INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Studies in a number of different populations have demonstrated familial aggregation of lung function (1, 2), with genetic heritability estimates for FEV₁ ranging from 30% (3) to 41-47% (4). An additional amount of variance in lung function, ranging from 1 to about 30%, depending on the age group and variables considered, has been explained by common familial environmental factors (4). Several twin studies, which have compared the relation of lung function parameters in monozygotic and dizygotic twins, have also concluded that a large proportion of variability in lung function can be accounted for by genetic influences with additional environmental effects. For example, heritability estimates for FEV₁ as high as 77% were noted in monozygotic twins (5), while comparisons of similarities in lung function among relatives differing in their degree of shared environment suggested a multifactorial mode of inheritance (6).

Despite the many studies suggesting a familial component in the determination of lung function, there have been few investigations of potential modes of inheritance. Such investigations have applied segregation analysis, using regressive models to assess Mendelian modes of inheritance. They have shown evidence for a major codominant gene for FEV₁ in families with chronic obstructive pulmonary disease (COPD) (7) or suggested polygenic control of FEV₁ or common environmental factors (8).

The aim of this study is to search for Mendelian inheritance patterns in the determination of lung function in family data. Also, since the phenotypic expression of lung function can be considered a potential subphenotype of the asthmatic condition, we have examined inheritance mechanisms for lung function in families with and without asthmatic members.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The subjects for this study were drawn from families enrolled in the Tucson Children's Respiratory Study (CRS), a large longitudinal study investigating the risk factors for asthma and other acute and chronic lower respiratory illnesses in infancy and childhood (9). Between 1980 and 1984, 1,246 healthy newborns from 1,151 families were enrolled in the CRS. All were members of one of the largest health maintenance organizations in Tucson at that time, which was a requirement for enrollment. This study was approved by the Human Subjects Committee of the University of Arizona, and parents have signed consent forms at all phases of the study. When enrolled children were approximately 6 yr old, all older siblings and parents who consented were given an in-depth evaluation, which included lung function measurement. Enrolled children and younger siblings had an in-depth evaluation that included lung function testing when they were approximately 10 yr old.

Spirometry, conforming to the American Thoracic Society Snowbird criteria (10), was performed with a pneumotachograph as described by Knudson and colleagues (11). The pulmonary function index used in this analysis was FEV₁. A total of 674 mothers, 513 fathers, 545 siblings, and 424 enrolled children had lung function tests before or during March 1994. Only those families in which pulmonary function tests were available for the mother, father, and at least one child are included in this analysis. In addition, since one of our objectives was to assess genetic heterogeneity across ethnic groups, only families where the biological parents declared they were either non-Hispanic white (NHW) or Hispanic were included; there were not enough representatives of other ethnic groups to consider them separately. A total of 309 families with 1,163 individuals met these criteria (Table 1), comprising 51 families (202 members) with both parents Hispanic, 219 families (814 members) with NHW parents, and 39 families (147 members) with one Hispanic and one NHW parent. To reduce the effect of height on any given measurement of FEV₁, observed values were divided by the square of the height. This correction has been demonstrated to be most effective in adults and adolescents and to be an approximation for younger children (12). Observed values divided by height squared were further adjusted for age and height within each sex and ethnic group, using continuous regression equations, applying previously determined age breakpoints (13). Separate equations were derived for NHW and Hispanic parents and children and for children of an NHW and a Hispanic parent. Individual values were expressed as Z-scored residuals. There were no significant differences in adjusted FEV₁ between parents and children, nor by sex or ethnic group, since age, sex, and ethnicity had all been adjusted for. In addition, neither weight at the time of spirometry, nor a measure of body habitus (height/weight^1/3) were significantly related to the adjusted FEV₁, accounting for an additional 0.1% of the variance both individually and when combined.

Questionnaires administered at the time of the pulmonary function tests, completed by parents for themselves and on behalf of family members, provided information on currently active physician-diagnosed asthma, and smoking status (current or former smoker). Also, children over age 10 were asked privately if they smoked cigarettes currently or in the past. Current and past environmental tobacco smoke (ETS) exposure variables were created for children, being positive if either parent was a current or former smoker.

