Introduction

Abstract Introduction Materials & Methods Results Discussion References

Ethanol consumption during pregnancy is one of the leading known causes of congenital mental retardation (for a review, see Whitfield and Martin, 1996). However, adverse infant outcome after intrauterine exposure to alcohol is highly variable. Of the offspring of women who continue to drink heavily while pregnant, <10% will have the most severe form: fetal alcohol syndrome (Sokol et al., 1986). More commonly, offspring may exhibit one or more effects termed alcohol-related birth defects (Abel and Dintcheff, 1985); these include developmental deficits documented by studies demonstrating a correlation between intrauterine alcohol exposure and lower scores on the MDI of the Bayley Scales of Infant Development (Golden et al., 1982; Streissguth et al., 1980).

Epidemiological studies controlling for ethanol intake have identified African American ethnicity, older maternal age and greater maternal parity as risk factors for adverse outcome after intrauterine ethanol exposure (Sokol et al., 1986, 1980). Multiple animal studies verifying the profile of blood alcohol concentrations as a critical risk factor support intersubject variation in ethanol metabolism as a determinant of susceptibility to ethanol teratogenicity. In animal models, the degree of microcephaly correlates with peak blood alcohol concentrations (Bonthius et al., 1988); adverse effects occur even from a single-dose exposure associated with high blood alcohol concentrations (Goodlett et al., 1990). Maternal age appears to increase risk (Abel and Dintcheff, 1985), and progressive increases in blood alcohol concentrations are found with increasing maternal age (Church et al., 1990). Among animal strains selected for diversity in the enzymes responsible for ethanol metabolism, differences in maternal blood ethanol concentrations correlated positively with fetal abnormalities and with differences in ethanol-metabolizing enzymes (Chernoff, 1980). Significant variation in human ethanol metabolism has been reported (Wagner et al., 1989; for a review, see Holford, 1987), and estimates of the genetic portion of the variation in ethanol metabolism have ranged from 49% to 98% (Martin et al., 1985; Vesell et al., 1971).

Ethanol is metabolized by ADH and CYP2E1 to acetaldehyde, which is metabolized by ALDH to acetate. The conversion of ethanol to acetaldehyde is believed to be the rate-limiting step, and ADH is believed by some to be the more important enzyme in this step (Mezey, 1976). At least four classes of human ADH have been identified (for a review, see Jornvall and Hoog, 1995); however, based on the enzyme quantity and catalytic activity, the class I ADH isoforms appear to be most important in the oxidation of ethanol. The class I ADH isoenzymes, heterodimers composed of subunit chains designated alpha, beta, and gamma, have been isolated and sequenced. The responsible genetic loci have been sequenced and labeled ADH1, ADH2 and ADH3, respectively (Jornvall and Hoog, 1995). Polymorphisms have been observed at the ADH2 and ADH3 loci, but it is likely the importance of these polymorphisms to variation in ethanol metabolism differs. The enzymes encoded by the two possible alleles at the ADH3 locus are very similar in their in vitro kinetic constants (Bosron et al., 1983a). In contrast, expression of the three possible alleles at the ADH2 locus results in enzymes that are very different in activity (Bosron et al., 1983a; Burnell et al., 1989; Ehrig et al., 1990). The most common allele at that locus, ADH2*1, occurs with varying frequencies in Oriental, Caucasian, native American and African American populations (Bosron et al., 1988; Bosron and Li, 1987; Chen et al., 1992; Rex et al., 1985). The majority of Oriental individuals and a smaller percentage of Caucasians express the ADH2*2 allele, whereas the ADH2*2 gene frequency is low (2%) among native Americans. The ADH2*3 allele, which encodes for the beta-3 isozyme subunit, appears to be unique to African Americans, occurring with a frequency of ~15% to 20% (Bosron et al., 1983b; Bosron and Li, 1987). The kinetic parameters of the beta-3 beta-3 isozyme (K_m = 36 mM ethanol, V_max = 300 min¹) differs from those of the isoform determined by the common ADH2*1 allele by >30-fold (beta-1 beta-1 K_m = 0.049 mM, V_max = 9.2 min¹) (Bosron et al., 1983a; Burnell et al., 1989).

