Main

Pulmonary surfactant is essential for normal lung function and consists of a complex mixture of lipids and SPs (SP-A, B, C, and D). SP-A, the major SP, is involved in varied aspects of surfactant biology(1, 2), and also appears to play a role in host defense and inflammatory processes of the lung(3). Two functional SP-A genes(SP-A1 and SP-A2) and one SP-A pseudogene have been characterized in humans(4–6). SP-A1 and SP-A2 are in linkage disequilibrium, and the SP-A locus has the characteristics (heterozygosity index and polymorphism information content values) of a good genetic marker(7). Both SP-A gene products appear to be necessary for a fully functional and stable mature SP-A(8). A number of alleles have been identified and characterized for each SP-A gene(7, 9–11). The naming of the alleles is based on the nomenclature of the cDNAs (6A and 1A) and thus the alleles for the SP-A1 gene are denoted 6An, and the alleles for the SP-A2 gene are denoted 1An. In addition, splice variability of the 5′-untranslated exons in the major transcripts of both genes has been observed(9, 12). Variability of SP-A mRNA levels among adults was previously observed(13), and a recent study suggested a correlation of this variability with SP-A1 and SP-A2 allelic variants(14).

Deficiency of surfactant can result in RDS in the prematurely born infant(15). Race, GA, and sex appear to be risk factors in the pathogenesis of RDS(16, 17). Black infants have a lower incidence of RDS when adjusted for their higher prematurity rate(18) and also appear to have a significantly lower incidence of RDS at every GA(19). It has been suggested that the racial differences in incidence of RDS are not due to quantitative differences in lecithin/sphingomyelin ratios(17, 20) or in TDx fetal lung maturity surfactant to albumin assay(21), but due to differences in the quality of surfactant(17, 21) or other race-related factors(20). In addition, several studies have implicated a genetic component in the development of RDS(22–24). A study of twins showed a high concordance of RDS among monozygotic twins, but not among dizygotic twins(23). Lankenau(24), has suggested a genetic contribution to the etiology of RDS(24). Thus the etiology of RDS is considered multifactorial/multigenic(10).

Although it has been known for decades that surfactant deficiency can result in RDS in the prematurely born infant, it is only recently that SPs have been implicated in the etiology of RDS or other pulmonary diseases(25–28). Immunohistochemistry of lungs of infants who died from RDS have shown that TM is absent in the RDS lungs(26). The amounts of SP-A and SP-B were reduced in the lungs of RDS patients, with SP-A either absent or at very low levels(26, 27). Decreased levels of SPs (35- and 6-kD proteins) in tracheal aspirates of neonates with RDS, early in the course of RDS, have been observed(29). A correlation between the level of SP-A and the severity of RDS has been noted(30, 31). A polymorphism in intron 4 of the human SP-B gene that associates with RDS has been characterized as well(25). Although the exact role of SPs in the development of RDS is unclear, findings of the above-cited studies suggest a role for the SPs in the etiology of RDS.

It has been shown in other complex diseases that mutations/polymorphisms at different genetic loci can result in/contribute to the same phenotype (genetic heterogeneity), or the same genotype can result in different phenotypes(32, 33). The latter case may arise from 1) incomplete penetrance and phenocopy, where the genotype at a particular locus may affect the probability of disease, but not fully determine the outcome or 2) polygenic inheritance, where the disease trait requires the simultaneous presence of mutations in multiple genes without the strict requirement of any single locus(34, 35). Further, environmental factors may be necessary for the expression of the disease phenotype, and interactions between genes may be synergistic (epistasis)(36). Interactions between three susceptibility loci linked to airway hyperresponsiveness in mice have been identified by using quantitative trait loci analysis(37). Recently, asthma was shown to be linked to different susceptibility loci in different ethnic groups(38).

In the present study, we investigated the association of SP-A alleles or genotypes and RDS and the possible modification of any such association by race, GA, or sex. We also investigated the association of the combined polymorphisms of SP-A (from this study) and SP-B(25) loci and RDS.