Statistical Methods

Familial correlation coefficients were calculated using the program FCOR within the Statistical Analysis for Genetic Epidemiology (SAGE) software package (14). Heritability is defined as the proportion of variation directly attributable to genetic differences among individuals relative to the total variation in a population (15). Empirical heritability, however, makes no assumptions about biologic mechanisms accounting for the proportion of variation in an offspring's trait directly attributable to the trait values in the parents. Such mechanisms may involve genetic and/or environmental factors. Empirical heritability was calculated as twice the regression coefficient of the child on one parent, or the coefficient itself from a regression of the child's values on the average of both parents (15). This measure can be used to predict offspring phenotype given a parent's phenotype.

To examine the fit of postulated genetic and nongenetic (environmental) models to the data, we applied segregation analysis using class D regressive models for continuous traits (16) as described by Martinez and colleagues (17). Briefly, segregation analysis for continuous variables is a method of modelling data to test whether a mixture of normal distributions fits the data better than a single distribution and to test whether Mendelian ratios can be detected in the transmission of a trait between generations. The model translates assumptions about genetic and environmental trait determinants into mathematical equations with a number of model parameters, shown in Tables 345, which are estimated. These models, which are incorporated into the SAGE software package (14), assume the trait to be a linear function of the major genotypes, the phenotypes and genotypes of antecedents, and other covariates. Mendelian inheritance, if present, is presumed to be through a single autosomal locus with two alleles, A and B, in our case with the putative A allele being associated with a high level of the pulmonary function index. Three types of individual are assumed, which for genetic models correspond to the three possible genotypes and the means, µ_AA, µ_AB, and µ_BB, of these distributions are estimated by the model. Random mating and Hardy-Weinberg equilibrium proportions are assumed (18). The type frequencies are defined in terms of q_A, the frequency of allele A, where the type frequencies () are as follows:

(1)

In addition, since the assumption of normality, conditional on type, is critical for the methods described above (18), the FEV₁ Z-scores were simultaneously normalized using the standardized Box-Cox transform that was implemented within the SAGE software package (14). In the results, the means of the type distributions are expressed as Z-scores that have been recalculated from the means of the Box-Cox transforms obtained from the different models, as described by Martinez and colleagues (17).

Model parameters were constrained (fixed) to define each of three hypothetical genetic (Mendelian) models and three nongenetic models, i.e., no major type or one distribution, and two environmental models that both disallow intergenerational transmission, but allow type distributions with or without heterogeneity between the generations. The addition of residual family effects (RFE), such as spouse- spouse, parent-offspring, or sibling-singling, to these models tests for the presence of a major gene plus an additional multifactorial component, representing genetic effects that may be non-Mendelian and/or the effects of a shared environment, beyond major gene segregation.

An unrestricted model in which parameters are unconstrained represents the best fit to the observed data. If the pattern predicted by the model is not significantly different from the pattern of the observed data, as represented by the unrestricted model, then the model fits the data and provides evidence in support of the model. Alternatively, if the model differs significantly from the observed data, then that model is rejected. The likelihood ratio test was used to evaluate specific hypotheses (17), with the degrees of freedom being the difference in the number of parameters estimated between models. In cases where parameter estimates converged to bounds, probability was assessed at the midpoint of a range of degrees of freedom as described previously (19). Briefly, comparing parameter to parameter, the maximum degrees of freedom was calculated by subtracting from the number of parameters in the unrestricted model the number of parameters maximized in the restricted model, but ignoring parameters that have the same boundary values in both models. For example, if a transmission probability is maximized in the unrestricted model at the same value at which it is fixed (constrained) under the hypothesis, one is cancelled against the other. The minimum was obtained by subtracting from this the number of parameters that went to bounds in the unrestricted model and adding the number that went to bounds in the restricted model, again ignoring parameters that have the same boundary values in both models; if this number was negative it was taken to be zero. In general, this procedure has the effect of reducing the number of degrees of freedom between model comparisons, making it more likely that specific hypotheses will be significantly different from the unrestricted model and therefore rejected. An example of the calculation of the degrees of freedom is given as a footnote to Table 3. Akaike's information criteria (AIC), which weights the ln likelihood by the number of parameters estimated, was also calculated, as previously described (17) to assess which model best fit the data in cases of competing models where one was not a strict subset of the other, with the same number of parameters. The AICs were calculated with and without parameters fixed at a bound (AIC_max and AIC_min, respectively). Since conclusions were the same with both AICs, only AIC_min is presented in the results.

