5-HTTLPR X Stress in Adolescent Depression: Moderation by MAOA and Gender
- First Online:
- Cite this article as:
- Priess-Groben, H.A. & Hyde, J.S. J Abnorm Child Psychol (2013) 41: 281. doi:10.1007/s10802-012-9672-1
Depression surges in adolescence, especially among girls. Most evidence indicates that the short allele of a polymorphism in the promoter region of the serotonin transporter gene (5-HTTLPR) interacts with stress to influence the onset of depression. This effect appears to be less robust in adolescents, particularly among boys, and may be moderated by other genetic polymorphisms. Seeking to explain the adolescent gender difference in depression, this study examined the effects of 5-HTTLPR (rs25531), the monoamine oxidase A-upstream variable number tandem repeat (MAOA-uVNTR), and negative life events (NLE). A community-based longitudinal sample of 309 adolescents reported depressive symptoms and NLE at ages 11, 13, and 15. 5-HTTLPR and MAOA-uVNTR genotypes were ascertained via buccal swabs. A significant four-way interaction of 5-HTTLPR, MAOA-uVNTR, NLE at age 13, and gender predicted depressive symptoms at age 15. Girls were most likely to exhibit elevated depressive symptoms when experiencing NLE if they possessed low-expression MAOA-uVNTR alleles and short 5-HTTLPR alleles, whereas low-expression MAOA-uVNTR alleles but long 5-HTTLPR alleles were implicated in boys. The results indicate that the commonly reported 5-HTTLPR by stress interaction for depression may be limited to individuals with low-expression MAOA-uVNTR alleles. These data also provide new evidence that the short allele of 5-HTTLPR confers susceptibility to stress differently for females compared with males.
KeywordsDepressionAdolescent developmentGender differencesBehavioral genetics
The World Health Organization has identified depression as a major cause of disease-related disability worldwide and the number one cause of disability among women (Kessler 2003). Notably, in childhood the incidence of depression is relatively low and typically does not exhibit a gender difference. With the onset of adolescence, however, depressive symptoms and diagnoses increase dramatically, particularly for girls (Angold 1988; Rutter et al. 1976). Indeed, one of the most consistent findings in depression research is that women are twice as likely to be depressed as men (Kessler et al. 1993; Nolen-Hoeksema 1987; Weissman and Klerman 1977), a difference found cross-culturally in both community and clinical samples, which cannot be explained by gender differences in help-seeking behavior (Nolen-Hoeksema 1987). Longitudinal studies indicate that this gender difference appears at about ages 13–15 (Hankin et al. 1998). Thus, adolescence is a crucial time to study the development of depression and the emergence of a gender difference in depression.
Numerous hypotheses have been advanced to explain depression in adolescence and the emergence of the gender difference in depression (Hyde et al. 2008). One promising approach is gene by environment interaction studies, and in particular, how variation in a polymorphism in the serotonin transporter gene (5-HTTLPR) may confer vulnerability to stress (Caspi et al. 2003). However, there have been several failures to replicate this effect (Risch et al. 2009), perhaps because the effect of 5-HTTLPR on depression may be moderated not only by stress but also by gender and by other genetic variants. Thus, we believe, more complex models must be tested. The current study drew upon longitudinal data to examine the development of depressive symptoms in adolescents. We adopted a theoretically guided approach to selecting the genes analyzed here, choosing variants that may influence serotonin levels given the important role of serotonin in depression and that have been previously shown to contribute independently and interactively to depression and other forms of psychopathology (discussed below). Specifically, we tested whether 5-HTTLPR interacted with negative life events, gender, and the monoamine oxidase A-upstream variable number tandem repeat (MAOA-uVNTR) to predict adolescent depressive symptoms.
Serotonin Transporter Polymorphism
In research concerned with the etiology of depression, much attention has been devoted to genes associated with the monoamine neurotransmitter serotonin. Serotonin, particularly in low levels, has been linked to depression based on varied evidence, including low levels of serotonin byproduct in the brains or cerebrospinal fluid of people with depression and victims of suicide, the effectiveness of antidepressant medications that increase serotonin, and the ability to induce depressive symptoms in a laboratory setting by decreasing serotonin itself or by reducing tryptophan, the amino acid used to synthesize serotonin (Elhwuegi 2004; Veenstra-VanderWeele et al. 2000). There are dozens of genes directly related to serotonin synthesis and neurotransmission, although one gene, the serotonin transporter (5-HTT, or SLC6A4) has captured the most attention, in part because the serotonin transporter is the target of one of the most effective classes of antidepressant medications, the selective serotonin reuptake inhibitors (SSRIs). Serotonin reuptake represents one of the primary ways in which serotonin neurotransmission is terminated, through the reuptake of serotonin into the presynaptic neuron. The 5-HTT gene codes for the protein responsible for reuptake and thus may play an important role in determining the amount of serotonin available in the synaptic cleft (Ramamoorthy et al. 1993). The hypothesis, then, is that genetic variation in 5-HTT may lead to variation in individual susceptibility to depression given the gene’s role in influencing serotonin levels.