Noting the increased incidence of alcohol-related birth defects in African Americans (Centers for Disease Control, 1993), the importance of alcohol concentrations and ethanol metabolism and the likely impact of a genetic polymorphism unique to African Americans, we tested the hypothesis that allelic differences at the ADH2 locus affect offspring susceptibility to intrauterine ethanol exposure among African Americans. We evaluated the neurobehavioral status and growth outcome of prospectively selected offspring of mothers with a range of both maternal alcohol intake during pregnancy and maternal ADH2 genotype.

	Materials and Methods

Abstract Introduction Materials & Methods Results Discussion References

Design. Pregnant women were selected for recruitment using a stratified design to ensure a range of both alcohol intake and ADH2 genotypes. Women attending their first antenatal visit at the obstetrical clinic at the University Health Center (Detroit, MI) were interviewed as described below to determine their alcohol intake during the previous 2 weeks and the periconceptional period. For enrollment stratification, those with heavy periconceptional alcohol intake, defined as 0.5 oz average absolute alcohol daily, and those with light alcohol intake, defined as <0.5 oz average absolute alcohol daily intake, were selected for ADH2 genotype determinations. Only African American women were recruited because, as previously stated, ADH2*3 occurs only in that population. Based on both periconceptional maternal alcohol intake and maternal ADH2 genotype, women were selected for subsequent offspring developmental evaluation. Because a subset of these women were selected for a study of ethanol metabolism (reported elsewhere), women were excluded from enrollment if they had liver disease, gastrointestinal bleeding, diabetes or hyperthyroidism or were taking any medications known to affect drug metabolism. Each woman gave informed consent for evaluation of her alcohol intake in pregnancy and her genotype. A second consent for infant genotype and developmental evaluation was obtained at the time of infant evaluation. This study was approved by the institutional review board.

Determination of alcohol intake. The periconceptional alcohol intake was obtained at the initial antenatal visit by interviewer-directed patient recall of a day-by-day history of a typical week of alcohol consumption immediately before pregnancy. The type of alcohol (beer, wine, wine cooler or liquor) and the amount of each in ounces were obtained. Quantities were converted to ounces of absolute alcohol using the following multipliers: liquor, 0.4; beer, 0.04; wine, 0.20 and wine coolers, 0.05 (Bowman et al., 1975). A similar day-by-day history was obtained for the 2 weeks before the first antenatal visit. At each subsequent antenatal visit, a similar interviewer-directed recall of both the type and amount of alcohol consumed in 2 previous weeks was performed.

Assessment of other maternal substance use. Mothers were considered illicit drug users if either maternal history, maternal or infant urine screening or infant meconium screening was positive. Urine and meconium samples were tested by radioimmunoassay for benzoylecogonine, the major metabolite of cocaine, opiates and cannabinoids.

ADH2 genotype determination. Blood (10 drops), drawn by fingerstick or venipuncture (maternal) or heel stick (infant), was placed on diagnostic filter paper for ADH2 genotyping. DNA was isolated and denatured, and in vitro amplification of DNA by the polymerase chain reaction was performed using primers specific for the alleles of interest (Xu et al., 1988). Subsequently, allele detection at the ADH2 locus was done by autoradiographic detection of the hybridization of sample DNA segments to allele-specific oligonucleotide probes for ADH2*1 and ADH2*3.

Selection of offspring for study. Women of both genotypes and with a wide range of alcohol intake were selected for recruitment for evaluation of their offspring. After delivery, the infant's medical record was reviewed. Infant exclusion criteria included birth weight of <1500 g, gestational age of <32 weeks, chromosomal anomalies, severe perinatal asphyxia and central nervous system structural anomalies. The obstetrical estimate of gestational age, based on the date of the last menstrual cycle and/or fetal ultrasound, was accepted as valid unless it differed by >2 weeks from that obtained using the newborn Ballard examination. For infants with >2-week disparity, the estimate from the Ballard examination was accepted as the gestational age. To minimize attrition within the study population, each mother was contacted by telephone or mail every month both before and after delivery.

