Association between maternal childhood maltreatment and mother-infant attachment disorganization: Moderation by maternal oxytocin receptor gene and cortisol secretion

Hormones and Behavior 102 (2018) 23–33

Contents lists available at ScienceDirect

Hormones and Behavior journal homepage: www.elsevier.com/locate/yhbeh

Association between maternal childhood maltreatment and mother-infant attachment disorganization: Moderation by maternal oxytocin receptor gene and cortisol secretion

T

Jaclyn A. Ludmera, Andrea Gonzalezb, James Kennedyc,d, Mario Masellise, Paul Meinzf, ⁎ Leslie Atkinsona, a

Department of Psychology, Ryerson University, 350 Victoria Street, M5B 2K3 Toronto, Ontario, Canada Department of Psychiatry and Behavioural Neurosciences, McMaster University, Offord Centre for Child Studies, 1280 Main Street West, Hamilton, Ontario, Canada Department of Psychiatry, University of Toronto, 250 College Street, M5T 1R8 Toronto, Ontario, Canada d Psychiatric Neurogenetics Section, Center for Addiction and Mental Health, 250 College Street, M5T 1R8 Toronto, Ontario, Canada e Department of Neurology, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, M4N 3M5 Toronto, Ontario, Canada f Cosumnes River College, 8401 Center Parkway, Sacramento, CA 95823-5704, USA b c

A R T I C LE I N FO

A B S T R A C T

Keywords: Attachment Attachment disorganization Cortisol Infant Mother Maternal maltreatment history Oxytocin OXTR Strange situation

This study examined maternal oxytocin receptor (OXTR, rs53576) genotype and cortisol secretion as moderators of the relation between maternal childhood maltreatment history and disorganized mother-infant attachment in the Strange Situation Procedure (SSP). A community sample of 314 mother-infant dyads completed the SSP at infant age 17 months. Self-reported maltreatment history more strongly predicted mother-infant attachment disorganization score and disorganized classification for mothers with more plasticity alleles of OXTR (G), relative to mothers with fewer plasticity alleles. Maltreatment history also more strongly predicted mother-infant attachment disorganization score and classification for mothers with higher SSP cortisol secretion, relative to mothers with lower SSP cortisol secretion. Findings indicate that maltreatment history is related to disorganization in the next generation, but that this relation depends on maternal genetic characteristics and cortisol.

Disorganized mother-infant attachment is one of the earliest measurable risk factors for maladaptive developmental trajectories in biologically normal individuals (Lyons-Ruth and Jacobvitz, 2016), having been linked to physiological stress reactivity (Bernard and Dozier, 2010; Luijk et al., 2010), attention and information processing (Claussen et al., 2002; Dykas and Cassidy, 2011), hostility and dissociation (Sroufe, 2005), and psychopathology (Fearon et al., 2010). Despite the importance of disorganization for child development, the origins of disorganized attachment are not completely clear (LyonsRuth and Jacobvitz, 2016). It is therefore “essential” (Bernier & Meinz, 2008, pp. 969) to study the factors contributing to disorganized motherinfant attachment. The current study addresses this issue by examining maternal maltreatment history, cortisol secretion, and OXTR genotype. 1. Attachment disorganization Main and Solomon (1990) identified disorganized attachment based

on the observation that some infants did not meet criteria for any of Ainsworth's (Ainsworth et al., 1978) organized attachment classifications (avoidant, secure, resistant), as assessed in the Strange Situation Procedure (SSP). These disorganized infants tended to engage in i) sequential displays of contradictory behaviors; ii) simultaneous displays of contradictory behaviors; iii) undirected, misdirected, incomplete, and interrupted movements and expressions; iv) stereotypies, asymmetrical movements, mistimed movements, and anomalous postures; v) freezing, stilling, and slowed movements or expressions; vi) direct indices of apprehension of the parent; and vii) direct indices of disorganization and disorientation. These disorganized behaviors are thought to reflect an approach/avoidance conflict as a result of motherinfant interactions in which maternal behavior evokes fear in the infant, or infant behavior evokes fear in the mother (Lyons-Ruth and Jacobvitz, 2016; Main and Hesse, 1990). In other words, infants are thought to exhibit these behaviors because their mother is simultaneously their source of comfort and fear.

⁎

Corresponding author. E-mail addresses: [email protected] (J.A. Ludmer), [email protected] (A. Gonzalez), [email protected] (J. Kennedy), [email protected] (M. Masellis), [email protected] (L. Atkinson). https://doi.org/10.1016/j.yhbeh.2018.04.006 Received 28 September 2017; Received in revised form 10 April 2018; Accepted 13 April 2018 0018-506X/ © 2018 Elsevier Inc. All rights reserved.

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

Keebaugh et al. (2015) found that suppressing OXTR expression in female prairie voles reduced alloparental behavior (i.e., time spent licking and grooming pups). The impact of OXTR on alloparental behavior appears to be “organizational” early in development, as adults over-expressing OXTR in the nucleus accumbens from weaning (but not from adulthood) display increased alloparental behavior (Keebaugh and Young, 2011). Furthermore, early life adversity (i.e., social isolation) reduces the pair bonding of female prairie voles as adults, but only among those with low nucleus accumbens oxytocin receptor densities (Barrett et al., 2015). In other words, female prairie voles are differentially susceptible to early life adversity, depending on variation in oxytocin function. Consistent with such animal research, a single nucleotide polymorphism (SNP) in the third intron, rs53576 (G/A), of the OXTR gene, located on chromosome 3p25, containing four exons and three introns, has been found to impact maternal interactive behavior in humans. For example, in a low-risk community sample, mothers with OXTR genotypes that signify more efficient oxytocin function (i.e., GG genotypes) display more observable sensitivity to their toddlers (BakermansKranenburg and van IJzendoorn, 2008). Bradley et al. (2011) examined childhood maltreatment history, OXTR, and self-reported attachment on the Adult Attachment Prototype Questionnaire in a sample of 284 low-income African American men and women. They found that individuals with the GG genotype, relative to individuals with other genotypes, had higher levels of disorganized attachment in the context of more maltreatment. Thus, given the role of OXTR in maternal bonding, and susceptibility to the influences of early life adversity (specifically the influences of childhood maltreatment on self-reported attachment), the OXTR gene is an excellent candidate for moderating the degree to which maternal history of maltreatment impacts motherinfant disorganization in the next generation, as assessed with the gold standard measure of attachment, the SSP. Relevant to this hypothesis are the definitions of G × E interaction models. Diathesis-stress refers to genetic characteristics exclusively conferring vulnerability to the adverse effects of negative rearing environments. Differential susceptibility refers to genetic characteristics conferring susceptibility to the adverse effects of negative rearing environments and the positive effects of enriched rearing environments (Belsky and Pluess, 2009). Finally, vantage sensitivity refers to genetic characteristics exclusively conferring susceptibility to the positive effects of enriched rearing environments. It is currently unknown how differential context associates with each of the three G × E models (Del Giudice, 2017; Ludmer et al., 2015). However, to address this issue, Roisman et al. (2012) proposed statistical criteria for differentiating between the models, including: i) Regions of Significance on environmental factors (RoS on X): demonstration that the outcome variable and the plasticity genotype are correlated at high and/or low ends of the environmental variable bounded by ± 2SD from the mean; ii) Proportion of interaction index (PoI): ratio of improved outcomes for the plasticity genotype over the sum of improved outcomes and harmful outcomes (this criterion was recently revised by Del Giudice, 2017); and iii) Linearity: repeating analyses when introducing quadratic effects (i.e., the environmental variable squared, and the product of the environmental variable squared and the moderator). Although these statistics are an important step toward clarifying the contexts in which each G × E model occurs, the RoS on X test is biased by sample size, sample characteristics, power, and environmental ranges, and the PoI index lacks clear cut-off guidelines (Del Giudice, 2016, 2017). Thus, it appears that, at this time, the most crucial piece of information is the fact that there is an interaction, as opposed to the type of interaction. In addition to the Roisman criteria, we also present the current results without “binning” alleles. Binning alleles is the typical method in G × E research and it involves creating dichotomous groups of “plasticity genotype” and “non-plasticity genotype” individuals. This is problematic because this unjustifiably assumes allele dominance in heterozygous individuals (Ludmer et al., 2015). To avoid such

