Default mode network mechanisms of repeated transcranial magnetic stimulation in heroin addiction

Repetitive transcranial magnetic stimulation (rTMS) over the left dorsolateral prefrontal cortex (DLPFC) has been shown to reduce cravings in heroin-dependent (HD) individuals, but the mechanisms underlying the anti-craving effects of rTMS are unknown. Abnormalities in the default mode network (DMN) are known to be consistent findings in HD individuals and are involved in cravings. We assessed the effect of rTMS on DMN activity and its relationship to the treatment response. Thirty HD individuals were included in this self-controlled study, and all HD participants received 10-Hz rTMS 7-session during a week. Data for cravings and withdrawal symptoms and resting-state functional magnetic resonance imaging data were collected before and after rTMS treatment. Thirty demographically matched healthy individuals who did not receive rTMS were included as controls. We focused on changes in coupling seeded from the medial prefrontal cortex (MPFC), posterior cingulate cortex (PCC), and bilateral inferior parietal lobe (IPL), which are the core regions of the DMN. The craving and withdrawal symptom score of HD individuals decreased significantly after rTMS treatment. The left IPL-left middle frontal gyrus coupling and the left IPL-right inferior occipital gyrus coupling decreased significantly, and the changes in the left IPL-left middle frontal gyrus coupling were positively correlated with changes in drug-cue induced cravings. rTMS could modulate the coupling between the DMN and executive control network (ECN). Alterations of the left IPL-left middle frontal gyrus coupling may play an important mechanistic role in reducing drug cue-induced cravings.