There are many known genetic polymorphisms in 5-HTT, but the most promising one to date is the serotonin transporter-linked polymorphic region (5-HTTLPR). This polymorphism has two common alleles, the so-called long and short alleles, differing by the presence or absence of a 44-base pair insertion/deletion. The short allele reduces the efficiency with which the 5-HTT gene is transcribed (Heils et al. 1996; Lesch et al. 1996) and has been associated with depression, particularly in interaction with stress, as evidenced by Caspi and colleagues’ (2003) groundbreaking study, which found that the short allele did not exert a main effect on depressive symptoms and episodes, but instead interacted with stress, such that carriers of the short allele were more vulnerable to the effects of stress on depression.
Since the publication of the Caspi paper, 5-HTTLPR by stress interactions on depression have been tested in over 50 studies and subjected to several meta-analyses. One highly publicized meta-analysis did not find support for the short allele as conferring vulnerability to depression under stressful conditions (Risch et al. 2009), whereas a subsequent and more comprehensive meta-analysis that included all child, adolescent, and adult studies to date upheld the findings of Caspi and colleagues (Karg et al. 2011). More recent findings regarding the role of the short allele in child and adolescent depression have also been mixed. Several methodologically sound studies that assessed stress broadly and with methods such as carefully designed stress questionnaires, contextual stress interviews, or maternal reports of stress, have found support for 5-HTTLPR X stress interactions in the prediction of adolescent depression (e.g., Hankin et al. 2011a; Jenness et al. 2011; Petersen et al. 2012). In contrast, studies that examined specific forms of stress, such as socioeconomic stress or lack of parental support, have failed to find that these stressors interacted with 5-HTTLPR to predict adolescent depression (e.g., Uddin et al. 2011; van Roekel et al. 2011).
One reason for the discrepancies across individual studies and meta-analyses may be that the effects of 5-HTTLPR on depression are heterogeneous and may depend on specific characteristics of the population under investigation, most notably age and gender (Uher and McGuffin 2008). For instance, Uher and McGuffin (2008) noted that in studies of adolescents, the short allele has been found to interact with stress more often in girls than in boys, whereas the long allele has been implicated in boys more than girls (e.g., Eley et al. 2004; Sjöberg et al. 2006). Furthermore, although most 5-HTTLPR by stress interaction studies examine adult depression, we found that two of the five studies included in the Karg et al. (2011) meta-analysis that implicated the long allele were studies exclusively of adolescents, one study examined adolescents and very young adults (age range 13–23 years), and a fourth study assessed stress in adolescence (but depression in adulthood). Thus, although the short allele confers vulnerability to depression in most samples studied, this effect may be moderated by age and gender.
Monoamine Oxidase A
Serotonin reuptake represents one mechanism whereby serotonin transmission is terminated. Another important mechanism of serotonin clearance from the synaptic cleft is through serotonin degradation. Serotonin (and other monoamine neurotransmitters) are degraded by the monoamine oxidase type A enzyme, encoded by the monoamine oxidase A (MAOA) gene, which reduces available serotonin. Thus, variation in MAOA may play a role in the development of depression, just as the serotonin transporter does, by affecting serotonin levels.
One variant of the MAOA gene, in particular, has been associated with various forms of psychopathology. MAOA-uVNTR is a 30-base pair repeat polymorphism in the promoter region of the MAOA gene, with variants including 2R, 3R, 3.5R, 4R, 4.5R, or 5R copies of the repeat (Sabol et al. 1998). The most common forms, 3R, 3.5R, and 4R, have been found to differentially affect the efficiency with which the MAOA gene is transcribed (Deckert et al. 1999; Denney et al. 1999; Sabol et al. 1998). The 3.5R and 4R alleles are associated with increased transcriptional efficiency (and presumably more MAOA enzyme, which would lead to increased serotonin degradation) compared to the 3R allele. Therefore, the 3.5R and 4R alleles are often grouped together when analyzed and are termed “high-expression,” whereas the 3R allele is termed “low-expression.” Given the lack of association between MAOA-uVNTR genotype and MAOA enzyme activity in the adult brain, but evidence that MAOA-uVNTR genotype interacts with early adversity to predict problem behaviors in humans and other animals, some researchers have speculated that the effects of MAOA-uVNTR on monoamine oxidase A levels may be stronger in prenatal development than in adulthood (Åslund et al. 2011; Fowler et al. 2007). The role of MAOA-uVNTR on the adolescent brain is unknown.