Offspring developmental assessment. Psychometric testing was done by one of three testers (concordance correlation coefficient r² > .9) at 12 months ± 2 weeks postnatal age using the Bayley Scales of Infant Development. Age of testing was corrected for infants born between 32 and 38 weeks' gestation. The developmental examiners were masked to the history of antenatal alcohol exposure, illicit substance use and ADH2 genotype data. Other infant outcome variables obtained by infant examiners blinded to fetal exposure status were birth weight, birth length and birth head circumference. An infant was considered small for gestational age if the birth weight was less than the 10th percentile for gestational age. Microcephaly was defined as a head circumference of <10th percentile.

Statistical methods. All outcome variables were evaluated for normality. Variables that were normally distributed were described using mean and standard deviation values. Those that were skewed were described using median and range values. For inclusion into testing, variables with skewed distributions were either transformed to a normal distribution by taking the log of the value or stratified into discrete variables. To evaluate the possible effect of genotype and drinking on offspring development, the population was divided into four groups based on the presence of maternal drinking in the 2 weeks before the first antenatal visit and the presence of a maternal ADH2*3 allele. Differences among these four groups was tested using ANOVA and Duncan's post hoc testing. Each maternal demographic and substance use variable and maternal and offspring genotypes, as well as interaction terms, were tested for association with the offspring growth and development outcome variables using stepwise linear regression. All variables with even minimal association with the outcome variable (P < .1) were retained in the model. Multiple logistic regression was used to evaluate risk factors for categorical outcomes.

	Results

Abstract Introduction Materials & Methods Results Discussion References

Sample characteristics. The study population included 243 mothers (table 1) and their offspring (table 2). By design, the sample population was not representative of the general African American population because mother-infant dyads were selected based on both maternal periconceptional alcohol intake and maternal ADH2 genotype. This design resulted in a nonrepresentative distribution for both of these variables, as well as for infant ADH2 genotype (table 2). The frequency of the ADH2*3 and ADH2*1 allele in the maternal sample population was 0.33 and 0.66, respectively, whereas that of the infants was 0.30 and 0.70, respectively. In this sample population, the use of alcohol at the time of the first antenatal visit was associated with cocaine use and cigarette smoking but not with opioid or cannabinoid use (table 3). As anticipated from the study design, alcohol intake and maternal ADH2 genotype were not correlated (P > .1).

Offspring mental development. The presence of the ADH2*3 allele was associated with higher Bayley scores in the offspring of drinking women but not in the offspring of nondrinking women. Drinking during pregnancy was associated with lower Bayley MDI scores, but this effect was significant only in the offspring of mothers without an ADH2*3 allele (fig. 1, P < .01, ANOVA, Duncan's post hoc test). The offspring of drinking women with at least one ADH2*3 allele had MDI scores similar to those of nondrinking women of each ADH2 genotype. A similar analysis of offspring genotype showed comparable results (fig. 2). MDI scores were lower among drinking women, but this effect was accounted for by lower scores in the offspring without an ADH2*3 allele (P < .05, ANOVA, Duncan's post hoc test). Offspring with at least one ADH2*3 allele had similar MDI scores, regardless of maternal drinking history.

View larger version (36K): [in this window] [in a new window]   Fig. 1.   Impact of maternal ADH2 genotype on the outcome of offspring of women abstaining (hatched bars) or drinking (filled bars) during pregnancy. Offspring neurobehavioral outcome was measured as the MDI score of the Bayley Scales of Infant Development at 1 year of age (total N = 243). Bayley scores were significantly lower among offspring whose mothers lacked an ADH2*3 allele and consumed ethanol during pregnancy. (*P < .01, ANOVA, Duncan's post hoc). Data shown as mean ± S.D.

View larger version (37K): [in this window] [in a new window]   Fig. 2.   Impact of offspring ADH2 genotype on the outcome of offspring of women abstaining (hatched bars) or drinking (filled bars) during pregnancy. Offspring neurobehavioral outcome was measured as the MDI score of the Bayley Scales of Infant Development at 1 year of age (total N = 243). Bayley scores were significantly lower among offspring without an ADH2*3 allele whose mothers consumed ethanol during pregnancy (P < .05, ANOVA, Duncan's post hoc). Data shown as mean ± S.D.

