Neural Signatures of Bipolar Disorder and Psychotropic Medication Effects: A Multimodal PET–MRI Study

Abstract

Background:

Synaptic loss and altered network function in fronto-limbic brain regions have been implicated in the neurobiology of bipolar disorder (BD), but in vivo evidence remains limited. Synaptic vesicle glycoprotein 2A (SV2A), a biomarker of synaptic density, can be quantified using [¹¹C]UCB-J positron emission tomography (PET).

Methods:

Nineteen individuals with BD– including BD1 (n=13) and BD2 (n=6)– and healthy controls (HC; n=26) completed SV2A PET imaging in parallel with structural and functional MRI to assess fronto-limbic synaptic density, gray matter volume (GMV), and intrinsic connectivity, respectively.

Results:

Individuals with BD showed lower fronto-limbic SV2A density, which exploratory analyses revealed to be seemingly driven by those taking psychotropic medications (BD-med). Conversely, group differences in GMV were observed only when accounting for medication status, with higher GMV in BD-med relative to unmedicated BD individuals (BD-none). Connectivity was higher in BD overall, but lower in BD-med relative to BD-none. SV2A and GMV were positively correlated in HC. However, this relationship was absent in BD, with lower SV2A/GMV ratios being associated with greater impulsivity in BD.

Conclusions:

These findings provide the first in vivo evidence of synaptic deficits in BD and offer preliminary evidence suggesting psychotropic medications may differentially influence brain structure and function. These results underscore the importance of accounting for medication effects when interpreting neuroimaging findings in BD and the need to disentangle illness-related changes from medication effects on brain and behavior in BD.

INTRODUCTION

Bipolar disorder (BD) is a debilitating psychiatric condition characterized by recurrent episodes of mania and depression, affecting approximately 1 in 150 adults (0.53% of the global population)(1). BD is associated with significant functional impairment and carries one of the highest suicide risks among psychiatric disorders (2, 3). Decades of neuroimaging research have consistently identified structural and functional abnormalities in the fronto-limbic circuitry, a network critically involved in emotion regulation (4, 5). Despite these advances, the underlying neuropathological mechanisms of BD remain poorly understood. Clinically, there is currently no cure for BD. First-line pharmacological treatments, including mood stabilizers, antidepressants, and antipsychotics, offer only modest efficacy, with marked heterogeneity in patient response, high relapse rates, and burdensome side effects (6). Moreover, the neural mechanisms through which these medications exert their effects on brain function and clinical symptoms are not well characterized.

One proposed mechanism contributing to both the structural and functional abnormalities in BD is synaptic dysfunction. In support of this hypothesis, postmortem studies provide molecular evidence of lower levels of synaptic markers in fronto-limbic regions, including the prefrontal cortex and hippocampus(7, 8) (9, 10). However, synaptic alterations have not yet been studied in the living brain of people with BD. Recent advances in molecular imaging now allow for in vivo measurements of synaptic density with radiotracers targeting the synaptic vesicle protein 2A (SV2A), such as [¹¹C]UCB-J(11). SV2A is a ubiquitously and uniformly expressed presynaptic vesicle protein(12, 13) that correlates strongly with synaptophysin and other markers traditionally used in postmortem studies, making it a reliable biomarker of synaptic density(14). Quantification of [¹¹C]UCB-J binding to SV2A thus provides a molecular proxy for estimating synaptic density in the living human brain. While our lab and others have used this molecular imaging approach to identify and characterize synaptic deficits in other neuropsychiatric disorders including major depressive disorder, PTSD (15-17), and schizophrenia(18-20), to our knowledge, no study to date has used SV2A PET to measure synaptic density in BD. However, given that SV2A PET measures only one dimension of synaptic biology, incorporating structural and functional neuroimaging measures through a multimodal imaging approach can significantly help contextualize SV2A PET findings and provide additional, exploratory insights into the neurobiological underpinnings of BD.

Extensive research has documented gray matter (GM) abnormalities in BD, particularly within the fronto-limbic circuitry. Large-scale studies have consistently reported reduced GMV in subcortical structures such as the hippocampus, as well as cortical thinning in prefrontal regions(21, 22). However, findings across studies remain mixed(23-26), which may be partly attributable to heterogeneity in sample demographics, clinical characteristics, and medication status(25, 26). Among pharmacological treatment, lithium, a first-line treatment for BD, has been consistently associated with increased GMV(27-29). In contrast, the structural effects of other medications commonly used to treat BD are less well understood.