Introduction
Heroin addiction, which is defined by compulsive heroin seeking and consumption behaviors despite awareness of the serious negative consequences, can result in serious social and personal problems such as individual morbidities, economic losses, a public health crisis, and overdoseinduced mortality (Kolodny et al., 2015). Several studies have provided direct evidence of the safety and effectiveness of methadone maintenance treatment (MMT) for heroin addiction (Kolodny et al., 2015;Preston et al., 2000). MMT is associated with reductions in heroin craving and withdrawal symptoms, but patients receiving MMT still have a substantially high risk of relapse and respiratory depression (Kolodny et al., 2015;Zhou & Zhuang, 2014). Therefore, efforts are being made to identify safe and effective new methods to treat heroin addiction. Non-invasive brain stimulation techniques such as repetitive transcranial Long Jin and Menghui Yuan contributed equally to this work. * Wei Wang tdwangw@126.com * Qiang Li tdqiangqiang@foxmail.com magnetic stimulation (rTMS) have shown promise for the treatment of substance use disorder (SUD) (Antonelli et al., 2021;Ekhtiari et al., 2019;Lefaucheur et al., 2020;Zhang et al., 2019). Furthermore, several lines of direct evidence have indicated that 10-Hz rTMS targeting the left dorsolateral prefrontal cortex (DLPFC) could effectively decrease drug cravings in heroin-dependent (HD) individuals Shen et al., 2016). An increasing number of functional magnetic resonance imaging (fMRI) studies have focused on the neural mechanisms underlying the effects of rTMS treatment for SUD. In this regard, previous studies have revealed that rTMS could result in decreased functional connectivity between the left DLPFC and the medial orbitofrontal cortex in nicotine-dependent individuals . A recent neuroimaging study also showed that rTMS could increase frontal-parietal functional connectivity, which was associated with craving reductions in methamphetamine-dependent individuals (Su et al., 2020). However, the mechanisms underlying these neuroimaging data for the effects of rTMS on heroin addiction are largely not clear. Thus, there is an urgent clinical need to understand how rTMS reduces cravings in HD individuals. Previous studies have already proven that the human brain is intrinsically divided into spatially and temporally dissociable functional networks (Fox & Raichle, 2007;Fox et al., 2005). Menon (Menon, 2011) proposed a triple brain network model consisting of the default mode network (DMN), salience network (SN), and executive control network (ECN), and this network has been shown to be consistently abnormal in heroin addiction . Notably, tightly coupled negative correlation activity was also found between the ECN and DMN, and the SN plays an important role in modulating the relative activity of the DMN vs. ECN (Fox et al., 2005(Fox et al., , 2009). The DMN, which mainly includes the medial prefrontal cortex (MPFC), posterior cingulate cortex (PCC), and bilateral inferior parietal lobe (IPL), is implicated in attention, self-monitoring, and introspective thinking (Greicius et al., 2003). Spontaneous and drug cueinduced cravings are critical behavioral drivers for relapse in HD individuals, and are both important therapeutic targets in the treatment of heroin addiction (Li et al., 2012). Our previous studies as well as studies by other authors have revealed that drug cue-induced cravings are involved in the neuronal activity of the DMN in HD individuals Li et al., 2013a;Lou et al., 2012). In a recent study, we also found that HD individuals exhibited lower cue-induced brain activity in the PCC and left IPL during protracted abstinence, with a tendency toward that of healthy controls (HCs) (Wei et al., 2020). Furthermore, an earlier study indicated that rTMS could normalize depression-related hyperconnectivity in the DMN but did not modulate connectivity in the CEN (Liston et al., 2014). These findings suggest that normalization of impaired DMN activity may result in reduced cravings, and may even reduce the risk of relapse in HD individuals.
The ECN, which includes the bilateral dorsolateral prefrontal cortex (DLPFC) and bilateral posterior parietal cortex (PPC), has been implicated in attention processing, working memory, and decision-making (Li et al., 2018). Our studies and those reported by other authors have provided promising evidence that HD individuals exhibit abnormal brain activity and functional connectivity within and between the DMN and ECN (Goldstein & Volkow, 2011;Li et al., 2016Li et al., , 2018Ma et al., 2015). An increasing number of recent studies have also shown that rTMS targeting the left DLPFC could modulate the coupling of brain networks. An fMRI study showed that rTMS over the DLPFC could modulate the functional connectivity between the DMN and ECN (Chen et al., 2013) in HCs. Liston et al. (2014) found that rTMS could modulate the functional connectivity between and within the DMN and ECN in depression patients. Su et al. (Su et al., 2020) found that rTMS could decrease the coupling between the DMN and insula in methamphetamine-dependent individuals. However, the ability of rTMS to modulate the functional connectivity within and between the DMN and ECN in HD individuals is poorly understood. Moreover, the underlying mechanisms of the anti-craving effect of rTMS are still largely not clear.
In the current study, we performed functional magnetic resonance imaging (fMRI) to analyze the functional connectivity within and between the DMN and other brain networks in HD individuals before and after a 7-session rTMS course. We first defined the patterns of coupling of the DMN in the HD individuals, and then analyzed the alterations in coupling after rTMS. We hypothesized that rTMS could reduce the cravings for heroin by normalizing the abnormal patterns of the DMN.

Participants
Thirty-five HD individuals and 30 matched HC individuals were recruited in this study after obtaining informed consent from the participants and approval from the Ethics Committee of Tangdu hospital. The inclusion criteria for HD individuals were as follows: (1) meeting the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5) criteria for opioid use disorder; and (2) undergoing MMT for more than one year, with no less than one month of a stable dose of methadone. Inclusion criteria for all participants were as follows: (1) right-handedness; (2) eligible for MRI scanning; and (3) smokers. Exclusion criteria for all individuals included (1) a history of major brain illness, (2) a history of neurologic disease or neurologic signs, (3) current medical illness or recent medicine use, (4) meeting the criteria for any neurological or mental disorder (except heroin and nicotine use disorder). This study is a part of a larger clinical trial (NCT04086459).

