were carried out on a Thermal Cycler (Appligene. France). and PCR products dot-blotted onto nylon filters. which were hybridized with ten 1 8-mer sequence-specific oligonucleotide (SSO) probes (from BSHI) end-labelled with Y'P-ATP. detecting eight DQAI alleles [*0101 (*0101 + *0104). *0102. *0103. *0201. *03 (*03011 + *0302). *0401. *05 (*05011 + *0502 + *0503) and *0601]. The filters were scanned for radioactivity on an InstantImager (Canberra Packard. Berks. UK). Positive hybridization was scored by comparing test vs negyative control spots (typically a tenfold difference in counts). DQAI alleles were assigned by comparing SSO patterns with published information (see Marsh reference to Website). The following pairs of alleles could not be distinguished from each other with the probes used in this study: *0101 from *0104: *03011 from *0302 and subtypes of *05: *0601 homozygotes gave the same result as *0401/*0601 heterozycotes: and *0103 homozygotes gave the same result as *0102*0103 heterozygotes. DOB1 molecular typing Patients and control subjects were typed for DQBI alleles by single-strand conformation polymorphism (SSCP) analysis as described previously (Dearden et al. 1996). Data analysis Differences in patient and control allele frequencies are expressed as odds ratios (OR) (Altman. 1991 ) together with 95% confidence intervals (CI) derived using Miettinens method (Breslow and Day. 1980). Patient and control DQAI allele frequencies were also compared using 2 x 2 analysis. and tested for significance by twosided Fisher's exact tests. DQAl and DQB1 exon 2 polymorphic amino acid frequencies (see Marsh) were compared in patients and control subjects using ORs and 2 x 2 tests. Observed and expected heterozygosity was compared using allelic diversity (h) (Nei and Roychoudhury, 1974). Sample size (i.e. statistical power) calculations were performed using nQuery Advisor release 2.0 (Statistical Solutions. Cork. Ireland).

Study group
The patients consisted of a prospective series of 38 boys and 22 girls with c-ALL aged between 1.6 and 12.9 years, with a mean age at diagnosis of 5 years 3 months, and a median of 4 years 4 months. There was a 26% excess of male to female patients. giving an M:F of 1.73:1. The mean (median) age of the male patients was 4 years 8 months (4 years 3 months), compared with the mean (median) age of the female patients of 6 years 2 months (4 years 10 months).
Apart from a small deficit in DQAJ *0201. there were no differences between female c-ALL and female control subjects. DOA1 and DOB1 alleles in c-ALL Patients and control subjects were classified according to whether they typed for both DQBI *0501 and DQAJ *01011*0104. Table 2 shows that the greatest difference was between male c-ALL and male control subjects (OR = 3.73: 1.19-10.3). This was absent in girls with c-ALL and in four other pairs of DQAIDQBI alleles known to be in linkage disequilibrium.

British Journal of Cancer
The significance of these associations disappears following correction for the number of alleles tested We therefore performed simulations based on allele and haplotype frequencies in the present study to estimate case and control sample sizes required to obtain statistical significance. We assumed 90% power to detect a P = 0.005 in a two-sided test, before correction for the number of alleles in equal numbers of cases and control subjects. For DQAJ *01011*0104. the total number of patients and control subjects is 475. and for boys it is 181. For DQAI*0101/*0104,DQBI*0501 haplotypes. total patient and control series require 388 in each group whereas boys require 122 patients and controls.