Strong support exists for the association of the low-expression MAOA-uVNTR allele with antisocial behavior, and to a lesser extent, other problem behaviors such as attention-deficit hyperactivity disorder, when the gene is considered in interaction with childhood maltreatment and other forms of stress (for reviews, see Kim-Cohen et al. 2006; Taylor and Kim-Cohen 2007). An association of MAOA-uVNTR with depression might be expected given the importance of serotonin and other neurotransmitters in the development of depression. However, tests of this association have been inconclusive, with numerous findings implicating the high-expression alleles (Gutiérrez et al. 2004; Rivera et al. 2009; Schulze et al. 2000; Yu et al. 2005) or the low-expression alleles (Brummett et al. 2007; Cicchetti et al. 2007; Doornbos et al. 2009), whereas at least a dozen studies have failed to find an association (e.g., Eley et al. 2004). One reason for the inconsistency in findings may be that results appear to depend on whether researchers included a stress interaction in their analyses. A meta-analysis of studies that omitted effects of stress implicated high-expression alleles in depression for at least some populations (Fan et al. 2010). In contrast, most MAOA-uVNTR by stress interaction papers have found that low-expression alleles interacted with stress to affect depression. Furthermore, to our knowledge only three studies have examined adolescents, but with different findings. Eley and colleagues (2004) found a main effect “trend” in boys implicating the high-expression allele but no effect in girls, whereas Marmorstein and Hart (2011) found that the low-expression allele interacted with one form of social stress, receipt of public assistance, to predict depressive symptoms in their male-only study. In contrast, Cicchetti and colleagues (2007) found that the low-expression allele interacted with 5-HTTLPR and one form of stress (out of several assessed) to predict depression for both genders. Specifically, the latter study found that adolescents who had experienced sexual abuse by a caregiver were most likely to develop depression if they possessed two short 5-HTTLPR alleles and the low-expression MAOA-uVNTR allele. However, this is the only study to date that has considered the interaction of 5-HTTLPR and MAOA-uVNTR on depression, despite the possibility that these genes may influence one another interactively given the similar roles of the 5-HTT and MAOA genes in serotonin clearance (5-HTT through reuptake; MAOA through degradation). Experimental evidence involving mice provides further support for interactions between these genes, as these genes appear to interactively influence serotonin levels and serotonin metabolism in mice (Murphy et al. 2003). Finally, 5-HTTLPR and MAOA-uVNTR have been shown to interact in their prediction of other forms of psychopathology, including anorexia nervosa (Urwin and Nunn 2005) and binge drinking (Herman et al. 2005). Therefore, the association of MAOA-uVNTR with depression is unclear yet promising, and is in need of further examination.
Gender Differences in Pathways to Depression
As noted above, adolescence is a time when depression surges, particularly for girls. One important question pertaining to the gender difference in depression is whether there are different pathways to depression for girls and boys, or if pathways are similar but with stronger effects for one gender than the other, as evidenced by gender interactions. There are several ways in which gender may moderate the effects of genes on depression, all of which have been reported in the literature (Hyde et al. 2008).
First, the gene-depression association or gene by stress interaction on depression may differ for adolescent girls and boys. With regard to 5-HTTLPR, some studies have shown that the short allele confers vulnerability to stress among adolescent girls but not boys (Eley et al. 2004; Hammen et al. 2010; Sjöberg et al. 2006; Åslund et al. 2009) or among women but not men (Brummett et al. 2008). Although some of these studies reported that 5-HTTLPR was unrelated to male depression, two studies (Sjöberg et al. 2006; Brummett et al. 2008) found that the long allele conferred vulnerability to stress in boys and men. Indeed, although most studies support the association of the short allele with depression, adolescent boys are overrepresented in studies that implicate the long allele as the one that confers vulnerability to stress (Sjöberg et al. 2006; Laucht et al. 2009). Thus, the effect of the short allele on depression may be less robust in adolescent boys than in other populations. Variation in the effects of 5-HTTLPR across gender may in turn explain some of the inconsistencies in research findings on the effects of 5-HTTLPR.
The effect of MAOA-uVNTR may also be moderated by gender. Numerous studies have reported that a main effect of MAOA-uVNTR on depression is specific to women (e.g., Rivera et al. 2009; Schulze et al. 2000; Yu et al. 2005). Adult studies that tested MAOA-uVNTR by stress interactions did not examine whether effects were moderated by gender. Only one study has examined this question among adolescents and found that MAOA-uVNTR marginally significantly predicted depression among boys but did not interact with stress for boys or girls (Eley et al. 2004). However, MAOA-uVNTR has been reported to exert effects of opposite direction on externalizing problems among adolescent girls and boys with a history of maltreatment (Åslund et al. 2011), and an interaction of 5-HTTLPR and MAOA-uVNTR on binge drinking was significant only for girls (Herman et al. 2005), suggesting that for at least some types of adolescent psychopathology, the effects of MAOA-uVNTR may vary by gender. Thus, gender differences in the association of MAOA-uVNTR with depression may exist but are unclear and understudied, particularly with regard to MAOA-uVNTR by stress interactions.
Finally, gender differences in gene-depression associations may be due to certain stressors being more depressogenic for one gender than the other. For instance, Sjöberg and colleagues (2006) reported that 5-HTTLPR interacted with family conflict only in girls but with parental separation only in boys, highlighting how the genders were vulnerable to different types of stress. Thus, the literature suggests that gene by stress interactions may differ for adolescent girls and boys due to the ways in which genetic variants or particular stressors affect girls and boys differently. The reason for gender differences in the association between genetic variants and depression is not yet understood, however. Possible explanations include gender differences in hormonal and other biological factors that influence gene expression (Ellegren and Parsch 2007; Gundlah et al. 2002), neural development and physiology (Bethea et al. 1998; Meyer-Lindenberg et al. 2006), and serotonergic function (Benmansour et al. 2012). More recent research has begun to identify gene by hormone interactions (e.g., Josephs et al. 2011 reported 5-HTTLPR X testosterone interactions predicting cortisol responses to threatening stimuli), which may further elucidate how hormones influence gene-depression associations. Explanations for gender differences in genetic susceptibility to depression remain an important direction for future research.