Functional connectivity, measured via resting-state functional magnetic resonance imaging (rs-fMRI), has also revealed abnormalities in fronto-limbic regions in BD. However, the direction of effects has varied, with some studies reporting hypoconnectivity and others hyperconnectivity(30-33). One major limitation of rs-fMRI studies in BD is the limited consideration of medication effects. Few studies have included medication-free participants, and many fail to control adequately for medication status, introducing potential confounds into interpretations of functional connectivity alterations(32).

The cerebral cortex exhibits spontaneous fluctuations in neuronal activity that, when synchronized across regions, are reflective of functional connectivity using fMRI(34). Given that GM contains neuronal cell bodies, and synaptic density reflects the number of synaptic connections formed by these neurons, integrating structural (GMV), molecular (SV2A PET), and functional (rs-fMRI) measures can provide a more comprehensive understanding of brain structure–function relationships in BD. This multimodal approach is well-suited to explore the neuropathological processes underlying BD and to examine how psychotropic medications affect brain structure and function. However, few studies to date have leveraged this multimodal approach in BD and even fewer have evaluated how these imaging modalities relate to clinical symptoms or differ as a function of psychotropic medication use.

To address these gaps, we conducted a multimodal neuroimaging study in a cohort of individuals with BD—both medicated and unmedicated—and healthy controls (HCs). We assessed regional synaptic density using SV2A PET, GMV using structural MRI, and intrinsic functional connectivity using rs-fMRI. Our primary aim was to identify core molecular, structural and functional differences within predefined regions of interest (ROIs) between individuals with BD and HCs. Due to the modest sample size, all subsequent analyses were exploratory, including testing for relationships between synaptic density, GMV, and intrinsic connectivity within the same brain regions across the entire sample and as a function of diagnosis. As further exploratory analyses we examined differences in imaging outcome measures between medicated (BD-med), unmedicated (BD-none), and HC groups, and finally, evaluated whether relationships between imaging measures are linked to clinical behavioral outcomes in individuals with BD.

METHODS AND MATERIALS

Participants

Data from a total of 45 individuals were included in the present study, for whom demographic and clinical information is provided in Table 1. Participant recruitment methods included flyers, word of mouth, community outreach, and referrals from local clinics/hospitals. Imaging and data collection took place between November 2014 to August 2024. At screening, diagnosis was confirmed using the Structured Clinical Interview for DSM-5 (35). Exclusion criteria for individuals with bipolar disorder (BD; n =19) were diagnosis of substance use disorder, (except nicotine use disorder) in the past 12 months; positive urine toxicology or pregnancy tests; history of loss of consciousness for more than 5 min; significant medical condition; and contraindications to MRI or PET. Exclusion criteria were the same for the psychiatrically HCs (n = 26) group, with the addition of no current, history of or first-degree family history of any DSM-5 diagnosis, not including nicotine use disorder. Participants underwent physical and neurological examination to exclude presence of active medical or neurological illness. Screening involved electrocardiography, hematology, blood chemistries, urinalysis and urine toxicology screening and plasma pregnancy tests.

Table 1.

Clinical, demographic, and imaging modality-specific features.

