Meta-analysis of genome-wide linkage studies of asthma and related traits

To identify published studies for inclusion in the GSMA of asthma and related phenotypes we completed a systematic literature review in September 2006. We used PubMed and the following search string (Asthma OR BHR OR bronchial hyper responsiveness OR bronchial hyperreactivity OR AHR OR airway hyper responsiveness OR respiratory hypersensitivity OR histamine OR slope OR methacholine OR atopy OR atopic OR dermatitis, atopic OR IgE OR immunoglobulin E OR SPT OR skin prick tests OR skin tests) AND linkage AND genome-scan OR scan OR genome OR genomewide OR genome-wide OR LOD OR microsatellite). Limits were set on the search including; published in English, human studies, published 1996–2006 and the exclusion of reviews. This initial search identified 516 matches of which 488 were discarded as not containing genome-wide linkage data. A further eight studies were discarded as they were in non-Caucasian populations and we wished to avoid any population stratification issues leaving 20 potential Caucasian studies for inclusion. Genome-wide linkage analyses for asthma related traits in the Hutterite Founder population [6] was not included in the current analyses as limited data was available and the focus of the present study was Caucasian out-bred populations.

Of the 20 manuscripts identified a further nine were removed from the analyses for a combination of the following reasons; the study was superseded by another including the families from the original, LOD score plots in the manuscript were not labelled and/or unreadable, no genome-wide data was presented e.g. in the manuscript describing the positional cloning of ADAM33, linkage analyses in 460 families for asthma, IgE and BHR phenotypes were performed but has never been published in full [7] or the phenotypes studied did not meet our criteria. All authors were contacted and invited to provide complete datasets.

There was a large degree of heterogeneity in phenotype definitions and so these were standardised for inclusion. Asthma was defined using doctor diagnosis and/or currently taking asthma medication, however we did include data from the Dutch population which used an algorithm based on asthma symptoms, the presence of BHR, reversibility to β₂-adrenergic receptor agonist and smoking history to define asthma [8]. Analyses were completed with and without the Dutch families. Total IgE levels were analysed in the genome scans using quantitative data generated by Pharmacia CAP system [9], Pharmacia IgE EIA [10], Phadebas PRIST [11] and ELISA techniques [12, 13] which have shown good inter assay correlation [14, 15], therefore all studies were included. Positive skin prick response to one or more allergen was used as a marker of atopy and for inclusion in the GSMA. However, allergens used in each study varied; Dermatphagoides pteronyssinus, mixed grass pollen [9], Dermatphagoides pteronyssinus, Cladosporium herbarum, Alternaria tenuis, timothy grass, olive, birch, Parieteria judaica, ragweed, Aspergillus, Blatella Germancia [11], mixed grass and tree pollens, mixed weeds, Dermatphagoides pteronyssinus, dog, cat, a mixture of guinea pig and rabbit, horse, Aspergillus fumigatus, Alternaria alternate [10], house dust mite [12], Dermatphagoides pteronyssinus and 10 others [13] and Dermatphagoides pteronyssinus, D. farinae, dog, cat, grass mix, pollen and alternaria [16]. BHR was measured in multiple ways using different provocation stimuli e.g. histamine or methacholine providing categorical and/or quantitative analyses. These provocation stimuli have shown a significant correlation (r = 0.95) in the responses induced [17], however it is worth noting that this has not been reproduced in all studies. Studies with BHR data were included in the GSMA irrespective of the criteria used in the original manuscript.

GSMA was used as it is able to combine linkage data from studies with different marker sets and analysed by different methods including permutated p-values. GSMA was implemented using GSMA software [4, 5, 18]. Briefly, the genome was divided into 120 bins of approximately 30 cM, for each study the maximum evidence for linkage e.g. LOD score or p-value was identified for each bin and these bins were then ranked relative to their evidence for linkage in that study. These ranks were summed across studies and the summed rank (SR) forms the basis of the test statistic [4]. An ordered rank (OR) statistic was also generated which gives a genome wide interpretation of significance by comparing the n-th highest summed rank with the distribution of the n-th highest summed ranks obtained through simulation [5]. We completed an unweighted and weighted analyses using information content (√(no. pedigrees × no. markers)) as a weighting factor.

Simulation studies have shown that any bin with a p(SR) < 0.05 and a p(OR) < 0.05 has a high probability of containing a true susceptibility gene [5]. Applying Bonferroni correction a p < 0.000417 provides evidence for genome wide significance for linkage and a p < 0.0083 provides suggestive evidence for linkage [5].

