ORIGINAL ARTICLE

Open Access

Neural substrates of predicting anhedonia symptoms in major depressive disorder via connectome-based modeling

Tingyu Yang

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Department of Child Healthcare, Hunan Children's Hospital, Changsha, China

Search for more papers by this author

Yangpan Ou,

Yangpan Ou

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Huabing Li,

Huabing Li

Department of Radiology, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Feng Liu,

Feng Liu

orcid.org/0000-0002-3570-4222

Department of Radiology, Tianjin Medical University General Hospital, Tianjin, China

Search for more papers by this author

Ping Li,

Ping Li

Department of Psychiatry, Qiqihar Medical University, Qiqihar, China

Search for more papers by this author

Guangrong Xie,

Guangrong Xie

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Jingping Zhao,

Jingping Zhao

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Xilong Cui,

Corresponding Author

Xilong Cui

[email protected]

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Correspondence

Wenbin Guo and Xilong Cui, Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, Hunan 410011, China.

Email: [email protected] and [email protected]

Search for more papers by this author

Wenbin Guo,

Corresponding Author

Wenbin Guo

[email protected]

orcid.org/0000-0002-1626-2465

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Correspondence

Email: [email protected] and [email protected]

Search for more papers by this author

Tingyu Yang,

Tingyu Yang

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Department of Child Healthcare, Hunan Children's Hospital, Changsha, China

Search for more papers by this author

Yangpan Ou,

Yangpan Ou

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Huabing Li,

Huabing Li

Department of Radiology, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Feng Liu,

Feng Liu

orcid.org/0000-0002-3570-4222

Department of Radiology, Tianjin Medical University General Hospital, Tianjin, China

Search for more papers by this author

Ping Li,

Ping Li

Department of Psychiatry, Qiqihar Medical University, Qiqihar, China

Search for more papers by this author

Guangrong Xie,

Guangrong Xie

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Jingping Zhao,

Jingping Zhao

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Search for more papers by this author

Xilong Cui,

Corresponding Author

Xilong Cui

[email protected]

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Correspondence

Email: [email protected] and [email protected]

Search for more papers by this author

Wenbin Guo,

Corresponding Author

Wenbin Guo

[email protected]

orcid.org/0000-0002-1626-2465

Department of Psychiatry, National Clinical Research Center for Mental Disorders, and National Center for Mental Disorders, The Second Xiangya Hospital of Central South University, Changsha, China

Correspondence

Email: [email protected] and [email protected]

Search for more papers by this author

First published: 22 July 2024

https://doi.org/10.1111/cns.14871

Citations: 1

The first two authors contributed equally to this work

Share a link

Email
Wechat
Bluesky

Abstract

Main Problem

Anhedonia is a critical diagnostic symptom of major depressive disorder (MDD), being associated with poor prognosis. Understanding the neural mechanisms underlying anhedonia is of great significance for individuals with MDD, and it encourages the search for objective indicators that can reliably identify anhedonia.

Methods

A predictive model used connectome-based predictive modeling (CPM) for anhedonia symptoms was developed by utilizing pre-treatment functional connectivity (FC) data from 59 patients with MDD. Node-based FC analysis was employed to compare differences in FC patterns between melancholic and non-melancholic MDD patients. The support vector machines (SVM) method was then applied for classifying these two subtypes of MDD patients.

Results

CPM could successfully predict anhedonia symptoms in MDD patients (positive network: r = 0.4719, p < 0.0020, mean squared error = 23.5125, 5000 iterations). Compared to non-melancholic MDD patients, melancholic MDD patients showed decreased FC between the left cingulate gyrus and the right parahippocampus gyrus (p__bonferroni = 0.0303). This distinct FC pattern effectively discriminated between melancholic and non-melancholic MDD patients, achieving a sensitivity of 93.54%, specificity of 67.86%, and an overall accuracy of 81.36% using the SVM method.

Conclusions

This study successfully established a network model for predicting anhedonia symptoms in MDD based on FC, as well as a classification model to differentiate between melancholic and non-melancholic MDD patients. These findings provide guidance for clinical treatment.

1 INTRODUCTION

Anhedonia, characterized by the inability to experience pleasure or loss of interest in previously enjoyable activities,¹ is a pivotal symptom in the diagnostic criteria for major depressive disorder (MDD). Its significance is emphasized by both the International Classification of Diseases-10² and the Diagnostic and Statistical Manual of Mental Disorders-IV Edition (DSM-IV). Anhedonia in MDD is associated with increased disease severity, higher risk of relapse and suicide, and poor response to interventions.^{1, 3, 4} Comprehending the neural mechanisms that underlie anhedonia holds significant promise for elucidating the pathophysiology of MDD and could potentially guide the refinement of treatment approaches for MDD.

Resting-state functional magnetic resonance imaging (fMRI) is a feasible approach to explore the neural mechanisms of anhedonia in MDD. It can noninvasively reveal the functional connectivity (FC) between different brain regions.⁵ Previous studies have found a correlation between abnormal FC patterns and anhedonia in MDD. For example, Lu et al.⁶ observed increased FC in the bilateral superior frontal gyrus in MDD patients with anhedonia compared to MDD patients without anhedonia. Yang et al.⁷ discovered a positive correlation between the severity of anhedonia and the extent of ventral caudate FC with the cuneus and middle temporal gyrus in MDD patients. Other studies have also reported heightened FC in the frontal-limbic circuits of MDD patients with anhedonia.^{8, 9} However, the research outcomes have been inconsistent, and these abnormal FC are widely distributed across various brain regions.^{10, 11} Therefore, this study aimed to establish a predictive networks model for anhedonia symptoms using whole-brain FC patterns. The improvement in diagnosing and predicting anhedonia in MDD patients can be achieved by identifying abnormal FC features and constructing predictive models. This facilitates early intervention and treatment. However, it is imperative to acknowledge the existing dearth of research in this particular domain.