To account for the development of this fear, Main and Hesse (1992) proposed that when a mother's traumatic experiences are “unresolved”, the memories and emotions associated with the trauma are reactivated by the infant's attachment cues, thereby triggering a dissociative state during which the mother displays inappropriate behavior while interacting with her infant. Supporting this idea, meta-analytic data found an effect size of r = 0.32 (k = 6; n = 325) linking “frightened or frightening” (FR) behaviors (e.g., aggressive facial expressions, dissociative expressions, submissive parental behavior) to disorganized mother-child attachment (Madigan et al., 2006). Building upon FR behaviors, Lyons-Ruth et al. (1999) identified disrupted communication behaviors (e.g., mocking infant, eliciting reassurance from infant) that can provoke unmodulated infant fear and contribute to disorganized attachment over and above FR behaviors. The effect size linking disrupted communication behaviors to disorganized mother-child attachment is r = 0.35 (k = 4; n = 384, Madigan et al., 2006). Direct child maltreatment (e.g., abuse, neglect) also contributes to disorganization (van IJzendoorn et al.'s, 1999), with a large effect size of r = 0.74 (k = 10; n = 456, Cyr et al., 2010). Importantly, the presence of maternal childhood maltreatment history alone is enough to confer significant risk for disorganization (e.g., even when considering mother's unresolved state of mind, Berthelot et al., 2015). About 15% of middle-class, nonclinical dyads are classified as disorganized in the SSP (van IJzendoorn et al., 1999), but this number triples in the context of maternal childhood maltreatment history, regardless of the dose of maltreatment (i.e., any maltreatment constitutes substantial risk, Berthelot et al., 2015). As such, researchers have suggested that a mother's history of childhood maltreatment is one of the most important causes of disorganization, and that it is sufficient in and of itself to give rise to disorganization in the next generation (Bernier and Meins, 2008). On the other hand, a substantial proportion of mothers with histories of childhood maltreatment do not develop disorganized attachment relationships with their infants (Berthelot et al., 2015), pointing to the role of moderating factors. Bernier and Meins (2008) proposed a threshold model for the development of disorganization that can account for such moderating factors. According to this model, certain maternal characteristics can both alter the dyad's threshold level for developing disorganization, and potentially induce the atypical parenting behaviors that can breach the threshold. As Bernier and Meins (2008) suggest, two such moderating factors may be genetic characteristics and maternal stress reactivity. To our knowledge, the current study is the first to examine these factors as moderators of the relation between maternal childhood maltreatment history and mother-infant disorganization in the SSP. 2. Maternal genetics and mother-infant attachment disorganization There has been increasing interest in examining gene × environment (G × E) interactions as they predict disorganization, although with limited replication and inconsistent results (Lyons-Ruth and Jacobvitz, 2016). For example, maternal unresolved loss was found to only be associated with mother-infant attachment disorganization for infants with the long allele of DRD4 (i.e., 7-repeat) (van IJzendoorn and Bakermans-Kranenburg, 2006), and maternal affective communication was found to only be associated with disorganization for infants with the short allele of DRD4 (Gervai et al., 2007). A notable limitation of G × E studies as they predict mother-infant attachment disorganization is their neglect of maternal genotypes. This oversight is important because maternal genotype is related to maternal features pertinent to mother-child interactions (Mileva-Seitz et al., 2016), and these are reflected in attachment classification. In this regard, a probable maternal genetic candidate is the OXTR gene. Animal models support the role of OXTR in impacting parental behavior and susceptibility to early life adversity. For example, 24

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

ambiguities, we tally the number of candidate alleles (i.e., an individual homozygous for plasticity alleles would be scored 2, an individual heterozygous for plasticity and non-plasticity alleles would be scored 1, and an individual homozygous for non-plasticity alleles would be scored 0).

order to utilize the Roisman et al. (2012) statistics for differentiating between diathesis stress, differential susceptibility, and vantage sensitivity. However, because dichotomous classification is a primary interest in the literature, we replicated all findings when examining disorganization categorically.

3. Maternal cortisol secretion and mother-infant attachment disorganization

5. Method 5.1. Participants

Similar to genetic characteristics, cortisol secretion has almost exclusively been studied in the infant (rather than in the mother) as it relates to mother-infant attachment disorganization (e.g., Bernard and Dozier, 2010; Luijk et al., 2010). This is an important oversight for a number of reasons. Firstly, maternal childhood maltreatment history is associated with heightened maternal cortisol secretion (Bublitz and Stroud, 2013; Schechter et al., 2004). Secondly, as Bernier and Meins (2008) suggest, heightened maternal physiological stress responses may lower the thresholds for mothers to interact with their infants in the insensitive and atypical ways that promote disorganization (Gonzalez et al., 2012). Furthermore, mothers' cortisol levels are associated with their infants' cortisol levels and arousal modulation (e.g., Atkinson et al., 2013), which may contribute to infant disorganization (LyonsRuth and Jacobvitz, 2016). As such, maternal cortisol secretion may play a role within the association between maternal childhood maltreatment history and mother-infant attachment disorganization. Here we examine maternal cortisol secretion as a moderator of the relation between maternal childhood maltreatment history and motherinfant attachment disorganization. We hypothesize a moderating role given that, parallel to genetic characteristics, trait-like physiological characteristics serve as markers of plasticity to environmental influences (van IJzendoorn and Bakermans-Kranenburg, 2012). That is, individuals with high physiological sensitivity (e.g., a tendency to secrete high levels of cortisol), relative to individuals with lower physiological sensitivity, show the most positive outcomes in positive environments and/or the most negative outcomes in negative environments, with outcomes including prosocial behavior and school engagement (Obradovic et al., 2010), intergenerational transmission of cortisol secretion (Khoury et al., 2016), respiratory illness (Boyce et al., 1995), memory (Quas et al., 2004), and internalizing and externalizing difficulties (El-Sheikh et al., 2001). As such, a maternal proclivity toward high levels of cortisol secretion may confer heightened plasticity (in accordance with diathesis-stress, differential susceptibility, or vantage sensitivity) to the impact of maternal childhood maltreatment history on mother-infant attachment disorganization.

We used a community sample of 314 demographically low-risk mother-infant dyads (52% male infants) recruited through postings in community centers and in-person visits to activity centers in Toronto (Ludmer et al., 2015). Infants were full-term and healthy. Mothers were 18 years or older at childbirth, with no known hormonal or psychiatric disorders and with sufficient English to complete questionnaires. Data were obtained when infants were 16 and 17 months old. Maternal age at the 16-month visit ranged from 21 to 46 years (M = 32.94; SD = 4.51). Median family income was $114,000–149,999 Canadian (25th and 75th percentiles were $92,000–113,999 and $150,000–199,999). Maternal highest education levels were: primary (1.0%), secondary (7.7%), community college (23.4%), university (47.8%), and post-graduate degree (21.2%). Most mothers did not smoke (95.5%), were working (60.1%), and were in a relationship (95.0%). Mothers self-reported their ethnicity as Caucasian (67.1%), Asian (14.7%), African American (3.1%), Hispanic (5.9%) and other (including Mixed, East Indian, Middle Eastern, Persian; 9.2%). Maternal report of number of hours of infant out-of-home-care per week (assessed at the 16-month visit) ranged from zero to 35 (Median = 2, Interquartile Range = 8). 5.2. Procedure The Research Ethics Boards at the Centre for Addiction and Mental Health and Ryerson University granted approval for this study. At infant age 16-months, two female experimenters observed the mother and infant in the home and distributed maternal inventories and demographic questionnaires. At 17 months in the laboratory, dyads participated in the SSP. At the end of the visit, buccal cells were collected. Saliva was collected at baseline, and at 20- and 40-minutes post-SSP. 5.3. Measures 5.3.1. Childhood maltreatment Maternal childhood maltreatment was assessed with the CTQ Short Form, which has good psychometric properties (CTQ-SF; Bernstein and Fink, 1998). The CTQ-SF is a 28-item screener examining five types of childhood maltreatment: emotional abuse, physical abuse, sexual abuse, emotional neglect, and physical neglect. Here we utilized a total CTQ-SF score, in which higher scores reflect greater maltreatment. There is evidence for convergent validity with non-self report measures such as therapist ratings based on, e.g., Child Protective Service investigations and court appearances (Bernstein et al., 1997). For the current sample, Cronbach's α = 0.87.