METHODS

Clinical sample collection and genotyping. Collection of blood specimens from newborns with and without RDS from various centers in the United States and Canada was carried out in accordance with the Declaration of Helsinki of the World Medical Association. The study protocol was approved by the ethical committee of the Pennsylvania State University College of Medicine as well as the committees of the respective centers. Information regarding race, GA, sex, and maternal and neonatal clinical histories was abstracted from hospital medical records. The diagnosis of RDS was made by clinical criteria (grunting, retraction, and flaring) and verified by radiographic analysis (reticulogranular pattern). Patients treated with surfactant prophylatically were excluded from the study. Caucasian and African Americans were included in the study. The GAs ranged from 24 wk to term. The distribution of the GAs in the various patient subpopulations is shown in Table 1. The mean ± SD of GA in the >28-wk (the largest sample group) white control and white RDS subgroups were 33.75± 2.22 and 32.83 ± 2.82, respectively. The male to female ratios in the >28-wk white control and white RDS subgroups were 1.33 and 1.86, respectively. The number of patients out of the total number in each of the subgroups of >28 wk who received maternal steroid administration were as follows: control white, 36/84; RDS white, 28/82; control black, 7/10; and RDS black, 5/20. Due to the small number of samples in the steroid-unresponsive RDS and steroid-treated asymptomatic infant groups, meaningful analysis of possible differences in frequencies of allele/genotypes in these groups could not be performed.

Table 1 Race, GA, and sex distribution of the study population
Full size table

DNA was extracted from cord or venous blood of newborns, and the preparation of DNA, PCRs, hybridization conditions for SP-A1 and SP-A2 alleles, and the scoring of the alleles were performed as described previously(7). The term allele refers to the presence of a given SP-A allele in an individual. For example, “1A0 allele” indicates the presence of at least one 1A0 allele in the subject. The term “1A0 genotypes” is an inclusive one referring to both 1A0 homozygotes and 1A0 heterozygotes (i.e. 1A0/1A0 and 1A0/*, where * is any other SP-A2 allele). The term genotype refers to both alleles at the given SP-A locus in an individual.

Database. A database consisting of the SP-A1 and SP-A2 genotypic and clinical data of all the study samples was generated using the Microsoft Access Database software. Using the database, the samples were categorized into various combinations of control/RDS with race, GA, and sex. These subgroups were then analyzed for patterns of disease association.

Approach and statistical analyses. First, we compared the overall distribution of genotypes for each of the two SP-A genes and the frequencies of individual genotypes and alleles between control and RDS populations among different races (black and white), age groups (≤28 and>28 wk of gestation), and sex. Fisher's exact test was performed for these comparisons using the statistical package StatXact3, Cytel Software Corp., Cambridge, MA(39). Second, taking into consideration the findings from the first set of analyses, we focused on the frequencies of all alleles/genotypes in control and RDS populations of >28 wk of GA (this subgroup had the largest sample size). A given allele/genotype was considered to have an association with the disease trait if the allele/genotype occurred at a higher frequency among the RDS compared with control individuals. Fisher's exact test (StatXact) was used for the comparisons of the frequencies of alleles/genotypes between control and RDS. To adjust for a risk factor such as age or sex, we used stratified exact tests in StatXact(39). In the case of the 1A0 genotype, we tested for trend using the Cochran-Armitage trend test in StatXact (see“Results”)(39). This part of the study was performed to look for associations that would be of potential value in subsequent investigations, such as the one shown below for the combined SP-A and SP-B polymorphisms. Furthermore, adjusting p values for the number of statistical tests that were performed precludes providing statistical significance. Third, we investigated the frequency of the combined SP-A and SP-B polymorphisms in RDS and control groups for white race >28-wk GA. Namely, we examined the frequency of the SP-A allele 1A0 (which, in the present study, was found to be in higher frequency in the >28-wk white RDS population) in combination with the SP-B variant described previously(25). The analysis of the SP-A/SP-B combination included only those samples that were analyzed for the SP-B locus (published data) as well as for the SP-A locus. Fisher's exact test was used for testing the association; a p value of <0.05 was considered significant. Thus the data from the published and the present study were analyzed for RDS association using two marker loci.

Because race, GA, and sex are risk factors for RDS, it is possible that these factors modify the association between a given allele/genotype and RDS. To test this possibility we used the exact HOR test(39). HOR tests whether the odds ratio between the risk of disease and a given allele/genotype is the same across all strata of a risk factor such as age or sex. When the HOR gives a statistically significant (p < 0.05) result, we conclude that this odds ratio differs across the strata of the risk factor. For example, if an age modification effect exists, the odds ratio of the particular allele/genotype in RDS versus control would differ in individuals ≤28 wk of gestation compared with those of >28 wk. A corollary to the observation of an effect modification by a risk factor is that the subgroups of that risk factor should be analyzed separately for tests of association of allele/genotype and RDS.