Prior evidence supports the possibility that melancholic and non-melancholic MDD represent two distinct biological subtypes. For instance, our previous research revealed differences in spontaneous brain activity at the location of the left anterior cingulate cortex between melancholic and non-melancholic MDD patients,¹² and distinct voxel-mirrored homotopic connectivity in the precentral gyrus and precentral/postcentral gyrus was observed between these two subtypes.¹³ Regarding treatment response, the presence of melancholic symptoms has been positively associated with the effectiveness of electroconvulsive therapy, tricyclic antidepressants,¹⁴ and fluoxetine treatment.¹⁵ We aimed to identify differences in the predictive networks model for anhedonia symptoms between melancholic or non-melancholic MDD, thereby uncovering potential neurobiological mechanisms underlying these two subtypes of MDD, which has significant implications for accurate diagnosis and personalized treatments.

Machine learning-based classification and prediction demonstrate immense potential in the identification of neurobiological biomarkers. These biomarkers not only hold clinical significance but also have the potential to serve as prognostic indicators for MDD.¹⁶ Recently, various machine learning techniques have been applied to predict clinical outcomes of psychiatric disorder patients based on MRI data.^{17, 18} Connectome-based predictive modeling (CPM) is a data-driven, machine learning technique, and has been successfully used to predict treatment response in cocaine use disorder,^{19, 20} treatment response in depression disorder,^{17, 21} and severity of anxiety symptoms.²² This indicates that CPM is suitable for using FC to predict anhedonia symptoms in MDD patients. These FC can be referred to as connectome “fingerprints”,²³ which can serve as important features to distinguish between melancholic and non-melancholic MDD patients. Moreover, support vector machine (SVM) algorithm, a classification machine learning method, was widely used in MDD researches due to its ability to process high-dimensional data and high classification accuracy.^24-26 The advantages of choosing SVM as the classification model are as follows. First, SVM is suitable for both linear and nonlinear problems and offers a variety of built-in kernel functions. Second, SVM performs well with small samples. Third, SVM differentiates between samples of different classes by finding the optimal hyperplane, which has a clear geometric interpretation. Finally, SVM focuses mainly on support vectors, making it less susceptible to outliers, which is crucial for small sample data.

Therefore, the present study aimed to establish a predictive model for anhedonia symptoms using CPM in MDD. Subsequently, we explore the differences in predictive networks for anhedonia symptoms between the two subtypes of MDD (melancholic and non-melancholic MDD). Finally, the SVM method was employed to determine whether these different FC patterns could serve as potential neurobiological markers to classify melancholic and non-melancholic MDD patients. This study will contribute to uncovering the underlying neurobiological mechanisms of anhedonia symptoms in MDD and explores the intrinsic FC patterns of its two distinct subtypes.

2 METHODS

2.1 Participants

The study recruited 64 patients diagnosed with MDD and 32 healthy controls (HCs). Five MDD patients were excluded due to excessive head movement. The patients were recruited from the Second Xiangya Hospital of Central South University. Diagnosis of MDD was based on the DSM-IV criteria. Within the MDD patient group, there were 31 patients with melancholic MDD and 28 patients with non-melancholic MDD. Inclusion criteria for melancholic MDD patients included the pervasive anhedonia or unreactive mood and at least three characteristic symptoms such as depression distinct from grief or loss, worsened morning mood, early morning awakening, appetite or weight disturbances, psychomotor difficulties, and excessive guilt. HCs were recruited from the community and matched with the patient group in terms of age and sex. Exclusion criteria for both the MDD and HCs groups included (1) no previous or current psychiatric disorders (except for MDD in MDD group), neurological disorders, and substance abuse; (2) excluding brain structural abnormalities, severe physical illnesses, pregnancy, and other contraindications to fMRI; (3) HCs did not have first-degree relatives with any psychiatric disorders. All MDD patients included in the study were experiencing their first episode of MDD and were not receiving any medication prior to the neuroimaging session. Ethical approval for data collection was obtained from the Medical Research Ethics Committee of the Second Xiangya Hospital, China, and an informed consent was obtained from all participants.

2.2 Symptom assessment

Depression severity was evaluated by the psychiatrist-assessed Hamilton Depression Rating Scale (HAMD, 17 items).²⁷ Higher HAMD scores indicate more severe depressive symptoms. Anhedonia was evaluated by the Temporal Experience of Pleasure Scale (TEPS).^{28, 29} Lower TEPS scores indicate more severe anhedonia. See Data S1 for details.

2.3 Image acquisition and preprocessing

fMRI scanning was performed at the Second Xiangya Hospital of Central South University using a 3.0-T Siemens MRI Scanner. During the experiments, the participants were instructed to relax with their eyes closed, without falling asleep. Resting-state functional imaging data were obtained utilizing a gradient echo-planar imaging (EPI) sequence (repetition time: 2500 ms, echo time: 25 ms, field of view: 240 × 240 mm, flip angle: 90°, matrix size: 64 × 64, 3.5 mm slice thickness, no gap, 39 slices, and 200 volumes). See Data S1 for details.

2.4 FC matrices

The fMRI data were parcellated into 246 regions of interest (ROIs) utilizing the Human Brainnetome Atlas (BNA),³⁰ which delineates the brain into 210 cortical and 36 subcortical subregions based on anatomical and FC information. Subsequently, representative time series were computed for each ROI by averaging the fMRI time series across all voxels within the respective region. This resulted in the generation of a participant-specific 246 × 246 FC matrix, representing the connectivity strengths between each pair of ROIs.