4. The current study The aims of this study were twofold. (i) We assessed whether maternal OXTR genotype moderates the relation between maternal history of childhood maltreatment and mother-infant attachment disorganization in the SSP. It was hypothesized that maltreatment history would more strongly predict mother-infant attachment disorganization for mothers with more plasticity alleles of OXTR (G allele), relative to mothers with fewer plasticity alleles. (ii) We assessed whether maternal cortisol secretion moderates the relation between maternal history of childhood maltreatment and mother-infant attachment disorganization in the SSP. It was hypothesized that maltreatment history would more strongly predict mother-infant attachment disorganization for mothers with higher total cortisol output in the SSP, relative to mothers with lower total cortisol output in the SSP. As reviewed, specific hypotheses regarding which interaction model (i.e., diathesis stress, differential susceptibility, or vantage sensitivity) would emerge could not be made a priori due to the dearth of research differentiating the different models with statistical rigor (e.g., Ludmer et al., 2015; Roisman et al., 2012; Del Giudice, 2016, 2017). Our results are primarily presented with mother-infant attachment disorganization rated continuously, in

5.3.2. Strange Situation Procedure (SSP) The SSP (Ainsworth et al., 1978) consists of eight episodes, each three minutes in duration, wherein the infant is repeatedly separated and reunited with the mother. As is conventional, SSP episodes (but not procedures) were stopped if the infant cried hard for 20 s or was in danger (e.g., standing on a chair while alone in the room). For example, if the infant cried hard for 20 s during SSP episode 6 (alone in room), episode 7 (stranger returns) would begin. Mainly due to reluctance to complete the second visit, which was in the laboratory, rather than in the home, 265 tapes were available for coding. Nine of these tapes were only recorded on VHS tapes that had broken, or had a reunion episode 25

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

describing maternal interactive behavior. Based on a two-hour home observation of mother-infant interaction at infant age 16 months, two blind and trained sorters arranged the items into nine piles of ten items each, ranging from pile 1 (Least Like the Mother) to pile 9 (Most Like the Mother). Each mother's final score is the correlation between the scores of her Q-sort with those of a theoretically derived sort of a prototypically sensitive mother. The mean scores of both sorters were used here. The MBQS has strong psychometric properties (Pederson et al., 1990). Interobserver reliability was high, ICC = 0.88, p < 0.001.

that was not captured on camera and thus could not be assigned a classification. As such, attachment classifications were available for 256 mother-infant dyads. Dyads without classifications did not differ from dyads with classifications with respect to maternal marital status, breast feeding status, self-reported ethnicity, OXTR genotype, childhood maltreatment history, cortisol levels, depressive symptoms, parenting stress, sensitivity, or history of maternal care (see below for descriptions of measures). They also did not differ with respect to family income, infant sex, number of siblings, and number of hours per week in out of home care. Dyads with classifications, relative to dyads without classifications, were more likely to have older mothers (t (279) = −2.04, p < 0.05) and mothers currently working (χ2 (1) = 10.95, p < 0.01). A trained coder (JL) who achieved reliability with Dr. Elizabeth Carlson and Dr. Alan Sroufe coded the tapes. Forty-two (16.4%) tapes were independently double coded by a second trained coder (PM) who achieved reliability with Dr. Elizabeth Carlson and Dr. Alan Sroufe. Raters achieved 81% agreement (34/42 cases) on the disorganized/not disorganized classifications. Kappa was 0.62 (p < 0.001), which is considered “substantial” (Landis and Koch, 1977). Attachment disorganization is also coded on a continuous scale ranging from 1 (No Signs of Disorganization/Disorientation) to 9 (Definite Qualification for D Attachment Status). The intraclass correlation coefficient for reliability on the continuous disorganization scores was 0.91, p < 0.001.

5.3.6. Maternal Depressive Symptoms The Beck Depression Inventory-II (BDI-II, Beck et al., 1996) is a selfreport questionnaire assessing depressive symptoms, and it has strong psychometric properties. It was used here as a potential covariate. Cronbach's alpha was 0.89 in this sample. 5.3.7. Parenting stress Parenting stress, a potential covariate, was assessed with the Parenting Stress Index-Short Form (PSI-SF; Abidin, 1995), consisting of three, 12-item subscales. The Parental Distress subscale examines perception of parenting competence, restriction of personal activities, depressive symptoms, and social support. The Parent-Child Dysfunctional Interaction subscale examines parent's expectations of child behavior and level of reinforcement derived from child interaction. The Difficult Child subscale examines child behavioral characteristics that may impact parenting. The PSI-SF total score (sum of the three subscales) was used here. The PSI-SF has adequate reliability and validity (Abidin, 1995; Haskett et al., 2006). Cronbach's alpha for the current sample was 0.88.

5.3.3. Maternal cortisol To address time-of-day issues, all visits commenced between 0900 h and 1000 h. Saliva was collected at baseline, and at 20- and 40-minutes post SSP. At each time point, two Sorbettes (Salimetrics, State College PA) were put in the mouth for 60 s. Samples were centrifuged for 10minutes at 3000 rpm at 4 °C to extract saliva and then sealed and stored at −70 °C. Each sample was assayed twice using a salivary cortisol enzyme immunoassay kit (Salimetrics, State College, PA). Average values were used in analyses. Principal component analyses have revealed that repeated cortisol measurements can be grouped into two components, “total cortisol production” and “change in cortisol levels” (Khoury et al., 2015; Fekedulegn et al., 2007). Of the various cortisol indices utilized in the literature, Pruessner et al.'s (2003) area under the curve with respect to ground (AUCG) and area under the curve with respect to increase (AUCI) consistently load highly on these two components, respectively (Khoury et al., 2015). More specifically, AUCG captures both the difference between single measurement points and the distance of single measurement points from the ground, whereas AUCI captures only the change in measurement points. AUCG is more stable over time than AUCI (Hankin et al., 2015). Thus, in the current study, we utilize AUCG in order to assess cortisol secretion as a trait-like marker of environmental plasticity, following van IJzendoorn and Bakermans-Kranenburg (2012). AUCG was computed from the trapezoid formula (Pruessner et al., 2003) using the three time points: {[(20 min value + baseline value) / 2] × time} + {[(40 min value + 20 min value) / 2] × time}.

6. Results 6.1. Data preparation and analytic approach The continuous disorganization score was slightly positively skewed, however, log transformations exacerbated the skew. We therefore utilized the original variable in analyses and examined the variable for outliers that may bias results. (We also replicated analyses with disorganization treated as classificatory variable, as described below). < 5% of scores had standardized residuals above 2 SD from the mean, and Cook's values, examined in all analyses, were consistently below 1, indicating no undue influence of outliers on results (Field, 2009). Continuous predictors (i.e., CTQ and AUCG) were centered for analyses. In terms of potential covariates, we utilized correlations or univariate ANOVAs to assess relationships between disorganization scores and marital status, employment, age, breast feeding status, self-reported ethnicity, depression, parenting stress, and sensitivity. We also examined family income, infant sex, number of siblings, and hours per week in out of home care (as assessed by maternal report at infant age 16 months). Only infant sex was related to disorganization (F(1, 255) = 11.98, p < 0.05), such that males (M = 3.96, SD = 1.60) had higher disorganization scores than females (M = 3.52, SD = 1.42). Thus, infant sex was included in all analyses as a covariate. Keller (2014) suggests including covariate × environment (CTQ) and covariate × moderator interaction terms in the model for a more stringent and conservative approach to covariates. Thus, we also replicate our results with this procedure. In regard to the potential effects of population stratification, although ethnicity was not associated with the outcome of interest, which some have argued precludes the need to control for it (e.g., Drury et al., 2012; Hutchinson et al., 2004), we nevertheless controlled for maternal ethnicity as an additional covariate for the OXTR analyses. To account for missing data with respect to 8 (2.66%) CTQ scores, 45 (14.95%) disorganization scores, and 49 (16.28%) AUCG scores, we used multiple imputation. Multiple imputation is valid where as many as 50% of the data are missing (Collins et al., 2001). Twenty multiple imputations were conducted (Collins et al., 2001) using SPSS 24. Based on Little's (1988) missing completely at random (MCAR) test, the data were suitable for imputation, χ2 (9) = 13.03, p = 0.16. For main

5.3.4. Maternal genotypes DNA was assessed from buccal cells at each visit. Four paper swabs (Whatman Omniswab, Fisher Scientific Company) were rubbed four times on each cheek and expelled into polypropylene tubes, which were sealed and stored at 4°C. DNA isolation and analysis was conducted at the Center for Addiction and Mental Health in Toronto. A 1ul volume of total genomic buccal swab DNA was amplified using the manufacturer's standard TaqMan genotyping protocol, scaled to a total reaction volume of 10 μL. Genotype calls were initially made in GeneMapper v4 and confirmed manually. Genotypes met Hardy-Weinberg equilibrium. 5.3.5. Maternal sensitivity As a potential covariate, maternal sensitivity was examined with the Maternal Behavior Q-Sort (MBQS, Pederson et al., 1990), a set of 90 items 26

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

analyses, each moderator (OXTR and AUCG) was examined separately in order to reduce the number of predictors and interactions in each model, given that genetic analyses require very large sample sizes (Duncan and Keller, 2011). For continuous disorganization scores as outcome, hierarchical multiple regression analyses were conducted with covariates, CTQ score, moderator (OXTR or AUCG), and CTQ × moderator as predictors. Based on best practice (Ludmer et al., 2015), our OXTR analyses did not bin alleles, i.e., we scored the OXTR moderator as the number of G alleles ranging from 0 (no G alleles) to 2 (two G alleles). When replicating primary analyses with the dichotomous disorganization classification as outcome, binary logistic regression analyses were conducted with covariates, CTQ score, moderator (OXTR or AUCG), and CTQ × moderator as predictors. We then replicated all results without imputation. For exploratory purposes, data were analyzed for diathesis-stress, vantage sensitivity, and differential susceptibility with Roisman et al.'s (2012) RoS on X, PoI, and linearity tests. As in Ludmer et al. (2015), the Roisman et al. (2012) analyses were conducted with the non-imputed data, given that Roisman et al.'s (2012) online calculators are not compatible with imputed data.