Clinical Characteristics Study Sample (N=45)	BD (n = 19)	HC (n = 26)	Statistical Test (p-value)
BD1 BD2 (n : n)	13 : 6	--	--
Depression
MADRS	13.6±9.9	0.65±1.0	< 0.001
HAMD-17	11.4±6.5	0.75±1.2	< 0.001
Mania (YMRS)	2.4±2.7	--	--
Impulsivity (BIS)	71.4±13.2	50.1±6.57.6	< 0.001
Anxiety (STAI)
State	43.1±7.5	28.9±6.2	< 0.001
Trait	49.2±11.7	29.9±6.2	< 0.001
Verbal memory (ISLDR)	10.4±4.2	9.37±2.4	0.38
Working memory (GML)	52.3±28.8	57.8±23.3	0.53
Synaptic Density: [11C]UCB-J PET, VT	BD (n = 17)	HC (n = 26)	Statistical Test (p-value)
Sex (female)	9, 52.9%	12, 46.2%	0.66
Age (years)	39.2±13.2	42.4±16.1	0.51
BMI (kg/m2)	30.3±6.2	26.2±3.8	0.01
Medication (anya)	14, 70.6%	--	--
Injected dose (MBq)	633.7±107.2	584.0±158.0	0.27
Injected mass (ug/kg)	0.02±0.01	0.02±0.01	0.44
Plasma free fraction (fP)	0.27±0.03b	0.28±0.03	0.54
Gray Matter Volumetrics: sMRI, GMV	BD (n = 19)	HC (n = 21)	Statistical Test (p-value)
Sex (female)	9, 47.4%	11, 52.4%	0.75
Age (years)	37.7±13.2	43.7±16.9	0.23
BMI (kg/m2)	30.2±6.2	26.2±3.7	0.02
Medication (anya)	14, 73.7%	--	--
Intercranial volume (mm3)	1.4±0.2	1.4±0.1	0.79
Intrinsic Connectivity: rs-fMRI, ICD	BD (n = 16)	HC (n = 19)	Statistical Test (p-value)
Sex (female)	9, 56.3%	10, 52.6%	0.83
Age (years)	36.6±13.8	44.0±17.4	0.32
BMI (kg/m2)	29.6±6.1	26.3±3.9	0.06
Medication (anya)	14, 73.7%	--	--

Note: data are provided as mean±SD or as number (n) and percent (%)

^aDetails of medications use by participants in the BD group is provided in Table 2 and Table S1.

^bf_p in BD-med (0.27±0.03) and BD-none (0.28±0.3) were not statistically different (t₁₄=0.51, p=0.62)

The Yale University Human Investigation Committee and the Radioactive Drug Research Committee approved the study. All participants provided written informed consent prior to study enrolment and participation.

Clinical and cognitive assessments

The Young Mania Rating Scale (YMRS)(36) was administered by a trained rater to assess manic symptoms in individuals with BD. Depressive symptoms were assessed in all participants using the Montgomery-Åsberg Depression Rating Scale (MADRS)(37) and Hamilton Depression Rating Scale (HAMD-17)(38). The State-Trait Anxiety Inventory (STAI) was used to measure state (STAI-S) and trait (STAI-T) anxiety, respectively(39).

Impulsivity and cognitive deficits are often associated with BD(40, 41). Therefore, impulsivity was assessed using the Barratt Impulsiveness Scale (BIS)(42). The Urgency-Premeditation-Perseverance-Sensation Seeking-Positive Urgency (UPPSP) Impulsive Behavior Scale was used to additionally assess the five dimensions of impulsivity(43) in BD. All participants completed a brief computerized cognitive testing battery (Cogstate: https://Cogstate.com/computerized-tests), including of the International Shopping List-Delayed Recall (ISLDR), a test of verbal short-term memory, and the Groton Maze Learning (GML) test that measures executive control.

Brain imaging

Regions of interest (ROIs)

A priori ROIs for PET and MRI were based on previous neuroimaging findings in BD and their role in emotional and cognitive processing(17, 23, 28, 44-47). Specifically, corticolimbic regions of anterior cingulate cortex (ACC), dorsolateral prefrontal cortex (dlPFC), orbitofrontal cortex (OFC), ventromedial prefrontal cortex (vmPFC), hippocampus, and amygdala (exploratory, see Supplement). ROIs were delineated using the Anatomical Automatic Labeling (AAL) atlas(48).

PET imaging

Details of PET imaging methodology have been previously reported(15) and are available in the Supplement. Briefly, [¹¹C]UCB-J was synthesized onsite and administered intravenously as a bolus. Subjects were scanned on a High-Resolution Research Tomograph (CTI/Siemens). The injected radioactivity was within a range that produced good image quality (mean: 602.91±141.47 MBq) and an injected mass dose (mean: 0.021±0.012 μg/kg) expected to produce less than 1% occupancy (i.e., “tracer” level)(49).

