Dynamic functional connectivity changes in the triple networks in patients with mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes

Introduction

Mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes (MELAS) is a heterogeneous mitochondrial disorder associated with mutations in mitochondrial or nuclear genes (1). Approximately 80% of MELAS patients carry m.3243A>G gene mutation in the MT-TL1, which encodes leucine-specific transfer RNA with anticodon UUR [tRNA^Leu(UUR)] (2). Epidemiologically, MELAS has been estimated to be approximately 2 to 20 per 100,000 individuals (3,4). This disease usually presents in childhood or early adulthood, with 65–76% of patients experiencing their initial onset before the age of 20 years (3). Stroke-like episode (SLE) is one of the core manifestations of MELAS-acute patients that usually present with headache, seizure, motor weakness, cortical blindness and cognitive impairment (1). SLEs are characterized by recurrent episodes followed by remissions, resulting in the gradual accumulation of neurological impairment in MELAS patients (5). As one of the main clinical symptoms of MELAS patients, cognitive impairment occurs in 40–90% of MELAS patients (3), however, the brain network mechanism of cognitive impairment is still unclear and further exploration is needed.

Resting-state functional magnetic resonance imaging (rs-fMRI) can quantify intrinsic functional brain organization by noninvasively measuring synchronizations between spontaneous activities in different, even non-spatially adjacent, brain regions (6). Rs-fMRI data have been widely utilized to investigate temporal properties of functional connectivity (FC) in brain networks across various diseases (7-10). Given the dynamic property of brain events, more research has found time-varying properties of FC, called temporal dynamic functional connectivity (dFC) (11,12). Recently, an increasing number of studies have focused on the dynamic alterations of FC within brain networks in healthy subjects and patients with various neurological diseases, such as major depressive disorder (MDD), MELAS, Parkinson’s disease (PD), and Alzheimer’s disease (AD) (13-17). Especially, Wang et al. (13) investigated dFC in MELAS using a whole-brain parcellation strategy to characterize global connectivity patterns and network topology. Previous global dFC study has provided a broad overview of network topology, while the present study focused on the triple network, comprising the default mode network (DMN), salience network (SN), and central executive network (CEN), identified via independent component analysis (ICA) and constrained by anatomical masks, and examined state-specific temporal metrics [dwell time, recurrence fraction (RF), number of transitions (NT)] to reveal stage-dependent alterations and offer distinct insights into triple network dysregulation in MELAS patients.

Menon proposed a highly influential “triple network model”, which provides a conceptual framework for understanding cognition through the interplay of core neurocognitive networks (18). This brain network model integrates DMN, SN and CEN into a unified model as the basis for understanding cognition and behavior (18). Specifically, the DMN is typically deactivated during most stimulus-driven cognitive tasks, and is primarily associated with self-referential processing, introspection, and internally-generated thought; in contrast to the DMN, the CEN is responsible for actively maintaining and manipulating information in working memory, problem-solving and decision-making; the SN plays a critical role in salience attribution, detecting and orienting to important external stimuli or internal events, thus mediating the interplay between introspective and executive functions. Given that MELAS frequently presents with impairments in execution, attention, and cognition (3), alterations of dFC across these core networks may underlie its cognitive phenotype. Previous various studies have indicated alterations of dFC based on the triple network, e.g., cerebral small vessel disease (19), hemodialysis patients (HD) (20), MDD (21), and AD (22). These findings have yielded new insights into the neural network mechanisms underlying cognitive impairment in these patients. To our knowledge, no prior studies have investigated alterations of dFC within the triple network in MELAS patients.

Given the importance of the triple network in cognition, this study hypothesized that MELAS patients had altered temporal properties of dFC in the triple network, which might be key to understand their cognitive impairment. To test our hypotheses, we employed a combination of group spatial ICA, k-means clustering algorithm and sliding window approach (SWA) to investigate the differences of temporal properties of dFC within the triple network among MELAS-acute patients, MELAS-chronic patients and healthy control volunteers (HCs). Furthermore, we explored potential correlations between temporal properties of dFC states and stroke-like lesions (SLLs) volumes. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-807/rc).

Methods

Participants

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Huashan Hospital (No. KY2019-432) and informed consent was obtained from all individual participants. This cross-sectional study analyzed data from a single-center, ongoing, prospective cohort study of genetically confirmed MELAS patients. Diagnosis for MELAS patients was genetically confirmed by detecting the m.3243A>G mutation in MT-TL1 via gene testing (4). This diagnosis was further validated by a professor (J.L.) with approximately 20 years of experience in MELAS. Concretely, MELAS patients who carried m.3243A>G gene mutation were included. Thirty-five MELAS-acute patients (MELAS-acute group, who were included within 1 week after SLE or seizure) and 34 MELAS-chronic patients (MELAS-chronic group, who were included about 6 months after SLE) were registered from June 2019 to November 2023. These two MELAS subgroups were established as independent cohorts. Moreover, 30 healthy control volunteers (HC group), matched for age and gender, were recruited from the community. The exclusion criteria were: (I) psychiatric or neurodegenerative diseases (e.g., autism, MDD, bipolar disorder, PD, etc.); (II) presence of other organic brain lesions; (III) history of head trauma; (IV) alcohol addiction; (V) inability to complete the MRI examination.

