Telomere Length and Psychopathology: Specificity and Direction of Effects within the Bucharest Early Intervention Project

Abstract

Objective:

Telomere length has been linked to several psychiatric conditions in children and adults. Telomere shortening is accelerated by early adversity, including maltreatment and psychosocial deprivation. These experiences also increase the risk of psychopathology in many domains. Two fundamental issues remain unresolved. The first concerns the specificity of the relations between TL and different dimensions of psychopathology; and the second relates to the direction of association between TL and psychopathology.

Method:

This study addresses these shortcomings in a two-fold manner. First, we examined the association between TL and statistically-independent general, internalizing, and externalizing psychopathology factors to determine the specificity of this relation. Second, we used a two-wave longitudinal cross-lagged model to explicitly examine the direction of the relation between telomere length and each psychopathology factor. Data were drawn from the Bucharest Early Intervention Project (BEIP), a longitudinal study exploring the impact of severe psychosocial deprivation on child health and development (N = 195). At age 8–10 and 12–14, buccal DNA were collected and teachers and/or caregivers reported on different domains of psychopathology.

Results:

Longitudinal path analyses revealed that shorter telomere length was specifically associated with higher internalizing psychopathology at age 8–10. In contrast, at age 12–14, shorter telomere length was associated with higher general psychopathology. Most telling, internalizing psychopathology at age 8–10 predicted shorter telomere length at age 12–14, with no reciprocal effects.

Conclusion:

Results suggest that telomere erosion may be a consequence of distress-related psychopathology rather than a selection mechanism for later psychiatric problems.

Trial Registration:

Clinicaltrials.gov Identifier: NCT00747396

Lay Summary:

This study used data from an ongoing prospective study exploring the long-term outcomes of children raised in Romanian institutions or removed and placed into families, the Bucharest Early Intervention Project. We were interested in the relation between mental health problems and a biological marker of cellular aging, telomere length. Telomeres typically shorten with age, but many factors can influence this rate. A growing literature suggests that early adversity is one potent factor influencing telomere length trajectory. In the current study we show that children experiencing more internalizing problems in late childhood have shorter telomeres in early adolescence. This data indicates that early internalizing problems may influence the rate of cellular aging, providing insights into one possible mechanism linking early mental illness to later physical health problems.

Introduction

Telomeres are nucleoprotein structures consisting of repetitive DNA sequences, proteins, and non-coding RNAs that preserve chromosome integrity. Telomeres naturally shorten with each mitotic division due to the “end-replication problem” of DNA polymerase.¹ Once telomeres reach a critical length, the unprotected ends trigger cells to enter a senescent state or undergo apoptosis. In neural precursor cells, telomere length influences the timing and direction of cellular differentiation, suggesting a role for telomeres in cellular aging, differentiation, and survival.²

Erosion of telomeres is accelerated by processes such as oxidative stress, which partially account for the gap between observed rates of telomere shortening and those predicted by the end-replication problem.² Environmental stress, including perceived psychological distress, has been associated with shorter telomere length in a number of human studies, including meta analyses.^3–5 In both children and adults, different types of childhood adversity are associated with shorter telomere length.^6–8 For example, institutionally-reared children raised under conditions of profound psychosocial deprivation have been shown to have shorter telomeres as a function of time spent in institutions,⁹ as well as accelerated telomere shortening from age 6 to 15 compared to never-institutionalized children.¹⁰ These results support the hypothesis that stress-related telomere attrition begins in childhood and persists across the life span.