PET images were coregistered to each participant’s MRI and then coregistered to the MRI template. Given the lack of a true reference region for SV2A targets(50), volume of distribution (V_T) was used as the outcome measure, which was computed using the one tissue compartment model with the metabolite-corrected arterial input function, as validated previously (51, 52) and detailed in the Supplement. PET data could not be processed for two BD participants for whom arterial blood sampling failed (Table 1).

MR Imaging

High-resolution structural magnetic resonance imaging (sMRI) data as well as resting-state functional MRI data were collected on 3T Siemens Prisma scanners (see full details of MR imaging processing in Supplement).

Statistical Analyses

All statistical analyses were performed using SPSS Statistics v22 (IBM). Data were checked for homoscedasticity, normality, and outliers (Grubs p < 0.01). For the SV2A/GMV ratio, imaging outcome measures were t-score transformed to have SV2A and GMV be in the same positive scale.

Multivariate analysis of variance (MANOVA) was used to assess group differences in the primary imaging outcome measure, including age, sex, BMI as covariates, and intercranial volume as an additional covariate in the GMV analysis. Post-hoc univariate tests were Bonferroni-corrected for multiple comparisons with a significance threshold of p < 0.05. Relationships between measures were examined using Pearson’s correlations (two-tailed); given the exploratory nature, significance was set at uncorrected p < 0.05, with p < 0.1 considered trend-level significant. Finally, post-hoc power analyses were performed and provided as part of the Supplement.

RESULTS

Group differences in synaptic density as measured by SV2A PET

The MANOVA testing for differences in fronto-limbic SV2A density (Figure 1A) revealed a main effect of diagnosis (F_5,34=3.17, p=0.019). Specifically, synaptic density was significantly lower in BD relative to HC in dlPFC (−13.4%, p_adj=0.046), vmPFC (−13.2%, p_adj=0.044), and hippocampus (−12.3%, p_adj=0.020).

Figure 1.

Multivariate Analysis of Variance (MANOVA) analyses testing for differences in A,B] SV2A PET measures of synaptic density (i.e., [¹¹C]UCB-J volume of distribution, V_T), C,D] Magnetization-Prepared Rapid Gradient Echo (MPRAGE) MRI measurement of gray matter volumes (GMV), and E,F] functional connectivity as measured by the Intrinsic Connectivity Distribution (ICD). A,C,E] Differences between the Healthy Comparison (HC) group and all individuals with Bipolar Disorder (BD) are visualized, along with B,D,F] differences within the BD group as a function of individuals with no current psychotropic medication use (BD-none) versus those currently taking any medication (BD-med). Individual datapoints are shown with the bar at the group mean ± standard deviation. Significant Bonferroni-corrected pairwise comparisons (p_adj) are denoted as *< 0.05, ***< 0.001 relative to HC; * < 0.05, **< 0.01 relative to BD-none. MANOVA tables are provided as in the Supplemental Results.

To explore the potential effect of psychotropic medications on SV2A density, the BD group was subdivided based on those who were unmedicated for at least one year (6.3±9.1 years) prior to study participation (BD-none) and individuals who reported taking any psychotropic medication (BD-med). A full list of medications used by BD participants is provided in Table 2. The secondary MANOVA testing the impact of psychotropic medication status– that is, HC vs. BD-none (n=5) vs. BD-med (n=12)– on SV2A density was statistically significant (F_10,64=2 .08, p_adj=0.039; Figure 1B), with significant group effects observed in dlPFC (p_adj=0.05), vmPFC (p_adj=0.025), and hippocampus (p=0.012). Post-hoc analyses revealed the overall group differences were driven by the BD-med group, with significantly lower SV2A density in the dlPFC (−17.3%, p_adj=0.047), vmPFC, (−17.8%, p_adj=0.025), and hippocampus (−16.3%, p_adj=0.011) relative to HC. The BD-med group also had lower SV2A density relative to BD-none, but these differences did not achieve statistical significance (with p_adj ranging from 0.46 in dlPFC to 0.19 in hippocampus), nor were there significant differences between BD-none and HC (all p_adj>0.99).

Table 2.

Psychotropic medication-use among participants with bipolar disorder (n = 19).

