Hippocampal subregion volume alterations at baseline and following antidepressant treatment in youth at high risk for bipolar disorder

Abstract

Abstract

Understanding hippocampal neurostructural alterations at baseline and following antidepressant treatment in high-risk youth, particularly within distinct subregions, is critical for elucidating bipolar disorder (BD) pathophysiology and treatment planning. In this prospective randomized clinical trial, 105 high-risk youth with a first-degree BD-I relative(s) and moderate to severe depression and/or anxiety disorder, along with 51 healthy controls (HCs), were recruited. All participants underwent baseline structural MRI and clinical evaluations for depression, anxiety, and adverse events. High-risk youth were then randomized to receive psychotherapy and with either escitalopram or placebo. After 4 weeks of treatment, high-risk youth underwent repeated structural MRI and clinical assessments. These clinical assessments continued biweekly over the remaining 12 weeks of treatment (up to week 16). Volumetric measurements of hippocampal subregions were analyzed at baseline and 4 weeks. Compared to HCs, high-risk youth had significant baseline volume increases in the right cornu ammonis (CA)3, bilateral CA4, bilateral granule cell and molecular layers of the dentate gyrus (GC-ML-DG), and bilateral molecular layer, but reduced volumes in the bilateral fimbria and left parasubiculum. Volume reduction in the left hippocampal tail after 4 weeks of escitalopram treatment was associated with improvements in disinhibition and impulsivity symptoms from baseline to 4 weeks, and this association remained significant from baseline to 16 weeks. CA3-4 and GC-ML-DG enlargement may reflect adaptive changes, while fimbria and parasubiculum atrophy indicate vulnerabilities in BD-risk youth. Escitalopram-related volume reduction in the left hippocampal tail may reflect neural plasticity associated with emotional symptom improvement in BD high-risk youth.

Clinical trial registration information

Mechanism of Antidepressant-Related Dysfunctional Arousal in High-Risk Youth; https://clinicaltrials.gov/; NCT02553161.

Introduction

Antidepressants are commonly prescribed to manage mood and anxiety disorders in youth, in spite of offering moderate benefits and being associated with potentially consequential psychiatric adverse events, such as treatment-emergent mania [1, 2]. The risk for such adverse events is particularly pronounced in youth with a family history of bipolar disorder (BD), where the intersection between antidepressant use and familial risk factors may exacerbate inherent vulnerabilities toward adverse outcomes within key neurobiological systems that regulate emotion [3]. The hippocampus, critical for the regulation of emotion, cognitive functions and response to stress [4, 5], plays a key role in understanding the pathophysiology and treatment of mood disorders, and undergoes significant plasticity with antidepressant exposure [6]. However, the specific impact of antidepressants on hippocampal structure — both in terms of efficacy and adverse events—remains inadequately explored in youth at familial risk for BD.

Hippocampal volumes in BD range from being reduced or not significantly different from healthy controls (HCs) [7–9]. Whole volume analyses may have missed unique functional variations in hippocampal subregions, which play distinct roles in the pathophysiology of BD [10, 11]. With advances in the resolution and accuracy of automatic hippocampal segmentations, fine-grained volume reductions have been observed in widespread hippocampal subfields such as cornu ammonis (CA)1-4, the granule cell and molecular layers of the dentate gyrus (GC-ML-DG), molecular layer, hippocampal-amygdala transition area (HATA), subiculum, and hippocampal tail in patients with BD [12–14]. Such patterns have also been supported by post-mortem neuropathological investigations of individuals with BD [15].

BD is among the most familial psychiatric disorders with high heritability rates; however, specific biomarkers of familial risk for BD have yet to achieve biomarker validation. Examination of anatomical variations in the hippocampus and its subregions in youth at high-risk (HR) for BD could determine whether these subregions could serve as risk factors for BD, potentially advancing predictive capabilities and preventative strategies for BD [16]. Like studies described above in BD, prior HR studies focused on total hippocampal volumes have reported inconsistent results, possibly due to volumetric averaging of its subregions. For example, in one study, increased volume in left parahippocampal/hippocampal gyrus was observed in HR youth compared to HCs [17], while another study noted trends for larger hippocampus volumes in unaffected co-twins discordant for BD compared to low-risk twins [16]. In a third study, Nery et al. reported decreased gray matter volumes in the left hippocampus and parahippocampal gyrus, and thicker right hippocampus in BD unaffected first-degree relatives compared to HCs [18]. However, Hajek et al. reported comparable hippocampal volumes among unaffected relatives, affected relatives for BD, and HCs [19]. These studies underscore the significant variability of hippocampal volume differences among those at risk for BD. More fine-grain examinations of hippocampal subregions in youth at risk for BD, and in the context of symptom manifestations and treatment, may uncover subtle predispositions for BD than reported in the extant literature.