The Current Study
Although gene by environment interactions have been studied extensively for depression, gene by gene by stress interaction studies are rare. Almost certainly depression is a polygenic trait, like height (Yang et al. 2010). In the next generation of gene by environment interaction studies, it will be crucial to study two or more genes, chosen in a theory-guided manner. We have done that in this study by choosing two genes that are meaningfully linked in the serotonin system and that have been shown to interact in experimental animal research (Murphy et al. 2003). Moreover, insufficient attention has been given, in past research, to variations in gene action depending on age and gender. We address these issues as well.
The current study drew upon data from the Wisconsin Study of Families and Work (WSFW), a longitudinal study of youth and their families, to test developmental models of depressive symptoms and gender differences in the development of depressive symptoms. This sample has demonstrated the expected emergence of a gender difference in depressive symptoms in adolescence, as reported elsewhere (Priess et al. 2009). With the goal of specifying and testing a more complex model of genetic effects, we examined whether the effect of 5-HTTLPR on depressive symptoms was moderated by MAOA-uVNTR, negative life events, and gender.
Participants were 309 adolescents (129 female) from the longitudinal Wisconsin Study of Families and Work (formerly named the Wisconsin Maternity Leave and Health Project; see Hyde et al. 1995). Participants were originally recruited from the Madison and Milwaukee, Wisconsin areas when women were pregnant with the adolescents who are the focus of this study. Participants currently reside in a range of communities, including a large Midwestern city, a small Midwestern city, several small towns, and rural areas. Participants have completed 13 waves of data collection: at the fifth month of pregnancy, 1, 4, 12, and 24 months postpartum and when children were 3.5, 4.5, 6, 7, 9, 11, 13, and 15 years of age.
The 309 adolescents included in the present study were participants drawn from the larger Wisconsin Study of Families and Work sample who met several criteria for inclusion. First, participants in the present study completed at least one of the three adolescent assessment waves (at ages 11, 13 and 15), which is 74.8 % of the families who participated when the children were 1 month of age (N = 232 at age 11, 283 at age 13, and 250 at age 15; the lower participation at age 11 reflects a funding shortage that kept us from collecting data with all potential participants during that assessment wave). Participants who remained in the study at adolescence did not differ from those who discontinued participation prior to adolescence in terms of previously assessed race/ethnicity, family income, or parents’ depressive symptoms. Likewise, those who participated in a given assessment during adolescence did not differ significantly from those who had participated in the previous but not current assessment in terms of race/ethnicity, family income, adolescents’ depressive symptoms or NLE, or parents’ depressive symptoms.
Participating families in the Wisconsin Study of Families and Work have diverse socioeconomic backgrounds that are representative of the United States. Average annual family income at the beginning of the study (1990–91), at which time children’s parents were married or living together (a condition of initial recruitment), was $49,500 (median = $45,000, range $2,000–$200,000). In 1991, the median income of married couple families in the United States with the wife in the paid labor force was $48,169 (U. S. Department of Labor 1993). Therefore, the sample matches the national figure well. The mothers’ occupations were coded using the Bose index of occupational status (Bose 1985), which ranges from 10 to 100 (physician = 100, registered nurse = 65, bookkeeper = 43, private household cleaner = 16). The mean for our sample was 52.7 (range 25–100); the national mean for women based on the census of occupations is 50.3 (Bose 1985, p. 193). Again, our sample matches national statistics well.
Among those participants who remained in the study at adolescence, 89.3 % were White, 1.7 % African American, 1.7 % Hispanic, 1.2 % American Indian/Alaskan, and 6.15 % biracial or multiracial. Participants are ethnically representative of the communities from which they were recruited. However, given concerns with population stratification in genetic studies (Lander and Schork 1994), inclusion in the present study was limited to adolescents whose parents both identified as White and non-Hispanic when first recruited to the study. Participants identified as White/non-Hispanic reported depressive symptoms and episodes that were similar to participants of other ethnicities but came from families with slightly but significantly higher incomes and education levels.
Given unknown patterns of X-inactivation in girls who were heterozygous for the MAOA-uVNTR low- and high-expression alleles, and thus unknown MAOA function in these girls (see Åslund et al. 2011 for discussion), analyses were also limited to girls who were homozygous for gene function (that is, girls with 3R/3R, 3.5R/3.5R, 3.5R/4R, and 4R/4R genotypes and all boys were included, whereas 48 girls with 3R/3.5R and 3R/4R were excluded), as is common in studies involving MAOA genetic variants. Because MAOA is X-linked, boys only had one copy of this allele, and therefore all boys could be included in analyses. The resulting sample size after these criteria for inclusion were imposed was 309.
Adolescent participants reported depressive symptoms at ages 11 and 15 and negative life events at age 13. These reports were provided on laptop computers during visits to the adolescents’ homes. Children and adolescents are more likely to provide accurate reports of sensitive information when this information is assessed with computers rather traditional forms of questionnaire administration, such as paper-and-pencil tests (Turner et al. 1998). DNA samples were collected via buccal swabs during the age 15 assessment and stored at −20 °C prior to genotyping.