Medication Count	n	%
Unmedicated	5	26.3
One	4	21.1
Two	4	21.1
Three	2	10.5
Four	3	15.8
Five	1	5.3
Medication Class	n	%
Antidepressant (SSRI, SNRI, NDRI)	8	42.1
Atypical Antipsychotic	5	26.3
Benzodiazepine	2	10.5
Mood Stabilizer/Anxiolytic/Anticonvulsant	11	57.9
Stimulant	1	5.3

Group differences in regional gray matter volume (GMV) as measured by structural MRI

Among participants who completed PET imaging , structural MRI (sMRI)-based volumetric data was available for forty individuals (Table 1). The MANOVA testing for group differences in regional GMV between HC (n=21) and BD (n=19) was not statistically significant (F_6,29=1.07, p=0.40; Figure 1C), with no significant differences in GMV for any of the fronto-limbic ROIs.

The secondary MANOVA comparing GMV in HC relative to BD-none (n=5) and BD-med (n=14) groups to explore the effect of psychotropic medication status showed a trend-level group effect of higher GMV in BD group as compared with HC (F_10,58=1.88, p = 0.068; Figure 1D). Despite the nonsignificant multivariate result, univariate analyses revealed statistically significant group differences in GMV for the ACC (p_adj=0.004), dlPFC (p_adj=0.03), OFC (p_adj=0.017), and vmPFC (p_adj=0.014), with a trend for hippocampal GMV (p_adj=0.077). Post-hoc analyses indicated that GMV was highest in the BD-med group, including being significantly higher than that observed in the BD-none group in ACC (p_adj=0.005), dlPFC (p_adj=0.026), OFC (p_adj=0.014), and vmPFC (p_adj=0.020). GMV in the BD-med group was also higher than that observed in HC adults across all ROIs. After correcting for multiple comparisons, this difference was marginally significant only in the ACC (p_adj=0.050).

Group differences in intrinsic connectivity as measured by resting-state fMRI

A MANOVA testing for group differences in intrinsic connectivity distribution (ICD), representing connectivity of the corticolimbic ROIs to the rest of the brain, was statistically significant (F_5,26=66.93, p<0.001; Figure 1E). Specifically, ICD from ACC (p_adj=0.020), dlPFC (p_adj< 0.001), vmPFC (p_adj=0.033) and hippocampus (p_adj<0.001) to the whole brain was significantly higher in the BD group (n=16) relative to HC (n=19), with a trend-level difference observed in the OFC (p_adj=0.081). Removal of identified outlier values from the BD group did not significantly impact these findings (see Supplement).

The exploratory MANOVA assessing the effect of psychotropic medication status by comparing connectivity in HC relative to BD-none (n=4) and BD-med (n=12) was also significant (F_5,34=18.39, p<0.001; Figure 1F), with group effects observed in intrinsic connectivity from the dlPFC (p_adj<0.001), ACC (p_adj=0.009), vmPFC (p_adj=0.009), and hippocampus (p_adj<0.001), and a trend-level effect in the OFC (p_adj=0.057). Post-hoc analyses revealed that connectivity was higher in the BD-none group as compared with HC in the dlPFC (p_adj<0.001). Additionally, the BD-med group displayed significantly lower intrinsic connectivity compared to the BD-none group in the dlPFC (p_adj=0.002), ACC (p_adj=0.029), vmPFC (p_adj=0.029), and hippocampus (p_adj=0.021). The BD-med group also showed significantly higher intrinsic connectivity than the HC group in the dlPFC (p_adj<0.001) and hippocampus (p_adj=0.001).

Correlations between neuroimaging imaging modalities

To understand relationships between molecular (SV2A PET), structural (sMRI GMV), and functional imaging (rs-fMRI ICD) derived measures, we performed exploratory correlation analyses for each pair of outcome measures. Across individuals with both PET and sMRI (HC, n=21; BD, n=17), no significant relationships were observed between SV2A density and GMV (all r≤0.232, p≥0.162). However, when limiting the analysis to just the HC group, there were significant positive correlations between SV2A density and GMV across almost all pairs of fronto-limbic ROIs (Figure 2, left), with statistically significant positive correlations ranging from r=0.617, p=0.003 (ACC SV2A relative to ACC GMV) to r=0.440, p=0.046 (hippocampus SV2A relative to OFC GMV). Conversely, no significant relationships were observed in the BD group (Figure 2, middle) and remained non-significant when limiting the analysis to just the BD-med group (n=12; Figure 2, right), that is, excluding the five BD-none individuals. Overall, SV2A-GMV correlations (r) were significantly greater in HC relative to BD (see Supplement).