Study design
In this self-controlled study, HD individuals received 10-Hz rTMS targeting the left DLPFC. At baseline before rTMS, all HD individuals were analyzed by an experienced psychologist and were required to complete a self-reported administrated handbook. The handbook included questions regarding demographic characteristics (age, education, weight, and height), heroin use history, MMT history, spontaneous and drug-cue induced cravings, and withdrawal symptoms, which were assessed using a published scale for HD individuals. After a 7-session course of rTMS, all HD individuals were only required to undergo evaluations of withdrawal symptoms and spontaneous and drug-cue induced cravings. Withdrawal symptoms were assessed by evaluating three components: somatization, negative mood, and dyssomnia; each of these components was assessed by a separate questionnaire (Shi et al., 2007). Drug-cue induced cravings for heroin were assessed by the event-related drugcue exposure paradigm, which consisted of 24 heroin-related picture cues of heroin injection, paraphernalia, and preparation and 24 neutral picture cues of household objects or chores. Spontaneous and drug-cue induced cravings were assessed by our published 0-10 visual analog scale (Li et al., 2012(Li et al., , 2013aWang et al., 2011) with the following question: "To what extent do you feel the urge to use heroin?" (0 for the least and 10 for the strongest craving) (Li et al., 2015a). Before and after the 7-session rTMS course, restingstate functional and structural MRI scans were also obtained immediately after collecting demographic and behavioral data. The standard experimental protocol is described in supplemental Fig. 1.

rTMS procedure
All HD individuals underwent rTMS once a day for seven consecutive days in the MMT clinic. The rTMS group received a 10-Hz stimulation targeting the left DLPFC for a duration of 10 min (5s on, 10s off, 2000 pulses) (Shen et al., 2016). We used the M-100 ultimate device (YingChi CO., Shenzhen, China) equipped with a figure-8-shaped BY-90 A stimulation coil for accurate stimulation. The resting motor threshold (RMT) was defined by stimulating the left motor cortex to identify the minimum intensity produced by a visually detectable thumb movement in the right abductor pollicis brevis muscles (APB), which produced motor evoked potential (MEP) responses of at least 50 µV in 50% of the trials. The approximate location of the left DLPFC was determined by moving the rTMS coil 6 cm anterior to M1 along a parasagittal line Li et al., 2017Li et al., , 2013b. On the subsequent rTMS, we repeated the RMT program, and the DLPFC position was reproduced by using YingChi rTMS repositioning caps. The rTMS was applied at 100% of the RMT. In addition, all HD individuals regularly received MMT in the clinic.

Imaging acquisition
Alcohol, caffeine, tea, and any other drugs were prohibited 12 h before the MRI scan. All images were acquired on a Siemens 3.0T scanner equipped with a 12-channel head coil. In the formal scanning process, each participant was asked to close the eyes and remain quiet while avoiding thinking about anything specific or falling asleep. All images were acquired using a gradient echo planar imaging sequence (repetition time/echo time = 2000 ms/30 ms, field of view = 256 × 256 mm 2 , imaging matrix = 64 × 64, slice thickness/gap = 4 mm/0 mm, number of slices = 36 [covering the whole brain], flip angle = 90°, spatial resolution = 4 × 4 × 4 mm 3 ) to yield 180 brain volumes over 6 min. High-resolution 3D T1-weighted images were also collected with the MPRAGE sequence (repetition time/ echo time = 1900 ms/2.43 ms, inversion time = 900 ms, field of view = 256 × 256 mm 2 , imaging matrix = 256 × 256, slice thickness = 1 mm, number of sagittal slices = 192, flip angle = 9°, spatial resolution = 1 × 1 × 1 mm 3 ). The T1-weighted images were assessed by an attending radiologist to rule out potential lesions.

Imaging data processing
MRI data were preprocessed using SPM12 software (http:// www. fil. ion. ucl. ac. uk/ spm) and MATLAB, which involved discarding the first five brain volumes, followed by routine slice timing, realignment, spatial normalization (resliced resolution, 3 × 3 × 3 mm 3 ), and smoothing (full width at half maximum = 6 mm). Translational head motion thresholds were set at 3 mm, while rotational head motion thresholds were set at 3°. The signals for cerebrospinal fluid, white matter, and 24 head motion parameters were regressed out (Friston et al., 1996). Finally, we applied linear detrending and temporal band-pass filtering of 0.01-0.1 Hz on the data.

