Major histocompatibility locus genetic markers of beryllium sensitization and disease

Study population and case definition

A population of 639 individuals from a beryllium manufacturing plant in Midwestern USA was enrolled after informed consent into a genetic study for susceptibility to berylliosis, in conjunction with the enrollment into a beryllium disease surveillance programme carried out by the plant's Medical Department, which included blood BeLPT carried out in two laboratories following published protocols 4. This population had been described previously by Kreiss et al. 10. For the purpose of the present study, individuals with two abnormal BeLPTs in two separate blood samples were defined as BeLPT-positive or “sensitized” and were offered medical evaluation. The diagnosis of berylliosis was established upon a fibreoptic bronchoscopic TBB showing noncaseating granulomas. Individuals with a single abnormal BeLPT and “sensitized” individuals who declined medical evaluation were not included in the study. Twenty-three BeLPT-positive TBB-negative individuals were identified and defined as “sensitized”; they included 21 males and two females, mean±sd age 42±9 yrs, average years of employment 14±9, all Caucasians. Twenty-two BeLPT-positive TBB-positive individuals were identified and defined as “disease cases”; they included 21 males and one female, mean±sd age 41±8 yrs, average years of employment 15±6, 20 Caucasians and two African-Americans. Ninety-three Be-exposed BeLPT-negative individuals were randomly selected from the whole Be-LPT-negative workers on the basis of a nested case-control study design, and therefore evaluated as controls; they included 78 males and 15 females, mean±sd age 44±9 yrs, average years of employment 16±11, 88 Caucasians, two Hispanics, two African-Americans, one Asian. The sample of the control population was selected on a 1:2 ratio and on the basis of previously published HLA-DP gene association data, in order to assure a sufficient statistical power to the study. The “sensitized” individuals and the “disease cases” were not statistically different from the control group for sex, age, time on the job and race.

Genetic studies

The study population was evaluated for the frequencies of two tumour necrosis factor (TNF)-;α alleles (−308 biallelic polymorphism), TNF-;α-;308*1 and TNF-;α-;308*2 and HLA class II (HLA-DPB1, HLA-DQB1, HLA-DRB1 and HLA-DRB3/4/5) alleles.

Tumour necrosis factor-;α gene polymorphism analysis

The analysis of the biallelic polymorphism (G to A) at position −308 of the promoter of the TNF-;α gene was carried out by polymerase chain reaction (PCR) amplification of 500 ng of genomic deoxyribonucleic acid (DNA) using the primers TNF-;α-;308-;A (CAAACACAGGCCTCAGGACT) and TNF-;α-;308-;B (AGGGAGCGTCTGCTGGCTG) with the following PCR cycles (GeneAmp 9600, Perkin Elmer Corporation, Roche Molecular Systems, Branchburg, CT, USA): one denaturation cycle (95°C for 1 min with a pause at 80°C for the hot start technique), 30 amplification cycles (94°C for 30 s, 55°C for 30 s, 72°C for 30 s), one extension cycle (72°C for 7 min). The expected molecular weight of the amplified product was 545 base pairs (bp). Ten μL of the PCR product were then analysed on a 2% agarose gel stained with ethidium bromide. Twenty µL of the PCR product were probed using the oligonucleotide probes TNF-;1 (Biotin-AGGGGCATGGGGACGGG) and TNF-;2 (Biotin-AGGGGCATGAGGACGGG) in the nonradiometric hybridization DNA enzyme immunoassay (DEIA) method (Gen-ETI-;K DEIA, Sorin Biomedica, Saluggia, Vercelli, Italy). The DEIA experimental conditions were as follows: PCR product: 20 µL PCR product per well; probe concentration: 5 ng probe per well; hybridization conditions: 1 h at 50°C. The hybridization results were obtained as optical density values on an enzyme-linked immunosorbent assay (ELISA) fast reader (Sorin Biomedica). In order to confirm the hybridization results, 20 DNA samples were sequenced for the TNF-;α-;308 genotyping. Briefly, 250 ng of genomic DNA were PCR amplified with the primers TNF308AM13 (TGTAAAACGACGGCCAGTGAAACAGACCACAGACCTG) and TNF308BB (Biotin-CTTCTGTCTCGGTTTCTTC) using one denaturation cycle (95°C for 1 min), 35 amplification cycles (94°C for 1 min, 54°C for 20 s, 72°C for 20 s), one extension cycle (72°C for 7 min) to generate a 250 bp fragment. The PCR product was directly cycle-sequenced with the fluorescein-labelled M13 universal sequencing primer (GTAAAACGACGGCCAGT), (Pharmacia Biotech, Uppsala, Sweden) (3 pmol) using the Thermo Sequenase fluorescent-labelled primer cycle sequencing kit (Amersham Life Sciences, Little Chalfont, UK) for 20 PCR cycles (98°C for 30 s, 60°C for 30 s, 72°C for 30 s) in conjunction with the AutoLoad Solid Phase Sequencing Kit (Pharmacia Biotech), following the manufacturer's instructions and analysed on an automated laser fluorescence (ALF) DNA sequencer (Pharmacia Biotech).