Given that antidepressant exposure may be a major confound in the evaluation of bipolar risk, directly examining the effects of antidepressants on the hippocampus and its subregions provide a unique opportunity to both advance understanding of risk for developing BD and the effect of antidepressant treatment on hippocampal plasticity. Indeed, it is well known that antidepressant treatment can alter the structure of the hippocampus across species, and in specific subregions [20]. Antidepressants can both lead to adverse events or promote neurogenesis and enhance synaptic plasticity in the hippocampus, as antidepressants increases the dendritic spine density of neurons in CA pyramidal cell subfields and stimulates neurogenesis in the dentate gyrus [21–23], suggesting high potential for structural remodeling. Indeed, antidepressant exposure can lead to significant alterations in hippocampal subregion volumes (such as dentate gyrus) in rodent models, modulating function in areas critical for stress response and emotional regulation [24, 25]. These insights provide a foundation for understanding how antidepressants might similarly affect the hippocampal architecture in HR youth for BD given that it is not yet unknown whether antidepressant-related hippocampal plasticity in humans is adaptive (toward improved symptoms) or linked to a risk of psychiatric adverse events. Direct evidence linking these neurobiological mechanisms to treatment outcomes and psychiatric adverse events in this vulnerable group is now needed. Youth at risk for BD are uniquely suited to examine adaptive versus maladaptive effects of antidepressants, as they may be at high risk for treatment-emergent mania compared to the general population [2].

Given these considerations, we examined hippocampal subregion volume differences compared to typically developing youth in large cohort of HR youth for BD with no prior antidepressant exposure, and then investigated early hippocampal changes from baseline to 4 weeks in HR youth randomized to escitalopram or placebo to treat symptoms of depression and anxiety. Based on the extant literature, we hypothesized that: 1) HR youth will exhibit abnormal volumes of the hippocampus and its subregions (such as CA1-4, dentate gyrus, and subiculum) compared to HC youth at baseline [12–14]; 2) Antidepressants will result in volumetric increases in subregions (such as CA and dentate gyrus) in HR youth who show early (within 4 weeks) improvements in depression or anxiety symptoms, and volume reductions in the other hippocampal subregions, in HR youth who develop antidepressant-related adverse events from baseline to 4 weeks; 3) Volumetric changes will correlate with changes in treatment-related efficacy or adverse events within respective treatment-responsive and adverse event subgroups over a 16-week full course of treatment. By conducting a thorough examination of the hippocampus and its subregions in the context of a clinical trial that can examine early and sustained treatment effects, our study aims to ultimately advance understanding of the earliest neurobiological underpinnings of BD risk and shed light on how antidepressant-driven changes in hippocampal structure might influence clinical outcome acutely and after a course of standard treatment.

Methods

Participants

Youth with a first-degree relative diagnosed with bipolar I disorder (n = 119), and age- and sex-matched HCs (n = 55) without mood or psychotic disorders in any first or second-degree relatives, were recruited from the University of Cincinnati and Stanford University. This study followed the protocol [3] as registered on ClinicalTrials.gov (NCT02553161, URL: https://clinicaltrials.gov/ct2/show/NCT02553161) on December 1, 2015. The structural MRI analyses presented in this manuscript are reported as secondary/exploratory outcomes of the trial. The inclusion criteria the HR youth were: (1) confirmed family history of BD based on the Family History-Research Diagnostic Criteria-Modified (FH-RDC) [26], the Structured Clinical Interview for DSM (SCID) [27], and/or the Kiddie Schedule of Affective Disorders and Schizophrenia (KSADS) [28]; (2) notable current depressive and/or anxiety symptoms, assessed by the Childhood Depression Rating Scale-Revised (CDRS-R) over 35 [29] and/or the Pediatric Anxiety Rating Scale (PARS) above 15 [30], (3) aged 12–17 years; and (4) no history of antidepressants. Exclusion criteria for all participants included: (1) a history of bipolar I or II disorder, (2) patients with any DSM-5 psychiatric comorbid disorders, such as autism spectrum, obsessive-compulsive, post-traumatic stress, or Tourette’s disorders, (3) intellectual disability (full scale IQ < 70), (4) drug and/or alcohol abuse or dependence.