Emerging evidence suggests that shorter telomeres are associated with multiple psychiatric conditions, including stress-related disorders such as PTSD and anxiety, mood disorders including depression and bipolar, externalizing disorders, and neurodevelopmental disorders such as autism.^11–16 These findings suggest that telomere length may be transdiagnostically linked to psychopathology. However, in assessing the link between telomere length and discrete disorders, most previous studies failed to control for other domains of psychopathology. Thus, it is unclear whether telomere length is a transdiagnostic marker of general psychopathology risk, or whether it is related to a particular psychiatric syndrome which may have been masked due to confounding by other disorders. A recent comprehensive review provides support for the hypothesis that telomere length is most strongly related to stress-related psychiatric disorders and indexes both adversity exposure and future health risk.¹⁷

Recent reports on the latent structure of psychopathology posit both shared causal influences and unique liabilities to different psychiatric problems.¹⁸ This presents an opportunity for an innovative approach to resolve the issue of specificity with respect to telomere length and psychopathology. The first goal of the current study was to examine the association between telomere length and latent general and specific internalizing and externalizing psychopathology factors. The presence of these factors has been replicated in adults, adolescents, and children.^19,20 The general factor is believed to reflect high negative emotionality and emotion dysregulation that underpins many conditions,²¹ while the internalizing factor captures elements of fear and distress, and the externalizing factor indexes approach-related traits such as sensation-seeking and surgency.^22,23 By simultaneously estimating general, internalizing, and externalizing factors, we are able to examine the specificity of the association between telomere length and each latent factor, thus improving our understanding of the fundamental traits related to telomere length.

A second challenge in the extant literature is the direction of the relation between telomere length and psychopathology. Two competing models have been proposed.²⁴ The first suggests that psychopathology leads to telomere erosion over time (“behavioral causation”). This model is consistent with the idea that stress related to psychopathology is associated with physiological and cellular alterations related to telomere length, such as elevated oxidative stress or inflammation.²⁵ The second model is one of reverse causation, in which individuals with shorter telomeres adopt patterns of behavior or neurobiologically-driven responses to the environment (“selective adoption”). This model predicts that shorter telomere length predisposes individuals to later psychopathology. For instance, Gotlib and colleagues have demonstrated that healthy daughters of depressed women have shorter telomeres compared to daughters of never-depressed woman, suggesting that telomere shortening antedates, and may be a risk factor for, depression.²⁶

Unfortunately, the preponderance of research on the relation between telomere length and psychopathology has been cross-sectional, thus precluding directional hypotheses. To date, no study in children or adolescents has assessed both telomere length and psychopathology across multiple waves of data collection and examined their cross-lagged associations. Cross-lagged models test for reciprocal relations between constructs once within-time covariance terms and cross-time stability coefficients are controlled, leading researchers closer to establishing causal effects.²⁷ In the current study, we used a two-wave longitudinal cross-lagged model to examine associations between telomere length and latent general, internalizing, and externalizing psychopathology factors from age 8 to 14, thereby permitting a direct simultaneous test of the competing “behavioral causation” and “selective adoption” hypotheses.

Method

Participants

Participants were children enrolled in the Bucharest Early Intervention Project (BEIP), a longitudinal randomized controlled trial (RCT) that compared foster care to care-as-usual for children raised in Romanian institutions, the details of which are described elsewhere.²⁸ Briefly, between 6 and 31 months of age, 136 children residing in six institutions in Romania underwent baseline assessments and were then randomly assigned to a care-as-usual group (CAUG; n = 68) or foster care group (FCG; n = 68). Together, these groups comprise the ever-institutionalized group (EIG). Following randomization, all subsequent decisions regarding placement were made by the child protection commissions in Romania. Ethical considerations for this project have been described elsewhere.²⁹ A comparison group (n = 72) of never-institutionalized children (NIG) were recruited from the same maternity hospitals in which EIG children were born or from area schools at later time points. Children in BEIP have been followed longitudinally in approximately four-year intervals. The current study used data from the 8 and 12 year follow-up assessments, at which point data on psychopathology and telomere length were collected (described below). There was some variability in the age at which data were collected. Psychopathology data were collected at a mean age of 8.51 (SD = .70; range = 7 to 10) and 13.0 (SD = .41; range = 11 to 15). DNA for telomere analyses was collected at 4 time points, with time 1 and 2 corresponding to a mean age of 8.68 (SD = .60; range = 7.3 to 11.1) and time 3 and 4 corresponding to a mean age of 13.04 (SD = .74; range = 11.1 to 14.5). We henceforth use the labels “age 8–10” and “age 12–14” to describe these periods of assessment.

