METHODS

Study subjects

We carried out a study of IUGR in relation to thrombophilic polymorphisms.¹⁷ Cases were newborns whose birth weight was below the 10th centile for gestational age and sex, based on Canadian standards.¹⁸ All cases seen at our centre between May 1998 and June 2000 who were born alive after the 24th week of gestation and without severe congenital anomalies were eligible for the study if the mother agreed to participate. The project was approved by the Institutional Review Board of the hospital. Informed consent was signed by the mother to collect cord and maternal blood. During that period, 505 newborns met the criteria for cases and 493 participated in the study (97.6%). In the original study, we also included controls, which are not used in the present report. Midway through this study, we started to collect buccal swabs from the fathers of babies to be included in the final phase of the study. The goal was to analyse case-parental trios (mother, father, newborn) to test for association and linkage.¹⁹ Among the fathers contacted, 86% provided genetic material. We genotyped 449 newborns and 440 mother-newborn pairs for APOE (89% of all case pairs); there was enough DNA from the buccal swabs remaining to genotype APOE in 194 fathers (78% of fathers providing DNA). Genotyping of all three family members was complete for 170 trios.

Laboratory investigation

Human genomic DNA was extracted from whole blood samples (mothers and newborns) or from buccal swabs (fathers), as previously described.¹⁷ Briefly, PCR reactions were performed with a final reaction volume of 50 l, using 50–100 ng of DNA template per tube under the following conditions: an initial DNA denaturation step at 94°C for three minutes before adding the mixture containing the Taq DNA polymerase enzyme; this was followed by a 40 cycle sequence of primer annealing at 62°C for 30 seconds, extension at 72°C for one minute, and denaturation at 94°C for 30 seconds with a 10 minute final extension step at 72°C. The PCR APOE primer sequences were as follows: 5′ CGGGCACGGCTGTCCAAGGA 3′ (forward) and 5′ CGGGCCCCGGCCTGGTACAC 3′ (reverse).²⁰ Allele specific oligonucleotide hybridisation assays were performed as described by others.²¹ PCR products were denatured, divided into aliquots, and blotted onto nitrocellulose membranes. Positive and negative controls were included on each membrane. Specimens of family members were assigned randomly to membranes. After hybridisation and washing, the membranes were read using PhosphoImager (Molecular Dynamics, Sunnyvale, CA) and an automatic scanning program. Membranes were also read visually by two independent observers and again by three observers together. Disagreements after this step were resolved by reamplification, digestion with HhaI restriction enzyme, and gel electrophoresis.²⁰

Statistical analysis

The general goal of the analysis was to determine if transmission of an allele from the parents to the newborn departs from the expected probability (50%). A statistically significant departure from the expected (or preferential transmission) is indicative of linkage and association. First, we used the transmission disequilibrium test (TDT) that analyses the transmission of alleles from heterozygous parents to their affected child. To analyse the trait defined as presence of IUGR, we used the Family Based Association Test (FBAT) program.²² FBAT replaces missing values for the unobserved parental genotype using the distribution of offspring genotype.²³ We also used the TDT program from STATA,²⁴ which does not replace missing values; both programs use a McNemar test for matched pairs and gave exactly similar statistical test results and p values. We report the results of the former. The FBAT test provides correct results regardless of population admixture, the true genetic model, and the sampling strategy.^22,23 With FBAT, we also analysed the trait as quantitative, using birth weight, adjusted for sex, gestational age, and race; the residuals standardised to variance equal to 1 were used as the quantitative trait in this analysis. In each instance (using IUGR or birth weight as the trait), biallelic tests were carried out (each allele against the others) as well as a global chi-square test for all the alleles (multi-allelic test); an additive model, which counts the number of alleles the offspring have, was used. In the second approach, we used a logistic regression model as an extension of the TDT. Here, the outcome is defined as the transmission of the high risk allele (coded as 1) versus the transmission of the other allele (coded as 0).²⁵ With a multi-allelic locus, one can define one of the alleles as high risk and contrast its transmission with that of the other alleles. We chose the ε2 allele as the high risk allele. In this analysis, a model including only the intercept is equivalent to a TDT test. Other factors can be included in the model to determine if the transmission probabilities are modified by these factors. The former TDT analyses treat transmissions from parents independently. Schaid and Sommer^26,27 proposed a method to study the association of a candidate gene with disease using case and parental data; the program GASSOC,²⁸ which was developed to test for the transmission of alleles from both parents together, was used in our third analysis.

Finally, we carried out an analysis using FBAT without 32 trios where the placenta, on routine pathological examination showed, as previously reported,¹⁷ signs of infarction and where the proband was genotyped for APOE. The second and third analyses (logistic regression and GASSOC) use only complete trios; there were 10 trios where a proband with placental infarction was involved and we removed these.

RESULTS

Table 1 shows the allele and genotype frequency distributions in 449 probands genotyped for APOE according to maternal racial groups. Allele ε2 was found more frequently in blacks than in whites as was allele ε4. The other racial groups included small numbers of subjects. The average birth weight in the genotyped group of newborn cases was 2425 g (±577); 54.8% were girls and 20.5% were born before the 37th week of gestation.

