Skip Navigation


Human Molecular Genetics Advance Access originally published online on April 21, 2004
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
13/11/1095    most recent
ddh132v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (24)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Rova, M.
Right arrow Articles by Hallman, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rova, M.
Right arrow Articles by Hallman, M.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Human Molecular Genetics, 2004, Vol. 13, No. 11 1095-1104
DOI: 10.1093/hmg/ddh132
Human Molecular Genetics, Vol. 13, No. 11 © Oxford University Press 2004; all rights reserved

Data mining and multiparameter analysis of lung surfactant protein genes in bronchopulmonary dysplasia

Meri Rova1, Ritva Haataja1,*, Riitta Marttila2, Vesa Ollikainen3, Outi Tammela4 and Mikko Hallman1

1Department of Paediatrics and Biocenter Oulu, University of Oulu, PO Box 5000, FIN-90014 Oulu, Finland, 2Seinäjoki Central Hospital, FIN-60220 Seinäjoki, Finland, 3CSC, The Finnish IT Centre for Science, PO Box 405, FIN-02101 Espoo, Finland and 4Tampere University Hospital, PO Box 2000, FIN-33521 Tampere, Finland

Received January 9, 2004; Revised March 12, 2004; Accepted April 1, 2004


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 RESULTS
 DISCUSSION
 MATERIALS AND METHODS
 REFERENCES
 
Bronchopulmonary dysplasia (BPD), the most common chronic lung disease in infancy, is influenced by a number of antenatal and postnatal risk factors and is mostly preceded by respiratory distress syndrome (RDS) in the newborn. Surfactant protein (SP-A, -B, -C and -D) gene variations may play a role in both BPD and RDS. An association study between these candidate genes and BPD was performed. A total of 365 preterm Finnish infants in a high-risk population with gestational age <=32 weeks were genotyped for all SP genes. A multiparameter analysis was performed using Agrawal's algorithm based data mining and conventional methods of statistical allelic association. In singletons and presenting multiples, the frequency of SP-B intron 4 deletion variant allele was increased in BPD versus controls (P=0.008, OR=2.0, 95%CI 1.2–3.4). The presence of the SP-B intron 4 deletion variant was a risk factor for BPD even when essential external confounding factors were included in the analyses. No other SP polymorphisms associated with BPD, and the SP-B intron 4 variation did not associate with RDS. Transcription Element Search Software predicted allele-specific differences at several putative transcription factor binding sites that may be important in SP-B regulation. The present multiparameter analysis demonstrates the presumable direct involvement of the SP-B intron 4 deletion variant allele as a genetic risk factor to BPD. We propose that two separate SP-B gene polymorphisms have a phenotypic significance via separate molecular mechanisms: the intron 4 length variation affecting transcriptional regulation, and the exonic Ile131Thr variation affecting post-translationally.


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
RESULTS
 DISCUSSION
 MATERIALS AND METHODS
 REFERENCES
 
Bronchopulmonary dysplasia (BPD), which affects infants born very prematurely, is the most common severe, chronic lung disease diagnosed in infancy. The pathologic findings in BPD include an arrest in lung growth and alveolar development, atelectasis, fibrosis, dysplasia in the airway lining and a severe early inflammatory response in the peripheral airways and air spaces. The clinical diagnosis of BPD is based on respiratory failure requiring oxygen therapy at least until 36 post-conceptional weeks after very premature birth. Among extremely low birth weight (ELBW, <1000 g) infants, the incidence of BPD is about 30% (1). The disease is uncommon in infants of more than 1500 g birth weight or with gestational age exceeding 32 weeks.

BPD is considered to be a multifactorial disease with possible genetic involvement (2,3). A number of genes together with environmental factors, such as those related to premature birth, very low birth weight and birth asphyxia, may influence the susceptibility to this disease (3–5). In most cases, BPD is preceded by respiratory distress syndrome (RDS), which is characterized by acute respiratory failure due to deficiency of a pulmonary surfactant at birth. The important role of a complex inflammatory reaction in the pathogenesis of BPD is also evident (5,6). Intrauterine inflammatory syndrome—also called chorioamnionitis—is associated with premature rupture of fetal membranes (PROM), spontaneous premature birth and early maturation of the pulmonary surfactant. Accordingly, it is associated with a decreased incidence of RDS, but on the other hand, with an increased risk of BPD. Chorioamnionitis has been shown to predict BPD, especially in those infants who are exposed to events that potentiate pulmonary inflammation after birth (i.e. requirement of mechanical ventilation or post-natal sepsis) (7). In multifactorial diseases the interaction between genetic and environmental factors is considered to be important. In this case, however, the molecular genetic background of BPD is very poorly understood, although some evidence of genetic association has been published (8).

Lung surfactant is a lipid–protein complex that lowers the surface tension in the lining of the terminal pulmonary air spaces, preventing atelectasis (collapse) and its devastating consequences leading to respiratory failure. Lung surfactant has additional roles in host defence. Surfactant components are involved in the innate immunity and they may serve as anti-inflammatory factors. In particular, lung surfactant protein genes (SP-A1, -A2, -B -C and -D) may affect individual susceptibility to BPD. They are all well characterized, polymorphic, and have been studied extensively as candidates for several multifactorial pulmonary diseases both in infants and in adults (9–12). SP-A and SP-D are hydrophilic C-type collectins that are involved in surfactant metabolism and particularly in innate immunity of the lung (13). SP-B and SP-C are hydrophobic proteins involved in reducing surface tension at the air–liquid interface of alveoli (14,15). SP-B has additional roles in maintaining the integrity of alveolar epithelial cells. According to recent data, it also has an important role in inhibiting endotoxin-induced lung inflammation (16).

Identification of disease-related genes requires a laborious examination of several candidate genes, which interact with each other and with non-genetic factors. Data mining method provides a powerful way to search a large hypothesis space for dependencies in multidimensional data. We believe that it can be a particularly useful tool for semi-automatic, explorative analysis in case–control studies involving both genetic and environmental effects, even in cases where non-genetic risk factors account for a significant, but limited, portion of disease susceptibility.

In the present study, data mining and conventional statistical allelic association study methods were used. All surfactant protein genes (SP-A1, -A2, -B, -C and -D) and several confounding environmental variables were examined as potential susceptibility factors in the aetiology of BPD in a homogenous Caucasian high-risk population of 365 preterm Finnish infants with gestational age of 32 weeks or less.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 RESULTS
DISCUSSION
 MATERIALS AND METHODS
 REFERENCES
 
Allele frequency comparisons
The whole data set included 86 infants with BPD and 279 control infants. The control infants were further divided into two groups on the basis of their RDS status. Of the control infants, 188 were RDS-positive and 91 had neither BPD nor RDS. The allele distributions of all surfactant protein genes are presented in Table 1. The observed genotype frequencies did not deviate from Hardy–Weinberg equilibrium.


View this table:
[in this window]
[in a new window]
  		Table 1. SP-allele frequencies (%) among infants with BPD and their premature controls
 
The frequency of the SP-B intron 4 deletion variant allele was increased (P=0.042) in the infants with BPD when compared with the controls with RDS (Table 1). We also separately tested whether the SP-B intron 4 polymorphism was associated with RDS alone instead of BPD, as judged by the allele or genotype frequency, but no association with RDS was found.

Our previous study of differences between singletons and twins revealed a significant confounding effect of birth order on allelic associations in RDS (17). Therefore, we separately analyzed the data including only the singletons and presenting multiples (n=249 individuals). They are exposed to the uterine cervix and also presumably to ascending infections and inflammatory reaction, whereas the non-presenting multiples are more protected from inflammatory events. In this setting, both the SP-B intron 4 allele (P=0.008) and the genotype (P=0.007) distributions as a whole differed significantly between the BPD (n=67) and control (n=178) infants (four samples were excluded due to failure in SP-B intron 4 genotyping). The association of the SP-B intron 4 deletion variant allele with BPD was found to be significant (Table 2). Furthermore, if the infants requiring supplemental oxygen for 28 days after birth were included together with the cases with BPD, defining a broader phenotype (n=107), the deletion variant allele was overrepresented (P=0.0036) compared with the controls (n=138). Neither the BPD nor the SP-B intron 4 alleles were associated with birth order. The other surfactant gene polymorphisms did not associate with BPD in any of these settings.


View this table:
[in this window]
[in a new window]
  		Table 2. SP-B intron 4 allele and genotype frequencies (%) among singletons and presenting multiples with BPD and their premature controls
 
Association of other alleles of the hydrophobic surfactant protein genes with respiratory failure
When the BPD infants were compared with the healthy control infants without RDS (Table 1), differences in allele frequencies were detected for the SP-B Ile131Thr polymorphism (P=0.043), SP-C Asn138Thr polymorphism (P=0.040) and SP-C Asn186Ser polymorphism (P=0.012). However, these allele frequencies did not differ significantly between the BPD infants and the controls with RDS. This indicates that if not due to type I error, these differences in allele frequencies are probably associated with RDS or with other factors related to RDS, rather than with BPD. In the group of RDS infants, an interactive association of SP-A alleles and the SP-B genotype was evident as the SP-A1 allele 6A2 was associated with RDS among the carriers of the SP-B Thr/Thr genotype (10). However, there were no detectable gene–gene interactions associated with BPD among this population.

Data mining
In order to study the complex interactions between genetic and environmental factors, the data mining method was utilized. Again, the whole data set was first analyzed, taking into account all individuals (n=365). The results indicated that low gestational age (below 197 days), low birth weight (below 1050 g) and RDS in different combinations were associated with the broadly defined phenotype (BPD or the requirement for supplemental oxygen for at least 28 days). None of the genetic factors had a clear effect on the phenotype per se; at best they slightly improved the associations discovered without them. This minor improvement is likely to be due to chance.

After the analysis of the whole data set, a separate analysis including all singletons and presenting multiples was performed. As data mining is an explorative method of analysis, which allows large numbers of hypotheses to be tested and a large number of results to be produced (including many false positives), we provide three interpretations of the results. First, Table 3 lists all rules where up to three explanatory variables are very strongly associated with the phenotype, irrespective of the genetic component (n=3, threshold for P=10–6). The results point out the strong association of low gestational age and low birth weight with the phenotype, but, in the light of the observed frequencies, the genetic components in these rules may well be artefacts.


View this table:
[in this window]
[in a new window]
  		Table 3. Data mining results: the most significant association rules for various rule lengths (n) and significance thresholds (P for the entire rule and Pgen for the combination of genetic variables within the rule)a
 
In Table 3, we focus on rules with no more than two explanatory variables, which require a strong association of the entire antecedent part and a moderate association of the genetic variables alone (n=2, threshold for P=0.001, Pgen=0.01). It is of particular interest that low birth weight (<1050 g) together with presence of the SP-B intron 4 deletion variant allele predicted the broadly defined disease phenotype in 22 of the 24 cases, pointing out an important subgroup in the data. Other known risk factors that with SP-B intron 4 deletion variant predicted BPD, included RDS and birth asphyxia. In other words, even though low birth weight, birth asphyxia or RDS are individual major predictors of BPD, their impact seems to be especially drastic in individuals who additionally carry the SP-B intron 4 deletion variant allele. It should be noted that, in such subgroups, the scale of {chi}2 values is expected to be lower than in Table 3 due to the much lower observed frequencies.

Table 3 also lists the rules with a very high overall association and a moderate association of the genetic component alone (n=3, threshold for P=10–5, Pgen=0.01). The association of the SP-B intron 4 deletion variant allele seems to be slightly strengthened by different combinations of environmental factors, but this small improvement may be due to chance.

To conclude, the data mining analysis presented confirms and extends the association of the SP-B intron 4 deletion variant allele with BPD and prolonged requirement for oxygen after birth in very premature infants. The procedure also confirms that there are no unrecognized combinations of up to three explanatory variables that include genetic variables and are strongly associated with the disease phenotype.

Logistic regression analysis
We used logistic regression analysis to investigate whether the SP-B intron 4 alleles or genotypes explained the risk of BPD regardless of the confounding factors related to the aetiology of BPD. The presence of the SP-B intron 4 deletion variant allele, either as a heterozygote or as a homozygous genotype (P=0.036), and the homozygous deletion variant genotype (P=0.008) were risk factors for BPD even when the following dichotomous confounding factors were included in the regression analysis: gestational age (<28 or 28–32 weeks of gestation), asphyxia (yes/no), RDS (yes/no) and birth order (presenting/not presenting).

Transcription factor search
The regulation of gene expression at transcriptional and post-transcriptional levels involves the interaction between short DNA tracts and the corresponding trans-acting protein factors. Detection of such cis-acting elements, not only in the promoters but also in intronic sequences, may significantly contribute to our understanding of the regulatory mechanisms.

The variable region of SP-B intron 4 is composed of unique short sequence motifs that are separated by stretches of (CA)n microsatellite repeats. Several putative transcription factor binding sites were identified in the SP-B intron 4 invariant allele sequence (Table 4). Special attention was given to the potential binding elements that are lost or reorganized in the deletion variant allele when compared with the invariant allele. With this in mind, several transcription factors can be highlighted (Fig. 1). Putative single binding sites for the C/EBPß, IL-6 RE-BP, Zeste and AML1 factors that are present in the invariant allele are lost in the deletion variant lacking the sequence motifs 5–9. Furthermore, deletion of the motifs 5–9 could affect HNF-3/TTF-1 binding by bringing a potential downstream HNF-3 site relatively close to another HNF-3 site located 11 bp apart from the CAAG core motif for TTF-1.


View this table:
[in this window]
[in a new window]
  		Table 4. Number of putative transcription factor binding sites in the variable region of SP-B intron 4
 


View larger version (37K):
[in this window]
[in a new window]
  		Figure 1. Locations of putative TTF-1, HNF-3, C/EBPß, IL-6 RE-BP, Zeste and AML1 binding sites in SP-B intron 4. M1–M11 are the unique sequence motifs characteristic of the variable region, with CAn microsatellite repeats located between them. The sequence lost in the most common deletion variant allele (M5–M9, 170 nucleotides) is indicated underlined. TTF, thyroid transcription factor; HNF, hepatocyte nuclear factor; C/EBPß, CCAAT/enhancer-binding protein beta; IL-6 RE-BP, interleukin-6 response element-binding protein; Zeste, Drosophila bithorax complex regulatory gene; AML, acute myeloid leukaemia gene.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 RESULTS
 DISCUSSION
MATERIALS AND METHODS
 REFERENCES
 
Preterm infants likely to develop BPD are born with structurally immature lung. A variety of environmental factors related to the care of very premature infants may cause delayed alveolarization and other pathologic features characteristic of BPD (5). There is accumulating evidence to suggest that inflammatory processes influence the risk of BPD. Chorioamnionitis and increased concentrations of pro-inflammatory mediators IL-1ß, IL-6 and IL-8 in amniotic fluid and cord blood are antenatal risk factors, whereas post-natal risk factors include mechanical ventilation, infection, and elevated levels of proinflammatory cytokines in airway specimens (6,7,18,19). Potential candidate genes affecting BPD are those involved in the regulation of differentiation, growth, alveolarization and host defence mechanisms of the lung.

In the present study, Agrawal's algorithm based data mining in combination with conventional statistical allelic association study methods was used to demonstrate an association of the SP-B gene with BPD in a racially homogeneous Finnish population of 365 very preterm infants. The frequency of the SP-B intron 4 deletion variant allele was increased in the group of BPD infants, diagnosed at either 28 days post-partum or 36 weeks post-menstruation. Among the various environmental and genetic factors analyzed with the data mining procedure, a joint association of low birth weight and the SP-B intron 4 deletion variant allele with the BPD phenotype was evident. This association was strong only among the singletons and presenting multiples, who are exposed to the intrauterine cervix and more prone to ascending infections than the non-presenting multiples.

Although the SP-B exon 4 Ile131Thr and the intron 4 polymorphisms are in linkage disequilibrium, only the latter was associated with BPD. By dividing the control infants into two groups, those with RDS and those without RDS, we could demonstrate that the association of the SP-B intron 4 polymorphism existed between SP-B and BPD, but not between SP-B and RDS. Logistic regression analysis confirmed the association in the presence of confounding factors related to the aetiology of BPD. In a recent study consisting of 140 preterms and 58 term neonates, an increased frequency of BPD diagnosed at 28 days of age was observed among the SP-B intron 4 deletion variant carriers (8). Currently, the generalizability of the present data in different ethnic populations remains unknown. There are remarkable race-specific differences in the distribution and nucleotide sequences of the deletion and insertion variants (10,20,21). Therefore, it could be expected that similar associations between the deletion variants and BPD are probably present in Caucasian populations in general but may not be evident in black populations.

Candidate gene studies often underestimate the significance of non-coding sequence polymorphisms by focusing on non-synonymous or frame-shift changes in exons. However, the intron regions of the genome and repetitive DNA sequences have an important regulatory role in gene expression, e.g. at the mRNA level. A search for transcription factor binding sites showed that allelic variation of SP-B intron 4 results in significant changes in putative binding sites that may be important for SP-B gene regulation. Most of the sites predicted by the search are redundant, but a few potential sites were identified that may well be targets of transcription factors activated by specific proinflammatory cytokines. Of the transcription factors, at least C/EBPß, IL-6 RE-BP, TTF-1 and HNF-3 may serve as interesting targets for functional characterization. C/EBP isoforms are involved in the control of inflammatory functions and have lung-specific target genes, including at least SP-A and SP-D (22,23). IL-6 is a well-known proinflammatory cytokine involved in a variety of cellular responses, but its potential role in SP-B regulation is currently unknown. TTF-1 and HNF-3 are both regulators of SP-B promoter activity (24) and the inflammatory response (25,26). It is also possible that, instead of an enhancer effect of particular sequences in intron 4, the repetitive nature of the DNA regions as such has a quantitative or qualitative impact on transcription through a different mechanism. The size variations in the complex microsatellite motif region of SP-B intron 4 could therefore result in altered mRNA levels, thus influencing the expression of the protein. According to experimental evidence in mice, decreased SP-B levels are associated with acute lung injury and other physiological abnormalities (27–29). It has been presumed that surfactant function may be impaired or defective in patients with acute respiratory failure in chronic obstructive pulmonary disease (COPD) and squamous cell lung carcinoma, and increased SP-B intron 4 variant allele frequencies have been observed among them (30,31).

In addition to the major role of SP-B as a protein essential for surfactant function, accumulating evidence strongly supports the possibility that SP-B also acts as an inflammatory modulator. It has been shown to regulate the induction of immune responses via the airways in mice (32). According to recent findings of transgenic animal studies, SP-B has a protective effect against endotoxin-induced lung inflammation by enhancing surfactant function. This results in reduced lung injury, decreased influx of inflammatory cells and lower cytokine levels (16). Previous studies have shown decreased SP-B mRNA levels after intratracheal administration of TNF-{alpha} (33). Furthermore, in lung explants from moderately premature or newborn rabbits, the effects of TNF-{alpha}, IL-1 and bacterial endotoxin (LPS) suppressed the expression of surfactant proteins (34). If the length variations of the SP-B intron 4 do have an effect on mRNA levels under either normal or altered conditions, this could actually be a functional link with a cytokine-mediated inflammatory cascade leading to BPD.

Based on the present and previous population studies, we hypothesize that there are two closely linked polymorphisms in the SP-B gene that both have important functional significance, leading to altered genetic susceptibility to a variety of pulmonary diseases via separate molecular mechanisms. We propose that length variation in intron 4 plays a role in the transcriptional regulation of the SP-B gene, resulting in altered mRNA levels. On the other hand, the Ile131Thr polymorphism affects post-translationally the glycosylation of the amino terminus of proSP-B and has been proposed to affect the intracellular processing and sorting of SP-B (10,35). This may relate to a dual function of SP-B both as a surfactant protein and as an anti-inflammatory mediator and could be a basis towards understanding the divaricate role of SP-B in healthy subjects and patients with different diseases.