The Institutional Review Boards of Boston Children’s Hospital, University of Maryland, and Tulane University approved this study. Furthermore, as mandated by Romanian law, informed consent for all participants was provided through the Commission on Child Protection.

Psychopathology assessment

The MacArthur Health and Behavior Questionnaire (HBQ) was used to collect data pertaining to several domains of psychopathology.³⁰ At age 8–10, reporters were the children’s teachers. At age 12–14, both teachers and caregiver reported on problems, which were combined into a composite score to reduce rater bias. Teachers and caregivers responded to several items on 3-point Likert scales: 0 (“never or not true”), 1 (“sometimes true”), and 2 (“often or very true”). Subscales included: depression (18 items), overanxious (14 items), social anxiety/withdrawal (11 items), oppositional defiant (9 items), conduct problems (14 items), overt aggression (8 items), relational aggression (7 items), and ADHD (18 items). Items details can be found on the MacArthur Foundation Research Network for Psychopathology and Development website. Means were taken across the items to generate a scale score for each psychopathology domain. These psychopathology ratings were then subjected to latent bifactor models to estimate general, internalizing, and externalizing factors, which were derived from a previous investigation.³¹ This previous study established measurement invariance on these factors over time. Moreover, we examined differential item functioning (DIF) in a multiple-cause multiple-indicator (MIMIC) model by testing direct associations between institutional rearing (NIG = 0; EIG = 1) and each indicator of the psychopathology factors.³² There was a single direct association between institutionalization status and relational aggression at age 8 only, generally supporting invariance across groups.

Telomere assessment

Collection of telomere data followed the approach previously described by Drury and colleagues.⁹ DNA was extracted from Isohelix buccal swabs (Cell Projects, Kent, UK) collected in Romania and shipped to the United States. Swabs were air dried, stored with a desiccator pellet, and frozen until extracted. Integrity of genomic DNA and purity was assessed via nanodrop, QuBit, and agarose gel electrophoresis. Average relative buccal TL was determined from the T/S ratio using an adapted monochrome multiplex quantitative real-time PCR (MMP-qPCR) and a BioRad CFX96.⁹ Samples were performed in triplicate, with a 7-point standard curve (0.0313ng to 2ng) derived from a single pooled control buccal DNA sample. Triplicate plates were repeated with all samples in a different well position on the duplicate plate and all samples from the same individual were run on the same plate. All time points from each individual were run on the same plates to decrease batch or plate effects.

PCR efficiency criteria for both reactions were 90–110%. Coefficients of variations were calculated within each triplicate (CV criteria ≤ 10%) and between plates (CV criteria ≤ 6%). Samples with unacceptably high CVs were removed from analysis or repeated. The telomere length ratio was derived by the average of the triplicates from both plates. The CV for samples was 2.4%.

To account for variability in chronological age, we residualized telomere length scores for child age within time, but not across time, allowing for cross-time associations to be evaluated. As telomere length is also associated with a child’s physical development, we further residualized telomere length scores for pubertal status at the 12–14 year assessment, assessed using the Tanner stage self-assessment measure.³³ Thus, the telomere scores were age- and puberty-adjusted within time.

Statistical analyses

Both EIG and NIG children were included in the analyses. For telomere length measurement, individuals contributed anywhere from 1 to 4 data points. As noted above, telomere data at time 1 and 2 were averaged to provide age 8–10 telomere length values, and data at time 3 and 4 were averaged to provide age 12–14 values. At age 8–10, 195 children contributed psychopathology data; at age 12–14, 162 children contributed data. There was some missing data across participants over time. To handle this without compromising power, the analyses were conducted using full-information maximum likelihood estimation (FIML) in Mplus 7.3. FIML has been shown to outperform other methods for handling missing data, including listwise deletion, pairwise deletion, and mean substitution in terms of parameter bias, model convergence, and model fit.³⁴ We used a maximum-likelihood with robust standard errors (MLR) estimator, which yields parameter estimates with standard errors and a chi-square that are robust to non-normality when missing data are present.