DISCUSSION
Evidence suggesting that the same HLA class II polymorphic sequences contribute both to the binding of antigenic peptides and disease susceptibility (Hammer et al. 1995: Kwok et al. 1996 suggests that an HLA-DQAJDQBI haplotype association with childhood c-ALL could be construed as evidence of an infectious aetiology. In this study we found an increased frequency of DQAI *01011*0104. and a deficit of DQAI *0201 in c-ALL. suggesting roles in susceptibility and resistance to c-ALL respectively. Further analysis showed that DQAI *01011*0104 was increased in boys but not girls with c-ALL. a finding which would not have been expected by chance alone. Analysis of patients and control subjects classified by the presence or absence of DQAI *01011*0104 and DQBI *0501 showed that an increase in this 'haplotype' in patients was confined to boys. This haplotype (and certain others) encodes DQctSer52 and DQIVal57. Both amino acids were increased in c-ALL. but this was also confined to male patients. These results need to be treated with caution as they were not corrected for the number of alleles. However. we used the results to simulate the number of cases and control subjects required to repeat the study in an independent series.
The patients in this study were an unselected series with c-ALL.
in which 63% were boys and 37% were girls (M:F 1.73:1). Analysis of 199 cases of childhood c-ALL in the Manchester Children's Tumour Registry (MCTR) for 1983-94 showed 118 boys and 81 girls (M:F 1.5:1). suggesting that the male excess was not due to chance. McKinney et al (1993) found no difference in the rate of c-ALL in boys and girls aged 1-9 in a UK study. but Buckley et al (1994) reported an M:F of 1.2:1 in 312 cases of c-ALL in a US study. We found no evidence that inclusion of only verified c-ALL patients. and exclusion of unclassified ALL favoured boys over girls. There was no difference in the age or gender of patients donating and not donating blood samples to the study. The DQAI SSO probes used here define polymorphisms confined to exon 2 but do not distinguish between DQAI *0101 and *0104. These alleles differ for single base substitutions in codons 2 (exon 1) and 199 (exon 4) (Yasunaga et al. 1996). It remains to be seen whether the difference between *0101 and *0104 has any influence on susceptibility to c-ALL.
Our results contrast with Dorak et al ( 1995) who used a restriction fragment length polymorphism (RFLP)-based method to type DQAI alleles in childhood ALL. They found no increase in DQAI-JA in ALL. nor any difference between male and female patients typing for this allele. However. they found a significant increase in allele 3 in male compared with female ALL patients. We found an increase in DQA1 *03. which was confined to boys with c-ALL. However. there was no difference in patients and control subjects typing for both DQA I*03 and DQBI*0302. which are in linkage disequilibrium (Imanishi et al. 1992). Furthermore. there was no increase in the frequency of DQAI *03 homozygotes in c-ALL.
As DQA1 is tightly linked to DQBJ. the DQAI*0101/*0104 association with c-ALL could be explained by linkage disequilibrium with DQBI *0501. The present study confirms an increased frequency of both alleles in boys with c-ALL. but not of other DQAIJDQBI haplotypes. Our previous results showed a reduced frequency of aspartic acid at position 57 in c-ALL (Dearden et al. 1996). suggesting similarities with the DQ~lAsp57motif associated with IDDM (Tosi et al. 1994). However. analvsis of the IDDM susceptibility haplotype DQalArg52.DQf1Asp57showed no evidence of a role in c-ALL.
Gene transfection studies bv Kwok et al (1993) showed that the expression of DQBI *0501-encoded a-chains is facilitated by DQAI*0101 et-chains and is influenced by amino acids coded at the 3' end of DQBJ. corresponding to positions 60 and 91 of the DQIBl subunit. Furthermore. DQAI *0101-DQBI *0501 is one of the most common DQAI-DQBI haplotypes in the UK population (Doherty et al. 1992) and the second most common haplotvpe in French. Danish and Spanish populations (Imanishi et al. 1992).
The gene expression and population genetic data thus suggest that the DQAI *0101.DQBI *0501 haplotype may have had a selective advantage. possible by protecting against infectious diseases. Our results sugaesting that susceptibility to c-ALL is increased in boys with a common DQAIl-DQBJ haplotvpe could be explained bya greater contribution of DQalSer52IDQlVal57 peptide-binding motifs to the protection of boys against certain types of childhood infection. There is increased susceptibility of male children to infections (Washburn et al. 1965: Purtilo andSullivan. 1979)i. which suggests that certain HLA haplotvpes maycounteract X-linked defects in immunity in boys (Immunological Reviews. 1994). by promoting the efficiency of antigen presentation. If confirmed. our results would imply that a common DQAJ-DQBJ haplotvpe that mayr increase resistance to infection in boys has an important influence on susceptibility to childhood c-ALL.
This would be consistent with the predictions of the Greaves hypothesis (Greaves. 1988: Greaves and. and further suggests that the candidate infection involved in c-ALL exhibits low-pathogenicitybut strong immunogenicity.  KS and Sasazuki T (1996) Different contribuion of HLA-DR and -DQ genes in suscepability and resistance to insulendent diabetes mellitus (IDDM). Tissue Antigens 47: