Psychological Medicine, Page 1 of 11.

© Cambridge University Press 2016 OR I G I N A L A R T I C L E

doi:10.1017/S0033291716002403

Effects of early life stress on depression, cognitive

performance and brain morphology

A. Saleh1, G. G. Potter2, D. R. McQuoid2, B. Boyd1, R. Turner1, J. R. MacFall3 and W. D. Taylor1,4*

1
The Center for Cognitive Medicine, Department of Psychiatry, Vanderbilt University Medical Center, Nashville, TN 37212, USA
2
Department of Psychiatry and Behavioral Sciences, Duke University Medical Center, Durham, NC 27710, USA
3
Department of Radiology, Duke University Medical Center, Durham, NC 27710, USA
4
The Geriatric Research, Education, and Clinical Center (GRECC), Department of Veterans Affairs Medical Center, Tennessee Valley Healthcare
System, Nashville, TN 37212, USA

Background. Childhood early life stress (ELS) increases risk of adulthood major depressive disorder (MDD) and is
associated with altered brain structure and function. It is unclear whether specific ELSs affect depression risk, cognitive
function and brain structure.

Method. This cross-sectional study included 64 antidepressant-free depressed and 65 never-depressed individuals. Both
groups reported a range of ELSs on the Early Life Stress Questionnaire, completed neuropsychological testing and 3T
magnetic resonance imaging (MRI). Neuropsychological testing assessed domains of episodic memory, working mem-
ory, processing speed and executive function. MRI measures included cortical thickness and regional gray matter
volumes, with a priori focus on the cingulate cortex, orbitofrontal cortex (OFC), amygdala, caudate and hippocampus.

Results. Of 19 ELSs, only emotional abuse, sexual abuse and severe family conflict independently predicted adulthood
MDD diagnosis. The effect of total ELS score differed between groups. Greater ELS exposure was associated with slower
processing speed and smaller OFC volumes in depressed subjects, but faster speed and larger volumes in non-depressed
subjects. In contrast, exposure to ELSs predictive of depression had similar effects in both diagnostic groups. Individuals
reporting predictive ELSs exhibited poorer processing speed and working memory performance, smaller volumes of the
lateral OFC and caudate, and decreased cortical thickness in multiple areas including the insula bilaterally. Predictive
ELS exposure was also associated with smaller left hippocampal volume in depressed subjects.

Conclusions. Findings suggest an association between childhood trauma exposure and adulthood cognitive function
and brain structure. These relationships appear to differ between individuals who do and do not develop depression.

Received 3 December 2015; Revised 15 August 2016; Accepted 17 August 2016

Key words: Brain volumes, depression, magnetic resonance imaging, neuropsychological testing, trauma.

Introduction associated with exposure to specific stressors or

whether any childhood trauma could increase vulner-
Childhood stress and trauma significantly influence
ability to depression. Although stressors in childhood
the risk of adult psychopathology. Early life stress
or adolescence contribute to a wide range of adulthood
(ELS) exposure lowers the threshold for depressive reac-
psychopathology, some studies associate major depres-
tions to stressors later in life (Harkness et al. 2006), while
sive disorder (MDD) with exposure to certain types
the intensity of ELS predicts symptom severity of mood
of ELSs including sexual abuse (Kaplow & Widom,
episodes (Martins et al. 2014). ELS exposure can have
2007), emotional abuse (Martins et al. 2014) and family
long-lasting effects on hypothalamic–pituitary–adrenal
conflict (Kessler & Magee, 1994). In contrast, others
(HPA) axis regulation (McEwen, 2003), which may in
concluded there is insufficient evidence associating
part explain the relationship between ELSs and depres-
specific childhood adversities with specific psychiatric
sion. It is currently unclear whether depression is
disorders (Gershon et al. 2013). Such potential ELS–dis-
order specificity is probably related to more than just
the occurrence of the stressor and may be related to
* Address for correspondence: W. D. Taylor, M.D., M.H.Sc., the intensity, duration, developmental stage of the vic-
Vanderbilt University, 1601 23rd Avenue South, Nashville, TN 37212, tim and physiological stress response at the time of the
USA.
stressor. Presumably only stressors resulting in signifi-
(Email: warren.d.taylor@vanderbilt.edu)
This research was previously presented as a poster at the American
cant or sustained stress responses characterized by
Psychiatric Association’s 168th annual meeting in Toronto, ON, HPA axis or other immune system activity would
Canada, in March 2015. influence vulnerability to psychiatric illnesses. Such

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 2 A. Saleh et al.

stressors would also be expected to be associated with (M.I.N.I., version 5.0) (Sheehan et al. 1998) and inter-
cognitive performance and magnetic resonance im- view with a psychiatrist. Additional inclusion criteria
aging (MRI) differences. included onset of first depressive episode before age
Poorer cognitive performance is increasingly recog- 35 years and a Montgomery–Åsberg Depression
nized as an important aspect of MDD, characterized Rating Scale (MADRS) (Montgomery & Åsberg, 1979)
by poor performance on measures of executive func- score of 15 or greater. Inclusion criteria specified no
tion, processing speed and episodic memory (Snyder, psychotropic medication use in the last month; most
2013). ELS exposure is associated with poorer adult subjects reported no use for at least 3 months or longer.
cognitive function in memory domains and executive Eligible control subjects had neither a history of psychi-
function in populations with and without psychopath- atric disorders nor a history of psychotropic medica-
ology (Navalta et al. 2006). However, not all studies as- tion use. Although not an entry criterion, medical
sociate ELS exposure with poorer cognitive performance co-morbidity was quantified using the Cumulative
and much of this work focuses on post-traumatic stress Illness Rating Scale (CIRS) (Miller et al. 1992).
disorder. It is thus unclear whether ELS exposure Exclusion criteria included other lifetime DSM-IV
influences cognitive performance in MDD. Axis I disorders including substance abuse or depend-
More consistently, even in individuals without psy- ence, although co-morbid anxiety symptoms occurring
chiatric disorders, neuroimaging studies associate in context of depressive episodes were allowable.
ELS exposure with volumetric and functional altera- Subjects were excluded for Axis II disorders assessed
tions in brain regions including the anterior cingulate by the Structured Clinical Interview for DSM-IV Axis
cortex (ACC) (Cohen et al. 2006; Udo et al. 2012), med- II Personality Disorders (SCID-II) (First et al. 1997).
ial prefrontal cortex (van Harmelen et al. 2010), caudate Additional exclusion criteria included: history of
(Cohen et al. 2006) and insula (Baker et al. 2013). In con- psychosis, acute suicidality, use of illicit substances in
trast, depressed patients exposed to ELSs exhibit smaller the last month, electroconvulsive therapy in the last 6
volumes of the orbitofrontal cortex (OFC) and prefrontal months, a family history of bipolar disorder, any un-
cortex (Frodl et al. 2010; Udo et al. 2012) and the hippo- stable medical condition, any history of neurological
campus (Cohen et al. 2006; Udo et al. 2012). Jointly, illness or head injury, or MRI contraindications.
these findings suggest that ELS effects on brain structure Both the Duke University and the Vanderbilt
may differ between healthy and depressed populations. University Institutional Review Boards approved this
Although it is challenging to disentangle the effects of study. All subjects provided written informed consent.
ELS v. the effects of depression itself, such population-
specific findings may provide clues related to depres- Assessment of ELS and estimation of total time
sion vulnerability or resilience. depressed
We hypothesized that specific ELSs are associated
Exposure to childhood stressors was assessed using the
with a diagnosis of MDD in adulthood. We further
self-report Early Life Stress Questionnaire (ELSQ). The
hypothesized that those ELSs associated with MDD
ELSQ was developed from the Child Abuse and
would also be associated with poorer performance on
Trauma Scale (Sanders & Becker-Lausen, 1995),
cognitive tests and structural alterations in brain
which has strong internal consistency, validity, and
regions involved in mood regulation. As those ELSs
test–retest reliability. The ELSQ consists of 19 traumat-
by definition would increase the risk of MDD, we
ic items answered yes or no occurring during child-
also tested for statistical interactions between ELS
hood and adolescent ages up to 17 years. We
and MDD diagnosis to determine whether the effect
modified the ELSQ to ask about the ages during the
of ELS exposure on cognition and brain structure dif-
time exposed for each stressor, allowing us to estimate
fered between depressed and non-depressed groups.
the duration of exposure in years.
Following procedures similar to past reports (Sheline
et al. 1999), duration of lifetime depression was
Method
assessed through detailed clinical interview and acqui-
Subjects sition of medical records. We used a life-charting ap-
proach to anchor each episode, applying diagnostic
Subjects were between 20 and 50 years of age and en-
criteria to each episode.
rolled at Duke University (n = 112) and Vanderbilt
University (n = 17) between April 2008 and December
Neuropsychological testing
2013. Depressed subjects had a Diagnostic and
Statistical Manual of Mental Disorders, 4th edition A trained psychometric technician supervised by a
(DSM-IV) diagnosis of recurrent MDD, as assessed licensed neuropsychologist administered neuropsycho-
by the Mini-International Neuropsychiatric Interview logical testing. Similar to our approach in geriatric

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 Early life stress and depression, cognitive performance and brain morphology 3

depression (Sheline et al. 2010), we created rationally pallidum, putamen and thalamus). Intracranial vol-
constructed composite domain variables from a ume was assessed using the method implemented in
broad test battery. To combine tasks, we created FreeSurfer. We visually inspected the data by overlay-
Z-scores for each measure based on the performance ing the surfaces and subcortical segmentations over the
of all subjects, then averaged the Z-scores for all tests T1 data. Individual slices in each orientation were
within each domain. Internal consistency for each do- assessed for errors. No manual corrections were
main was assessed using Cronbach’s coefficient α needed.
(CoA). This resulted in four composite neuropsycho-
logical measures: (a) episodic memory (logical memory Thickness analyses
1 and 2; Benton Visual Retention Test, number correct;
This was an exploratory approach to test for differ-
Rey’s Verbal Learning Test, total I–V and total VII;
ences not captured using our atlas-based comparisons.
CoA = 0.87); (b) executive function (Controlled Oral
We tested for differences in cortical thickness using
Word Association Test, total score; Trail Making B
FreeSurfer’s QDEC module. We used a general linear
time, reverse scored time to completion; verbal
model (GLM) to test for differences in cortical thick-
fluency, total phonological and semantic; Stroop
ness between groups exposed or not exposed to
color–word interference condition, number completed;
ELSs, including age as a nuisance variable. In this
CoA = 0.75); (c) processing speed (symbol–digit modal-
method, cortical thickness is computed as the shortest
ity, number completed; Trail Making A, reverse scored
distance between any point on the pial surface and the
time to completion; Stroop color naming condition,
gray/white boundary and vice versa; these two values
number completed; CoA = 0.70); and (d) working mem-
are averaged (Fischl & Dale, 2000). Maps were smoothed
ory (digit span forward, number of trials correctly
using the standard Gaussian kernel of 10 mm. We used a
completed; digit span backward, number of trials cor-
GLM to test for differences in cortical thickness between
rectly completed; CoA = 0.75).
diagnostic groups, including age as a nuisance variable.
Correction for multiple comparisons was carried out
MRI acquisition using the Monte Carlo simulation method using an ini-
Due to differences in MRI manufacturers, only MRI tial cluster threshold set at p < 0.01. Data were tested
data acquired at Duke University were included in against an empirical null distribution of maximum clus-
analyses. Cranial MRI was performed using the ter size by running 10 000 synthesized Gaussian noise
eight-channel parallel imaging head coil on a whole- simulations, producing clusters fully corrected for mul-
body MRI system (Trio, Siemens Medical Systems, tiple comparisons. Right and left hemispheres were
USA). Parallel imaging was employed with an acceler- tested separately.
ation factor of 2. Duplicate T1-weighted image sets
were acquired during the scan session using a sagittal Statistical analyses
MPRAGE sequence with repetition time/echo time = All analyses were conducted using SAS 9.4 (USA). We
2300/3.46 ms, a 240 Hz/pixel bandwidth, a 256 × 256 tested for univariate differences between diagnostic
matrix, a 240 mm diameter field of view, 160 slices groups in demographic and clinical variables using χ2
with a 1.2 mm slice thickness, yielding an image with tests for categorical variables and two-tailed t tests
voxel sizes of 0.9 × 0.9 × 1.2 mm. In eight cases, subjects for continuous variables. Initial tests for differences in
did not complete MRI or scan quality was not suitable the report of ELSs between depressed and non-
for image processing. depressed cohorts were conducted using χ2 tests, or
Fisher’s exact test when cell sizes were low.
Structural MRI analyses ELSs that differed between groups in univariate tests
were incorporated into GLMs predicting diagnosis
Volumetric MRI analyses
(MDD or non-depressed) while controlling for age,
Regional volumes and cortical thickness were calcu- sex, education and medical morbidity (CIRS score).
lated using FreeSurfer (version 5.1) software. The Retaining these demographic variables, we conducted
FreeSurfer methods used to derive cortical and sub- backward regression to develop a parsimonious
cortical brain volumes have been previously described model, removing each ELS item based on its statistical
(Dale et al. 1999; Fischl et al. 2002, 2004a, b). Cortical significance level of p = 0.05. The remaining ELS items
parcellation used an anatomical mask derived from that subsequently predicted a MDD diagnosis were
the Desikan–Killiany Atlas (Desikan et al. 2006); termed ‘predictive ELSs’.
in each hemisphere, this method identified 33 cortical We next planned a hypothesis-driven approach to
and seven subcortical gray matter regions (nu- reduce the number of comparisons. For neuropsycho-
cleus accumbens, amygdala, caudate, hippocampus, logical analyses, we consolidated individual test

