Effects of childhood adversity on the volumes of the amygdala subnuclei and hippocampal subfields in individuals with major depressive disorder

Abstract

Background

Reductions in total hippocampus volume have frequently been reported in MRI studies in major depressive disorder (MDD), but reports of differences in total amygdala volume have been inconsistent. Childhood maltreatment is an important risk factor for MDD in adulthood and may affect the volume of the hippocampus and amygdala. In the present study, we examined associations between the volumes of the amygdala subnuclei and hippocampal subfields and history of childhood maltreatment in participants with MDD.

Methods

We recruited 35 patients who met the DSM-IV criteria for MDD and 35 healthy controls. We acquired MRI data sets on a 4.7 T Varian Inova scanner. We manually delineated the amygdala subnuclei (lateral, basal and accessory basal nuclei, and the cortical and centromedial groups) and hippocampal subfields (cornu ammonis, subiculum and dentate gyrus) using reliable volumetric methods. We assessed childhood maltreatment using the Childhood Trauma Questionnaire in participants with MDD.

Results

In participants with MDD, a history of childhood maltreatment had significant negative associations with volume in the right amygdala, anterior hippocampus and total cornu ammonis subfield bilaterally. For volumes of the amygdala subnuclei, such effects were limited to the basal, accessory basal and cortical subnuclei in the right hemisphere, but they did not survive correction for multiple comparisons. We did not find significant effects of MDD or antidepressant treatment on volumes of the amygdala subnuclei.

Limitations

Our study was a cross-sectional study.

Conclusion

Our results provide evidence of negative associations between history of childhood maltreatment and volumes of medial temporal lobe structures in participants with MDD. This may help to identify potential mechanisms by which maltreatment leads to clinical impacts.

Introduction

Major depressive disorder (MDD) causes disability around the world,¹ and its etiology remains uncertain, although changes in the hippocampus and amygdala have been suggested as potentially important.² Volumetric reductions in the hippocampus have often been reported in MRI studies of MDD.^2–5 Although the amygdala is an important structure in neuronal circuits of emotion, fear and stress,^6,7 the results of volumetric MRI studies of the amygdala in MDD have been inconsistent.^2,8–10 The amygdala consists of several functionally different subnuclei groups⁷: the basolateral complex, including the lateral, basal and accessory basal nuclei; the centromedial group, including the central and medial nuclei; and the cortical group. Generally, the basolateral amygdala is involved in learning and memory through its connections with the prefrontal cortex and the hippocampus.¹¹ Animal studies indicate that during chronic stress, the basolateral amygdala undergoes adaptive plastic changes.¹² The centromedial amygdala is involved in the regulation of behavioural, autonomic and hormonal responses to emotional stimuli via its connections with the hypothalamus.¹¹ The cortical group is involved in olfactory-related responses.^6,13 Until recently, MRI studies of the amygdala have examined the amygdala only at the level of its total volume,² but recent advances in MRI have enabled researchers to measure individual amygdala subnuclei in vivo.¹⁴

Childhood adversity, including trauma and maltreatment, has been recognized as an important risk factor for developing MDD in adulthood.¹⁵ It can induce biological changes in stress-related brain structures, which may then become maladaptive in adult environments and make people more vulnerable to psychiatric disorders.¹⁶ Previous MRI studies have found that childhood adversity is associated with smaller volumes in the anterior cingulate, dorsolateral prefrontal and orbitofrontal cortices, and in the hippocampus.¹⁷ However, the effects of childhood adversity might not be uniform across the entire structure of the hippocampus.^18,19 The hippocampus can be divided into subregions along its anteroposterior axis (head, body and tail), and into cellular subfields along its cross-sectional axis, including the dentate gyrus, the cornu ammonis (CA1–3) and the subiculum.²⁰

Recent MRI studies that measured the volumes of hippocampal subfields reported equivocal findings: patients with MDD had smaller volumes in the dentate gyrus,²¹ CA1–3^21,22 or all subfields.²³ In contrast, other studies found no differences²⁴ or reported that volume reductions were present only in unmedicated patients.²¹ The volumetric results for hippocampal subfields from the MDD cohort in the present study have been reported elsewhere.^21,25

Animal models of adult chronic stress indicate that the CA3 subfield is most susceptible to cellular changes associated with prolonged stressors and glucocorticoid exposure.²⁶ Additional changes include dendritic retraction and suppressed adult neurogenesis in the dentate gyrus subfield.^27,28

Previous human studies have suggested that the adverse effects of psychological stress and childhood adversity were more pronounced in the anterior hippocampus^18,29 and the CA2–3 and CA4–dentate gyrus subfields.¹⁹ Although the results of individual studies for the relationship between total amygdala volume and childhood adversity have been inconsistent, ^30–34 2 meta-analyses showed that people who experienced maltreatment, regardless of the absence or presence of psychiatric disorders, had significantly smaller total amygdala volumes than people who were not maltreated.^35,36

Most of the work on the effect of stress on the amygdala and hippocampal substructures has been conducted in animals; ^12,26 direct testing of preclinical stress models in humans has been impossible to date. However, recent advances in high-resolution MRI of the hippocampal subfields²⁰ and amygdala subnuclei¹⁴ have allowed researchers to test these models in vivo in humans for the first time.

In this study we aimed to investigate volumetric differences in amygdala subnuclei between participants with MDD and healthy controls; and to determine the relationships between childhood maltreatment (as measured using the Childhood Trauma Questionnaire [CTQ]) and the volumes of the amygdala subnuclei and hippocampal subfields in people with MDD using ultra-high-resolution 4.7 T MRI methods developed by our group.^14,20

Considering the findings of 2 previous meta-analyses on total amygdala volumes in MDD^2,9 that did not show a significant effect of MDD, we did not expect to find significant differences in amygdala volume. However, based on previous animal^12,26,37 and human studies,^{17,18,29,35,36} we hypothesized that childhood adversity would be associated with volumetric reductions in the basolateral amygdala and in the CA1–3 and dentate gyrus subfields — especially in the anterior hippocampus — in participants with MDD. Furthermore, considering the results of previous studies^34–36,38 showing that the effects of childhood maltreatment were more pronounced in the right amygdala, we hypothesized that the effects of childhood adversity on the basolateral amygdala would be more pronounced in the right hemisphere.

Methods

Participants

A total of 35 participants with MDD were included in the study: 12 males and 23 premenopausal females aged 18 to 49 years who met the DSM-IV criteria for MDD with moderate or severe episodes, based on a full clinical assessment and the Anxiety Disorders Interview Schedule for DSM-IV–Lifetime Version.³⁹ We also recruited 35 age-, sex- and education-matched healthy controls (12 males, 23 premenopausal females) through local advertisements. Written informed consent was obtained, and the study was approved by the University of Alberta Health Research Ethics Board.

Participants with MDD were recruited from local notices or referred directly from the outpatient psychiatry department by the authors (N.J.C. and P.H.S.). Among the participants with MDD, 25 reported continuous use of antidepressant treatment for more than 6 months, and 10 were either antidepressant-naïve or medication-free for 6 months or more.

Exclusion criteria for participants with MDD were mild depressive episodes; psychotic or atypical features; seasonal affective disorder; lifetime schizophrenia, bipolar disorder, alcohol or substance dependence, anorexia nervosa, or predominant personality or anxiety disorders; antipsychotic or mood stabilizer treatment; corticosteroid use; or significant medical or neurologic diseases. Healthy controls had no lifetime psychiatric disorders, as assessed with the Anxiety Disorders Interview Schedule for DSM-IV–Lifetime Version, and no reported psychosis or mood disorders in first-degree relatives.

We assessed the severity of depressive symptoms using the 17-item Hamilton Depression Rating Scale. We assessed childhood maltreatment in participants with MDD using the 28-item CTQ, which includes 5 subscales of 5 items each (emotional abuse, physical abuse, sexual abuse, emotional neglect and physical neglect) and 3 validity items.⁴⁰

MRI data acquisition and analysis

We acquired images using a 4.7 T MRI system (Varian). We used a T₂-weighted fast spin echo MRI sequence (echo/repetition time: 39/11 000 ms; field of view: 20 × 20 cm; acquisition time: 13.5 minutes; native resolution: 0.52 × 0.68 × 1.0 mm³) to obtain 90 contiguous coronal slices perpendicular to the anterior–posterior commissure. To estimate intracranial volume (ICV), we acquired a whole-brain T₁-weighted 3D magnetization-prepared rapid gradient echo (MPRAGE) sequence (axial, echo/repetition time: 8.5 ms/4.5 ms; field of view: 256 × 200 × 180 mm³; voxel size: 1 × 1 × 1 mm³). We used the program DISPLAY (Montreal Neurological Institute) to manually trace ICV using the method of Eritaia⁴¹ on the T₁-weighted MPRAGE images.

We have published detailed protocols for the manual delineation of the total amygdala,⁴² amygdala subnuclei,¹⁴ hippocampal subregions⁴² and hippocampal subfields (CA1–3, dentate gyrus and subiculum;²⁰ Fig. 1⁴³ and Appendix 1, available at jpn.ca/200034-a1). Two raters measured the amygdala (AAS) and hippocampus (YH) in the T₂-weighted fast spin echo images using DISPLAY. Inter- and intra-rater reliabilities and Dice’s kappa coefficients for volumetric measurements have been reported previously (Table 1; see Appendix 1 for more details).^14,20,42 We then normalized the raw volumetric measurements of the amygdala and hippocampus to ICV using a proportional method: normalized volume = (raw volume [mm³]/ICV [cm³]) × sample average ICV (cm³).

Fig. 1.

Three-dimensional reconstructions of the hippocampal subfields and subregions, as well as the amygdala and its subnuclei, from a healthy individual. Three-dimensional images from the (a) medial, (b) superior and (c) lateral views were created using ITK-SNAP software (version 3.8.0).⁴³ (d–i) Segmentation of the hippocampal subfields and subregions, as well as the amygdala subnuclei on coronal views, are shown on T₂-weighted fast spin echo images with inverted contrast. For the hippocampal subfields, the cornu ammonis (CA 1–3) is outlined in red, the dentate gyrus is outlined in blue and the subiculum is outlined in green; hippocampal subfields along the anteroposterior axis of the hippocampus (head, body and tail) are shown with different colour intensities. For the amygdala subnuclei, the lateral nucleus is outlined in pink, the basal nucleus is outlined in dark purple, the accessory basal nucleus is outlined in light purple, the cortical group is outlined in orange and the centromedial group is outlined in yellow. Dotted lines indicate the locations of corresponding MRI slices.

Table 1.

Reliability results

Region	Intraclass correlation coefficients		Dice’s kappa
	Intra-rater	Inter-rater	Intra-rater	Inter-rater
Lateral nucleus	0.95	0.93	0.86	0.82
Basal nucleus	0.94	0.87	0.81	0.76
Accessory basal nucleus	0.96	0.95	0.77	0.71
Cortical group	0.97	0.87	0.72	0.72
Centromedial group	0.86	0.85	0.79	0.76
Total amygdala	0.93	0.95	0.93	0.91
Hippocampus head	0.92	0.95	0.90	0.89
Hippocampus body	0.93	0.83	0.87	0.86
Hippocampus tail	0.88	0.95	0.82	0.80
Cornu ammonis	0.92	0.92	0.75	0.73
Dentate gyrus	0.84	0.86	0.81	0.81
Subiculum	0.95	0.87	0.74	0.74
Total hippocampus	0.97	0.95	0.90	0.89

Source: Aghamohammadi-Sereshki and colleagues;¹⁴ Malykhin and colleagues²⁰,⁴²

Statistical analysis

All analyses were carried out in SPSS Statistics 25.0 (IBM). We used 1-way analysis of variance to compare age, education, MDD characteristics and ICV in participants with MDD and controls. We used analysis of covariance (ANCOVA) to compare amygdala volumes between participants with MDD and healthy controls, and between participants with medicated and unmedicated MDD and healthy controls. The ANCOVAs included volume as the dependent variable, diagnosis (or treatment) as a between-subject factor, hemisphere as a within-subject factor, interaction between diagnosis or treatment and hemisphere, and ICV as covariates. For diagnosis, analyses were 2-group; for treatment, analyses were 3-group, followed by pair-wise ANCOVA of subgroups. We tested correlations between normalized volumetric measurements, CTQ scores and clinical characteristics in participants with MDD using Pearson or Spearman correlations, as appropriate. Because of our a priori hypotheses, for the amygdala we examined effects of childhood adversity first on the volume of the total amygdala and its subnuclei (hypothesis-driven analysis) and second on the volumes of the individual basolateral amygdala subnuclei (exploratory analysis). Similarly for the hippocampus, we examined the effects of childhood adversity first on the volume of the total hippocampus and the total volume of its subfields (CA1–3, dentate gyrus, subiculum) and subregions (head v. body and tail; hypothesis-driven analysis), and second on the volume of subfields in the hippocampal subregions (exploratory analysis). We also conducted exploratory analyses for all CTQ factors with individual regions of interest. We used Holm–Bonferroni correction for type I error inflation because of multiple hypothesis testing: for all amygdala groups including centromedial, cortical and basolateral (n = 3); for basolateral subnuclei (n = 3); for bilateral hippocampi (n = 2); for total hippocampal subfields (n = 3); for hippocampal subfields in each hippocampal subregion (n = 3); and for all CTQ factors (n = 5). We used Levene’s test to check homogeneity of variance; significance was set at p < 0.05 using a 2-tailed test for all analyses, except for associations between childhood adversity and amygdala-and hippocampus-related volumes, for which we used p < 0.05, 1-tailed (i.e., negative correlations were hypothesized).

Results

Demographic and clinical characteristics

We found no significant differences in age, sex, education or ICV between participants with MDD and healthy controls, or in demographic and clinical characteristics between unmedicated and medicated participants with MDD (Table 2). Analysis of variance showed no significant association between sex and CTQ measurements (all p > 0.1; Table 3), indicating that male and female participants with MDD did not differ from each other.

Table 2.

Demographic and clinical characteristics

Characteristics	Healthy controls (n = 35)	MDD			F	p values
		Total (n = 35)	Medicated (n = 25)	Unmedicated (n = 10)		Controls v. MDD	3 groups*
Demographic characteristics
Female/male, n	23/12	23/12	17/8	6/4	—	1.0	0.91
Age, yr	32.3 ± 10.0	34.9 ± 8.7	36.1 ± 7.7	31.9 ± 10.8	—	0.25	0.25
Education, yr	15.7 ± 1.7	15.4 ± 1.8	15.7 ± 1.5	14.6 ± 2.2	—	0.55	0.19
Intracranial volume, cm3	1599 ± 172	1561 ± 154	1537 ± 156	1622 ± 141	—	0.33	0.24
MDD clinical characteristics
HAM-D score	—	—	17.0 ± 8.4	20.0 ± 2.6	3.04	0.09†
Duration of MDD, yr	—	—	4.6 ± 3.7	4.5 ± 4.7	0.004	0.95
Recurrent/nonrecurrent, n	—	—	20/5	7/3	NA	0.66‡
Major depressive episode, n	—	—	2.5 ± 1.1	2.3 ± 1.1	0.30	0.59

HAM-D = Hamilton Depression Rating Scale; MDD = major depressive disorder; NA = not applicable.

Values are mean ± standard deviation unless otherwise indicated.

^*Three-group comparison between healthy controls, medicated MDD and unmedicated MDD.

^†Adjusted for Welch’s F-test because of violation of the homogeneity of variance.

^‡Adjusted for Fisher’s exact test because of violation of Pearson’s χ² assumption.

Table 3.

Childhood Trauma Questionnaire scores

Total CTQ and CTQ factors	Patients with MDD			p values*
	Total (n = 33)	Male (n = 12)	Female (n = 21)
Total	50.1 ± 19.0	44.7 ± 17.8	53.2 ± 19.4	0.18
Emotional abuse	11.7 ± 5.4	9.7 ± 5.8	12.9 ± 5.0	0.08
Physical abuse	8.3 ± 4.1	7.4 ± 3.3	8.8 ± 4.5	0.35
Sexual abuse	8.4 ± 6.2	7.8 ± 5.8	8.8 ± 6.5	0.65
Emotional neglect	13.5 ± 5.4	12.9 ± 4.9	13.8 ± 5.8	0.66
Physical neglect	8.2 ± 3.8	6.9 ± 3.3	9.0 ± 3.9	0.06

CTQ = Childhood Trauma Questionnaire; MDD = major depressive disorder.

^*Total scores and scores for all factors on the CTQ but emotional neglect were not normally distributed; we compared findings for males and females using the Mann–Whitney U test.

Amygdala volumetric analyses

Diagnosis (participants with MDD v. controls), hemisphere, treatment (unmedicated MDD v. medicated MDD v. controls) and hemisphere interactions were not significant for any amygdala measurement (all p > 0.23), indicating that any group differences were not lateralized. We found no main effects of diagnosis in the 2-group analyses for total amygdala volume or the total volume of each subnucleus group (Table 4). Exploratory analysis of the individual basolateral subnuclei also revealed no significant effect of diagnosis or treatment on the total volume of each subnucleus (all p > 0.16). Correlational analyses revealed no significant relationships between the volumes of the amygdala or its subnuclei and Hamilton Depression Rating Scale score, MDD duration or number of major depressive episodes (all p > 0.16).

Table 4.

Analysis of covariance and descriptive statistics for amygdala measurements^*

Region of interest	Healthy controls v. MDD				Healthy controls v. unmedicated MDD v. medicated MDD
	Healthy controls (n = 35)	MDD (n = 35)	F1,67	p value†	Unmedicated MDD (n = 10)	Medicated MDD (n = 25)	F2,66	p value‡	Unmedicated MDD v. healthy controls		Medicated MDD v. healthy controls		Unmedicated MDD v. medicated MDD
									F1,42	p value	F1,57	p value	F1,32	p value
Total amygdala	1994 ± 330	1880 ± 307	1.35	0.25	1933 ± 285	1858 ± 319	0.67	0.52	0.88	0.35	1.10	0.30	0.02	0.90
Centromedial group	287 ± 59	278 ± 71	0.06	0.81	306 ± 101	267 ± 54	0.77	0.47	0.45	0.51	1.04	0.31	0.81	0.37
Cortical group	157 ± 29	144 ± 33	1.03	0.31	149 ± 38	142 ± 32	0.51	0.60	0.69	0.41	1.00	0.32	0.02	0.90
Basolateral complex	1552 ± 282	1457 ± 238	1.45	0.23	1477 ± 205	1448 ± 253	0.78	0.46	1.41	0.24	0.83	0.37	0.01	0.94
Lateral nucleus	642 ± 120	598 ± 94	2.02	0.16	596 ± 90	599 ± 98	1.17	0.32	2.22	0.14	0.91	0.34	0.01	0.92
Basal nucleus	618 ± 122	581 ± 105	1.08	0.30	597 ± 100	574 ± 108	0.54	0.59	0.64	0.43	0.84	0.36	0.00	0.96
Accessory basal nucleus	291 ± 57	277 ± 59	0.38	0.54	283 ± 53	274 ± 62	0.23	0.80	0.54	0.47	0.24	0.63	0.05	0.83

MDD = major depressive disorder.

^*Analysis of covariance of volumes (mm³) of the total amygdala, basolateral, cortical and centromedial groups of the amygdala, as well as the lateral, basal and accessory basal nuclei of the basolateral group (covariate, intracranial volumes). Group (diagnosis or treatment) × hemisphere interactions were not significant; volumes are mean ± standard deviation of both hemispheres.

^†Diagnosis main effect in 2-group comparison: healthy controls and MDD.

^‡Treatment main effect in 3-group comparison: healthy controls, unmedicated MDD and medicated MDD.

Childhood adversity and amygdala volumes in MDD

The volume of the right total amygdala (but not the left; p > 0.35) showed a significant negative association with total CTQ-25 score (Table 5). Similarly, the volume of the right basolateral amygdala showed a negative association with total CTQ-25 score, but this finding did not survive Holm–Bonferroni correction for multiple comparisons. The volumes of the centromedial and cortical groups in the left and right hemispheres (all p > 0.09) and left basolateral amygdala (p > 0.29) were not correlated with total CTQ-25 score.

Table 5.

Summary of p values for relationships between Childhood Trauma Questionnaire scores and volumetric measurements in participants with MDD (n = 33)

Score	Amygdala	Hippocampus			Subfields within hippocampal subregions
		Total	Subregions	Subfields	Head	Body	Tail
Total score*	Right amygdala: rs: −0.324, p < 0.033 Right basolateral complex: rs: −0.311, p < 0.040 Right accessory basal nucleus: rs: −0.361, p < 0.020 Right basal nucleus: rs: −0.288, p < 0.05	Left hippocampus: rs: −0.347, p < 0.024; p < 0.048† Right hippocampus: rs: −0.283, p < 0.06	Left head: rs: −0.518, p < 0.001; p < 0.002† Right head: rs: −0.369, p < 0.018; p < 0.018†	Left cornu ammonis: rs: −0.560, p < 0.001; p < 0.002† Right cornu ammonis: rs: −0.395, p < 0.012	Left cornu ammonis: rs: −0.633, p < 0.001; p < 0.001† Right cornu ammonis: rs: −0.412, p < 0.009; p < 0.043† Right dentate gyrus: rs: −0.299, p < 0.046 Left subiculum: rs: −0.319, p < 0.036	—	—
Emotional abuse*	Right amygdala: rs: −0.359, p < 0.021 Right basolateral complex: rs: −0.343, p < 0.026 Right accessory basal nucleus: rs: −0.448, p < 0.005 Right basal nucleus: rs: −0.375, p < 0.016 Left basal nucleus: rs: −0.293, p < 0.049 Right cortical group: rs: −0.383, p < 0.015	Left hippocampus: rs: −0.246, p < 0.09 Right hippocampus: rs: −0.236, p < 0.09	Left head: rs: −0.419, p < 0.008 Right head: rs: −0.311, p < 0.040	Left cornu ammonis: rs: −0.423, p < 0.007 Right cornu ammonis: rs: −0.381 p < 0.015	Left cornu ammonis: rs: −0.557, p < 0.001; p < 0.012† Right cornu ammonis: rs: −0.386, p < 0.014 Left subiculum: rs: −0.254, p < 0.08	—	—
Physical abuse*	—	Left hippocampus: rs: −0.229, p < 0.1 Right hippocampus: rs: −0.386, p < 0.014	Left head: rs: −0.396, p < 0.012 Right head: rs: −0.453, p < 0.004; p < 0.037†	Left cornu ammonis: rs: −0.393, p < 0.012 Right cornu ammonis: rs: −0.479 p < 0.002	Left cornu ammonis: rs: −0.403, p < 0.011 Right cornu ammonis: rs: −0.476, p < 0.003 Right dentate gyrus: rs: −0.481 p < 0.002 Left subiculum: rs: −0.248, p < 0.08	—	—
Sexual abuse*	—	Left hippocampus: rs: −0.288, p < 0.053	Left head: rs: −0.245, p < 0.09	NS	Right dentate gyrus: rs: −0.432, p < 0.006	Left dentate gyrus: rs: −0.345, p < 0.025	—
Emotional neglect‡	—	Left hippocampus: rs: −0.367, p < 0.018 Right hippocampus: r: −0.295, p < 0.049	Left head: r: −0.439, p < 0.005; p < 0.043† Right head: r: −0.280, p < 0.06	Left cornu ammonis: r: −0.539, p < 0.001; p < 0.018† Right cornu ammonis: r: −0.258, p < 0.07	Left cornu ammonis: r: −0.536, p < 0.001; p < 0.018† Right cornu ammonis: r: −0.277, p < 0.06 Left subiculum: r: −0.281, p < 0.06	Right dentate gyrus: r: −0.248, p < 0.08	—
Physical neglect*	—	—	Left head: rs: −0.519, p < 0.001; p < 0.01† Right head: rs: −0.334, p < 0.029	Left cornu ammonis: rs: −0.536, p < 0.001; p < 0.019† Right cornu ammonis: rs: −0.330, p < 0.031	Left cornu ammonis: rs: −0.600, p < 0.001; p < 0.004† Right cornu ammonis: rs: −0.410, p < 0.009 Left subiculum: rs: −0.298, p < 0.046	—	—

MDD = major depressive disorder.

^*Spearman’s correlation.

^†Corrected p values for those that survived Holm–Bonferroni correction; p value for total amygdala was not adjusted for multiple comparisons because of our initial hypothesis.

^‡Pearson’s correlation.

Exploratory analysis in the basolateral group showed that total CTQ-25 score had negative correlations with the right accessory basal and basal nuclei (Table 5), but not with the left nuclei (both p > 0.11).

Exploratory analysis of CTQ factors showed that only emotional abuse was negatively correlated with the volumes of the total amygdala and the accessory basal, basal and cortical nuclei, mainly in the right hemisphere. However, the exploratory correlations for the amygdala subnuclei did not survive Holm–Bonferroni correction.

Childhood adversity and hippocampal volumes

We found correlations between CTQ scores and hippocampal measurements in both hemispheres (Table 5).

Total hippocampal volume showed a negative relationship with total CTQ-25 score. Among hippocampal subregions, only the head was negatively correlated with total CTQ-25 score, but not the body or tail (both p > 0.13). Analysis of the hippocampal subfields showed that CA1–3 had a significant negative correlation with total CTQ-25 score (Table 5), but not the total dentate gyrus or subiculum (both p > 0.17).

Exploratory analysis of the subfields in the hippocampal head showed that the bilateral CA1–3, right dentate gyrus and left subiculum had a significant negative correlation with total CTQ-25 score (Table 5), but not the left dentate gyrus and right subiculum volumes (both p > 0.31).

Exploratory analysis for the CTQ factors showed that emotional abuse, physical abuse, emotional neglect and physical neglect had strong negative relationships with the volume of the hippocampal head and CA1–3 in the hippocampal head. We found weaker negative associations between physical and sexual abuse and the dentate gyrus of the hippocampal head and body, and between physical neglect and the subiculum in the hippocampal head, but these findings did not survive Holm–Bonferroni correction.

Discussion

The current volumetric in vivo MRI study examined the effects of childhood adversity on the amygdala subnuclei and hippocampal subfields in MDD and the effects of MDD on volumes of the amygdala subnuclei. Although we did not find any significant effect of MDD or long-term antidepressant treatment on the amygdala subnuclei, we did find that childhood adversity was negatively associated with both hippocampal and amygdala volumes. The negative effects of childhood adversity as measured by the CTQ-25 were observed bilaterally in the volumes of the anterior hippocampus (i.e., the hippocampal head) and the total CA1–3 subfield; in the amygdala, these effects were limited to the basolateral amygdala in the right hemisphere. These are structures shown to be affected by chronic stress in preclinical studies.^12,26

Effects of adverse experiences in childhood/adolescence and MDD on amygdala subnuclei

Both the amygdala and the hippocampus are regarded as targets of childhood adversity^17,44,45 because they exhibit protracted postnatal development, a high density of glucocorticoid receptors and postnatal neurogenesis.^44,45

The present study confirmed the negative effects of childhood adversity on the right amygdala and suggested that these effects might also affect the basolateral amygdala.