MRI data

MR imaging was acquired using a 3.0T GE scanner with an 8-channel head coil (Discovery MR750, General Electric, Boston, MA, USA). Rs-fMRI was performed within one week of neurologic deficits or seizure. All participants were instructed to keep their eyes closed during scanning but to remain relaxed and avoid falling asleep. Firstly, to exclude possible organic brain lesions, we obtained T1-weighted, T2-weighted, and T2-weighted fluid-attenuated inversion recovery (FLAIR) imaging sequences. The rs-fMRI data were acquired using a single-shot gradient-recalled echo planar imaging (EPI) sequence with the following parameters: echo time (TE) =30 ms, repetition time (TR) =2,000 ms, flip angle =90°, slice thickness =4 mm, slices =35, matrix size =64×64, field of view (FOV) =240 mm × 240 mm, number of volumes =210. Using a brain volume (BRAVO) sequence, we obtained high-resolution three-dimensional T1-weighted anatomical images with the parameters: TE =3.2 ms, TR =8.2 ms, flip angle =12°, slice thickness =1.2 mm, slices =170, FOV =240 mm × 240 mm, matrix size =256×256.

MRI data preprocessing

The rs-fMRI data were preprocessed by using DPABI toolbox (http://rfmri.org/dpabi) (23) and statistical parametric mapping (SPM12) (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) within MATLAB (version R2023b; The MathWorks, Natick, MA, USA, https://www.mathworks.com/). Specifically, we deleted the first 10 volumes, applied slice-timing correction, and corrected for head motion via realignment. Then, we calculated mean frame-wise displacement (FD) proposed by Power et al. (24). Hence, five MELAS-acute patients and four MELAS-chronic patients were excluded with mean FD >0.25 mm, rotation >2.5° or translation >2.5 mm. The final sample included 90 subjects (30 MELAS-acute patients, 30 MELAS-chronic patients, 30 HCs). Using the DARTEL algorithm (25), the preprocessed MRI data were normalized to the Montreal Neurological Institute (MNI) space, and then were resampled to 3×3×3 mm³ voxel size. Finally, using a Gaussian kernel, normalized imaging data were smoothed with a 6-mm full-width.

Group ICA and identification of intrinsic networks

Group ICA was performed using the GIFT toolbox (Group ICA of fMRI Toolbox, version 4.0b; http://mialab.mrn.org/software/gift). Preprocessed rs-fMRI data were decomposed into different independent components (ICs). The ICA included reduction of MR images, decomposition of ICs, and reconstruction. Firstly, for the group-level ICA, we applied a two-step principal component analysis (PCA) to reduce the dimensionality of the spatially normalized T1W images. Subsequently, we utilized the minimum description length (MDL) criterion to objectively determine the optimal number of ICs for decomposition (26). The MDL criterion, a statistically principled method for model order selection, indicated that 36 components provided the best balance between model complexity and data fitting. This number was selected to capture a sufficient number of distinct spatially independent networks while avoiding an overly fragmented representation of the data. Secondly, ICs estimation was implemented by applying the Infomax algorithm (27). Using the ICASSO (http://www.cis.hut.fi/projects/ica/icasso) (28), the reliability and stability of the decomposition were assessed by repeating 100 times. In this step, blood oxygenation level-dependent (BOLD) signals were used to generate time courses and spatial maps for each IC. Finally, using the back reconstruction algorithm in the group ICA, subject-specific time courses and spatial maps were generated for each IC, and subsequently transformed into z-scores (29).

On the basis of the criteria of previous studies (11,30), all ICs were assessed based on the group IC maps: (I) peak activations of spatial maps were located in gray matter; (II) low spatial overlap with motion, susceptibility artifacts, vascular, and ventricular; (III) time courses were created by low-frequency signals and high dynamic range. Based on the GIFT with spatial sorting function, we selected five ICs corresponding to posterior DMN (pDMN), anterior DMN (aDMN), right CEN (rCEN), left CEN (lCEN), and SN (31).

Finally, we performed the following post-processing steps on the time courses of 36 ICs to remove noise: (I) removing linear trends; (II) multiple regression, including temporal derivatives and the six realignment parameters; (III) removing outliers; (IV) low-pass filtering with a 0.15 Hz cut-off.