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 4 A. Saleh et al.

Table 1. Demographics and cognitive functiona

Variable Depressed (n = 64) Control (n = 65) Test statistic df p

Age, years 35.1 (8.9) 29.7 (9.2) t = 3.40 127 0.0009

Sex, women, % (n) 60.9 (39) 66.2 (43) χ2 = 0.38 1 0.5382
Education, years 15.2 (2.4) 16.0 (1.9) t = 1.94 127 0.0550
CIRS total 0.6 (1.0) 0.022 (0.5) t = 2.89 127 0.0043
MADRS total score 25.1 (4.5) 0.7 (0.9) t = 41.97 64.15 <0.0001
Depression duration: lifetime, yearsb 5.8 (4.5) – – – –
ELSQ total score 3.5 (2.7) 1.8 (1.7) t = 4.36 106.13 <0.0001
Race, white, % (n) 67.2 (43) 55.4 (36) χ2 = 1.89 1 0.1689
Processing speed −0.13 (0.73) 0.31 (0.73) t = 3.41 127 0.0009
Adjusted F = 3.50 1,125 0.0639
Working memory −0.09 (0.86) 0.17 (0.95) t = 1.60 127 0.1123
Adjusted F = 0.21 1,125 0.6453
Episodic memory −0.21 (0.82) 0.26 (0.66) t = 3.55 127 0.0006
Adjusted F = 2.82 1,125 0.0954
Executive function −0.06 (0.78) 0.26 (0.70) t = 2.45 127 0.0158
Adjusted F = 0.94 1,125 0.3350

Data are given as mean (standard deviation) unless otherwise indicated.

df, Degrees of freedom; CIRS, Cumulative Illness Rating Scale; MADRS, Montgomery–Åsberg Depression Rating Scale;
ELSQ, Early Life Stress Questionnaire.
a
Comparison of demographic variables utilized pooled, two-tailed t tests for continuous measures with equal variances and
Satterthwaite’s t tests for unequal variances. Comparison of categorical variables utilized χ2 tests. Cognitive measures were
Z-transformed and presented both as unadjusted (pooled t tests) and adjusted for age and education level (general linear
models with F values).
b
Note n = 52 for depression duration as not all participants could provide data for past episodes.

results into domain scores as described above. For MRI predictive ELS exposure. If the interaction did not
analyses, we selected a priori regions associated with reach statistical significance, this suggested that there
ELS in existent literature: the ACC (Cohen et al. 2006; was no significant difference in the relationship be-
Udo et al. 2012), OFC (Frodl et al. 2010; Udo et al. tween ELSs, cognition and brain morphology between
2012), amygdala (Tottenham et al. 2010), hippocampus diagnostic groups.
(Sheline et al. 1996; Cole et al. 2011; Teicher et al. 2012) Finally, for predictive ELSs, we conducted explora-
and caudate (Cohen et al. 2006; Udo et al. 2012). For tory analyses using models similar to those described
exploratory analyses of ELS effects on other regions above, but examining the effect of duration of stressor
identified by FreeSurfer, we controlled for multiple exposure. For these analyses, individuals who denied
comparisons using false discovery rate (FDR), imple- each stressor were scored as a duration of zero. Due
mented within SAS. to the high number of people reporting the absence
We first examined the effect of the total ELS expos- of trauma, we utilized a log transformation of the
ure (defined as total ELSQ score) on cognitive domains years of reported duration plus 1.
and brain volumes. For models examining cognitive
domains, we controlled for diagnosis, age, sex and
Results
education. For models examining MRI variables, we
controlled for diagnosis, age, sex and intracranial vol- We examined 129 subjects: 64 with MDD and 65 non-
ume. A similar approach was used for examining the depressed controls. The diagnostic groups differed in
effects of predictive ELSs on cognition and brain struc- age and medical morbidity (Table 1), with the
ture, with participants dichotomized as exposed or not depressed group being older, having more medical ill-
exposed. nesses, and higher total ELSQ score (range: depressed
To determine whether ELS effects differed between 0–11, non-depressed 0–8). In univariate analyses,
diagnostic groups, we added a term coding for an depressed groups exhibited poorer performance in epi-
interaction between ELS and diagnosis. This was sodic memory, executive function and processing
done for analyses of both total ELSQ score and speed. However, these were no longer statistically