Our results also suggested that the accessory basal, basal and cortical nuclei could be potential targets for these effects in the amygdala. These results could be explained by differences in the connectivity profile and functions of the amygdala subnuclei. Previous studies have shown that childhood adversity induces abnormal changes in frontolimbic regions. ^35,36 The majority of the amygdala reciprocal connections with other frontolimbic regions — including the CA1–3 hippocampal subfield, the caudate, the nucleus accumbens and the orbitofrontal, medial prefrontal and anterior cingulate cortices — run mainly through the accessory basal and basal nuclei and only to a lesser degree through the cortical group.^11,13,46 In addition, differences in functional specialization of the amygdala subnuclei could explain their differential vulnerability. Our previous high-resolution functional MRI study of emotional processing by the amygdala subnuclei demonstrated that both the centromedial and basal amygdala groups were most sensitive to negative emotional stimuli.⁴⁷ In contrast, we reported that the lateral nucleus was not particularly sensitive to negative emotions, suggesting that this structure might play a less important role in this process. Previous functional MRI studies have consistently shown hyperactivity of the amygdala in response to threatening stimuli in maltreated individuals,¹⁷ but which subnuclei are responsible for these functional changes remains to be determined. In contrast to functional studies, the findings of structural studies in the amygdalae of maltreated individuals have been heterogenous. Some studies have reported volumetric reduction^35,36 or volumetric increase,^33,34 but others found no significant effects of maltreatment on amygdala volume in healthy people^30,31 or in psychiatric patients.³² Nevertheless, the timing and type of exposure to the early adverse environment could account for these heterogeneous results. ¹⁷ It has been suggested that volumetric enlargement has been reported by studies that investigated the effects of early maltreatment (mainly emotional and/or physical neglect) on the amygdala, and volumetric decrease has been reported by studies that investigated older adolescents or adults with more psychopathological symptoms exposed to different types of maltreatment across their development.¹⁷ As a result, an early adverse environment might increase amygdala volume during childhood, but further exposure to a stressful environment during adulthood could cause a volumetric reduction because of the biological embedding induced by exposure to an early adverse environment.¹⁷

Our findings related to the negative relationship between amygdala volume and childhood adversity were in line with those of our previous study,⁴⁸ which showed that amygdala volumes in participants with MDD with a history of childhood sexual or physical abuse were smaller than in participants with MDD who did not have such a history.

Stress-integrative functions are also localized in the amygdala subnuclei.⁶ The centromedial amygdala is the primary regulator of emotional and physiologic responses to acute stressors, and the basolateral amygdala plays a role in the integration of chronic stress.⁶ Animal studies have linked chronic stress to depression-related behaviours.¹² During stress, the basolateral amygdala undergoes plastic changes, with increased dendritic spine density in response to acute stress and increased length and arborization of pyramidal cell dendrites following chronic stress.^12,37,49 In contrast, chronic immobilization stress did not change the dendrites of the central amygdala,^37,49 and in the medial amygdala it was associated with both a reduction in dendritic spine density⁵⁰ and no significant change.⁵¹ Although our findings on the negative associations between basolateral volume and CTQ scores differed from animal studies, most of those studies used rodents whose ages were relevant to human adulthood and not to childhood. Therefore, future animal studies for comparison could investigate the effects of chronic stress in adult animals exposed to stress during their weaning and adolescent periods.

In the present study the negative associations between total CTQ score, emotional abuse (and total abuse history) and amygdala volumes were observed mainly in the right hemisphere. This finding was in line with a recent structural MRI study in emerging adults that reported negative correlations between adverse childhood experience and the volume of the basolateral and centromedial amygdala (including central, medial and cortical nuclei) in the right hemisphere only.³⁸ Furthermore, basolateral volumetric reduction in the right hemisphere was associated with an increase in symptoms of anxiety and depression, and of alcohol use.³⁸ Moreover, in response to sad stimuli, activation of the right amygdala was positively correlated with CTQ factors, especially physical abuse.⁵² A recent meta-analysis demonstrated that those who experienced childhood maltreatment had smaller amygdalae in both hemispheres, but the effect sizes for the right amygdala were larger in men.³⁶ Our findings were consistent with a meta-analysis that showed that people who experienced childhood maltreatment had significantly smaller right amygdala volumes than people who were not exposed to childhood maltreatment.³⁵

In line with these findings, previous studies reported hemispheric asymmetries for emotional processing in the amygdala⁵³ and suggested that the right amygdala is involved in the rapid detection of threatening stimuli or the early processing of affective stimuli via right hemisphere subcortical circuits, and the left amygdala is preferentially activated during decoding of cognitive-related emotional stimuli via the left hemisphere’s slower cortical feedback mechanisms.⁵³ Therefore, we speculate that the observed negative relationships between right basal and accessory basal nuclei volumes and a history of abuse that is considered a threat⁵⁴ might be due to the critical role of the right amygdala in processing threat-related stimuli.⁵³ Such stimuli sensitize the right amygdala to stressful situations later in life, causing volumetric reduction.^17,36

Although previous meta-analyses found no significant differences in amygdala volume between participants with MDD and controls,^2,3,8–10 individual studies have reported an increase in total amygdala volume with MDD,⁴⁸ a decrease⁵⁵ and no effect.⁵⁶ Similarly, postmortem studies have reported no volumetric or neuronal density difference in the amygdala between participants with MDD and controls,⁵⁷ a trend toward amygdala volume reduction in people with MDD,⁵⁸ or a modest (11%) volumetric increase in the lateral nucleus with an increase in the number of neurovascular cells in the basolateral group in people with MDD, without significant differences in the total number and densities of neurons and glia in the basolateral group.⁵⁹ Some of the confounding variables that might contribute to these conflicting results include the demographic and clinical characteristics of the people with MDD, as well as the anatomic definition of the amygdala.^2–4,9,10 We have found larger amygdalae in participants with MDD.⁴⁸ However, the results of the present study were in agreement with previous studies that demonstrated similar amygdala volumes in participants with MDD and healthy controls.^2,9 Although in the present study we employed an identical anatomic definition of the amygdala and used the same criteria to recruit participants with MDD, differences in the field strength of the MRI scanners between our 2 studies (4.7 T v. 1.5 T), and the sex ratio of participants with MDD might have contributed to the discrepancies. Similar to previous studies,^2,55,56 we found no correlations between amygdala volume and MDD duration or severity. Finally, we found no associations between amygdala volumes and antidepressant treatment, in agreement with our previous findings.⁴⁸ Meta-analyses^2,9 and a postmortem study⁵⁹ found no association between medication status and amygdala volume or effects of antidepressant treatment on the basolateral group, respectively.