Assessment of dFC

Dynamic functional connectivity was estimated using the Temporal dFNC module within the GIFT toolbox. A SWA was used for the estimation of dFC. In each window, Pearson correlation coefficients among time courses were computed, resulting in a 5×5 correlation matrix.

Following previous studies (7,20,31), resting-state time series were segmented into 30-TR windows (60 s) convolved with a Gaussian kernel (σ=3 TR) in this study. This duration could offer a good balance between resolving network dynamics and ensuring stable FC estimates. Validation studies [e.g., Shirer et al. (31)] indicated that cognitive states can be classified with higher accuracy from 30–60 s data. We used a tapered window, generated by convolving a rectangle (width=30 TR) with a Gaussian (σ=3 TR) and the beginning of each window was slid by 1 TR relative to the previous one, generating 140 consecutive windows in total.

We used the inverse covariance matrix or regularized precision matrix to assess covariance matrix (32). To improve sparsity in estimation, an additional L1 norm with 10 repetitions was constraint in graphic least absolute shrinkage and selection operator (LASSO) framework (33). Using Fisher’s r-to-z switch, results of FC matrices were transformed to z-scores. To reduce the impact of covariates, we applied multiple linear regression to residualize the z-scores with gender and age.

dFC states analysis

We applied the k-means clustering algorithm to capture FC states by clustering all 5×5 FC matrices from the windows of all participants. To increase the probability of escaping local minima, the k-means clustering algorithm was repeated 500 times with a random initialization of the centroid positions. Based on the elbow criterion, the optimal number of clusters was identified as four (k=4) (11,34). These centroids were applied to cluster windows of dFC for all the participants, and we acquired a vector of state transition on behalf of state status across time. Finally, four states were clustered from time windows of all participants.

We obtained temporal properties of dFC from state vector of each participant, including mean dwell time (MDT), RF, and NT. MDT is the average time of continuous windows assigned to each state. RF is proportion of time spent for each state. NT characterizes times transformed from one state to the other (9).

To calculate the centroid for subject-specific, we applied median z-score of FC matrix. Then, we obtained the centroids for group-specific from averaged centroids of subject-specific to confirm differences in FC strength between-groups.

Volume of SLL

SLL is defined as acute focal neurological deficits accompanied by fresh lesions on imaging. These lesions, observed on T2-FLAIR sequences, typically manifest as gyral swelling in the cerebral cortex and subcortical white matter hyperintensity (35). By applying medical imaging interaction toolkit (MITK) (36), the SLLs of MELAS-acute patients were recognized and segmented slice-by-slice on T2-FLAIR by a radiologist with 5 years’ experience in imaging of MELAS (R.W.), which were reviewed by a specialized radiologist in MELAS with more than 10 years’ experience (Y.L.). Finally, SLL volumes were acquired using MATLAB software.

Statistical analysis

One-way analysis of variance (ANOVA) test was applied to analyze group differences for age. A chi-square test was used to analyze gender-ratio differences between the groups. An independent two-sample Student’s t-test was applied to compare group differences of disease onset and duration. For temporal properties of dFC states and FC strength of each state, the Kruskal-Wallis test was performed to assess group effects between the MELAS-acute group, MELAS-chronic group and HC group, followed by Dunn’s test for post-hoc multiple comparisons. The Pearson correlation coefficient was computed to evaluate the relationship between dFC properties and SLL volumes, with P values subsequently corrected for multiple comparisons using the Benjamini-Hochberg (BH) method to control the false discovery rate (FDR).

Statistical analysis was performed using GraphPad Prism (version 9.1.0 for Windows, San Diego, CA, USA) and MATLAB software. An adjusted P value of less than 0.05 was considered statistically significant.

Results

Demographic and clinical characteristics

Table 1 summarizes the demographic and clinical features. There were no significant group differences in age (MELAS-acute: 24.7±7.6; MELAS-chronic: 25.8±7.5; HC: 26.4±6.2; P=0.677), gender (P=0.838), age of disease onset (MELAS-acute: 22.3±5.8; MELAS-chronic: 23.7±6.2; P=0.361), or disease duration (MELAS-acute: 2.5±2.2; MELAS-chronic: 2.1±1.5; P=0.425), defined as the time from initial diagnosis to the date of MRI acquisition. MELAS-acute patients presented with SLE symptoms, including headache, seizure, motor weakness, vomiting, hearing loss, cortical blindness, and aphasia.

Intrinsic FC networks

As displayed in Figure 1, the five ICs, aDMN (IC19), pDMN (IC10), rCEN (IC28), lCEN (IC30), and SN (IC17), were selected from the 36 ICs. The spatial maps resembled previous studies (20,37).