As a conclusion, our multiparameter analysis demonstrates the presumable direct involvement of the SP-B intron 4 deletion variant allele as a genetic risk factor to BPD. To support this, allele-specific differences were predicted at putative transcription factor binding sites that could potentially act as SP-B gene enhancer elements involved in critical inflammatory pathways leading to chronic lung disease. We also wish to highlight intronic regions in general as potentially important causative genetic factors in explaining complex trait variation and susceptibility to multifactorial diseases.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 RESULTS
 DISCUSSION
 MATERIALS AND METHODS
REFERENCES
 
Study population
The study population consisted of 365 infants born between 24 and 32 weeks of gestation in Oulu University Hospital Seinäjoki Central Hospital and Tampere University Hospital. Umbilical cord blood samples (n=156) were collected prospectively during the years 1997–2002, and buccal smear samples (n=209) were obtained from retrospectively recruited premature infants born during the years 1991–1999. All infants were of Finnish origin. The maternal and neonatal medical histories were evaluated for clinical data concerning various pre- and post-natal factors (Table 5). The diagnostic criterion of BPD used in the study was the continuous requirement for supplemental oxygen due to deficient lung function at the post-menstrual age of 36.0 weeks (gestation <32 weeks at birth) or at 56 days of age for those born at 32 weeks. All BPD cases required supplemental oxygen at the age of 28 days (2). In addition, all infants requiring supplemental oxygen due to respiratory deficiency at 4 weeks of post-natal age were identified. Oxygen saturation of hemoglobin was allowed to range between 87 and 97%. The severity of respiratory distress was scored according to Toce et al. (36). The clinical Toce score was at least 5 in infants requiring supplemental oxygen at the age of 4 weeks. The diagnosis of RDS was made as described (37). The infants who died before the diagnosis of BPD could be made and the infants from whom high-quality DNA samples were not available were excluded. Altogether 86 infants with BPD and 279 control infants were included in the study. The clinical characteristics of the infants are shown in Table 5. The study protocol was approved by the ethical committees of the participating centres. The parents of the preterm infants gave informed written consent for the genetic analyses.


View this table:
[in this window]
[in a new window]
  		Table 5. Clinical characteristics of premature infants with or without BPD
 
DNA sample preparation and genotyping of SP alleles
Altogether 16 SNPs and a length polymorphism were genotyped (Table 6). Genomic DNA was extracted from frozen, EDTA-anticoagulated blood samples using the Puregene DNA Isolation kit (Gentra Systems, Minneapolis, MN, USA). DNA was then diluted to 50 ng/µl. Genomic DNA was extracted from buccal smears using Chelex 100 medium (Bio-Rad, Hercules, CA, USA).


View this table:
[in this window]
[in a new window]
  		Table 6. SP variations and references to genotyping procedures
 
SP-A1, SP-A2, SP-B, SP-C and SP-D genotyping was performed according to previously published methods (10,11,37–39). The SP-B intron 4 length variant alleles were determined according to Haataja et al. (10). Owing to occasional failures in genotyping, slight variation occurred in the number of loci analyzed per sample.

Allele frequency comparisons and logistic regression analysis
The allele and genotype frequencies were calculated using SPSS for Windows versions 9 or 11 (SPSS Inc., Chicago, IL, USA). Comparisons of the allele and genotype frequencies were performed by the two-sided {chi}2 test using 2x2 or 2x3 contingency tables, respectively. The observed genotype frequencies of all polymorphisms were tested for Hardy–Weinberg equilibrium by {chi}2 analysis.

Logistic regression analysis was used to investigate whether SP-B intron 4 alleles or genotypes explained the risk of BPD (SPSS for Windows version 11) in the presence of confounding independent environmental factors. Variables in the regression analyses were included using forced entry, and the presence or absence of BPD was set as the dependent variable.

Data mining methods
To carry out data mining analysis of our data, we used Christian Borgelt's freeware implementation of Agrawal's Apriori algorithm (40). The algorithm is used to find association rules of the form A1{wedge}A2{wedge}···{wedge}An{Rightarrow}X, where A1, ..., An are combinations of binary explanatory variables, X is the disease phenotype and{wedge}denotes conjunction (‘and’). An example of an association rule is ‘If the individual is homozygous for the deletion variant allele at the SP-B{Delta} i4 locus, and the mother had antenatal steroid treatment, then the individual carries the BPD phenotype’. Data mining can be basically considered an extension of association analysis, which also takes into account the simultaneous effects of multiple genetic and environmental variables.

The following ante- and neonatal variables were included in the analysis: antenatal steroid treatment, PROM, gestational age at birth, pre-eclampsia (toxaemia), gender, sepsis, birth asphyxia (APGAR scores <7 at 5 min of age), RDS, birth weight and intrauterine growth retardation. Our data mining process was carried out in four steps:

  1. Categorizing continuous variables: the values of the continuous variables (gestational age and birth weight) were first divided into three ranges with equal observed frequencies. For each of the three ranges (high, medium, low), we generated a separate indicator variable. For the individuals whose values fell into a given range, the value of the corresponding indicator variable was set to 1.
  2. Deriving genetic variables: for each occurrence of a genotype a/b at each locus L, the following binary variables were set to the value of 1 : ‘L : a–b’ (presence of genotype a/b), ‘L : a–?’ (presence of allele a) and ‘L-HETa’ (presence of allele a in a heterozygous genotype, only if a!=b).
  3. Extracting the association rules: we searched for all association rules of the form A1{wedge}A2{wedge}···{wedge}An{Rightarrow}X, where n is 3 or lower, and X corresponds to either BPD or the broadly defined phenotype (BPD or supplemental oxygen for at least 28 days). The limit of 3 or fewer explanatory variables is feasible for a data set of this size: with more variables, the rules are likely to adapt to the data too closely.
  4. Finding interesting rules: out of the collection of association rules obtained, we selected the rules where the nominal P-value computed from {chi}2 statistics using a 2x2 contingency table (presence of the rule antecedent part A1{wedge}A2{wedge}···{wedge}An versus the disease phenotype) does not exceed P. To select rules where genetic factors are important contributors, it is further required that, for the derived rule B1{wedge}B2{wedge}···{wedge}Bm{Rightarrow}X, where the Bis are the genetic variables present in the rule R, the P-value does not exceed another threshold Pgen. In addition, we rejected redundant rules that can be derived by an equally or more significant rule by adding variables to the antecedent part. It should be borne in mind that the selection of relevant association rules is a challenging task with no single best solution. Owing to the multiple testing inherent in explorative analysis of this kind, we use the {chi}2 values and the corresponding P-values only as measures of association to display the rules and do not use them to draw far-fetching conclusions of statistical significance.

Computer-assisted sequence analysis
Transcription Element Search Software (TESS) (41) was utilized to search for putative transcription factor binding sites as potential regulators in the SP-B intron 4 sequence. TESS can predict binding sites using site or consensus strings and positional weight matrices from the TRANSFAC, IMD and CBIL-GibbsMat databases. The input SP-B invariant and deletion variant sequences were the same as described (10). In this context, the deletion variant refers to the most common variant allele lacking sequence motifs M5–M9, which accounts for >80% of all the deletion variant alleles in the Finnish population.


   ACKNOWLEDGEMENTS
 
The authors are grateful to Ms Maarit Hännikäinen and Ms Mirkka Ovaska for excellent technical assistance. Ms Eija Rautio is acknowledged for the collection of clinical data and Dr Marja-Leena Pokela for collection of blood samples. The study was supported by grants from Biocenter Oulu (M.R. and M.H.), the Sigrid Juselius Foundation (M.H.) and the Foundation for Paediatric Research in Finland (R.M.).


   FOOTNOTES
 
* To whom correspondence should be addressed. Tel: +358 85376602; Fax: +358 85376600; Email: rhh{at}cc.oulu.fi


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 RESULTS
 DISCUSSION
 MATERIALS AND METHODS
 REFERENCES
 

  1. Stevenson, D.K., Wright, L.L., Lemons, J.A., Oh, W., Korones, S.B., Papile, L.A., Bauer, C.R., Stoll, B.J., Tyson, J.E., Shankaran, S., Fanaroff, A.A., Donovan, E.F., Ehrenkranz, R.A. and Verter, J. (1998) Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Research Network, January 1993 through December 1994. Am. J. Obstetr. Gynecol., 179, 1632–1639.[CrossRef][ISI][Medline]

  2. Jobe, A.H. and Bancalari, E. (2001) Bronchopulmonary dysplasia. Am. J. Respir. Crit. Care Med., 163, 1723–1729.[Free Full Text]

  3. Parker, R.A., Lindstrom, D.P. and Cotton, R.B. (1996) Evidence from twin study implies possible genetic susceptibility to bronchopulmonary dysplasia. Semin. Perinatol., 20, 206–209.[CrossRef][ISI][Medline]

  4. Hallman, M. and Haataja, R. (2003) Genetic influences and neonatal lung disease. Semin. Neonatol., 8, 19–27.[CrossRef][Medline]

  5. Jobe, A.H. (2003) Antenatal factors and the development of bronchopulmonary dysplasia. Semin. Neonatol., 8, 9–17.[CrossRef][Medline]

  6. Yoon, B.H., Romero, R., Jun, J.K., Park, K.H., Park, J.D., Ghezzi, F. and Kim, B.I. (1997) Amniotic fluid cytokines (interleukin-6, tumor necrosis factor-alpha, interleukin-1 beta, and interleukin-8) and the risk for the development of bronchopulmonary dysplasia. Am. J. Obstetr. Gynecol., 177, 825–830.[CrossRef][ISI][Medline]

  7. Van Marter, L.J., Dammann, O., Allred, E.N., Leviton, A., Pagano, M., Moore, M. and Martin, C. (2002) Chorioamnionitis, mechanical ventilation, and postnatal sepsis as modulators of chronic lung disease in preterm infants. J. Pediatr., 140, 171–176.[CrossRef][ISI][Medline]

  8. Makri, V., Hospes, B., Stoll-Becker, S., Borkhardt, A. and Gortner, L. (2002) Polymorphisms of surfactant protein B encoding gene: modifiers of the course of neonatal respiratory distress syndrome? Eur. J. Pediatr., 161, 604–608.[CrossRef][ISI][Medline]

  9. Haataja, R. and Hallman, M. (2002) Surfactant proteins as genetic determinants of multifactorial pulmonary diseases. Ann. Med., 34, 324–333.[CrossRef][ISI][Medline]