The primary model examined the association between telomere length and general, internalizing, and externalizing factors both within and across time using a two-wave longitudinal cross-lagged path model in which reciprocal relations between constructs at age 8–10 and age 12–14 were freely estimated. This longitudinal path model also permitted within-time correlations between telomere length and each psychopathology factor to freely vary. Institutionalization status (NIG = 0; EIG = 1) was included as a covariate, with all variables regressed onto this variable.

Results

Descriptive statistics and group differences across study variables are presented in Table 1. Bivariate correlations are presented in Table 2. As shown in these tables, there were group differences in birth weight and IQ. Also, birth weight, IQ, and sex were variably associated with psychopathology. Thus, we re-ran separate path models with each of sex, IQ, and birth weight as covariates and the results did not differ substantively. As seen in Table 1, there were no direct associations between telomere length at either time point and group status – either comparing NIG to EIG, or FCG to CAUG. As seen in Table 2, there was rank-order stability in TL over time, meaning individuals with shorter telomeres at age 8–10 tended to have shorter telomeres at age 12–14 compared to those with longer telomeres. Moreover, consistent with previous studies, there was telomere shortening over time across the whole sample, t (df = 76) = 3.01, p = .004.

Table 1:

Characteristics and Descriptive Statistics Among Institutionalized and Never-institutionalized Children

Age 8–10 Measures
	All participants			Institutionalized children only
	EIG (n= 96)	NIG (n= 99)	t or χ2	CAUG (n= 44)	FCG (n= 52)	t or χ2
Sex (% Male)	52.1%	48.5%	.25	52.3	51.9	.001
Birth weight (g)	2755.2	3208.0	5.06***	2847.5	2675.0	1.33
Full Scale IQ	80.2	100.2	9.30***	78.2	81.9	1.29
General factor	.45	−.44	6.84***	.51	.40	.57
Internalizing factor	.04	−.04	.63	.12	−.02	.69
Externalizing factor	.00	.00	.06	.06	−.04	.44
Telomere Length	1.60	1.59	.06	1.53	1.65	.74
Age 12–14 Measures
	All participants			Institutionalized children only
	EIG (n= 112)	NIG (n= 50)	t or χ2	CAUG (n= 56)	FCG (n= 56)	t or χ2
Sex (% Male)	51.8	46.0	.46	51.8	51.8	<.001
Birth weight (g)	2799.5	3244.9	4.29***	2878.6	2722.0	1.31
Full Scale IQ	72.3	98.1	9.00***	68.6	75.8	2.19*
General factor	.20	−.44	4.29***	.34	.06	1.52
Internalizing factor	−.05	.11	1.22	−.01	−.09	.58
Externalizing factor	.16	−.37	4.44***	.23	.10	.63
Telomere Length	1.40	1.43	.64	1.38	1.41	.59

Note: All statistics are means except for sex (%). General, internalizing, and externalizing variables are factor scores with a total sample mean of zero.

IQ was assessed using the WISC-IV. Birth weight is measured in grams. Telomere length scores are unadjusted in this table.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Table 2:

Bivariate Correlations Between Study Variables and Other Child Characteristics

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.
1. General factor (age 8–10)	–
2. Internalizing factor (age 8–10)	.00	–
3. Externalizing factor (age 8–10)	.00	.02	–
4. Telomere length (age 8–10)	−.02	−.19§	.03	–
5. General factor (age 12–14)	.29**	.00	−.09	.18	–
6. Internalizing factor (age 12–14)	−.23**	.12	.04	.03	.15§	–
7. Externalizing factor (age 12–14)	.37***	−.04	.06	−.02	.04	−.25**	–
8. Telomere length (age 12–14)	−.05	−.25**	.08	.35**	−.22**	−.05	−.01	–
9. Sex (male)	.15*	−.07	−.05	.01	.16*	.01	.24**	−.01	–
10. Birth weight	−.23**	−.09	.23**	.01	−.22**	−.13	−.11	−.16§	.09	–
11. FSIQ (age 8–10)	−.54***	−.10	.05	.04	–	–	–	–	−.08	.34***	–
12. FSIQ (age 12–14)	–	–	–	–	−.31***	.12	−.32***	.02	−.08	.29**	.90***