Figure 1 Spatial maps of the five ICs. aDMN, anterior default mode network; ICs, independent components; lCEN, left central executive network; pDMN, posterior default mode network; rCEN, right central executive network; SN, salience network.

Dynamic FC states analysis

Four recurring FC states were identified across all participants using k-means clustering (Figure 2). State 1 (occurring in 18% of windows) was characterized by positive FC between the DMN and CEN, and exhibited stronger positive FC between the SN and both the DMN and CEN compared with other identified states. State 2 (observed in 20% of windows) was characterized by the strongest positive FC between the DMN and CEN. Crucially, in this state, both the DMN and CEN exhibited negative FC with the SN. State 3 (accounting for 19% of windows) demonstrated positive FC between the DMN and lCEN and weaker negative FC between SN and both the DMN and CEN compared with state 2. State 4 (accounting for 42% of windows) exhibited the weakest overall FC across the analyzed networks, with notable differences in the patterns compared with the other states.

Figure 2 Cluster centroid and number of occurrence (%) for each state. The color bar indicates the z-score of FC. aDMN, anterior default mode network; FC, functional connectivity; ICs, independent components; lCEN, left central executive network; pDMN, posterior default mode network; rCEN, right central executive network; SN, salience network.

Figure 3 demonstrates the group-specific median z-score matrices for each identified FC state. Significantly, not all participants allocated the windows to each state, which resulted in altered number of subject-specific matrices for each state (see Figure 3). There were no significant between-groups differences in FC strength in each state.

Figure 3 Group-specific centroid matrices for each state (A-C). The color bar indicates the z-score of FC. aDMN, anterior default mode network; FC, functional connectivity; HC, healthy control; ICs, independent components; lCEN, left central executive network; pDMN, posterior default mode network; rCEN, right central executive network; SN, salience network.

Table 2 and Figure 4 illustrate temporal properties of FC states and between-groups differences. MELAS-acute group had strikingly lower RF in state 1 compared with MELAS-chronic group (P=0.0229, Dunn’s correction) and HC group (P=0.0093, Dunn’s correction), and shorter MDT compared with MELAS-chronic group (P=0.0414, Dunn’s correction). And MELAS-acute group had higher RF (P=0.008, Dunn’s correction) and longer MDT (P=0.0162, Dunn’s correction) in state 3 than the MELAS-chronic group. MELAS-chronic group had lower RF (P=0.0141, Dunn’s correction) and shorter MDT (P=0.0137, Dunn’s correction) in state 3 but higher RF (P=0.0499, Dunn’s correction) in state 4 compared with HC group. In addition, the MELAS-chronic group had significantly lower NT than the HC group (P=0.0347, Dunn’s correction).

Figure 4 Between-group differences for temporal properties of FC (A-C). Asterisks show the significance of between-group difference (*, P<0.05; ** P<0.01; Dunn’s correction). FC, functional connectivity; HC, healthy control; MELAS, mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes; MELAS-acute, MELAS patients at the acute stage; MELAS-chronic, MELAS patients at the chronic stage.

Correlations between acute SLL volumes and temporal properties of dFC states

As shown in Table 3, a comprehensive analysis of correlations between acute SLL volumes and temporal properties of dFC states was conducted. Following BH correction for multiple comparisons, no significant correlations were found between SLL volumes and temporal properties of dFC states.

Discussion

In this study, we investigated the temporal properties of dFC states within the triple network (DMN, CEN, SN) in MELAS patients. Our findings revealed distinct FC states, characterized by varying patterns of inter-network coupling. Notably, MELAS patients exhibited a significant redistribution of time spent across these states compared with HCs, spending more time in the weaker connectivity states, and had lower NT, especially MELAS-chronic patients.

State 4, characterized by the weakest FC across all networks, was the most prevalent state in all participants. The weaker FC in a prevalent state resembles the stationary FC, likely signifies the average of a large number of additional states that are not sufficiently distinct or frequent to be separated (11). FC patterns in states 1–3 were observed less frequently (approximately 18–20%), but represented connectivity diverging substantially from the mean. Furthermore, among the less frequent but more divergent from the mean connectivity states (states 1–3), state 1 demonstrated stronger positive DMN-CEN coupling, and stronger positive coupling between the SN and both the DMN and CEN. This enhanced SN-DMN/CEN integration is considered a feature of state 1. The SN’s role in detecting and orienting attention to salient internal and external stimuli, along with its function as a switch between the DMN and CEN (18,38), suggests that in state 1, there is a more effective integration of salience detection with both internally focused cognition and executive control processes. State 2 exhibited the strongest positive coupling between the DMN and CEN. This heightened co-activation is consistently associated with internally-directed thought, self-referential processing, and executive control, tasks that necessitate significant integration between these large-scale networks (18). The pronounced DMN-CEN coupling in state 2 indicates an evident engagement of internally focused cognition and executive functions. Concurrently, this state characteristically presented with the negative SN-DMN/CEN integration, implying a reduced responsiveness to external stimuli during periods of intense introspection or execution. State 3 was characterized by a weaker positive DMN-lCEN coupling and weaker negative SN-DMN/CEN integration compared with state 2. This state might represent an intermediate level of network coordination, where neither strong integration nor broad segregation is evident.