The study was approved by the Institutional Review Boards at both the University of Cincinnati and Stanford University (IRB ID: 2021-0102, approval date: 3/24/2021), and both participants and their legal guardians provided written assent and informed consent, respectively. All methods were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki.

Study design

All participants received Magnetic Resonance Imaging (MRI) scans and assessments for psychiatric symptoms and potential adverse events at baseline. Following initial evaluations, HR youth were then randomized to receive psychotherapy plus escitalopram or psychotherapy plus placebo (2:1). The study was double-blind; participants, treating clinicians, and outcome assessors were blinded to treatment allocation. Full details of the randomization and blinding procedures are described in the trial protocol (NCT02553161).Treatment adherence was closely monitored through weekly clinical follow-ups during dose titration, followed by biweekly visits through week 16, pill counts at each visit (with adherence defined as taking ≥80% of prescribed doses), participant self-report, and legal guardian interviews, consistent with our prior studies demonstrating high adherence in this population [31, 32]. After 4 weeks of treatment, HR youth underwent follow-up MRI scans and assessments for psychiatric symptom changes and monitoring for any evolving acute adverse events or evidence of early treatment response. Then, assessments for psychiatric symptoms and adverse events were collected biweekly in the remaining 16 weeks of treatment to assess for treatment effects after a full 16 week course of treatment. Psychiatric symptoms were assessed using the clinician-rated Young Mania Rating Scale (YMRS) [33], the clinician-rated CDRS-R, and the clinician-rated PARS. Adverse events were rated using the clinician-rated Pediatric Adverse Events Rating Scale (PAERS) [34] and clinician-rated Treatment-Emergent Activation and Suicidality Assessment Profile (TEASAP) [35]. Specifically, from the PAERS scale, 13 items (items 1, 2, 6–11, 14, 17, 28, 29, and 36) deemed to be related to treatment-emergent mania were selected, and their clinician-rated scores were aggregated to define our primary outcome, assessing antidepressant-related adverse events. Please see Table S1 for a detailed information on these items. As a secondary outcome, the TEASAP measured adverse event severity across five dimensions: irritability, akathisia and somatic anxiety, disinhibition and impulsivity, mania, and a combined dimension of self-injury, suicidality, and harm to others.

MRI acquisition

The MRI scans were conducted on 3-Tesla scanners equipped with 8-channel phased-array head coils at two centers: a Phillips scanner at the University of Cincinnati and a GE scanner at Stanford University. High-resolution T1 images were acquired with similar imaging parameters across both centers, as detailed in Table S2. Images were visually inspected by an experienced pediatric neuroradiologist to identify scans with visible cerebral abnormalities, distortions or artifacts for exclusion.

Neuroimaging data preprocessing and volumetric analysis of hippocampal subregions

Structural image preprocessing was performed using the open-source FreeSurfer software (version 7.0). To reduce intra-subject variability, we applied FreeSurfer’s specialized longitudinal stream [36], which involves creating a common template using all scans of a subject, and reprocessing each scan with this shared template. The longitudinal segmentation pipeline for these regions utilized a probabilistic atlas based on high-resolution ex vivo MRI data ( ~ 0.1 mm isotropic), producing an automated joint segmentation of these substructures. This longitudinal segmentation stream, increased test-retest reliability and reduced measurement noise compared to independent segmentations, offering more consistent segmentations over time [37].Two hierarchical segmentation levels were produced: (a) a broader categorization into head, body, and tail; and (b) a detailed division into parasubiculum, presubiculum, subiculum, CA1, CA3, and CA4, GC-ML-DG, molecular layer, HATA, fimbria, tail, and fissure. Finally, we extracted volumes of bilateral hippocampi and their subregions. Figure S1 presents an example of the hippocampal segmentation in a healthy participant.

