INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The eosinophil uniquely characterizes the airway obstruction found in asthma (1). In this disorder, eosinophils are believed to be major contributors in the development of airway inflammation and bronchial damage (2). Eosinophils are released into the peripheral blood for short periods after a process of maturation and differentiation that occurs in the bone marrow. Circulating eosinophilia is frequently observed in asthma and has been found to be directly correlated with airway hyperresponsiveness, which is considered a hallmark of the disease (3). The number of circulating eosinophils is thus an important component of the asthmatic inflammatory response.

Although genetic susceptibility to the development of asthma has been clearly established (4), surprisingly few studies have addressed the possible role of heredity in determining eosinophilia. As part of a genome-wide scan, Daniels and colleagues (5) reported that the log of eosinophil blood counts showed evidence for linkage with markers on chromosome 6.

The long arm of chromosome 5 contains a large number of cytokine genes, many of which may be involved in determining eosinophilia. The cytokines granulocyte macrophage-colony-stimulating factor (GM-CSF), interleukin (IL3), and IL-5 are important in promoting eosinophilopoiesis from stem cells (6), and their structural genes are all located in close proximity to each other in chromosome 5q31-35. These three cytokines also block eosinophil apoptosis in vitro (7). Studies of asthmatic families have suggested that both total serum IgE (8) and bronchial hyperresponsiveness (BHR) (9) are controlled by genes located in chromosome 5q, and that the locus controlling BHR exerts its influence independent of that controlling total serum IgE (9). Since, as stated earlier, airway hyperresponsiveness is associated with increased numbers of eosinophils and their secondary products both in circulation and in the airways (10), it is possible that genes controlling for circulating eosinophils could also be involved in determining BHR in asthmatic patients.

We conducted a study in which we performed single-point and multipoint sib-pair analyses to assess linkage between markers on chromosome 5q and circulating eosinophils (as a percent of total leukocytes) in nuclear families living in Tucson, Arizona.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The families involved in the study were enrolled in the Tucson Children's Respiratory Study, a long-term longitudinal assessment of risk factors for asthma and allergies in childhood. Details of the population under study and the methods used for enrollment have been provided elsewhere (11). A total of 1,151 families were enrolled at the time of the index children's birth. Owing to the marked ethnic heterogeneity in the genetic determination of asthma (12), only nuclear families for which both parents defined themselves as non-Hispanic whites were eligible for this study (n = 674). The study was approved by the Institutional Review Board of the University of Arizona. Informed consent was obtained from parents.

When the index children were approximately 6 yr old, a blood sample was obtained by phlebotomy for all members of the nuclear families who were  5 yr of age and who gave consent and cooperated. Blood smears were made for all subjects and stained with standard hematologic techniques. A differential cell count based on 300 leukocytes was performed. A questionnaire was also administered to parents. Among other inquiries, parents were asked whether they or their children had asthma, and if so, whether they had had episodes of asthma and/or wheezing during the previous year. Subjects were considered to have asthma if both questions were answered positively. Parents were also asked about their smoking habits. When the index child was approximately 11 yr old, all available members of the nuclear family were again examined. Blood smears and differential leukocyte counts were obtained for those subjects who were not available or had not consented to participation during the previous survey. Approximately 10% of blood samples were obtained at this later examination. However, there was no difference in eosinophil levels by age or month of year of blood sample (p = 0.37, and p = 0.18, respectively). Total serum IgE levels were measured using the paper radioimmunosorbent test (PRIST) technique (13) at the same age at which eosinophil counts were performed. A total of 562 non-Hispanic white families had at least two children, and of these, 123 families had both blood for smears available for at least two siblings and genomic DNA available for both parents and for at least two siblings.

DNA Extraction and Genotyping

For all participating subjects, genomic DNA was extracted from blood according to conventional methods, wherein: (1) the cells were lysed with a nonionic detergent; (2) the nuclei were pelleted and then lysed with a hypertonic buffer; (3) nuclear and residual cellular proteins were solubilized with sodium dodecyl sulfate (SDS) and removed with phenol/chloroform, and (4) DNA was precipitated with cold ethanol.

