Sex-specific resting state brain network dynamics in patients with major depressive disorder

Abstract

Sex-specific neurobiological changes have been implicated in Major Depressive Disorder (MDD). Dysfunctions of the default mode network (DMN), salience network (SN) and frontoparietal network (FPN) are critical neural characteristics of MDD, however, the potential moderating role of sex on resting-state network dynamics in MDD has not been sufficiently evaluated. Thus, resting-state functional magnetic resonance imaging (fMRI) data were collected from 138 unmedicated patients with first-episode MDD (55 males) and 243 healthy controls (HCs; 106 males). Recurring functional network co-activation patterns (CAPs) were extracted, and time spent in each CAP (the total amount of volumes associated to a CAP), persistence (the average number of consecutive volumes linked to a CAP), and transitions across CAPs involving the SN, DMN and FPN were quantified. Relative to HCs, MDD patients exhibited greater persistence in a CAP involving activation of the DMN and deactivation of the FPN (DMN + FPN-). In addition, relative to the sex-matched HCs, the male MDD group spent more time in two CAPs involving the SN and DMN (i.e., DMN + SN- and DMN-SN + ) and transitioned more frequently from the DMN + FPN- CAP to the DMN + SN- CAP relative to the male HC group. Conversely, the female MDD group showed less persistence in the DMN + SN- CAP relative to the female HC group. Our findings highlight that the imbalance between SN and DMN could be a neurobiological marker supporting sex differences in MDD. Moreover, the dominance of the DMN accompanied by the deactivation of the FPN could be a sex-independent neurobiological correlate related to depression.

Introduction

Major depressive disorder (MDD) is a prevalent psychiatric disorder with higher prevalence in females than males [1]. Sex differences in depressive symptoms, comorbidity and antidepressant efficiency in MDD are also commonly reported [2]. These possible sex differences in MDD have spurred interest in uncovering the potential sex-specific neural underpinnings of MDD. Emerging preclinical evidence shows that neurobiological abnormalities may differ between males and females with MDD [3, 4]. Sex-specific findings have emerged from neuroimaging studies in MDD [5–8], predominantly in limbic/striatal-prefrontal regions. Our prior study found that only female MDD patients exhibited atypical hyperactivity of limbic and striatal regions when experiencing psychosocial stress [9]. Another recent structural study revealed lower surface area in the ventrolateral prefrontal cortex, and lower cortical volume in the rostromedial prefrontal cortex, in female MDD patients compared to healthy females, whereas the male MDD patients showed opposite structural patterns when compared to healthy males [6]. Collectivity, prior neuroimaging findings thus revealed that sex-specific neural mechanisms in MDD do not only pertain to the magnitude, but also to the direction of the effects, highlighting the need for comprehensive investigations. However, this has not yet been achieved, particularly regarding sex-specific alterations in large-scale resting-state neural networks [10].

Mounting evidence suggests that MDD patients are characterized by a dysfunction of several large-scale networks [i.e., default mode network (DMN), frontoparietal network (FPN) and salience network (SN)]. The most common outcome is that individuals with MDD exhibit hypoconnectivity within the FPN [11, 12], hyperconnectivity within the DMN [13], and hyperconnectivity between the DMN and FPN [12]. Of note, hypoconnectivity within the DMN has been reported among patients with recurrent episodes [14], suggesting that the trajectory of MDD-related DMN alterations may change with increasing number of major depressive episodes. Moreover, in healthy individuals, sex differences in network organization were also observed in terms of DMN, SN and FPN, with emerging evidence suggesting that females display more intra-network connectivity, while males exhibit more inter-network connectivity [15, 16], implicating potential sex differences in information processing.

Critically, few studies have evaluated the role of sex assigned at birth on large-scale functional brain network alterations in MDD. One amygdala-based functional connectivity (FC) study found that compared to healthy peers of the same gender, females with depressive symptoms had greater amygdala FC with the insula and the mid-posterior cingulate cortex (PCC), whiles males with depressive symptoms showed weaker FC between the amygdala and subcallosal prefrontal cortex [17]. Another FC study found that males with MDD exhibited increased anterior cingulate FC with the PCC and decreased anterior cingulate FC with the anterior insula, temporal pole, and lateral prefrontal cortex; this study also revealed within-DMN hyperconnectivity in males with MDD, whereas females with MDD featured a modest within-DMN hypoconnectivity [18]. Together, these findings thus highlight increased functional coupling within the DMN and decreased functional coupling between the SN and DMN in males with MDD relative to healthy males, as well as modestly decreased functional coupling within the DMN and increased functional coupling between the SN and DMN in females with MDD relative to healthy females.