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 Early life stress and depression, cognitive performance and brain morphology 5

Table 2. Reported ELS exposure between depressed and non-depressed participantsa

Reported trauma Depressed (n = 64) Control (n = 65) p

Emotional trauma 37.5 (24) 7.7 (5) <0.0001

Physical abuse 18.7 (12) 3.0 (2) 0.0045
Sexual abuse 28.1 (18) 4.6 (3) 0.0003
Domestic violence 9.4 (6) 6.1 (4) 0.5305
Severe family conflict 39.1 (25) 12.3 (8) 0.0006
Neglect 15.6 (10) 1.5 (1) 0.0043
Divorce 21.9 (14) 13.9 (9) 0.2581
Separated 18.8 (12) 12.3 (8) 0.3407
Death in family 39.1 (25) 44.6 (29) 0.5227
Major illness in family 28.1 (18) 13.9 (9) 0.0536
Fire destroyed home 3.1 (2) 1.5 (1) 0.6191
War 3.1 (2) 3.1 (2) 1.0000
Natural disaster 1.6 (1) 3.1 (2) 1.0000
Major personal illness 6.3 (4) 7.7 (5) 1.0000
Hospitalization/surgery 21.9 (14) 18.5 (12) 0.6290
Bullied 37.5 (24) 13.9 (9) 0.0025
Premature birth 10.9 (7) 3.1 (2) 0.0958
Adoption 1.6 (1) 1.5 (1) 1.0000
Other events 9.4 (6) 6.2 (4) 0.5305

Data are given as percentage (number) of subjects exposed to each trauma type.
ELS, Early life stress; df, degrees of freedom.
a
Due to small cell sizes, ELSs were compared using Fisher’s exact test except χ2 tests were used for death in family
(χ2 = 0.41, 1 df) and hospitalization/surgery (χ2 = 0.23, 1 df).

significant after controlling for age and education level Effect of ELS on cognition
(Table 1).
We found no significant main effect of total ELSQ
score on any Z-transformed cognitive domain. After
ELSs: predicting MDD diagnosis
controlling for covariates of age, sex and education,
Depressed patients reported significantly higher rates we observed an interaction between total ELSQ score
of six ELSs (Table 2). While controlling for covariates, and diagnosis on processing speed (F1,122 = 5.28, p =
we incorporated those ELSs into a model predicting 0.0232) and a trend for an effect on working memory
MDD diagnosis. After backwards regression, in the (F1,122 = 3.64, p = 0.0588), but not episodic memory or
final parsimonious model three ELS variables signifi- executive function. On analysing the interaction,
cantly and independently predicted a diagnosis of depressed patients exhibited worsening processing
MDD: emotional trauma (F1,121 = 6.79, p = 0.0103), sex- speed with increasing ELSQ score (Fig. 1a), while non-
ual abuse (F1,121 = 6.00, p = 0.0157) and severe family depressed subjects exhibited faster processing speed
conflict (F1,121 = 7.85, p = 0.0059). performance with increased ELSQ score.
We next dichotomized the sample based on whether In contrast, exposure to predictive ELSs was asso-
they reported one or more of those three ELSs: emo- ciated with progressively poorer performance on
tional abuse, sexual abuse and severe family conflict working memory (F1,123 = 5.08, p = 0.0260) and process-
(‘predictive’ ELSs). Of the study population, 40% (ap- ing speed (F1,123 = 7.74, p = 0.0062), but not executive
proximately 75% of the depressed sample and 33% of function or episodic memory. To better elucidate
the non-depressed sample) reported one or more of these relationships, we found that increasing predictive
these predictive ELSs (χ2 = 24.34, 1 degree of freedom, ELS exposure was associated with worsening perform-
p < 0.0001). Women were more highly represented in ance on mean Z-transformed cognitive domain scores
the group reporting predictive ELSs (78.4%, compared of working memory (0 predictive ELS = 0.24, S.D. =
with 53.9% of those denying predictive ELSs; χ2 = 8.05, 0.93; 1 ELS = −0.13, S.D. = 0.96; 2 ELS = −0.28, S.D. =
1 degree of freedom, p = 0.0046). These groups did 0.64; 3 ELS = −0.70, S.D. = 0.50) and processing speed
not otherwise significantly differ on age, education, mean (0 predictive ELS = 0.24, S.D. = 0.68; 1 ELS = 0.09,
medical morbidity or duration of depression (online S.D. = 0.77; 2 ELS = −0.26, S.D. = 0.83; 3 ELS = −0.70, S.D. =
Supplementary Table S1). 0.68). Importantly, we found no statistically significant

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 6 A. Saleh et al.