All segmentations were thoroughly reviewed in all three orthogonal planes, slice by slice, to ensure accuracy. Using standard ENIGMA protocols [38], any segmentations that mistakenly included non-target tissue or missed portions of the hippocampus were excluded. To address variances in volume measurements attributed to different sites, we applied calibration techniques from the Functional Biomedical Informatics Research Network and conducted a traveling subjects study. Scan parameters were optimized for comparability across sites using recommendations from the Functional Biomedical Informatics Research Network for multi-site fMRI studies [39]. Three volunteers were scanned on 2 different days at each site and showed high test-retest consistency between sites and good comparability of fMRI results across sites. Finally, scanner site was conservatively included as a covariate in all imaging analyses.

Statistical analysis

Volumetric differences in the hippocampus and its subregions between High-risk BD and healthy controls at baseline

We first examined baseline volume differences in the hippocampus and its subregions between HR youth and HCs using General Linear Model, with age, sex, and total intracranial volume (TIV) as covariates. The significance thresholds were set at p < 0.05, with false discovery rate (FDR) to correct for multiple comparisons. The sample size for this analysis was based on the availability of usable imaging data from the parent study, and no a priori power calculation was performed.

Post-treatment volumes changes of hippocampus and its subregions

To evaluate whether antidepressants treatment had significant effects on the volumes of hippocampus and its subregions in the HR youth, we employed linear mixed models to estimate the treatment (escitalopram vs. placebo)-by-time (baseline vs. week 4) interaction effects on volumes of hippocampus and its subregions, controlling for age, sex, and TIV. We maintained p < 0.05 as the threshold for significance, applying FDR correction for multiple comparisons. A post-hoc analysis was conducted to analyze differences between the escitalopram and placebo treatments at baseline and week 4, and within-group changes from baseline to week 4, for each hippocampal subregion where significant interactions were found.

Associations between post-treatment volume alterations in the hippocampus and its subregions with clinical outcomes

We employed linear regression models to investigate the relations between volumetric changes in hippocampal subregions and clinical ratings from baseline to week 4 and from baseline to week 16 after starting escitalopram or placebo administration in HR youth. These clinical ratings include both adverse event (including PAERS and TEASAP, as well as their sub-score categories) and clinical efficacy (CDRS/PARS). In addition, we examined whether these correlations were present within each treatment group, and whether they displayed distinct patterns between the escitalopram and placebo groups. Correlations between imaging changes and clinical ratings were exploratory; thus, p-values are reported without multiple comparison correction.

Results

From December 2015 to December 2022, we recruited 120 HR youth and 55 HCs. One HR youth was subsequently excluded, and the remaining 119 were randomized to receive either psychotherapy plus escitalopram (n = 75) or psychotherapy plus placebo (n = 44). For the primary analysis, MRI data were included from 105 of these HR participants (66 from the escitalopram group and 39 from the placebo group) and 51 HCs (see Fig. 1).

Fig. 1. Participant recruitment, randomization, and analysis flowchart.

A total of 120 HR youth and 55 HCs were recruited. One HR youth was excluded, and 119 were randomized to psychotherapy plus escitalopram (n=75) or psychotherapy plus placebo (n=44). For primary MRI analysis, 105 HR youth (66 escitalopram, 39 placebo) and 51 HCs were included. HR high-risk youth for bipolar disorder, HC healthy control.

Demographic and clinical characteristics for all participants

The HR youth and HCs did not significantly differ in age, sex, handedness, race, or ethnicity; similarly, HR youth randomized to escitalopram were demographically similar to those randomized to placebo (Table 1). At baseline, there were no significant differences in diagnosis (including depression, anxiety, and depression with anxiety), or clinical rating scales such as CDRS, PARS, PAERS and TEASAP between the two HR treatment groups (Table 1). For most HR youth in both treatment groups, clinical rating scores showed significant decreases over time from baseline to both week 4 and week 16 (all p < 0.001). However, no significant difference was observed between the two treatment groups in changes of those clinical measures (all p > 0.05) (Table 1, Table S3).

Table 1.

Demographic and clinical characteristics of high-risk youth for bipolar disorder with escitalopram or placebo.