One potential limitation of previous work probing sex-specific brain network mechanisms in MDD pertains to the focus on static functional network properties. Dynamic network analytical techniques capture changes in functional coordination among distributed brain regions over time [19–21], which is likely important regarding functional network alterations in MDD. Relatedly, one study comparing statistical models involving static versus dynamic functional network properties found that the latter enabled better MDD identification [22]. Several studies have examined brain dynamics in MDD without probing potential sex differences using co-activation pattern (CAP) analysis [23, 24], which parses resting-state data into CAPs and allows for the computation of several network properties. One CAP study in adolescents found that higher levels of depressive symptoms severity was associated with greater expression and larger persistence of a frontoinsular-DMN CAP and more transitions between this network and a canonical DMN network [24]. Additionally, adult females diagnosed with MDD spend more time in an FPN-posterior DMN CAP and transition more frequently between this CAP and a canonical DMN CAP [23]. These reports revealed that MDD patients may exhibit atypical functional coordination within the DMN and between the FPN and DMN, which supports prior findings observed in MDD using static FC. However, sex-specific resting-state network dynamics in MDD have not been sufficiently evaluated.

Here, we tested sex differences in resting dynamic functional network properties in a relatively large sample of male and female (sex assigned at birth) healthy controls and first-episode unmedicated MDD patients using a data-driven CAP approach [25, 26]. Since prior static FC analyses revealed sex-specific functional coupling among SN and DMN regions (i.e., males with MDD exhibited greater within-DMN FC and lower SN-DMN FC relative to male HCs; females with MDD exhibited lower within-DMN FC and greater SN-DMN FC), we speculated that male and female MDD patients would exhibit distinct CAP properties (i.e., time spent and persistence) in CAPs involving the SN and DMN. Specifically, the female MDD patients would spend less time/persist less in CAPs involving the coactivation of DMN regions or an opposite polarity between SN and DMN regions (since the SN and DMN are anti-correlated in the resting-state as revealed by static FC), whereas the male MDD patients would spend more time/persist more in these same CAPs.

Materials and methods

Participants

Patients meeting DSM-IV-TR Axis I disorders criteria for their first episode were recruited, with exclusion criteria for potential confounding effects of antidepressant medications, multiple episodes, and comorbidities (see Supplemental Methods for details). All participants were aware of the study’s purpose and provided informed written consent. The study was conducted in accordance with the Declaration of Helsinki and was approved by the ethics committee of the Second Xiangya Hospital of Central South University. Fourteen MDD patients and 22 HCs were excluded because of excessive head movement (see fMRI Preprocessing for details), leaving 243 HCs (137 female) and 138 MDD (83 female) patients for analysis. Clinical and demographic characteristics of MDD patients and HCs are summarized in Table 1 and Supplemental Results.

Table 1.

Clinical and demographic characteristics of MDD patients and healthy participants.

Characteristics	HC Male	HC Female	MDD Male	MDD Female	Diagnosis		Sex		Diagnosis × Sex
	(n = 106)	(n = 137)	(n = 55)	(n = 83)	F	p	F/t	p	F	p
Age (Years)	20.76 (2.92)	21.09 (3.32)	24.31 (4.76)	25.66 (7.79)	61.51	< 0.001	2.65	0.104	0.98	0.323
Education (Years)	14.33 (1.45)	14.63 (1.68)	14.38 (2.26)	13.95 (2.32)	2.39	0.123	0.10	0.753	3.15	0.077
Mean FD	0.12 (0.04)	0.12 (0.05)	0.11 (0.06)	0.13 (0.06)	0.19	0.660	6.61	0.011	1.94	0.164
Illness duration (Months)	-	-	12.66 (21.50)	9.35 (11.54)	-	-	1.03	0.307	-	-
BDI	5.6 (5.97)	5.65 (5.84)	26.89 (10.59)	30.80 (10.18)	754.98	< 0.001	5.47	0.020	5.23	0.023
HAM-D	_	_	19.96 (6.47)	22.69 (4.80)	_	_	−2.85	0.005	_	_