Segregation analyses were performed with and without active smoking, ETS exposure, and current asthma status as covariates, within ethnic groups, and for the total group. Analyses were also conducted, with ascertainment correction, for the 87 families (328 members) who had at least one family member with current physician-diagnosed asthma and in the remaining 222 families (835 members) with no asthmatic members. Tests for genetic heterogeneity between ethnic groups and between families with and without asthmatic members were performed as described by Martinez and colleagues (17). In addition, for certain models mean FEV₁ values in liters (L) were calculated from these Z-scores at the mean height for boys and girls, respectively (mean height ± SD: 143.6 ± 6.5 cm, boys; 143.2 ± 6.8 cm, girls) at age 10 yr.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 shows characteristics of the 309 families included in the segregation analyses and the 842 families not included, according to demographic and diagnostic criteria. In addition, for NHW and Hispanic individuals with lung function tests, there were no significant differences in the proportion with current asthma, parents' smoking status, adjusted FEV₁, or age at time of testing between those individuals who were included in analyses (n = 1,163), and those with lung function tests who were not included (n = 661). Mean age (± SD) of parents at the time of lung function testing was 36.4 ± 5.3 yr for fathers and 33.7 ± 4.2 yr for mothers; children were tested at a mean age of 9.7 ± 2.0 yr. The prevalence of active physician-diagnosed asthma in the analysis group was 9.8%. Individuals with currently active physician-diagnosed asthma had a lower adjusted FEV₁ Z-score (0.63) than those without (0.06, p < 0.0001). Overall, 18.0% of parents were current smokers (20.8% of fathers and 15.2% of mothers); current smokers had a lower, but not significantly lower, adjusted FEV₁ Z-score compared with others (0.16 versus 0.002, p = 0.120).

Tests for genetic heterogeneity between families who were NHW, Hispanic, or NHW/Hispanic indicated no significant difference in the pattern of genetic inheritance (² for heterogeneity, 27.2; df = 20, p = 0.130) between the three groups. Groups were therefore combined for analytic purposes. The 309 families comprised 87 (28.2%) with at least one family member who had current physician-diagnosed asthma and 222 families with no asthmatic family members. There was significant genetic heterogeneity between families with and without asthmatic members (² for heterogeneity, 27.4; df = 13, p = 0.011). Therefore, separate analyses were performed for these two groups.

Table 2 shows the familial correlation coefficients for FEV₁ in the 87 families with at least one member with asthma, the 222 families with no asthmatic members, and the two groups combined. The spouse-spouse correlations are low and not statistically significant, whereas the higher and statistically significant parent-offspring and sibling-sibling correlations indicate intergenerational transmission. The sibling-sibling correlations are higher than parent-offspring correlations, suggesting either dominance variance or additional family associations, beyond common parentage. In the families with asthmatic members the mother-offspring correlation ( = 0.26) is statistically significant, and higher than the statistically nonsignificant father-offspring correlation ( = 0.06). These data are shown in Figures 1 and 2 with plots of the child's FEV₁ Z-score versus the mother's and father's FEV₁ Z-scores, respectively, for families with asthmatic members. The slopes and intercepts of the regression lines were calculated using the method of Press and colleagues (20), which minimizes the ² goodness of fit for data points having errors on both axes. Empirical heritability (15) ranged from 0.09 to 0.41 when considering father-offspring or mother-offspring, respectively, and was 0.28 when the mid-parent (average of both parents) value was used. There was no significant heterogeneity in familial correlations by ethnic group for families without asthmatic members (p = 0.439). There were insufficient Hispanic families (9) and NHW/Hispanic families (8) with asthmatic members to assess heterogeneity for families with asthmatic probands, although trends in familial correlations were the same by ethnic group.

View larger version (20K): [in this window] [in a new window]   Figure 1.   Plot and regression line of child's FEV1 Z-score versus mother's FEV1 Z-score in families with at least one asthmatic member. Pearson's correlation coefficient 0.26, p < 0.01. Equation of line: expected child FEV1 Z-score = 0.110 + (0.298 · mother FEV1 Z-score).

