1. Introduction

Bipolar disorder (BD) is characterized by the presence of manic and depressive episodes in patients [1] and contributes significantly to disability among the working population, ranking 20th among all diseases [2]. It is a disorder with high genetic predisposition and heritability [3, 4], with clinical manifestations typically debuting at a young age [5]. Notably, the incidence rate in the 20–24 age group has gradually increased over the past 30 years, from 51.76 to 58.37 per 100,000 [6]. In the diagnostic paradigm of the prevailing categorical approach, the misdiagnosis rate for BD remains relatively high (ranging from 40% to 70%) [7, 8]. Differential diagnosis is often complicated by insufficient attention from clinicians and patients to hypomanic states, the higher presentation rates of depressive phases, or the misinterpretation of manic symptoms as psychotic features. Additionally, frequent comorbidity with other disorders [9] exacerbates these challenges. In at least one-third of cases, it takes 10–15 years from the initial manifestations to the diagnosis of BD [10].

Some contemporary scientific hypotheses propose a continuum [11], wherein BD and major depressive disorder manifest across three key domains of human functioning: activity, cognition, and emotion [12]. According to these proposed models, diagnostic difficulties may be mitigated by focusing on the severity of dysfunction within each domain [13]. In addition, there is growing evidence for the importance of transnosological symptoms in patients with affective disorders, such as anhedonia and impulsivity, as well as predisposing factors, which include specific personality traits [14–17].

Of particular importance in responding to external stressors is an individual’s temperament, whose relationship with various aspects of health and mental disorders has been repeatedly demonstrated in population studies [18, 19]. Threre is also some evidence that altered temperament traits are associated with affective disorders [19–22]. Moreover, some of these changes are also found in relatives of patients with affective disorders [23, 24]. Numerous theoretical models of temperament traits have been developed, most of which emphasize a predisposition to developing affective-spectrum disorders [23]. One of the most well known and extensively studied models over the past 35 years is R. Cloninger’s model [25]. This model is based on a biopsychosocial approach, linking personality traits to genetic predisposition, behavior, and neurotransmitter system functioning [26, 27].

At the same time, the search for biomarkers of affective disorders, particularly neuroimaging biomarkers, could facilitate addressing diagnostic challenges [28]. Systematizing the existing data from original studies suggests the identification of the most involved brain regions in the pathogenesis of affective disorders [29–32]. These findings are gradually being reflected in studies aimed at refining our understanding of pathogenesis [33], improving differential diagnosis [34, 35], identifying high-risk groups prior to disorder onset [36, 37], and determining the most appropriate psychopharmacotherapy [38, 39] using advanced magnetic resonance imaging (MRI) techniques.

Furthermore, neuroimaging has been actively employed to objectify temperamental traits, with the relative stability of neural network activity demonstrating a trait-like nature [40]. A large meta-analysis, encompassing data from over 3,000 respondents, revealed structural (in the parietal and temporal lobes) and functional (in the parietal, frontal, and temporal lobes) changes associated with varying degrees of Harm Avoidance in healthy individuals [41]. A similar original study conducted on 360 students enabled the development of a model to predict temperamental traits based on functional connectivity patterns in frontal-subcortical circuits [42]. Emerging studies with comparable designs are now being conducted on patients with mental disorders. For instance, an evaluation of changes in functional connectivity across different brain regions in stable outpatients with psychoses (including BD and schizophrenia) revealed correlations with specific personality features such as sensory hypersensitivity, negative emotional balance, impaired attentional control, avolition, and social mistrust [43]. Interestingly, character and temperament traits in patients compared to control groups accounted for the majority of differences in functional connectivity.

The aim of this study was to compare temperamental traits in patients with BD and healthy controls (HC) and potentially identify functional connectivity changes in large-scale brain networks associated with these differences. We hypothesized that patients with BD would exhibit distinct temperamental traits compared to healthy controls and that these differences, in turn, would be linked to alterations in the functional connectivity of key large-scale brain networks, specifically those most commonly affected in BD: the Default Mode Network (DMN), Salience Network (SN), FrontoParietal Network (FPN), and Sensorimotor Network.

2. Materials and Methods

This study was conducted using a shared neuroimaging dataset from the UCLA Consortium for Neuropsychiatric Phenomics, which included imaging and clinical data for 49 BD patients in various phases and 130 healthy controls (HC) (from which we selected 49 respondents matched for sex and age with the study group) [44]. According to the data listed in the dataset, the established diagnosis with last episode suggested patients with BD in the following states: 22 in full/partial remission (of which with last manic/hypomanic—9, depressive—13), 18 with current mild/moderate illness (of which with manic/hypomanic—6, mixed—6, depressive—6) and 9 in current severe episode (of which with manic/hypomanic—1, mixed—3, and depressive—5). Demographic: sex, age; psychometric: Hamilton Depression Rating Scale (HAMD) for depressive symptoms severity, Young Mania Rating Scale (YMRS) for manic symptoms severity, The Temperament and Character Inventory (TCI-125) for temperament traits, Barrat Impulsivity Scale (BIS11) for impulsivity severity; and structural and resting-state functional MRI data were extracted for all 98 respondents.