Our dFC states analysis revealed significant differences in the temporal dynamics of network engagement between MELAS patients and HCs. The HC group showed a relatively balanced distribution across all states, characterized by a fairly even frequency of transitions between states. In contrast, MELAS patients exhibited a significant redistribution across these states, spending more time in the weaker connectivity states. Specifically, MELAS patients spent a disproportionately longer duration in state 4, characterized by the weakest FC across all three core networks (DMN, CEN, SN). Consistent with findings in other neurodegenerative conditions, such as AD (10), bipolar disorder and MDD (39), HD (20), this observation indicates that spending more time in a more sparsely connected state, and consequently a reduced capacity to maintain dynamic and integrated neural communication, is a common feature. A deeper analysis of group-specific temporal properties and the distinct network coupling patterns of each state revealed critical distinctions. MELAS-acute group displayed a significantly lower RF and shorter MDT in state 1. And that state 1 represents a state of more effective integration of salience detection with internally focused cognition and execution. This reduced occurrence suggests a diminished capacity to enter or sustain this highly integrated state, particularly at the acute phase. Conversely, the MELAS-acute group exhibited a higher RF and longer MDT in state 3, representing an intermediate level of network coordination. This indicates that MELAS-acute patients are more prone to occupying this less integrated state for extended periods, which may reveal a state of network destabilization and struggle to achieve either strong integration or efficient segregation. Furthermore, the MELAS-chronic group exhibited lower RF and shorter MDT in state 3 but a higher RF in state 4 compared with HC group, suggesting that MELAS-chronic patients are more likely to spend more time in the globally sparsely connected state 4, potentially representing a more established, less dynamically responsive, state of network dysfunction. The observed alterations of temporal properties in MELAS patients, particularly the prolonged engagement in sparsely connected states and reduced presence in highly integrated states, strongly suggest a compromised ability for sustaining efficient inter-network communication and dynamic reconfigurations. This deficit is likely underpinned by the mitochondrial dysfunction inherent in MELAS patients. The critical role of adenosine triphosphate (ATP) as the primary energy currency for neuronal function means that mitochondrial dysfunction, leading to inadequate ATP availability (40), would directly impede the energy-intensive processes required for dFC. We hypothesize that the brain’s struggle to maintain efficient connectivity among the DMN, CEN, and SN, particularly after a salient or stressful stimulus, leads to a gradual weakening of inter-network coupling. This metabolic deficit may manifest as a progressive descent towards less coupled states, with the most energy-demanding states being less accessible. In conclusion, the dFC states analysis highlights not only altered patterns of network coupling but also fundamental differences in how MELAS patients engage with and transition between these states over time. These temporal disruptions, particularly the reduced access to highly integrated states and prolonged occupation of sparsely connected ones, coupled with the underlying mitochondrial dysfunction, provide novel insights into the neurophysiological basis of cognitive and functional impairments in MELAS.

Furthermore, the observed reduction in the NT between dFC states in MELAS patients, particularly at the chronic stage, is a key finding suggesting impaired network flexibility. Dynamic transitions across states are essential for adapting to changing cognitive demands, recruiting optimal network configurations, and maintaining efficient information processing in healthy individuals (11). This dynamic re-configuration reflects the concept of metastability, referring to the brain’s intrinsic ability to sustain overall functional integrity while enabling transient, localized fluctuations and adaptive reorganization across its complex networks. A reduced NT implies a diminished capacity for this dynamic adaptation. MELAS patients may be ‘locked’ in certain states or transition less readily, hindering their ability to efficiently reconfigure neural resources for effective information transfer and problem-solving capabilities.

In addition, Pearson’s correlation analysis demonstrated, in MELAS-acute patients, no significant correlations were found between SLL volumes and temporal properties of dFC states after BH correction for multiple comparisons. While rigorous statistical analysis showed no significant correlations were found, the uncorrected correlation (r=−0.375) with state 3 (characterized by a positive DMN-lCEN coupling and weaker negative SN-DMN/CEN integration) temporal dynamics warrants further discussion. It also aligns with the broader understanding that structural integrity influences large-scale brain network dynamics, which in turn support cognition. This finding, even at a trend level, might provide a valuable hypothesis for future research with larger sample sizes or optimized power.