View larger version (20K): [in this window] [in a new window]   Figure 2.   Plot and regression line of child's FEV1 Z-score versus father's FEV1 Z-score in families with at least one asthmatic member. Pearson's correlation coefficient 0.06, p > 0.40. Equation of line: expected child FEV1 Z-score = 0.008 + (0.069 · father FEV1 Z-score).

Table 3 shows the results of fitting class D regressive models to the data of the families without asthmatic members. Parameter estimates with RFE (spouse-spouse, parent-offspring, and sibling-sibling) but without covariates are shown. This segregation analysis demonstrated that all models were significantly different from the unrestricted model and could therefore be rejected. However, comparison of the "no major type" model (line 2) with an equivalent model that excluded parent- offspring, spouse-spouse, and sibling-sibling effects (sporadic model) indicated a strong familial component (² = 45.9, df = 3, p < 0.0001). This familial component was statistically significant over and above all six models tested. Also, comparison of three distributional genetic and nongenetic models with the one distribution model indicated no significant differences (line 7 versus 2, ² = 5.4, df = 4, p = 0.249; line 6 versus 2, ² = 1.6, df = 3, p = 0.659; line 3 versus 2, ² = 5.8, df = 3, p = 0.122), suggesting a polygenic/multifactorial mode of inheritance. Further, AIC values indicated very little difference between the no-major-type, Mendelian arbitrary, and recessive models and the second environmental model, which allows for heterogeneity between generations. The addition of smoking covariates (current, former, ETS-current, and ETS-former) did not appreciably change these findings.

In families with at least one asthmatic member, there was a significant parent-offspring correlation (p = 0.034), a significant sibling correlation (p < 0.0004), and a nonsignificant spouse-spouse correlation compared with the sporadic model without RFE. Analyses without RFE indicated that the Mendelian arbitrary, recessive, and environmental ( = q_A) models fit the data as well as the unrestricted model (² = 1.1, df = 1, p = 0.294; ² = 1.4, df = 2, p = 0.497; ² = 5.7, df = 3, p = 0.127). Table 4 shows the parameter estimates from segregation analysis of FEV₁, with ascertainment correction in these families with RFE (spouse-spouse, parent-offspring, sibling-sibling), plus covariates. These analyses allowed the rejection of both environmental hypotheses (p = 0.005, 0.024). The Mendelian arbitrary, dominant, and recessive hypotheses could not be rejected (p = 0.202, 0.079, 0.158). Although the Mendelian arbitrary model was not significantly different from the other Mendelian models, AICs indicated that the Mendelian arbitrary and recessive models fit the data better than the Mendelian dominant model (AIC = 639.9, 639.9, and 641.5, respectively). However, the no-major-type, one-distribution hypothesis was also not rejected (p = 0.055), although it was borderline, indicating that a polygenic/multifactorial model fits the data as well as the Mendelian single-locus model plus additional genetic/multifactorial effects ("mixed" model). This is further supported by a comparison of the Mendelian model and the polygenic/multifactorial model (model 3 versus 2) (² = 7.6, df = 4, p = 0.110), which indicates that the Mendelian component is not significant within a polygenic/multifactorial context. Furthermore, there was no significant difference between the first environmental model ( = q_A) with RFE compared with that without (² = 2.9, df = 4, p = 0.575), suggesting that nontransmissible environmental factors responsible for the mixture of distributions could not be ruled out. In addition, RFEs were not significant over and above either the Mendelian arbitrary or recessive models (² = 5.9, df = 3, p = 0.117; ² = 4.2, df = 3, p = 0.241). Results were similar in the absence of smoking covariates. In the absence of all covariates, the environmental model ( = q_A) fit the data as well as the unrestricted model (² = 2.3; df, 0-3; p = 0.129).