2.5 CPM analysis

CPM analysis was conducted using MATLAB scripts.²³ Please refer to the Data S1 for specific step-by-step instructions.

2.6 Node-based FC analysis

In the CPM framework, multiple edges are usually present for selection. To address this issue, we adopted a strategy that involved the identification of high-degree nodes, or nodes with a greater number of edges.¹⁹ These highest-degree nodes were then utilized as seeds, and their FC with other nodes was computed.

2.7 SVM analysis

In this study, the LIBSVM software³¹ was employed to evaluate the discriminative capability of abnormal FC values in brain regions. Specifically, the performance of SVM was evaluated in distinguishing melancholic MDD from non-melancholic MDD. The SVM analysis involved these steps: data collection, splitting samples into training and test sets (using five-fold cross-validation), normalizing features to [−1, +1], selecting gaussian radial basis function kernels for classification. These kernels have two parameters, c and g, optimized using grid search and cross-validation. Validation was done via five-fold cross-validation, where samples were randomly split into five groups. Four groups were for training and one for testing.³²

2.8 Statistical analyses

The normality test for continuous data was performed using histograms and the Shapiro–Wilk test. All data conform to a normal distribution. Among the melancholic MDD, non-melancholic MDD, and HCs groups, between-group comparisons of continuous variables were conducted using one-way analysis of variance, while gender distribution was examined via chi-square test. Post-hoc tests were performed for pairwise comparisons. An independent samples t-test was utilized to investigate differences in node-based FC between melancholic and non-melancholic MDD patients, with Bonferroni correction applied. Partial correlation analysis was applied to analyze the correlation between the abnormal node-based FC and clinical variables of the MDD patients with age, education level, illness duration, and framewise displacement serving as covariates, also adjusted with Bonferroni correction. The significance level was set at a two-tailed α of 0.05.

3 RESULTS

3.1 Demographic and clinical variables of participants

Demographic and clinical data are presented in Table 1. Post-hoc analysis showed that, except for TEPS contextual anticipatory score (p = 0.688), the MDD group exhibited significantly lower scores than the HCs across all other TEPS subscales and total scores (all p < 0.05). Additionally, there was no significant difference in HAMD scores between the melancholic and non-melancholic MDD patients (p = 0.779). The scores for total TEPS, TEPS anticipatory, TEPS abstract anticipatory and TEPS contextual anticipatory were higher in non-melancholic MDD patients compared to melancholic MDD patients (all p < 0.05).

TABLE 1. Demographic and clinical characteristics of MDD and HCs groups.

Variables	MDD (n = 59)		HCs (n = 32)	F, t or χ²	p	Follow-up tests
Variables	Melancholic (n = 31)	Non-melancholic (n = 28)	HCs (n = 32)	F, t or χ²	p	Follow-up tests
Age (years)	28.65 ± 5.30	32.04 ± 8.18	29.59 ± 5.00	2.291	0.107
Gender (male/female)	10/21	10/18	15/17	1.550	0.461
Education (years)	15.16 ± 3.195	12.54 ± 3.00	14.59 ± 2.82	6.143	0.003
Illness duration (months)	6.75 ± 4.26	5.96 ± 4.64		0.462	0.500
FD	0.05 ± 0.03	0.04 ± 0.02	0.05 ± 0.02	0.373	0.690
HAMD	21.77 ± 3.79	21.00 ± 3.14	0.94 ± 0.95	527.891	<0.001	MDD > HCs
TEPS anticipatory	26.30 ± 7.83	33.71 ± 5.91	35.13 ± 6.11	24.686	<0.001	MDD < HCs, Melancholic < Non-melancholic
Abstract anticipatory	13.17 ± 4.71	17.04 ± 3.85	19.94 ± 2.7	5.951	0.004	MDD < HCs, Melancholic < Non-melancholic
Contextual anticipatory	13.13 ± 3.90	16.68 ± 3.64	15.19 ± 4.31	18.186	<0.001	Melancholic < Non-melancholic
TEPS consummatory	32.00 ± 7.34	35.75 ± 7.12	42.53 ± 6.63	14.863	<0.001	MDD < HCs
Abstract consummatory	20.20 ± 5.12	22.39 ± 5.28	26.78 ± 4.23	10.947	<0.001	MDD < HCs
Contextual consummatory	11.80 ± 3.18	13.36 ± 3.27	15.75 ± 3.65	19.527	<0.001	MDD < HCs
TEPS total score	58.30 ± 13.95	69.46 ± 11.16	77.66 ± 11.6	64.191	<0.001	MDD < HCs, Melancholic < Non-melancholic

Abbreviations: FD, framewise displacement; HAMD, 17-item Hamilton Rating Scale for Depression; HCs, healthy controls; TEPS, temporal experience of pleasure scale.

3.2 Prediction of the anhedonia symptoms in MDD patients

CPM could successfully predict anhedonia symptoms in MDD (positive network: r = 0.4719, p < 0.0020, mean squared error (MSE) = 23.5125, 5000 iterations). The Figure 1 showed that 26 edges were positively related to the TEPS abstract consummatory score with a typical significance threshold of p = 0.0001. The predictive networks model primarily focuses on connections between the temporal, frontal, parietal, and occipital lobes. The nodes in right inferior temporal gyrus (ITG) and right parahippocampal gyrus (PhG_R) with five or more edges and were the highest-degree nodes (Table 2, Table S1 and Figure 1). We designated these two nodes as seed points and calculated node-based FC.

