Functional connectivity abnormalities underlying mood disturbances in male abstinent methamphetamine abusers

Abstract Anxiety and depression are the most common withdrawal symptoms of methamphetamine (METH) abuse, which further exacerbate relapse of METH abuse. To date, no effective pharmacotherapy exists for METH abuse and its withdrawal symptoms. Therefore, understanding the neuromechanism underlying METH abuse and its withdrawal symptoms is essential for developing clinical strategies and improving patient care. The aims of this study were to investigate brain network abnormalities in METH abusers (MAs) and their associations with affective symptoms. Forty‐eight male abstinent MAs and 48 age‐gender matched healthy controls were recruited and underwent resting state functional magnetic resonance imaging (fMRI). The severity of patient anxiety and depressive symptoms were measured by Hamilton anxiety and depression rating scales, which decreased across the duration of abstinence. Independent component analysis was used to investigate the brain network functional connectivity (FC) properties. Compared with healthy controls, MAs demonstrated hypo‐intra‐network FC in the cerebellar network and hyper‐intra‐network FC in the posterior salience network. A whole‐brain regression analysis revealed that FC strength of clusters located in the right rostral anterior cingulate cortex (rACC) within the ventromedial network (VMN) was associated with affective symptoms in the patients. Importantly, the intra‐network FC strength of the rACC in VMN mediated the association between abstinence duration and the severity level of affective symptoms. Our results demonstrate alterations in brain functional networks underlying METH abuse, and that the FC of rACC within VMN serve as a neural substrate in the association between abstinence length and affective symptom severity in the MAs.

METH abuse, and that the FC of rACC within VMN serve as a neural substrate in the association between abstinence length and affective symptom severity in the MAs. METH abuse produces a variety of adverse effects, including physiological, affective and cognitive dysfunction, which are usually demonstrated in terms of withdrawal symptoms during abstinence from continuous METH exposure (Hellem, Lundberg, & Renshaw, 2015;Rothman & Baumann, 2003;Thompson et al., 2004). Among the affective symptoms, anxiety and depression are the most common disorders experienced by MAs (Darke, Kaye, McKetin, & Duflou, 2008). Importantly, METH abuse co-occurring with anxiety or depressive symptoms have greater overall impairment, higher suicide attempts and worse treatment adherence and outcomes relative to those without affective symptoms (Glasner-Edwards et al., 2008;Glasner-Edwards et al., 2010).
The severity level of the affective symptoms decreases in line with the duration of METH abstinence, but the symptoms can extend from several days to 2-5 years following the last use of METH (Rawson et al., 2002;Zweben et al., 2004). Failure to manage affective withdrawal symptoms also exacerbates relapse of METH abuse (Bao et al., 2013;Zorick et al., 2010). Thus, addressing affective withdrawal symptoms and syndromes in MAs may help to optimize treatment outcomes and develop strategies for relapse prevention.
The affective withdrawal symptoms observed in MAs are presumed to reflect brain functional and structural abnormalities in drug abusers (Rothman, Blough, & Baumann, 2008). Pharmacological and neuroimaging findings suggest that acute METH use activates brain reward systems including limbic, paralimbic, and basal ganglia circuits (including nucleus accumbens [NAcc] and ventral tegmental area), and leads to molecular and neurochemical changes in these brain regions resulting in feelings of pleasure, confidence, and euphoria (Cruickshank & Dyer, 2009;Koob & Volkow, 2016;Panenka et al., 2013). However, repeated use of METH results in neurotoxic effects, such as depletion of monoamine stores, downregulation of monoamine receptors and transporters, and neurite degeneration in the reward system, especially in the striatum and limbic and paralimbic regions, which contribute to the anxiety and depressive symptoms associated with acute withdrawal and protracted abstinence (Cruickshank & Dyer, 2009;Koob & Volkow, 2016;London et al., 2004;Volkow et al., 2001). In addition, chronic drug administration increases stress function in the motivational circuits of the ventral striatum and extended amygdala, and contributes to the negative emotional states in acute withdrawal and protracted abstinence (Koob & Volkow, 2016).
Recently, neuroimaging studies have confirmed that different cortical areas are functionally integrated as brain networks supporting different neurobehavioral functions (Fox et al., 2005), and recent advances in psychoradiology (Gong, Kendrick, & Lu, 2021) including conducting resting-state functional connectivity (FC) analysis of brain networks to investigate brain functional changes in patients with psychiatric disorders have been widely used (Haber et al., 2020;Lui, Zhou, Sweeney, & Gong, 2016). While previous resting state fMRI (rs-fMRI) studies mainly focused on dysfunction of a single network or circuit (Ipser et al., 2018;Kohno et al., 2018;Mansoory et al., 2020) and disrupted topological graph properties (Siyah Mansoory, Oghabian, Jafari, & Shahbabaie, 2017) in MAs, whole-brain network resting-state FC changes among MAs are still unclear, especially those underlying their anxiety and depressive symptoms.
Our study therefore aimed to use resting state fMRI to characterize abnormal functional brain networks in male MAs compared with age-gender matched healthy controls (HCs), and also to identify FC changes associated with affective withdrawal symptoms in the MAs.
The severity levels of anxiety and depressive symptoms in patients were measured by Hamilton anxiety (HAMA) and depression (HAMD) rating scales. To avoid hypothesis bias, we utilized data-driven independent component analyses (ICA) to identify whole-brain networks, and compared the FC differences of each network between the MAs and HCs. In addition, whole-brain regression analyses were performed to investigate the association between affective symptoms and brain network FC in the patients. Subsequently, mediation analyses were used to explore the potential relationships among affective symptoms, clinical characteristics and brain network FC in the MAs. We hypothesized that (a) FC alterations in brain networks related to reward or stress circuits may underlie the affective symptoms in the MAs, and (b) these network FCs may also mediate the association between the duration of abstinence and affective symptom severity.  participants were excluded if they had any drug use history other than nicotine or neurologic or psychiatric illness as assessed by medical records or structural brain defects on T1-weighted images.

