Published ahead of print on February 13, 2003, doi:10.1164/rccm.200209-1097OC
© 2003 American Thoracic Society Sarcoidosis Susceptibility and Resistance HLA-DQB1 Alleles in African AmericansDivision of Pulmonary and Critical Care Medicine, Mount Sinai Medical Center, New York, New York; and Department of Pulmonary, Critical Care, Allergy, and Sleep Medicine and Department of Biostatistics and Research Epidemiology, Henry Ford Health System, Detroit, Michigan Correspondence and requests for reprints should be addressed to Michael C. Iannuzzi, M.D., Division of Pulmonary and Critical Care Medicine, Mount Sinai Medical Center, One Gustave Levy Place, Box 1232, NY, NY 10029. E-mail: michael.iannuzzi{at}mountsinai.org
Sarcoidosis, in the United States, more commonly and severely affects African Americans. HLA associations with sarcoidosis have been reported, but most studies used case-control designs, which may produce biased results because of population stratification. We examined transmission of HLA-DQB1 alleles in 225 African American families with at least one offspring with sarcoidosis. Of five low-resolution HLA-DQB1 alleles, *02 and *06 showed significant deviation in transmission patterns to affected offspring. High-resolution typing of these allelic subsets revealed that HLA-DQB1*0201 was transmitted to affected offspring half as often as expected (p = 0.001), whereas DQB1*0602 was transmitted to affected offspring about 20% more often than expected (p = 0.029). Examining interactions between *0201 and *0602 alleles and environmental exposures showed that *0602 varied little with respect to exposure, but sarcoidosis risk associated with *0201 often depended on exposure status. Alternatively, the *0602 allele in affected probands was associated with radiographic disease progression, but the *0201 allele showed no significant correlation with phenotype. Major differences in the amino acid sequences encoded by *0201 and *0602 alleles exist, which may explain the differential effects these alleles have on sarcoidosis susceptibility and progression in African Americans.
Key Words: African Americans • disease susceptibility • genetic predisposition • HLA antigens • sarcoidosis Sarcoidosis results from marked macrophage and CD4+ helper T cell activity, immune dysregulation, and formation of noncaseating granulomas in affected organs. Although the initiating antigen(s) remains unknown, familial aggregation and ethnic predominance suggest that an inherited susceptibility to sarcoidosis exists (1, 2). Attempts to identify sarcoidosis susceptibility genes have focused on the genes residing in the major histocompatibility complex (MHC), particularly the HLA genes. The focus on HLA genes in sarcoidosis is warranted because T cell activation likely occurs after antigen processing and presentation in conjunction with the appropriate HLA Class II cell surface molecule (3). Further support derives from the role of HLA in chronic beryllium disease. Chronic beryllium disease (CBD), with clinical presentation and immunopathologic mechanisms similar to those of sarcoidosis, is thought to be an excellent model with which to study genetic susceptibility to lung granulomas (4). CBD occurs most commonly in carriers of an HLA-DPB1 allele with glutamate in position 69 (4–6). Although HLA-DPB1 associations have been reported in sarcoidosis, they are not as strong or consistent as that reported for CBD (7–11). Other HLA associations with sarcoidosis have been reported. These include HLA-A*01, A*26, A*30, B*08, B*13, B*27, DRB1*03, DRB1*11, DRB1*13, DRB1*14, and DRB3*01 with susceptibility (8, 12–17), HLA-B*35 and HLA-DRB1*11 with early onset (18, 19), HLA-DRB1*03 with acute onset (20–23), HLA-DRB1*14 and DRB1*15 with chronic disease (24), and HLA-DRB1*13 with severe disease (14). For many of these earlier studies, HLA antigens were typed by serological methods. These methods have been replaced by DNA testing to identify sequence variants responsible for the different structure of the antigens (25). An increased density of Class II MHC antigen expression on alveolar macrophages in pulmonary sarcoidosis has also been reported (26–28). More recently, familial sarcoidosis was found to be linked to the major histocompatibility complex region in German families (11, 29). Association studies have narrowed the MHC region where sarcoidosis susceptibility genes likely reside to the Class II region. Associations were observed with TAP-2 (transporter associated with antigen processing-2), but not with TAP-1 and HLA-DP (30). We completed a fine mapping study of the MHC region, on the basis of African American families including members with sarcoidosis (31), and concur that the sarcoidosis susceptibility locus resides telomeric to TAP-2, but our data suggested that the susceptibility locus is more closely associated with HLA-DQB1 than with DRB1. We conducted a family-based genetic association analysis of sarcoidosis and HLA-DQB1 alleles. The term family-based indicates that the data consist of nuclear families rather than unrelated individuals as cases and control subjects. The strength of the family-based design is that it controls for genetic background unrelated to the disease of interest that can potentially confound traditional case control designs. Although case-control samples are generally easier to collect and provide more statistical power than family-based samples (32), case-control samples may give rise to spurious disease associations because of ethnic admixture within a population or other problems relating to inappropriate controls (33).