Details are in the caption following the image — **FIGURE 1**
Open in figure viewer PowerPoint

Visualizing selected connectivity features. Image (A) (Node Threshold > 0) represents 26 edges positive correlated with the TEPS Abstract Consummatory scores with a typical significance threshold of p = 0.0001. Image (B) shows nodes with edge connections greater than 5. (C) and (D) Matrix plots: The cells of the matrices plots represent the total number of positive edges connecting the nodes in different brain regions. The images (C) and (D) represent seven different brain lobes and 24 different brain gyrus of the Human Brainnetome Atlas (BNA) template, respectively. (E): The CPM model predicted the TEPS Abstract Consummatory scores of the MDD patients (r = 0.4719, p < 0.0020, 5000 iterations). For details on the division of brain regions, please refer to the http://atlas.brainnetome.org/bnatlas.html. CPM, connectome-based predictive modeling; MDD, major depressive disorder; TEPS, temporal experience of pleasure scale.

TABLE 2. Highest-degree nodes and their connections in the positive network predictive of TEPS abstract consummatory score in MDD group.

Network	Connections
Network	Node	Frontal Lobe	Temporal Lobe	Parietal Lobe	Insular Lobe	Limbic Lobe	Occipital Lobe	Subcortical Nuclei
Positive	ITG_R		STG_R (3)	IPL_ R	INS_R
			MTG_R
			ITG_L
			FuG_L
			pSTS_R
	PhG_R	SFG_R	PhG_R	IPL_L		CG_L
				PCun_L		CG_ R
				PCun_R

Note: The division of brain regions is based on the Human Brainnetome Atlas (BNA) template.
Abbreviations: CG, cingulate gyrus; FuG, fusiform gyrus; INS, insular gyrus; IPL, inferior parietal lobule; ITG, inferior temporal gyrus; MDD, major depressive disorder; MTG, middle temporal gyrus; Pcun, precuneus; PhG, parahippocampal gyrus; pSTS, posterior superior temporal sulcus; SFG, superior frontal gyrus; STG, superior temporal gyrus.

3.3 Comparison of node-based FC between the melancholic and non-melancholic MDD patients

Compared to non-melancholic MDD patients, melancholic MDD patients showed decreased FC between the left cingulate gyrus (CG_L) and the PhG_R (p__bonferroni = 0.0303) (Table S2 and Figure 2).

3.4 SVM analyses

The SVM analysis revealed that FC values between PhG_R and CG_L could effectively discriminate between melancholic and non-melancholic MDD patients, yielding a sensitivity of 93.54%, a specificity of 67.86%, and an accuracy of 81.36% (Figure 3). Under optimal parameters, the standard errors (SEs) of the accuracy, sensitivity, and specificity of the SVM classification are 0.031, 0.028, and 0.067, respectively.

3.5 Correlations between the FC values and anhedonia symptoms

It was found that after controlling for age, education level, illness duration, and framewise displacement, FC values between PhG_R and CG_L were positively correlated with TEPS consummatory scores (p__bonferroni = 0.0016) and TEPS total scores (p__bonferroni = 0.0200) in MDD patients (Figure 4).

4 DISCUSSION

This study applied the CPM method based on FC of the whole-brain to establish a predictive model for anhedonia symptoms in MDD. There are several meaningful findings. First, CPM could successfully predict anhedonia symptoms in MDD. The positive predictive networks primarily focus on connections between the temporal, frontal, parietal, and occipital lobes. Second, compared to non-melancholic MDD patients, melancholic MDD patients showed decreased FC between the default mode network (DMN, CG_L) and the limbic network (LN, PhG_R) in positive predictive networks, according to Yeo 7 network division criteria.³⁰ Finally, the abnormal FC patterns between the DMN and the LN in the predictive networks serve as potential neurobiological markers to distinguish between melancholic and non-melancholic MDD patients.

The present study indicates that the FC values can be utilized to predict the anhedonia symptoms in MDD patients. Specifically, a higher FC value represents a higher TEPS score, indicating fewer anhedonia symptoms. The right ITG and the right PhG are the two highest-degree nodes based on the prediction network of the anhedonia symptoms. The ITG is related to emotion regulation and expression processes.^{33, 34} The PhG is not only important for memory formation but also involved in regulating emotional processing and emotional memory.³⁵ The PhG is believed to be involved in the formation of contextual associations and episodic memory during mood initiation in MDD patients. This may contribute to the tendency of patients to engage in repetitive rumination on negative thoughts rather than experiencing pleasure.^{36, 37} Consistent with our findings, previous studies have identified abnormal widespread FC patterns in the ITG and PhG of MDD patients with anhedonia symptoms.³⁸ Previous research also found that MDD patients exhibits abnormality in brain structure in ITG and the PhG.³⁹ Lower gray matter volume and cortical thickness were showed in the right ITG among MDD patients compared with HCs,^{39, 40} and were associated with more severe anhedonia severity.⁴¹ Gray matter loss in the PhG is the most frequently observed morphometric abnormality in patients with MDD.⁴² Furthermore, structural MRI research have found that a higher risk of anhedonia is associated with thinner PhG.⁴³ The diminution of cortical thickness in the ITG and PhG could potentially trigger compensatory elevations in FC among these regions and other cerebral areas, effectively mitigating the functional deficits stemming from structural alterations. It is noteworthy that the MDD subjects recruited for this study were experiencing their initial episode and had not received any pharmacological intervention, thereby minimizing confounding factors such as recurrent episodes and medication effects. These augmented FC patterns may be interpreted as etiological factors rather than mere outcomes of anhedonia symptoms in patients with MDD.