| Drug use and affective symptom measures
A detailed interview including questions on socio-demographics, drug use history, and anxiety and depressive symptoms were recorded by experienced psychiatrists. The HAMA and HAMD clinical rating scales were performed by the interviewer to assess the anxiety and depressive symptoms of the MAs. The HAMA is a 14-item questionnaire which includes both psychic and somatic anxiety items with the total score ranging from 0 to 56 (Hamilton, 1959;Vaccarino, Evans, Sills, & Kalali, 2008). The HAMD is a 17-item questionnaire with the total score ranging from 0 to 50 (Hamilton, 1960). For both HAMA and HAMD scales higher total score indicate more severe anxiety or depressive symptoms.

| MR imaging acquisition
All participants underwent a rs-fMRI examination by using a 3.0 T system (Tim Trio; Siemens Healthineers, Erlangen, Germany) equipped with a 12-channel phased-array head coil. A gradient-echo echo-planar imaging sequence was used for obtaining blood oxygen leveldependent-sensitive MR images with the following parameters: repe-

| Resting state fMRI image preprocessing
Preprocessing of individual rs-fMRI data with the first five volumes excluded consisted of brain extraction, motion correction, slice timing correction, high-pass temporal filtering equivalent to 100 s (0.01 Hz) and spatial smoothing (5 mm FWHM Gaussian kernel). The global signal was not regressed out. Functional MRI data were registered to the individual's structural image and the 2 mm MNI152 standard space template using Boundary-Based Registration (BBR) method implemented in FMRIB's Linear Image Registration Tool (FLIRT). The FMRIB's ICA-based Xnoiseifier-FIX (v1.061 beta) Salimi-Khorshidi et al., 2014) was applied on each individual's rs-fMRI data to largely control for any influences of head motion and other nuisance noise (e.g., respiration, cardiac pulse), and to produce cleaned data sets for the subsequent analyses. The level of head motion related noise was significantly reduced (comparison between the mean framewise displacement [FD] before and after denoising in each group both p < .0001), but without significant group differences either before or after denoising (both ps > .05). Full details of the cleaning steps are provided in the Supplemental Materials.
To create the spatial templates, the cleaned individual resting state data of the 48 HCs were fed into the MELODIC package in FSL (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/MELODIC) for group-level decomposition by a temporal concatenation approach with a determined number of output components (i.e., 25 components) (Beckmann & Smith, 2004 Figure S1). The naming of the networks in the current study was based on previous reports (Huang et al., 2015;Jiang et al., 2018) and on the correlation analysis with other reported brain networks in Shirer et al., 2012 and. The obtained brain networks were used as spatial templates for subsequent analyses.