There are some limitations in this study. First, the relatively small sample size in this study could have impacted statistical power, potentially limiting the significance of the findings. Future study, ideally a longitudinal cohort or multi-center collaborations, will be crucial to replicate and extend our findings, allowing for the detection of more subtle effects and improving the generalizability of the results. Second, MELAS patients commonly have cognitive impairment, and it is widely known that “triple network model” plays a key part in cognitive assessment. Therefore, further study will establish a longitudinal cohort study and conduct comprehensive cognitive assessment on all subjects to better explain the contribution of abnormal FC based on the triple network to cognitive impairment in MELAS patients and to directly correlate dFC alterations with cognitive outcomes. Furthermore, these longitudinal data will enable us to investigate whether alterations in dFC can predict declines in cognition over time. An ongoing longitudinal cohort will acquire MMSE and MoCA at baseline and follow-up. Third, the selected window length (30 TRs, 60 s) represents a compromise between temporal resolution and estimation reliability. Future studies could test classification performance with shorter durations (e.g., 30 s) to evaluate sensitivity to rapid network changes. Additionally, we should acknowledge the effects of related-medications on resting state data, which might impact the significance of the observations. In future studies, it would be important to dissociate MELAS-related abnormality from medication-related effects on dFC properties by comparing with drug-naïve patients.

Conclusions

In summary, this study highlighted the temporal properties of dFC states within the triple network in MELAS patients. Our results revealed distinct FC states, characterized by varying patterns of inter-network coupling. Especially, MELAS patients exhibited a significant redistribution of time spent across these states compared with HCs, spending more time in the weaker connectivity states, and had lower NT. Our findings indicated the decreased ability of information transfer and processing, which might provide us novel insights to elucidate neural network mechanisms of cognitive impairment in MELAS patients.

Acknowledgments

We sincerely thank all patients with MELAS and their families, as well as the healthy volunteers who agreed to participate in this study.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-807/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-807/dss