Region of interest definition and functional connectivity
We used the group-independent component analysis applied with GIFT (http:// mialab. mrn. org/ softw are/ gift/) on the smoothed data to define the DMN. The component number of 25 was determined by estimation and the informax algorithm. The spatial maps of group-independent components were converted into a Z score map and thresholded at 2.3. The DMN was defined through visual inspection by two experienced researchers. Four regions of interest (ROIs) of the DMN, including the MPFC, PCC, and bilateral IPL were defined (supplemental Fig. 2). In terms of functional connectivity, a mean time series for each ROI was calculated for a reference time series in each participant. A cross-correlation analysis was then conducted between the mean time series in the ROI and the time series for each voxel in the whole brain. Fisher's z-transform was then applied to improve the normality of the correlation coefficient. Finally, the Fisher's z-transformed data were submitted to the subsequential statistical analysis.

Statistical analysis
We performed a paired t test (HD [pre-TMS] vs. HD [post-TMS]) to test the effects of rTMS on functional connectivity based on each ROI in the HD group, taking into account the covariance associated with age, sex, education, smoking duration, number of cigarettes, and head motion. We also performed a two-sample t test (HD [pre-TMS] vs. HC and HD [post-TMS] vs. HC) to test the differences in functional connectivity based on each ROI before and after TMS treatment, taking into account the covariance associated with age, sex, education, smoking duration, number of cigarettes, and head motion. The differences were significant at P < 0.05 (corrected by Gaussian Random Field, GRF). Statistical analyses of behavioral and demographic data were performed using SPSS19.0 (α < 0.05). We only used the fMRI data of the HC group that did not receive rTMS as control to show the trend of changes in the HD group.
A post-hoc analysis of Pearson correlation data was performed between the alterations in withdrawal symptoms and spontaneous and drug cue-induced cravings and the changes in functional connectivity values extracted from clusters showing significant effects in the HD group for exploration purposes. Three analytical steps were adopted to examine these correlations: (1) identifications of functional connectivity that were significantly altered in the HD group (post-vs. pre-rTMS); (2) extraction of the mean signal of significant clusters at post-and pre-rTMS time points; and (3) Pearson correlation analysis of the changes in functional connectivity and alterations in withdrawal symptoms and spontaneous and drug-cue induced cravings (post-vs. pre-rTMS).

Participants
Thirty-five HD individuals and 30 HCs were enrolled in this study. However, only 30 (5 HD individuals dropped out due to the headache) finished the entire course of rTMS. The HD and HC groups showed no significant differences in terms of age, years of education and smoking duration, cigarettes per day, and head motion ( Table 1).

Effects of rTMS on withdrawal symptoms and cravings
The result of two-paired t tests showed that the withdrawal symptom score (t = 4.31; P < 0.001), spontaneous craving score (t = 2.27; P < 0.05), and drug-cue induced craving score (t = 4.36; P < 0.001) of HD individuals significantly decreased after a 7-session course of rTMS (Fig. 1). Fig. 1 The withdrawal symptom score, spontaneous craving score and drug-cue induced craving score of HD group was significantly decreased after finishing a 7-session course of rTMS. Error bars = S.E.M. * = P < 0.05, ** = P < 0.01, *** = P < 0.001