Because the father-offspring correlation was not statistically significant in asthmatic families, an analysis was conducted in which both parent-offspring correlations were introduced separately. Mother-offspring effects were statistically significant over and above the sporadic model (² = 3.9, df = 1, p = 0.048), while father-offspring effects were not (² = 1.2, df = 1, p = 0.273). A further segregation analysis, therefore, fixed the residual father-offspring correlation to zero. Parameter estimates from this segregation analysis are shown in Table 5. A comparison of AICs between these models and those in Table 4 indicates very little difference in the absence of residual father-offspring effects. However, in Table 5, both the one-distribution and environmental hypotheses were rejected, and the Mendelian hypotheses were not. There was a borderline significant difference between the one-distribution (polygenic/multifactorial) hypothesis and the Mendelian arbitrary hypotheses (² = 7.7, df = 3, p = 0.053), providing weak evidence for a Mendelian component. The Mendelian recessive model plus RFE was significantly different from the one-distribution polygenic model (² = 6.6, df = 2, p = 0.037), suggesting the possibility of a genetic component expressed in a recessive fashion. The Mendelian unconstrained (arbitrary) model also reflected a recessive allele. The recessive allele identified in the Mendelian recessive model is allele A, with an estimated gene frequency (± SD) of 0.30 ± 0.10. It is associated with a homozygous phenotype of higher FEV₁ in about 9% of this population (i.e., _AA = 0.09 [Equation 1]). Figure 3 shows the predicted distribution of FEV₁ according to this model (Table 5, line 5). Also, the recessive locus plus RFE (mixed model) was not significantly different from a single recessive locus (² = 4.4, df = 3, p = 0.221) when model 5 was compared with an equivalent model that excluded RFE, implying that the single locus accounts for all the familial aggregation. However, RFE were again not significant in the first environmental model ( = q_A) (² = 2.8, df = 4, p = 0.591), failing to rule out nontransmissible environmental factors. Results were similar when sex was considered; that factor was significant as a covariate (p = 0.032).

View larger version (36K): [in this window] [in a new window]   Figure 3.   Histogram of the adjusted FEV1 Z-score for families with at least one asthmatic member. The two solid lines represent the predicted normal distribution of the trait for each genotype under the Mendelian recessive model (line 5 in Table 5). Means: µAA = 1.36; µAB = µBB = 0.23. The dotted line corresponds to the sum of the two normal distributions.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our results indicate significant heterogeneity in the control of lung function (FEV₁) between families with asthmatic members and those without. This suggests that there are different genetic and/or other etiologic determinants associated with the lung function phenotype in families with and without current asthmatic members. In the families without asthmatic members, all Mendelian models were rejected, as were the one-distribution and environmental models. This rejection indicates the absence of a major gene determining lung function in these families. Nevertheless, a significant familial component was demonstrated, indicating a strong polygenic/multifactorial mode of inheritance; both maternal- and paternal- offspring correlations were statistically significant. In the group of families with at least one asthmatic member, our results also suggested that polygenic/multifactorial components explained the familial aggregation, with some evidence for a Mendelian (recessive) element or common environmental factors. However, it was evident that the maternal, and not the paternal, effect was significant over and above the sporadic model, suggesting that a main component of intergenerational correlation in these families is associated with a maternal influence, either genetically or environmentally mediated. Still, the absence of a well-defined model in these families may indicate this study's inadequate power to reject certain hypotheses. The differences in familial correlations contribute to the heterogeneity observed between families with and without at least one asthmatic member.

Interestingly, others (21) have shown a greater correlation in FEV₁ and other lung function parameters between mothers and offspring compared with fathers and offspring, while some have not (1). However, it is difficult to compare our results with these studies as the other populations were not subdivided into families with and without asthmatic members; in the combined group we found that the maternal- and paternal-offspring correlations were similar. Although we have no definite explanation for the observed maternal influence on lung function among families with asthmatic members, there are a number of potential interpretations founded in the literature. First, several studies have shown a strong maternal influence on children's immune parameters, such as IgE and allergic sensitization, which has not been observed between fathers and offspring. For example, maternal, but not paternal, allergic history was associated with higher umbilical cord IgE levels in our longitudinal study of asthma (22). Also, both maternal asthma and allergic rhinitis were stronger predictors of asthma in the child than paternal asthma or allergic rhinitis (23); similar findings have been reported by others (24). Further, it has been shown that, for some allergens, sensitization in a child occurs significantly more frequently in children of mothers who are sensitized to the same allergen (25). In utero sensitization is implicated by proliferative responses of cord blood mononuclear cells to three cow's milk proteins (26). Intrauterine environmental factors and early allergen exposure have been suggested as important determinants of asthma and the allergic phenotype (27). Studies showing a strong positive relationship between head circumference and the development of atopy have led Holgate and colleagues (28) to the hypothesis that fetal nutrition via the placenta is a major determinant of asthma and allergy, possibly by redirecting the immune response to naturally occurring allergens. It has been suggested that while both the father and mother may contribute to the genetic transmission of susceptibility to atopic disease, additional environmental effects with a maternal influence may modify the expression of the genetic factors (29). In light of these ideas, the maternal influence on FEV₁, which we observe only among families with asthmatic members, could be connected with the maternal environment in utero. Evidence suggests that environment is associated with the development of atopy or asthma in the childconditions that can influence the phenotypic expression of the child's lung function. For example, asthma is known to affect the development of lung function and to be associated with decrements of FEV₁ (30). However, whether these observed differences are the cause or consequence of a diagnosis of asthma is not well understood.