If the HOR is not statistically significant, we cannot conclude that effect modification is present. In this case, we assume the odds ratio between disease and allele/genotype is constant across the strata of the risk factor,i.e. the risk of disease given a particular allele/genotype is the same in both age groups. We denote this constant odds ratio as the common odds ratio. In the case of no effect modification, a risk factor may confound the common odds ratio. That is, controlling for the risk factor may decrease or increase the common odds ratio. This is analogous to multiple regression, where regression slopes may change depending on whether other variables are included in the regression model. In the case of the common odds ratio, we estimated the common odds ratio and confidence interval and tested its significance with the Fisher exact test not adjusting for the risk factor, and then compared these results with the common odds ratio estimate and corresponding exact confidence interval and tested adjusting (i.e. controlling) for the risk factor(39). If the p values for the two tests differed (e.g. one was significant and one was not) we then concluded that the risk factor confounded the odds ratio between disease and the allele/genotype. If the p values were similar, we concluded that confounding did not exist. If a risk factor is confounding, i.e. separate correlations exist (p < 0.05) between a particular allele/genotype and the confounding factor, as well as between the confounding factor and RDS, a spurious association between the allele and RDS may be observed if no adjustment is made for the confounder(third variable).

RESULTS

Distribution of SP-A Alleles and Genotypes

Race. The overall distribution of the genotypes for each SP-A gene was not significantly different between the control and RDS populations within each race (SP-A1: for white p = 0.27 and for black p = 0.23; SP-A2: for white p = 0.58 and for black p = 0.82). The frequencies of the alleles and individual genotypes of SP-A1 and SP-A2 within each race (data not shown) were also not different between control and RDS. Tables 2 and 3 show the odds ratios and their corresponding confidence intervals for the alleles and genotypes in controls and RDS. The odds ratios of several alleles and genotypes, e.g. 6A2, 1A0, 6A3/6A3, 1A0/1A0, and others, were in the opposite directions for black(control and RDS groups) and white (control and RDS groups)(Tables 2 and 3). For example, the odds ratios for 6A2 were 0.57 (<1) in black and 1.29 (>1) in white subjects. This observation points to possible differences between the two race groups. The frequencies of alleles/genotypes with odds ratios in the opposite direction are shown in Figure 1 to illustrate the differences between the races. It is of note that these differences in odds ratios between the races are not general phenomena but specific to certain alleles and genotypes. However, the HOR that tests whether the odds ratios for a particular allele/genotype are the same in two sets of groups reached statistical significance only in the case of 6A3/6A3(Table 3). The 95% confidence interval for 6A3/6A3 was fairly large and this was probably due to sample size (n = 11 control) in black and/or due to the low frequency of the genotype in white subjects. The low sample size in the control black group may contribute to the lack of statistical significance with the other alleles/genotypes despite the differences in odds ratios. Effect modification by race cannot be excluded. Therefore, black and white subjects were analyzed separately.

Table 2 Effect of race on SP-A1 and SP-A2 allelic frequencies
Full size table
Table 3 Effect of race on SP-A1 and SP-A2 genotypic frequencies
Full size table
Figure 1
figure 1

Frequencies of SP-A1 and SP-A2 alleles and genotypes in control/RDS of black and white subjects. The relative frequencies of the SP-A1 and SP-2 alleles and genotypes (with odds ratios in the opposite directions in control/RDS of black and white populations) are shown. The sample sizes(n) of each subgroup are shown above the corresponding bar. The asterisk (*) indicates that the comparison between 6A3/6A3 black and 6A3/6A3 white subjects was statistically significant(p = 0.046).