Funding: This work was supported by National Nature Science Foundation of China (No. 82372048); Shanghai Municipal Commission of Science and Technology (Nos. 22TS1400900, 22ZR1409500, 23S31904100, 24SF1904200, and 24SF1904201); and Huashan Hospital Foundation (Basic Research Youth Cultivation Program) (No. 2024JC018).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-807/coif). All authors report that this work was supported by National Nature Science Foundation of China (No. 82372048); Shanghai Municipal Commission of Science and Technology (Nos. 22TS1400900, 22ZR1409500, 23S31904100, 24SF1904200, and 24SF1904201); and Huashan Hospital Foundation (Basic Research Youth Cultivation Program) (No. 2024JC018). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable 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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Huashan Hospital (No. KY2019-432) and informed consent was obtained from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Gorman GS, Chinnery PF, DiMauro S, Hirano M, Koga Y, McFarland R, Suomalainen A, Thorburn DR, Zeviani M, Turnbull DM. Mitochondrial diseases. Nat Rev Dis Primers 2016;2:16080. [Crossref] [PubMed]
2. Song M, Ye L, Yan Y, Li X, Han X, Hu S, Yu M. Mitochondrial diseases and mtDNA editing. Genes Dis 2024;11:101057. [Crossref] [PubMed]
3. Fan HC, Lee HF, Yue CT, Chi CS. Clinical Characteristics of Mitochondrial Encephalomyopathy, Lactic Acidosis, and Stroke-Like Episodes. Life (Basel) 2021;11:1111. [Crossref] [PubMed]
4. Yatsuga S, Povalko N, Nishioka J, Katayama K, Kakimoto N, Matsuishi T, Kakuma T, Koga YTaro Matsuoka for MELAS Study Group in Japan. MELAS: a nationwide prospective cohort study of 96 patients in Japan. Biochim Biophys Acta 2012;1820:619-24. [Crossref] [PubMed]
5. Ng YS, Bindoff LA, Gorman GS, Klopstock T, Kornblum C, Mancuso M, McFarland R, Sue CM, Suomalainen A, Taylor RW, Thorburn DR, Turnbull DM. Mitochondrial disease in adults: recent advances and future promise. Lancet Neurol 2021;20:573-84. [Crossref] [PubMed]
6. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med 1995;34:537-41. [Crossref] [PubMed]
7. Yu Q, Erhardt EB, Sui J, Du Y, He H, Hjelm D, Cetin MS, Rachakonda S, Miller RL, Pearlson G, Calhoun VD. Assessing dynamic brain graphs of time-varying connectivity in fMRI data: application to healthy controls and patients with schizophrenia. Neuroimage 2015;107:345-55. [Crossref] [PubMed]
8. Liao W, Zhang Z, Mantini D, Xu Q, Ji GJ, Zhang H, Wang J, Wang Z, Chen G, Tian L, Jiao Q, Zang YF, Lu G. Dynamical intrinsic functional architecture of the brain during absence seizures. Brain Struct Funct 2014;219:2001-15. [Crossref] [PubMed]
9. Kim J, Criaud M, Cho SS, Díez-Cirarda M, Mihaescu A, Coakeley S, Ghadery C, Valli M, Jacobs MF, Houle S, Strafella AP. Abnormal intrinsic brain functional network dynamics in Parkinson's disease. Brain 2017;140:2955-67. [Crossref] [PubMed]
10. Schumacher J, Peraza LR, Firbank M, Thomas AJ, Kaiser M, Gallagher P, O'Brien JT, Blamire AM, Taylor JP. Dynamic functional connectivity changes in dementia with Lewy bodies and Alzheimer's disease. Neuroimage Clin 2019;22:101812. [Crossref] [PubMed]
11. Allen EA, Damaraju E, Plis SM, Erhardt EB, Eichele T, Calhoun VD. Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex 2014;24:663-76. [Crossref] [PubMed]
12. Choe AS, Nebel MB, Barber AD, Cohen JR, Xu Y, Pekar JJ, Caffo B, Lindquist MA. Comparing test-retest reliability of dynamic functional connectivity methods. Neuroimage 2017;158:155-75. [Crossref] [PubMed]
13. Wang R, Sun C, Lin J, Chen N, Hu B, Liu X, Geng D, Yang L, Li Y. Altered Dynamic Functional Connectivity in Patients With Mitochondrial Encephalomyopathy With Lactic Acidosis and Stroke-Like Episodes (MELAS) at Acute and Chronic Stages: Shared and Specific Brain Connectivity Abnormalities. J Magn Reson Imaging 2021;53:427-36. [Crossref] [PubMed]
14. Foo H, Mather KA, Jiang J, Thalamuthu A, Wen W, Sachdev PS. Genetic influence on ageing-related changes in resting-state brain functional networks in healthy adults: A systematic review. Neurosci Biobehav Rev 2020;113:98-110. [Crossref] [PubMed]
15. Yang Y, Ye H, Yan H, Zhang C, Li W, Li Z, Jing H, Li X, Liang J, Xie G, Liang W, Ou Y, Li X, Guo W. Potential correlations between asymmetric disruption of functional connectivity and metabolism in major depressive disorder. Brain Res 2024;1838:148977. [Crossref] [PubMed]
16. Holtbernd F, Hohenfeld C, Oertel WH, Knake S, Sittig E, Romanzetti S, Heidbreder A, Michels J, Dogan I, Schulz JB, Schiefer J, Janzen A, Reetz K. The functional brain connectome in isolated rapid eye movement sleep behavior disorder and Parkinson's disease. Sleep Med 2024;117:184-91. [Crossref] [PubMed]
17. García-Colomo A, Nebreda A, Carrasco-Gómez M, de Frutos-Lucas J, Ramirez-Toraño F, Spuch C, Comis-Tuche M, Bruña R, Alfonsín S, Maestú F. Longitudinal changes in the functional connectivity of individuals at risk of Alzheimer's disease. Geroscience 2024;46:2989-3003. [Crossref] [PubMed]