Neuroimaging data were acquired using a 3T Siemens Trio scanner. Functional MRI data were obtained with a T2 ^∗-weighted echoplanar imaging (EPI) sequence using the following parameters: slice thickness = 4 mm, 34 slices, TR = 2 s, TE = 30 ms, flip angle = 90°, matrix = 64 × 64, field of view (FOV) = 192 mm. High-resolution T1-weighted anatomical scans (MPRAGE) were acquired with parameters: slice thickness = 1 mm, 176 slices, TR = 1.9 s, TE = 2.26 ms, matrix = 256 × 256, FOV = 250 mm. Diffusion-weighted imaging data were obtained with the following parameters: slice thickness = 2 mm, 64 diffusion directions, TR/TE = 9000/93 ms, flip angle = 90°, matrix = 96 × 96, axial slices, and b-value = 1000 s/mm². Further details on MRI acquisition parameters and preprocessing procedures are described in dataset article [45].

Neuroimaging data analysis was conducted using the CONN functional connectivity toolbox version 22.407 [46]. Preprocessing of native images included the following steps: realignment, susceptibility distortion correction, slice timing correction, co-registration, and normalization, followed by functional smoothing. After preprocessing, the compiled images underwent two stages of denoizing. The first stage involved linear regression of potential noise signals in the global BOLD contrast using the “aCompCor” method. The second stage involved applying a temporal frequency filter, removing frequencies below 0.008 Hz or above 0.09 Hz to retain only low-frequency oscillations, thereby minimizing the effects of head motion and other noise sources.

Seed-based connectivity maps were generated to characterize patterns of functional connectivity with 164 «HPC-ICA networks and Harvard-Oxford atlas ROIs» [47]. Functional connectivity strength was quantified using Fisher-transformed bivariate correlation coefficients derived from a weighted general linear model [46]. These coefficients were calculated for each seed-target pair, modeling associations between their respective blood oxygenation level dependence signal time series. To mitigate transient magnetization effects at the start of each scan, individual runs were weighted using a step function convolved with a statistical parametric mapping canonical hemodynamic response function, followed by rectification.

Group-level analyses employed a general linear model. For each voxel, a separate general linear model was estimated, with first-level connectivity measures as dependent variables and group or subject-level identifiers as independent variables. Hypotheses were evaluated using multivariate parametric statistics with random-effects across subjects and sample covariance estimation across measurements. Statistical inferences were performed at the cluster level, examining contiguous voxel groups. Cluster-level significance was assessed using parametric statistics based on Gaussian random field theory, with thresholds set at p  < 0.001 for voxel-level significance and p-FWE/p-FDR < 0.05 for familywise error- or false discovery rate-corrected cluster sizes [48].

To evaluate differences in functional connectivity between the control and study groups, based on temperament traits, a one-way ANCOVA with covariate interaction was employed, enabling regression comparison between groups. Seeds included key regions of interest from four large-scale brain networks: DMN: medial prefrontal cortex, bilateral lateral parietal cortex, posterior cingulate cortex; SN: anterior cingulate cortex, bilateral anterior insular cortex, bilateral supramarginal gyrus; FPN: bilateral lateral prefrontal cortex, bilateral posterior parietal cortex; SomatoSensory Network: regions involved in somatosensory processing.

Psychometric data were analyzed using Jamovi software [49]. Descriptive statistics, normality testing (Shapiro–Wilk test), and homogeneity of variance testing (Levene’s test) were conducted. Parametric data were analyzed using Student’s t-test (Welch’s t-test for unequal variances), one-way ANOVA (Fisher’s F-test; Welch’s F-test for unequal variances), and post hoc comparisons via Tukey’s test (Games-Howell for unequal variances). Correlations were assessed using Pearson’s correlation coefficient. Nonparametric data were analyzed with Mann–Whitney U test, Kruskal–Wallis one-way ANOVA, and pairwise comparisons using Dwass-Steel–Critchlow–Fligner tests. Results are presented as measures of central tendency: Mean and standard deviation M(SD) for parametric data; median and interquartile range Me[IQR] for non-parametric data; and percentage (%) or absolute frequency (n) for categorical variables.

3. Results

The comparative analysis of temperamental features in BD and HC revealed significant differences in all four traits. Thus, patients with BD demonstrated higher levels of Harm Avoidance (t-Welch’s (82.8) = 4.85; p < 0.001) and Novelty Seeking (t-Welch’s (87.3) = 4.37; p < 0.001). Conversely, Reward Dependence (t-Welch’s (87.1) = −2.50; p = 0.014) and Persistence (U = 998; p = 0.002) were significantly lower compared to the temperamental characteristics of HC (Table 1).

Higher Persistence in BD was associated with greater severity of manic symptoms (Spearman’s rho = 0.302, p = 0.018), while lower Reward Dependence was associated with greater severity of depressive symptoms (Pearson’s r = −0.388, p = 0.003). When assessing the internal correlation between temperament traits in patients with BD, we found that Harm Avoidance negatively correlates with Persistence (Pearson’s r = −0.525, p < 0.001) and positively with Reward Dependence (Pearson’s r = 0.259, p = 0.036) (Figure 1).