FD Framewise displacement, BDI Beck Depression Inventory, HAM-D Hamilton Depression Rating Scale, HC Healthy control, MDD Major depressive disorder.

fMRI acquisition

All scans were collected on the same 3 T Siemens Magnetom Skyra scanner (Siemens Healthineers, Erlangen, Germany) with a 16-channel head coil. Resting-state fMRI data were collected with an echo-planar imaging sequence and repetition time/echo time = 2000/30 ms, thickness/gap = 4/1 mm, field of view = 256 mm², flip angle = 80^°, matrix = 64 × 64, 32 slices. During the resting-state fMRI scan, participants were instructed to rest with their eyes closed without falling asleep. After finishing the scan, they were asked whether they stayed awake; only the data of subjects who stayed awake were considered for analysis. The total acquisition time for the resting-state fMRI data was 7 min 12 s (216 volumes).

fMRI preprocessing

Preprocessing was performed using fMRIPrep 1.5.8 [27], which is based on Nipype 1.4.1 [28]. We outline the main steps below (see Supplemental Methods for details). Following the removal of the first four volumes, realignment, slice-time correction, co-registration to the structural image and segmentation, blood oxygenation level-dependent (BOLD) time-series were resampled into MNI space, and spatially smoothed with a Gaussian kernel (6 mm full width at half-maximum). Independent component analyses (ICA-AROMA, [29]) were then conducted to identify and exclude motion artifacts. Lastly, the denoised BOLD time-series were high-pass filtered (f = 0.0067 Hz). Subjects were excluded if they had more than 20% of resting-state volumes with at least 0.5 mm framewise displacement (FD) and/or 1.5 standard deviation temporal derivative of timecourses of root mean square variance over voxels (DVARS).

Resting-state Co-activation Pattern (CAP) analysis

An ROI-wise seed-free whole-brain co-activation pattern analysis was conducted using custom scripts originating from the TbCAPs toolbox [26]. First, timecourses were extracted for each participant using 349 ROIs consisting of cortical (333) and subcortical (16) regions [30]. Second, consensus clustering was performed across the whole dataset to determine the optimal number of CAPs, which involves running k-means clustering over several folds on a randomly selected subsample of the data without replacement over the specified k range. A good clustering solution is one for which across folds, two volumes (one pair) are either always clustered together or never clustered together. The proportion of ambiguously clustered pairs (PAC) was used to quantify the quality of consensus clustering, with lower PAC values indicating more robust clustering. In the current study, k-means clustering was run over 50 folds on a randomly selected 80% subsample of the whole dataset for k = 2 to 20. The optimal PAC value was achieved for k = 10 (see Supplementary Fig. S1). Finally, k-means clustering using the cosine distance and random initialization of the algorithm (50 replicates) was run to partition all volumes into 10 CAPs.

Considering the hypothesis of the current study, four CAPs involving the DMN, FPN and SN (CAP₄, CAP₆, CAP₇, CAP₈) were included in the group analyses on CAP metrics (they are shown in Fig. 1). CAP₄ (DMN + SN-) involved activation of the DMN and deactivation of SN regions. CAP₆ (DMN-SN + ) included activation of SN regions and deactivation of DMN regions. CAP₇ (DMN + FPN-) involved activation of DMN regions and deactivation of FPN regions. Finally, CAP₈ (DMN-FPN + ) was characterized by activation of FPN regions and deactivation of anterior DMN regions. See Supplemental Information and Supplemental Fig. S2 for a detailed description of other CAPs. Two CAP metrics were calculated for each CAP of interest: time spent in CAP (total number of volumes spent in each CAP throughout the scan) and persistence (average total number of consecutive volumes participants remained in a given CAP). Transitions (total number of transitions from one CAP to another) were only calculated for CAPs exhibiting significant group effects. All CAP variables were inspected for normality or outliers (values outside the 1st quartile ± 3 × interquartile range), and any variables that violated assumptions were square root-transformed (see Supplementary Table S1 for the descriptive statistics of all CAPs).

Fig. 1. Co-activation patterns (CAPs) of interest.