In addition, the correlation between FC and anhedonia symptoms may be related to changes in plasma C-reactive protein levels. Research has shown that FC associated with plasma C-reactive protein levels, including right ITG and PhG, can predict anhedonia symptoms in patients with MDD.⁴⁴ Abstract consummatory pleasure refers to the emotional experience of something that is inherently more abstract or less concrete in nature (for example, “I enjoy taking a deep breath of fresh air when I walk outside.”). Connecting such spatial information (like “outside”) with emotional information is closely associated with the function of the PhG.^{35, 45} Hence, the reduced FC between the PhG and brain regions associated with anhedonia(such as the superior frontal gyrus⁴⁶ and precuneus⁴⁷) in MDD patients could result in compromised information processing, leading to disruptions in the perception and experience of pleasure. Of note, the FC of right ITG and right PhG also related to gene expression.⁴⁸ Further investigation is necessary to elucidate the intricate mechanisms underlying relationships between the FC of right ITG and right PhG and anhedonia in MDD patients.

Compared with non-melancholic MDD patients, melancholic MDD patients showed decreased FC between the PhG_R and CG_L in positive predictive networks of anhedonia symptoms. The CG, the region of the DMN, plays a crucial role in the regulation of emotions and memory, receiving inputs from the orbitofrontal cortex and amygdala, and establishing connections with the PhG.^49-51 Our previous research has shown that higher FC corresponds to milder anhedonia symptoms. Compared to MDD patients, the increased FC between PhG_R and CG_L in HCs group also confirms this.⁵² The present study found that the FC between PhG_R and CG_L is associated with anhedonia and effectively distinguishes between melancholic MDD and non-melancholic MDD. This might suggest that the abnormal FC between PhG_R and CG_L could be an intrinsic neural substrate for the difference in the anhedonia symptoms between melancholic MDD and non-melancholic MDD patients. Furthermore, previous research suggested that aberrant FC in the nucleus accumbens is a critical node in melancholic MDD-related anhedonia.^{53, 54} The anterior cingulate cortex (a constituent of the CG) forms a neural circuit with the nucleus accumbens. Within this circuit, neurons encode signals responsive to reward acquisition and integrate effort expenditure signals with reward anticipation to influence subsequent behavior.⁵⁵ Nevertheless, prolonged exposure to stress disrupts this neural loop, resulting in motivational anhedonia.⁵⁵ Integrating these findings with our study outcomes, it indicates a potential involvement of the left CG as another critical node implicated in the anhedonia present in melancholic MDD patients.

CG is recognized as a crucial node of the DMN,^{50, 56} whereas PhG constitutes a component of the limbic system.⁵⁷ CG and PhG have been proposed as potential interfaces connecting the systems responsible for semantic retrieval, episodic encoding, and emotional processing.⁵⁸ This may indicate that the FC between the CG and PhG represents the integration process of the DMN and the LN. This integration determines whether patients clinically present with melancholic or non-melancholic MDD. Besides, it is worth noting that previous literature has found a close association between the melancholic MDD and the depression severity.^{59, 60} However, our current investigation did not reveal any statistically significant differences in depression severity between the two groups of MDD patients. This suggests that the FC between CG and PhG may serve as a specific biological marker for melancholic MDD, independent of the severity of depressive symptoms. Notably, MDD is considered a heterogeneous disease,^{61, 62} and previous research⁶³ combined with the present study suggests that MDD patients with melancholic traits may represent a subtype of MDD characterized by unique neuroimaging characteristics. Additionally, this study revealed no significant inter-group differences in TEPS abstract consummatory scores between melancholic and non-melancholic MDD patients, indicating that changes in related FC might occur prior to clinically evident abstract consummatory anhedonia in MDD patients. This further confirms that the FC between PhG_R and CG_L is an intrinsic neural substrate for the anhedonia symptoms among different subtypes of MDD.

This study has some highlights. First, CPM was used to establish a predictive model for anhedonia symptoms in MDD. Second, two machine learning methods, CPM and SVM, were respectively employed in our study for prediction and classification purposes. Next, both melancholic and non-melancholic MDD exhibit anhedonia symptoms, but we discovered distinct intrinsic neural substrates for anhedonia symptoms in these two subtypes of MDD, revealing differences in their FC patterns. Finally, abnormal FC patterns might serve as neural biomarkers to differentiate between these two subtypes of MDD.

However, there are some limitations to this study. First, the sample size in each group is relatively small, which may limit the generalizability of our findings. Second, the predictive model we have constructed is deficient in independent sample validation, which also limits its generalizability. Finally, the cross-sectional nature of this study limits its ability to establish a direct relationship between FC changes and anhedonia in MDD patients. Future longitudinal studies are needed to further explore the stability of FC in predicting anhedonia and classifying melancholic and non-melancholic MDD.

5 CONCLUSION

This study successfully established a positive predictive networks model for anhedonia symptoms in MDD patients. The different FC patterns between the DMN and the LN in the predictive networks serve as potential neurobiological markers to distinguish between melancholic and non-melancholic MDD patients. These research findings reveal distinct neural substrates for anhedonia symptoms in melancholic and non-melancholic MDD patients, offering guidance for clinical treatment approaches.

ACKNOWLEDGMENTS

This work was supported by the National Natural Science Foundation of China.

CONFLICT OF INTEREST STATEMENT

There are no financial conflicts of interest to disclose.