The selection criteria and data requests provided eleven studies of nine Caucasian asthma populations for our analyses (Table 1) including data from 1,267 pedigrees (n = 5,832) for asthma (80.2% of pedigrees available in the public domain or following request, missing 249 [19] and 65 pedigrees [20]), 989 pedigrees (n = 4,294) for BHR (79.9% of available, missing 249 pedigrees [19]), 1,093 pedigrees (n = 4,746) for SPT (81.5% of available, missing 249 pedigrees [19]) and 729 pedigrees (n = 3,224) for total IgE (65.9% of available, missing 249 [19], and 129 pedigrees [21]).

The weighted asthma analyses did not identify any chromosomal region with a p(SR) and p(OR) < 0.05 (Figure 1 and Table 2). No bin p(SR) met genome wide significance (p < 0.000417) or suggestive evidence for linkage (p < 0.0083) in these analyses however three regions demonstrated a p(SR) < 0.05; 6p22.3-p21.1, 10p14-q11.21 and 12q24.31-qter (Table 2). Eight regions met suggestive linkage criteria in the ordered rank analyses; 1p31.1-p13.3, 2p12-q22.1, 4p14-q13.3, 7q34-qter, 12pter-p12,1, 12p12.1-p11.21, 14q32.12-qter, 17pter-p12 and 20pter-p12.3 (Table 2). Analyses of the asthma phenotype using unweighted GSMA generated similar findings to the weighted analyses (Figure 2). To confirm that the inclusion of the Dutch linkage data for the asthma phenotype (defined by algorithm) had not confounded the analyses we completed GSMA without these data focusing on doctor diagnosed asthma only (1,067 pedigrees). Again, no chromosomal region with a p(SR) and p(OR) < 0.05 was identified (data not shown).

The weighted BHR analyses strongly suggested that 6p22.3-p21.1 contains BHR susceptibility gene(s) as a p(SR) and p(OR) < 0.05 was observed (p = 0.007, p = 0.049 respectively (Figure 1 and Table 3)). Two other regions showed suggestive evidence (p < 0.0083) for linkage to the BHR phenotype; 2p12-q22.1 (p(SR) = 0.006) and 11q24.1-qter (p(SR) = 0.005). In the unweighted analyses three regions showed evidence for linkage (p(SR) and p(OR) ≤ 0.05), i.e. 2q22.1-q23.3, 7q12.11-q31.1 and 5q23.2-q34 (Figure 2).

Weighted analyses of the SPT phenotype identified two regions that had a p(SR) and p(OR) ≤ 0.05 strongly suggesting these regions are susceptibility loci (Figure 1 and Table 4). These regions are 17p12-q24.3 (two adjacent bins, 17p12-q21.33 p(SR) = 0.00043, p(OR) = 0.050 and 17q21.33-q24.3 p(SR) = 0.047, p(OR) = 0.038) and region 3p22.1-q22.1 (three adjacent bins, 3p22.1-p14.1 p(SR) = 0.045, p(OR) = 0.063, 3p14.1-q12.3 p(SR) = 0.003, p(OR) = 0.0045 and 3q12.3-q22.1 p(SR) = 0.00084, p(OR) = 0.00406). The analyses of the unweighted SPT datasets identified chromosomes 3 and 17 as containing the major determinants (Figure 2).

Weighted analyses of the IgE phenotype strongly suggested 5q11.2-q14.3 (p(SR) = 0.031, p(OR) = 0.060) and 6pter-p22.3 (p(SR) = 0.033, p(OR) = 0.026) contain genes that influence IgE levels (Figure 1 and Table 5). The region adjacent to 6pter-p22.3, i.e. 6p22.3-p21.1 has a p(SR) = 0.00999 approaching suggestive linkage providing further evidence for this region. Analyses of the IgE phenotype using the unweighted GSMA showed similar overall findings (Figure 2).

In addition to the primary phenotype analyses we investigated overlapping chromosomal regions containing genetic determinants of asthma and asthma related traits consistent with gene(s) having pleiotrophic effects in asthma and allergy (Table 6). Interestingly, region 6p22.3-p21.1 which contains the HLA region showed a p(SR) < 0.05 in the asthma, BHR and IgE analyses potentially as expected due to the role of HLA restriction in many immunological mechanisms. Several other regions also showed overlapping concordance, in particular regions; 3p14.1-q12.3 (asthma, SPT), 5q23.2-q34 (asthma, BHR, IgE) and 7p21.1-14.1 (asthma, BHR, IgE).