Depressive symptoms were assessed with the Children’s Depression Inventory (CDI; Kovacs 1981). The CDI was designed to be used with children and consists of 27 items, each with three statements of increasing severity, to assess affective, behavioral, and cognitive symptoms of depression within the past 2 weeks. For each item, adolescents were asked to pick which of three statements best described them in the past 2 weeks, for instance, by choosing “I was sad once in a while,” “I was sad many times,” or “I was sad all the time.” The CDI has demonstrated good internal consistency (typically 0.71–0.89) and adequate test-retest reliability (typically 0.72–0.87) (e.g., see Kovacs 1981; Saylor et al. 1984; Smucker et al. 1986). In the present study, adolescents completed 24 of 27 items (the three items that asked for information about school were omitted since data collection occurred during the summer; scores were scaled to be comparable to the 27-item version). Internal consistency was α = 0.79 at age 11, 0.83 at age 13, and 0.86 at age 15. A continuous, symptom-based measure was used based on evidence that adolescent depression is best viewed as a continuum (Hankin et al. 2005; Ruscio and Ruscio 2000). Furthermore, compared with categorical diagnostic measures, symptom-based assessments may have greater sensitivity to identify depressed individuals and have increased statistical power to detect gene by stress interactions, which may be why a recent meta-analysis found that the interaction of 5-HTTLPR with stress was stronger for continuous measures of depression than for categorical measures (Karg et al. 2011). Finally, adolescents with elevated depressive symptoms tend to exhibit impairment in numerous domains, regardless of whether they meet criteria for a depressive episode (Gotlib et al. 1995).
Negative Life Events
The Adolescent Perceived Events Scale (APES; Compas et al. 1987) measured negative life events that occurred in adolescents’ lives over the past year (proximal to ages 11, 13, and 15). Using a shortened form of the APES, adolescents reported which of 59 events they had experienced in the preceding year and rated each experience on a scale ranging from −4 (extremely bad) to 0 (neither good nor bad) to +4 (extremely good). In the current study, negative life events refer to the sum of the number of events rated negatively. The questions assessed NLE that ranged greatly in severity and covered domains that have been found to be important to adolescents, such as family, friends, school, and romantic relationships. The most commonly reported NLE in this study included “getting punished by parents” (N = 167), “doing poorly on an exam or paper” (N = 149), and “fight with or problem with friend” (N = 125). Examples of more severe but less frequently experienced NLE include “death of a family member” (N = 62), “arrest of a family member” (N = 14), “family move” (N = 12), and “parents getting divorced” (N = 3).
Several studies have found the APES to be a valid measure of adolescent negative life events. The measure was designed in consultation with 658 adolescents to assess all domains of life that adolescents report as stressful. Adolescent responses on the APES are corroborated to a high degree by other people familiar with the adolescents’ experiences (Compas et al. 1987). In addition, APES scores are predictive of physical and mental health outcomes (Caputo et al. 1998; Coleman et al. 1998). The APES exhibits good test-retest reliability over 2 weeks (0.81–0.84; Compas et al. 1987). Measures of internal consistency are not meaningful for the APES because occurrences of significant life events are largely independent.
DNA samples were collected from adolescents via buccal swabs. 5-HTTLPR/rs25531 reactions were performed using a modified form of the procedure of Wendland et al. (2006). MAOA-uVNTR reactions were performed using a modified form of the procedure of Nilsson et al. (2008). For both polymorphisms, the forward primer was fluorescently labeled with 5′FAM; for 5′HTTLPR, the reverse primer was also labeled with 5′HEX. Primers used were 5′ FAM-TCC TCC GCT TTG GCG CCT CTT CC (5-HTTLPR/rs25531 forward), 5′ HEX-TGG GGG TTG CAG GGG AGA TCC TG (5-HTTLPR/rs25531 reverse), 5′ FAM-ACA GCC TGA CCG TGG AGA AG (MAOA-uVNTR forward), and 5′ GAA CGG ACG CTC CAT TCG GA (MAOA-uVNTR reverse).
Given evidence that functionality of 5-HTTLPR long alleles varies by rs25531, a single nucleotide polymorphism within 5-HTTLPR, LG alleles were reclassified as short alleles (Hu et al. 2006; Kraft et al. 2005). With regard to MAOA-uVNTR, 3-repeat alleles were classified as low-expression, and 3.5- and 4-repeat alleles were classified as high-expression, as suggested by research on the alleles′ functionality (Deckert et al. 1999; Sabol et al. 1998). One individual with a 4.5/5 genotype was excluded from analyses given the rarity of these alleles and the uncertainty regarding their functionality. No other alleles were detected. The distributions resulting from these classifications were consistent with Hardy-Weinberg equilibrium.