Each CAP was characterized by the activation (see warm colors) and deactivation (cold colors) of brain regions. Z-scored CAPs are displayed at 0.5 ≤ |Z | ≤ 3.0. A CAP₄ involving activations of DMN regions and deactivations of salience network regions. B CAP₆ involving activations of salience network regions and deactivations of DMN regions. C CAP₇ involving activations of DMN regions and deactivations of FPN regions. D CAP₈ involving activations of FPN regions and deactivations of anterior DMN regions. DMN Default mode network, FPN Frontoparietal network, SN Salience network.

Group-level analysis

Group differences on time spent, persistence and number of transitions

With the aim to examine group differences in time spent and persistence in each hypothesized CAP, a series of Diagnosis (MDD/HC) × Sex (male/female) ANCOVAs were conducted with age as a covariate for each CAP metric (i.e., time spent, persistence) with only CAPs involving SN, DMN and FPN networks. False discovery rate (FDR) correction was applied for multiple ANCOVAs for each effect (i.e., Diagnosis, Sex, Diagnosis × Sex interaction) in each metric (i.e., time spent, persistence). For the CAPs showing group effects in terms of either time spent or persistence, numbers of transitions were calculated and compared across groups using Diagnosis (MDD/HC) × Sex (male/female) ANCOVAs controlling for age. Post-hoc simple effects analyses (with Bonferroni correction) were conducted for significant Diagnosis × Sex interaction effects.

Results

Group differences in time spent in CAP and CAP persistence

For the “time spent in CAP” metric, a significant Sex × Diagnosis interaction emerged in CAP₄ [DMN + SN-] (F_(1,376) = 8.96, p_FDR = 0.012, η² = 0.023) and CAP₆ [DMN-SN + ] (F_(1,376) = 5.54, p_FDR = 0.038, η² = 0.015). For CAP₄ [DMN + SN-], upon follow-up simple effect analyses, time spent was greater in the male MDD group compared to the male HC group (p_Bonferroni = 0.034; Fig. 2A); additionally, time spent was also greater in the female HC group relative to the male HC group (p_Bonferroni = 0.008; Fig. 2A). For CAP₆ [DMN-SN + ], follow-up simple effect analyses revealed greater time spent by the male MDD group in comparison to the male HC group (p_Bonferroni = 0.01; Fig. 2C). No other significant group effects emerged for time spent in the 4 CAPs of interest (all p_FDR > 0.05; Supplementary Table S2). In summary, the male MDD group spent more time in DMN + SN- and DMN-SN+ configurations relative to the male HC group, whereas female MDD patients did not exhibit these alterations.

Fig. 2. Group differences on time spent in CAPs and persistence.

A significant Diagnosis × Sex interaction effect emerged for the time spent (A) and persistence (B) of CAP₄ [DMN + SN-]. Follow-up analysis revealed a greater time spent in the CAP for the male MDD group compared to the male HC group, while the female MDD group showed lower persistence relative to the female HC group; in addition, time spent and persistence measures were also larger in the female HC group relative to the male HC group, and persistence was lower in the female MDD group relative to the male MDD group. C A significant Diagnosis × Sex interaction effect emerged for the time spent in CAP₆ [DMN-SN + ] involving activations of SN regions and deactivations of DMN regions. Follow-up analysis found that the male MDD group spent more time in this CAP relative to the male HC group, whereas the female MDD group did not show significant alterations. D The MDD group exhibited higher persistence in CAP₇ [DMN + FPN-] involving activations of DMN regions and deactivations of FPN regions in comparison to the HCs. CAP co-activation pattern, DMN Default mode network, FPN Frontoparietal network, SN Salience network, HCs Healthy controls, MDD Major depressive disorder. Estimated means are plotted, and error bars represent standard error (SE). *p < 0.05, **p < 0.01.

For the persistence metric, a significant Sex × Diagnosis interaction in CAP₄ [DMN + SN-] emerged (F_(1,376) = 9.21, p = 0.003, p_FDR = 0.012, η² = 0.024). Follow-up simple effect analyses revealed lower persistence in the female MDD group relative to the female HC group (p_Bonferroni = 0.029) and the male MDD group (p_Bonferroni = 0.039; Fig. 2B). In addition, the female HC group exhibited greater persistence in CAP₄ [DMN + SN-] in comparison to the male HC group (p_Bonferroni = 0.022; Fig. 2B). For CAP₇ [DMN + FPN-], a significant main effect of Diagnosis emerged (F_{(1, 375)} = 6.94, p = 0.009, p_FDR = 0.036, η² = 0.018), with the MDD group exhibiting greater persistence than the HC group (Fig. 2D); additionally, a significant main effect of Sex also emerged (F_{(1, 375)} = 6.77, p = 0.010, p_FDR = 0.040, η² = 0.018), with the male group showing greater persistence relative to the female group. No other significant effects emerged for the 4 CAPs of interest (all p_FDR > 0.05; Supplemental Table S3). In summary, the MDD group showed greater persistence in a DMN + FPN- configuration relative to HCs; at the same time, females showed less persistence than males in a DMN + SN- configuration when suffering from MDD, while the opposite was seen for healthy subjects.