Typing of human leukocyte antigen-DRB1, -DRB3/4/5 and -DQB1 alleles

For high-resolution typing of HLA-DRB1, -DRB3/4/5 and -DQB1, either PCR with sequence-specific primers (PCR-SSP) or sequence-based typing (SBT) was used. First, a low resolution typing was performed by means of group-specific amplification of exon 2 using a 5′ or 3′ specific primer in combination with a generic primer. Depending on the allele groups that were identified, high-resolution typing was performed by a subsequent PCR-SSP using additional primer mixes, or by sequence-based typing 11, 12. For high-resolution typing of HLA-DPB1, sequence-based typing was used in all instances 13. High-resolution typing for DRB1/3/4/5, DQB1 and DPB1 was performed for all individuals studied, except in nine cases where high resolution of DRB3/4/5 and/or DQB1 was not possible due to a shortage of material. Since no subtype differences are located in exon 2 of DQB1*02 only low-resolution typing was performed.

The sequence-based typing method for the HLA class II alleles was carried out as described previously 11–13. For SBT of all class II genes, a PCR product was generated by amplification of exon 2 and, if needed, exon 3 using amplification and sequencing primers located at the 5′ and 3′ end of exon 2 or in introns 1 and 2. Amplification was performed in a final volume of 60 µL consisting of 600 ng DNA, PCR buffer (10 mM Tris-HCl, pH 8.3; 50 mM KCl; 1.5 mM MgCl₂), 5% glycerol, 6 µg cresol red, 200 µM each deoxyribonucleoside triphosphate (dNTP), 20 pmol biotinylated primer, 40 pmol unlabelled primer and 2.0 U AmpliTaq DNA polymerase (Perkin-Elmer Corporation). For amplification of DQB1, 1% dimethyl sulphoxide (DMSO) was added to the PCR reaction mixture. PCR amplifications were carried out in a Gene Amp PCR System 9600 (Perkin-Elmer Corporation). The PCR profiles consisted of an initial denaturation at 94°C (2 min), 10 cycles of 10 s at 94°C, 1 min at 65°C and finally 20 cycles (for DRB1/3/4/5 and DQB1) or 30 cycles (for DPB1) of 10 min at 94°C, 50 s at 61°C and 30 s at 72°C, followed by 10 min at 72°C. The sequencing reaction was based on a solid phase approach using attachment of the biotinylated PCR product to streptavidin-coated beads. The sequencing reaction was performed using the Autoread Sequencing Kit (Amersham Pharmacia), according to the suppliers protocol. The sequencing reaction mixture was analysed on a Pharmacia ALFexpress DNA sequencer (Pharmacia Biotech) using a 6% polyacrylamide/7 M urea gel. The gel was run at 1,500 V, 60 mA and 25 W at 5°C for 6 h. After manual evaluation of the processed data, the determination of the HLA subtype was performed with the use of the HLA-Sequityper software (Pharmacia Biotech).