Table 1 Correlations among main study variables.

1. CTQ 2. OXTR number of G allelesa 3. AUCG 4. Disorganization continuous score 5. Disorganization dichotomous classificationb 6. Infant sexc 7. Maternal ethnicityd

1.

2.

3.

4.

– 0.09

–

−0.04 0.03

5.

0.10 0.00

– −0.03

–

0.02

−0.02

−0.06

0.69⁎⁎⁎

–

0.01 −0.11

−0.08 −0.19⁎⁎

0.03 −0.03

−0.14⁎ −0.10

−0.19⁎⁎ −0.07

6.

– −0.02

Note: CTQ = childhood trauma questionnaire. AUCG = Area under the curve with respect to ground, i.e., total maternal cortisol output. a Coded as no G alleles = 0, one G allele = 1, two G alleles = 2. b Coded as not disorganized = 0, disorganized = 1. c Coded as male = 0, female = 1. d Coded as not Caucasian = 0, Caucasian = 1. ⁎ p = 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001.

6.2. Descriptive statistics Genotype distributions were as follows: 113 (37.8%) had 0 G alleles, 136 (45.5%) had 1 G allele, and 50 (16.7%) had 2 G alleles. SSP classifications were as follows: 34 (13.3%) avoidant, 121 (47.3%) secure, 51 (19.9%) resistant, and 50 (19.5%) disorganized. Descriptive statistics are provided here, with means and standard deviations reported for normally distributed variables, and with medians and interquartile ranges reported for variables that deviated from normality. Disorganization scores ranged from 1 to 8 (Median = 4.00, Interquartile Range = 1.50), where a score above 5 reflects a disorganized classification. CTQ scores ranged from 25 to 112 (Median = 32.00, Interquartile Range = 11.00), with a minority of mothers meeting the subscale cut-offs for “severe” maltreatment for emotional abuse (16+, 6.8%), physical abuse (13+, 3.7%), sexual abuse (13+, 7.1%), emotional neglect (18+, 4.7%), and physical neglect (13+, 3.0%). AUCG values ranged from 52.34 to 1088.58 (Median = 199.10, Interquartile Range = 164.88). PBI scores ranged from 0 to 36 (Median = 29.00, Interquartile Range = 10.00), with 60.2% of mothers meeting the cut-off score (27) of high care. MBQS scores ranged from −0.69 to 0.90 (M = 0.46, SD = 0.34). PSI scores ranged from 47 to 133 (M = 79.97, SD = 16.76), with 23.4% of participants meeting the cut-off score of the 90th percentile. BDI scores ranged from 0 to 39 (Median = 6.0, Interquartile Range = 7.00), with 9.8% of mothers meeting the clinical cut-off of 17+. Table 1 provides correlations among main variables.

equivalent when re-analyzed without multiple imputation, with the overall model reaching significance, F (5, 237) = 2.23, p = 0.05, R2 = 0.05 (Table 2). The RoS on X test revealed that the regression of disorganization on OXTR genotype was significant for all values of centered CTQ that fell below −26.67, which is within 2SD of the mean of CTQ. This is most consistent with a vantage sensitivity model (Roisman et al., 2012). The PoI index was 0.59, which is also consistent with vantage sensitivity. When added to the model, nonlinear terms were non-significant. Thus, these tests suggest vantage sensitivity, and mothers with more G alleles of OXTR, relative to mothers with fewer G alleles of OXTR, have infants with lower disorganization scores in the context of low maltreatment histories (Fig. 1). We then replicated the model when predicting the dichotomous disorganization categories (as opposed to the continuous disorganization score) using binary logistic regression. The overall model was significant, χ2 (5) = 18.44, p < 0.01, Nagelkerke R2 = 0.11. The interaction between CTQ and OXTR genotype significantly predicted disorganization classification, β = 0.03, SE = 0.01, p < 0.05 (Table 2).2 Results were equivalent when re-analyzed without multiple imputation (Table 2). 6.3.2. Maternal Cortisol AUCG as moderator of the relation between maternal childhood maltreatment history and mother-infant attachment disorganization The overall model was significant, F (4, 297) = 2.87, p < 0.05, R2 = 0.04. The interaction between CTQ and AUCG made a significant contribution to disorganization score, β = 0.14, p < 0.05 (Table 3, Fig. 2, note that a median split was used for AUCG in Fig. 2 for illustrative purposes, but AUCG was not median split in analyses). Specifically, CTQ more strongly predicted disorganization score for mothers with higher total cortisol output, relative to mothers with lower total cortisol output.3 Results were equivalent when re-analyzed without multiple imputation (Table 3). The RoS on X test revealed that the regression of disorganization on AUCG was significant for all values of

6.3. Main Analyses 6.3.1. Maternal OXTR genotype as moderator of the relation between maternal childhood maltreatment history and mother-infant attachment disorganization The overall model approached significance, F (5, 281) = 2.37, p = 0.06, R2 = 0.04, and the interaction between CTQ and OXTR genotype made a significant contribution to disorganization score, β = 0.19, p < 0.05 (Table 2, Fig. 1). Specifically, CTQ more strongly predicted disorganization score for mothers with more G alleles of OXTR, relative to mothers with fewer G alleles.1 Results were

2 When including Keller's (2014) stringent covariate terms (i.e., ethnicity × CTQ, ethnicity × OXTR, sex × CTQ, and sex × OXTR), the CTQ × OXTR interaction approached significance (p = 0.09). Again, it seems likely that the increase in p-value reflects loss of degrees of freedom with the addition of effects to the model, without diminishing the error term; we suggest this because none of the Keller (2014) additional interactions proved significant. 3 When including Keller's (2014) stringent covariate terms (i.e., sex × CTQ, and sex × AUCG), the CTQ × AUCG interaction remained significant.

1 When including Keller's (2014) stringent covariate terms (i.e., ethnicity × CTQ, ethnicity × OXTR, sex × CTQ, and sex × OXTR), the CTQ × OXTR interaction approached significance (p = 0.09). It seems likely that the increase in p-value reflects loss of degrees of freedom with the addition of effects to the model, without diminishing the error term; we suggest this because none of the Keller (2014) additional interactions proved significant.

27

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

Table 2 Multiple linear and binary logistic regression analyses to predict mother-infant attachment disorganization continuous scores and dichotomous categories (respectively) in the strange situation procedure from maternal OXTR genotype × maternal childhood maltreatment history interaction. Outcome; type of dataset

Ethnicity Infant sex CTQ OXTRa CTQ × OXTR

Continuous disorganization; nonimputed

Continuous disorganization; imputed

Dichotomous disorganization; nonimputed

Dichotomous disorganization; imputed

b(SE)

b(SE)

b(SE)

b(SE)

−0.33(0.22) −0.40(0.20)⁎ −0.01(0.01) −0.05(0.14) 0.02(0.01)⁎

−0.30(0.22) −0.37(0.20) −0.01 (0.01) −0.06(0.14) 0.02(0.01)⁎

−0.50(0.39) −1.34(0.39)⁎⁎⁎ −0.02(0.02) −0.29(0.26) 0.03(0.02)⁎

−0.49(0.35) −1.16(0.38)⁎⁎ −0.02(0.02) −0.24(0.27) 0.03(0.01)⁎

Note. CTQ = childhood trauma questionnaire. a Coded as no G alleles = 0, one G allele = 1, two G alleles = 2. ⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001. Fig. 1. Maternal OXTR genotype moderating the relation between maternal history of childhood maltreatment (CTQ score) and mother-infant attachment disorganization in the Strange Situation Procedure. Maternal history of childhood maltreatment more strongly predicted mother-infant attachment disorganization score for mothers with more G alleles, relative to mothers with fewer G alleles.

classification, β = 0.0002, SE = 0.00007, p < 0.05 (Table 3).4 Results were equivalent when re-analyzed without multiple imputation (Table 3).

centered CTQ that fell below −15.43 and above 16.08, which are within 2SD of the mean of CTQ. This is most consistent with a differential susceptibility model (Roisman et al., 2012). The PoI index was 0.50, which is typical of differential susceptibility (Roisman et al., 2012). When added to the model, nonlinear terms were non-significant. Thus, these tests suggest differential susceptibility, and mothers with higher total cortisol output, relative to mothers with lower total cortisol output, have infants with lower disorganization scores in the context of low maltreatment histories and have infants with higher disorganization scores in the context of high maltreatment histories. We then replicated the model when predicting the dichotomous disorganization categories (as opposed to the continuous disorganization score) using binary logistic regression. The overall model was significant, χ2 (4) = 21.01, p < 0.01, Nagelkerke R2 = 0.11. The interaction between CTQ and AUCG significantly predicted disorganized

7. Discussion Although maternal childhood maltreatment history is a potent risk factor for mother-infant attachment disorganization (Berthelot et al., 2015), a significant proportion of mothers with maltreatment histories do not develop disorganized relationships with their infants. This study examined whether maternal OXTR genotype and cortisol secretion 4 When including Keller's (2014) stringent covariate terms (i.e., sex × CTQ, and sex × AUCG), the CTQ × AUCG interaction remained significant.