resulting in a sample of 51 depressed and 53 non-

depressed individuals. After controlling for age, sex
and intracranial volume, we found no significant differ-
ences between diagnostic groups in a priori brain regions
(online Supplementary Table S2). We also found no dir-
ect effects of total ELSQ score on a priori brain regions.
However, we observed an interactive effect between
total ELSQ score and diagnosis on the left lateral OFC
(F1,97 = 4.05, p = 0.0469). Greater numbers of ELSs are
associated with increasing OFC volumes in non-
depressed subjects, but minimal differences in
depressed subjects (Fig. 1b). In exploratory analyses,
after controlling for multiple comparisons we did not
observe significant direct or interactive effects of total
ELSQ score on other brain regions.
Of the MRI sample, 40% (32 depressed and 10 non-
depressed subjects) reported one or more predictive
ELSs. In analyses of a priori regions, predictive ELSs
were associated with smaller left lateral OFC (F1,98 =
5.11, p = 0.0260) and smaller right caudate volumes
(F1,98 = 6.19, p = 0.0145). We found a significant inter-
action between diagnosis and predictive ELS exposure
only for the hippocampus, with predictive ELSs being
Fig. 1. Relationship of Early Life Stress Questionnaire associated with smaller left hippocampus volume
(ELSQ) total score with processing speed and lateral (F1,98 = 4.98, p = 0.0280) and a trend for smaller right
orbitofrontal cortex (OFC) volume. The figures show how hippocampus volume (F1,98 = 3.35, p = 0.0705), but
exposure to increasing numbers of early life stressors (ELSs;
only in the depressed group. After controlling for mul-
as total ELSQ score) have different effects on processing
tiple comparisons there were no statistically significant
speed performance and lateral OFC volume between
individuals with and without depression. (a) While an
findings in other brain regions.
increased number of ELSs resulted in progressively poorer In exploratory analyses, we examined the relation-
performance in the Z-transformed process speed domain in ship between the volumes of the a priori regions and
depressed patients, a greater number of ELSs is associated duration of exposure to the predictive ELSs. A statistic-
with better process speed performance in non-depressed ally significant relationship was only observed between
controls. (b) With increasing ELS exposure, non-depressed exposure to sexual abuse and caudate volume, wherein
subjects showed relative increases in OFC volume. a greater log-transformed duration of exposure
Conversely, depressed subjects exhibited a slight decline in was associated with smaller caudate volumes (left:
OFC volume with increasing ELS exposure. F1,96 = 4.92, p = 0.0292; right: F1,96 = 5.02, p = 0.0276).
Finally, we tested for relationships between predict-
interactions between diagnosis and predictive ELS expos- ive ELS exposure and cortical thickness. After control-
ure, suggesting that predictive ELSs affect cognition com- ling for sex and diagnosis, predictive ELS exposure
parably regardless of depression. was associated with reduced cortical thickness in sev-
Finally, in exploratory analyses, we examined the re- eral regions, including the bilateral insula, frontal and
lationship between domain scores and log-transformed parietal lobes (Fig. 2 and online Supplementary
duration of predictive ELS exposure. Poorer processing Table S3). We found no association between predictive
speed performance was associated with longer child- ELS exposure and increased cortical thickness.
hood exposure to emotional (F1,121 = 5.74, p = 0.0183)
and sexual abuse (F1,121 = 4.98, p = 0.0277) in both Discussion
cohorts. We did not find statistically significant asso-
ciations between predictive ELS duration and other Similar to past literature (Wise et al. 2001), depressed
cognitive domains. individuals report more ELSs than non-depressed
individuals. However, only emotional abuse, sexual
abuse and severe family conflict significantly predicted
Effect of ELS on brain structure
adult MDD. Exposure to these childhood stressors is
Due to differences in MRI manufacturers, MRI analyses associated with poorer cognitive performance and
included only data gathered at Duke University, alterations in brain morphology that did not differ

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 Early life stress and depression, cognitive performance and brain morphology 7

Fig. 2. Cortical thickness differences related to predictive early life stress (ELS) exposure. Whole-brain vertex-wise display
shows the direct effect of reported predictive ELSs (emotional abuse, sexual abuse or severe family strife) on cortical
thickness. Analyses controlled for diagnosis (major depressive disorder or non-depressed) and sex. Lighter blue color reflects
areas where ELS exposure is associated with thinner cortex. ELS exposure was not significantly associated with increased
cortical thickness in any region.

between diagnostic groups. Additionally, we found working memory was independent of diagnosis.
that exposure to predictive ELSs was associated with Similarly, duration of emotional and sexual abuse ex-
smaller hippocampal volumes, but only in depressed posure was associated with slower processing speed.
individuals. In contrast, when defined broadly, overall Our results did not support prior literature associat-
ELS exposure has a different relationship with process- ing poorer performance on executive function measure
ing speed and OFC volumes between depressed and or episodic memory with ELSs (Polak et al. 2012;
non-depressed adults. Brewin, 2014). The difference between our results and
Childhood trauma has prominent effects on adult past work associating ELS with episodic and semantic
mental health (Navalta et al. 2006). Past work shows memory impairment (Parks & Balon, 1995) could be
altered HPA axis regulation and secondary regional related to heterogeneity in samples or measures used
brain structure changes in children exposed to emo- to examine episodic memory. While prior studies
tional abuse, sexual abuse and aggressive families used the word-cueing technique and the Logical
(McEwen, 2003). Emotional abuse is common in child- Memory test (Parks & Balon, 1995), we additionally
hood and adversely affects self-esteem, interpersonal used the Benton Visual Retention and Rey’s Verbal
skills and personal autonomy and integrity (Vietze Learning Tests.
et al. 1980). Childhood sexual abuse has a worldwide When considering the relationship between ELS
prevalence of 20% (Jakubczyk et al. 2014) and is and processing speed, we may be observing an
associated with multiple psychiatric disorders includ- inverted U-shaped curve. In this model, stress expos-
ing MDD, addictions and increased suicide risk ure broadly defined is associated with improved pro-
(Jakubczyk et al. 2014). Significant family strife also cessing speed, but exposure to more severe (and
interferes with normal development and creates a vul- potentially more chronic) ELSs results in impaired
nerability to maladjustment and internalizing personal performance and vulnerability to depression. This is
problems (Luebbe & Bell, 2014). Admittedly, while concordant with studies in older adults associating
these specific stresses predicted adulthood MDD, childhood trauma with better processing speed
others propose that any significant childhood stress (Feeney et al. 2013). In our non-depressed population,
may increase the risk of depression (Brown & Harris, it is possible that less severe stresses result in
1978). Such effects may depend on stressor severity improved processing speed and contribute to a resili-
and chronicity, age of exposure and positive support ency mechanism (Wu et al. 2013). However, subjects
(Brown & Harris, 1978). predisposed to depression may have pre-existing cir-
cuit dysfunction where neurobiological changes
related to even milder stresses may result in poorer
Effects of ELS on cognition
cognitive performance.
ELS exposure also affected cognitive performance. A similar model may apply to working memory, al-
When defined broadly using the total ELSQ score, though we observed a relationship only with the more
ELS exhibited different effects between depressed severe predictive stressors that did not differ between
and non-depressed subjects on processing speed diagnostic groups. Past work supports negative effects
(Fig. 1a). We propose that these group differences of childhood stressors on working memory (Navalta
reflect different long-term adaptations to stress that et al. 2006). This may be related to altered function of
are related to either vulnerability to or resilience to stress-sensitive systems, as working memory deterio-
developing later depression. In contrast, the effect of rates with increased allostatic stress load (Evans &
predictive ELS exposure on processing speed and Schamberg, 2009).