Statistical analyses were conducted in Mplus (Muthén and Muthén 2010) with multiple regression, using full information maximum likelihood estimation with robust standard errors to account for the naturally skewed distribution of depressive symptoms. Mplus was chosen because it offers a solution to handle missing values using an approach proposed by Arbuckle (1996), in which separate means and covariances are calculated for each group of participants who have the same pattern of data completion and then combined to predict the log likelihood for the full sample using full information maximum likelihood. This technique is particularly useful given the common problem of missing values in longitudinal studies, and allowed data to be used from all participants who were included in the present study, even if some of them had missing data. Interaction terms were created by multiplying the factors (genotype, NLE, gender) from which they were composed.
Three adolescents were taking antidepressant medications at the time depressive symptoms were assessed, and four others had used these medications in the past. However, results were unchanged if those adolescents with current or lifetime antidepressant usage were omitted from analyses. Therefore, results are presented for the full sample.
Descriptive statistics: depressive symptoms, negative life events, and peer sexual harassment victimization, by gender
Depr Symp—age 11
Depr Symp—age 15*
Neg Life Ev—age 13
Depressive Symptoms as Predicted by Genes, Negative Life Events, and Gender
Regression coefficients and variance explained for depressive symptoms regressed onto previous symptoms, gender, 5-HTTLPR and MAOA-uVNTR genotype, age 13 negative life events, and associated interactions
Previous Depr. Symptoms
Negative Life Events (NLE)
5-HTTLPR X MAOA
5-HTTLPR X NLE
5-HTTLPR X Gender
MAOA X NLE
MAOA X Gender
NLE X Gender
5-HTTLPR X MAOA X NLE
5-HTTLPR X MAOA X Gender
5-HTTLPR X NLE X Gender
MAOA X NLE X Gender
5-HTTLPR X MAOA X NLE X Gender
R2 = 0.420
F(16,125) = 5.600, p < 0.001
The finding of a four-way interaction of 5-HTTLPR, MAOA-uVNTR, negative life events, and gender on depressive symptoms in adolescents may provide an important explanation for the sometimes inconsistent results of studies of the interaction of 5-HTTLPR and stress (Karg et al. 2011; Risch et al. 2009). The results of the current study show that the short allele of 5-HTTLPR confers vulnerability to stress only for adolescent girls who carry the low-expression allele of MAOA-uVNTR. Indeed, it is the long allele of 5-HTTLPR that confers vulnerability in adolescent boys, but only in the presence of the low-expression allele of MAOA-uVNTR. These findings reflect the phenomenon of epistasis, or gene by gene interactions, and are also consistent with evidence that gene expression can vary as a function of stress level, age, and gender (Casey et al. 2009; Ellegren and Parsch 2007; Karssen et al. 2007). Inconsistencies across studies of gene by stress interactions for 5-HTTLPR may be due to varying ages of the samples, different gender composition, or failure to take into account the MAOA-uVNTR polymorphism. Notably, both girls and boys appeared vulnerable to the types of negative life events we assessed in the current study, suggesting that gender differences in the prediction of depressive symptoms were due to gender differences in genetic effects rather than effects of stress.
The finding of a gender difference in the association of 5-HTTLPR with depression is consistent with previous research in adolescents, including two studies that implicated the short allele only in girls (Eley et al. 2004; Hammen et al. 2010), and one study that produced results similar to those found in the present study, in which the short allele was implicated in girls and the long allele in boys (Sjöberg et al. 2006). As noted in the introduction, it is possible that the effects of genetic variation on depression are moderated by gender, perhaps due to the ways in which gender differences in hormone levels affect gene expression, neural development, and serotonergic function. Another explanation is that gender differences are due to ways in which the genders may experience depression differently. For instance, depression is more likely to be comorbid with externalizing disorders in boys compared with girls, which may implicate different genetic variants (although in analyses not reported here, we found the same pattern of results when we controlled for mother-reported externalizing symptoms). Furthermore, some studies of genes other than the serotonin transporter gene have suggested that which genetic variant confers vulnerability to depression may vary across the lifespan. For instance, Hilt and colleagues (2007) reported that in the brain-derived neurotrophic factor (BDNF) Val66Met polymorphism, the Val/Val genotype was associated with adolescent-onset depression, whereas the Val/Met genotype was associated with adult-onset depression. Therefore, it is possible that the long allele confers vulnerability to earlier developing depression, whereas the short allele is associated with later developing depression, as discussed below. Gender differences in genetic effects, and the role of the long allele in adolescent-onset depression, are in need of further examination in future research.
Research regarding the effects of MAOA-uVNTR on depression has yielded mixed results. When studied as a main effect, without regard for stress, high-expression alleles have been most linked to depression. In gene by stress interaction designs, however, the low-expression allele has tended to confer vulnerability to stress in predicting depression. As expected, the results of the present gene by stress study are consistent with the latter pattern of results and indicate that the low-expression MAOA-uVNTR allele may confer vulnerability to depression in individuals experiencing stressful circumstances.