Open Research

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supporting Information

REFERENCES

1Pizzagalli DA. Toward a better understanding of the mechanisms and pathophysiology of anhedonia: are we ready for translation? Am J Psychiatry. 2022; 179(7): 458-469.
10.1176/appi.ajp.20220423
PubMed Google Scholar
2 World Health Organization. The ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines. World Health Organization; 1992.
Google Scholar
3Whitton AE, Pizzagalli DA. Anhedonia in depression and bipolar disorder. Curr Top Behav Neurosci. 2022; 58: 111-127.
10.1007/7854_2022_323
PubMed Google Scholar
4Whitton AE, Kumar P, Treadway MT, et al. Distinct profiles of anhedonia and reward processing and their prospective associations with quality of life among individuals with mood disorders. Mol Psychiatry. 2023; 28: 5272-5281.
10.1038/s41380-023-02165-1
PubMed Google Scholar
5van den Heuvel MP, Hulshoff Pol HE. Exploring the brain network: a review on resting-state fMRI functional connectivity. Eur Neuropsychopharmacol. 2010; 20(8): 519-534.
10.1016/j.euroneuro.2010.03.008
CAS PubMed Web of Science® Google Scholar
6Lu S, Shao J, Feng Q, et al. Aberrant interhemispheric functional connectivity in major depressive disorder with and without anhedonia. BMC Psychiatry. 2022; 22(1): 688.
10.1186/s12888-022-04343-x
PubMed Google Scholar
7Yang XH, Tian K, Wang DF, et al. Anhedonia correlates with abnormal functional connectivity of the superior temporal gyrus and the caudate nucleus in patients with first-episode drug-naive major depressive disorder. J Affect Disord. 2017; 218: 284-290.
10.1016/j.jad.2017.04.053
PubMed Web of Science® Google Scholar
8Wang X, Xia Y, Yan R, et al. The relationship between disrupted anhedonia-related circuitry and suicidal ideation in major depressive disorder: a network-based analysis. Neuroimage Clin. 2023; 40:103512.
10.1016/j.nicl.2023.103512
PubMed Google Scholar
9Bracht T, Mertse N, Walther S, et al. Link between structural connectivity of the medial forebrain bundle, functional connectivity of the ventral tegmental area, and anhedonia in unipolar depression. Neuroimage Clin. 2022; 34:102961.
10.1016/j.nicl.2022.102961
PubMed Google Scholar
10Misaki M, Tsuchiyagaito A, Guinjoan SM, Rohan ML, Paulus MP. Trait repetitive negative thinking in depression is associated with functional connectivity in negative thinking state rather than resting state. bioRxiv. 2023. doi:10.1101/2023.03.23.533932
Google Scholar
11Rzepa E, McCabe C. Anhedonia and depression severity dissociated by dmPFC resting-state functional connectivity in adolescents. J Psychopharmacol. 2018; 32(10): 1067-1074.
10.1177/0269881118799935
PubMed Google Scholar
12Zhang Y, Cui X, Ou Y, et al. Differentiating melancholic and non-melancholic major depressive disorder using fractional amplitude of low-frequency fluctuations. Front Psych. 2021; 12:763770.
10.3389/fpsyt.2021.763770
Google Scholar
13Shan X, Cui X, Liu F, et al. Shared and distinct homotopic connectivity changes in melancholic and non-melancholic depression. J Affect Disord. 2021; 287: 268-275.
10.1016/j.jad.2021.03.038
PubMed Google Scholar
14Rush AJ, Weissenburger JE. Melancholic symptom features and DSM-IV. Am J Psychiatry. 1994; 151(4): 489-498.
10.1176/ajp.151.4.489
CAS PubMed Web of Science® Google Scholar
15Lin CH, Huang CJ, Liu SK. Melancholic features in inpatients with major depressive disorder associate with differential clinical characteristics and treatment outcomes. Psychiatry Res. 2016; 238: 368-373.
10.1016/j.psychres.2015.11.009
PubMed Web of Science® Google Scholar
16Ebdrup BH, Axelsen MC, Bak N, et al. Accuracy of diagnostic classification algorithms using cognitive-, electrophysiological-, and neuroanatomical data in antipsychotic-naïve schizophrenia patients. Psychol Med. 2019; 49(16): 2754-2763.
10.1017/S0033291718003781
PubMed Google Scholar
17Ju Y, Horien C, Chen W, et al. Connectome-based models can predict early symptom improvement in major depressive disorder. J Affect Disord. 2020; 273: 442-452.
10.1016/j.jad.2020.04.028
PubMed Google Scholar
18Gao S, Calhoun VD, Sui J. Machine learning in major depression: from classification to treatment outcome prediction. CNS Neurosci Ther. 2018; 24(11): 1037-1052.
10.1111/cns.13048
PubMed Web of Science® Google Scholar
19Yip SW, Scheinost D, Potenza MN, Carroll KM. Connectome-based prediction of cocaine abstinence. Am J Psychiatry. 2019; 176(2): 156-164.
10.1176/appi.ajp.2018.17101147
PubMed Web of Science® Google Scholar
20Lichenstein SD, Scheinost D, Potenza MN, Carroll KM, Yip SW. Dissociable neural substrates of opioid and cocaine use identified via connectome-based modelling. Mol Psychiatry. 2021; 26(8): 4383-4393.
10.1038/s41380-019-0586-y
PubMed Google Scholar
21Fan S, Nemati S, Akiki TJ, et al. Pretreatment brain connectome fingerprint predicts treatment response in major depressive disorder. Chronic Stress (Thousand Oaks). 2020; 4:2470547020984726.