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 8 A. Saleh et al.

Effects of ELS on brain structure N-methyl-D-aspartate (NMDA) receptor function and

hemispheric differences in NMDA subunit distribu-
ELS exposure was also associated with altered tions (Teicher et al. 2012).
volumes of several regions involved in emotional regu-
lation (Udo et al. 2012). In parallel with our observa-
Limitations and conclusions
tions on processing speed, we observed diagnostic
group differences on the relationship between total Despite the strength of a large, well-characterized sam-
ELSQ score and lateral OFC volume (Fig. 1b). A similar ple, the study also has weaknesses. These include
inverted U-shaped model may also apply to this rela- using a retrospective, self-report scale for ELS, which
tionship. This theory is concordant with a primate does not measure severity, chronicity, or include stress-
study examining early-life maternal separation ful events with positive connotations. Self-report is
(Parker et al. 2005). This study associated separation subject to a memory bias, which may be significant
with increased adult OFC volume, a finding thought in depressed individuals, although others report con-
to be related to stress resiliency by learning extinction sistency between retrospective accounts and documen-
of fear through top-down regulation (Lyons et al. ted events (Martins et al. 2014). Further, our study is
2009). Although the underlying mechanism is unclear, cross-sectional so cannot address longitudinal develop-
the localization of our OFC finding to the left hemi- mental effects of ELS exposure on brain volume or cog-
sphere is supported by prior studies suggesting higher nition. It also does not inform us if the observed
left hemisphere sensitivity to emotional neglect during cognitive and volumetric differences observed in the
brain development (Frodl et al. 2010). MDD population persist with successful depression
Exposure to more severe predictive ELSs was asso- treatment.
ciated with smaller OFC and caudate volumes in There are additional limitations specifically related
both cohorts. This is concordant with past work show- to the depressed group. The depressed group is older
ing that physically and emotionally abused children than the never-depressed group, which is important
exhibit smaller OFC volumes (De Brito et al. 2013) as increased age is associated with changes on MRI
while domestic violence and sexual abuse are asso- and neuropsychological testing. Although this concern
ciated with smaller caudate volumes (Cohen et al. is ameliorated by controlling for age in statistical ana-
2006). Our results from cortical thickness analyses are lyses and the lack of an observed age difference be-
in line with a study associating decreased insula thick- tween individuals who were and were not exposed
ness with ELS exposure (Baker et al. 2013). As the in- to predictive ELSs, study findings do need replication
sula plays a role in salience network regulation, in a truly age-matched sample. Finally, it can be chal-
reported abnormalities may explain deterioration in lenging to disentangle the effects of ELS from the oc-
working memory and processing speed (Krishnadas currence of depression. As we did not observe
et al. 2014). significant differences in cognitive or MRI measures
Smaller hippocampal volumes are reported in MDD between groups after controlling for demographic vari-
(MacQueen & Frodl, 2011). We found that predictive ables (Table 1, online Supplementary Table S2), the
ELSs were associated with hippocampal volume, but predictive ELSs do not appear to be serving as a surro-
only in depressed individuals. Animal models demon- gate marker for depression diagnosis. However, it is
strate that controlled maternal separation results in possible that predictive ELS exposure may influence
decreased hippocampal volumes; however, those the duration or recurrence of depressive episodes, or
volumes may normalize in adulthood (Herpfer et al. be related to early antidepressant treatment. The com-
2012). Extending that finding to our data, we may be plexity of this relationship may require prospective
observing a vulnerability mechanism wherein subjects longitudinal studies to disentangle these effects and
who do not experience recovery of hippocampal identify potential benefits for early intervention.
neurogenesis are at increased risk of adult depression. It is important to note that analyses included numer-
This theory is concordant with past work demonstrat- ous comparisons. This included analyses of four
ing that depression itself contributes to hippocampal Z-transformed cognitive domains, seven a priori brain
volume reduction (Sheline et al. 1996) while smaller regions measured bilaterally, and numerous other
hippocampal volumes are also a risk factor for depres- regions examined in exploratory analyses. For analyses
sion (Cole et al. 2011). Moreover, our finding is consist- of exploratory brain regions, we controlled for multiple
ent with past observations that the left hippocampus comparisons using the FDR method. However, as
is more sensitive to stressful events than the right examination of cognitive domains and our a priori
hippocampus (Teicher et al. 2012). Although the regions were based on specific hypotheses and past lit-
underlying cause for this difference is unclear, it is erature, we did not control for multiple comparisons
hypothesized that it is related to cortisol’s effect on in those analyses. Despite this hypothesis-driven

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 Early life stress and depression, cognitive performance and brain morphology 9