Full size image

The distributions of SP-A1 genotypes in RDS white and RDS black were significantly different (n = 106 RDS white and n = 37 RDS black, p < 0.001), but the distributions in control white and control black were not different (n = 86 control white and n = 12 control black, p = 0.88). Similarly, the overall distribution of SP-A2 genotypes was significantly different between RDS white and RDS black (n = 106 RDS white and n = 35 RDS black,p < 0.0001), but not between control white and control black(n = 85 control white and n = 11 control black,p = 0.36). The frequencies of certain SP-A1 alleles were also significantly different between RDS white and RDS black populations compared with their corresponding control subjects (6A2, p = 0.0002; 6A3, p < 0.0001; 6A4, p = 0.019). Comparable results were obtained for SP-A2 alleles (1A0, p< 0.0001; 1A1, p = 0.0007; 1A2, p = 0.0087). 1A0, 6A2, and 6A4 alleles were overrepresented in the RDS white population, whereas 1A1, 1A2, and 6A3 were significantly more frequent in the RDS black population. It is possible that this differential distribution of SP-A alleles/genotypes in the two races may contribute to the above noted differences in the odds ratios.

Age. We examined the effect of GA (≤28 and >28 wk) on the genotypic and allelic associations with RDS separately for each race (data not shown). For white subjects, the test of HOR showed that age did not modify the effect of any of the alleles/genotypes on RDS (for ≤28 wk: n = 2 control and n = 24 RDS and for >28 wk: n = 84 control and 82 RDS). In view of the sample size (n = 2) for the ≤28-wk control sample, the absence of any age-related effect modification cannot be completely ruled out. However, 6A2/6A2 (p = 0.042) and 6A2 (p = 0.032) were overrepresented in the RDS >28 wk compared with the ≤28 wk of gestation RDS group, which hints at an effect modification. Also, the genotype 1A1/1A2 occurred at a significantly higher frequency in the ≤28 wk of age group compared with>28 wk in the RDS white population (p = 0.036), but no difference was seen in the comparable control groups (p = 0.99). Age was not a statistically confounding factor for alleles/genotypes and RDS. But, because effect modification of association of allele/genotype and RDS by GA could not be ruled out, the two age groups should be analyzed separately for disease association.

Although effect modification of the alleles/genotypes-RDS association due to GA was not observed in the sample study, the low sample numbers in black subjects [for ≤28 wk, n = 2 (SP-A1) and n = 1 (SP-A2) control, n = 17 (SP-A1) and n = 16 (SP-A2) RDS; for >28 wk, n = 10 control, n = 20 (SP-A1) RDS, and n = 19 (SP-A2) RDS] preclude definite conclusions.

Sex. The test of HOR revealed no effect modification by sex for SP-A1 and SP-A2 alleles/genotypes in the >28-wk white, ≤28-wk white,≤28-wk black, and >28-wk black subgroups (data not shown). As expected, sex was not an independent statistically confounding factor for the alleles/genotypes. Although the male to female ratios in the >28-wk control white and RDS white subjects were different [1.33 versus 1.86(Table 1)], the absence of effect modification and confounding indicates that this difference is unlikely to have caused any spurious results.

Association Studies for SP-A Alleles/Genotypes

Taking into account the above observations, we investigated differences in frequencies of SP-A alleles/genotypes in control versus RDS for age and race separately.

The >28-wk white group. The 1A0 allele was found in significantly higher frequency (p = 0.035) in RDS patients compared with control subjects (Table 4). On adjusting for age the p value was 0.039, and on adjusting for sex it was 0.057. Although consideration of sex or age as an independent confounder of the allele/genotype is not meaningful in biologic terms, the sex- and age-adjusted calculations were made to rule out sample bias. The age- and sex-adjusted p values were obtained by using the exact common odds ratio test that adjusts for the third variable (age or sex) and tests whether the common odds ratio (common to the subgroups of the third variable) is unity(39). Although the frequency of the homozygous 1A0/1A0 compared with all other genotypes (including the 1A0 heterozygotes) was not different in the control and RDS groups, the frequency of the homozygous 1A0/1A0 genotype relative to the non-1A0-containing genotypes was greater in RDS than in the control population. Although the frequency of the heterozygous 1A0/* relative to the non-1A0-containing genotypes was not different (p = 0.058), the odds ratio was 2.44, suggesting that even in the heterozygous state 1A0 may have some effect. The importance of confidence intervals of odds ratios compared with p values has been emphasized by other workers(40). The Cochran-Armitage trend test was used to test whether the odds ratio increases from non-1A0 genotypes to 1A0 heterozygotes and from 1A0 heterozygotes to 1A0 homozygotes. The findings (p = 0.039) suggested a trend for stronger association of RDS with the homozygous state. There was no difference in frequencies of homozygous 1A0 compared with heterozygous 1A0 in the control/RDS groups (p = 0.73). The presence of the 1A0 allele in the homozygous state appears to confer a disease susceptibility in this subpopulation of RDS.