Effects of adverse experiences in childhood/adolescence on hippocampal subfields

It has been suggested that the hippocampus might be the brain region most vulnerable to stress.¹⁷ Our findings indicated that of all hippocampal subfields and subregions, the CA1–3 and the anterior hippocampus showed the strongest negative associations with childhood adversity. In contrast to our findings for the amygdala, which were limited to a history of emotional abuse, our findings for the effects of childhood adversity on the hippocampus include histories of emotional and physical abuse and neglect.

Previous studies have reported hippocampal volumetric reductions in maltreated healthy individuals,³⁰ participants with MDD⁶⁰ and maltreated individuals regardless of the absence or presence of psychiatric disorders.^17,29,36 It is important to note that the values of the CTQ scores in our MDD sample were similar to those reported previously in participants with MDD.^48,52,61 The CTQ scores of participants with MDD were generally much higher than those of healthy participants in studies that compared these 2 groups to each other.^48,52,61 Because hippocampal volume can be affected by childhood adversity, MDD itself or both, future neuroimaging studies in MDD need to include a control group with a similar level of exposure to adverse childhood experiences to determine if any observed volumetric differences are associated with chronic stress or with MDD diagnosis.

The results of the present study agree with those of previous studies in MDD participants²⁹ and healthy controls¹⁸ that suggested that childhood adversity affects the anterior hippocampus. ^5,18,29 In addition, our previous study showed that cortisol levels were negatively correlated with anterior hippocampus and CA1–3 volumes in both participants with MDD and healthy controls.⁶²

Our results were also in agreement with those of a previous study that reported negative relationships between childhood maltreatment and volumes of the CA2–3 subfield.¹⁹ Although we did not find statistically significant associations between the dentate gyrus and total CTQ-25 score, we believe that these differences could be partially explained by different sample characteristics and the effects of antidepressant treatment on the dentate gyrus in our MDD cohort.²¹

Animal models of chronic stress have demonstrated that although chronic stress induced dendritic atrophy in both the dorsal and ventral hippocampus, the CA3 was the most vulnerable hippocampal subfield.²⁶ These findings are in agreement with our results. Dendritic retraction in CA1 and the dentate gyrus in animals occurred in response to more severe chronic stress.²⁶ We have demonstrated that the largest proportion of the CA1–3 subfield is in the anterior hippocampus, likely making this subregion a main target of chronic stress.²⁰ In contrast, the largest proportion of the dentate gyrus is in the posterior hippocampus (i.e., the body), which is particularly affected by MDD,²¹ suggesting that the effects of MDD diagnosis and stress history may be distinct.

In contrast to our findings for the amygdala, our findings of hippocampal changes related to childhood adversity were in accordance with preclinical studies. Explanations for these findings might arise from the different vulnerability windows in postnatal development of these structures. Although the hippocampus achieves dramatic growth in the first 2 years of life, the amygdala continues to develop until young adulthood.⁴⁴ Therefore, exposure to childhood adversity likely does not correspond to the postnatal window of hippocampal vulnerability but could occur simultaneously with amygdala development.

Limitations

Our study sample was cross-sectional. Although the significant negative associations we found between childhood maltreatment and the volume of the right amygdala were in accordance with the findings of previous studies,^35,38 our results for the amygdala subnuclei should be interpreted with caution, because none of those findings remained significant after Holm–Bonferroni correction. Studies that incorporate a treatment-naïve group of MDD patients with and without history of childhood adversity are needed to separate the effects of MDD, history of abuse and antidepressant treatment on hippocampal subfields and amygdala subnuclei. In general, methods used in volumetric studies of human hippocampal subfields⁶³ and amygdala subnuclei¹⁴ in vivo do not exactly match the intra-anatomic boundaries of these structures; instead, different geometrical rules are employed that approximately match the location and the orientation of these structures based on histological references.

Conclusion

This study provides in vivo evidence that childhood adversity might have negative effects on specific brain subregions in participants with MDD, namely the basolateral amygdala and the CA1–3 hippocampal subfield.