Study Sample and Data Collection The study protocol was approved by the Henry Ford Hospital (Detroit, MI) Institutional Review Board. Diagnosis was confirmed by tissue biopsy in 86% of the index cases and in the 14 patients with normal (Stage 0) chest X-rays. The patients without histologic confirmation had radiographic evidence of bilateral hilar adenopathy, a compatible clinical presentation, and were observed for 2 years or more with no other medical condition that could explain the clinical course. Of the 623 eligible probands, 359 (58%) were enrolled with one or more first-degree family members. Of these, 234 had 2 or more parents or sibs who donated a blood sample. An additional 10 African American sarcoidosis families were recruited outside of the Henry Ford Hospital system, resulting in 244 total families. Both parents were genotyped when available. When one or both parents were unavailable for genotyping, all available full sibs were genotyped. Using a panel of unlinked markers, we excluded 20 families (8.2%) with Mendelian segregation inconsistencies and 2 families (0.8%) where on further follow-up the proband was found not to have sarcoidosis from the analysis. Three of the remaining 222 families each had an additional nuclear family that was analyzed separately, resulting in 225 nuclear families or a total of 704 individuals for study. Approximately 8% of families (19 of 225) had 2 or more affected sibs. Sibs were considered unaffected by the self-reported absence of skin, eye, lung disease, or history of sarcoidosis.
Questionnaire Data
Radiography
Phenotyping DNA isolation was performed as previously described (37). HLA-DQB1 typing was performed by sequence-specific primer polymerase chain reaction (PCR) (Olerup SSP; GenoVision, West Chester, PA). HLA-DQA1 alleles were genotyped in a stepwise fashion. Low-resolution genotyping was performed on the entire study sample. High-resolution genotyping was performed on only those low-resolution allelic subsets that exhibited an association with the disease outcome.
Statistical Methodology
Potential gene–environment interactions were evaluated by a general estimating equations analytic approach with the Statistical Analysis System software program (SAS Institute, Cary, NC). Using the PROC GENMOD (40) procedure in SAS, genotypes of sibs were modeled as repeated measures to account for the correlation among sibs. For analyses of correlation between proband genotypes and phenotypes
HLA-DQB1 Low-resolution Alleles On the basis of low-resolution typing, we detected five DQB1 alleles in our study population. Table 1 shows the distribution of these alleles among offspring in the 225 sarcoidosis nuclear families that formed our study population. The most frequent allele, DQB1*06, represented 31.1% of the 1,018 chromosomes examined and was found in a slightly higher percentage in offspring with a history of sarcoidosis compared with those without a sarcoidosis history (32.8 versus 29.6%). The least frequent allele, DQB1*04, was found in only 8.2% of chromosomes examined and had a slightly higher percentage in offspring with a history of sarcoidosis compared with those without a sarcoidosis history (8.9 versus 7.4%). Of the remaining four alleles, only DQB1*02 appeared to be differentially distributed between offspring with a history of sarcoidosis and those without a sarcoidosis history (13.4 versus 19.1%).
Table 2 shows the transmission statistics for individual DQB1 alleles under both dominant (overall  2, 5 degrees of freedom = 16.3; p = 0.006) and additive (overall 2, 4 degrees of freedom = 15.9; p = 0.003) inheritance models. DQB1*02 was the most statistically significant allele (p = 0.0005, both dominant and additive models) with a marked undertransmission to affected offspring. The allele that deviated the most, next to DQB1*02, from its expected transmission pattern was DQB1*06, which was slightly overtransmitted to affected offspring. The overtransmission of DQB1*06 was statistically significant under both dominant (p = 0.019) and additive (p = 0.015) inheritance models. The transmission of the other four DQB1 alleles to sarcoid-affected offspring did not show any deviance from expected values.