Results of the TDT analysis using the FBAT program²² with the observed and expected numbers of transmitted alleles are shown in table 2. There was a significantly reduced transmission of allele ε2 and a marginally significant excess transmission of allele ε3. Transmission of allele ε4 did not depart from the expected. Overall, transmission of apo ε significantly departed from the expected probabilities. Removing the 32 trios where probands had signs of placental infarction did not change any of the conclusions: for allele ε2, 16 transmissions were observed while 27 were expected (p=0.002), whereas for allele ε3 these were 132 and 199, respectively (p=0.02), and 44 and 44.5, respectively, for allele ε4 (p=0.74). The global chi-square was 9.78 (p=0.007).

Table 3 shows the results of the analysis with the FBAT program when using the birth weight of the probands as the trait. Although there was a tendency towards a deficit in transmission of allele ε2, the results were not statistically significant. Removing the 32 trios where probands had signs of placental infarction did not change any of the conclusions (data not shown).

A logistic regression analysis with transmission of allele ε2 as the outcome of interest confirmed the results of all previous analyses showing a significantly reduced transmission of allele ε2 (results not shown). Introducing race, sex, birth weight, gestational age, and maternal smoking, each in turn in the model, did not affect the transmission probabilities as measured by the likelihood ratio (LR) chi-square statistic comparing nested models. Removing from this analysis, which uses complete trios, the 10 trios where probands had signs of placental infarction did not materially change the results.

Finally, using the GASSOC program, we rejected the null hypothesis that the genotype relative risks were null based on the LR statistic (7.7, 2 df, p=0.02). Under the additive model, allele ε2 was associated with half the risk of allele ε3, used as the reference allele (relative risk (RR) = 0.44, p=0.008), while the risk was close to null for allele ε4 (RR=0.82, p=0.39). The dominant model was also compatible with the data (score statistic = 6.8, 2 df, p=0.03). The recessive model could not be estimated well with only one case carrying the ε2/ε2 genotype. Removing from the analysis the 10 trios where the proband had signs of placental infarction did not materially change the results.

DISCUSSION

Using a family based study design and related statistical tests, we consistently found a significantly reduced transmission of allele ε2 to newborns affected with intrauterine growth restriction; in other words, allele ε2 seems protective against IUGR. Overall, results indicated significant linkage disequilibrium (linkage and allelic association) between the APOE polymorphisms and IUGR. On the other hand, using the probands’ birth weight as the trait, we did not observe significant deviation in the number of transmissions from the expected. One possible explanation for the latter results is the relatively limited variation in birth weight among cases. Removing the trios where placental infarction was found on routine pathological examination did not alter any of the conclusions, although the results for the presence of IUGR were even more statistically significant. This could point to a different cause for the growth restricted newborns with placental infarction because their inclusion in the analysis seems to dilute the effect; however, there was only a small number of such newborns and placental infarction reported on gross examination was not confirmed histologically in this study.

If we assume that allele ε2 is protective against later cardiovascular disease, the fact that it is less often transmitted to babies who are born small gives support to the Barker hypothesis. Indeed, the hypothesis suggests that newborns with small body size are more prone to later cardiovascular diseases. To our knowledge, there are no previous reports showing linkage disequilibrium between the APOE locus and IUGR.

The relation between the apo ε2 allele and cardiovascular disease is complex. A protective role for the allele in the development of CVD has been reported^29,30 and this seems particularly marked in younger people.³¹ The apo ε2 allele is also associated with lower LDL cholesterol levels,^15,16,32 as well as with a survival advantage^33,34 which could be the result of a reduced risk of cardiovascular disease. However, the apo ε2 allele is also associated with higher triglyceride levels and possibly with other adverse outcomes such as diabetic nephropathy in type I diabetes³⁵ and hypertension.³⁶ In addition, homozygosity for the apo ε2 allele predisposes to the development of type III hyperlipidaemia.³⁷ Our data suggested a small excess in the transmission of allele ε3, for which chance is still an explanation. Long term follow up studies conducted in mice overexpressing apoE3 showed clear retardation of atherosclerotic and xanthomatous lesions.³⁸ Finally, the role of the apo ε4 allele on CVD seems clearer: its presence has been associated with an increased risk in CVD as well as with higher total and LDL cholesterol³⁰ compared with the other apoE isoforms. We found no departure from expected in the transmission of allele ε4. It is possible that functions of the apoE isofoms other than those associated with lipid abnormalities are involved in IUGR. However, this study cannot address these issues. The group of IUGR newborns in this study and their mothers were very typical in their clinical and personal characteristics of similar published groups.¹⁷ There were population substructures in the studied group, but the family based analysis we used is robust against such a bias.

In conclusion, our results are indicative that the apo ε2 allele is transmitted significantly less often than expected among babies whose birth weight for gestational age and sex was below the 10th centile. Because the apo ε2 allele has been associated with a lower risk of cardiovascular disease, and babies born with growth restriction have been found to be at higher risk of cardiovascular disease, our data reconcile these two observations. IUGR is a complex disease about which we know little in terms of mechanisms. In a previous study, we have excluded the role of thrombophilic polymorphisms as potential contributing causes for IUGR status.^17,39 The results of the present study may suggest an underlying atherosclerotic mechanism for IUGR. Despite the plausibility of our results, they need to be replicated in independent studies.