| Group comparisons of brain network functional connectivity
The between group analysis of the rs-fMRI data were conducted by dual regression and permutation tests (Nickerson, Smith, Ongur, & Beckmann, 2017;Winkler, Ridgway, Webster, Smith, & Nichols, 2014) that allow voxel-wise comparisons of FC patterns. The spatial templates were used to generate subject-specific versions of the spatial maps and associated time-series, using dual regression. First, for each subject, the spatial templates were used as spatial regressors in a multiple regression analysis against the cleaned individual 4D data set.
This results in a set of subject-specific timeseries, one per spatial template. Next, those timeseries were used as temporal regressors, again in a multiple regression analysis against the same 4D data set, resulting in a set of subject-specific spatial maps, one per spatial template. Finally, FSL's Randomize nonparametric permutation-testing tool (10,000 permutations) was used to test for statistically significant group differences of intra-network connectivity controlling for demographic variables, that is, age and education years. A threshold-free cluster enhancement (TFCE) method was used for thresholding and family-wise error (FWE) correction with alpha level of .05 for controlling multiple comparisons across voxels of the whole brain (Smith & Nichols, 2009

| Mediation analysis between FC and affective symptoms measures
To examine the indirect effect of resting-state brain network FC strength on the association between clinical measurements of METH use, duration of abstinence and anxiety or depression scores, we conducted mediation analyses using the SPSS PROCESS v3.4 with a 5,000 bias-correction bootstrapping approach (Hayes & Preacher, 2014;Preacher & Hayes, 2008). Mediation analysis is a path analysis used to statistically evaluate how the independent variables (IVs) transmit their effects on the dependent variables (DVs) through potential intervening variables or mediators (M) (Hayes, 2018). To this end, the clinical measurements and duration of abstinence were considered as the IVs, HAMA/HAMD scores-associated intra-or between-network FC z-scores were considered as the M and the anxiety or depression levels (HAMA/HAMD scores) were considered as the DVs. Age and education years were used as the covariates in these mediation analyses.

| Exploration analysis for the affective symptoms
Since there is strong association between HAMA and HAMD scores, we used the dimension reduction method-principle component anal-

| Group differences in brain-network functional connectivity
We defined 19 brain networks from the rs-fMRI data of the HCs ( Figure S1) which were used as spatial templates. Age and education years were used as covariates in the group comparison analyses.
Results showed that the MAs relative to HCs had a significantly higher intra-network FC for the posterior cingulate cortex (PCC) in the posterior salience network and a lower intra-network FC for the cerebellum in the cerebellar network (p < .05, FWE corrected, cluster size >10 contiguous voxels) (Figure 1).

| Associations between affective symptoms and brain-network FC in the MA group
Whole-brain regression analyses were conducted to detect potential relationships between brain-network FC and affective symptoms as estimated by the HAMA, HAMD, and PCA1 scores in the MA group.
Significant negative correlations were revealed between HAMA, HAMD scores and intra-network FC in the ventromedial network (VMN) for clusters in the right rostral ACC (including part of BA 24 and 25) and medial prefrontal cortex (medPFC) (including part of BA 24, 25, and 32) respectively (p < .05, FWE corrected, see Table 2 and Figure 2).
Correlation analysis showed that higher intra-network FC of the right rACC in the VMN was associated with lower HAMA (Spearman r = −.42, p = .0032) and HAMD scores (Spearman r = −.46, p = .0010) in the MAs (Figure 2).
Using the PCA1 to represent affective symptoms, the whole-brain regression analysis found five clusters in the VMN exhibiting a significant negative correlation with the PCA1 value in the MAs (p < .05, FWE corrected, see Table 2 and Figure S2). The association between the FC of the cluster located in rACC within the VMN and PCA1 for the MA group (Spearman r = −.32, p = .025) is illustrated in Figure 2.
For between-network FC, coupling between the auditory network and VMN demonstrated a significant positive correlation with HAMA scores in the MAs (Spearman r = .38, p = .0071) (Figure 2).