All Diagnosis × Sex interaction findings remained significant in supplementary analyses controlling for BDI scores and mean FD (Supplementary Table S4). Supplementary analyses were also conducted to test group differences on other CAPs that were not part of our a priori hypotheses (i.e., CAP₁, CAP₂, CAP₃, CAP₅, CAP₉, CAP₁₀; see Supplementary Table S5). Significant Diagnosis × Sex interactions emerged for time spent and persistence in CAP₂ (activation of visual network); see Supplemental Results for more details.

Follow-up analyses on group differences in number of transitions between CAP₄, CAP₆, and CAP₇

Four Sex × Diagnosis ANCOVA analyses with age as a covariate were conducted to test group differences in the number of transitions between CAP₄ [DMN + SN-] and CAP₇ [DMN + FPN-] as well as between CAP₆ [DMN-SN + ] and CAP₇ [DMN + FPN-]. Since the mean number of transitions between CAP₄ [DMN + SN-] and CAP₆ [DMN-SN + ] is almost equal to zero (see Supplementary Fig. S3), group comparison was not performed. Significant interactions emerged for transitions from CAP₇ [DMN + FPN-] to CAP₄ [DMN + SN-] (F_(1,374) = 5.58, p = 0.017, p_FDR = 0.044, η² = 0.015; Supplementary Table S6) and from CAP₆ [DMN-SN + ] to CAP₇ [DMN + FPN-] (F_(1,373) = 5.27, p = 0.022, p_FDR = 0.044, η² = 0.014; Supplementary Table S6). In the former case, follow-up simple effect analyses revealed that the male MDD group exhibited a greater number of transitions in comparison to the male HC group (p_Bonferroni = 0.010; Fig. 3); the male MDD group also showed a greater number of transitions relative to the female MDD group (p_Bonferroni = 0.040; Fig. 3). In the latter case, none of the simple effect analysis findings survived multiple comparison correction (all p_Bonferroni > 0.05). No significant group effects on the number of transitions emerged for CAP₇ [DMN + FPN-] → CAP₆ [DMN-SN + ] and CAP₄ [DMN + SN-] → CAP₇ [DMN + FPN-] cases (i.e., opposite transitions; all p_FDR > 0.05). In summary, the male MDD group specifically transitions more frequently from a DMN + FPN- to a DMN + SN- configuration.

Fig. 3. Group differences on transition probability from CAP₇ to CAP₄.

A significant Diagnosis × Sex interaction emerged for the number of transitions from CAP₇ [DMN + FPN-] to CAP₄ [DMN + SN-]. Follow-up simple effect analyses revealed that the male MDD group exhibited a significantly higher number of transitions from CAP 7 to CAP 4 relative to the female MDD group and the male HC group, whereas the female MDD group did not exhibit significant alterations relative to HCs. DMN default mode network, FPN frontoparietal network, SN salience network. Estimated means are plotted, and error bars represent standard error (SE). *p < 0.05.

Associations between resting dynamic metrics and clinical measures

No significant associations were observed between resting dynamic metrics exhibiting significant group differences and clinical measures (all p_FDR > 0.05). See Supplemental Methods and Supplemental Table S7 for details.

Discussion

The overarching goal of the current study was to test the potential interaction between MDD and sex assigned at birth on resting-state brain network dynamics. Our findings revealed that MDD patients (irrespective of sex) spent more time in one CAP involving activation of the DMN and deactivation of the FPN (DMN + FPN-) when compared to HCs. In addition, we also unraveled potential sex-specific effects in CAPs involving the SN, DMN and FPN in MDD. Specifically, males with MDD spent more time in CAPs involving the SN and DMN (DMN + SN-/DMN-SN + ) relative to male HCs, whereas females with MDD exhibited greater persistence in a DMN + SN- CAP relative to female HCs; moreover, the male MDD group exhibited a greater number of transitions from the DMN + FPN- CAP to the DMN + SN- CAP relative to the male HCs and female MDD group. Collectively, these findings highlight the importance of considering sex as a variable when exploring the neural underpinnings of MDD and provide evidence of potential sex-specific brain network dynamic alterations in MDD.