Nine polymerase chain reaction (PCR)-based, highly polymorphic, dinuclotide repeat anonymous markers, spanning approximately 28 centimorgans (cM) in chromosome 5q31-5q33, were used, as follows: D5S642, D5S393, D5S479, D5S399, D5S500, D5S658, D5S210, D5S519 and D5S209. Distances between the markers (7 cM, 2 cM, 1 cM, 2cM, 1cM, 5cM, 7cM, and 3cM, respectively) were obtained from published maps (14, 15) or from our own data. Table 1 shows the sequence of the PCR primers, the number and size of the alleles, the heterozygosity, and the PCR conditions used to genotype the sibships for each of these markers. Fluorescent primers were obtained from commercial sources (Bioserve, Laurel, MD; Genosys, Woodlands, TX; Research Genetics, Huntsville, AL). The ABI 373 sequencer (Perkin Elmer, Foster City, CA) was used to automatically read the bands obtained by gel electrophoresis and to assign genotypes to each individual based on the marker phenotypes. Markers were checked for consistency with Mendelian inheritance within families, and ambiguous readings were regenotyped. Haplotypes were checked for double crossovers, but only those in which the apparent double crossover was clearly shown to be due to mistyping were corrected. All genotyping procedures were performed with blinding to the subjects' phenotypes.

Linkage Strategy

Recent reassessments of the best strategies for selecting sib pairs for the study of quantitative trait loci (QTLs) have shown that the power to detect linkage is concentrated in sib pairs that are concordant for low values (LC), concordant for high values (HC), or discordant (DC) for the quantitative trait (16). It has also been shown that the threshold for low or high values for a given QTL is a function of the mechanisms of inheritance and the expected allele frequencies for the gene in question, as is the best combination of HC, LC, and DC sib pairs to be used in linkage analysis of QTLs (16).

Since the mechanism of inheritance of circulating eosinophils has not been clearly established, we opted for a two-stage, cross-validating strategy (17), using selected sib pairs. We initially defined as having the HC phenotype those subjects who had 5% or more circulating eosinophils, a threshold that has often been used to define eosinophilia in epidemiologic studies (18). The choice of threshold for the LC phenotype (i.e.,  2% circulating eosinophils) was in accordance with our own results for the association between prevalence of asthma and percent circulating eosinophils (see RESULTS). We first assessed evidence for linkage using LC, HC, and DC sib pairs defined as such in a subgroup consisting of approximately 60% of all sib pairs available. This initial group was chosen by giving preference to families with three or more children (n = 71). In a second phase, we performed a cross-validating study in the remaining group of 52 families with at least two children selected for sib pairs (in our case, 19 LC sib-pairs) that showed significant evidence for linkage in the first phase.

Statistical Analysis

Single-point and multipoint linkage analyses were performed with the SAGE SIBPAL program (19) and the MAPMAKER/Sibs software (20). The SAGE SIBPAL program provides an estimate of the proportion of alleles shared identically by descent for both concordant and discordant sib pairs. This software also provides a statistical test for deviation from the expected value of 0.5 (toward values of < 0.5 in the case of discordant sib pairs and of > 0.5 in the case of concordant sib pairs).

SIBPAL also applies the Haseman-Elston regression method (21), in which the squared sib pair difference for the trait in question is regressed on the estimated proportion of alleles shared identically by descent, and in which a test for negative slope is performed. The regression coefficient is a function of both the additive genetic variance attributable to the "affection" locus and the recombination fraction between the marker and affection locus. The method has the advantage of permitting covariates to be included in the regression models, and we therefore used this method to adjust for the effects of the following covariates for sib pairs: age when blood was drawn for eosinophil assessment; gender; total serum IgE level adjusted for age and gender; current and past exposure to environmental tobacco smoke; and a report of any wheezing in the year prior to phlebotomy. As suggested by Wilson and Elston (22), the degrees of freedom (df) for this analysis are based on the effective sample size rather than on the total number of sib pairs in the analysis.

In our final assessment of linkage done with all available LC sib pairs, the multipoint techniques contained in the MAPMAKERS/Sibs software (20) were used. These techniques are based on the assumption that if a marker is linked to a locus controlling for the expression of a particular phenotype, the proportion of all affected pairs sharing 0, 1, or 2 alleles identical by descent (ibd) is expected to deviate significantly from the Mendelian ratios (i.e., 1:4, 1:2, and 1:4, respectively). The software computes the probabilities of sharing 0, 1, and 2 alleles ibd for each sib pair at preestablished (usually 1 cM) intervals according to the available data for all genetic markers. This increases the power of the analysis at each genotyped point because it allows probabilities of ibd sharing to be computed for uninformative sibships, and also allows calculation of the most likely ibd status at points between genotyped markers. The ratio of the likelihood of the data given biologic consistency constraints (23) to the likelihood under the Mendelian alternative can thus be calculated for each sib pair, and the addition of the logarithms of the odds favoring genetic linkage (LODs) of these ratios for all sib pairs is the LOD score. LOD scores can be determined with or without the assumption of no-dominance variance. To calculate the probability of the LOD score under the null hypothesis, one can calculate the statistic (LOD · 4.6)^1/2, which is asymptotically equivalent to a one-sided standardized normal distribution (24).