Characteristics	Escitalopram + Psy (n = 66)	Placebo + Psy (n = 39)	Statistics	p
Age (years)	14.88 ± 1.60	15.03 ± 1.57	−0.451	0.653a
Female, N (%)	36 (54.5%)	27 (69.2%)	1.633	0.201b
Diagnosis, N (%)			0.115	0.944b
anxiety-only	3 (4.5%)	2 (5.1%)
depression & anxiety	51 (77.3%)	29 (74.4%)
depression-only	12 (18.2%)	8 (20.5%)
Baseline CDRS	52.83 ± 11.37	56.33 ± 12.5	−1.433	0.156a
Baseline PARS	12.77 ± 5.22	12.59 ± 5.46	0.169	0.866 a
Baseline YMRS	8.20 ± 5.99	7.18 ± 4.98	−0.937	0.351a
Baseline PEARS	8.14 ± 7.18	7.49 ± 7.74	−0.427	0.671a
Baseline TEASAP
total	32.56 ± 20.16	34.08 ± 20.16	0.372	0.711a
irritability	13.09 ± 6.60	13.49 ± 7.39	−0.276	0.783a
akathisia and somatic anxiety	5.02 ± 4.06	5.67 ± 3.94	−0.810	0.420a
disinhibition and impulsivity	6.27 ± 5.13	6.26 ± 5.21	0.016	0.988a
mania	5.30 ± 4.99	5.31 ± 4.90	−0.005	0.996a
self-injury, suicidality, and harm to others	2.88 ± 3.36	3.36 ± 3.22	−0.727	0.469a
∆CDRS	−19.76 ± 11.70	−19.18 ± 11.08	−0.249	0.804a
∆PARS	−6.03 ± 5.27	−4.33 ± 4.00	−1.737	0.085a
∆PAERS	−5.80 ± 6.55	−4.03 ± 7.26	−1.291	0.200a
∆TEASAP
∆total	−20.60 ± 17.90	−21.40 ± 16.30	0.250	0.803a
∆irritability	−8.08 ± 6.73	−7.08 ± 6.23	−0.755	0.452a
∆akathisia and somatic anxiety	−3.06 ± 3.57	−3.82 ± 3.37	1.076	0.284a
∆disinhibition and impulsivity	−4.00 ± 4.18	−4.10 ± 4.13	0.122	0.903a
∆mania	−3.47 ± 4.45	−3.62 ± 3.90	0.170	0.866a
∆self-injury, suicidality, and harm to others	−1.95 ± 2.74	−2.82 ± 3.16	1.478	0.142a

Psy Psychotherapy, CDRS children’ depression rating scale, PARS pediatric anxiety rating scale, YMRS young mania rating scale, CGAS children’s global assessment scale, PAERS Pediatric adverse events rating scale, TEASAP Treatment-emergent activation and suicidality assessment profile.

∆ indicated changes from baseline to week 4 or a nearby assessment (week4-baseline).

^aIndependent-sample t test.

^bChi-squared test.

Volume Differences of Hippocampus and its Subregions in High-Risk Youth at Baseline

Compared to HCs, HR youth had significantly increased volumes in the right CA3, bilateral CA4, bilateral GC-ML-DG, and bilateral molecular layer, but reduced volumes in the bilateral fimbria and left parasubiculum at baseline (all FDR-corrected p < 0.05, Table 2). Figure S2 provides bar plots showing detailed baseline volume comparisons among the three groups (HCs, HR treated with escitalopram, and HR treated with placebo).

Table 2.

Baseline volume differences in hippocampus and its subregions between high-risk youth for bipolar disorder and healthy controls.