Effect of alcohol consumption during pregnancy on offspring growth measures. When considered as a single independent variable, ethanol intake during pregnancy was associated with significant decreases in offspring growth (fig. 3). Smaller birth weights were associated with shorter gestational age, the interaction between drinking at the first antenatal visit and absence of a maternal ADH2*3 allele and twin gestation (table 5, r² = .041, P < .0001, stepwise linear regression). With these variables in the model, no other variable (including alcohol intake during pregnancy, maternal ADH2 genotype, infant ADH2 genotype, other substance use and mother and infant demographic variables or any of the interaction terms) was significantly associated with differences in birth weight. Shorter gestation, drinking during pregnancy and twin gestation were associated with shorter birth length (table 5). Maternal and offspring ADH2 genotype and the interactions between each genotype and drinking, as well as all other maternal and infant variables, were not significant (P > .1). Shorter gestation and the absence of a maternal ADH2*3 allele were both associated with smaller head size at birth (P < .001 and .02, respectively). The impact of drinking during pregnancy on head circumference was marginal (P = .09).

View larger version (40K): [in this window] [in a new window]   Fig. 3.   Effect of alcohol intake during pregnancy on the growth parameters of newborns delivered to abstaining women (open bars), women with average daily absolute ethanol intakes from 0.01 to 0.49 oz (hatched bars) and those with intakes 0.5 oz (filled bars). Data shown as mean ± S.D.

	Discussion

Abstract Introduction Materials & Methods Results Discussion References

Our observation of an association between maternal and offspring ADH genotype and offspring outcome after intrauterine alcohol exposure is the first documentation of a specific genetic determinant of the risk for alcohol-related birth defects. Our hypothesis of such a genetic risk factor was fueled by observations of discordant outcomes in dizygotic twins (Christoffel and Salafsky, 1975; Crain et al., 1983; Riese, 1989; Streissguth and Dehaene, 1993). Further evidence supporting such a determinant in the African American population included the increased risk of alcohol-related birth defects in this population (Sokol et al., 1986) and, unique to this ethnic group, the presence of an ADH allele encoding an enzyme expected to result in markedly different ethanol elimination (for a review, see Bosron and Li, 1987). The only other risk assessment of a specific genetic polymorphism examined the ALDH2 mutation and the "atypical ADH" (ADH2*2), an isoform present in most Orientals and some Caucasians (Faustman et al., 1992). This genotypic analysis of 24 mother/offspring pairs was unable to verify or deny any association between these genotype variations and alcohol-related birth defects.

In contrast to the lack of study of alcohol-related birth defects, many studies have tested correlations among genetic differences in the ethanol-metabolizing enzymes, ethanol intake and susceptibility to alcohol-related liver disease. The most clear-cut demonstration of such a genetic association is that of the ALDH2 polymorphism. A point mutation in the ALDH2 gene present in Oriental populations produces a deficiency in the low-K_m, mitochondrial ALDH enzyme that results in accumulation of acetaldehyde during drinking. This mutation has been well linked to flushing after alcohol intake, lower alcohol intakes and a lower incidence of alcoholism (Maezawa et al., 1995; Shibuya and Yoshida, 1988; Tanaka et al., 1996; Thomasson et al., 1991). However, this ALDH variant does not occur in Caucasian or African American populations. Thus, this polymorphism does not account for alcohol-related susceptibility differences between or within these two populations and therefore was not evaluated in our study. The influence of the allele we studied, the ADH2*3 allele, on ethanol intake, risk of alcoholism and alcohol-related liver disease has not been reported. Unlike the apparently intake-related adverse effect of the ALDH2 mutation, the mechanism underlying our observation of altered ethanol susceptibility associated with the ADH2*3 allele was not related to ethanol intake. By design, recruitment was stratified prospectively to ensure a range of intake within each genotype. The resulting distribution of alcohol intake across genotype groups was consistent with this design. Thus, whether the presence of the ADH2*3 allele is associated with differences in drinking behavior cannot be tested with these data.