Table 4 Frequencies of 1A0 allele and genotypes
Full size table

The 1A1/1A1 genotype was fairly uncommon in control subjects and was absent in the RDS population. The frequencies of 1A1/1A1 genotype in control and RDS, compared with either all other genotypes or to non-1A1 genotypes, did not differ (data not shown). However, on adjusting for age, a negative association was observed between 1A1/1A1 genotype but not between 1A1 allele and RDS(1A1/1A1 versus other genotypes: p = 0.005; 1A1 allele versus other alleles: p = 0.13). The frequencies of 1A1 heterozygotes and 1A1 allele were not different between the control and RDS (data not shown). Thus, the association of 1A1/1A1 was observed only on adjusting for GA. Whether this observation reflects an effect due to underrepresentation of the genotype in the >28-wk white RDS subgroup or is a result of sample size is not clear. The 1A1 homozygous state appears to confer a protective effect, with a negative association between RDS and the genotype. None of the other alleles or genotypes showed an association with RDS in this age-race subgroup. Thus, out of the 9 known alleles and 10 possible SP-A1 genotypes and 15 possible SP-A2 genotypes, only a few of the alleles/genotypes appeared with different frequencies between control and RDS in the >28-wk white subgroup.

The >28-wk black group. No significant differences were seen in the overall genotypic distributions of control and RDS subgroups (SP-A1,n = 10 control and n = 20 RDS, p = 0.13; and SP-A2, n = 10 control and 19 RDS, p = 0.95). Unlike the>28-wk white population, the >28-wk black subgroup of this sample size did not show any associations between individual genotypes or alleles and RDS(the p values were 6A, 0.99; 6A2, 0.59; 6A3, 0.28; 6A4, 0.11; 1A, 0.99; 1A0, 0.99; 1A1, 0.99; 1A2, 0.75; and 1A3, 0.99).

RDS Association Using SP-A2 and SP-B Allelic Markers

When the 1A0 allele was used in combination with the previously characterized intron 4 SP-B insertion/deletion variant that associated with RDS(25), we observed a significant positive association between 1A0 + SP-B polymorphic variant and RDS in the >28-wk white population in spite of the small sample size (p = 0.036)(Fig. 2). (Previously published data were incorporated in Fig. 1 to emphasize this observation.) The odds ratios and the 95% confidence interval were 2.051 (1.010-4.15) for SP-B alone, 1.61(1.013-2.55) for 1A0 alone, and 4.55 (1.016-27.69) for SP-B + 1A0. Synergism is inferred if the observed odds ratio for SP-B + 1A0 minus one (3.55) is greater than the calculated sum (1.66) of odds ratios of 1A0 minus one (0.61) and SP-B minus one (1.051)(41). Thus, the association with RDS was stronger for SP-B + SP-A (p = 0.036, total n = 60) than for either SP-B(p = 0.041, total n = 219) or SP-A alone (0.035, total n = 164), indicating synergism between the two. The comparison of the combined polymorphisms was made between individuals with 1A0 + SP-B invariant versus those with 1A0 + SP-B variant alleles. The term variant denotes any SP-B allele that differs in size from the invariant allele at the intron 4 locus(25, 42). The 1A0 allele occurred at a frequency of 63% in RDS group, 35% of which also had the SP-B variant. Of note is that the SP-A2 genotype of all the samples that were also analyzed for SP-B contained at least one 1A0 allele. Although it may be argued that the observed 1A0 association is actually due to the association of RDS and the SP-B locus, our data indicate an association of 1A0 and RDS independent of the SP-B locus.

Figure 2
figure 2

Frequency of SP-B allelic variant, 1A0 allele, and SP-B + SP-A (1A0 allele) in control and RDS. The relative frequencies of SP-B variant in control and RDS of all races (previously published data, p = 0.041), 1A0 allele (p = 0.035) and the combination 1A0 allele + SP-B variant (p = 0.036) in white control and white RDS subjects are shown above. The sample sizes(n) of each subgroup are shown above the corresponding bar. The p values, odds ratio (OR), and corresponding 95% confidence intervals (CI) are shown below the histograms.