  10. Haataja, R., Ramet, M., Marttila, R. and Hallman, M. (2000) Surfactant proteins A and B as interactive genetic determinants of neonatal respiratory distress syndrome. Hum. Mol. Genet., 9, 2751–2760.[Abstract/Free Full Text]

  11. Lahti, M., Lofgren, J., Marttila, R., Renko, M., Klaavuniemi, T., Haataja, R., Ramet, M. and Hallman, M. (2002) Surfactant protein D gene polymorphism associated with severe respiratory syncytial virus infection. Pediatr. Res., 51, 696–699.[CrossRef][ISI][Medline]

  12. Lofgren, J., Ramet, M., Renko, M., Marttila, R. and Hallman, M. (2002) Association between surfactant protein A gene locus and severe respiratory syncytial virus infection in infants. J. Infect. Dis., 185, 283–289.[CrossRef][ISI][Medline]

  13. Crouch, E., Hartshorn, K. and Ofek, I. (2000) Collectins and pulmonary innate immunity. Immunol. Rev., 173, 52–65.[CrossRef][ISI][Medline]

  14. Pryhuber, G.S. (1998) Regulation and function of pulmonary surfactant protein B. Mol. Genet. Metab., 64, 217–228.[CrossRef][ISI][Medline]

  15. Glasser, S.W., Burhans, M.S., Korfhagen, T.R., Na, C.L., Sly, P.D., Ross, G.F., Ikegami, M. and Whitsett, J.A. (2001) Altered stability of pulmonary surfactant in SP-C-deficient mice. Proc. Natl Acad. Sci. USA, 98, 6366–6371.[Abstract/Free Full Text]

  16. Epaud, R., Ikegami, M., Whitsett, J.A., Jobe, A.H., Weaver, T.E. and Akinbi, H.T. (2003) Surfactant protein B inhibits endotoxin-induced lung inflammation. Am. J. Respir. Cell. Mol. Biol., 28, 373–378.[Abstract/Free Full Text]

  17. Marttila, R., Haataja, R., Ramet, M., Lofgren, J. and Hallman, M. (2003) Surfactant protein B polymorphism and respiratory distress syndrome in premature twins. Hum. Genet., 112, 18–23.[CrossRef][ISI][Medline]

  18. Hitti, J., Krohn, M.A., Patton, D.L., Tarczy-Hornoch, P., Hillier, S.L., Cassen, E.M. and Eschenbach, D.A. (1997) Amniotic fluid tumor necrosis factor-alpha and the risk of respiratory distress syndrome among preterm infants. Am. J. Obstetr. Gynecol., 177, 50–56.[CrossRef][ISI][Medline]

  19. Munshi, U.K., Niu, J.O., Siddiq, M.M. and Parton, L.A. (1997) Elevation of interleukin-8 and interleukin-6 precedes the influx of neutrophils in tracheal aspirates from preterm infants who develop bronchopulmonary dysplasia. Pediatr. Pulmonol., 24, 331–336.[CrossRef][ISI][Medline]

  20. Floros, J., Veletza, S.V., Kotikalapudi, P., Krizkova, L., Karinch, A.M., Friedman, C., Buchter, S. and Marks, K. (1995) Dinucleotide repeats in the human surfactant protein-B gene and respiratory-distress syndrome. Biochem. J., 305, 583–590.

  21. Veletza, S.V., Rogan, P.K., TenHave, T., Olowe, S.A. and Floros, J. (1996) Racial differences in allelic distribution at the human pulmonary surfactant protein B gene locus (SP-B). Exp. Lung Res., 22, 489–494.[ISI][Medline]

  22. Poli, V. (1998) The role of C/EBP isoforms in the control of inflammatory and native immunity functions. J. Biol. Chem., 273, 29279–29282.[Free Full Text]

  23. Cassel, T.N. and Nord, M. (2003) C/EBP transcription factors in the lung epithelium. Am. J. Physiol. Lung Cell. Mol. Physiol., 285, L773–L781.[Abstract/Free Full Text]

  24. Margana, R.K. and Boggaram, V. (1997) Functional analysis of surfactant protein B (SP-B) promoter. Sp1, Sp3, TTF-1, and HNF-3alpha transcription factors are necessary for lung cell-specific activation of SP-B gene transcription. J. Biol. Chem., 272, 3083–3090.[Abstract/Free Full Text]

  25. DeFelice, M., Silberschmidt, D., DiLauro, R., Xu, Y., Wert, S.E., Weaver, T.E., Bachurski, C.J., Clark, J.C. and Whitsett, J.A. (2003) TTF-1 phosphorylation is required for peripheral lung morphogenesis, perinatal survival, and tissue-specific gene expression. J. Biol. Chem., 278, 35574–35583.[Abstract/Free Full Text]

  26. Hromas, R. and Costa, R. (1995) The hepatocyte nuclear factor-3/forkhead transcription regulatory family in development, inflammation, and neoplasia. Crit. Rev. Oncol. Hematol., 20, 129–140.[ISI][Medline]

  27. Ingenito, E.P., Mora, R., Cullivan, M., Marzan, Y., Haley, K., Mark, L. and Sonna, L.A. (2001) Decreased surfactant protein-B expression and surfactant dysfunction in a murine model of acute lung injury. Am. J. Respir. Cell. Mol. Biol., 25, 35–44.[Abstract/Free Full Text]