Figure 2.

Heatmap visualizing relationships between regional measures of SV2A density and GMV in the 21 HC adults (left), all 17 BD participants with both imaging measures (center), and specifically the 12 BD participants with any current psychotropic medication use (right), excluding the 5 unmedicated BD participants. The cell color indicates the direction and magnitude of the correlation (i.e., Pearson’s r). Exact p-values are displayed for non-significant correlations, while significant correlations are denoted as *p<0.05, **p<0.01.

When testing for associations between SV2A density and intrinsic connectivity for the entire study sample with both imaging modalities (N=34), higher dlPFC intrinsic connectivity was found to be associated with lower SV2A density, ranging from r=−0.398, p=0.020 relative to vmPFC SV2A density, to r=−0.332, p = 0.055 relative to OFC SV2A density. However, these correlations were not significant when the analysis was limited to just the HC group (n=19), the total BD group (n=15), or the BD-med sub-group (n=12). Further, no other regional ICD measures were significantly correlated with SV2A density or GMV in the total study sample or any of the groups/subgroups (data not shown).

Exploratory Correlations between SV2A/GMV Ratios and Clinical and Cognitive Symptoms in BD

Most measures of clinical and cognitive symptoms were less severe in the BD-med group relative to BD-none (see more details in Supplement). To explore the possible clinical relevance of lack of correlation between SV2A density and GMV observed in the individuals with BD, we tested for relationships between regional SV2A/GMV ratios and clinical and cognitive measures. No associations were found between mood or anxiety symptom scales (data not shown). However, we observed significant correlations, with lower SV2A/GMV ratios (i.e., lower SV2A density per unit of GMV) associated with greater Negative Urgency measured by the UPPSP Impulsive Behavior Scale, across all frontal ROIs (ACC: r=−0.630, p=0.012; Figure 3A; dlPFC: r=−0.604, p=0.017; OFC: r=−0.611, p=0.016, and vmPFC: r=−0.558, p=0.031). SV2A/GMV ratios in frontal ROIs were additionally negatively correlated with Lack of Premeditation scores on this scale (ACC: r=−0.759, p=0.001; Figure 3B; dlPFC: r=−0.758, p=0.001; OFC: r=−0.695, p=0.004, and vmPFC: r=−0.691, p=0.004), and hippocampus (r=−0.541, p=0.004, and vmPFC: r=−0.691, p=0.004), and hippocampus (r=−0.541, p=0.037). Lower ACC SV2A/GMV ratios were also marginally negatively associated with performance on the GML Test, a measure of executive function/visual working memory(r=−0.483, p=0.068; Figure 3C). Notably, none of the aforementioned correlations remained significant after limiting the analysis to the BD-med subgroup.

Figure 3.

Correlations between the ACC SV2A/GMV ratio and dimensions of impulsivity as measured by the UPPS scale, specifically, A] Negative Urgency and B] (Lack of) Premeditation. C] The correlation between the ACC SV2A/GMV ratio and working memory difficulties as measured by errors made during the Groton Maze Learning Task (GMLT). Normalized values (t-scores) of SV2A density and GMV were used to calculate regional SV2A/GMV ratios such that a ratio < 1 (highlighted) indicates lower SV2A relative to GMV. Correlations in the full BD group are represented by solid the solid lines and correlations limited to the BD-med subgroup are represented by dashed lines.

DISCUSSION

Using multimodal neuroimaging, we identified molecular (SV2A PET), structural (GMV), and functional (ICD) synaptic alterations in individuals with BD compared with HCs. Exploratory analyses further provided preliminary evidence suggesting that psychotropic medication may modulate neuroimaging signatures. To our knowledge, this is the first study to apply SV2A PET imaging in BD and to report synaptic deficits using this molecular imaging approach.