rTMS affects functional connectivity based on the DMN
In analyses based on an ROI of the PCC, we found that the HD group exhibited increased PCC-left inferior frontal gyrus coupling and PCC-right precentral gyrus coupling than the HC group (HD [pre-TMS] vs. HC) ( Fig. 2A).
After finishing a 7-session course of rTMS, the abnormally increased PCC-left inferior frontal gyrus coupling in the HD group decreased, showing a similar tendency to that Fig. 2 The functional connectivity between the PCC and whole-brain networks after rTMS. A Compared to the HC group, the HD (pretreatment) group exhibited increased PCC-left inferior frontal gyrus coupling and PCC-right precentral gyrus coupling. Images depict t statistics for the contrast of HD (pre-treatment) versus HC group. B The abnormally increased PCC-left inferior frontal gyrus coupling of the HD group decreased after rTMS treatment, showing a similar tendency to that in the HC group, but the hyper PCC-right precen-tral gyrus coupling of the HD group persisted after finishing a 7-session course of rTMS. Images depict t statistics for the contrast of HD (post-treatment) versus HC group. C Quantification of data extracted from the different regions in panels A and B. For coordinates and statistics, see Table S1 in By evaluating the ROI of the left IPL, we found that the HD group exhibited increased left IPL-left precentral gyrus coupling and left IPL-right inferior frontal gyrus coupling in comparison with the HC group (HD [pre-TMS] vs. HC) (Fig. 3A). The abnormally increased left IPL-left precentral gyrus coupling and left IPL-right inferior frontal gyrus coupling tended to decrease and showed a tendency approaching that of the HC group after a 7-session course of rTMS (HD [post-TMS] vs. HC) (Fig. 3B). rTMS showed significant effects on left IPL-left middle frontal gyrus coupling and left IPL-right inferior occipital gyrus coupling in the HD group based on left IPL: the functional connectivity in both areas significantly decreased after a 7-session course of rTMS (HD [post-TMS] vs. HD [pre-TMS]) (Fig. 3C).
By evaluating the ROI of the right IPL, we found that the HD group exhibited increased right IPL-left precentral gyrus coupling and right IPL-right precentral gyrus coupling than the HC group (HD [pre-TMS] vs. HC) (Fig. 4A). Notably, these abnormally increased couplings decreased after a 7-session course of rTMS (HD [post-TMS] vs. HC) (Fig. 4B). However, rTMS had no significant effects on functional connectivity based on the right IPL in the HD group (HD [post-TMS] vs. HD [pre-TMS]).
Analyses of the ROI of the MPFC did not indicate any significant effects of rTMS on functional connectivity.

Correlation analyses
The correlation analyses showed that only the changes in functional connectivity between the left IPL and left middle frontal gyrus were positively correlated with changes in the drug-cue induced craving score (n = 30, P < 0.05, r 2 = 0.169) in the HD group (post-rTMS vs. pre-rTMS) (Fig. 5). However, no significant correlation was found between changes in the left IPL-right inferior occipital gyrus coupling and changes in behavioral data.