Analysis of the human leukocyte antigen-DPB Glu69 marker

In addition to SBT, genomic DNA samples were screened for the HLA-DPB1*AAG→GAG polymorphism coding the lysine to glutamate 69 amino acid residue using a simple hybridization test designed so that the sensitivity and specificity of each test could be validated by discriminant analysis as already described 6.

Statistical analysis

HLA class II allele and phenotype distributions in berylliosis-affected, Be-sensitized and healthy exposed control groups were compared by means of univariate analysis using odds ratio (OR) and 95% confidence interval (CI); probability values were calculated with the Fisher's exact test. In order to assess the relative strength of the different MHC associations with the disease and sensitization status, and to ascertain possible linkage disequilibria between the HLA-DP and HLA-DR markers and the TNF-;α gene, the statistical method described by Svejgaard and Ryder 14 was applied.

The method is based on two-by-two tests of the various components of a two-by-four table. Briefly, the frequencies of the phenotypic combinations of the two genetic factors (the DPGlu69, the DRArg74 and the TNF-;α-;308*2 markers) in patients and controls were calculated (“basic” table). These “basic” data were then analysed by means of various tests, i.e. two-by-two tables (“tests” table); for each test an OR was calculated using adequate modifications for “zero” entries and the statistical significance of the OR values was established by Fisher's exact test. Because this analysis involved multiple comparisons, the p-;values were corrected; in particular, due to the fact that the association of the DPGlu69 marker was known before the present study, no corrections were done for this marker. On the contrary, p-;values for the DRArg74 and the TNF-;α-;308*2 markers have been corrected for the number of polymorphisms and the number of groups analysed. Due to the fact that no association data were available for the TNF-;α-;308 polymorphism and berylliosis and no conclusive demonstration of linkage disequilibrium between HLA-DP and TNF-;α was found to exist, the other comparisons were corrected for six or by nine, accordingly to the recommendations given by Svejgaard and Ryder 14.

In the “tests” table, a series of two-by-two comparisons represents various stratifications of the data for the genetic markers testing (in)dependent contributions to disease susceptibility (table 1⇓, independent association), “interaction” of the factors, difference between the associations (table 1⇓, A association, B association and difference between A and B associations), the value of the combined action of the two factors (table 1⇓, combined association) and linkage disequilibrium of the genetic factors in sensitized/patients and controls (table 1⇓, association between A and B). The tests performed were aimed to verify various hypotheses and an interpretation is given based upon the significance values. According to Svejgaard and Ryder 14, p-;values 0.01–0.05 should be regarded as “probably significant” and only p-;values <0.01 should be considered as “significant”. In fact, the retrospective analysis of HLA-association studies suggests that only these latter p-;values have been confirmed in multiple independent studies 14. However, several authorities in the HLA-disease field 14 strongly recommend the use of phenotypic (or supratypic) markers instead of allelic frequencies for specific reasons: 1) to avoid the bias of increased statistical power determined by the artificial doubling of the population size inherent to allelic frequencies analysis (from N patients to 2×N genes); 2) the phenotype analysis appears to be more appropriate because usually the gene product and not the gene itself is responsible for disease susceptibility or protection 14. Accordingly, allelic variant analysis data are presented as phenotypic data and the above definitions of “probably significant” and “significant” are used throughout the manuscript.

Human leukocyte antigen-DR, -DQ and -DP marker frequencies

The frequencies of HLA-DRB1*0301, HLA-DRB3*0101, HLA-DQB1*02 and HLA-DPB1*0501 alleles were higher in the “sensitized” subgroup, while HLA-DRB3*0301 was higher and HLA-DRB4*0103 was lower in the “disease case” subgroup (tables 2–5⇓⇓⇓⇓). None of the differences was statistically significant after correction for the number of alleles tested.