10.1177/2470547020984726
Google Scholar
22Wang Z, Goerlich KS, Ai H, Aleman A, Luo YJ, Xu P. Connectome-based predictive modeling of individual anxiety. Cereb Cortex. 2021; 31(6): 3006-3020.
10.1093/cercor/bhaa407
PubMed Web of Science® Google Scholar
23Shen X, Finn ES, Scheinost D, et al. Using connectome-based predictive modeling to predict individual behavior from brain connectivity. Nat Protoc. 2017; 12(3): 506-518.
10.1038/nprot.2016.178
CAS PubMed Web of Science® Google Scholar
24Geng X, Xu J, Liu B, Shi Y. Multivariate classification of major depressive disorder using the effective connectivity and functional connectivity. Front Neurosci. 2018; 12:38.
10.3389/fnins.2018.00038
PubMed Web of Science® Google Scholar
25Bhadra S, Kumar CJ. An insight into diagnosis of depression using machine learning techniques: a systematic review. Curr Med Res Opin. 2022; 38(5): 749-771.
10.1080/03007995.2022.2038487
PubMed Google Scholar
26Mao Y, Chen Q, Wei D, et al. Resting-state functional connectome predicts individual differences in depression during COVID-19 pandemic. Am Psychol. 2022; 77(6): 760-769.
10.1037/amp0001031
PubMed Google Scholar
27Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960; 23(1): 56-62.
10.1136/jnnp.23.1.56
CAS PubMed Web of Science® Google Scholar
28Simon JJ, Zimmermann J, Cordeiro SA, et al. Psychometric evaluation of the temporal experience of pleasure scale (TEPS) in a German sample. Psychiatry Res. 2018; 260: 138-143.
10.1016/j.psychres.2017.11.060
PubMed Web of Science® Google Scholar
29Zhou H, Liu W, Fan J, Xia J, Zhu J, Zhu X. The temporal experience of pleasure scale (TEPS): measurement invariance across gender in Chinese university students. Front Psychol. 2019; 10:2130.
10.3389/fpsyg.2019.02130
PubMed Google Scholar
30Fan L, Li H, Zhuo J, et al. The human Brainnetome atlas: a new brain atlas based on connectional architecture. Cereb Cortex. 2016; 26(8): 3508-3526.
10.1093/cercor/bhw157
PubMed Web of Science® Google Scholar
31Chang C-C, Lin C-J. LIBSVM: a library for support vector machines. ACM Trans Intell Syst Technol. 2011; 2(3): 1-27.
10.1145/1961189.1961199
Web of Science® Google Scholar
32Chao-Gan Y, Yu-Feng Z. DPARSF: a MATLAB toolbox for “pipeline” data analysis of resting-state fMRI. Front Syst Neurosci. 2010; 4:13.
PubMed Google Scholar
33Lin Y-H, Young IM, Conner AK, et al. Anatomy and white matter connections of the inferior temporal gyrus. World Neurosurg. 2020; 143: e656-e666.
10.1016/j.wneu.2020.08.058
PubMed Web of Science® Google Scholar
34Hu B, Rao J, Li X, et al. Emotion regulating attentional control abnormalities in major depressive disorder: an event-related potential study. Sci Rep. 2017; 7(1): 13530.
10.1038/s41598-017-13626-3
PubMed Google Scholar
35Aminoff EM, Kveraga K, Bar M. The role of the parahippocampal cortex in cognition. Trends Cogn Sci. 2013; 17(8): 379-390.
10.1016/j.tics.2013.06.009
PubMed Web of Science® Google Scholar
36Harel EV, Tennyson RL, Fava M, Bar M. Linking major depression and the neural substrates of associative processing. Cogn Affect Behav Neurosci. 2016; 16(6): 1017-1026.
10.3758/s13415-016-0449-9
PubMed Web of Science® Google Scholar
37Zamoscik V, Huffziger S, Ebner-Priemer U, Kuehner C, Kirsch P. Increased involvement of the parahippocampal gyri in a sad mood predicts future depressive symptoms. Soc Cogn Affect Neurosci. 2014; 9(12): 2034-2040.
10.1093/scan/nsu006
PubMed Web of Science® Google Scholar
38Yu F, Fang H, Zhang J, et al. Individualized prediction of consummatory anhedonia from functional connectome in major depressive disorder. Depress Anxiety. 2022; 39(12): 858-869.
10.1002/da.23292
PubMed Web of Science® Google Scholar
39Gray JP, Müller VI, Eickhoff SB, Fox PT. Multimodal abnormalities of brain structure and function in major depressive disorder: a meta-analysis of neuroimaging studies. Am J Psychiatry. 2020; 177(5): 422-434.
10.1176/appi.ajp.2019.19050560
PubMed Web of Science® Google Scholar
40Schmaal L, Hibar DP, Sämann PG, et al. Cortical abnormalities in adults and adolescents with major depression based on brain scans from 20 cohorts worldwide in the ENIGMA major depressive disorder working group. Mol Psychiatry. 2017; 22(6): 900-909.
10.1038/mp.2016.60
CAS PubMed Web of Science® Google Scholar
41Kocsis K, Holczer A, Kazinczi C, et al. Voxel-based asymmetry of the regional gray matter over the inferior temporal gyrus correlates with depressive symptoms in medicated patients with major depressive disorder. Psychiatry Res Neuroimaging. 2021; 317:111378.
10.1016/j.pscychresns.2021.111378
PubMed Google Scholar
42Arnone D, Job D, Selvaraj S, et al. Computational meta-analysis of statistical parametric maps in major depression. Hum Brain Mapp. 2016; 37(4): 1393-1404.
10.1002/hbm.23108
PubMed Web of Science® Google Scholar
43Zhu X, Ward J, Cullen B, et al. Phenotypic and genetic associations between anhedonia and brain structure in UK biobank. Transl Psychiatry. 2021; 11(1): 395.