  28. Clark, J.C., Weaver, T.E., Iwamoto, H.S., Ikegami, M., Jobe, A.H., Hull, W.M. and Whitsett, J.A. (1997) Decreased lung compliance and air trapping in heterozygous SP-B-deficient mice. Am. J. Respir. Cell. Mol. Biol., 16, 46–52.[Abstract]

  29. Tokieda, K., Iwamoto, H.S., Bachurski, C., Wert, S.E., Hull, W.M., Ikeda, K. and Whitsett, J.A. (1999) Surfactant protein-B-deficient mice are susceptible to hyperoxic lung injury. Am. J. Respir. Cell Mol. Biol., 21, 463–472.[Abstract/Free Full Text]

  30. Seifart, C., Plagens, A., Brodje, D., Muller, B., von Wichert, P. and Floros, J. (2002) Surfactant protein B intron 4 variation in German patients with COPD and acute respiratory failure. Dis. Markers, 18, 129–136.[ISI][Medline]

  31. Seifart, C., Seifart, U., Plagens, A., Wolf, M. and von Wichert, P. (2002) Surfactant protein B gene variations enhance susceptibility to squamous cell carcinoma of the lung in German patients. Br. J. Cancer, 87, 212–217.[CrossRef][ISI][Medline]

  32. van Iwaarden, J.F., Claassen, E., Jeurissen, S.H., Haagsman, H.P. and Kraal, G. (2001) Alveolar macrophages, surfactant lipids, and surfactant protein B regulate the induction of immune responses via the airways. Am. J. Respir. Cell. Mol. Biol., 24, 452–458.[Abstract/Free Full Text]

  33. Pryhuber, G.S., Bachurski, C., Hirsch, R., Bacon, A. and Whitsett, J.A. (1996) Tumor necrosis factor-alpha decreases surfactant protein B mRNA in murine lung. Am. J. Physiol., 270, L714–L721.

  34. Vayrynen, O., Glumoff, V. and Hallman, M. (2004) Inflammatory and anti-inflammatory responsiveness of surfactant proteins in fetal and neonatal rabbit lung. Pediatr. Res., 55, 55–60.[ISI][Medline]

  35. Marttila, R., Haataja, R., Guttentag, S. and Hallman, M. (2003) Surfactant protein a and B genetic variants in respiratory distress syndrome in singletons and twins. Am. J. Respir. Crit. Care Med., 168, 1216–1222.[Abstract/Free Full Text]

  36. Toce, S.S., Farrell, P.M., Leavitt, L.A., Samuels, D.P. and Edwards, D.K. (1984) Clinical and roentgenographic scoring systems for assessing bronchopulmonary dysplasia. Am. J. Dis. Child., 138, 581–585.[Abstract]

  37. Ramet, M., Haataja, R., Marttila, R., Floros, J. and Hallman, M. (2000) Association between the surfactant protein A (SP-A) gene locus and respiratory-distress syndrome in the Finnish population. Am. J. Hum. Genet., 66, 1569–1579.[CrossRef][ISI][Medline]

  38. DiAngelo, S., Lin, Z., Wang, G., Phillips, S., Ramet, M., Luo, J. and Floros, J. (1999) Novel, non-radioactive, simple and multiplex PCR–cRFLP methods for genotyping human SP-A and SP-D marker alleles. Dis. Markers, 15, 269–281.[ISI][Medline]

  39. Lahti, M., Marttila, R. and Hallman, M. (2004) Surfactant protein C gene variation in the Finnish population—association with perinatal respiratory disease. Eur. J. Hum. Genet., 12, 312–320.[CrossRef][ISI][Medline]

  40. Agrawal, R, Imielinski, T and Swami, A. (1993) Mining association rules between sets of items in large databases. Proceedings of the 1993 ACM SIGMOD Conference, Washington DC, USA. pp. 207–216.

  41. Schug, J. and Overton, C. (1997) TESS, Transcription Element Search Software on the WWW. Technical Report CBIL-TR-1997-1001-v0.0; Computational Biology and Informatics Laboratory School of Medicine, University of Pennsylvania, PA, USA.


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Arch. Dis. Child. Fetal Neonatal Ed.Home page
D. Harding
Impact of common genetic variation on neonatal disease and outcome
Arch. Dis. Child. Fetal Neonatal Ed., September 1, 2007; 92(5): F408 - F413.
[Abstract] [Full Text] [PDF]

Home page
 NeoReviewsHome page
V. Bhandari and J. R. Gruen
The Genomics of Bronchopulmonary Dysplasia
NeoReviews, August 1, 2007; 8(8): e336 - e344.
[Abstract] [Full Text] [PDF]

Home page
 PediatricsHome page
H.-C. Lin, C.-H. Tsai, and F.-J. Tsai
Association Study of Gene Polymorphism and Bronchopulmonary Dysplasia
Pediatrics, January 1, 2005; 115(1): 198 - 199.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
13/11/1095    most recent
ddh132v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (24)
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Rova, M.
Right arrow Articles by Hallman, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rova, M.
Right arrow Articles by Hallman, M.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?