Although the ADH2*2 and ADH2*3 alleles generate very different enzymes and are present in different ethnic groups, the direction of the effect of these mutations on alcohol-related disease would be expected to be similar because both alleles are associated with an increased ethanol elimination rate in vitro (Ehrig et al., 1990). Conflicting results have been obtained regarding the effect of the presence of the ADH2*2 allele on ethanol-related risk. Although an increased incidence of the ADH2*2 allele correlated with lower drinking behavior in some reports (Muramatsu et al., 1995; Tanaka et al., 1996; Thomasson et al., 1991; Yamauchi et al., 1995b), others have reported either no relationship to intake (Vidal et al., 1993) or an increased incidence of the allele among alcoholics (Tanaka et al., 1996; Yamauchi et al., 1995a). Similarly, a reported association between this allele and alcoholic liver disease (Pares et al., 1994; Yamauchi et al., 1995a, 1995b) has been controversial (Couzigou et al., 1990; Shibuya and Yoshida, 1988; Vidal et al., 1993). The effect of the ADH3 polymorphism, a locus whose alleles determine enzymes with similar kinetic constants, on alcohol-related disease is unclear. Some authors report an association between genotypic differences and alcoholism (Thomasson et al., 1991), whereas others suggest no effect (Couzigou et al., 1990; Pares et al., 1994; Poupon et al., 1992). A multivariate analysis of genetic polymorphisms in ALDH2, ADH2, ADH3 and CYP2E1 demonstrated a positive correlation between alcoholic cirrhosis and the presence of the ADH2*2 allele but no effect from the ALDH2 or ADH3 polymorphisms (Yamauchi et al., 1995a).

Our results have implications for variable offspring susceptibility to intrauterine ethanol exposure within the African American population but do not address the mechanism of the increased incidence of alcohol-related birth defects in African Americans compared with Caucasians (Centers for Disease Control, 1993). Because the ADH2*3 allele is unique to African Americans (Bosron et al., 1983b; Bosron and Li, 1987), a contribution of the ADH2 genetic polymorphism to the differences in vulnerability between Caucasians and African Americans would have yielded poorer offspring outcome in the presence of the ADH2*3 allele.

We speculate that the mechanism for our observation is the difference in the metabolism of alcohol by isoenzymes encoded by these alleles. The isoenzyme encoded by the ADH2*3 allele (ADH beta-3 beta-3) has a lower enzymatic efficiency than that encoded by ADH2*1 (ADH beta-1 beta-1), if compared on the basis of V_max/K_m ratios (8.8 and 187.8, respectively). However, it is unlikely that a lower catalytic efficiency is protective. Binge drinking has been associated with an increased risk of adverse offspring outcome (Bonthius et al., 1988; Goodlett et al., 1989; West et al., 1990). At the blood alcohol concentrations expected with binge drinking (20-40 mM), ADH beta-1 beta-1 would be saturated (K_m = 0.049 mM) with a velocity approaching its V_max value (9.2 min¹) (Bosron and Li, 1987). ADH beta-3 beta-3, on the other hand, would not be saturated (K_m = 34 mM) but would have a velocity far in excess of that of beta-1 beta-1 isoenzyme (beta-3 beta-3 V_max = 300 min¹ Burnell et al., 1989). Thus, the availability of the low-affinity, high-capacity beta-3 beta-3 form encoded by the ADH2*3 allele would increase alcohol metabolism when high blood alcohol concentrations occur. Based on the in vitro kinetic constants of the beta beta isoenzymes, individuals with the ADH2*3 alleles would be expected to have significantly higher ethanol elimination rates. Recently, Thomasson et al. (1995) confirmed this predicted result in an in vivo study of ethanol metabolism in 112 African Americans. Individuals with at least one ADH2*3 allele had more rapid ethanol elimination than those who were homozygous ADH2*1. Interestingly, African Americans had slower rates of ethanol elimination than Caucasians. Taken together, the observation of better offspring outcome in the presence of the ADH2*3 allele and the association between the allele and more rapid ethanol elimination suggest that ethanol rather than acetaldehyde is the more important proximate teratogen.