HLA-DQB1 High-resolution Alleles Given the statistically significant differences of the HLA-DQB1*02 and *06 alleles (Table 2), we undertook high-resolution typing of these two allelic subsets. A total of three different *02 alleles and eight different *06 alleles were detected in our study population. Of these 11 alleles, 5 alleles (DQB1*0203, *0608, *0619, *06052, and *06111) were not found frequently enough to generate transmission statistics. Transmission statistics for the remaining six alleles are depicted in Table 3 . Two of these six alleles showed significant deviation in transmission patterns to affected offspring under both models, with the DQB1*0201 allele, transmitted to affected offspring half as often as expected, being the allele with the strongest association. In addition, DQB1*0602, the other high-resolution allele with a statistically significant transmission pattern (under an additive model), was transmitted to affected offspring about 20% more often than expected.
We next investigated whether sarcoidosis risks associated with certain exposures were significantly stronger (or weaker) in the presence of the DQB1*0201 and *0602 alleles. This was done through fitting a gene–environment interaction model in the subset of data that comprised sibships of size two or greater with one or more affected sib. A total of 186 of the 225 nuclear families under study had sibships that met this criterion. Although the results for the dominant and additive inheritance models from the previous FBAT analyses were not markedly different, the results for the additive model had a somewhat higher statistical significance. This, coupled with the modest gain in statistical power that the additive model would afford in gene–environment interaction testing, led us to model genotype effects for both the DQB1*0201 and *0602 alleles on an additive scale. All models tested also had covariate terms for age and sex because sarcoidosis is known to vary according to both. Only exposure terms that had an association with sarcoidosis on the univariate level in sibships with a p value of 0.15 or less and a prevalence in sibs of 10% or greater were tested in gene–environment interaction models. Of the 61 exposure terms tested, 10 qualified for interaction model testing on the basis of these criteria. The interaction odds ratios for these 10 exposures with the DQB1*0201 and *0602 alleles are shown in Table 4 with the corresponding genotype-stratified exposure odds ratios shown in Table 5 . The interaction odds ratios shown in Table 4 can be interpreted as the ratio of the proportion of individuals with disease who have the exposure and genotype of interest over the proportion of individuals with disease who do not have the exposure and genotype of interest. Interaction odds ratios greater than one indicate a positive interaction (i.e., presence of the allele increases the exposure odds ratio), whereas odds ratios less than one indicate a negative interaction (i.e., presence of the allele decreases the exposure odds ratio). The interaction odds ratios are not an actual measure of relative risk, but rather a ratio of relative risk measures. The odds ratios in Table 5 are shown to give some idea of how exposure directly affects disease risk in families where the genotype is either present or absent.
For the DQB1*0201 allele, the most striking positive interactions were with vegetable dust exposure (OR, 4.11) and water damage (OR, 3.99). Only the DQB1*0201–water damage exposure interaction was statistically significant at the 0.05  level (Table 4). However, the odds ratio for vegetable dust (Table 5) in individuals with the DQB1*0201 allele was much higher (OR, 18.60) than the corresponding odds ratio for water damage exposure (OR, 4.44). Several negative interactions with the DQB1*0201 allele were noted (Table 4), the most prominent being the DQB1*0201–high humidity exposure interaction (OR, 0.34; p = 0.026). As shown in Table 5, for individuals with neither the DQB1*0201 nor DQB1*0602 allele, exposure to high humidity in the workplace for greater than 1 year was a risk factor for sarcoidosis (OR, 1.62), but in individuals with the DQB1*0201 allele this exposure had a protective effect (OR, 0.47). No statistically significant exposure interactions were observed for the DQB1*0602 allele, with the interaction odds ratios for this allele all relatively close to one (Table 4). Correspondingly, the exposure odds ratios were comparable between individuals with the DQB1*0602 allele and individuals without the DQB1*0602 and DQB1*0201 alleles (Table 5). We tested whether the presence of the DQB1*0201 or *0602 allele was associated with disease presentation and clinical course (Table 6) . Incomplete medical record data at the time of presentation and/or during follow-up precluded complete phenotyping of all probands. Data completeness ranged from 85% of probands for presenting chest X-rays to 59% for slit lamp examination for detection of eye involvement. The strongest phenotypic association with the DQB1*0201 allele was observed for presenting chest X-ray (8.0% with a Stage 0 or I chest X-ray had the *0201 allele, compared with only 4.6% with a Stage II or greater chest X-ray), but this difference did not reach statistical significance (p = 0.334). No other associations between DQB1*0201 allele and proband phenotypic characteristics approached statistical significance. For the DQB1*0602 allele, the most striking associations were with follow-up chest X-ray (p = 0.051) and the change in chest X-ray between presentation and follow-up (p = 0.032). Most recent follow-up radiographs were performed up to an average of 7.9 ± 6.4 years after time of presentation. For those with both presentation and follow-up chest X-rays, the average time between radiograph at time of presentation and most recent chest radiograph was 6.5 ± 5.1 years. A suggestive positive association was also observed between the DQB1*0602 allele and clinical course, presenting chest X-ray and presence of any extrathoracic disease.