28

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

Table 3 Multiple linear and binary logistic regression analyses to predict mother-infant attachment disorganization continuous scores and dichotomous categories (respectively) in the strange situation procedure from maternal AUCG × maternal childhood maltreatment history interaction. Outcome; type of dataset

Infant sex CTQ AUCG CTQ × AUCG

Continuous disorganization; nonimputed

Continuous disorganization; imputed

Dichotomous disorganization; nonimputed

Dichotomous disorganization; imputed

b(SE)

b(SE)

b(SE)

b(SE)

−0.36(0.20) 0.01(0.01) 0.00005(0.001) 0.0001(0.00004)⁎⁎

⁎

⁎⁎

−1.11(0.38)⁎⁎ 0.01(0.01)⁎⁎⁎ −0.002(0.002) 0.0002(0.0001)⁎

−1.15(0.38) 0.02(0.01) −0.002(0.002) 0.0002(0.0001)⁎

−0.39(0.19) 0.01(0.01) −0.00009(0.001) 0.00009(0.00004)⁎

Note. CTQ = childhood trauma questionnaire. AUCG = area under the curve ground, i.e., total maternal cortisol output. ⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.001.

Fig. 2. Maternal total cortisol output (AUCG) moderating the relation between maternal history of childhood maltreatment (CTQ score) and motherinfant attachment disorganization in the Strange Situation Procedure. Maternal history of childhood maltreatment more strongly predicted mother-infant attachment disorganization score for mothers with higher total cortisol output, relative to mothers with lower total cortisol output. Note that total cortisol output is presented dichotomously with a median split in the Figure for illustrative purposes, but it was examined continuously in analyses.

disorganization. In this regard, our findings suggest that the G allele of OXTR and relatively high cortisol secretion can potentially exacerbate the link between maternal maltreatment history and infant disorganized behavior. Indeed, oxytocin and/or cortisol are associated with various correlates of disorganization (e.g., Ellenbogen et al., 2012; Vreeburg et al., 2010), including attention and information processing issues (Claussen et al., 2002; Dykas and Cassidy, 2011), hostility and dissociation (Sroufe, 2005), and psychopathology (Fearon et al., 2010). There are multiple mechanisms through which maternal cortisol secretion and OXTR genotype can influence the relationship between maternal maltreatment history and mother-infant attachment disorganization. With respect to cortisol, high maternal cortisol levels may directly heighten infant cortisol secretion (e.g., Khoury et al., 2016), which may in turn result in the unmodulated infant arousal underlying disorganized behaviors (Lyons-Ruth and Jacobvitz, 2016).

moderate the relation between maternal history of childhood maltreatment and mother-infant attachment disorganization in the SSP. Results revealed that maternal history of childhood maltreatment more strongly predicts disorganization score and classification for mothers with more plasticity alleles of OXTR (i.e., G allele), relative to mothers with fewer plasticity alleles. Likewise, maternal history of childhood maltreatment more strongly predicts disorganization score and classification for mothers with higher total cortisol output in the SSP, relative to mothers with lower total cortisol output in the SSP. The current results support the idea that both genetic characteristics and cortisol secretion are markers of environmental plasticity (van IJzendoorn and Bakermans-Kranenburg, 2012). These findings are consistent with Bernier and Meins's (2008) threshold model of the development of disorganization wherein maternal genetic and physiological characteristics can alter a dyad's threshold for developing 29

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

In terms of the Roisman et al. (2012) analyses, we found that the OXTR × maternal maltreatment history interaction best reflected vantage sensitivity (i.e., in the context of a history of low maltreatment, mothers with G alleles, relative to mothers without G alleles, had infants with lower disorganization). By contrast, the maternal cortisol × maltreatment history interaction best reflected differential susceptibility (i.e., mothers with higher total cortisol output, relative to mothers with lower total cortisol output, had infants with lower disorganization in the context of a history of low maltreatment, and had infants with higher disorganization in the context of a history of high maltreatment). It is difficult to know why OXTR and cortisol secretion reflect different interactional models, and perhaps this can only emerge with the accumulation of data. As Del Giudice (2016, 2017) argued, the Roisman et al. (2012) criteria are limited by power and restricted environmental ranges and thus may not accurately categorize the types of interaction models. Although the quest to characterize interactions remains an important one, perhaps the most crucial piece of information in the present context is the fact that there is an interaction, as opposed to what type of interaction it is. In terms of clinical implications, the current findings may augment our understanding of attachment-based interventions, which are effective at enhancing the quality of mother-child attachment (Berlin et al., 2016). Interestingly, such interventions have been shown to be most effective for children with the 7-repeat allele of the dopamine receptor genotype, that is, for children with differential susceptibility markers (Bakermans-Kranenberg et al., 2008). Although OXTR has not been assessed in the context of interventions, one may expect that mothers with G alleles of OXTR and/or high cortisol secretion may benefit the most from attachment-based interventions (relative to mothers without G alleles and mothers with lower cortisol secretion). Furthermore, interventions that include attachment-based components have been shown to lower and normalize maternal cortisol secretion (e.g., Field et al., 1998; Toth et al., 2015; Urizar and Munoz, 2011). Thus, based on the current findings, it is possible that attachment-based interventions are effective at improving the quality of mother-infant attachment because they lower maternal cortisol secretion and thereby reduce the impact of a mother's difficult childhood experiences on her attachment with her infant. Future studies should explore these hypotheses. In terms of study limitations, our method relied on maternal reports of childhood maltreatment. However, self-reports are standard methodology in the area, perhaps because other methodologies also have shortcomings. For example, although Child Protective Service records are less vulnerable to memory bias, the methodology is susceptible to false positives and negatives (Pinto and Maia, 2013). Furthermore, maltreatment is a dimensional construct, not a categorical one; for example, it may vary from infrequent spankings to frequent beatings, both reprehensible but the latter far more damaging than the former. This dimensional aspect is lost when maltreatment is assessed as a conviction, no conviction dichotomy. In any event, Hardt and Rutter (2004) found that, although adult retrospective reports of childhood adverse experiences may be biased by false negatives, when abuse and neglect are reported, these positive reports have likely occurred. Accordingly, the CTQ has previously been shown to interact with OXTR to predict self-reported attachment disorganization in a sample of lowincome African American men and women (Bradley et al., 2011). The utilization of the CTQ here allowed for conceptual replication of Bradley et al.'s (2011) results, which is particularly important given that less than one third of G × E replication attempts are statistically significant (Duncan and Keller, 2011). Under ideal circumstances, CTQ, OXTR, cortisol, and their interactions should be examined in a single equation, however, given all the covariates involved, this would translate into at least 13 effects and would require a sample size well beyond that presented here to ensure both sensitivity of analysis and reduced probability of Type I and Type II errors. Relatedly, the current sample size is relatively small to test G × E interactions (Duncan and Keller, 2011). In regard to population