Specifically, we observed lower SV2A binding in fronto-limbic regions in individuals with BD, suggesting lower synaptic density. This in in vivo finding complements prior postmortem brain evidence(7, 8) (9, 10) of synaptic deficits in these regions and supports the hypothesis that fronto-limbic synaptic dysfunction is a possible pathophysiological hallmark of BD. However, this interpretation is complicated by the observation that SV2A density was lowest among medicated individuals with BD, raising the possibility that this reduction reflects a medication effect. This finding is somewhat unexpected, as preclinical studies have shown that commonly prescribed mood stabilizers and atypical antipsychotics, such as lithium, lamotrigine, and aripiprazole (the top three medications used in the BD-med group, at n=7, n=4, and n=4, respectively), have neuroprotective and synaptogenic properties in both in vitro(53-55) and in vivo(56-58) models. A possible explanation is that the medicated group represents individuals with more severe illness; however, BD-med participants presented with less severe symptoms. Thus, longitudinal studies with more detailed information regarding illness duration and history are needed to clarify these preliminary findings and to determine their clinical relevance.

Although no overall group differences in GMV were observed, secondary analyses revealed significantly greater GMV in fronto-limbic regions—including the ACC, dlPFC, OFC, and vmPFC—in the BD-med group compared to both the BD-none and HC groups. This is consistent with prior studies showing increased GMV associated with lithium treatment(27-29). Notably, individuals in the BD-med group in the current study were on various psychotropic medications, suggesting that GMV increases may reflect broader medication effects not limited to lithium. This interpretation would also be consistent with the preclinical findings showing neuroprotective and pro-synaptogenic effects of a wider variety of medication classes(53-58).

Functionally, we found that intrinsic connectivity in fronto-limbic regions was altered in BD, with the BD-none group showing increased connectivity—particularly in the dlPFC—relative to both the medicated BD and HC groups. Although prior studies of functional connectivity in BD have reported mixed findings, some have similarly described hyperconnectivity in the fronto-limbic circuitry underlying emotion regulation networks in BD(31). Notably, many of these studies did not control for psychotropic medication status. Our use of ICD, which considers the full range of connection strengths rather than applying an arbitrary correlation threshold(59) may also partly explain our ability to detect more nuanced connectivity differences. Taken together, these findings suggest that elevated fronto-limbic connectivity may be a feature of unmedicated BD, whereas psychotropic medications may attenuate such hyperconnectivity in key regulatory regions.

Each neuroimaging modality provides distinct but complementary insight into synaptic biology, including molecular, brain structural, and functional measures. In the current study, we had the unique opportunity to examine correlations between these imaging measures and evaluate whether these relationships are perturbed in BD. In an exploratory analysis, we observe significant positive correlations between synaptic density and GMV in the fronto-limbic circuitry in HC. However, no such relationships were observed in BD, regardless of medication status. Given that synaptic density is a key component of GM —and could reflect the richness and complexity of neural connections — these findings suggest a disruption in this relationship as part of BD neuropathology that may not be restored by medication. Furthermore, higher dlPFC intrinsic connectivity was modestly associated with lower SV2A density across the full sample, indicating a potential decoupling between synaptic structure and function in BD, independent of medication use.

Clinically, BD-med individuals showed significantly lower depressive, manic, and anxiety symptoms compared to their unmedicated counterparts, with trend-level group differences observed on several clinical measures. Although no significant group differences were found in cognitive performance or behavioral (impulsivity) measures, the BD-med group consistently showed numerically lower scores, suggesting possible benefits of psychotropic treatment on BD-related symptoms and behavior(6, 60).

Lower SV2A/GMV ratios in frontal-limbic regions were associated with greater impulsivity—specifically, higher levels of negative urgency and lack of premeditation—and showed a marginal association with poorer executive function in BD. This finding was not limited to the unmediated subgroup, suggesting that disrupted synaptic-structural integrity, or more specifically, the degree of dissociation between SV2A density and GMV, may contribute to impulsivity in BD. While these findings are preliminary and should be interpreted with caution, lower SV2A/GMV ratios could serve as a neurobiological marker of impulsivity and related cognitive deficits in BD, and further, may help identify individuals at higher risk for impulsive behaviors (e.g., risky decision-making, suicidality, substance misuse) independent of medication status. Notably, within each modality, we observed no correlations with any mood symptoms, cognitive measures, or impulsivity, which may be a reflection of divergent medication effects across modalities (i.e., lower SV2A density and intrinsic connectivity, but higher GMV with medication), which may not directly align with symptom relief mechanisms.