Brain regions	HC (n = 51)	HR (n = 105)	F	FDR corrected p values
	Mean ± SD	Mean ± SD
L whole HP	3514 ± 304	3521 ± 389	0.346	0.557
R whole HP	3595 ± 331	3642 ± 422	1.541	0.216
Broad HP subregions
L head	1742 ± 178	1745 ± 189	0.441	0.746
L body	1200 ± 120	1218 ± 132	1.651	0.402
L tail	572 ± 62	576 ± 88	0.244	0.746
R head	1798 ± 187	1830 ± 196	2.982	0.258
R body	1202 ± 122	1237 ± 146	3.797	0.258
R tail	594 ± 65	596 ± 100	0.089	0.765
Detailed HP subregions
L parasubiculum	74 ± 11	67 ± 11	11.372	0.003*
L presubiculum	321 ± 40	309 ± 38	2.376	0.214
L subiculum	443 ± 51	435 ± 50	0.482	0.587
L CA1	651 ± 68	626 ± 76	3.754	0.102
L CA3	205 ± 26	213 ± 35	3.934	0.098
L CA4	247 ± 25	282 ± 34	64.914	< 0.001*
L GC-ML-DG	290 ± 29	313 ± 34	30.544	< 0.001*
L molecular layer	567 ± 53	588 ± 62	8.317	0.011*
L HATA	60 ± 9	57 ± 8	1.679	0.278
L fimbria	85 ± 15	71 ± 15	31.696	< 0.001*
L tail	572 ± 62	576 ± 88	0.244	0.652
L fissure	149 ± 27	154 ± 29	1.870	0.276
R parasubiculum	71 ± 11	69 ± 10	0.413	0.595
R presubiculum	302 ± 32	295 ± 36	0.587	0.562
R subiculum	436 ± 50	425 ± 44	1.537	0.289
R CA1	688 ± 78	667 ± 85	1.784	0.276
R CA3	223 ± 27	240 ± 39	12.084	0.003*
R CA4	255 ± 27	300 ± 38	85.788	< 0.001*
R GC-ML-DG	298 ± 30	330 ± 39	43.951	< 0.001*
R molecular layer	585 ± 58	614 ± 66	12.693	< 0.001*
R HATA	60 ± 7	59 ± 9	0.240	0.652
R fimbria	83 ± 16	67 ± 13	44.399	< 0.001*
R tail	594 ± 65	596 ± 100	0.089	0.765
R fissure	147 ± 25	159 ± 26	7.968	0.011*

HR high-risk youth for bipolar disorder, HCs healthy controls, SD standard deviation, FDR false discovery rate, Lleft R, right, HP hippocampus, CA cornu ammonis, GC-ML-DG granule cell and molecular layers of the dentate gyrus, HATA hippocampal-amygdaloid transition area.

*indicates adjusted p < 0.05, signifying significant results.

Post-treatment Volume Alterations of Hippocampus and its Subregions in High-Risk Youth

A significant treatment-by-time interaction was observed in the volume of the left hippocampal tail (F = −2.767, FDR-corrected = 0.040, Fig. 2, Table S4) from baseline to 4 weeks. Post-hoc analysis revealed a significant decrease in the volume of this subregion in the escitalopram treatment group (t = 3.080, p = 0.003, Fig. 2), while the placebo group had no significant changes from baseline to week 4.

Fig. 2. Volume change of left hippocampal tail over the course of treatment and correlation to change of clinical arousal rating in high-risk youths for bipolar disorder.

The green line and dots represent patients treated with psychotherapy plus escitalopram (esc), and the pink line and dots represent patients treated with psychotherapy plus placebo (pla). The correlations in the escitalopram group at week 4 and week 16 were exploratory and did not survive correction for multiple comparisons (uncorrected p values reported). L left, HP hippocampus, esc psychotherapy plus escitalopram, pla psychotherapy plus placebo, TEASAP Treatment-Emergent Activation and Suicidality Assessment Profile, dis disinhibition and impulsivity. ∆^{(4w-baseline)} represents the change after 4 weeks treatment. ∆^{(16w-baseline)} represents the change after 16 weeks treatment. ** p < 0.01: Significant volume changes after 4 weeks treatment. The green line and dots represent patients treated with pla, and the pink line and dots represent patients treated with esc.

Correlation between post-treatment volume alterations and changes in clinical outcomes

A positive correlation was observed between the change in volume of the left hippocampal tail and the change in TEASAP disinhibition and impulsivity subscales after 4 weeks escitalopram treatment (r = 0.267, p = 0.030, Fig. 2), but not in the placebo group. This volume change was also significantly associated with change in TEASAP disinhibition and impulsivity subscales after 16 weeks of treatment in the escitalopram group (r = 0.317, p = 0.034, Fig. 2).

Discussion

This current study, utilizing a relatively large sample of youth at high risk for BD, assessed the morphological differences and treatment effects in the hippocampus and its subregions compared to HCs early and after a 16-week course of treatment. We found larger hippocampal sub-regional volumes in the right CA3, bilateral CA4, bilateral GC-ML-DG, and bilateral molecular layer, but reduced volumes in the bilateral fimbria and left parasubiculum in the HR youth at baseline in comparison to healthy comparison youth. Importantly, we investigated the differential volumetric changes in the hippocampus and its subregions following treatment with psychotherapy combined with either escitalopram or a placebo in HR youth, as well as how these changes correlate with treatment outcomes at 4 and 16 weeks. We found that, volume reduction in the left hippocampal tail after 4 weeks of escitalopram was associated with normalization (reduction) of a subset of mood symptoms reported in TEASAP disinhibition and impulsivity subscales. This structural change also significantly predicted symptom change from baseline to week 16.