| Intra-network FC of rACC in VMN linking abstinent days and affective symptoms
The correlations between number of days of abstinence and HAMA and HAMD scores were significant in the MAs ( Figure S3). Mediation  analyses were performed to investigate whether FC strengths of the brain networks underlie the association between affective symptoms and the duration of abstinence after controlling for age and years of education. The results revealed that the association between affective symptoms and the intra-network FC of right rACC in VMN significantly mediated the effect of abstinence duration on HAMA scores (β = −.16, p = .010) and the PCA1 value (β = −.14, p = .0092) ( Figure 3). The detailed results of the mediation analyses are illustrated in Table S1 and S2.

| DISCUSSION
In this study, we investigated differences in the resting state FC of whole-brain networks between male MAs and age-gender matched HCs. The MAs compared with HCs demonstrated hyper-intranetwork connectivity in the posterior salience network and hypointra-network connectivity in the cerebellar network. Furthermore, the intra-network FC strength of different clusters located in the right rACC within the VMN was associated with anxiety and depressive symptoms in the MAs. Importantly, the intra-network FC of the rostral ACC in the VMN mediated the association between abstinence duration and the level of affective symptoms. Overall, these findings suggest that the FC of brain networks is disturbed in the male MAs during abstinence and that the FC of rostral ACC within the VMN serves as a neural substrate underlying the link between the abstinence length and the affective symptoms in the male abstinent MAs.

| Brain network FC differences between MAs and HCs
Our findings revealed abnormalities in intra-network FC in the PCC of posterior salience network and the cerebellum of cerebellar network in the MAs compared with HCs. The finding of significant hyper-intranetwork FC of the right PCC in posterior salience network in the MAs compared with HCs was in line with previous studies reporting structural deficits and metabolic increase in this region in chronic MAs Thompson et al., 2004). The posterior salience network, including bilateral supramarginal gyrus, parietal operculum cortex, precuneus and PCC extending to the middle portion of cingulate and insular cortices, is responsible for processing behaviorally salient events and for task initiation and switching between introspection and executive function (Dunlop, Hanlon, & Downar, 2017;Sridharan, Levitin, & Menon, 2008). The PCC is anchored within the posterior salience network and proposed as a functional integrative hub playing an important role in cognitive function by mediating information flow around the brain (Leech & Sharp, 2014). Based on the PCC's predominant contribution to cognitive function, we postulate that the hyper-intra-network FC in the PCC reported here may F I G U R E 1 Group differences in brain network FC using age and education years as covariates. Compared with HC, the MA group demonstrated a significantly higher intra-network FC for the posterior cingulate cortex in the posterior salience network (a) and a lower intra-network FC for the cerebellum in the cerebellar network (b). The graphs illustrate the standard residual value of intra-network FC (z scores) after regressing out age and education years as covariates. FC, functional connectivity; HC, healthy control; MA, methamphetamine abuser; R, right. **p < .01; ***p < .001 underlie a variety of cognitive deficits in MAs (Darke et al., 2008). In line with this implication, a task fMRI study has provided evidence that the PCC exhibits higher activation in the context of the Stroop effect in MAs compared with the controls (Ghavidel et al., 2020).
Interestingly, we also found three clusters located in the right dentate nucleus, cerebellar lobules I-IV and V in the cerebellar network demonstrating hypo-intra-network connectivity in the MAs compared with HCs. The primary function of cerebellum has been recognized as being involved in motor function and anatomical tract tracing in nonhuman primates (Kelly & Strick, 2003) together with functional neuroimaging observations in humans (Buckner, Krienen, Castellanos, Diaz, & Yeo, 2011) have revealed that the anterior lobe of the cerebellum (including lobules I-IV and V) is anatomically and functionally mapped to the motor cortex. Clinical observations also showed that right cerebellar lobules I-IV and V demonstrated reduced FC with motor cortex in patients with essential tremor compared with healthy controls (Buijink et al., 2015). Therefore, the hypo-intranetwork connectivity we observed in the lobules I-IV and V of cerebellum may imply METH-induced motor deficits in the MAs including stereotypic motor activity and locomotor hyperactivity (Ferrucci et al., 2007;Seiden, Sabol, & Ricaurte, 1993). On the other hand, recent accumulating evidence has revealed that the cerebellum is also engaged in other high-level functions, such as emotional memory and experience (Sacchetti, Baldi, Lorenzini, & Bucherelli, 2002;Schutter & van Honk, 2006) and cognitive functions (Kuper et al., 2011). Studies on nonhuman primates have demonstrated that the cerebellar dentate nuclei structurally project to, and are functionally related to, the prefrontal cortex involved in working memory (Middleton & Strick, 2000), while animal model (Bauer, Kerr, & Swain, 2011) and patient studies (Thoma, Bellebaum, Koch, Schwarz, & Daum, 2008) suggest that lesions in cerebellar dentate nuclei result in motivational deficits. Our finding of hypo-intra-network connectivity in the cerebellar dentate nuclei may therefore reflect cognitive deficits usually seen in MAs, such as in executive function and working memory (Potvin et al., 2018). Moreover, our findings are also in line with observations suggesting the involvement of the cerebellum in many T A B L E 2 Clusters demonstrate significant group differences and association with affective symptoms in intra-network FC in the MAs Group differences in intra-network FC with age and education years as covariates