The current findings revealed that MDD patients, irrespective of sex, exhibited greater persistence in a DMN + FPN- CAP compared to HCs [31]. Consistently, a CAP study also found that time spent in a DMN + FPN- CAP was positively associated with depressive symptom severity [32]. The DMN is related to self-referential mental activity, and the FPN is crucial for goal-related behavior and emotion regulation [10]. Thus, the longer persistence in a DMN + FPN- CAP observed in MDD could reflect a dysfunction of shifting the attention from self-referential thoughts to goal-related behavior. Relatedly, other studies have found that individuals with MDD spend more time and/or persist longer in mixed CAPs involving the DMN, including a frontoinsular-DMN CAP in a largely medicated adolescent MDD sample [24] and an FPN-PCC CAP (activation of PCC and FPN) in an unmedicated adult female sample [23]. Together, these findings suggest that DMN dominance and altered DMN-FPN interactions could be a core neurobiological feature of MDD.

Consistent with our hypothesis, more time spent in CAPs (DMN + SN-, DMN-SN + ) exhibiting an opposite polarity between the SN and DMN, and a higher number of transitions from the DMN + FPN- to the DMN + SN- CAP, were observed in males, whereas lower persistence in the DMN + SN- CAP was observed in females. These findings agree with prior literature showing weaker static FC between SN- and DMN-related regions in males with depressive symptoms and greater FC between SN- and DMN-related regions in female MDD patients [17, 18]. The current study further supports the sex-specific SN and DMN coupling alterations in MDD using a dynamic resting-state approach and provide new insights into the sex-specific SN and DMN coupling in MDD. To supplement these findings, we also quantified group differences in static FC between the SN and DMN (see more details in Supplementary Methods and Results; Supplementary Table S8); a trend of significant Diagnosis × Sex interaction was observed, which further supports CAP findings. Static FC averages together different contributions, which may make it less sensitive to detecting group differences in terms of between-network coordination relative to the dynamic FC approach.

Our findings revealed sex-specific functional coordination among SN- and DMN-related brain regions in MDD. The SN mainly plays a role in salience detection, and the DMN in self-referential processing [33]. For HCs, females spent more time and persisted more in the DMN + SN- CAP (and a similar trend was observed for the DMN-SN + CAP) relative to males. A prior study also found that females persisted longer in certain CAPs and switched less frequently, showing a less flexible functional substrate whereas the males exhibited higher dynamic fluidity [34]. Our findings complement the existing literature and highlight this pattern specifically for SN/DMN functional coordination. Less time in and persistence of CAPs exhibiting an opposite polarity between two networks relative to externally focused attention and self-focused processes in the male HCs could be beneficial to quick problem solving. In line with this, one prior study found that less persistence (higher dynamic fluidity) in CAPs was associated with males’ higher abilities in problem solving [34]. For females, higher persistence in a DMN + SN- configuration in the resting-state could be adaptive for reducing attention to external stimulus as well as external stimulus induced-rumination. Together, we could speculate that either the lower time spent/persistence in males or greater time spent/persistence in females is adaptive.

Following this reasoning, it is then reasonable why opposite patterns were observed in male and female MDD patients in CAPs exhibiting an opposite polarity between the SN and DMN: it is more likely a collapse of their original adaptive spontaneous functional substrate. Stress is a well-acknowledged casual factor of MDD [35]; overwhelming stress contributing to the onset of MDD could damage/reverse the resting functional substrates beneficial to stress adaptation in a sex-dependent way. However, these speculations need to be further verified. Noteworthily, our current findings reinforced the concern that similar neural substrates could contribute to distinct depressed conditions in males as opposed to females. For instance, in the current study, male MDD patients exhibited similar CAP patterns to female HCs although they also displayed significant differences in depressive severity. These findings highlight the importance of clarifying the sex-specific neural mechanisms of depression, which could provide guidance for precision treatment, especially for region-targeted neuromodulation therapy. Together, our results highlight that the SN and DMN interactions could be key neuroimaging markers for distinguishing male and female MDD patients and also point to putative neural targets to explore factors contributing to sex-specific effects in MDD.