A 5-HTTLPR by MAOA-uVNTR by stress interaction on depression was reported previously in a high-risk sample of adolescents (Cicchetti et al. 2007). That study found that the low-expression MAOA-uVNTR allele interacted with the short allele of 5-HTTLPR and one form of stress (the most extreme form the researchers considered—sexual abuse by a caregiver) to predict depression. The present study partially replicated these results to the extent that the 5-HTTLPR by stress interaction was found only in adolescents who carried low-expression MAOA-uVNTR alleles. However, we found that which 5-HTTLPR allele predicted depression varied by gender, thus only replicating the exact gene by gene by stress interaction in girls. One reason for this discrepancy may be because the previous study did not test for a gender interaction and therefore may have missed a gender difference in effects had one existed. Another possibility is that the different results reflect the different ages of adolescents included in the studies. The adolescents in the present study were somewhat younger, and the long allele seems more likely to be implicated in depression among younger boys. Other methodological differences may have also contributed to discrepant results, including that the current study examined a community-based, racially homogenous sample that was experiencing milder, typical forms of stress, as opposed to the racially diverse, at-risk sample used in the Cicchetti study.
The selection of 5-HTTLPR and MAOA-uVNTR as genetic variants that may confer vulnerability to depression was theoretically guided and based on evidence that these polymorphisms may influence serotonin levels in the synaptic cleft. Paradoxically, however, the variants most commonly associated with depression are the variants that would presumably increase serotonin: the short allele of 5-HTTLPR through reduced serotonin reuptake, and the low-expression allele of MAOA-uVNTR through reduced serotonin degradation. These associations are surprising given evidence that depression is characterized by low serotonin. The direction of this effect is not understood but may be due to the way in which the brain adjusts to high serotonin levels early in development, for instance, by down-regulating certain serotonin receptors (see, e.g., David et al. 2005; Mickey et al. 2008) so that serotonin transmission is relatively low despite high serotonin levels. A related possibility is that high serotonin levels early in development, such as during the prenatal period, cause the brain to develop in ways that make it more vulnerable to adversity across the lifespan; that is, serotonin may function in a neurotrophic manner when it influences brain development prenatally, but may cause different effects when functioning as a neurotransmitter later in development (Nordquist and Oreland 2010). These developmental effects may explain why MAOA-uVNTR is associated with various forms of psychopathology yet is unrelated to MAOA activity in the adult brain (Fowler et al. 2007). Whatever the exact mechanism, the association of short 5-HTTLPR and low-expression MAOA-uVNTR alleles with depression may be due to lasting effects of serotonin on neural development rather than short-term effects on serotonin transmission.
Along these lines, one possible reason why the long allele may be sometimes implicated in depression in boys is that genetic effects on depression may be somewhat delayed developmentally, such that genetic variants associated with lower serotonin levels (e.g., long and high-expression alleles in this study) may increase depression early in development given the role of low serotonin in depression, but that genetic variants associated with high serotonin levels (e.g., short and low-expression alleles) may increase depression later in life, after high serotonin levels have made the brain develop in a way that is more reactive to stress (e.g., through down-regulation of serotonin receptors). To the degree that neural development proceeds at a different rate for girls and boys, there may be gender differences in genetic effects on depression for a limited period of time. To date, most research has considered the developmental effects of serotonin on brain development only in the prenatal or early post-natal periods. Thus, the effects of genetic variants on adolescent brain development are unknown but may be important given that gene expression can vary across the life course. Investigating genetic effects on depression from a developmental viewpoint will no doubt provide important insights into the etiology of depression.
Finally, it is important to note that the effects of the short 5-HTTLPR and low-expression MAOA-uVNTR alleles on the brain and behavior may not be solely deleterious. The short 5-HTTLPR allele has been associated not only with depression and anxiety but also with enhanced cognitive and social functioning, perhaps due to hypervigilance of one’s environment (Homberg and Lesch 2011). In addition, 5-HTTLPR may be related not only to greater sensitivity to stressful situations but also to greater sensitivity to positive experiences (Belsky and Pluess 2009). For instance, among 5-HTTLPR short allele carriers, neuroticism was found to be above average in individuals who were experiencing multiple negative life events, but below average in individuals who were experiencing multiple positive life events, whereas significant life events were unrelated to neuroticism in long allele carriers, (Pluess et al. 2010). Likewise, adolescents with two 5-HTTLPR short alleles were found to have high positive affectivity in the presence of supportive parenting but low positive affectivity in the presence of unsupportive parenting, whereas parenting style had little effect on adolescents with at least one 5-HTTLPR long allele (Hankin et al. 2011b). These studies suggest that the short allele may confer sensitivity to both negative and positive environments and highlight the importance of considering the broader implications of the short 5-HTTLPR allele and low-expression MAOA-uVNTR allele on psychological functioning in future studies.
Strengths and Limitations
Strengths of the present study include its longitudinal design, which allowed us to control for pre-existing depressive symptoms at age 11 and to ensure that stress preceded depressive symptoms. Assessment of stress at age 13 and symptoms at age 15 avoided stress-depression correlations that are due to depressed individuals being more likely to recall, experience, and even generate their own stress (Hammen 1991). In contrast to most research examining genetic effects on adolescent depression, the present study consisted of a community sample of adolescents who were experiencing stressors normative to the adolescent period. Thus, the study demonstrates how the interaction of genetic polymorphisms and stress influence depressive symptoms even among community samples of adolescents. The study is also strengthened by examining 5-HTTLPR and MAOA-uVNTR concurrently, which may be important since our results and those of others show that the effect of 5-HTTLPR on depression, depressive endophenotypes, and other forms of psychopathology may be limited to certain MAOA genotypes.