10.1038/s41398-021-01522-4
PubMed Web of Science® Google Scholar
44Yin L, Xu X, Chen G, et al. Inflammation and decreased functional connectivity in a widely-distributed network in depression: centralized effects in the ventral medial prefrontal cortex. Brain Behav Immun. 2019; 80: 657-666.
10.1016/j.bbi.2019.05.011
PubMed Web of Science® Google Scholar
45Peng X, Burwell RD. Beyond the hippocampus: the role of parahippocampal-prefrontal communication in context-modulated behavior. Neurobiol Learn Mem. 2021; 185:107520.
10.1016/j.nlm.2021.107520
PubMed Web of Science® Google Scholar
46Fang Z, Mu Q, Wu C, et al. The impacts of anhedonia on brain functional alterations in patients with major depressive disorder: a resting-state functional magnetic resonance imaging study of regional homogeneity. J Psychiatr Res. 2022; 156: 84-90.
10.1016/j.jpsychires.2022.10.028
PubMed Google Scholar
47Luo Y, Qi S, Chen X, You X, Huang X, Yang Z. Pleasure attainment or self-realization: the balance between two forms of well-beings are encoded in default mode network. Soc Cogn Affect Neurosci. 2017; 12(10): 1678-1686.
10.1093/scan/nsx078
PubMed Google Scholar
48Li Y, Yu X, Ma Y, et al. Neural signatures of default mode network in major depression disorder after electroconvulsive therapy. Cereb Cortex. 2023; 33(7): 3840-3852.
10.1093/cercor/bhac311
PubMed Google Scholar
49Rolls ET. The cingulate cortex and limbic systems for emotion, action, and memory. Brain Struct Funct. 2019; 224(9): 3001-3018.
10.1007/s00429-019-01945-2
PubMed Web of Science® Google Scholar
50Greicius MD, Supekar K, Menon V, Dougherty RF. Resting-state functional connectivity reflects structural connectivity in the default mode network. Cereb Cortex. 2009; 19(1): 72-78.
10.1093/cercor/bhn059
PubMed Web of Science® Google Scholar
51Yang J, Chen K, Zhang J, et al. Molecular mechanisms underlying human spatial cognitive ability revealed with neurotransmitter and transcriptomic mapping. Cereb Cortex. 2023; 33(23): 11320-11328.
10.1093/cercor/bhad368
PubMed Google Scholar
52Cheng W, Rolls ET, Qiu J, et al. Increased functional connectivity of the posterior cingulate cortex with the lateral orbitofrontal cortex in depression. Transl Psychiatry. 2018; 8(1): 90.
10.1038/s41398-018-0139-1
PubMed Web of Science® Google Scholar
53Hu Y, Zhao C, Zhao H, Qiao J. Abnormal functional connectivity of the nucleus accumbens subregions mediates the association between anhedonia and major depressive disorder. BMC Psychiatry. 2023; 23(1): 282.
10.1186/s12888-023-04693-0
PubMed Google Scholar
54Liu R, Wang Y, Chen X, Zhang Z, Xiao L, Zhou Y. Anhedonia correlates with functional connectivity of the nucleus accumbens subregions in patients with major depressive disorder. Neuroimage Clin. 2021; 30:102599.
10.1016/j.nicl.2021.102599
PubMed Web of Science® Google Scholar
55Fetcho RN, Parekh PK, Chou J, et al. A stress-sensitive frontostriatal circuit supporting effortful reward-seeking behavior. Neuron. 2024; 112(3): 473-487.e4.
10.1016/j.neuron.2023.10.020
CAS PubMed Google Scholar
56Hagmann P, Cammoun L, Gigandet X, et al. Mapping the structural core of human cerebral cortex. PLoS Biol. 2008; 6(7):e159.
10.1371/journal.pbio.0060159
CAS PubMed Web of Science® Google Scholar
57Catani M, Dell'acqua F, Thiebaut de Schotten M. A revised limbic system model for memory, emotion and behaviour. Neurosci Biobehav Rev. 2013; 37(8): 1724-1737.
10.1016/j.neubiorev.2013.07.001
PubMed Web of Science® Google Scholar
58Binder JR, Desai RH, Graves WW, Conant LL. Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. Cereb Cortex. 2009; 19(12): 2767-2796.
10.1093/cercor/bhp055
PubMed Web of Science® Google Scholar
59Tondo L, Vázquez GH, Baldessarini RJ. Melancholic versus nonmelancholic major depression compared. J Affect Disord. 2020; 266: 760-765.
10.1016/j.jad.2020.01.139
CAS PubMed Web of Science® Google Scholar
60Dold M, Bartova L, Fugger G, et al. Melancholic features in major depression – a European multicenter study. Prog Neuropsychopharmacol Biol Psychiatry. 2021; 110:110285.
10.1016/j.pnpbp.2021.110285
CAS PubMed Google Scholar
61Nguyen TD, Harder A, Xiong Y, et al. Genetic heterogeneity and subtypes of major depression. Mol Psychiatry. 2022; 27(3): 1667-1675.
10.1038/s41380-021-01413-6
CAS PubMed Google Scholar
62Drysdale AT, Grosenick L, Downar J, et al. Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nat Med. 2017; 23(1): 28-38.
10.1038/nm.4246
CAS PubMed Web of Science® Google Scholar
63Wang Y, Tang S, Zhang L, et al. Data-driven clustering differentiates subtypes of major depressive disorder with distinct brain connectivity and symptom features. Br J Psychiatry. 2021; 219(5): 606-613.
10.1192/bjp.2021.103
PubMed Web of Science® Google Scholar

Citing Literature

Volume30, Issue7

July 2024

e14871

Neural substrates of predicting anhedonia symptoms in major depressive disorder via connectome-based modeling