^§p < .10.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Given previous findings of sex-specific relations between duration of institutional care and telomere length, we present correlations between the length of time spent in institutions among EIG (percentage of one’s life up to a given age) and telomere length for male and female participants in Table S1, available online. Greater duration of institutional care by 54 months and 8 years of age was associated with shorter telomere length at age 8–10 in male participants. In female participants, greater duration of institutional care at baseline and 54 months was associated with shorter telomere length at age 12–14. The amount of time spent in institutions over time was correlated (Table S2, available online). When duration of institutional care and telomere length were examined within FCG and CAUG (accounting for sex), greater exposure to institutional care at 54 months, 8 years, and 12 years was associated with shorter telomere length at age 12–14 among CAUG, but not FCG (Table S3, available online). Sex differences were not examined within the FCG and CAUG groups due to power limitations. We expand on these findings in Supplement 1. Thus, there are associations between duration of institutionalization and telomere length that vary based on sex and age of assessment rather than direct associations between institutionalization group and telomere length. The strong relation between institutionalization status (NIG = 0; EIG = 1) and psychopathology justifies its inclusion as a covariate in the primary model, outlined below.

Association between telomere length and psychopathology

The longitudinal path model describing the relation between telomere length and the general, internalizing, and externalizing factors is presented in Figure 1. Model fit was acceptable: RMSEA = .069 [.02, .11], PCLOSE = .22, CFI = .90, and SRMR = .05. EIG children had significantly higher general psychopathology at age 8–10 than NIG, yet institutionalization status was unrelated to internalizing or externalizing at age 8–10 after accounting for the general factor. Furthermore, there was relative stability in general psychopathology over time, regardless of institutional status, but little stability in internalizing or externalizing after accounting for stability in the general factor.

Figure 1. Longitudinal Path Model Linking Telomere Length and Psychopathology Factor Scores at Age 8–10 and Age 12–14.

Note. Parameters are standardized β coefficients. Psychopathology factors are factor scores saved from the latent bi-factor model and used as manifest variables. Only relations between telomere length and psychopathology are shown. See Table 1 for associations between psychopathology factor scores. EIG = ever-institutionalized group; NIG = never-institutionalized.

*p < .05; **p < .01; ***p < .001.

In terms of within-time associations, shorter telomere length at age 8–10 was significantly associated with higher internalizing problems, but was unrelated to general or externalizing psychopathology. On the other hand, shorter telomere length at age 12–14 was associated with higher general, but not internalizing or externalizing, psychopathology. Finally, across-time effects (i.e., cross-lagged associations) revealed that higher internalizing psychopathology at age 8–10 predicted shorter telomere length at age 12–14 after controlling for telomere length at age 8–10. This effect can be interpreted as higher internalizing psychopathology predicting change in telomere length from age 8–10 to age 12–14. Neither general nor externalizing psychopathology at age 8–10 were associated with telomere length at age 12–14, and no reciprocal effects of telomere length on psychopathology emerged. All effects were tested simultaneously and conditional on all other effects in the model. As these were not independent tests, no correction for multiple testing was required.

Finally, consistent with the original intent of the RCT, we examined whether associations between the psychopathology factors and telomere length differed between: (i) EIG and NIG, and (ii) CAUG and FCG. The latter model offered a poor fit to the data, likely due to inadequate power given the small sample size of the CAUG and FCG groups. Results of the former model are provided in Supplement 1, available online. As there were no observable differences in the paths linking psychopathology and telomere length between EIG and NIG children, the primary model above collapsed across all children.