Brain networks
Regions Vox x y z Note: The MNI coordinates of the peak voxel and number of voxels in the cluster are reported to express group differences in intra-network FC and significant associations between intra-network FC and affective symptoms in the MAs. The brain area corresponding to peak voxel is in bold and the anatomical areas included in the cluster are written in regular font. The clusters were FWE corrected and thresholded at p < .05 with larger than 10 contiguous voxels. of the brain functions affected in drug addiction (Miquel et al., 2016;Moulton, Elman, Becerra, Goldstein, & Borsook, 2014). In addition, neuroimaging studies have found that the cerebellum responds robustly to acute drug exposure and craving for the psychostimulants cocaine and methylphenidate (Moulton et al., 2014). In cocaine abusers, cue-elicited craving is also correlated with increases in the regional metabolic rate for glucose in the cerebellum (Grant et al., 1996). Together with our finding of hypo-intra-network connectivity of cerebellum in the MAs these observations indicate an important detrimental influence of METH on cerebellar function in humans.

| Affective symptom association with brain network FC in MAs
In line with previous reports (Bao et al., 2013;Zorick et al., 2010;Zweben et al., 2004), we found anxiety and depressive symptoms co-occurred frequently in the MAs, and some individuals continued to experience these affective symptoms from weeks to months after abstaining from METH use. Our correlation analysis also revealed that the severity of these affective symptoms decreased with increasing durations of METH abstinence.
To further explore affective symptom-related brain networks, we conducted whole-brain regression analyses using the severity level of affective symptoms as regressors in the MA group. Results showed that the intra-network FC strengths of the clusters located in the ros- assignment and incentive salience (Dunlop et al., 2017). Our finding supported our hypothesis and previous reports that repeated METH abuse results in dysfunction in the reward system, especially in limbic regions, which contribute to affective symptoms in abstinent patients (Koob & Volkow, 2016). Addictive patients usually demonstrate distorted incentive salience (Goldstein & Volkow, 2011), which is also a common feature shown in patients with depression (Dunlop et al., 2017). For example, substance use disorder patients attribute excessive salience to drug and drug cues and decreased sensitivity to nondrug reinforcers, which links to abnormal activation of the VMN (Diekhof, Falkai, & Gruber, 2008;Goldstein & Volkow, 2011 tions with hypersensitivity to immediate gains rather than long-term outcomes, which is also consistent with dysfunction in the brain reward system (Paulus et al., 2002). Similarly, abnormal decisionmaking links to the presence of prominent depressive symptoms in mood disorders, which ascribes to a reward processing impairment, characterized by a hyposensitivity to positive outcomes in depression (Mukherjee, Filipowicz, Vo, Satterthwaite, & Kable, 2020;Whitton, Merchant, & Lewandowski, 2020) and anxiety disorders (Reilly et al., 2020).
In terms of a key specific brain region demonstrating abnormal FC in the VMN, the rostral ACC-an important node in the VMNreceives a dense dopaminergic innervation  which has been implicated in emotional processes, and dysfunction of this region occurs in major depressive disorders (Drevets, 2000). Previous PET studies have demonstrated that negative affective states, including anxiety and depressive symptoms, are associated with abnormal regional metabolism in the rostral ACC , which could serve as an index for predicting eventual medication treatment response in major depression (Mayberg, 2002). In addition, our finding that the rostral ACC exhibited functional