There is considerable debate about the LOD score threshold to be used for a false-positive rate of 5%, the traditional level of statistical significance. Lander and Kruglyak (24) have suggested a threshold of 3.6 for genome-wide searches, but recent reassessments based on simulations suggest that the true threshold for genome-wide studies is closer to the traditional LOD score of 3.0 (25). There is consensus, however, that for studies of candidate regions (like our study), a lower threshold is more appropriate. A threshold of 2.3, with a nominal significance level of 0.0005, was recently adopted by Morahan and associates (26) in a study of linkage between insulin-dependent diabetes mellitus and a candidate region in chromosome 2q.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There was a strong correlation between physician-diagnosed active asthma and percent of circulating eosinophils among all available members of the enrolled families (n = 2,345; Figure 1). A threshold value of 2% eosinophils seemed to define increased risk for asthma, and there was an apparent linear association between risk for asthma and percent circulating eosinophils beyond this threshold (Figure 1).

View larger version (12K): [in this window] [in a new window]   Figure 1.   Association of circulating eosinophils (as % of total white blood cells) with prevalence of asthma among all enrolled subjects (including index children, their parents, and their siblings).

A comparison of the characteristics of the 123 non-Hispanic white families included in the analysis with the 551 families not included indicated that the parents of both sets of families were equally likely to be married at the time of the index child's birth (98% for those included versus 96% for those not included, p > 0.7), and that at least one parent was likely to have more than a high-school education (91% versus 85%, p > 0.09). There was also no significant difference between families included and not included in the percent of families with at least one parent who had physician-diagnosed asthma (26% versus 25%, p > 0.75), or physician-diagnosed hay fever (63% versus 61%, p > 0.65). In addition, mothers and fathers, respectively, in each set of families were just as likely to smoke (18% versus 19% for mothers, p > 0.70; 30% versus 28% for fathers, p > 0.50).

Out of 246 sib pairs available for the nuclear families selected for the first phase of this study (see METHODS), 35 were classified as LC (both sibs had 2% or less circulating eosinophils), 18 were defined as HC (both sibs had 5% or more circulating eosinophils), and 26 were defined as DC (one sib had 2% or less and the other sib had 5% or more circulating eosinophils). Results of sib-pair analyses done with the nine markers located on chromosome 5q are shown in Table 2. No significant evidence for linkage was found for HC sib pairs or for DC sib pairs. In contrast, LC sib pairs shared 60% or more alleles ibd for five of the nine markers in this chromosomal region. A peak was observed for markers D5S500 and D5S658, with mean proportions of alleles shared ibd of 0.68 and 0.65, respectively. On the basis of these results, the second, cross-validating study was concentrated on all other LC sib pairs in the study population sample of non-Hispanic white children (n = 19 sib pairs) for whom both parents had genomic DNA available. As shown in Table 3, these LC sib pairs shared 60% or more of alleles ibd for five of the nine markers reassessed on chromosome 5q, and nominal significance levels (p < 0.05) were reached for four of these markers. Since the cross-validating study showed the same trends as those observed in the first phase described earlier, the two data sets were merged with results shown in Table 4 for the 54 resulting sib pairs. The mean proportion of allele sharing ibd was greater than 0.55 for all markers assessed except for D5S209, with a peak of 0.64 for markers D5S500 and D5S658.

Table 5 shows the results of the linear regression analysis done to further test for possible linkage. The negative slope was statistically significant for markers D5S500 (p < 0.002) and D5S658 (p = 0.005). The association with these two markers remained statistically significant in the presence of all covariates tested (see METHODS), including total serum IgE and reports of wheezing during the previous year (data not shown). A nonlinear relationship with the covariates total serum IgE and age at the time of blood sampling was not significant and did not affect the statistical significance for the two markers. There was no significant correlation between the differences between members of the sib pairs in eosinophil levels and the differences between those same sib-pair members in the levels of any of the covariates. In addition, the use of medications did not explain the linkage findings. We estimate that approximately 5% of children overall used asthma or cough medications, and that only 1% in the LC group used medications.

Figure 2 shows the results of a multipoint analysis of LC sib pairs with respect to a map of the studied markers in chromosome 5q. A clear peak is observed in the area corresponding to marker D5S658, with an LOD score of 2.4 (p = 0.0004). Results were very similar with or without the assumption of no-dominance variance.