Discussion

The present study was designed to address two issues related to the association between telomere length and psychopathology in a unique sample of children with histories of institutionalization and a cohort of never-institutionalized control children: (i) the specificity of this association with respect to the dimension of psychopathology; and (ii) the direction of association. We provide the first evidence that internalizing problems are associated with shorter telomere length in late childhood (age 8–10); yet in early adolescence (age 12–14), shorter telomeres are associated with higher general psychopathology. The former finding is consistent with work showing that shorter telomeres are cross-sectionally related to more internalizing problems in adults,³⁵ and studies showing that shorter telomeres are associated with persistent internalizing problems from age 11 to 38.³⁶ The current study complements these existing studies by demonstrating the association between internalizing psychopathology and shorter telomere length in middle childhood. These effects remain significant after accounting for what is shared between all domains of psychopathology, institutional rearing, and other covariates.

The finding that shorter telomere length is associated with higher general psychopathology at age 12–14 is especially novel, as this was the first study to examine telomere length in relation to a general factor that captures the shared variance across domains of psychopathology. These results suggest that the relation between telomere length and psychopathology may become more diffuse and non-specific across development. Given that the general psychopathology factor is conceptualized as a latent liability to psychopathology that is characterized by high negative emotionality and deficits in emotion regulation, our results suggest that, by early adolescence, shorter telomeres may be a marker of poor emotion regulation or negative emotionality that is associated with several psychiatric difficulties. Supporting this idea, negative emotionality has recently been shown to mediate the link between family risk in adolescence and shorter telomere length in early adulthood.³⁷ These results also help to explain previous findings linking telomere length to multiple psychiatric conditions by suggesting that, during the transition to adolescence, telomere length is transdiagnostically linked to psychopathology through the general factor. In other words, in adolescence telomere length is related to those features of psychopathology that are shared across disorders, rather than features unique to internalizing or externalizing problems.

The second major goal of the current study was to determine the direction of the association between telomere length and psychopathology. Our results revealed a single directional effect between the internalizing factor at age 8–10 and telomere length at age 12–14, with no unique prediction of the general of externalizing factors, and no reciprocal effects of telomere length on later psychopathology. These effects were not explained by children’s history of institutional rearing, indicating that the predictive relation between early internalizing psychopathology and later telomere length is applicable to both typically-developing children and those with histories of significant adversity. In addition, these results were robust to consideration of pubertal status, sex, and other relevant covariates. These results provide strong evidence of a directional relation between distress-related psychopathology and telomere attrition, and thus more firmly support the “behavioral causation” over “selective adoption” hypothesis.

On the surface, the current results differ from those of Gotlib and colleagues.²⁶ In their study, the daughters of depressed mothers, who had not developed depression themselves, had shorter telomeres compared to daughters of non-depressed mothers, suggesting that telomere length may be a vulnerability marker for future psychopathology, a result not supported by our data. A contrasting hypothesis is that, within families at elevated genetic risk for internalizing psychopathology, this same genetic predisposition overlaps with genetic factors influencing telomere dynamics. Alternatively, depression liability within high genetic risk families may be less mechanistically linked to the stress-activated pathways associated with telomere length and thought to influence negative health trajectories under conditions of adversity. Importantly, however, the current study cannot rule out long-term transactional effects between telomere length and psychopathology, with shorter telomeres as a function of increased internalizing psychopathology forecasting later risk for psychopathology. Prospective studies with multiple waves of data collection are required to address these questions. Also, since telomere attrition is greatest during the first years of life,³⁸ future studies with younger children are needed to inform how the relation between psychopathology and telomere length varies across development, and whether these effects are more pronounced earlier in life or during sensitive periods. The particular importance of experience during the first years of life for telomere length is demonstrated by our finding that telomere length in middle childhood and early adolescence was associated with the amount of time children spent in institutions before the transition to adolescence.