HLA Class II functions to trap peptide antigenic fragments derived from endocytosed or newly synthesized protein in the endosome of antigen-presenting cells (42). This HLA antigen–peptide complex ends up at the cell surface, where it is presented to CD4+ T lymphocytes. Peptide binding to Class II proteins depends on interactions between HLA anchor pockets and peptide anchor residues and on the formation of many hydrogen bonds between the peptide and the HLA protein (43). Thus, an individual's immune responsiveness to a peptide relates directly to the polymorphisms in these pockets. Schurmann and coworkers reported an overrepresentation of DQB1*0603 and DQB1*0604 alleles shared by affected first-degree relatives with sarcoidosis (44). DQB1*0301 and *0501 were found to be increased in Japanese patients compared with healthy control subjects (45) and DQB1*0201/*0202 (alleles not separated) were found to be increased in Scandinavians (24). Haslam and coworkers (27) reported greater increases in levels of HLA-DQ and DP on alveolar macrophages in sarcoidosis compared with HLA-DR and further noted an inverse correlation between HLA-DQ levels on alveolar macrophages and lung function in sarcoidosis. Sato and colleagues evaluated HLA-DQB1 alleles in U.K. and Dutch patients with sarcoidosis (46). These authors found no difference in the frequency of DQB1*0201 between U.K. and Dutch patients and control subjects, but found a decreased frequency of the DQB1*0202 allele in the UK patients with sarcoidosis. This decrease in *0202 was not confirmed in the Dutch patients with sarcoidosis. No difference was found for *0602 in either population. The DQB1*0201 allele was found associated with milder disease at presentation and with resolution on chest X-ray, whereas the *0602 allele was found associated with unfavorable clinical characteristics such as uveitis and higher chest X-ray stage. These results are also consistent with that reported by Visser and coworkers (47), who reported that DQB1*0201–DRB1*0301 haplotypes were associated with acute sarcoid arthritis that was self-limiting and associated with a good prognosis. Although we found no association between the DQB1*0201 allele and sarcoidosis presentation or prognosis in our population, we did find a suggestive association of DQB1*0602 with higher stage of chest X-ray at presentation and progression of chest X-ray at follow-up. Because our study was observational with retrospective data collection methods, medical record data on a substantial proportion of probands were incomplete and our sample may not be representative. Allelic frequencies of HLA Class II molecules are known to vary widely between populations (48) The MHC Class II DQB1 gene has 36 known alleles with serologic equivalents (49). In both white and African American populations, DQB1 allelic frequencies are highly variable (48), which may lead to detection of population association rather than disease associations in case-controls studies. While in theory, selection of a control sample from the same population from which cases are drawn should avoid this pitfall, in practice uniform case/control sampling is difficult to achieve. Statistical methods to adjust for population stratification using population-specific alleles have been proposed (50), but these methods have yet to be applied widely and may be prone to overcorrection of potential biases. Family-based studies provide a sampling procedure by which to avoid the potential bias of case-control sampling without making any elaborate statistical corrections (51). In our study, we applied a unified family-based association test (38) in an effort to examine whether one or more DQB1 alleles were associated with sarcoidosis in African American nuclear families. Our study design controlled for genetic background that can lead to spurious associations in the presence of genetic admixture and our statistical approach maximized the allelic transmission data in families where parental information was often incomplete. A potential limitation of using both affected and unaffected sibs in family-based association testing, as we did in our study, is that some unaffecteds may actually be misclassified undiagnosed, subclinical cases. This limitation is offset by reports that subclinical sarcoidosis in African Americans is not as common as in white individuals (52, 53) In addition, even for diseases with a relatively low prevalence, such as sarcoidosis, the statistical power of a family-based association test that uses both affected and unaffected sibs is comparable to that of tests using the traditional nuclear family trio with one affected offspring (39). Despite controlling for genetic admixture within families, if a study sample is genetically heterogeneous, the ability to detect genetic associations can be limited. In an unpublished analysis of data used to estimate linkage disequilibrium in our African American population (54), we estimated that 90% of our study sample had between 16.6 and 22.5% white admixture, which suggests less within-sample heterogeneity than in African American samples drawn from different geographic areas of the United States (55). In addition to susceptibility, we also examined the potential of environmental interaction with HLA Class II molecules that are encoded by the DQB1*0201 and *0602 alleles. No significant gene–environment interactions emerged for the DQB1*0602 