View larger version (11K): [in this window] [in a new window]   Figure 2.   Multipoint sib-pair linkage analysis of circulating eosinophils to markers on chromosome 5q. A total of 53 sib pairs classified as low concordant for circulating eosinophils were included. For definition of low concordant, see text.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Epidemiologic studies have suggested that circulating eosinophilia is associated with resistance to parasitic infestation (27, 28), with prevalence and severity of asthma (2, 10), and with the likelihood of survival in subjects undergoing radiation therapy for tumors (29). In this study we have shown significant evidence for linkage between markers located on chromosome 5q31-33 and circulating eosinophils (as a proportion of total leukocytes) in an unselected population sample. These results suggest that a variant or variants in one or more genes present in this area of chromosome 5q control for the number of circulating eosinophils.

We chose this specific region of chromosome 5q because it contains the structural genes for the main three cytokines that have been shown to regulate eosinophil production, activation, and localization, namely: IL-5, IL-3, and GM-CSF. These three genes map to a small segment of chromosome 5q31 centromeric from the marker D5S393 used in this study (9). Of these three cytokines, IL-5 seems to be the most important determinant of peripheral eosinophilia. This is supported by data showing that anti-IL-5 antibody completely inhibits eosinophil production (30) and that transgenic mice expressing IL-5 only in the bronchial epithelium show marked eosinophilia (31). It is thus logical to surmise that the linkage signal observed in this study could be due to a variation in the structural gene for IL-5 or in its promoter region. However, the highest LOD value in our data was reached for marker D5S658, which is located several centimorgans distal from the putative location of the IL-5 gene. Although this does not exclude the gene for IL-5 as a possible candidate gene for our finding, other known or unknown loci may exist in this region that explain our results.

Of particular interest is that linkage of asthma-related phenotypes to markers located in this same segment of chromosome 5q has been recently reported. Meyers and colleagues (8) found linkage between marker D5S436 and total serum IgE levels in a group of Dutch families selected through an asthmatic proband. Using this same group of families, Postma and associates (9) found linkage of D5S436 with BHR, and this signal seemed to be independent of total serum IgE levels. Interestingly, our linkage signal with circulating eosinophils was also independent of total serum IgE. Doull and coworkers (32) reported significant association between one allele in a marker located in the IL-9 gene on chromosome 5q and total serum IgE levels. A recent report by Noguchi and colleagues (33) suggests the presence of loci for atopy and asthma on chromosome 5q31-33. Results of the Collaborative Study for the Genetics of Asthma (12) also suggest that among non-Hispanic white families, markers located on chromosome 5q show evidence for linkage with the risk of having asthma. The combined evidence from all these studies thus strongly suggests that more than one gene on chromosome 5q31-33 may contain polymorphisms that determine increased susceptibility for asthma and for asthma-related traits.

Genetic variations in chromosome 5q31-33 also seem to be important in determining susceptibility to parasitic infestation. Marquet and coworkers (34) recently reported the results of a genome-wide search for loci determining the intensity of infection by Schistosoma mansoni in Brazilian families. The only area of the genome in which evidence of linkage was observed was located in the region between markers D5S393 and D5S410 on chromosome 5q. The point at which the highest LOD score value was observed was located approximately 6 cM distal to marker D5S436. These findings have been recently replicated by Muller-myhsok and colleagues (35). Since resistance to Schistosoma mansoni has been shown to be associated with eosinophilia in African populations (27), it is tempting to speculate that our study and that of Marquet and coworkers may have detected a genetic variation or variations in the same gene. Further studies leading to the identification of the gene or genes responsible for these two linkage signals will be needed to resolve this issue.

Our linkage signal was observed only among sib pairs who were LC for circulating eosinophils, and this was corroborated in our cross-validating study. Recent simulation studies (16) have suggested that high frequency dominant genes controlling for high values of a quantitative trait may be best detected by using LC sib pairs. Rodrigues and associates (36) performed a segregation analysis of IL-5 responses to specific and nonspecific stimuli by peripheral blood mononuclear cells of subjects infected with Schistosoma mansoni. They observed that the inheritance pattern of this phenotype was compatible with the presence of a recessive gene with a low frequency controlling for low responses. One or more recessive, rather infrequent alleles may thus be present in genes on chromosome 5q controlling for these different, low response, eosinophil-related phenotypes.

Identification of the genes and gene products responsible for these linkage signals may have important implications for understanding of the mechanisms of defense against parasitic infestation and of genetic susceptibility to the development of asthma.