18. Menon V. Large-scale brain networks and psychopathology: a unifying triple network model. Trends Cogn Sci 2011;15:483-506. [Crossref] [PubMed]
19. Wei HL, Wei C, Yu YS, Yu X, Chen Y, Li J, Zhang H, Chen X. Dysfunction of the triple-network model is associated with cognitive impairment in patients with cerebral small vessel disease. Heliyon 2024;10:e24701. [Crossref] [PubMed]
20. Cao J, Liu G, Li X, Yue Z, Ren J, Zhu W, Wu B. Dynamic functional connectivity changes in the triple networks and its association with cognitive impairment in hemodialysis patients. Brain Behav 2021;11:e2314. [Crossref] [PubMed]
21. Beckmann FE, Gruber H, Seidenbecher S, Schirmer ST, Metzger CD, Tozzi L, Frodl T. Specific alterations of resting-state functional connectivity in the triple network related to comorbid anxiety in major depressive disorder. Eur J Neurosci 2024;59:1819-32. [Crossref] [PubMed]
22. Feng Q, Wang L, Tang X, Hu H, Ge X, Liao Z, Ding Z. Static and dynamic functional connectivity combined with the triple network model in amnestic mild cognitive impairment and Alzheimer's disease. Front Neurol 2023;14:1284227. [Crossref] [PubMed]
23. Yan CG, Wang XD, Zuo XN, Zang YF. DPABI: Data Processing & Analysis for (Resting-State) Brain Imaging. Neuroinformatics 2016;14:339-51. [Crossref] [PubMed]
24. Power JD, Mitra A, Laumann TO, Snyder AZ, Schlaggar BL, Petersen SE. Methods to detect, characterize, and remove motion artifact in resting state fMRI. Neuroimage 2014;84:320-41. [Crossref] [PubMed]
25. Ashburner J. A fast diffeomorphic image registration algorithm. Neuroimage 2007;38:95-113. [Crossref] [PubMed]
26. Li YO, Adali T, Calhoun VD. Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp 2007;28:1251-66. [Crossref] [PubMed]
27. Bell AJ, Sejnowski TJ. An information-maximization approach to blind separation and blind deconvolution. Neural Comput 1995;7:1129-59. [Crossref] [PubMed]
28. Himberg J, Hyvärinen A, Esposito F. Validating the independent components of neuroimaging time series via clustering and visualization. Neuroimage 2004;22:1214-22. [Crossref] [PubMed]
29. Calhoun VD, Adali T, Pearlson GD, Pekar JJ. A method for making group inferences from functional MRI data using independent component analysis. Hum Brain Mapp 2001;14:140-51. [Crossref] [PubMed]
30. Cordes D, Haughton VM, Arfanakis K, Wendt GJ, Turski PA, Moritz CH, Quigley MA, Meyerand ME. Mapping functionally related regions of brain with functional connectivity MR imaging. AJNR Am J Neuroradiol 2000;21:1636-44.
31. Shirer WR, Ryali S, Rykhlevskaia E, Menon V, Greicius MD. Decoding subject-driven cognitive states with whole-brain connectivity patterns. Cereb Cortex 2012;22:158-65. [Crossref] [PubMed]
32. Smith SM, Miller KL, Salimi-Khorshidi G, Webster M, Beckmann CF, Nichols TE, Ramsey JD, Woolrich MW. Network modelling methods for FMRI. Neuroimage 2011;54:875-91. [Crossref] [PubMed]
33. Friedman J, Hastie T, Tibshirani R. Sparse inverse covariance estimation with the graphical lasso. Biostatistics 2008;9:432-41. [Crossref] [PubMed]
34. Rashid B, Blanken LME, Muetzel RL, Miller R, Damaraju E, Arbabshirani MR, Erhardt EB, Verhulst FC, van der Lugt A, Jaddoe VWV, Tiemeier H, White T, Calhoun V. Connectivity dynamics in typical development and its relationship to autistic traits and autism spectrum disorder. Hum Brain Mapp 2018;39:3127-42. [Crossref] [PubMed]
35. Bhatia KD, Krishnan P, Kortman H, Klostranec J, Krings T. Acute Cortical Lesions in MELAS Syndrome: Anatomic Distribution, Symmetry, and Evolution. AJNR Am J Neuroradiol 2020;41:167-73. [Crossref] [PubMed]
36. Nolden M, Zelzer S, Seitel A, Wald D, Müller M, Franz AM, Maleike D, Fangerau M, Baumhauer M, Maier-Hein L, Maier-Hein KH, Meinzer HP, Wolf I. The Medical Imaging Interaction Toolkit: challenges and advances : 10 years of open-source development. Int J Comput Assist Radiol Surg 2013;8:607-20. [Crossref] [PubMed]
37. Liu J, Li X, Xue K, Chen Y, Wang K, Niu Q, Li Y, Zhang Y, Cheng J. Abnormal dynamics of functional connectivity in first-episode and treatment-naive patients with obsessive-compulsive disorder. Psychiatry Clin Neurosci 2021;75:14-22. [Crossref] [PubMed]
38. Seeley WW, Menon V, Schatzberg AF, Keller J, Glover GH, Kenna H, Reiss AL, Greicius MD. Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci 2007;27:2349-56. [Crossref] [PubMed]
39. Wang J, Wang Y, Huang H, Jia Y, Zheng S, Zhong S, Chen G, Huang L, Huang R. Abnormal dynamic functional network connectivity in unmedicated bipolar and major depressive disorders based on the triple-network model. Psychol Med 2020;50:465-74. [Crossref] [PubMed]
40. Tsujikawa T, Yoneda M, Shimizu Y, Uematsu H, Toyooka M, Ikawa M, Kudo T, Okazawa H, Kuriyama M, Kimura H. Pathophysiologic evaluation of MELAS strokes by serially quantified MRS and CASL perfusion images. Brain Dev 2010;32:143-9. [Crossref] [PubMed]