Full size image

DISCUSSION

The etiology of RDS is considered multifactorial/multigenic(10, 43), and surfactant deficiency can lead to RDS. Because the SPs play important roles in surfactant biology/physiology(1, 44), we hypothesized that the SPs under at least certain circumstances constitute some of the contributors to the etiopathogenesis of RDS. The purpose of the present study was 2-fold. First, we wished to identify associations between SP-A alleles/genotypes and RDS that would be of potential value in subsequent investigations. For this work the use of multiple tests precludes proving statistical significance. Second, we wished to use the SP-A alleles found in association with RDS and a previously described(25) SP-B polymorphism to study the frequency of the combined SP-A and SP-B polymorphisms in control and RDS groups. We identified associations between SP-A2 alleles/genotypes and RDS. The frequency of the 1A0 was higher in the RDS group in white subjects >28 wk compared with the control group (p = 0.035). The frequency of the combined 1A0 and the SP-B intron 4 polymorphisms(25) was significantly higher in the RDS group(p = 0.036). The use of the combined SP-A and SP-B marker loci in association studies increased the association between the marker alleles and RDS. These observations suggest that polymorphisms within these SP genes could serve as markers to distinguish various subgroups of RDS. Alternatively, certain SP-A and/or SP-B alleles/genotypes may confer or are linked to loci that confer disease susceptibility such that these loci in the presence of other genetic or environmental factors could lead to the disease state. We have observed possible modifications of the SP-A allelic association with RDS by race and GA, indicating that caution should be exercised regarding the composition of control and RDS groups to avoid erroneous conclusions.

The association of 1A0 allele in the >28-wk RDS white group is rather intriguing in view of a recent report that the 1A0 allele associates with low levels of SP-A mRNA in human lung tissue of unrelated individuals(14). Taken together, these data suggest that the interaction of 1A0 allele with other genetic and/or environmental determinants could lead to RDS. Previously, we have shown that SP-B polymorphisms occur at a higher frequency in the RDS group(25). The significantly higher frequency of the combination of 1A0 and SP-B intron 4 variant in the RDS group, compared with either locus alone, suggests a synergistic interaction between SP-A and SP-B loci in the development of RDS phenotype in >28-wk white population. The frequency of the 1A0 allele as well as the distribution of SP-A alleles/genotypes differed between the RDS white and the RDS black groups. The distribution of SP-B alleles has also been shown to differ between black and white populations(42). Taking these data together, we hypothesize that the variability in the frequency of the SP-A and SP-B alleles accounts, in part, for differences in the incidence of RDS between blacks and whites.

The association of multiple genetic alterations of the SP-B locus with two diseases, congenital alveolar proteinosis(45, 46) and RDS(25), and the correlation of these alterations with the phenotypic variability in congenital alveolar proteinosis(46) support the notion that different alleles of the SP-B locus contribute to the diversity of disease phenotypes. In the case of the RET protooncogene, different mutations were shown to associate with Hirschsprung's disease, multiple endocrine neoplasia, sporadic medullary thyroid carcinoma, and familial medullary thyroid carcinoma in humans, and defects in the kidney and enteric nervous system in mice(47). Phenotypic diversity can also arise from additional mutations(48), polymorphisms in the same disease gene(49) or in unlinked genes(50). The phenotype of a particular cystic fibrosis mutation was also shown to be modified by the presence of an inefficient CFTR splice variant(48). Although the functional significance of SP-A splice variants remains to be determined, it is conceivable that low levels of SP-A mRNA and the predominance of a specific splice variant in 1A0 allele(14) may, in combination with other factors, lead to the development of RDS. It has been reported that the simultaneous presence of mutations in two unlinked yet functionally interacting photoreceptor-specific genes resulted in retinitis pigmentosa(50). Because SP-A and SP-B proteins interact functionally, e.g. both are essential for the in vitro formation of TM and contribute to the adsorption of phospholipids to the surfactant layer at the alveolar air-liquid interface(1, 2, 51), the synergistic association of SP-A and SP-B variants with RDS is an interesting finding.