The observed baseline enlargement in specific hippocampal subregions (such as the right CA3, bilateral CA4, and GC-ML-DG) in youth at high-risk for BD compared to HCs, contrast with selectively reported atrophy of both the whole hippocampus and its subregions in BD [10, 12, 14]. This enlargement may reflect an adaptive mechanism in the background of clinical, familial, and environmental risk factors for BD. A prior high-risk study [16] similarly reported increases in hippocampal volume, including similar enlargements in specific subregions such as CA3-4 and GC-ML-DG. The CA3 and CA4 subregions plays a pivotal role in memory consolidation, spatial navigation and stress response [40]. The enlargement of CA3 and CA4 might enhance these cognitive functions, potentially protecting youth from the memory and spatial processing deficits often observed in BD and in the context of early depression and anxiety symptoms. The GC-ML-DG and molecular layer are key sites for neurogenesis (i.e., the process of generating new neurons), which is closely associated with mood regulation and cognitive flexibility [41]. Thus, the enlargement of GC-ML-DG and molecular layer in youth at high risk for BD could be protective, priming youth to cope with stress through this potential neurobiological mechanism of resilience.

In contrast, reduced volumes in the bilateral fimbria and left parasubiculum in HR versus HC may represent a neuroanatomical vulnerability within the hippocampal circuitry that functions to promote effective emotional and stress response regulation. The fimbria, part of the fimbria-fornix complex, and the parasubiculum are pathways for hippocampal outputs, facilitating connections to other brain regions such as the thalamus, cingulate cortex, and nucleus accumbens [42, 43]. These neural circuits are essential for the regulation of the hypothalamic-pituitary-adrenal axis and mood and are broadly implicated in affective disorders. Smaller fimbria and reductions in the parasubiculum may disrupt these pathways leading to emotional dysregulation characteristic of BD or mania, as supported by the decrease in fractional anisotropy in the fornix (which is continuous with the fimbria) in adolescents with bipolar I disorder [44].

In this study, we evaluated the tolerability and efficacy of psychotherapy combined with either escitalopram or placebo in youth with a family history of BD, presenting with depression and anxiety symptoms but no prior lifetime manic episodes. After 4 weeks of treatment, both groups showed significant improvements in managing symptoms of depression and anxiety, without any emergence of significant adverse events compared to baseline. A lack of significant group differences in emergence of adverse events, as measured by PAERS and TEASAP scores, indicates that escitalopram and placebo showed comparable tolerability and no increased risk of antidepressant-related adverse events such as mania. Moreover, parallel reductions in CDRS and PARS scores across treatment groups demonstrate that escitalopram and placebo are similarly effective when combined with psychotherapy for treating depression and anxiety. Interestingly, volume reductions in left hippocampal tail observed following 4 weeks of escitalopram treatment, in contrast to no change with placebo, suggest a differential neurobiological response to treatment [45]. Given its unique structural and vascular properties—such as highly folded CA1 regions and relatively poor vascular supply—the hippocampal tail may be particularly responsive to stress modulation and serotonergic effects [46]. The positive correlation between volume reduction in the left hippocampal tail and the normalization of emotional symptoms, as reflected in TEASAP disinhibition and impulsivity subscales, may similarly point to escitalopram-specific hippocampal neuroplasticity through mechanisms such as synaptic homeostasis [47]. This neuroadaptive process may enhance synaptic efficiency and reorganize neural networks that improves emotional regulation and reduces risk for impulsivity and disinhibition characteristic of mania. Thus, this volume reduction may reflect a neuroadaptive process, potentially representing an increase in synaptic efficiency and a reorganization of neural networks that improves emotional regulation. Notably, this early volume reduction also significantly predicted symptom improvement at 16 weeks, suggesting that it has sustained clinical relevance rather than being a transient phenomenon. Monitoring early hippocampal tail volume changes may therefore serve as a potential indicator of both short-term and long-term antidepressant treatment outcomes. This finding adds to efforts using neuroimaging measures to characterize biological features of treatment response[48]. Volumetric changes in the hippocampal tail could also be attributable to concurrent psychotherapy which may itself enhance the capacity for emotional regulation and stress resilience independent of pharmacological effects, potentially stabilizing or augmenting putative neuroplastic changes induced by escitalopram [49]. Future research should aim to further elucidate the specific neuroplastic mechanisms involved and explore the full trajectory of early and late structural changes in the context of treatment.