This study has some limitations. First, the modest sample size limits statistical power for MRIbased analyses, especially for subgroup comparisons. While an N of 45 is not atypical for PET studies, we recognize this is quite modest for MR-based analyses. In particular, our sample of unmedicated individuals was particularly small. While we applied rigorous correction to group-wise post-hoc tests, sub-group analyses were underpowered (see Supplement) and considered to be exploratory. Further, correlation analyses lack multiple comparison correction, thus findings should be interpreted with caution due to increased risk of Type I error. In addition, the present sample size and instances of polypharmacy prohibited any meaningful investigation of specific medication classes and their effects imaging outcome measures. Overall, future studies with larger samples are critical for testing how diagnostic group or medication status moderate relationships between molecular, structural, and functional imaging measures, and application of formal moderation models may help clarify the mechanisms through which psychotropic medications influence brain circuitry in BD. Second, we used volume of distribution (V_T) as the PET outcome measure, as in our previous work(15, 17). Binding potential (BP_ND) is often the preferred outcome measure but must be calculated using a reference region with negligible specific binding. Some groups will use centrum semiovale (CS) white matter, as a pseudo-reference region for the calculation of BP_ND with radioligand [¹¹C]UCB-J. However, given substantial evidence of white matter pathology in BD(45, 61, 62), in addition to observed differences in CS V_T values (Figure S2: HC vs BD p = 0.06; HC vs BD-med p = 0.036), CS was determined to be invalid as a pseudo-reference region for the present study. Lastly, the cross-sectional study design restricts causal inference, underscoring the importance of future longitudinal studies to clarify the directionality of relationships among synaptic measures, medication effects, and clinical outcomes.

Notwithstanding these limitations, a major strength is the integration of three neuroimaging modalities – SV2A PET, structural MRI, and resting-state fMRI – within the same cohort, enabling a more holistic assessment of brain alterations, as well as the effects of psychotropic medication in BD. Inclusion of both medicated and unmedicated individuals with BD enhances ecological validity and provided critical insights into how psychotropic medications may influence synaptic measures and clinical symptoms.

In summary, this multimodal neuroimaging study provides converging evidence of structural and functional brain abnormalities in BD and offers the first in vivo evidence of synaptic deficits in the disorder. While psychotropic medications were associated with increased GMV and reduced clinical symptom severity, they were also linked to lower synaptic density, suggesting complex and divergent effects on brain structure and function. The absence of a positive SV2A-GMV correlation in BD and the association of lower SV2A/GMV ratios with greater impulsivity highlight potential disruptions in synaptic-structural coupling in BD. In addition, while psychotropic medications may support improvements in brain function and clinical symptoms, they may not fully restore underlying structural deficits. These findings underscore the importance of accounting for medication effects in neuroimaging findings in BD and highlight the need for longitudinal studies to disentangle illness- from treatment-related changes in brain structure and function in BD.

Supplementary Material

Supplement Methods, Results, Figures S1-S2, Tables S1-S4

KEY RESOURCES TABLE

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab.	catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs.	Include any additional information or notes if necessary.
Chemical Compound, radiotracer	(4R)-1-{[3-(11C)Methylpyridin-4-yl]methyl}-4-(3,4,5-trifluorophenyl)pyrrolidin-2-one; [11C]UCB-J	In-house radiosynthesis	PubChem CID: 74539761
Software	Interactive Data Language (IDL) v8.6	NV5 Geospatial Solutions, Inc		https://www.nv5geospatialsoftware.com/Products/IDL
Software	FSL v6.0.7.18	Analysis Group, FMRIB, Oxford, UK	RRID:SCR_002823	https://fsl.fmrib.ox.ac.uk/fsl/docs/install/index.html
Software	Bioimage Suite v1.2.0	Neuroimageing Tools & Resources Collaboratory	RRID:SCR_002986	https://bioimagesuiteweb.github.io/webapp/
Software	IBM SPSS Statistics v31.0.0.0	IBM Corporation	RRID:SCR_016479	https://www.ibm.com/products/spss-statistics