Furthermore, Bernier and Meins (2008) suggest that high levels of maternal cortisol secretion during mother-infant interactions may promote the atypical parenting behaviors linked to disorganization. Although maternal disrupted communication has been linked to blunted maternal cortisol secretion in the context of infant stress, this association has only been examined within samples at high socioeconomic risk (Crockett et al., 2013; Schechter et al., 2004), and was only significant for mothers demonstrating the most extreme levels of disrupted communication (Crockett et al., 2013). Given that high socioeconomic risk can cloud the impact of maltreatment history on disorganization (Cyr et al., 2010), it is possible that in low risk populations, high (as opposed to blunted) maternal cortisol secretion relates to insensitive, FR and disrupted communication behaviors (Gonzalez et al., 2012). Blunted levels of cortisol are, after all, the product of unsustainably high and chronic activation of the HPA system. Future studies are needed to test this hypothesis. Elevated maternal cortisol secretion may further impact the association between maltreatment history and mother-infant attachment disorganization by influencing maternal anxiety symptoms (Vreeburg et al., 2010), which are in turn associated with elevated rates of disorganized mother-child attachment (Manassis et al., 1994). The maternal depressive symptoms associated with elevated cortisol secretion (Dienes et al., 2013) may also play a role, although perhaps only when severe (Lyons-Ruth and Jacobvitz, 2016). With respect to OXTR, animal models suggest that increased OXTR expression can promote social processing (Ferguson et al., 2000; Johnson and Young, 2015) and alloparental behavior (Keebaugh et al., 2015; Keebaugh and Young, 2011). This may occur via the role of oxytocin as a social reinforcement signal within the nucleus accumbens (e.g., Dolen et al., 2013; Keebaugh and Young, 2011). As such, we speculate that mothers with the G allele (relative to mothers without this allele), who grew up in environments with low levels of maltreatment, may be most likely to respond warmly (Klahr et al., 2015) and sensitively (Bakermans-Kranenburg and van IJzendoorn, 2008) to infant cues, and least likely to engage in the atypical behaviors associated with maltreatment history and disorganization, perhaps as a result of enhanced oxytocin reward signaling in the nucleus accumbens (e.g., Dolen and Malenka, 2014; Kohls et al., 2013). In the context of childhood abuse, however, increased oxytocin has been shown to reduce social reward learning and memory (Bhandari et al., 2014), which may impact the maternal attachment representations and dissociative symptoms (Reiner et al., 2016) that promote disorganization (Madigan et al., 2006). Furthermore, in the context of childhood abuse, adults with OXTR GG genotypes (relative to those without these genotypes) report more difficulty regulating their emotions (Bradley et al., 2011). Maternal displays of such dysregulated emotion in the context of mother-infant interactions could potentially trigger infant fear and disorganized behaviors (Lyons-Ruth and Jacobvitz, 2016). For exploratory purposes, we examined whether the OXTR × maltreatment and cortisol × maltreatment interactions reflect diathesis stress, differential susceptibility, or vantage sensitivity models using the Roisman et al. (2012) statistical criteria. However, it is important to note that we examined maternal history of maltreatment as the environmental variable, and low history of maltreatment does not necessarily constitute an “enriched” environment. Differential susceptibility theory has typically been treated dichotomously, with susceptibility markers being theorized to lead to either exceptionally good or exceptionally bad outcomes in enriched and negative environments, respectively (Belsky and Pluess, 2009). The current findings suggest that a “less negative” environment (i.e., lack of maltreatment), in combination with susceptibility markers, can lead to good outcomes. As such, differential susceptibility theory may be best conceptualized as dimensional: exceptionally good (or bad) environments in combination with susceptibility markers may breed exceptionally good (or bad) outcomes, whereas relatively good (or bad) environments in combination with susceptibility markers may breed relatively good (or bad), but not exceptional, outcomes. 30

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

Conflicts of interest

stratification, although we controlled for ethnicity within our models (Keller, 2014), and found no association between ethnicity and disorganization (which some have argued precludes the need to control for it, e.g., Drury et al., 2012; Hutchinson et al., 2004), others have argued in favor of replicating results for each ethnicity separately (e.g., Alexander et al., 2011), or including a 3-way interaction term, G × E × Ethnicity (e.g., Kim et al., 2010), as the effect of interest may differ within each ethnic group. Given that we do not have the sample size to support such analyses, we cannot conclude with certainty that population stratification did not play a role here. However, our OXTR findings do conceptually replicate Bradley et al. (2011), who used a sample consisting entirely of African American participants. This successful replication, in addition to our use of the gold standard measure of mother-infant attachment (i.e., SSP), and our reduction in error from utilizing non-binned genetic terms, does attenuate the probability of Type I error. Another study limitation is that, due to reluctance to complete the second visit, which was in the laboratory, dyads with attachment classifications, relative to dyads without classifications, were more likely to have older mothers and mothers currently working. Thus, results may not generalize to younger or unemployed mothers. However, we replicated all results when analyzing both the original dataset (with missing data deleted listwise) and the dataset with imputed variables. Of further note, future studies might test the current model in high risk (e.g., low income) samples, which typically have high rates of disorganization (Lyons-Ruth and Jacobvitz, 2016). It will also be important for future studies to test the potential mediating roles of maltreating parenting, FR behaviors, reflective function, and disrupted communication behaviors (Berthelot et al., 2015; Lyons-Ruth et al., 1999; Madigan et al., 2006). Further studies could examine other genetic markers as well as epigenetic effects (e.g., impacting glucocorticoid reception function) as they relate to atypical parenting behaviors and mother-infant attachment disorganization (Mileva-Seitz et al., 2016).

None. References Abidin, R.R., 1995. Parenting Stress Index. Professional Manual, 3rd ed. Psychological Assessment Resources, Odessa, FL. Ainsworth, M.D.S., Blehar, M.C., Waters, E., Wall, S., 1978. Patterns of Attachment: A Psychological Study of the Strange Situation. Erlbaum, Hillsdale, NJ. Alexander, N., Osinsky, R., Mueller, E., Schmitz, A., Guenthert, S., Yvonne, K., Hennig, J., 2011. Genetic variants within the dopaminergic system interact to modulate endocrine stress reactivity and recovery. Behav. Brain Res. 216, 53–58. http://dx.doi.org/ 10.1016/j.bbr.2010.07.003. Atkinson, L., Gonzalez, A., Kashy, D.A., Basile, V.S., Masellis, M., Pereira, J., et al., 2013. Maternal sensitivity and infant and mother adrenocortical function across challenges. Psychoneuroendocrinology 38, 2943–2951. http://dx.doi.org/10.1016/j.psyneuen. 2013.08.001. Bakermans-Kranenberg, M.J., van IJzendoorn, M.H., Mesman, J., Alink, L.R.A., Juffer, F., 2008. Effects of an attachment-based intervention on daily cortisol moderated by dopamine receptor D4: a randomized control trial on 1- to 3-year-olds screened for externalizing behavior. Dev. Psychopathol. 20, 805–820. Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., 2008. Oxytocin receptor (OXTR) and serotonin transporter (5-HTT) genes associated with observed parenting. Soc. Cogn. Affect. Neurosci. 3, 128–134. http://dx.doi.org/10.1093/scan/nsn004. Barrett, C.E., Arambula, S.E., Young, L.J., 2015. The oxytocin system promotes resilience to the effects of neumatal isolation on adult social attachment in female prairie voles. Transl. Psychiatry 5, e606. http://dx.doi.org/10.1038/tp.2015.73. Beck, A.T., Steer, R.A., Brown, G.K., 1996. Beck Depression Inventory Manual, second ed. The Psychological Corporation, San Antonio, TX. Belsky, J., Pluess, M., 2009. Beyond diathesis stress: differential susceptibility to environmental influences. Psychol. Bull. 135, 885–908. http://dx.doi.org/10.1037/ a0017376. Berlin, L.J., Zeanah, C.H., Lieberman, A.F., 2016. Prevention and intervention programs to support early attachment security: a move to the level of the community. In: Cassidy, J., Shaver, P.R. (Eds.), Handbook of Attachment: Theory, Research, and Clinical Applications. Guildford, New York, pp. 666–697. Bernard, K., Dozier, M., 2010. Examining infants' cortisol responses to laboratory tasks among children varying in attachment disorganization: stress reactivity or return to baseline? Dev. Psychol. 46, 1771. http://dx.doi.org/10.1037/a0020660. Bernier, A., Meins, E., 2008. A threshold approach to understanding the origins of attachment disorganization. Dev. Psychol. 44, 969–982. http://dx.doi.org/10.1037/ 0012-1649.44.4.969. Bernstein, D.P., Fink, L., 1998. Childhood Trauma Questionnaire: A Retrospective SelfReport. Professional Manual. The Psychological Corporation, New York. Bernstein, D.P., Ahluvalia, T., Pogge, D., Handelsman, L., 1997. Validity of the childhood trauma questionnaire in an adolescent psychiatric population. J. Am. Acad. Child Psychiatry 36, 340–348. Berthelot, N., Ensink, K., Bernazzani, O., Normandin, L., Luyten, P., Fonagy, P., 2015. Intergenerational transmission of attachment in abused and neglected mothers: the role of trauma-specific reflective functioning. Infant Ment Health J. 36, 200–212. http://dx.doi.org/10.1002/imhj.21499. Bhandari, R., van der Veen, R., Parsons, C.E., Young, K.S., Voorthuis, A., BakermansKranenburg, M.J., ... van IJzendoorn, M.H., 2014. Effects of intranasal oxytocin administration on memory for infant cues: moderation by childhood emotional maltreatment. Soc. Neurosci. 9, 536–547. http://dx.doi.org/10.1080/17470919.2014. 932307. Boyce, W.T., Chesney, M., Alkon, A., Tschann, J.M., Adams, S., Chesterman, B., Cohen, F., Kaiser, P., Folkman, S., Wara, D., 1995. Psychobiologic reactivity to stress and childhood respiratory illnesses: results of two prospective studies. Psychosom. Med. 57, 411–422. Bradley, B., Westen, D., Mercer, K.B., Binder, E.B., Jovanovic, T., Crain, D., Heim, C., 2011. Association between childhood maltreatment and adult emotional dysregulation in a low-income, urban, African American sample: moderation by oxytocin receptor gene. Dev. Psychopathol. 23, 439–452. http://dx.doi.org/10.1017/ S0954579411000162. Bublitz, M.H., Stroud, L.R., 2013. Maternal history of child abuse moderates the association between daily stress and diurnal cortisol in pregnancy: a pilot study. Stress 16, 706–710. http://dx.doi.org/10.3109/10253890.2013.825768. Claussen, A.H., Mundy, P.C., Mallik, S.A., Willoughby, J.C., 2002. Joint attention and disorganized attachment status in infants at risk. Dev. Psychopathol. 14, 279–291. Collins, L.M., Schafer, J.L., Kam, C.M., 2001. A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychol. Methods 6, 330–351. http:// dx.doi.org/10.1037//1082-989X.6.4.330. Crockett, E.E., Holmes, B.M., Granger, D.A., Lyons-Ruth, K., 2013. Maternal disrupted communication during face-to-face interaction at 4 months: relation to maternal and infant cortisol among at-risk families. Infancy 18, 1111–1134. http://dx.doi.org/10. 1111/infa.12015. Cyr, C., Euser, E.M., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., 2010. Attachment security and disorganization in maltreating and high risk families: a series of meta-analyses. Dev. Psychopathol. 22, 87–108. http://dx.doi.org/10.1017/ S0954579409990289. Del Giudice, M., 2016. The evolution of interaction shape in differential susceptibility. Child Dev. http://dx.doi.org/10.1111/cdev.12710.