Although our sample is socioeconomically diverse and community-based, in the current study we only examined participants from one ancestral group to minimize the possibility that our results arose from population stratification, and the use of this homogenous sample limits the generalizability of these findings beyond non-Hispanic Whites. There may also be additional heterogeneity in ancestry not accounted for in the study design that may attenuate or otherwise alter the findings.
The study design may also be limited by its use of measures, including the reliance on self-report from adolescents for both depressive symptoms and NLE, the adoption of the CDI, and the adoption and use of the APES. One concern regarding self-report is that the association between NLE and depressive symptoms may be biased because adolescents are the source of both pieces of information. However, although we used adolescent reports because we believe adolescents themselves are the best source of this information, we also asked parents to report adolescents’ depressive symptoms using the Child Behavior Questionnaire (CBCL; Achenbach and Rescorla 2001) and mothers to report on a subset of their adolescents’ NLE using the APES (Compas et al. 1987). All correlations between adolescent and parent reports were positive and significant, which reduces the concern that our results were simply the artifact of single-informant self-report measures. Furthermore, some researchers (e.g., Saylor et al. 1984) have criticized the CDI for providing a better measure of general distress than of depressive symptoms. Again, the correlation between CDI scores and depression items on the CBCL, as well as the expected emergent gender difference in CDI scores in our sample, the prominent use of this measure, and its excellent psychometric properties (Klein et al. 2005) support our use of the CDI as a measure of depressive symptoms, although this concern should be noted. With regard to the APES, many studies use the APES to record NLE for time periods shorter than the one-year span we used. The validity of the APES over a one-year time span has not been fully established, although adolescents and mothers in our sample provided similar reports of NLE when recollecting events over a one-year period. Our use of the APES at age 13 and CDI at age 15 does not allow us to fully account for NLE across the full adolescent period for our sample. Finally, research indicates that stronger gene-depression associations are found when stress is assessed by a clinical interview or relatively more objective documentation (such as court or medical records) than by self-report (e.g., Karg et al. 2011), and indeed, studies of adolescents that have used contextual stress interviews have found significant interactions of 5-HTTLPR with stress on depression (Hammen et al. 2010; Jenness et al. 2011). Although we still found significant effects of stress, in interaction with genetic variants and gender, on depressive symptoms, our use of a self-report measure may have attenuated associations that are actually stronger than those we have reported.
Perhaps the most significant limitation of the current study is its sample size, which is relatively small for genetic studies and limits the number of participants we have in each gender, genotype, and NLE category. As numerous researchers (e.g., Asherson and Price 2012; Duncan and Keller 2011) have noted, it is important to draw upon as large of sample sizes as possible and to replicate results so that we can be more confident in the results that are reported. In some ways, our results replicate previous work, including findings that the effects of 5-HTTLPR on depression and its correlates may be limited to those individuals with low-expression MAOA-uVNTR and that 5-HTTLPR and MAOA-uVNTR may influence one another in an interactive manner. Our work also provides new support for the possibility that the effects of 5-HTTLPR on depression appear to be less robust in younger boys. In addition, we have adopted strategies suggested to maximize statistical power and reduce unwanted heterogeneity in effects (see McGuffin et al. 2011) by studying an ethnically homogenous population, taking into account the rs25531 SNP within 5-HTTLPR, and considering gender effects given the age of our sample. The presence of the four-way interaction of 5-HTTLPR, MAOA-uVNTR, negative life events, and gender despite the current sample size indicates that these effects have at least moderate effect sizes. However, given our limited sample size, and as with all genetic studies, it will be imperative to replicate these effects in additional samples of adolescents.
In conclusion, the current study indicates that there may be important developmental and gender effects that influence how genes affect the onset of adolescent depression. In particular, the effect of 5-HTTLPR on depression may vary by age and gender. The study also contributes to ongoing debate about whether the low- or high-expression MAOA-uVNTR alleles are associated with depression and adds support to evidence that the low-expression allele may confer vulnerability to stressful experiences. This research replicates and extends a previous study (Cicchetti et al. 2007) to suggest that the 5-HTTLPR by stress interaction may be limited to those individuals with low-expression MAOA-uVNTR alleles. More generally, this finding means that inconsistencies across gene association studies of depression may be due in part to other, omitted, genetic polymorphisms and thus indicates the need for more complex models, such as the present one, that allow for gene-gene interactions, gene-stress interactions, and interactions with gender.
Given that only girls can be heterozygous for MAOA-uVNTR, we cannot analyze these girls in a model that also contains boys. However, when we analyzed girls separately and included heterozygous girls (increasing the sample size by 48), our finding remained unchanged. That is, the interaction of 5-HTTLPR and NLE held only for girls homozygous for low-expression MAOA alleles.
This research was supported, in part, by Award Number F31MH084476 from the National Institute of Mental Health to Heather A. Priess-Groben. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Mental Health or the National Institutes of Health. This research was also supported by a University of Wisconsin Graduate School grant to Janet Shibley Hyde.
We thank Lyn Abramson and Luke Hyde for their comments on an earlier draft.