Several potential mechanisms have been proposed to explain the link between distress-related psychopathology and telomere length. For example, inflammation has been shown to decrease leukocyte telomere length by increasing leukocyte turnover and accelerating replicative senescence.³⁹ A related mechanism involves changes in cellular redox potential (e.g., oxidative stress), in which reactive oxygen species preferentially damage DNA at telomeres, disrupt telomerase activity (the enzyme that lengthens telomeres), and/or influence epigenetic factors such as methylation and phosphorylation.^40,41 These processes have also been implicated in the pathophysiology of distress-related psychopathology (i.e., anxiety and depression).⁴² Consistent with the current pattern of results, it has been suggested that both pro-inflammatory processes and oxidative stress mediate the relation between depression and telomere erosion.⁴³ Finally, cortisol-mediated stress reactivity is associated with accelerated telomere shortening,²⁶ the effects of which may operate by disrupting telomerase activity or increasing oxidative stress.^3,44 These results suggest that the complex and inter-related biological processes of inflammation, oxidative stress, and neuroendocrine functioning may together contribute to telomere erosion in the context of internalizing psychopathology.

Some limitations of this study are noteworthy. First, with respect to telomere measurement, DNA was not available for all study participants at all time points. This was the result of differential participation across the course of the study, as well as our stringent quality control requirements for the telomere assay.¹⁰ Briefly, in order for any telomere measurements from an individual to be included, all six replicates, from all time points, had to pass our quality control metrics simultaneously. If any time point or any replicate failed these metrics, no time points were included. Second, telomere length was measured through buccal DNA, and other studies have used salivary- or blood-derived DNA. While tissue specificity is an ongoing challenge for telomere studies, there is no evidence to date that one biological source is superior, and there is also a high correlation in telomere length measured across tissue types.⁴⁵ Although true correlation with telomere length in the brain has not been established, buccal swabs measure telomeres in cells with the same embryologic origin as the neural system, the ectoderm, and therefore may be more reflective of epigenetic processes in the brain than blood or saliva, which contain immune cells that are mesodermally-derived. Third, psychopathology was assessed at age 8–10 using teacher-only reports, whereas at age 12–14 was measured using a combination of teacher and caregiver reports. The latter reduces rater bias, and sole reliance on teacher reports for earlier psychopathology is not ideal. However, we were able to successfully reproduce the latent psychopathology factors at both time points and have previously demonstrated measurement invariance.³¹ Nonetheless, replication using interviewer-based diagnostic assessments is encouraged. Fourth, missing data across participants may have introduced bias. The potential for bias was minimized by using best-practice methods for handling missing data. Post-hoc analyses using bootstrap re-sampling methods yielded similar, though more conservative, results, supporting the robustness of effects. Finally, despite employing an RCT design, we did not demonstrate a direct relation between group status (EIG versus NIG, or FCG versus CAUG) and telomere length at either time point. However, consistent with our previous findings in early childhood, telomere length in middle childhood and early adolescence was related to greater amount of time spent in institutional care, suggesting a possible dose-repose relation between the extent of early deprivation and telomere length later in development.

Psychiatric illness frequently co-occurs with age-related medical illnesses such as cardiovascular disease and diabetes. Since telomere erosion is associated with increased morbidity and mortality from a number of causes,⁴⁶ it is plausible that telomere attrition as a function of distress-related psychopathology contributes to increased rates of premature mortality among individuals with psychiatric conditions.⁴⁷ Interventions designed to reduce the stress-related effects of psychopathology or target telomere length directly (e.g. exercise, meditation, and antioxidants) may engender positive effects on physical and mental health by attenuating telomere erosion and/or increasing telomerase activity.^48–50 More research is needed to determine how both pharmacological and psychosocial treatments protect against telomere attrition, and the downstream consequences of this for health and development.

Clinical Guidance:

• Children experiencing early neglect and social deprivation are at risk for more mental health problems, and early internalizing difficulties may accelerate cellular aging and predict future physical health risk.

• In addition to social, family, and academic difficulties that accompany higher levels of psychopathology, there may also be biological consequences of heightened internalizing problems, particularly with respect to cellular aging, which are often over looked in clinical settings.

• Clinicians should be aware of the bi-directional relation between physical and mental health, particularly for children with early internalizing disorders. For pediatricians, routine screening for early behavioral and socioemotional problems may provide an important indicator of risk for later physical health difficulties. Similarly, for child mental health professionals, awareness and monitoring of physical health trajectories is warranted, with the expectation that addressing mental health may translate into improved physical health outcomes across the life span.