8. Conclusions In summary, maternal OXTR genotypes and cortisol secretion moderate the relation between maternal history of childhood maltreatment and mother-infant attachment disorganization in the context of the SSP. Maternal history of childhood maltreatment more strongly predicts increases in disorganization scores and the probability of a disorganized classification for mothers with more plasticity alleles of OXTR (G), relative to mothers with fewer plasticity alleles of OXTR, and for mothers with higher total cortisol output, relative to mothers with lower total cortisol output. Acknowledgments The authors would like to acknowledge the contributions of Dr. Susan Goldberg, a co-principal investigator on the grant that supported this work. She died prior to completion of the project, but her influence pervades the manuscript. Thank you to Sandi Bento and Dr. David Pederson for their help in attachment coding training. Thank you to Dr. Chris Fraley and Dr. Glenn Roisman for their help with the statistical analyses. We would like to thank Emilie Boucher and Monica Tan, as well as other research assistants and students who contributed a great many hours to this project. And of course, we appreciate the time and effort of the mothers and babies who volunteered to participate in this project. Funding This research was supported by funds from the Canadian Institutes of Health Research (MOP-64301) and Ryerson University with no stipulations other than we complete the work. 31

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

pnas.1010830107. Klahr, A.M., Klump, K., Burt, A., 2015. A constructive replication of the association between the oxytocin receptor genotype and parenting. J. Fam. Psychol. 29, 91–99. http://dx.doi.org/10.1037/fam0000034. Kohls, G., Perino, M.T., Taylor, J.M., Madva, E.N., Cayless, S.J., Troiani, V., ... Schultz, R.T., 2013. The nucleus accumbens is involved in both the pursuit of social reward and the avoidance of social punishment. Neuropsychologia 51, 2062–2069. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics. 33, 159–174. Little, R.J., 1988. A test of missing completely at random for multivariate data with missing values. J. Am. Stat. Assoc. 83, 1198–1202. Ludmer, J.A., Levitan, R., Gonzalez, A., Kennedy, J., Villani, V., Masellis, M., Basile, V.S., Atkinson, L., 2015. DRD2 and SLC6A3 moderate impact of maternal depressive symptoms on infant cortisol. Psychoneuroendocrinology 62, 243–251. http://dx.doi. org/10.1016/j.psyneuen.2015.08.026. Luijk, M.P., Saridjan, N., Tharner, A., van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., Jaddoe, V.W., ... Tiemeier, H., 2010. Attachment, depression, and cortisol: deviant patterns in insecure-resistant and disorganized infants. Dev. Psychobiol. 52, 441–452. http://dx.doi.org/10.1002/dev.20446. Lyons-Ruth, K., Jacobvitz, D., 2016. Attachment disorganization: genetic factors, parenting contexts, and developmental transformation from infancy to adulthood. In: Cassidy, J., Shaver, P.R. (Eds.), Handbook of Attachment: Theory, Research, and Clinical Applications. Guildford, New York, pp. 666–697. Lyons-Ruth, K., Bronfman, E., Parsons, E., 1999. Maternal frightened, frightening, or atypical behavior and disorganized infant attachment patterns. Monogr. Soc. Res. Child Dev. 64, 67–96. Madigan, S., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., Moran, G., Pederson, D.R., Benoit, D., 2006. Unresolved states of mind, anomalous parental behavior, and disorganized attachment: a review and meta-analysis of a transmission gap. Attach Hum. Dev. 8, 89–111. http://dx.doi.org/10.1037/a0028793. Main, M., Hesse, E., 1990. Parents’ unresolved traumatic experiences are related to infant disorganized attachment status: is frightened and/or frightening parental behavior the linking mechanism? In: Greenberg, M.T., Cicchetti, D., Cummings, E.M. (Eds.), Attachment in the Preschool Years: Theory, Research, and Intervention. University of Chicago Press, Chicago, pp. 161–182. Main, M., Hesse, E., 1992. Disorganized/disoriented infant behavior in the strange situation, lapses in the monitoring of reasoning and discourse during the parent's adult attachment interview, and dissociative states. In: Attachment and Psychoanalysis, pp. 86–140. Main, M., Solomon, J., 1990. Procedures for identifying infants as disorganized/disoriented during the Ainsworth strange situation. In: Greenberg, M.T., Cicchetti, D., Cummings, E.M. (Eds.), Attachment in the Preschool Years: Theory, Research, and Intervention. University of Chicago Press, Chicago, pp. 121–160. Manassis, K., Bradley, S., Goldberg, S., Hood, J., Swinson, R.P., 1994. Attachment in mothers with anxiety disorders and their children. J. Am. Acad. Child Psychiatry 33, 1106–1113. Mileva-Seitz, V.R., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., 2016. Genetic mechanisms of parenting. Horm. Behav. 77, 211–233. http://dx.doi.org/10.1016/j. yhbeh.2015.06.003. Obradovic, J., Bush, N.R., Stamperdahl, J., Adler, N.E., Boyce, W.T., 2010. Biological sensitivity to context: the interactive effects of stress reactivity and family adversity on socioemotional behavior and school readiness. Child Dev. 81, 270–289. http://dx. doi.org/10.1111/j.1467-8624.2009.01394.x. Pederson, D.R., Moran, G., Sitko, C., Campbell, K., Ghesquire, K., Acton, H., 1990. Maternal sensitivity and the security of infant-mother attachment: a Q-sort study. Child Dev. 61, 1974–1983. Pinto, R.J., Maia, A.C., 2013. A comparison study between official records and self-reports of childhood adversity. Child Abuse Rev. 22, 354–366. http://dx.doi.org/10.1002/ car.2232. Pruessner, J.C., Kirschbaum, C., Meinlschmid, G., Hellhammer, D.H., 2003. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28, 916–931. http://dx.doi.org/10.1016/S0306-4530(02)00108-7. Quas, J.A., Bauer, A.M., Boyce, W.T., 2004. Physiological reactivity, social support, and memory in early childhood. Child Dev. 75, 797–814. http://dx.doi.org/10.111/j. 1467-8624.00707.x. Reiner, I., Frieling, H., Beutel, M., Michal, M., 2016. Gene-environment interaction of the oxytocin receptor polymorphism (rs53576) and unresolved attachment status predict depersonalization symptoms: an exploratory study. Psychol. Stud. 61, 295–300. Roisman, G.I., Newman, D.A., Fraley, R.C., Haltigan, J.D., Groh, A.M., Haydon, K.C., 2012. Distinguishing differential susceptibility from diathesis-stress: recommendations for evaluating interaction effects. Dev. Psychopathol. 24, 389–409. http://dx. doi.org/10.1017/S0954579412000065. Schechter, D.S., Zeanah, C.H., Myers, M.M., Brunelli, S.A., Liebowitz, M.R., Marshall, R.D., Hofer, M.A., 2004. Psychobiological dysregulation in violence-exposed mothers: salivary cortisol of mothers with very young children pre- and post-separation stress. Bull. Menn. Clin. 68, 319–336. Sroufe, L.A., 2005. Attachment and development: a prospective, longitudinal study from birth to adulthood. Attach Hum. Dev. 7, 349–367. http://dx.doi.org/10.1080/ 14616730500365928. Toth, S.L., Sturge-Apple, M.L., Rogosch, F.A., Cicchetti, D., 2015. Mechanisms of change: testing how preventative interventions impact psychological and physiological stress functioning in mothers in neglectful families. Dev. Psychopathol. 27, 1661–1674. Urizar, G.G., Munoz, R.F., 2011. Impact of a prenatal cognitive behavioral stress management intervention on salivary cortisol levels in low-income mothers and their infants. Psychoneuroendocrinology 36, 1480–1494.