Second, a number of other parameters show increased mother-offspring correlations compared with father-offspring effects. For example, correlation coefficients between body weight and body mass indices are higher in mother-offspring pairs than father-offspring pairs (31). In addition to in utero environmental effects, such studies suggest a greater maternal role in the determination of calorie intake in offspring (31), representing a confounding (environmental) effect in genetic analyses, although a genetic influence cannot be discounted. Thus, lung function could be another factor affected by postnatal maternal influences on growth and development.

Finally, genetic transmission could also be involved in the observed maternal influence on the child's lung function. In this context, reports by Cookson and colleaques (32) indicating linkage between a locus on chromosome 11q (the beta subunit of the high-affinity receptor for IgE [33]) and an atopic phenotype are noteworthy. Their findings indicated the transmission of atopy to be detectable only through the maternal line, and consistency with paternal genomic imprinting or maternal modification of immune responses was suggested. Other studies have subsequently confirmed this linkage and maternal inheritance in different populations (34), although others have not (35).

Segregation analysis has been applied by others to assess Mendelian modes of inheritance for lung function. Rybicki and colleagues (7) showed evidence for a major codominant gene for FEV₁ in chronic obstructive pulmonary disease (COPD) families, which accounted for all of the familial aggregation. In contrast, there was no significant familial correlation for FEV₁ in non-COPD families. As a consequence, significant heterogeneity was evident between the two groups of families. We also find significant genetic heterogeneity for FEV₁, apparent in our case, between families with and without asthmatic members. Another more recent segregation analysis of lung function in the Humboldt family study suggested polygenic control of FEV₁ or common environmental factors (8). Both the family ages and methodology of the latter study are similar to ours; indeed, its results are similar to our finding of polygenic/multifactorial control in families without asthmatic members.

The concept of genetic heterogeneity in the determination of a trait as complex as lung function is realistic. Lung function is probably affected by a number of genetic determinants, including the size of the airways and lungs, the intrinsic contractile and relaxing capacities of the airway smooth muscles, the lung elastic recoil and resistance properties, and the control of airway tone. With the additional influence of environmental factors on lung function development (36), it seems reasonable to assume that multiple genetic loci plus further environmental effects (both associated with each parent) may be involved in the normal phenotypic manifestation of lung function, as suggested by the polygenic/multifactorial models demonstrated in this study and others (8). In contrast, when considering alterations in lung function connected with disease (e.g., asthma, COPD) pathogenesis, it is more likely that anomalies would be observed and a single or few genes unrelated to those determining lung function in families without asthmatic members would be identified or differing parental contributions would be detected. In this context, the weak evidence we find for a recessive genetic component in families with at least one asthmatic member, as well as the detection of a codominant mode of inheritance in COPD families (7), are noteworthy. However, we were unable to rule out common environmental factors accounting for the familial aggregation. Still, the pathogenesis of asthma is also very complex. Its mode of inheritance has remained elusive and is likely not the result of a single, two-allele locus, but an oligogenic or polygenic influence with a strong environmental determination (19). Although we have shown genetic heterogeneity of lung function between families with and without asthmatic members, it is likely, given the lack of any well-defined model for FEV₁, that there is additional etiologic and genetic heterogeneity within these groups.