Several limitations should be mentioned. First, our depressed sample consisted of first-episode medication-naive individuals with MDD and no comorbidities, which may make it less representative of the whole community. Replication of the current findings in a clinically more heterogeneous sample is needed. Second, lack of group matching for age required entering age as a covariate in the analyses. Third, the current female MDD group had more severe depressive symptoms than the male MDD group; nevertheless, the Sex × Diagnosis interaction was confirmed when controlling for depressive severity (see Supplementary Table S4 for details). Fourth, we analyzed effects of sex assigned at birth. However, sex effects are often distributional and overlap [36]. In addition, this study does not consider gender identity and associated environmental impact. Fifth, the relatively small number of per-subject volumes and the usage of a 16-channel head coil may affect the effectiveness of the CAP approach; further replication of the current findings is thus warranted. Sixth, the extraction of CAPs does not prove non-stationarity in the data, or the existence of a state-based data organization. Indeed, approximately similar CAPs and metrics of temporal dynamics can be obtained from a static null model with identical power distribution and covariance structure as the original data [37], CAPs should thus strictly be regarded as momentary co-activation patterns. Seventh, CAP analysis operates under the limiting assumptions of (1) a shared pool of CAPs across all subjects from the analyzed population, and (2) the expression of just one CAP at each moment in time. While similar assumptions are made in other popular dynamic fMRI analytical approaches [38, 39], there are also alternatives that work under different assumptions, such as through first-order autoregressive models [39] or the extraction of co-activation patterns that can overlap not only spatially, but also in temporal expression [40]. Here, we chose CAP analysis because (1) it enables a frame-wise analysis, (2) it shows direct links to the well acknowledged whole-brain and seed-based static FC approaches, and (3) it has been leveraged in several past reports investigating MDD [23, 41], easing cross-study interpretations.

Despite several limitations, to the best of our knowledge, the current study is the first to investigate the potential interaction between sex and MDD psychopathology in terms of resting-state brain dynamics. Both male and female MDD patients spent more time in a CAP involving the activation of DMN regions and deactivation of FPN regions during the resting-state, which provides evidence for a general (sex-nonspecific) neural abnormality related to depression. In addition, male and female MDD patients exhibited sex-specific neurobiological features in resting-state CAPs involving the SN and the DMN, highlighting the critical role of SN and DMN interactions in distinguishing between the male and female MDD patients and providing evidence for sex differences in the psychopathology of depression.

Author contributions

SY, ELB, DAP, and XW conceptualized the study; SY, XW, DD, XZ, CL, XS, GX, and CC collected the data; DD, ELB, TAWB, DAP, and MI analyzed the data and interpreted the data. DD drafted the manuscript with critical revisions from ELB, DAP, TAWB, and MI. All authors approved the final manuscript and are accounted for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding

This project was supported by the National Natural Science Foundation of China (82071532 to SY). Dr. Dong was funded by the Scientific Research Launch Project for new employees of the Second Xiangya Hospital of Central South University. Dr. Ironside has additional funding from the National Institute of Mental Health (NIMH; R01MH132565). Dr. Belleau is supported by funding from the National Institute of Mental Health (K23MH122668) and the Klingenstein Third Generation Foundation. The content is solely the responsibility of the authors. The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Competing interests

Over the past three years, Dr. Pizzagalli has received consulting fees from Albright Stonebridge Group, Boehringer Ingelheim, Compass Pathways, Engrail Therapeutics, Neumora Therapeutics (formerly BlackThorn Therapeutics), Neurocrine Biosciences, Neuroscience Software, Otsuka, Sunovion, and Takeda; he has received honoraria from the Psychonomic Society and American Psychological Association (for editorial work) and from Alkermes; he has received research funding from the Brain and Behavior Research Foundation, Dana Foundation, Millennium Pharmaceuticals, and Wellcome Leap; he has received stock options from Compass Pathways, Engrail Therapeutics, Neumora Therapeutics, and Neuroscience Software. There are no conflicts of interest with the work conducted in this study. No funding from these entities was used to support the current work, and all views expressed are solely those of the authors. The other authors report no financial relationships with commercial interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-024-01799-1.