Limitations

We note several limitations of our study. First, the 4-week MRI follow-up period, while sufficient to detect early structural changes in the left hippocampal tail, may be limited in capturing the full trajectory of antidepressant-related hippocampal plasticity. Second, as the majority of participants were living with first-degree relatives with BD, this familial context could differentially influence their BD risk, depending on how well adjusted the family environment might be. However, the uniform application of family-focused therapy across all groups aimed to mitigate the impact of any differences in familial context on study outcomes. Third, HCs were not included in the follow-up phase, limiting our ability to compare the observed changes with those in typically developing youth, as significant changes in hippocampal volumes over a 4-week observational period would not be expected in typically developing youth. Fourth, given that both treatment protocols included psychotherapy combined with either escitalopram or a placebo, differences in volumetric changes between the two groups could be influenced by the psychotherapy component. This makes it challenging to exclusively attribute the differential changes to the effects of escitalopram or placebo alone. However, the uniform application of family-focused therapy across all groups aimed to mitigate the impact of psychotherapy on study outcomes. Fifth, serum medication levels were not available for the current analysis, precluding examination of dose-response relationships. However, treatment adherence was rigorously monitored through pill counts ( ≥ 80%), clinical follow-ups, and participant/guardian interviews, with high adherence demonstrated in our prior studies [31, 32]. Lastly, our study was limited to analyzing the structural features of the hippocampus, hypothesized to be relevant to BD and the effects of antidepressants. Exploring additional brain regions and employing functional imaging could identify additional neurobiological biomarkers, potentially enhancing our understanding of antidepressant treatment response on the developing brain in the context of bipolar risk.

Conclusion

Our finding that enlargement in the CA3-4 and GC-ML-DG in youth at risk for BD may indicate an adaptive response to factors intrinsic to BD risk. Conversely, fimbria and parasubiculum atrophy might reflect neuroanatomical vulnerabilities in HR youth. Additionally, volumetric reduction in the left hippocampal tail following 4 weeks of antidepressant treatment could suggest that early hippocampal tail plasticity may serve as a potential biomarker of escitalopram response in youth at risk for BD.

Author contributions

LRP, TJB, JAW, JRS, KMB, MKS, and MPD designed the study and collected clinical and image data. BT and WZ analyzed the image data and drafted the manuscript. SL, MKS, and MPD provided guidance and suggestions to data analysis. BT, WZ, SL, JRS, MKS, and MPD contributed to manuscript revision. All authors reviewed and approved the manuscript for submission.

Funding

This study was supported by NIMH grant R01 MH097818 to Dr. DelBello, R01 MH105469 and the Brain and Behavior Research Foundation to Dr. Singh, and the National Natural Science Foundation of China (Grants 82120108014 and 82441007) to Dr. Lui.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy or ethical restrictions.The trial protocol and statistical analysis plan are also available from the corresponding author upon reasonable request.

Competing interests

Dr. Patino and Dr. DelBello have received research support from NIMH, Patient Centered Outcomes Research Institute (PCORI), Acadia, Alkermes, Allergan, Janssen, Johnson and Johnson, Lundbeck, Myriad, Otsuka, Pfizer, Sunovion and Shire and Dr. DelBello has provided consultation or advisory board services for Alkermes, Allergan, CMEology, Janssen, Johnson and Johnson, Lundbeck, Medscape, Myriad, Pfizer, Sage, Sunovion. Dr. Singh has received research support from the National Institutes of Health, PCORI, and the Brain and Behavior Research Foundation. She is on the national advisory board for Skyland Trail and receives honoraria from the American Academy of Child and Adolescent Psychiatry (AACAP), and royalties from American Psychiatric Association Publishing. Dr. Strawn has received research support from AbbVie, PCORI, and the National Institutes of Health. He has provided consultation to Cerevel, Intracellular Therapeutics Otsuka and the FDA. He receives royalties from Springer Publishing, Cambridge and UpToDate and receives material support from Myriad. He has also received honoraria from Medscape Live, the Neuroscience Education Institute, the American Academy of Pediatrics, and AACAP. Dr. Zhang consult to VeraSci. All other authors declare that they have no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04314-w.