3.1. Neuroimaging

The lower Reward Dependence in patients with BD, compared to the HC, was associated with increased connectivity of the anterior cingulate cortex (SN) with a cluster of structures in the left primary visual cortex (147 voxels with a center at −42, −60, +48; p-FDR < 0.05). A larger number of voxels with altered functional connectivity were located in the DMN and FPN (Figure 2).

An increased level of the Novelty Seeking in BD patients compared to the HC was also associated with changes in the SN. Specifically, it was linked to reduced connectivity of the anterior cingulate cortex with a cluster of structures in the left primary visual cortex (234 voxels centered at −26, −86, −18; p-FWE < 0.01) and reduced functional connectivity between the left anterior insular cortex and the right lateral occipital cortex (260 voxels centered at +18, −64, +54; p-FWE < 0.01). In contrast, increased connectivity was observed between the right anterior insular cortex and the bilateral paracingulate gyrus (172 voxels centered at +2, +24, +36; p-FWE = 0.02) and a cluster in the frontal pole and middle frontal gyrus (205 voxels centered at +34, +30, +30; p-FWE < 0.01). Functional connectivity of the left anterior insular cortex was elevated with a cluster in the superior frontal and paracingulate gyri bilaterally (310 voxels centered at 0, +28, +48; p-FWE < 0.01) and with the left middle frontal gyrus (203 voxels centered at −34, +24, +28; p-FWE < 0.01). At the inter-network level, Novelty Seeking was characterized by increased functional connectivity between SN and FPN and decreased functional connectivity between SN and Visual Network as well as Dorsal Attention Network (Figure 3).

A decreased level of the temperament trait Harm Avoidance in BD patients compared to the HC was associated with enhanced connectivity of the posterior parietal cortex (FPN) with a cluster of subcortical structures, including the putamen, globus pallidus, and amygdala on the left side (287 voxels centered at −26, +2, −2; p-FWE < 0.01) (Figure 4).

Finally, a lower Persistence in BD patients compared to the HC was also associated with changes in the FPN. This included a decreased functional connectivity of the left posterior parietal cortex with a cluster in the frontal pole and superior frontal gyrus on the right side (214 voxels centered at +26, −40, +34; p-FWE < 0.01); reduced functional connectivity between the left lateral prefrontal cortex and the right orbitofrontal cortex (118 voxels centered at −32, +24, −20; p-FDR = 0.03), as well as with a cluster in the frontal pole and anterior cingulate cortex on the right side (234 voxels centered at +6, +54, +18; p-FWE = 0.02). At the inter-network level, Persistence was primarily characterized by decreased functional connectivity between FPN and DMN (Figure 5).

4. Discussion

The results revealed a specific spectrum of temperament trait alterations of patients with BD compared to HC: lower scores on the dimensions of Reward Dependence and Persistence, and higher scores on Novelty Seeking and Harm Avoidance. Neuroimaging data analysis identified alterations in functional connectivity within the SN and FPN, as well as in their interactions with the DMN, Dorsal Attention Network, Visual Network, and basal nuclei, associated with these temperament traits in BD patients compared to HC. The primary role of the SN is its involvement in attentional control through the detection of subjectively salient events and the provision of control signals to the FPN for goal-directed actions or to the DMN for self-referential processes, social cognition, episodic and autobiographical memory, language and semantic memory, and mind-wandering [50–52]. Of particular interest, warranting further investigation, is the striking similarity of the temperament trait alterations observed in our study with those reported in a French cohort of 570 patients with Parkinson’s disease [53].

5. Limitations

This study has several limitations. First of all, the study’s cross-sectional design precludes causal inferences regarding the relationship between temperamental traits and functional connectivity alterations. Longitudinal studies are needed to explore the temporal dynamics of these associations and their potential role in the progression of BD. As well as the sample size may limit the generalizability of the findings to broader BD populations. A larger and more diverse sample, including participants from different age groups, cultural backgrounds, and phases of the disorder, could enhance the robustness of the results. Temperamental traits and impulsivity were evaluated using self-report measures, which may be subject to response biases, such as social desirability or recall bias. We did not control the effect of psychotropic medications and comorbidities which could also influence brain connectivity. Despite these limitations, the findings provide important insights into the neural correlates of temperamental traits in BD and lay the groundwork for future research.

6. Conclusions

We identified some specific temperament traits related in the pathophysiological processes of BD. It reflects in functional connectivity alterations within the SN and FPN and their interactions with the DMN, Dorsal Attention Network, Visual Network, and basal nuclei, closely linked to clinical symptoms severity. The association between increased impulsivity and temperament traits, which are critical for its development and expression, underscores the importance of impulsivity in the structure of BD. These findings highlight the need for assessing crossdiagnostic phenomena to improve diagnostic accuracy.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding

The author(s) declare this research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.