allele, but the effect of several exposures on sarcoidosis susceptibility appeared to differ with respect to the DQB1*0201 allele. Exposure to high humidity and water damage in the work environment both showed significant interaction odds ratios with the DQB1*0201 allele. Exposure to vegetable dust had the highest odds ratio for DQB1*0201-positive individuals, but the interaction odds ratio was not statistically significant. To our knowledge, no previous studies have reported gene–environment interaction acting in sarcoidosis susceptibility. Although our results are provocative, they must be interpreted within the context of the limitations of our study hypotheses and exposure assessment methodology. The exposure questionnaire was based on the survey instrument used in the ACCESS study (34), which was constructed to elicit a minimum amount of information on a maximum amount of exposures. As such, no specific exposure hypotheses or mechanism to validate self-reported exposures existed. For example, vegetable dust is defined as respirable fine particles generated from crop production. Individuals who reported vegetable dust exposures typically had occupational histories in agriculture-related industries, but the actual exposure occurrence was based solely on the respondent's perception. Timing of these exposures is also an issue in that the preclinical period for sarcoidosis is unknown, and therefore it was not possible to determine whether the exposure occurred exactly when disease pathogenesis was initiated. Despite the limitations of our exposure data, our gene–environment results do provide some avenues for future investigations that might involve more in-depth exposure assessments and determination of HLA Class II molecule-binding specificities with respect to exposures associated with DQB1*0201 and sarcoidosis susceptibility. Foley and coworkers (8) concluded that MHC associations in sarcoidosis reflect involvement of HLA-DR rather than other Class II loci. This was supported by the finding that associations detected between single-nucleotide polymorphisms at the TAP-2 locus are due to linkage disequilibrium with the HLA-DR1 allele in both U.K. and Polish populations. Further, it was concluded that involvement of the HLA-DQ locus is also unlikely, considering that the presence of a specific HLA-DR allele in a haplotype, rather than an HLA-DQ allele, appears to determine disease susceptibility. Because linkage disequilibrium complicates the interpretation of the HLA associations, controversy exists about whether several HLA factors or a single HLA allele or genotype can explain sarcoidosis associations with the MHC region. For instance, in one report DQB1 associations with sarcoidosis were explained by the tight linkage disequilibrium of the DQB1 gene with DRB1 (45). In white populations, the DQB1 and DRB1 loci are in tight linkage disequilibrium, with the allele at one locus generally predicting the allele at the other locus (56). The degree of linkage disequilibrium between the DQB1 and DRB1 loci in African Americans may not be as strong (57). For example, more diversity in HLA-DRB1/DQA1/DQB1 haplotypes has been found in African Americans compared with white individuals (58). In the same population used in the present study, we reported an association scan of the HLA Class II region that suggested DQB1 associations were distinct, and not serving as surrogates for DRB1-associated alleles (31). Validating which HLA molecules are directly involved and defining their functional role in sarcoidosis await more direct experimental studies such as developing transgenic mice that carry only specific human MHC molecules, as has been done for diabetes (59, 60). We took a stepwise conservative approach in our analysis of HLA-DQB1 alleles. In first examining associations of sarcoidosis with low-resolution alleles, and then subjecting to high-resolution typing only the associated low-resolution allelic subsets, we lessened the possibility of reporting spurious associations. Although it is possible that we did miss some real associations with high-resolution *03, *04, and *05 alleles, it is unlikely that these alleles had particularly strong effects on sarcoidosis susceptibility, as otherwise we would expect that their effects would have been detected in our analysis or the low-resolution alleles. Our study, along with those of others, support that notion that HLA-DQ is important in sarcoidosis. It appears that although *0201 is protective, this may depend on an individual's exposure status. We also found that the presence of the *0602 allele was associated with radiographic progression of disease, but that the *0201 allele showed no significant correlation with phenotype. Further work is needed to assess how HLA-DQ interacts with environmental factors and other genes in sarcoidosis.
The authors thank the patients and their families for participating in this study; Wanda Bibbs, R.N., for work in enrolling study participants; and Marcie Major, R.N., research coordinator.
Supported by National Heart, Lung, and Blood Institute grant RO1 HL54306–01. This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Received in original form September 25, 2002; accepted in final form January 31, 2003
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
with sarcoidosis in 225 african american nuclear families under dominant and additive inheritance models