Del Giudice, M., 2017. Statistical tests of differential susceptibility: performance, limitations, and improvements. Dev. Psychopathol. 29, 1267–1278. Dienes, K.A., Hazel, N.A., Hammen, C.L., 2013. Cortisol secretion in depressed, and at-risk adults. Psychoneuroendocrinology 38, 927–940. http://dx.doi.org/10.1016/j. psyneuen.2012.09.019. Dolen, G., Malenka, R.C., 2014. The emerging role of nucleus accumbens oxytocin in social cognition. Biol. Psychiatry 76, 354–355. http://dx.doi.org/10.1016/j. biopsych.2014.06.009. Dolen, G., Darvishzadeh, A., Huang, K.W., Malenka, R.C., 2013. Social reward requires coordinated activity of nucleus accumbens oxytocin and serotonin. Nature 501, 179–184. http://dx.doi.org/10.1038/nature12518. Drury, S.S., Gleason, M.M., Theall, K.P., Smyke, A.T., Nelson, C.A., Fox, N.A., Zeanah, C.H., 2012. Genetic sensitivity to the caregiving context: the influence of 5hhtlpr and BDNF val66met on indiscriminate social behavior. Physiol. Behav. 106, 728–735. http://dx.doi.org/10.1016/j.physbeh.2011.11.014. Duncan, L.E., Keller, M.C., 2011. A critical review of the first 10 years of candidate geneby-environment interaction research in psychiatry. Am. J. Psychiatry 168, 1041–1049. http://dx.doi.org/10.1176/appi.aip.2011.11020191. Dykas, M.J., Cassidy, J., 2011. Attachment and the processing of social information across the life span: theory and evidence. Psychol. Bull. 137, 19–46. http://dx.doi.org/10. 1037/a0021367. Ellenbogen, M.A., Linnen, A., Grumet, R., Cardoso, C., Joober, R., 2012. The acute effects of intranasal oxytocin on automatic and effortful attentional shifting to emotional faces. Psychophysiology 49, 128–137. http://dx.doi.org/10.1111/j.1469-8986.2011. 01278.x. El-Sheikh, M., Harger, J., Whitson, S.M., 2001. Exposure to interparental conflict and children's adjustment and physical health: the moderating role of vagal tone. Child Dev. 72, 1617–1636. http://dx.doi.org/10.1111/1467-8624.00369. Fearon, R.P., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., Lapsley, A.M., Roisman, G.I., 2010. The significance of insecure attachment and disorganization in the development of children's externalizing behavior: a meta-analytic study. Child Dev. 81, 435–456. http://dx.doi.org/10.1111/j.1467-8624.2009.01405.x. Fekedulegn, D.B., Andrew, M.E., Burchfiel, C.M., Violanti, J.M., Hartley, T.A., Charles, L.E., Miller, D.B., 2007. Area Under the Curve and Other Summary Indicators of Repeated Waking Cortisol Measurements. Ferguson, J.N., Young, L.J., Hearn, E.F., Matzuk, M.M., Insel, T.R., Winslow, J.T., 2000. Social amnesia in mice lacking the oxytocin gene. Nat. Genet. 25, 284–288. http:// dx.doi.org/10.1038/77040. Field, A., 2009. Discovering Statistics Using SPSS. SAGE Publications, London. Field, T.M., Scafidi, F., Pickens, J., Prodromidis, M., Pelaez-Nogueras, M., Torquati, J., ... Kuhn, C., 1998. Polydrug-using adolescent mothers and their infants receiving early intervention. Adolescence 33, 117–143. Gervai, J., Novak, A., Lakatos, K., Toth, I., Danis, I., Ronai, Z., et al., 2007. Infant genotype may moderate sensitivity to maternal affective communications: attachment disorganization, quality of care, and the DRD4 polymosphism. Soc. Neurosci. 2, 1–13. http://dx.doi.org/10.1080/17470910701391893. Gonzalez, A., Jenkins, J.M., Steiner, M., Fleming, A.S., 2012. Maternal early life experiences and parenting: the mediating role of cortisol and executive function. J. Am. Acad. Child Adolesc. Psychiatry 51, 673–682. Hankin, B.L., Badanes, L.S., Smolen, A., Young, J.F., 2015. Cortisol reactivity to stress among youth: stability over time and genetic variants for stress sensitivity. J. Abnorm. Psychol. 124, 54–67. Hardt, J., Rutter, M., 2004. Validity of adult retrospective reports of adverse childhood experiences: review of the evidence. J. Child Psychol. Psychiatry 45, 260–273. http:// dx.doi.org/10.1111/j.1469-7610.2004.00218.x. Haskett, M.E., Ahern, L.S., Ward, C.S., Allaire, J.C., 2006. Factor structure and validity of the parenting stress index-short form. J. Clin. Child Adolesc. Psychol. 35, 302–312. http://dx.doi.org/10.1207/s15374424jccp3502_14. Hutchinson, K.E., Stallings, M., McGeary, J., Bryan, A., 2004. Population stratification in the candidate gene study: fatal threat or red herring? Psychol. Bull. 130, 66–79. http://dx.doi.org/10.1037/0033-2909.130.1.66. Johnson, Z.V., Young, L.J., 2015. Neurobiological mechanisms of social attachment and pair bonding. Curr. Opin. Behav. Sci. 3, 38–44. http://dx.doi.org/10.1016/j.cobeha. 2015.01.009. Keebaugh, A.C., Young, L.J., 2011. Increasing oxytocin receptor expression in the nucleus accumbens of pre-pubertal female prairie voles enhances alloparental responsiveness and partner preference formation as adults. Horm. Behav. 60, 498–504. http://dx. doi.org/10.1016/j.yhbeh.2011.07.018. Keebaugh, A.C., Barrett, C.E., Laprairie, J.L., Jenkins, J.J., Young, L.J., 2015. RNAi knockdown of oxytocin receptor in the nucleus accumbens inhibits social attachment and parental care in monogamous female prairie voles. Soc. Neurosci. 10, 561–570. http://dx.doi.org/10.1080/17470919.2015.1040893. Keller, M.C., 2014. Gene-by-environment interaction studies have not properly controlled for potential confounders: the problem and the (simple) solution. Biol. Psychiatry 75, 18–24. http://dx.doi.org/10.1016/j.biopsych.2013.09.006. Khoury, J.E., Gonzalez, A., Levitan, R., Pruessner, J.C., Chopra, K., Basile, V.S., ... Atkinson, L., 2015. Summary cortisol reactivity indicators: interrelations and meaning. Neurobiol. Stress. 2, 34–43. http://dx.doi.org/10.1016/j.ynstr.2015.04. 002. Khoury, J.E., Gonzalez, A., Levitan, R., Masellis, M., Basile, V., Atkinson, L., 2016. Maternal self-reported depressive symptoms and maternal cortisol levels interact to predict infant cortisol levels. Infant Ment Health J. 37, 125–139. http://dx.doi.org/ 10.1002/imhj.21554. Kim, H.S., Sherman, D.K., Sasaki, J.Y., Xu, J., Chu, T.Q., Ryu, C., ... Taylor, S.E., 2010. Culture, distress, and oxytocin receptor polymorphism (OXTR) interact to influence emotional support seeking. PNAS 107, 15717–15721. http://dx.doi.org/10.1073/

32

Hormones and Behavior 102 (2018) 23–33

J.A. Ludmer et al.

org/10.1037/a0027536. van IJzendoorn, M.H., Schuengel, C., Bakermans-Kranenburg, M.J., 1999. Disorganized attachment in early childhood: a meta-analysis of precursors, concomitants, and sequale. Dev. Psychopathol. 11, 225–249. Vreeburg, S.A., Zitman, F.G., van Pelt, J., DeRijk, R.H., Verhagen, J.N.M., van Dyck, R., ... Pennix, B.W.J.H., 2010. Salivary cortisol levels in persons with and without different anxiety disorders. Psychosom. Med. 72, 340–347.

van IJzendoorn, M.H., Bakermans-Kranenburg, B.K., 2006. DRD4 7-repeat polymorphism moderates the association between maternal unresolved loss or trauma and infant disorganization. Attach Hum. Dev. 8, 291–307. http://dx.doi.org/10.1080/ 14616730601048159. van IJzendoorn, M.H., Bakermans-Kranenburg, B.K., 2012. Differential susceptibility experiments: going beyond correlational evidence: comment on beyond mental health, differential susceptibility articles. Dev. Psychol. 48, 769–774. http://dx.doi.

33