Discussion
We found that rTMS could decrease withdrawal symptoms, spontaneous cravings, and heroin-cue induced cravings in the HD group after a 7-session course of rTMS. Our findings are consistent with the results of previous studies showing that 10-Hz rTMS can reduce withdrawal symptoms and craving in HD individuals Mahoney et al., 2020;Shen et al., 2016). In terms of neuroimaging results, we found that rTMS selectively reduced the DMN (left IPL) -ECN (left middle frontal gyrus) coupling and the DMN (left IPL) -right inferior occipital gyrus coupling after rTMS treatment. In addition, this change in functional connectivity between the left IPL and left middle frontal gyrus was positively correlated with changes in the drug-cue induced craving score in HD individuals.
The DMN, which consists of the MPFC, PCC, and bilateral IPL, is thought to be related to attention, self-monitoring, and introspective thinking (Buckner et al., 2008;Greicius et al., 2003;Raichle et al., 2001;Volkow et al., 2015). Our previous studies as well as studies by other authors have shown that abnormal alterations in the intrinsic functional pattern of the DMN are a hallmark of heroin use disorder (Kuo et al., 2019;Li et al., 2016Li et al., , 2015bMa et al., 2015). The left middle frontal gyrus, a region in the left DLPFC, is a key hub of the ECN, which plays a key role in decisionmaking, problem-solving, and memory-working (Goldstein & Volkow, 2011;Sridharan et al., 2008;Volkow et al., 2015). Several studies have revealed that functional changes within the ECN are associated with impaired executive control and are an important indicator in SUD (Fu et al., 2008;Goldstein & Volkow, 2011). Moreover, increasing evidence suggests that abnormal intrinsic functional patterns and interaction of the ECN, DMN, and SN (involved in detection and attention capture) may underlie the mechanism of SUD (Geng et al., 2017;Lerman et al., 2014;Li et al., 2018;Liang et al., 2015;Volkow et al., 2015). Early studies have shown an anti-correlation between the DMN and ECN in healthy individuals (Fox & Raichle, 2007;Fox et al., 2005Fox et al., , 2009). In our previous fMRI study, the relapsed HD group showed a weaker correlation between the DMN and ECN (Li et al., 2018). In a TMS/fMRI study, Chen found that single-pulse excitation over the DLPFC could induce negative coupling between the ECN and DMN in HCs (Chen et al., 2013). Listen found that 10-Hz TMS could induce anti-correlated coupling between the DMN (MPFC) and ECN (left DLPFC) (Liston et al., 2014). Meanwhile, Li found that 10-Hz rTMS could decrease the functional connectivity between the DMN (medial orbitofrontal cortex) and ECN (left DLPFC) Fig. 3 The functional connectivity between the Left IPL and wholebrain networks after rTMS. A Compared to the HC group, the HD (pre-treatment) group exhibited increased left IPL-left precentral gyrus coupling and left IPL-right inferior frontal gyrus coupling. Images depict t statistics for the contrast of HD (pre-treatment) group versus HC group. B The abnormally increased left IPL-left precentral gyrus coupling and left IPL-right inferior frontal gyrus coupling of the HD group decreased after rTMS treatment, showing a tendency approaching that of the HC group after finishing a 7-session course of rTMS. Images depict t statistics for the contrast of HD (posttreatment) group versus HC group. C Two-paired T tests revealed significant effects of rTMS on the left IPL-left middle frontal gyrus coupling and the left IPL-right inferior occipital gyrus coupling. In both areas, the functional connectivity significantly decreased after a 7-session course of rTMS. Images depict t statistics for the contrast of HD (post-treatment) group versus HD (pre-treatment) group. D Quantification of data extracted from the different regions in panels A − C. For coordinates and statistics, see Table S2 in Supplement 1. Error bars = S.E.M. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, corrected with Bonferroni's multiple comparison test ◂ in nicotine-dependent individuals . A recent study on smokers undergoing transcranial direct current stimulation (tDCS)/fMRI also indicated that stimulation of the left DLPFC could change the ECN (left DLPFC)-DMN (bilateral PHG) coupling, and the coupling between the left DLPFC and right PHG was negatively correlated with smoking cravings . Moreover, another recent study showed that the coupling between the prefrontal and parietal cortices could shift visuospatial attention and that 5-Hz continuous theta burst stimulation (cTBS) over the right frontal gyrus could suppress the blood oxygen leveldependent (BOLD) signal in the lateral frontal region and IPL. Our findings showed similar results, with 10-Hz rTMS over the left DLPFC reducing the functional connectivity Fig. 4 The functional connectivity between the Right IPL and wholebrain networks after rTMS. A Compared to the HC group, the HD (pre-treatment) group exhibited increased right IPL-left precentral gyrus coupling and right IPL-right precentral gyrus coupling. Images depict t statistics for the contrast of HD (pre-treatment) versus HC group. B The abnormally increased couplings decreased after a 7-session course of rTMS. C Quantification of data extracted from the different regions in panel 3A. For coordinates and statistics, see Table S3 in Supplement 1. Error bars = S.E.M. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, corrected with Bonferroni's multiple comparison test between the DMN (left IPL) and ECN (left DLPFC) with a similar tendency to that in the HC group. Our findings suggest that rTMS could decrease the craving for heroin by modulating the functional connectivity between the DMN and ECN, which is supported by the positive relationship between changes in DMN (left IPL) -ECN (left DLPFC) coupling and changes in craving.