associations with affective symptoms in the MAs has also been reported in treatment-resistant depression patients (La Torre et al., 2020). This could possibly explain why no pharmacological approaches to date have been found to be effective in treating anxiety (Hellem, 2016) or depressive (Hellem et al., 2015)  Interestingly, we noted that the brain regions demonstrating FC differences between groups and associated with affective symptoms were located in the right but not the left hemisphere of the brain, which is in line with a previous study reporting right lateralization of reinforcing and conditioned drug responses in cocaine abusers (Volkow et al., 1999). This phenomenon merits further investigation to determine hemispheric lateralization effects in addiction.

F I G U R E 3
The FC for rACC within the VMN significantly mediated the association between abstinence duration and affective symptoms after controlling for age and years of education in the MAs. The FC of right rACC in the VMN (mediator) mediates the association between abstinence duration (IV) and HAMA scores (DV) and PCA1 value (DV

| Limitations
This study has several limitations. First, this study is a cross-sectional investigation, and cannot establish causal relationship between brain function and drug addition, as well as its relationship with affective symptoms in the MAs. Thus, longitudinal studies are needed to help address these questions in future. Second, the subjects were recruited from a compulsory isolation and rehabilitation center and several unmeasured factors might have contributed to between-group brain network differences, such as living environment, nutrition, and activity levels. Third, we only recruited male MAs and the results cannot represent the brain alterations in female patients. Since previous studies have demonstrated gender effects on brain disturbances in MAs, future works on female patients are needed. Fourth, subjects' education durations were not matched between the two groups, mainly due to difficulties in recruiting healthy young subjects with low educational levels in China currently. Although linear regression was used to exclude the potential effects produced by the education durations variances on the results a better designed community-based study with a larger sample size should be conducted in the future to validate the functional neural mechanisms for affective symptoms associated with METH abuse. Finally, we could not identify and exclude subjects who fell asleep during eyes-closed rs-fMRI data acquisition, which may have influenced on our final results. Therefore, rs-fMRI data obtained with subjects eyes open are needed to confirm our observation in the future.

| CONCLUSION
By investigating whole-brain network FC differences between MAs and HCs and characterizing affective symptoms severity associated with them in the MAs, we have identified changes in intra-network FC of cerebellum and posterior cingulate cortex in the cerebellar and posterior salience networks respectively. Furthermore, we have demonstrated that the rostral ACC is a key region underlying the association between abstinence duration and affective symptom severity.
Our findings shed light onto alterations in brain functional networks that underlie METH abuse and its withdrawal psychiatric symptoms and indicate that the rostral ACC may serve as a neural substrate in the association between abstinence length and affective symptom severity in MAs.

ACKNOWLEDGMENTS
This study was supported by the National Natural Science Foundation Huaiqiang Sun and Song Wang for their help on data analysis and we appreciate all the subjects participated in our study for their contribution to our project.

CONFLICT OF INTERESTS
All authors declare no potential conflicts of interest.

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