approach, this does increase the risk for false-positive Brown WG, Harris TO (1978). Social Origins of Depression: a
findings. Adjusting those analyses for multiple com- Study of Psychiatric Disorder in Women. The Free Press:
parisons would have limited cognitive findings to pro- New York.
cessing speed with no a priori brain region surviving Cohen RA, Grieve S, Hoth KF, Paul RH, Sweet L, Tate D,
Gunstad J, Stroud L, McCaffery J, Hitsman B, Niaura R,
correction. Thus our findings should be viewed cau-
Clark CR, McFarlane A, Bryant R, Gordon E, Williams
tiously and need replication in context of the broader
LM (2006). Early life stress and morphometry of the adult
literature. anterior cingulate cortex and caudate nuclei. Biological
This study supports that not all childhood stressors Psychiatry 59, 975–982.
increase risk of depression in adulthood. We found a Cole J, Costafreda SG, McGuffin P, Fu CH (2011).
complex relationship between ELSs, cognitive function Hippocampal atrophy in first episode depression: a
and regional brain structures that in some cases dif- meta-analysis of magnetic resonance imaging studies.
fered between diagnostic groups. As we defined pre- Journal of Affective Disorders 134, 483–487.
dictive ELSs based on their relationship with MDD, Dale AM, Fischl B, Sereno MI (1999). Cortical surface-based
these findings require replication in independent analysis. I. Segmentation and surface reconstruction.
populations. Future longitudinal human studies are NeuroImage 9, 179–194.
De Brito SA, Viding E, Sebastian CL, Kelly PA, Mechelli A,
required to incorporate physiological measures of
Maris H, McCrory EJ (2013). Reduced orbitofrontal and
stress reactivity, investigate what factors contribute to
temporal grey matter in a community sample of maltreated
the cognitive deficits observed with exposure to ELS, children. Journal of Child Psychology and Psychiatry 54,
and to clarify the association with volumetric brain 105–112.
changes. Such studies should also examine the effect Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC,
of emotional processes on cognitive performance, and Blacker D, Buckner RL, Dale AM, Maguire RP, Hyman
how ELS exposure may influence those relationships. BT, Albert MS, Killiany RJ (2006). An automated labeling
system for subdividing the human cerebral cortex on MRI
scans into gyral based regions of interest. NeuroImage 31,
Supplementary material 968–980.
Evans GW, Schamberg MA (2009). Childhood poverty,
The supplementary material for this article can be
chronic stress, and adult working memory. Proceedings of
found at http://dx.doi.org/10.1017/S0033291716002403 the National Academy of Sciences USA 106, 6545–6549.
Feeney J, Kamiya Y, Robertson IH, Kenny RA (2013).
Cognitive function is preserved in older adults with a
Acknowledgements reported history of childhood sexual abuse. Journal of
This project was supported by the National Institute of Traumatic Stress 26, 735–743.
Mental Health (NIMH) grants R01 MH077745 and K24 First MB, Gibbon M, Spitzer RL, Williams JB, Benjamin LS
(1997). Structured Clinical Interview for DSM-IV Axis II
MH110598; and the National Center for Advancing
Personality Disorders (SCID-II). American Psychiatric Press,
Translational Sciences (NCATS) award UL1TR000445.
Inc.: Washington, DC.
The authors would like to acknowledge Dr Hakmook Fischl B, Dale AM (2000). Measuring the thickness of the
Kang for guidance on statistical analyses. This project human cerebral cortex from magnetic resonance images.
was conducted using the resources of the Advanced Proceedings of the National Academy of Sciences USA 97,
Computing Center for Research and Education at 11050–11055.
Vanderbilt University, Nashville, TN. Fischl B, Salat DH, Busa E, Albert M, Dieterich M,
Haselgrove C, van der Kouwe A, Killiany R, Kennedy D,
Klaveness S, Montillo A, Makris N, Rosen B, Dale AM
Declaration of Interest (2002). Whole brain segmentation: automated labeling of
neuroanatomical structures in the human brain. Neuron 33,
None. 341–355.
Fischl B, Salat DH, van der Kouwe AJ, Makris N, Segonne
F, Quinn BT, Dale AM (2004a). Sequence-independent
References
segmentation of magnetic resonance images. NeuroImage 23
Baker LM, Williams LM, Korgaonkar MS, Cohen RA, (Suppl. 1), S69–S84.
Heaps JM, Paul RH (2013). Impact of early vs. late Fischl B, van der Kouwe A, Destrieux C, Halgren E,
childhood early life stress on brain morphometrics. Brain Segonne F, Salat DH, Busa E, Seidman LJ, Goldstein J,
Imaging and Behavior 7, 196–203. Kennedy D, Caviness V, Makris N, Rosen B, Dale AM
Brewin CR (2014). Episodic memory, perceptual memory, (2004b). Automatically parcellating the human cerebral
and their interaction: foundations for a theory of cortex. Cerebral Cortex 14, 11–22.
posttraumatic stress disorder. Psychological Bulletin 140, Frodl T, Reinhold E, Koutsouleris N, Reiser M, Meisenzahl
69–97. EM (2010). Interaction of childhood stress with

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 10 A. Saleh et al.