An alternative mechanism for the effect of the interaction between ethanol intake and ADH genotype might be an ethanol-associated alteration in the ADH-mediated metabolism of an endogenous substrate. ADH catalyzes a number of steps in the metabolism of several neurotransmitters. Class I ADH oxidizes norepinephrine glycols (Mardh et al., 1985) and can catalyze the interconversion of the intermediary alcohols and aldehydes of dopamine metabolism (Mardh and Vallee, 1986). The dopamine intermediate reactions are inhibited by 4-methylpyrazole (Kassam et al., 1989) and appear to occur at the same active site as ethanol oxidation, and the presence of ethanol shifts dopamine metabolism toward the formation of reduced metabolites (Mardh and Vallee, 1986). However, the relative rate of oxidation is lower than that of ethanol. ADH also plays a minor role in the metabolism of serotonin; however, ADH2*1 and ADH2*2 allelic differences are not associated with differences in serotonin metabolism (Helander et al., 1994). Class I ADH enzymes are capable of metabolizing cytotoxic intermediates in the peroxidation of polyunsaturated fatty acids (aliphatic 4-hydroxyalkenals) (Boleda et al., 1993; Mitchell and Petersen, 1987; Sellin et al., 1991), and this reaction is competitively inhibited by ethanol (Sellin et al., 1991). Hypotheses regarding the importance of variation in ADH-mediated metabolism of these endogenous substrates are less attractive as a mechanistic explanation of better offspring outcome in view of the local nature of these mechanisms and the limited distribution of ADH expression within the central nervous system.

Class I ADH catalyzes retinol oxidation to retinal, which is the rate-limiting step in the conversion to retinoic acid that controls gene expression at the transcriptional level and appears to be involved in differentiation of epithelial cells (Ang et al., 1996). Multiple authors have proposed ethanol-mediated alterations in retinoic acid metabolism as a mechanism in fetal alcohol syndrome (Duester, 1991; Grummer and Zachman, 1990; Pullarkat, 1991). Based on evidence supporting this hypothesis (Grummer et al., 1993), one might suspect that our observation of a significant interaction between ADH2 genotype and ethanol intake would be retinoic acid related. However, despite the variation in retinoic acid catalytic activity from the polymorphism at the ADH2 locus, the relative inefficiencies of the beta alloenzymes compared with other ADH classes refutes this idea as an explanation of our findings (Yang et al., 1994). The two most efficient ADH isoforms for retinol oxidation, the class II and class IV ADH, have poor affinity for ethanol and would not be expected to be sensitive to ethanol inhibition. The polymorphism at the ADH3 locus is an attractive candidate for a retinoid-mediated mechanism for ethanol toxicity because this gene contains a retinoic acid response element (Duester et al., 1991). Although transcriptional activity of the ADH3 promoter has been demonstrated in a transgenic mouse embryo model (Zgombic-Knight et al., 1994), the expression of the gamma isoenzymes encoded at this locus has been demonstrated postnatally only in humans (Smith et al., 1973, 1971).

The significance of the ADH2 offspring genotype on outcome was somewhat surprising because the beta isoenzyme does not appear until about 25 weeks' gestation (Smith et al., 1971), apparently in response to specific transcription factors (Van Ooij et al., 1992). The ADH2 promoter has been shown to respond to CCAAT/enhancer binding protein-alpha, a transcription factor particularly active during late fetal and early postnatal liver development (Stewart et al., 1991). These molecular observations are consistent with earlier studies showing that early human fetal ADH activity is less than one-10th of adult activity (Pikkarainen and Raiha, 1967). Although ADH activity increases during gestation, maturation is not complete until early childhood (Pikkarainen and Raiha, 1967). Biologically, offspring genotype is not independent of maternal genotype; therefore, the effect of offspring ADH2 genotype cannot be totally separated statistically from that of maternal genotype. However, because ethanol crosses the placenta (Dilts, 1970), the impact of offspring metabolism and its determinants are relevant, and offspring genotype was included in the analysis.

Our results are the first specific support for genetic risk for alcohol-related birth defects after alcohol use in pregnancy. The presence of the ADH2*3 allele was protective for both offspring neurobehavioral outcome and intrauterine growth. Continued delineation of important risk factors, both genetic and environmental, will be important to target high-risk populations for public health prevention efforts. Ongoing studies of postpartum women are evaluating the direct role of maternal intersubject variation in ethanol and acetaldehyde metabolic ability as risk factors for alcohol-related birth defects among women who drink during pregnancy.