Complex diseases such as diabetes, hypertension, heart disease, and schizophrenia do not exhibit classic Mendelian recessive or dominant inheritance attributable to a single gene locus. Although there may be a genetic component to RDS, it seems unlikely that a single major gene will be identified as a cause of RDS, given the multiple factors involved in its pathogenesis. It is possible that RDS, like other complex diseases, results from the accumulation and interaction of multiple genetic and nongenetic factors. The identification of genetic components of complex diseases has been facilitated by disease subgrouping based on clinical phenotype, e.g. colon cancer; age of onset, e.g. Alzheimer's disease; family history, e.g. breast cancer; and severity, e.g. hypertension(34). A correspondence between clinical heterogeneity (in terms of severity of RDS) and SP-A levels has been reported(30, 31). Other indirect evidence for the role of SP-A as a disease modifier comes from the SP-A-deficient mouse model(52). The postnatal survival and pulmonary function in the SP-A knockout mouse were unaltered. However, surfactant from the homozygous (-/-) SP-A mice exhibited decreased surface tension lowering properties at low phospholipid concentrations, especially in the absence of Ca2+. The formation of TM was also impaired in these mice. In addition,in vitro studies have also shown that SP-A increases adsorption of surfactant lipids to the air-liquid interface and enables surfactant function at very low lipid concentrations(51). These observations together suggest that SP-A plays a vital role in maintaining surfactant function and alveolar surface area in compromised conditions such as circumstances of low surfactant concentrations that may develop in the presence of prematurity and/or other genetic or environmental determinants. Interestingly, in an another complex disease, the nonsyndromic cleft lip and palate, an association was found with TGFA but linkage with traditional linkage studies was lacking. However, the TDT was positive and TGFA alleles appeared to modify the severity of this disease in an Asian population(53). The TGFA knockout mouse showed no evidence of craniofacial abnormality supporting a modifier role for TGFA(54). Thus, for complex disease traits that involve interactions of multiple genes, the combined use of genetic epidemiologic studies with studies of functional and regulatory aspects of candidate genes can help unravel the disease pathways.

Because race, GA, and sex are risk factors of RDS we investigated the potential modification effect of each of these factors on the association of SP-A allele/genotype with RDS. The statistically significant positive association (increased likelihood of the genotype in RDS) with RDS black and a negative association with RDS white for 6A3/6A3 populations, and the finding that the odds ratios of black and white subjects are in opposite directions suggest effect modification. Consistent with the race modification effect is the observation that the distributions of SP-A1 and SP-A2 genotypes and the frequencies of most SP-A1 and SP-A2 alleles differed significantly between RDS black and RDS white subjects. Therefore, to avoid erroneous conclusions, caution should be exercised regarding the racial composition when associations of SP-A alleles/genotypes and disease are considered in groups of this sample size. In fact, association studies in racially/ethnically nonhomogeneous groups have led to conflicting results for other diseases(34, 55).

GA did not modify the association between alleles/genotypes and RDS. In view of the observation that the 6A2 allele and 6A2/6A2 genotype are overrepresented in the RDS white >28-wk group and that other developmental/prematurity factors could be involved in the etiology of RDS in the ≤28-wk group, an effect modification is likely to exist and thus age should be taken into consideration in the various analyses. Sex did not show any modifying or confounding effect on the genotypic or allelic distributions for the SP-A genes, suggesting that the SP-A locus does not contribute to the slightly higher relative risk for RDS in males. The reader should, however, keep in mind the potential consequences (increasing the probability of type I error) of performing many hypothesis tests of associations and effect modification.

In summary, we show that there are associations of the 1A0 allele as well as the 1A0/1A0 genotype with RDS and that the combined polymorphisms of 1A0 and SP-B intron 4 increases the specificity and the power of association. Population association studies of unrelated individuals are highly sensitive in identifying disease risk factors. After a significant association is found, the next appropriate step is to confirm the association by TDT(34, 35, 56, 57). In TDT, the nontransmitted parental allele is used as an internal control; this type of analysis overcomes problems arising from differences in genetic backgrounds of unrelated controls(56). Our data suggest that SP-A and SP-B are good candidate loci and that these loci either contribute to the etiology of RDS or can be used as genetic markers to identify disease subgroups, such as severity, clinical outcomes, and steroid responsiveness versus nonresponsiveness. Moreover, the data discussed above suggest that additional experimentation is warranted to both confirm and to help unravel the genetic complexities of RDS. Furthermore, elucidation of the complexity of the structural, functional, and regulatory aspects of the SP genes may provide a better understanding of their role in the pathogenesis of RDS.