hippocampus and prefrontal cortex volume reduction in women. Journal of Neuropsychiatry and Clinical Neurosciences
major depression. Journal of Psychiatric Research 44, 799–807. 18, 45–53.
Gershon A, Sudheimer K, Tirouvanziam R, Williams LM, Parker KJ, Buckmaster CL, Justus KR, Schatzberg AF, Lyons
O’Hara R (2013). The long-term impact of early adversity DM (2005). Mild early life stress enhances
on late-life psychiatric disorders. Current Psychiatry Reports prefrontal-dependent response inhibition in monkeys.
15, 352. Biological Psychiatry 57, 848–855.
Harkness KL, Bruce AE, Lumley MN (2006). The role of Parks ED, Balon R (1995). Autobiographical memory for
childhood abuse and neglect in the sensitization to stressful childhood events: patterns of recall in psychiatric patients
life events in adolescent depression. Journal of Abnormal with a history of alleged trauma. Psychiatry 58, 199–208.
Psychology 115, 730–741. Polak AR, Witteveen AB, Reitsma JB, Olff M (2012). The role
Herpfer I, Hezel H, Reichardt W, Clark K, Geiger J, Gross of executive function in posttraumatic stress disorder: a
CM, Heyer A, Neagu V, Bhatia H, Atas HC, Fiebich BL, systematic review. Journal of Affective Disorders 141, 11–21.
Bischofberger J, Haas CA, Lieb K, Normann C (2012). Sanders B, Becker-Lausen E (1995). The measurement of
Early life stress differentially modulates distinct forms of psychological maltreatment: early data on the Child Abuse
brain plasticity in young and adult mice. PLOS ONE 7, and Trauma Scale. Child Abuse and Neglect 19, 315–323.
e46004. Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J,
Jakubczyk A, Klimkiewicz A, Krasowska A, Kopera M, Weiller E, Hergueta T, Baker RW, Dunbar GC (1998). The
Slawinska-Ceran A, Brower KJ, Wojnar M (2014). History Mini-International Neuropsychiatric Inventory (M.I.N.I.):
of sexual abuse and suicide attempts in alcohol-dependent the development and validation of a structured diagnostic
patients. Child Abuse and Neglect 38, 1560–1568. interview for DSM-IV and ICD-10. Journal of Clinical
Kaplow JB, Widom CS (2007). Age of onset of child Psychiatry 59 (Suppl. 20), 22–33.
maltreatment predicts long-term mental health outcomes. Sheline YI, Pieper CF, Barch DM, Welsh-Bohmer K,
Journal of Abnormal Psychology 116, 176–187. McKinstry RC, MacFall JR, D’Angelo G, Garcia KS,
Kessler RC, Magee WJ (1994). Childhood family violence and Gersing K, Wilkins C, Taylor W, Steffens DC, Krishnan
adult recurrent depression. Journal of Health and Social RR, Doraiswamy PM (2010). Support for the vascular
Behavior 35, 13–27. depression hypothesis in late-life depression: results of a
Krishnadas R, Palaniyappan L, Lang J, McLean J, Cavanagh 2-site, prospective, antidepressant treatment trial. Archives
J (2014). Psychoticism and salience network morphology. of General Psychiatry 67, 277–285.
Personality and Individual Differences 57, 37–42. Sheline YI, Sanghavi M, Mintun MA, Gado MH (1999).
Luebbe AM, Bell DJ (2014). Positive and negative family Depression duration but not age predicts hippocampal
emotional climate differentially predict youth anxiety and volume loss in medically healthy women with recurrent
depression via distinct affective pathways. Journal of major depression. Journal of Neuroscience 19, 5034–5043.
Abnormal Child Psychology 42, 897–911. Sheline YI, Wang PW, Gado MH, Csernansky JG, Vannier
Lyons DM, Parker KJ, Katz M, Schatzberg AF (2009). MW (1996). Hippocampal atrophy in recurrent major
Developmental cascades linking stress inoculation, arousal depression. Proceedings of the National Academy of Sciences
regulation, and resilience. Frontiers in Behavioral USA 93, 3908–3913.
Neuroscience 3, 32. Snyder HR (2013). Major depressive disorder is associated
MacQueen G, Frodl T (2011). The hippocampus in major with broad impairments on neuropsychological measures
depression: evidence for the convergence of the bench and of executive function: a meta-analysis and review.
bedside in psychiatric research? Molecular Psychiatry 16, Psychological Bulletin 139, 81–132.
252–264. Teicher MH, Anderson CM, Polcari A (2012). Childhood
Martins CM, Von Werne Baes C, Tofoli SM, Juruena MF maltreatment is associated with reduced volume in the
(2014). Emotional abuse in childhood is a differential factor hippocampal subfields CA3, dentate gyrus, and subiculum.
for the development of depression in adults. Journal of Proceedings of the National Academy of Sciences USA 109,
Nervous and Mental Disease 202, 774–782. E563–E572.
McEwen BS (2003). Early life influences on life‐long patterns Tottenham N, Hare TA, Quinn BT, McCarry TW, Nurse M,
of behavior and health. Mental Retardation and Developmental Gilhooly T, Millner A, Galvan A, Davidson MC, Eigsti
Disabilities Research Reviews 9, 149–154. IM, Thomas KM, Freed PJ, Booma ES, Gunnar MR,
Miller MD, Paradis CF, Houck PR, Mazumdar S, Stack JA, Altemus M, Aronson J, Casey BJ (2010). Prolonged
Rifai AH, Mulsant B, Reynolds CF (1992). Rating chronic institutional rearing is associated with atypically large
medical illness burden in geropsychiatric practice and amygdala volume and difficulties in emotion regulation.
research: application of the Cumulative Illness Rating Scale. Developmental Science 13, 46–61.
Psychiatry Research 41, 237–248. Udo D, Anja S, Victoria B, Peter Z, Thomas L, Dominik G,
Montgomery SA, Åsberg M (1979). A new depression scale Katharina D, Christa H, Patricia O, Jochen B, Christian L,
designed to be sensitive to change. British Journal of Christian P, Carsten K, Volker A, Walter H, Thomas S,
Psychiatry 134, 382–389. Harald K (2012). Limbic scars: long-term consequences of
Navalta CP, Polcari A, Webster DM, Boghossian A, Teicher childhood maltreatment revealed by functional and
MH (2006). Effects of childhood sexual abuse on structural magnetic resonance imaging. Biological Psychiatry
neuropsychological and cognitive function in college 71, 286–293.

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403
 Early life stress and depression, cognitive performance and brain morphology 11

van Harmelen AL, van Tol MJ, van der Wee NJ, Veltman of child maltreatment. Infant Mental Health Journal 1,
DJ, Aleman A, Spinhoven P, van Buchem MA, Zitman 248–261.
FG, Penninx BW, Elzinga BM (2010). Reduced Wise LA, Zierler S, Krieger N, Harlow BL (2001). Adult onset
medial prefrontal cortex volume in adults reporting of major depressive disorder in relation to early life violent
childhood emotional maltreatment. Biological Psychiatry victimisation: a case–control study. Lancet 358, 881–887.
68, 832–838. Wu G, Feder A, Cohen H, Kim JJ, Calderon S, Charney DS,
Vietze P, Falsey S, Sandler H, O’Connor S, Altemier WA Mathe AA (2013). Understanding resilience. Frontiers in
(1980). Transactional approach to prediction Behavioral Neuroscience 7, 10.

Downloaded from http:/www.cambridge.org/core. Boston University Libraries, on 22 Nov 2016 at 00:56:53, subject to the Cambridge Core terms of use, available at http:/www.cambridge.org/core/terms.
http://dx.doi.org/10.1017/S0033291716002403