The inspection of the polymorphic amino acid residues in the sensitization-associated alleles HLA-DRB1*0301 and HLA-DRB3*0101 showed 82% homology, suggesting that shared polymorphic residues could in fact be associated with sensitization to Be. The comparison between polymorphic residues expressed by the alleles, positively or negatively associated with disease or sensitization, showed that residues DRTyr26 and DRArg74 were unique to sensitization-associated alleles.

Strikingly, the frequency of the DRArg74 marker was significantly higher in the “sensitized” subgroup than in controls (OR 3.96, p=0.005, 95% CI 1.5–10.1). In contrast, in the “disease case” subgroup it was similar to controls (OR 0.89, nonsignificant, 95% CI 0.31–2.6; fig. 1⇓). Also, the DRArg74 marker frequency in the “sensitized” subgroup was significantly higher than in the “disease case” subgroup (χ² test, p=0.044).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.—

Frequencies (%) of the genetic markers tumour necrosis factor (TNF)-;α-;308*2, human leukocyte antigen (HLA)-;DRArg74 and HLA-DPGlu69 in the “disease case” () the “sensitized” (□) and the control () groups.

Consistent with previous study results, the frequency of the DPGlu69 marker was significantly higher in the “disease case” subgroup than in controls (OR 3.7, p=0.016, 95% CI 1.4–10.0; fig. 1⇑). In contrast, its frequency in the “sensitized” subgroup was not different from that in controls (OR 0.89, nonsignificant, 95% CI 0.30–2.22). As expected, the frequencies of several DPGlu69-positive HLA-DPB1 alleles were also increased in the “disease case” subgroup: HLA-DPB1*0201 27.3% (controls: 17.2%), *0601 2.3% (controls: 1.1%), *0901 2.3% (controls: 0%), *1001 6.8% (controls: 2.7%), *1601 2.3% (controls: 0%), *1701 4.6% (controls: 0%), *1901 2.3% (controls: 0.5%). Similarly, the frequencies of the DPGlu69-positive HLA-DPB1 allele phenotypes were also increased in the “disease case” subgroup (table 6⇓). None of these differences, possibly due to the small size of the study group, were statistically significant in comparison to the phenotypic frequencies in the control group.

Tumour necrosis factor-;α-;308 polymorphism analysis

The frequency of the TNF-;α-;308*2 allele was significantly increased in the whole BeLPT-positive group compared to BeLPT-negative controls (BeLPT-positive 51.1% versus BeLPT-negative 16.1%, OR 7.8, corrected p<0.0001, 95% CI 3.2–19.1). When the “sensitized” and the “disease case” subgroups were compared separately to the controls, the frequency of TNF-;α-;308*2 was significantly higher in both subgroups (fig. 1⇑), and it was similar in the “sensitized” and the “disease case” subgroup (χ² test=1.8, nonsignificant; fig. 1⇑).

Analysis for gene interactions

The evaluation of the interaction between the DPGlu69, DRArg74 and TNF-;α-;308*2 markers, using the analysis of multiple HLA associations with disease described by Svejgaard and Ryder 14, showed a positive interaction between TNF-;α-;308*2 and DRArg74 in the association with the “sensitized” subgroup (table 1⇑, combined association) in the absence of linkage disequilibrium (table 1⇑, association between A and B), but no interaction in the association with the “disease case” subgroup (table 1⇑, combined association). Conversely, while there was no interaction between TNF-;α-;308*2 and DPGlu69 in the association of the TNF-;α-;308*2 marker with the “sensitized” subgroup (table 1⇑, combined association) there was a positive interaction in the association with the “disease case” subgroup (table 1⇑, combined association), in the absence of linkage disequilibrium (table 1⇑, association between A and B). In addition, there was an association of DPGlu69 with the “sensitized” DRArg74-negative subjects (table 1⇑, B association −+ versus −−), but no positive interaction between the two markers either in the “sensitized” or the “disease case” subgroups (table 1⇑, A association ++ versus −+).