Besides, our findings also showed a reduction in coupling between the DMN (left IPL) and right inferior occipital gyrus in the HD group after rTMS treatment. As a core region of the DMN, the IPL is mainly involved in visuospatial attention (Bokde et al., 2006;Hahn et al., 2007). Our previous studies and studies by other authors have also shown that the IPL is overactivated when exposed to drug-induced cues in comparison with neutral cues (Wei et al., 2020;Yang et al., 2009). Several studies have implicated the involvement of the IPL in the treatment of addictive disorders. Cocainedependent individuals undergoing cognitive behavioral treatment showed lower task-related neural activity in the IPL (DeVito et al., 2017). Our recent study also found that HD individuals undergoing protracted abstinence showed lower IPL activity when exposed to drug-related cues in comparison with HD individuals under MMT (Wei et al., 2020). The occipital gyrus, which is believed to relate to visuospatial processing, exhibited high activity in cue-induced tasks in SUD (Kleinhans et al., 2020;McClernon et al., 2008;Qiu & Wang, 2021). Several studies have also reported that the activity of the occipital gyrus is positively correlated with self-reported cravings (Charboneau et al., 2013;McClernon et al., 2008). Thus, the IPL and occipital gyrus are both visual regions and exhibit high activity in cue-induced tasks, implying that these regions may mediate the visual cue aspects of drug craving (Charboneau et al., 2013). To our knowledge, the visual mental imagery associated with cognitive processes is involved in a brain network consisting of the prefrontal, parietal, inferotemporal, and occipital cortices (Ragni et al., 2020). Top-down connectivity during visual imagery was found from the parietal to occipital cortex using granger causality and dynamic causal modeling analyses (Dentico et al., 2014;Ragni et al., 2020). Meanwhile, several diffusion tensor imaging (DTI) studies showed that the parietal lobe connects the frontal lobe in the superior longitudinal fasciculus (SLF) and the occipital gyrus connects the frontal lobe in the inferior frontal-occipital fasciculus (IFOF), both of which are involved in visual attention and working memory (Ffytche et al., 2010;Migliaccio et al., 2012;Yu & Shim, 2017). Moreover, the rTMS can modulate remote brain activities distal to the target regions (Fox et al., 2012;Kobayashi & Pascual-Leone, 2003). For example, an interleaved TMS/fMRI study showed that cTBS over the left frontal pole could significantly reduce the BOLD signal in the parietal cortex of cocaine-dependent individuals (Hanlon et al., 2017). Our results showed a reduction in coupling between the DMN (left IPL) and right inferior occipital gyrus in the HD group, which might be the remote secondary effects caused by rTMS targeting the left DLPFC. These findings suggest that a reduction in left IPL-right inferior occipital gyrus coupling might attenuate visuospatial attention to visual imagery for drug-related cues.
Our findings should be interpreted in light of the potential limitations of this investigation. First, this study lacked a sham-rTMS control, and we could not exclude the possibility that some of the coupling changes after rTMS might be attributable to the placebo effect. Second, our study was based on the DMN and we did not investigate whether additional networks could be modulated by the rTMS. Third, traditional skull-based landmark stimulation location was used in this study. More accurate location techniques such as MRI-guided neuro-navigation are warranted in future studies. Fourth, we did not evaluate the changes in daily methadone dosage or perform a urine test for heroin use during this study, although many of the HD individuals hoped to reduce the dosage of methadone. Fifth, although a 6-min resting-state fMRI scan is considered to be a little short to obtain reliable data for restingstate functional connectivity (Birn et al., 2013), we used it in order to reduce the likelihood of poor image quality due to poor cooperation by HD individuals. Finally, we did not account for the possibility that the degree of craving and withdrawal symptoms may be related to the time since last methadone use, but instead maintained a constant time for the pre-and post-rTMS assessments for each HD participant. Fig. 5 The correlation between the changes of functional connectivity and the changes of drug-cue induced craving score. The changes in functional connectivity between the left IPL and left middle frontal gyrus were positively correlated with the changes in the drug-cue induced craving score (n = 30, P < 0.05, r 2 = 0.169) in the HD individuals

Conclusion
In summary, our findings demonstrated that rTMS targeting the left DLPFC could modulate the DMN (left IPL)-ECN (left DLPFC) coupling and decrease cravings in the HD group. These findings may support the hypothesis that rTMS alters abnormal coupling of the DMN to reduce cravings in HD individuals. These data clarify the mechanism underlying the effects of rTMS through resting-state fMRI and will inform future efforts to optimize TMS treatment protocols and enhance their therapeutic effects for heroin addiction.